clinical waste handling and obstacles in malaysia - Centro de ...

EDITORIAL TEAM 

Editors 

Celso Augusto Guimarães Santos, Federal University of Paraíba, Brazil 

Masuo Kashiwadani, Ehime University, Japan 

Dragan Savic, University of Exeter, United Kingdom 

Vicente L. Lopes, Texas State University, United States 

Associate Editors 

Koichi Suzuki, Ehime University, Japan 

Hafzullah Aksoy, Istanbul Technical University, Turkey 

António Pais Antunes, University of Coimbra, Portugal 

Roberto Leal Pimentel, Federal University of Paraíba, Brazil 

Max Billib, Hannover University, Germany 

Bernardo Arantes do Nascimento Teixeira, Federal University of São Carlos, Brazil 

Generoso de Angelis Neto, State University of Maringá, Brazil 

FOCUS and SCOPE 

Journal of Urban and Environmental Engineering (JUEE) provides a forum for original papers and for the exchange of 

information and views on significant developments in urban and environmental engineering worldwide. The scope of the 

journal includes: 

(a) Water Resources and Waste Management: This topic includes (i) waste and sanitation; (ii) environmental 

issues; (iii) the hydrological cycle on the Earth; (iv) surface water, groundwater, snow and ice, in all their physical, 

chemical and biological processes, their interrelationships, and their relationships to geographical factors, atmospheric 

processes and climate, and Earth processes including erosion and sedimentation; (v) hydrological extremes and their 

impacts; (vi) measurement, mathematical representation and computational aspects of hydrological processes; (vii) 

hydrological aspects of the use and management of water resources and their change under the influence of human 

activity; (viii) water resources systems, including the planning, engineering, management and economic aspects of 

applied hydrology. 

(b) Constructions and Environment: Buildings and infrastructure constructions (bridges/footbridges, pipelines etc) 

are part of every urban area. In recent years there is a growing interest in seeking rationality of construction systems, in 

balance with environmental adequacy and harmony in an urban area. This involves, among others, adequacy of structural 

systems (shapes, functionality, rational design etc), use of alternative materials for construction (recycled, 

environmentally friendly materials etc) and solutions seeking energy efficiency. 

(c) Urban Design: This topic covers the arrangement, appearance and functionality of towns and cities, and in 

particular the shaping and uses of urban public space (e.g. streets, plazas, parks and public infrastructure), including also 

urban planning, landscape architecture, or architecture issues (e.g. thermic and acoustic comfort). 

(d) Transportation Engineering: This topic covers such area as Traffic & Transport Management, Rail Transport, Air 

Transport, International Transport, Logistics/Physical Distribution/Supply Chain Management, Management Information 

Systems & Computer Applications, Motor Transport, Regulation/Law, Transport Policy, and Water Transport.

SUMMARY 

CLINICAL WASTE HANDLING AND OBSTACLES IN MALAYSIA……………………………………….. 47-54 

Shaidatul Shida Razali 

Multi-criteria evaluation of the expansion of natural gas distribution network by the urban 

dynamics………………………………………..………………………………………..……………………………… 55-62 

Vanessa Meloni Massara, Miguel Edgar Morales Udaeta 

ANALYSIS OF NEIGHBORHOOD IMPACTS ARISING FROM IMPLEMENTATION OF 

SUPERMARKETS IN CITY OF SÃO CARLOS………………………………………………………………….. 63-73 

Pedro Silveira Gonçalves Neto, José Augusto de Lollo 

DETERMINING INDICATORS OF URBAN HOUSEHOLD WATER CONSUMPTION THROUGH 

MULTIVARIATE STATISTICAL TECHNIQUES……………………………………………………………….. 74-80 

Gledsneli Maria Lima Lins, Walter Santa Cruz, Zédna Mara Castro Lucena Vieira, 

Francisco de Assis Costa Neto, Érico Alberto Albuquerque Miranda 

COMPARATIVE ANALYSIS OF EROSION MODELING TECHNIQUES IN A BASIN OF 

VENEZUELA………………………………………………..……………………………………………………………. 81-104 

Adriana Marquez Romance, Edilberto Guevara Perez

Razali and Ishak 

47 

J U E E 

Journal of Urban and Environmental 

Engineering, v.4, n.2, p.47-54 

ISSN 1982-3932 

doi: 10.4090/juee.2010.v4n2.047054 

Journal of Urban and 

Environmental Engineering 

www.journal-uee.org 

CLINICAL WASTE HANDLING AND OBSTACLES 

IN MALAYSIA 

Shaidatul Shida Razali ∗ and Mohd Bakri Ishak 

Department of Environmental Management, Faculty of Environmental Studies, Universiti Putra Malaysia, 

Serdang, Selangor, Malaysia. 

Received 22 March 2010; received in revised form 22 April 2010; accepted 27 December 2010 

Abstract: 

Keywords: 

As in many other developing countries, the generation of clinical waste in Malaysia has 

increased significantly over the last few decades. Even though the serious impact of the 

clinical waste on human beings and the environment is significant, only minor attention 

is directed to its proper handling and legal aspects. This study seeks to examine the 

management of clinical waste in Selangor’s government hospitals as well as problems 

that arise from the current practice of clinical waste management. A depth interview 

with the responsible concession who handles the clinical waste management in those 

hospitals also has been taken. In general, it was found that the consortium’s 

administration was reasonably aware of the importance of clinical waste management. 

However, significant voids were presented that need to be addressed in future including 

efficient segregation, better handling and transfer means, as well as the need for 

training and awareness programs for the personnel. Other obstacles faced by 

consortiums were to handle the clinical waste including the operational costs. Waste 

minimizing and recycling, as well as the alternative treatment methods for incineration 

are regarded to be major challenges in the future. 

clinical waste, management, obstacles 

© 2010 Journal of Urban and Environmental Engineering (JUEE). All rights reserved. 

∗ Correspondence to: Shaidatul Shida Razali, Tel.: +60102177483. 

E-mail: shida_upm@yahoo.com 

Journal of Urban and Environmental Engineering (JUEE), v.4, n.2, p.47-54, 2010


48 

INTRODUCTION 

Medicine is one of the important sectors showing 

enhancement throughout the recent decades (Birpinar et 

al., 2008). A hospital is midpoint that supplies various 

healthcare services to the community. Its activity may 

include curative, rehabilitation, preventive, and 

promoting of health education. In undertaking the 

activities, the hospital may also generate waste (Chaerul 

et al., 2008). However, the fraction of waste generated 

at medical institutions has not attracted the level of 

attention as other types of wastes, especially in 

developing countries (Birpinar et al., 2008). Even 

though concrete steps like proposing a policy and laws 

to regulate clinical waste management, there is still 

problems presence due to several factors. Lack of 

training, awareness, and financial resources had been 

seemed as factors contribute to the mismanagement of 

clinical waste. 

Clinical wastes differ from any other wastes that 

being produced in hospitals. Sharps, human tissues or 

body parts and other infectious materials contain in 

clinical waste poses potential health and environmental 

risks (Bareja et al., 2000). By weight, approximately 15 

– 25% of clinical waste is considered infectious (Shinee 

et al., 2008). Even though the current practices of 

clinical waste management are different from hospital to 

hospital, the problematic are similar for all healthcare 

institutional from segregation, collection, packaging, 

storage, transport, treatment and disposal (Tsanoka et 

al., 2007). 

Environmental pollution, unpleasant odors 

encourages insects, rodents and worms to breed that 

may lead to transmission of disease like cholera, 

hepatitis or typhoid through injuries from contaminated 

sharps (Abdulla et al., 2008). 

Proper manner of clinical waste management is 

greatly important to avoid health risks and damage to 

the flora, fauna and the environment (Yong, et al., 

2009). 

Clinical Waste Management in Malaysia 

Up until 1980s, Malaysia has no proper system for the 

management of clinical waste. With the emergence of 

HIV, Ministry of Health in collaboration with 

Department of Environment took an initiative to revise 

policies and guidelines for prevention and control of 

infectious disease and clinical waste handling. 

It is estimated that total bed’s strength is about 47 

000 and 35 000 came from government’s hospital with 

the occupancy rate 65%. Clinical waste in Malaysia 

may be defined as: 

i. Any waste which consist wholly or partly of human 

animal tissue, blood or other body fluids, excretions, 

drugs or other pharmaceutical products, swabs or 

dressings, syringes, needles or other sharp instruments, 

being waste which unless rendered safe may cause 

hazardous to any person coming into contact with it. 

ii. Any other waste arising from medical, nursing, 

dental, veterinary, pharmaceutical or similar practice, 

investigation, treatment, care, teaching or research, or 

the collection of blood for transfusion, being waste 

which may cause infection to any person coming into 

contact with it. 

Currently, clinical is reported together with 

pharmaceutical waste. The total clinical waste 

generated is about 8000 tonnes per year and DOE 

(2005) estimated that the generation rate for clinical 

waste varies from 0.3 to 0.8 kg per occupied bed per 

day. 

Clinical Waste Problems and Obstacles 

In Brazil, law of environment No. 7 (1982), law of the 

atmospheric and air protection (1992), and law of 

transport of hazardous materials (2005) deals with 

clinical waste management, but problems such as 

waste transported to on-site storage containers via 

uncovered trolleys, containers placed near the main 

street within the hospitals buildings or located outside 

at the street curb, waste simply dumped in the corner 

of hospital room until it could be transported off-site, 

use of open trucks and lack of training (Sawalem, 

2008). 

Same goes to Korea. Korea National Assembly 

customized the Waste Management Act in 1999 to 

enhanced control from the generation of medical waste 

to its final destination. The Korea Ministry of 

Environment (MOE) had given the responsibility for 

implementing the act. The Korea MOE spread several 

regulations for definition, segregation, packaging, 

tracking, and disposal of medical waste but yet, the 

mismanagement still arise that medical waste was 

often mixed and waste minimization and recycling are 

still not well-promoted (Jang et al., 2006). 

In 2002, Croatia endorsed the Directive on the 

management of waste produced during healthcare 

(Republic of Croatia, 2000). The Directive portrays an 

overall system of waste management; sorting at the 

point of generation, collection, transportation, storage 

and treatment but only a small number of medical 

institutions report their waste to the Registry due to the 

weakness of Registry function. 

These deprived management sides are due to the 

lack of sensitivity from the management of the 

facilities, meaning there seems to be some lack of 

awareness concerning health risks towards the 

community members as well as environmental issues, 

and due to economic problems in the country that 

prevent the government from adequately supporting a 

healthcare policy (Silva et al., 2005; Abdulla et al., 

2008). 



49 

Concerning this, this paper aimed to: 

(a) To examine the clinical waste management in 

Selangor (one of the state in Malaysia) government’s 

hospitals. 

(b) To analyze problems faced in the management 

means. 

METHODOLOGY 

The methodological framework can be described as 

follows (Fig. 1): 

Study site 

The study took place in Selangor, Malaysia. Located on 

the west coast of peninsular Malaysia and covering 

8000 square kilometers, Selangor is bounded on the 

north by Perak, on the east by Pahang and Negeri 

Sembilan, and on the west by the Straits of Malacca. 

Selangor has been called the gateway of Malaysia. It 

is also the industrial hub of Malaysia; the country’s 

largest industrial site is located in Shah Alam, the states 

capital, just 25 kilometers from Kuala Lumpur. It is the 

most populous state in the country with a total 

population of 2.7 million inhabitants. Despite being 

industry-based, the state is blessed with natural forests, 

waterfalls, hills, and lakes to complement its many manmade 

attractions. 

There are 10 government hospitals in Selangor, 

which are located at Kajang, Serdang, Kuala Kubu 

Bharu, Ampang, Selayang, Sungai Buloh, Banting, 

Sabak Bernam, Klang and Tanjung Karang. 

Data Collection 

The first step in collecting the data was to determine the 

panelists. The panelists were the clinical waste 

supervisor in each hospital. Panelists were then 

interviewed. Interviews were helpful in obtaining 

information about common practices in the management 

of the waste (Bdour, et al., 2007). Questions were asked 

basically about the management of clinical waste in 

terms of collection, transportation, treatment, disposal 

and training. The informants were also asked about the 

clinical waste management during H1N1 acute and the 

challenges involved as well as the problems faced. The 

number of wastes generated was also asked. 

A form for the panelists to provide the information 

regarding clinical waste generation and training was 

also distributed. 

Data Analysis 

The recorded interviews were then transferred into 

transcripts. Then, the transcripts were sent back to the 

interviewees for the process of validation. In this 

process, the panelists may make any correction if they 

wish to. 

RESULTS AND DISCUSSION 

Background information and clinical waste 

generation 

The quantity of clinical waste depends upon several 

factors such as the size of hospital, the segregation 

program of clinical waste and the medical activities. 

Table 1 and Fig. 2 presents clinical waste generated 

from all general hospitals in Selangor throughout the 

year 2009. There are 10 government hospitals in 

Selangor and the number of beds ranges from 93 to 864 

with a mean of 335.9. However, in this study, we will 

look upon to just nine hospitals because one of the 

hospitals refused to give cooperation. Considering the 

occupancy bed is 100%, the average generation rates of 

total clinical waste in Selangor’s government hospitals 

were estimated to be 1.355 kg/bed/day. 

Fig. 1 Research framework. 


50 


Table 1 Generated clinical waste for Selangor’s government hospital for 2009 (in kg) 

A B C D E F G H I 

Jan 13 022.6 10 935.3 2 184.2 24 749.37 16 978.1 23 304.2 2 672.0 2 253.3 30 424.0 

Feb 12 866.5 10 932.7 1 904.0 23 506.5 17 202.1 22 451.3 2 362.5 1 985.8 29 549.9 

March 14 683.3 11 871.6 2 001.7 26 101.3 18 539.6 21 095.4 2 813.4 2 210.8 33 020.4 

Apr 1 4707 11 898.1 2 126.8 25 939.3 20 259.4 23 611.2 2 950.3 2 290.7 34 368.2 

May 14 878.3 11 553.5 1 688.8 25 233.0 20 482.9 25 452.5 2 965.2 2 535.1 34 787.3 

June 14 298.3 10 366.3 2 077.1 24 002.3 19 600.1 36 896.3 2 739.1 2 415.4 32 917.0 

July 17 397.3 11 490.7 2 286.9 25 363.7 20 958.1 27 572.3 3 154.9 2 495.4 34 164.5 

Aug 1 6906.0 12 848.1 2 465.0 27 630.9 20 361.5 26 459.0 3 073.1 2 597.9 36 435.4 

Sept 14 587.5 11 498.8 1 676.7 23 141.3 18 306.3 23 121.2 2 840.5 2 519.0 31 556.9 

Oct 17 701.8 13 953.3 2 383.9 26 959.6 21 693.7 25 528.7 2 999.1 2 685.4 36 815.7 

Nov 16 498.9 12 753.2 1 981.0 25 414.1 19 948.7 23 678.4 2 853.2 2 595.9 33 005.9 

Dec 17 381.4 13 698.6 2 010.9 26 757.5 20 901.1 24 702.4 3 013.1 2 666.1 33 709.1 

Table 2. Specialties in each hospital 

Services 

Anaesthesiology 

Dermatology 

Herpetology 

Nephrology 

Paediatric 

Psychiatry 

Haematology 

Obstetric and Genecology 

Orthopaedic and Traumatology 

Ophthalmology 

Teeth & Mouth Operation 

Radiology 

Haemodialysis 

Otorhinolaryngology (ENT) 

Urology 

Cardiothoracic 

Ear, Mouth & Throat 

Plastic Surgery 

Paediatric and Neonatology 

Rehabilitation 

Neurosurgery 

■ indicates the service available at the hospital. 

Hospital 


Fig. 2 Reported generation of clinical waste in Selangor’s 

government hospital. 

Fig. 3 Generated clinical waste presented in months. 



51 

181 

306 

150 

778 

382 

340 

114 93 


Hospital's name 

Fig. 4 Number of beds in Selangor’s government hospital. 

Figure 4 illustrates the number of beds in Selangor’s 

government hospitals. 

Almost all hospitals offer same services; general 

medical, general surgery, emergency, pathology and 

maternity. Hospital F offers a lot of specialties that are 

not available in other hospitals. Table 3 presents the 

specialties offered in all hospitals. 

From the table, the number of waste generated is 

depending on the number of beds and departments in 

the hospitals. Hospital J, which has the highest number 

of beds (864 beds), produced the highest amount of 

clinical waste (406 753.7 kg). Also, the location of the 

hospital influenced the number of waste produced. For 

example, Hospital D, H and I produced wastes less than 

50 000 kg because they are located at the rural area, 

compare to other hospitals that are located in urban area. 

There is no specific reason to why the number of 

wastes decreased in September but H1N1 outbreak 

could be the possible explanation of why the number of 

wastes increased and became the month with the highest 

production of clinical wastes throughout the whole year. 

According to the clinical waste Supervisors, there are 

a few units that had been identified to generate the most 

wastes. 

Abu: Usually GOT (General Operation Theatre), 

maternity, and ICU (Intensive Care Unit). But the rest is 

normal; however, it is hard to say that. It depends on the 

patients. Sometimes, there is no patient entering ICU, 

which means there is no waste. 

Ahmad: Ward 2 (dengue cases), labor room, GOT, 

emergency. 

Based on the interview, it can be concluded that high 

production of clinical wastes comes from Operation 

Theatre, maternity, Intensive Care Unit, Emergency and 

dengue cases ward. 

Clinical waste collection, storage, transportation and 

incineration 

The collection usually starts at 08:30 am until 11:30 am. 

Porter with well-equipped Personal Protection 

864 

Equipment (PPE) then pushes the yellow cart and start 

doing collection from unit to unit. The PPE that are 

being used include apron, glove and boot. The fullyfilled 

yellow bags with the wastes will then be 

transferred to the cart and the process will be repeated 

until all the waste has been collected. Then, they will 

send the waste to the cold storage. Here, they will wait 

for the liaison officer or the hospital verifier to witness 

the weighing process. The weighing process is usually 

conducted from 11:30 am to 12:30 pm. While doing the 

collection, there are some porters that are being 

assigned to clean the bin. They will change the used bin 

with a new bin. To clean the bin, 100 ml of a chemical 

named SteriQuat were mixed with 250 L of water. 

Then, the bin will be dried. The bin must be washed 

daily. 

While for the disposable sharps and needles, a 

specific bin that is called sharp bin was used. After it is 

¾ filled, porter will close the cap and it will be collected 

and transferred to the cart. Unlike other yellow bins, this 

sharp bin will not going to be washed. It will be 

incinerated and the wastes inside it are not going to be 

transferred to other place. 

For a small hospital which produces less clinical 

waste, the waste will be store in a refrigerator with the 

temperature range of -1 to -5°C while for the hospital 

which produce large amount of clinical waste, there is 

no refrigerator provided. This is because for the hospital 

which produce small amount of waste, the lorry will 

collect the waste three times per week while for the 

hospital which produce large amount of waste, the lorry 

will come daily to collect the waste. The lorry will send 

the waste to the incinerator plant in Teluk Panglima 

Garang. The time for the lorry to come and collect the 

waste is from 12:00 noon to 05:30 pm. 

During the H1N1 outbreak, the management of 

clinical waste in the H1N1 ward is slightly different. 

The hospital briefed to the porter who was assigned to 

collect the waste. Only one person can be assigned. At 

one time, he then will be given a H1N1 immunization. 

The wastes that have a H1N1 contact will be placed in a 

yellow bag and being stored in a labeled yellow cart. 

When the lorry comes, this yellow cart has the priority 

to be collected and incinerated when they reach the 

incineration plant at Teluk Panglima Garang. 

A flowchart of the incineration system is being 

described in Fig. 4. The system consists of a waste 

feeder, a primary combustion chamber (rotary type), a 

secondary combustion chamber, a waste heat boiler, dry 

air pollution control, and a 24-h gas-emission 

monitoring device. The melting of the ash occurs in the 

secondary combustion chamber at 1 200°C (Azni et al., 

2005). 

With the capacity of 500 kg/h, this incinerator 

operates 24 hours. Table 3 summarizes the specification 

of the incinerator. 



52 

Fig. 5 Flow diagram of clinical waste incinerator. 

Table 3 Specifications of the incinerator 

Operation time 

24h 

Type of fuel 

Diesel 

Type of feed 

Hospital waste 

Treatment capacity 

500 kg/h 

Amount of flue gas 

20 000 m³/h 

Amount of slag produced 10% 

Temperature in Primary Chamber 

950° C 

Temperature in Secondary Chamber 

1 200° C 

Source: Azni et al. (2005). 

The clinical waste management in Selangor’s 

government hospital can be simplified as follows: 

Training and Awareness 

The concession company is the one who is responsible 

to conduct the trainings. There are two groups of people 

who are involved in the trainings. They are the porter 

and the user (hospital staffs). The porter will go for 

trainings at least three times a year. The training 

involves clinical waste collection procedure, procedure 

on wearing Personal Protection Equipment (PPE), 

clinical waste service management, procedure on 

washing the bag holder, procedure on filling the record 

sheet, procedure on filling bag holder washing form, 

customer relationship program, clinical waste 

segregation, procedure on handle the accident and 

health. The training will be given throughout the year or 

when accident occurs. 

For the user, the training will be given three times a 

year or upon request. Usually, the hospital will ask the 

concession company to conduct training for user when 

there is a new staff. The trainings involve identifying 

the types of waste, clinical waste segregation, clinical 

waste handling, safety and health. 

Unfortunately, there is no specific training for the 

public. The only way to educate the public, especially 

on the segregation of waste is through posters. The 

poster will be pinned-up on top of the bin to give 

awareness to the public regarding the segregation. 

Problems and Obstacles 

Fig. 6 Clinical Waste Pathway in Selangor’s Government Hospital. 

Although there is a management system in clinical 

waste handling in Malaysia, there are several problems 

and obstacles that had been found throughout the study. 



53 

Ahmad: Usually, doctors always throw needles 

everywhere, sometimes in the yellow bag. That will 

cause problems. 

Ali: The problems arise while we are doing the 

collection, the people from the district clinics come, and 

the problem is that they like to store the waste outside 

the fence but they are not allowed to do so. 

Abu: If every user is being given a room to place the 

bins and do the cleaning process, it would be easier, 

avoiding transportation in the elevator and contact with 

public, which would cause trouble to publics. 

Abu: The price should be increased because the 

groceries’ price is increasing, plastic and all the bins. 

The price is not the same as 10 or 15 years ago. Look at 

the current economy; I think we should increase the 

price. 

Abu: There is no drying machine. 

Aini: Porters have no infectious control towards patients 

or hospital. They open the door and collect the waste by 

using the same hand. 

Aini: The problem we face now is that even we had 

given the training, not all patients know about it. In the 

yellow bin, they are not allowed to throw the clinical 

waste. We cannot open the bag back because it is 

already mixed. There is no awareness from the public. 

Adila: If we do the collection in the ward, we will carry 

the plastic bag to the outside. The large wheel bin 

cannot be pull inside the ward. That is the weakness 

because we have to carry the yellow bag to the outside. 

It would spill on the floor. 

Mala: for the clinical waste, there are too much 

procedures and too much contact. There are too much 

things that we need to follow. 

From the interview, there are several problems in 

managing clinical waste. The user’s attitude has been 

identified as one of it. Even though training had been 

given, their self awareness still considered as low. The 

second problem is regarding the waste from district 

health centre. For district health centre, there is no 

specific concession company to manage clinical waste 

as government hospitals does. They need to send their 

waste to the nearest government hospital by their own. 

The problem arises when they tend to put the waste 

outside the fence of the clinical waste storage. This is 

dangerous because it will expose the waste to the public 

as well as animals such as dogs and cats. The vaporized 

waste might spread the infectious disease through the 

air. Space had been identified as one of the problems in 

managing clinical waste. 

There is no enough space provided for the hospital 

support services in the hospital. For clinical waste, they 

need a large space especially to wash and dry the bin. 

Because of this limitation, some bins are not being 

washed properly or not being washed at all. Also, using 

a traditional approach to dry the bins proved to be a 

major problem too. A drying machine for the bins is 

needed especially for the hospitals which produces high 

amount of wastes. 

When the agreement was first signed, both parties 

agreed to update the service charge from time to time. 

Unfortunately, since it was signed in 1992, there is no 

revision made regarding the price. The current price is 

RM5.20 per kilograms. The price should be revised due 

to the increasing price of raw materials and 

transportation costs. 

The problem also occurs from the handling of the 

clinical waste. By using the same hand to collect the 

waste and to open the door, the probability for the 

infectious disease to spread becomes higher. When the 

porter carries the bags to the clinical waste cart outside 

the ward, there is a probability of the infection 

spreading through the spillage on the floor occurs. 

Public awareness has been identified as one of the 

problem. Because of the low public awareness on the 

management or handling of clinical waste, they tend to 

mix up the clinical waste and the general waste. This 

will increase the disposal cost because the general waste 

will be treated as clinical waste once it is contacted with 

the clinical waste. 

There are several agencies that responsible to the 

clinical waste management. Due to this, a lot of 

requirements and procedures need to be followed. This 

seems as the problems to the concession company in 

dealing with clinical waste management. 

CONCLUSION 

This study helped in establishing database, information 

and statistics on the clinical waste sources, generation, 

collection, transportation, treatment and disposal. It also 

highlighted the problems and obstacles faced during its 

handling. Also, it has provided suggestions for policy 

makers and further information to facilitate policy 

development and improve clinical waste management. 

The management of clinical waste has been of major 

concern due to its potential high risks to human health 

and the environment. The current practices for the 

handling, transportation, storage and disposal of wastes 

generated at the hospitals needs to be changed and 

improved. Generally, the problems and obstacles arise 

from the handling of clinical waste are non-segregated 



54 

waste, awareness and attitudes among hospital’s staffs 

and patients, collection of waste from district healthcare 

centre, facilities and spaces provided, service charge 

and documentations. 

Hospital’s design is very important. It is the first step 

to allocate spaces for the clinical waste management; 

either from collection, storage and disposal. Abu had 

stated before, “If every user is being given a room to 

place the bins and do the cleaning process, it would be 

easier. No elevator’s problem, contact with public and 

causing trouble to publics”. It seems that training and 

supervision is very important to educate not only the 

porter, hospital’s staff but also all the patients to make 

sure there’s no misuse of bins that had being provided. 

Other than that, improvements and research and 

development activity (R&D) should be carried out 

continuously to combat the problem that might not 

occur before but tend to be hazard such as what has 

being highlighted by Ani and Adila about infectious 

control. The policy also should be revised to make it 

more understandable and will not cause the difficulty as 

what Mala had told, “For the clinical waste, there are 

too much procedures and too much contact. There are 

too much things that we need to follow.” 

Improvements of clinical waste management require 

all parties to involve. It also requires strategic and 

systematic planning that aim the goal to controlling 

costs, educate the users and publics, understandable 

policy and to manage the clinical waste in proper 

manner as it can reduce the hazards and risks to the 

ecosystem and the community. 

REFERENCES 

Abdulla, F., Qdais, H.A. & Rabi, A. (2008) Site investigation on 

medical waste management practices in Northern Jordan. Waste 

Manag. 28(2), 450–458. doi: 10.1016/j.wasman.2007.02.035. 

Azni, I., Katayon, S., Ratnasamy, M. & Johari, M.M.N.M (2005) 

Stabilization and utilization of hospital waste as road and asphalt 

aggregate. J. Mater. Cycles Waste Manag. 7(1), 33–37. doi: 

10.1007/s10163-004-0123-0. 

Baveja, G., Muralidhar, S. & Aggarwal, P. (2000) Medical waste 

management – an overview. Hospital Today 5(9), 485–486. 

Birpinar, M. E., Bilgili, M. S. & Erdorgan, T. (2008). Medical waste 

management in Turkey: a case study of Istanbul. Waste Manag. 

29(1), 445–448. doi: 10.1016/j.wasman.2008.03.015. 

Chaerul, M., Tanaka, M. & Shekdar, A.V. (2008) A Syatem dynamic 

approach for hospital waste management. Waste Manag. 28(2), 

442–449. doi: 10.1016/j.wasman.2007.01.007. 

Jang, Y.C., Lee, C., Yoon, O.S. & Kim, H. (2006) Medical waste 

management in Korea. J. Environ. Manag. 80(2), 107–115. doi: 

10.1016/j.jenvman.2005.08.018. 

Sawalem, M., Selic, E. & Herbell, J.-D. (2008). Hospital waste 

management in Libya: a case study. Waste Manag. 29(4), 1370– 

1375. doi: 10.1016/j.wasman.2008.08.028. 

Shinee, E., Gombajav, E., Nishimura, A., Hamajima, N. & Ito, K. 

(2008) Healthcare waste management in the capital city of 

Mangolia. Waste Manag. 28(2), 435–444. doi: 

10.1016/j.wasman.2006.12.022. 

Silva, C.E.D., Hoppe, A.E., Ravanello, M.M. & Mello, N. (2005) 

Medical Wastes Management in the South of Brazil. Waste 


Tsanoka, M., Anagnostopoulou, E. & Gidarakos, E. (2007) Hospital 

waste management and toxicity evaluation: a case study. Waste 


Yong, Z., Gang, X., Guanxing, W., Tao, Z. & Dawei, J (2009). 

Medical waste management in China (2009). Waste Manag. 

29(4), 1376–1382. doi: 10.1016/j.wasman.2008.10.023. 


Massara and Udaeta 

55 

J U E E 


Engineering, v.4 , n.2, p.55-62 

ISSN 1982-3932 

doi: 10.4090/juee.2010.v4n2.055062 




MULTI-CRITERIA EVALUATION OF THE EXPANSION OF 

NATURAL GAS DISTRIBUTION NETWORK BY THE URBAN 

DYNAMICS 

Vanessa M. Massara 1∗ , Miguel E. M. Udaeta 1 

1 Department of Petroleum and Natural Gas, Energy and Electrotechnics Institute, University of São Paulo, Brazil 

Received 28 June 2009; received in revised form 30 September 2009; accepted 02 January 2010 

Abstract: 

Keywords: 

The objective of this work is to analyze the expansion of the infrastructure of natural 

gas distribution, identifying priorities from large metropolis using the energy planning 

based on urban design tools like urban dynamics and techniques like AHP (analytic 

hierarchy process). The methodology proposed uses matrices considering the relations 

between the concept of urban dynamics, quality of life and the possibilities of natural 

gas displacing other energy forms. The matrices are made up of information about 

social and urban development, costs of establishing the infrastructure and projections of 

the consumption potential in various sectors. Relating the consumption to urban 

development parameters and the real estate future of the areas in study, the 

methodology allows indicating for each district, the viability of implementing a gas 

network. As conclusion, the model presents the integration between the cities profile 

and the natural gas use, by means of a growth natural gas on districts of São Paulo City 

as a specific case study. 

Energy, natural gas, infrastructure, urban development, analytic hierarchy process 


∗ Correspondence to: Vanessa Meloni Massara, Tel.: +55 11 9860-8116. 

E-mail: vanessa.massara@gmail.com 



56 

INTRODUCTION 

Natural gas (NG) is currently the third largest source of 

primary energy in the world, surpassed only by oil and 

coal (EIA, 2008). 

In the ANP (Brazilian National Agency of Oil, 

Natural Gas and Biofuels) surveys for the next years, 

new natural gas fields will be incoming in Brazil. For 

instance, the Manati field will duplicate de actual gas 

production in state of Bahia solving in the short-term 

the problem of natural gas unsatisfied demand. In the 

Espírito Santo state, the Peroá-Cangoa field emerges as 

an opportunity of complementing the gas for the 

Brazilian northeastern and southeastern regions and to 

reliability the Campos Basin production in the state of 

Rio de Janeiro. But the Mexilhão field in the São Paulo 

state, Santos Basin, as a prelude of the context of the 

pre-salt reservoirs, will be anticipating the gas 

production as definition of the federal Brazilian 

administration (BNDES, 2006). Also in January of 

2008, in the pre-salt hydrocarbons offshore Brazilian 

reservoirs of Santos Basin (state of São Paulo) another 

well was discovered. The reserve that was entitled 

Jupiter is a large natural deposit for natural gas and 

light-oil, in deep waters estimated in very large gas 

mega-field that, despite demanding large investments, 

will make Brazil double its reserves (GASBRASIL, 

2008). With this favorable forecast of natural gas supply 

and considering its various uses, this work focuses on 

market gas demand in sectors like residential, 

commercial, services and industrial. 

In this work, an analytical methodology that 

integrates the understanding of the urban dynamics to 

the strategies of expansion in the natural gas distribution 

network is considered, characterizing gas consumption 

possibilities and attractiveness for the districts, 

composing a city. 

The methodology is developed by gathering 

information such as family income, demographic 

density and construction area, percentage of land use, 

number of households as well as commercial, service 

and industrial establishments, number of real estate as 

well as indicative information released by the Urban 

Plan of the city regarding the increments in the 

peripheral districts. By relating the gas consumption 

estimated by each type of land occupation and the cost 

for expanding the gas distribution network, the model 

will indicate, for each neighborhood, the viability of 

implementing a gas network as well as the places with 

potential for growing density in the existing gas 

distribution system. In this paper, examples of essential 

information that compose the methodology are 

presented for six districts of São Paulo city (Brazil): 

Ipiranga,Tatuapé, Penha, Vila Matilde, Socorro and Vila 

Formosa, which have different socioeconomic and 

geographical profiles. Finally, the model tested for São 

Paulo will be generalized in a computer model, which 

allows its use in other Brazilian cities, pointing out the 

possibilities of natural gas as a final option of energy in 

the urban uses, besides presenting guidelines for the 

Urban Plans and the sustainable gas infrastructure 

incorporation in the cities (Udaeta et al., 2004). 

METHODOLOGY 

The methodology based on urban indicators (Massara, 

2007) has the main objective of developing procedures 

that allow to analyze and to guide the expansion and the 

growing density of the natural gas network inside a 

municipal district through the study of the development 

among several parameters and also to analyze the urban 

dynamics that determines the expansion of the 

metropolitan natural gas infrastructure (Williams, 1962; 

Forrester, 1969). The systemic model was elaborated 

according to the following stages: 

• Identification, characterization and organization of 

the main interventionary factors; 

• Definition of the study cells (streets, districts, 

suburbs) according to the availability of 

information about the parameters; 

• Hierarchization of quantitative and qualitative 

parameters through the attribution scale of 

priorities that unifies the parameters units, so that 

they can mathematically be treated; 

• Evaluation of the system with the purpose of 

directing the choice based in theoretical guidelines 

and also in the application of the Analytic 

Hierarchy Process (AHP) supporting the computer 

model (Saaty & Vargas, 1982); 

• Validation of the decision-making process for the 

planning of the expansion of natural gas 

distribution infrastructure through case study in 

the city of São Paulo and comparison with the 

results obtained through the model and the 

mapping of the existent implanted net. 

• Validation through comparison with the Analytic 

Hierarchy Process (Saaty, 2006). 

In this preliminary analysis, all the parameters have 

been used with influence distribution indices in intervals 

from 1 to 5. The attribution of this scale is related to the 

use of the cardinal and semantic scales (Allen, 1987). At 

the end of the study, it will be possible to determine 

whether this linear scale is valid or not, via test and 

comparison with analysis of multiple-criteria models 

that also use the priority scale as a tool in the decisionmaking 

process (Saaty, 1980). Similarly, the algorithm 

used in this paper for the determination of the 



57 

attractiveness index must be revised and compared with 

the results obtained according to the AHP methods. 

The chosen hierarchization is the same used by the 

AHP (Saaty, 1980; 2006), which allows the verification 

between algorithms of this method and that one of the 

proposal modeling (in this work), according to the 

association presented in Table 1. 

The proposed model and the information systems 

The sets of data are composed of four systems referring 

to the study area, according to the Brazilian official 

denomination “SEMPLA” (São Paulo, 2002). These 

systems establish an attractiveness index for natural gas 

infrastructure expansion with regards to the urban 

development of the cities. 

First System: Life Quality Indicators (LQ) 

This system is represented by three factors, all of which 

represent numerical values that are distributed in the 

five groups that have been initially described. 

Social Exclusion Index (SEI): it has the objective of 

identifying the social development degree of the 

districts (Sposati, 2000) considering the existence of 

social equipments (green areas, bus lines, number of 

schools and hospitals). The final index attributed to each 

one of the districts lies within the interval that ranges 

between -1.00 (reflecting the worst exclusion situation, 

i.e., the first group) and +1.00 (reflecting the best 

inclusion situation, i.e., the fifth group). 

Human Development Index (HDI): This is an adaptation 

of the index created by the United Nations Organization 

(UNO), with the objective of comparing the degree of 

human development according to the districts. It 

comprises factors such as education and basic 

conditions of health in each district (IBGE, 2008). 

These indicators are transformed in intervals that range 

between 0 (first group, the worst development 

condition) and 1 (fifth group, referring to the best 

conditions). 

Priority Infrastructure (WS, SS, SI): water supply, sewer 

system and streets illumination are considered “priority 

infrastructures” (São Paulo, 2002). This parameter 

corresponds to the idea that a given district will not be 

attractive to natural gas network if it still does not 

possess such infrastructure. The index considers the 

mean percentage of the actual conditions of the three 

nets, expressed in five groups in intervals of 20%. 

Table 1. Adaptation of AHP scale to natural gas study 

Group Semantic scale for natural gas 

AHP 

scale 

1 Low attractivity to network installation 1 

2 

Low to medium attractivity to the 

installation of the network 

3 

3 

Medium attractivity to the installation 

of the network 

5 

4 

Medium to high attractivity to the 

installation of the network 

7 

5 

High attractivity to the installation of 

the network 

9 

Source: Massara (2007) based on Saaty (2006). 

Second System: Urban Plan Indicators (UP) 

These indicators combine qualitative parameters (land 

use, urban development and zoning) and quantitative 

(urbanization rate and real estate). For the analysis of 

non-numerical values, the systems are based on the 

mapping and classification of the Urban Plan of São 

Paulo city and its adaptation for any city, in order to 

verify the types of land occupation (residential, 

commercial, industrial and services) and their expansion 

perspectives. This is carried out with the purpose of 

effecting the elaboration of a neighborhood profile with 

larger tendency to industrial developments (higher 

attractiveness for natural gas), and follows the natural 

gas consumption projection listed below. 

Land Use (Lures, LUcom, LUserv, LUind): considering 

the largest percentage of streets with certain use type, 

this concept is based in the occupation characteristic of 

the city districts represented in the urban plan map (São 

Paulo, 2006). Although the municipality considers 

several occupation categories in view of the natural gas 

attractiveness, groups have been classified according to 

five main uses: 

1st. Group: horizontal residential occupation; 

2nd. Group: mixed use (commercial and residential 

horizontal); 

3rd. Group: vertical residential occupation; 

4th. Group: mixed use (commercial, services and 

residential vertical); 

5th. Group: mixed use (residential and industrial). 

Urban Development (UD): the basis of this concept is 

the Urban Plan of São Paulo city (São Paulo, 2002; 

2006), corresponding to five “macroareas” that 

comprehend the present specifications and the future 

urban developments, as represented in the following 

five groups: 

1st. Group: environmental protection – limits of public 

areas and preservation areas; 


58 


2nd. Group: urbanization and urban qualification – areas 

predominantly occupied by low income 

families with high concentration of 

irregular constructions; 

3rd. Group: requalification – areas with good 

infrastructure although presenting many 

empty properties; 

4th. Group: urbanization in consolidation – areas in 

condition to attract real estate investments 

in residences, services and commercial 

establishments; 

5th. Group: consolidated urbanization – areas formed by 

consolidated neighborhoods inhabited by 

population of medium and high income 

and good urbanization conditions. 

Land Use Rule (Zoning - Z): This refers to the rule 

imposed by the Municipality (São Paulo, 2002; 2006) 

that limits land use and the destination of the several 

sections of the city for determined uses (trade, services, 

housing, industries). The natural gas attractiveness 

predominance is, in this case, represented by the 

following five groups: 

1st. Group: residential zone of low density / zone of 

environmental protection; 

2nd. Group: residential zone of medium density / 

mixed zone of low density; 

3rd. Group: residential zone of high density / mixed 

zone of medium density / special uses; 

4th. Group: mixed zone of high density; 

5th. Group: great industrial zone occupation. 

The definition of “mixed zone” corresponds to the 

combination of the residential use, services and 

commercial uses. 

Urbanization Rate (UR): corresponds to the percentage 

of the total district area occupied by urban use (blended 

residences with commerce, services and industries) as 

compared to the total population (São Paulo, 2006; 

IBGE, 2008), which is represented in intervals of 20%. 

Real Estate for residential and service market (REres, 

REserv): this concept is related to the Construction 

Code about building facilities for natural gas in new 

constructions (International Code Council, 2000), which 

must increase the consumption of natural gas. In order 

to attribute the attractiveness index, five groups have 

been selected which were based on the largest and the 

smallest number of district releases as reported in a 

survey by EMBRAESP (2008). 

Third System: Potential of Natural Gas 

Consumption Indicators (PNGC) 

In this data system, the characteristics of the 

neighborhoods in terms of street extensions are stored. 

All parameters are numerical values, thus simplifying 

the creation of the five groups using the numerical 

simple division of values that are obtained by the 

Brazilian institutions IBGE (2008) and São Paulo 

(2002; 2006), as follows. 

Demographic Density (DD): corresponds to the ratio 

between the number of resident people and the total area 

of the district (SEADE, 2008), considering that “larger 

people concentrations generate larger energy demand”. 

Family Income (FI): corresponds to the minimum wages 

in the homes of the district (IBGE, 2008; SEADE, 

2008), considering the relationship between “income 

and energy consumption possibilities”. 

Stratification according to the type of land use, 

households and economic activities (Sres, Scom, Sserv, 

Sind): the basis of this method considers that both the 

demand for natural gas and the attractiveness of the 

district receiving the canalized gas system increase with 

the number of domiciles or establishments with 

economic activities in a specific district (Groenendaal, 

1998). The division in establishments is made through 

the Economic Activities Cadaster (IBGE, 2003) and for 

the households through the “Census” (IBGE, 2008; 

SEADE, 2008). Those data enter in the file as “number 

of units” and the system automatically transforms them 

in influence indices from 1 to 9, according to the lowest 

and the highest numbers obtained for each district. 

Besides the internal scale, the program allows the 

insertion by the user of one external influence that is 

multiplied by the attributed value. This comes already 

as a function of the number of units that distinguish the 

activities that are more appropriate to the use of natural 

gas or, for other reasons, are more attractive to the 

natural gas concessionaire. 

In the following example, the external influence has 

not been attributed. All parameters have only the 

internal scale from 1 to 9. 

Fourth System: Civil Construction Indicators (CC) 

The function of this system is to represent values 

associated with the cost of the underground 

infrastructure implementation. As for other numerical 

parameters, the five groups have been elaborated by 

simple division of the acquired values expressed by the 

four factors below. 



59 

Natural Gas Infrastructure Extension (E): corresponds 

to the distance between areas that can be served and 

others that are effectively attended by natural gas. The 

rule for the influence attribution is “the smaller the 

distance, the larger the influence attributed for natural 

gas attractiveness”. 

Natural Gas Distribution Ramification (D): corresponds 

to the total sum of the internal streets that may be 

attended by natural gas supply systems. The parameters 

are also considered in the rule: “the smaller the distance, 

the larger the influence”. Thus, districts partially located 

in served areas are attributed “five” as the influence 

index, expressing the best attractiveness conditions. 

Built Density (BDres, BDcom, BDserv, BDind): this 

factor is obtained by the ratio between the built area 

according to the type of land use (residential, 

commercial, services and industrial) and the total area 

of the district (São Paulo, 2006). In districts with high 

industrial density, lower investments with ramifications 

are considered (“capillarity” of the distribution 

network). 

Avenues and Streets of Great Importance Traffic (T): 

even with the constructive process evolution, the 

parameter indicates the importance of the district as a 

connection among neighborhoods and other municipal 

districts and the special attention that must be given to 

the interdiction plan. The number of avenues (or streets) 

is obtained by simple consultation of any streets guide 

and by counting avenues and streets of larger extension 

within any given district. 

According to the description in the methodology, all 

the values gathered are unified to the same unit 

(percentage) and divided in five zones of attractiveness 

that are converted in the hierarchisation scale of 1 to 9. 

After the weights are attributed, they are submitted to 

the calculation of “Attractiveness index”, whose 

algorithm is based on the simple sum of the weights for 

each area of study as summarized in Table 2. 

Table 2. Summary of the algorithm for the Attractiveness Index. 

Attractivity Index by Information Systems 

LQ= (ISE+HDI+WS+SS+SI)/n 

UP= 

LUres+LUcom+LUserv+LUind+UD+Z+UR+REres+REserv)/n 

PNGC= (DD+FI+Sres+Scom+Sserv+Sind)/n 

CC= (E+D+BDres+BDcom+BDserv+BDind+T)/n 

General Attractivity Index 

IG= (LQ+UP+PNGC+CC)/n 

Source: Massara (2007). Note: n represents the number of 

parameters effectively used. The acronyms are described in 

methodology. 

RESULTS AND DISCUSSION 

As suggested in the recommendations about natural gas 

policy (Meier, 1997; IESP, 2004), the choice of São 

Paulo (capital of São Paulo State) as an all-gas city, is to 

demonstrate the complete utilization of natural gas in all 

its applications. In the context of this work, the choice 

has background in: 

• The possibility of increasing the density of use of 

the current grids: the increment of the factor of 

utilization of the infrastructure implanted in the 

expanded center, besides the extending the consumption 

market can leverage the financing of the network 

expansion. 

• The process of civil construction has evolved with 

the use of a no-dig method of cutting trenches, the 

trenchless technology (Istt, 2008; Najafi, 2005). It 

reduces the trouble caused by the interdiction of traffic 

ways to rebuild the pavement, thus making its execution 

cheaper, mainly in already consolidated urban areas. 

• As mentioned by Moutinho dos Santos et al. (2002, 

p. 162): “In the São Paulo municipality, along the roads 

at the margins of Tietê and Pinheiros rivers, the greatest 

commercial areas in the country are concentrated here, 

with various shopping centers and large office 

buildings. All of them are located less than 2 km from 

the high pressure Comgás (gas utility) pipe-ring, but 

rarely it is used”. 

This situation will be the same in the next years, 

according to the map (Comgás, 2007). The map of the 

gas utility shows the natural gas grid distribution in the 

city of São Paulo, in other words, which district is 

already served by the network, permitting the selection 

of areas not served for testing the model proposed in 

this work. 

Figure 1 shows the city of São Paulo within the 

metropolitan region of São Paulo. 

Fig. 1. City of São Paulo (ash area) inserted in the Metropolitan 

Region. Source: SEADE (2008). 


60 


Since the case study has been used in the city of São 

Paulo, it has been divided in 96 districts in conformity 

with the information reported by the local Urban Plan 

Department (São Paulo, 2002; Massara, 2002). 

Figure 2 introduces the positioning of the districts 

selected in this paper, emphasizing the different 

geographical and social characteristics of each one. 

The selected districts represent areas of the city with 

great urban transformations. Some of them face a 

process of intensification of the residential real estate 

market (Ipiranga, Tatuapé and Penha), attracting 

investments in several economic activities in order to 

respond to that new occupation style. Others have low 

family income but possess industrial areas that must 

consume natural gas (Vila Formosa and Socorro). 

Others present great urban complexity, functioning 

as a “bedroom cities”, with land use predominantly 

composed by habitational groups (Vila Matilde). 

In order to demonstrate the use of the model 

developed in this work, for instance, one application is 

presented across the Table 3. 

Therefore, in this example, the following 

simplifications were considered: 

• Calculation of the attractiveness index, using the 

“General” grouping (without the division in 

information systems) and its respective ranking; 

• Study area: districts; 

• Extra weight to emphasize the segments for the 

natural gas use = 1.0 for all the activities. 

• Algorithm application according to the Table 2. 

LEGEND: 

1 - District of Ipiranga 

2 - District of Tatuapé 

3 - District of Penha 

4 - District of Vila Matilde 

5 - District of Socorro 

6 - District of Vila Formosa 

Fig 2. The 6 selected districts in São Paulo municipality (capital 

city). (Note: unscaled figure). 

Source: SEADE (2008). 

Table 3. Attribution of weights according to the four information 

systems 

Selected Districts of São Paulo City 

Parameters Vila 

Ipiranga Tatuapé Penha 

Matilde Socorro Vila 

Formosa 

ISE 7 9 7 7 9 9 

HDI 3 5 5 5 7 7 

WS 9 9 7 9 9 9 

SS 9 9 9 9 1 9 

SI 9 9 9 5 9 7 

LU res 3 5 5 7 3 9 

LU com 5 7 7 5 3 3 

LU serv 3 9 9 5 1 1 

LU ind 5 1 1 1 7 1 

Z 9 7 7 5 5 7 

UD 9 5 5 5 9 9 

UR 9 9 9 9 9 9 

RE res 5 9 5 1 1 3 

RE serv 3 9 9 1 5 1 

DD 3 3 5 5 1 5 

FI 3 5 1 1 3 3 

S res 5 3 5 3 1 3 

S com 9 7 9 3 9 5 

S serv 5 7 9 5 5 5 

S ind 9 5 5 1 9 3 

D 9 9 3 5 9 3 

E 5 7 1 5 9 5 

T 1 1 1 3 1 3 

BD res 5 9 7 7 3 9 

BC com 9 3 7 5 3 9 

BC serv 9 3 7 5 3 9 

BC ind 1 5 3 1 9 1 

Source : Massara (2007) based on Seade (2008), São Paulo (2006). 

Table 4. Results for the case-study – The general attractiveness 

index and its respective ranking 

District 

AHP 

General 

General AHP 

index 

index 

Ranking ranking 

(%) 

Ipiranga 6,1 18,9% 2º 2º 

Tatuapé 6,3 22,2% 1º 1º 

Penha 5,9 16,8% 3º 3º 

Vila Matilde 4,7 10% 6º 6º 

Socorro 5,4 15,6% 5º 5º 

Vila Formosa 5,6 16,5% 4º 4º 

Source: Massara (2007) based on Saaty (2006). 

Table 4 resumes the attractiveness ranking for five 

municipal districts for the natural gas expansion 

projection. From the results listed therein, one may 

conclude that the districts with better positions 

contemplate mixed land use (Tatuapé is the best 

example of which) and equality of consumption 

projection in all uses, associated to a larger 

demographic density and larger family income. 

The first units in the ranking of the model are 

districts with the largest development in the real-estate 

market, largest leeway of zoning. These districts are 

nearest to the areas already served and, of course, have 



61 

the largest forecast of natural gas consumption, mainly, 

in the vertical residential use (with highlight for the 

following districts: Ipiranga, Tatuapé and Penha). 

The rest of the districts concentrate the least income, 

horizontal residential use and a small commercial, 

services and industrial concentration (Vila Formosa and 

Socorro). Apart from the characteristics of the 

intermediate group, the last classification (Vila Matilde) 

is in an area with environmental limitations to 

expansion and, therefore, of small urban development. 

The comparison between the ranking using AHP 

(Saaty, 2006) and the model ranking (Massara, 2007) 

has satisfactory results (presented in Table 4). 

The model based on urban indicators has presented 

coherent results when tested in the city of São Paulo. It 

has been demonstrated that the model proved to be a 

good calculation tool with a reasonable precision 

degree, being easily understood and functioning as an 

auxiliary in the decision-making process for the natural 

gas expansion infrastructure in the Brazilian cities. 

However, in order to have greater potential of 

expansion of the natural gas grid service, it is necessary 

that there exists strengthening of the relation between 

the utilities, network engineers of natural gas pipeline 

facilities and all the civil construction related to the 

urban plans. In this case, it is good looking for the 

development of equipment for the diverse purposes that 

can utilize natural gas, including the formal assisting 

from the public entities in the implementation of 

infrastructure and in advertising of natural gas use as an 

agent for urban development as well. 

Thus, the creation of a residential and commercial 

market is based on the current use of PLG (petroleum 

liquefied gas) and on the gradual introduction of natural 

gas in daily activities, thus inducing customers to the 

use of natural gas and enabling the development of 

natural gas canalization and distribution. 

Acknowledgment This work received financial 

support from the Brazillian agencies Agência Nacional 

do Petróleo – ANP (PRH-04), Financiadora de Estudos 

e Projetos – FINEP and Ministério da Ciência e 

Tecnologia – MCT and FAPESP (Fundação de Amparo 

à Pesquisa do Estado de São Paulo) through the research 

project Process 03/06441-7 (“New Instruments for 

Regional Energy Planning seeking Sustainable 

Development”) that permitted the utilization of the 

Decision Lens Program. 

REFERENCES 

Allen, E. (1987) Techniques of atitude scale construction, Appleton 

Century Crofts, New York. 

BNDES – Banco Nacional de Desenvolvimento Econômico e Social. 

(2006) Setor de Petróleo e Gás: Perfil dos Investimentos, (Gas 

and Petroleum Sector: Investiment Profile), BNDES, Brasília. (in 

portuguese). 

COMGAS – Companhia de Gas de São Paulo. (2007) Mapa da área 

Coberta pela Rede, (Distribution Network Map), COMGÁS, São 

Paulo. (in portuguese). 

EMBRAESP - Empresa Brasileira de Estudos de Patrimônio. (2008) 

Relatório anual sobre número de lançamentos imobiliários por 

região, região (Annual report on the number of real estate 

investments by region), EMBRAESP, São Paulo. (in portuguese). 

EIA – Energy Information Administration. (2008) Annual Energy 

Outlook 2008 with Projections to 2030. Available in: 

http://www.eia.doe.gov/. 

Forrester, J.W. (1969) Urban Dynamics,M.I.T. Press, Cambridge. 

Gas Brasil. (2008) Newsletter Semanal [on line]. Available in: 

http://www.gasbrasil.com.br. (in portuguese). 

Groenendaal, W.V. (1998) The economic appraisal of natural gas 

projects, Oxford University Press for the Oxford Institute for 

Energy Studies, New York. 

IBGE – Instituto Brasileiro de Geografia e Estatística. (2003) 

Cadastro Nacional das Atividades Econômicas (CNAE), 

(National cadaster of economic activities) IBGE, Rio de Janeiro. 

(in portuguese). 

IBGE – Instituto Brasileiro de Geografia e Estatística. (2008) 

Cidades, (Cities), IBGE, Rio de Janeiro. (in portuguese). 

IESP – Instituto de Engenharia de São Paulo. (2004) Recomendações 

sobre a política para o gás natural, (Recommendations about 

natural gas policy), IESP, São Paulo. (in portuguese). 

International Code Council. (2000) International building code. 

Falls Church, New York. 

ISTT - International Society for Trenchless Technology (2008) 

Trenchless techniques.Available in http://www.istt.com/. 

Massara, V.M. (2002) O perfil da infra-estrutura no Município de 

São Paulo e sua relação com as transformações de uso do solo: o 

centro expandido e a região de São Miguel Paulista, (The 

infrastructure profile in the city of São Paulo with regards to the 

transformations on the land uses: the expanded center and the 

region of São Miguel Paulista), Dissertação (Mestrado), Escola 

Politécnica da Universidade de São Paulo. (in portuguese). 

Massara, V.M. (2007) A Dinâmica Urbana na Otimização da Infra- 

Estrutura para o Gás Natural, (Urban Dynamics in the natural 

gas infrastructure), PhD Thesis, Instituto de Eletrotécnica e 

Energia da Universidade de São Paulo. 

Méier, E.P. (1997) Memória da infra-estrutura urbana na área 

central de São Paulo, (Urban Infrastructure Memory in tnhe 

central área of São Paulo City), Associação Viva o Centro, São 

Paulo. (in portuguese). 

Moutinho dos Santos E., Fagá, M.T.W., Villanueva, L.D. & 

Zamalloa, G. (2002) Gás natural: estratégias para uma nova 

energia no Brasil, (Natural Gas: strategy for a new energy in 

Brasil), Editora Anamblume, FAPESP, PETROBRAS, São Paulo. 

(in portuguese). 

Najafi, M. (2005) Trenchless technology: pipeline and utility design, 

construction, and renewal, McGraw-Hill, New York. 

Saaty, T. (2006) Decision Lens. Programa de Computador, London. 

Saaty T. (1980) The Analytic Hierarchy Process: Planning, Priority 

Setting, Resource Allocation, McGraw-Hill, London. 

Saaty, T.; Vargas, L.G. (1982). The logic of priorities, applications 

in business, energy, healthy, transportation. Nijhoff, Boston. 

São Paulo – Secretaria Municipal de Planejamento Urbano. (2002) 

Plano Diretor estratégico do Município de São Paulo 2002-2012, 

(Strategic directory planning in the citybof São Paulo), SEMPLA, 

São Paulo. (in portuguese). 

São Paulo – Secretaria Municipal de Planejamento Urbano. (2006) 

Dinâmica Urbana, SEMPLA, São Paulo. (in portuguese). 

SEADE – Fundação Sistema Estadual de Análise de Dados. (2008) 

Pesquisa Municipal Unificada, (Unified municipal research). 

SEADE, São Paulo. (in portuguese). 


62 


Sposati, A. (2000) Mapa da Exclusão Social da cidade de São 

Paulo, (Social Exclusion Map), São Paulo, CDRom. (in 

portuguese). 

Udaeta, M.E.M;., Grimoni, J.A.B. & Galvão, L.C.R. (2004) 

Iniciação a Conceitos de Sistemas Energéticos para o 

Desenvolvimento Limpo, (Initiation to Energy Systems Concepts 

for the Clean Development), EDUSP, São Paulo. (in portuguese). 

Williams, N. (1966) The structure of urban zoning and its dynamics 

in urban planning and development, Buttenheim Pub. Corp., New 

York. 


Gonçalves Neto and Lollo 

63 

J U E E 


Engineering, v.4 , n.2, p.63-73 

ISSN 1982-3932 

doi: 10.4090/juee.2010.v4n2.063073 




ANALYSIS OF NEIGHBORHOOD IMPACTS ARISING FROM 

IMPLEMENTATION OF SUPERMARKETS IN CITY OF SÃO 

CARLOS 

Pedro Silveira Gonçalves Neto* and José Augusto de Lollo 

Urban Engineering Post-graduation Program, São Carlos University, Brazil 

Received 18 March 2010; received in revised form 27 October 2010; accepted 16 November 2010 

Abstract: 

Keywords: 

The study included supermarkets of different sizes (small, medium and large - defined 

based on the area occupied by the project and volume of activity) located in São Carlos 

(São Paulo state, Brazil) to evaluate the influence of the size of the project impacts 

neighborhood generated by these supermarkets. It was considered the influence of 

factors like the location of enterprises, size of the building, and areas of influence 

contribute to the increased population density and change of use of buildings since it 

was post-deployment analysis. The relationship between the variables of the spatial 

impacts was made possible by the use of geographic information system. It was noted 

that the legislation does not have suitable conditions to guide the studies of urban 

impacts due to the complex integration between the urban and impacting components. 

Urban planning, Law 10257/2001, neighborhood impacts, supermarket cluster 




64 

INTRODUCTION 

São Carlos is a Brazilian municipality located in the 

state of São Paulo, near its geographic center and in a 

distance of 231 km from the state capital. With a 

population of around 220 463 inhabitants (IBGE, 2009), 

distributed a total area of 1 141 km², is the 14 th largest 

city in the state in number of residents. The city is an 

important regional center, with the economy based on 

industrial and agricultural activities (in this sector, there 

is the production of sugarcane, oranges, milk and 

chicken). 

Served by a several bus and rail systems, São Carlos 

has units of production of some multinational 

companies, among them Volkswagen, Faber-Castell, 

Electrolux, Husqvarna and Tecumseh and some units of 

domestic companies, among which Toalhas São Carlos, 

Tapetes São Carlos, Papel São Carlos, Brazil Prominas, 

Cardinali, Opto Electronics and Latina. 

In the light of local needs, and in certain aspects, 

regional, the city has a network of trade and services 

distributed in high street shops, gas stations 

convenience, several supermarket chains and a shopping 

center of Iguatemi group. In the field of research, they 

present the Federal University of São Carlos (UFSCar) 

and the University of São Paulo (USP), also are present 

in the city two centers of technical development at 

Embrapa. 

Despite these qualifications, the occupation of urban 

space took place in a disorderly way, due to high 

population growth and the absence of laws to propose a 

direction to the use and occupation. The result can be 

seen in the segregation experienced by the population 

and the ineffectiveness of urban infrastructure for 

subsequent deployment of large buildings without the 

proper review within the current legal parameters. 

According Lollo (2006), since the adoption of the 

Neighborhood Impact Study as a tool for urban 

management by the City Statute, the technical means 

has an instrument whose purpose is to analyze the 

behavior of high density, however, the fact there is an 

administrative mechanism of this magnitude, does not 

represent a significant advance in public administration 

since there is no standardized evaluation system for 

such analysis and the means are not known for the 

correct application of such instruments. 

Molina (2008) states that the City Statute is subject 

to many discussions of how to apply the study of 

neighborhood impacts and what requirements that 

should be considered for the analysis of impacts, since 

this document is of recent creation. It is emphasized that 

the proposition and use of instruments of control and 

supervision of urban densification activity in Brazil is 

typical of a certain size of cities (especially capitals and 

major cities within the states), while in small and 

medium cities, such instruments do not exist or are not 

applied. 

The Neighborhood Impact Study has been practiced 

by most government (Molina, 2008) to indicate the 

developments of greatest potential impact on the 

dimensions of these, but is overlooked aspects which 

relate to the use and occupation of land, traffic 

generation and urban planning. 

In view of this, this study addressed the urban 

impacts arising from Neighborhood Impact Study postdeployment 

supermarket which fit this grouping, 

(Decree n. 19.915/98 apud Lindau et al., 2009), in the 

municipality of São Carlos and propose strategies to 

support impact assessments of the neighborhood. 

For this, we used the methodology developed by 

Lollo (2006) which produced an impact study of 

neighborhood in the same region whose focus was on 

business information technology. It was based on the 

parameters required by Federal Law 10.257/2001, 

reporting and questionnaires to standardize the 

quantification of variables that had more than one 

interpretation. 

THE NEIGHBORHOOD IMPACT STUDY (NIS) 

The Neighborhood Impact Study under the City Statute 

guarantees the negotiation between the private interests 

of entrepreneurs and the right to urban environmental 

quality of city residents, especially those who live and 

travel in the vicinity of the project under study. 

The NIS, when established in the Organic Law of 

Municipalities, provides for the participation of society 

in the analysis of each new development to be built, 

thereby exercising democracy and inclusion of local 

communities that participate in the decisions of what is 

best for their city. Some authors such as Leão-Aguiar et 

al. (2005, p. 2) argue that “the charge required for the 

preparation of the NIS, fills an important gap in the 

legislation, with regard to buildings or activities that 

damage the local environment”. 

Urban growth has increased demand and need for 

new jobs, housing options public transportation and 

pollution control, as well as the provision of basic 

services like water, sanitation, education and health. 

Regarding the social content, urban growth contributed 

with the intensification of segregation since the 

inhabitants, according to Gregori (2004, p. 6) “are 

forced to live in poor neighborhoods of large cities, 

which have given the poor quality of life in cities, and 

directly contributed to environmental degradation and 

increasing poverty in urban society”. 



65 

It is noteworthy that the Statute, the collective 

interest should prevail over the use of private property 

(Brazil, 2001). With regard to the social function of the 

city and urban property, people’s participation in the 

public interest, must ensure democratic management 

and equitable distribution of benefits and burdens 

resulting from the urbanization process. 

Since the population should have access to 

intervention, it must be aware about the recovery of 

public investment that provided the valuation of urban 

property and appropriateness of economic policy 

instruments, tax and financial public spending on 

development urban to reach the goals. 

The aim of the Neighborhood Impact Study is to 

democratize the system for making decisions on major 

projects to be undertaken in the city, listening to 

neighborhoods and communities that are exposed to the 

impacts of large ventures. 

The Art. 36 of the City Statute (Brazil, 2001) states 

that “the bylaws must include criteria to define which 

constructions depend of a previous neighborhood 

impact study as a condition for approval. These criteria 

may vary depending on the characteristics and urban 

infrastructure of the municipality, and may be based on, 

for example, the impact of traffic generated, strain on 

infrastructure, population density, shadowing of 

neighboring properties, noise, etc.” (p. 1). 

Under the same Act, art. 37, “the NIS will run to 

consider the positive and negative effects of the activity 

on the quality of life of people living in the area and its 

vicinity, including the analysis, at least the following 

questions: (i) density of population, (ii) urban and 

community facilities, (iii) use and land cover, (iv) 

valuation of real property, (v) generation of traffic and 

demand for public transportation, (vi) ventilation and 

lighting; ( vii) urban landscape and natural and cultural 

heritage” (p. 1). 

The NIR (Neighborhood Impact Report) should be 

elaborated from a series of identifiers that indicate the 

level of impact on the deployment of any new venture 

on the urban landscape, human activities that can be 

installed, urban infrastructure and on natural resources 

of the neighborhood (Moreira, 1999). 

Moreira (1999) also says that the factors can be 

quantified and demonstrated through mathematical 

models, computational, and diagnostic reports (or 

preliminary studies). 

According to Moreira (1999), the roadmap for the 

development of a NIR must bear the thought of the 

following questions: “(i) the impact of new 

development on the urban landscape, (ii) located on 

human activities, (iii) on the move people and goods, 

(iv) on urban infrastructure, and (v) on the natural 

resources of the neighborhood” (p. 110). 

Moreira (1999) also describes what should be the 

end products of a NIR that the script needs to indicate: 

“(i) the demonstration of the compatibility of the road 

system and transport, (ii) the demonstration of the 

compatibility of the drainage system with increased 

volume and velocity of storm water generated by the 

sealing area of intervention, (iii) the demonstration of 

the feasibility of water supply, sewage collection, 

electricity supply, (iv) an indication of urban 

transformations induced by the project, and (v) the 

insertion of the work in the landscape” (p. 111). 

The observation of the points raised by Moreira 

(1999) compared to the analytical criteria of the City 

Statute (2001) suggests that the discussion on the NIR 

tends to spread, because the legislation is not complete 

on all possible parameters of analysis urban only serves 

“as a guideline for the preparation of bylaws that 

address the evaluation of neighborhood impacts” (Lollo 

& Röhm, 2005, p. 39). 

About these items of the Statute, Lollo & Röhm 

(2005) produced a material that points to the aspects 

that are not considerate in some neighborhood impact 

studies. It was found that the main problems occur or 

disability legislation or disability of the methodology 

applied. Following the reasoning of Lollo & Rohm 

(2005, p. 41), “The consequences of the development of 

Neighborhood Impact Studies that do not properly 

describe or evaluate the conditions of the project, 

neighborhood, or components subject to impact, create 

bad consequences in four spheres, namely: the 

environment, for local residents, for the general 

population and to the public government”. 

SUGGESTION TO URBAN IMPACT 

EVALUATION 

The NIR does not yet have a standardized methodology; 

therefore, the discussion over the horizon of reach of 

neighborhood impacts is still open. However, over 

recent years and given the speed at which technology 

evolves, one sees a tendency to simplify the procedures 

for the method of assessing impacts on urban areas, “in 

order to remove subjectivity from this process and thus 

shape, streamline decision-making processes” (Lollo & 

Röhm, 2005, p. 7). 

Lollo and Röhm (2005) also argue that this trend is 

not unique to the evaluation of neighborhood impacts, 

and falls within a global context of establishing criteria 

for evaluation and ranking of environmental impacts as 

an alternative to standardization of the language issue. 


66 


In this sense, the use of Geographic Information 

Systems tends to be an excellent alternative to 

technological development of such activities, because 

their use permits, in addition to encoding information 

that facilitates their treatment, large savings in storage 

process, updating and analysis data. 

The implementation of an urban model in a 

geographic information system is one of the best ways 

to convey to users of the facilities of the urban structure 

as the changes could interfere both positively and 

negatively to the inhabitants of any area depending on 

the facilities that current technology provides. 

An “Urban GIS” can be used as a tool for automation 

of the municipal administration, therefore, contributes to 

urban planning with descriptive analysis of urban space 

which serves to support local government which 

demands efficient data transmission due to the growing 

processing public services (Falcoski, 1997). 

Falcoski (1997, p. 12) describes that “in general, in 

studies of urban impacts, using the computational model 

proposed levels of study in the urban scale has 

considerable improvements in both macro-analysis well 

as in micro-analysis, therefore, it is possible to simulate 

potential of high density, indices to build, compare the 

volume and verify the proper application of coefficients 

determined by law and can create comparisons of urban 

models desirable, and likely present. With regards to 

urban design, the emphasis is on mechanisms of 

regulatory control and management (monitoring) of 

dynamic spatial information”. 

DATA COLLECTION 

The procedure for obtaining the data followed the 

proposal prepared by Lollo (2006), which suggests the 

delimitation of areas of influence through the process of 

comparison matrix between the variables selected for 

the study. According to Leopold et al. (1971) apud 

Lollo (2006, p. 173), the matrix system stands out 

among the available techniques of impact evaluation 

“due to the agility, flexibility and simplicity that allows 

the survey and impact assessment. It’s also widely used 

when trying to make identifications and assessments of 

environmental impacts if it is agile”. The basic proposal 

of an “impact matrix” (Lollo, 2006) suggested in this 

study is to cross the proposed actions with 

environmental factors, and assign these crosses 

(characterizing impacts urban) values that represent the 

relative importance. 

Aiming to represent the influences of the 

development phase of the project impacts in the urban 

environment and considering the peculiarities of each of 

these steps, in this work, it was chosen to follow the 

model proposed by Lollo (2006) who distinguished 

stage of development where the impact occurs in order 

to properly weigh the importance of each impact. 

The delimitation of the area of influence of the 

companies is extremely important for correct evaluation 

of the effects stemming from neighborhood impacts and 

implications of conflict in urban land use, and depends, 

according Lollo (2006, p. 20): “(1) conditions of the 

space, (2) the occupation in question, (3) impact 

analysis”. 

Lollo & Gonçalves Neto (2006) also highlight the 

importance of a thorough assessment of interventions 

and the means considered to establish specific criteria 

for defining the area of influence of each factor. 

According Lollo (2006, p. 20.), “the determining factors 

for defining the area of influence are those related to 

traffic and parking issues because they are the most farreaching”. 

For the quantification of the impacts were possible to 

be examined, was designed a questionnaire based on 

studies of Lollo (2006), able to quantify the variables 

regarded as quality by implementing a new 

development in the urban area. 

The step of the valuation of the impacts identified 

and listed in the questionnaire established the relative 

values for each result of each impact considering their 

order (direct or indirect), magnitude (high, medium and 

low) and duration (temporary or permanent) shown in 

Table 1. For each of the conditions described were 

established values (weights) that consider the relative 

importance of impact, allowing the assessment of 

impacts (Lollo, 2006). 

From this, it was possible to establish a scale for 

quantification of the data that were collected in the field 

by means of questionnaires. 

The field survey was the option to refine the data 

volume, because the experiments reported in the work 

of Lollo (2006) demonstrated that independent of 

cooperation by the companies (in this case, 

supermarkets), some kind of information was not 

reported, which required data collection by visiting the 

field. 

Table 1. Values adopted for quantifying the impacts identified 

Classification 

Classes 

Values 

Order 

Direct 3 

Indirect 1 

High 3 

Magnitude 

Medium 2 

Low 1 

Duration 

Permanent 

3 

Temporary 

1 

Source: Lollo (2006). 



67 

Table 2. Schematic representation of the structure of the matrix of impacts used 

Planned 

Phase of Company 

Impact Available Components Measures Proposed 

Intervention 

Info Info I+ H+ M- D+ Info 

Info Info T+ D- L+ P+ Info 

Planning 

Info Info M+ P- T- I- Info 

Info Info H- T+ I+ M- Info 

Info Info P+ I+ L+ D+ Info 

Info Info H+ M T+ P- Info 

Info Info D- T+ H+ I- Info 

Construction/Adaptation 

Info Info L- D- P+ H- Info 

Info Info I+ P+ D+ L- Info 

Info Info P- H- I- T+ Info 

Info Info P- L+ M+ D- Info 

Info Info M- I+ T- H- Info 

Operation 

Info Info D+ H+ P- M- Info 

Info Info H- L- T- I- Info 

Info Info T- I+ H+ L+ Info 

Source: Lollo (2006). 

Info Info M+ P+ L- D- Info 

Thus, based on the proposed system were considered 

three steps: (1) Planning, (2) Construction / Adaptation / 

Occupancy (as appropriate), and (3) Operation, each 

representing different possibility of generating potential 

impacts as shown below. 

Still, according with Lollo (2006, p. 9): “The 

structure in module of the matrix gives two clear 

advantages. The first is that it allows the impacts are 

assessed at each stage, allowing the discussion of the 

importance that each phase represents the environmental 

quality of the neighborhood, and providing alternatives 

to distinguish between businesses that represent 

construction, those resulting from adaptations or 

extensions to buildings already exist. This situation also 

allows treat equally the two situations described 

(construction × adaptation/upgrade), the impacts 

generated by the second group are despised. In case of 

upgrade/expansion, such a proposal still allows only 

interventions are valued consistent with the process”. 

Aiming to represent the influences of the 

development phase of the impacts identified in the 

project, and considering the peculiarities of each of 

these steps, we chose to distinguish the stage of 

development where the impact occurs in order to 

properly weigh the importance of each impact. 

The structure includes the categories of information 

(aspects of the occupation, its consequences, 

environmental components assessed, mitigation 

measures and compensation arrangements) in columns, 

while the actions themselves are described and their 

impacts are assessed in rows of the matrix. 

DATA PROCESSING 

This phase included the selection and refinement of 

information obtained and defined in the stage of data 

collection, in addition to proper storage of information 

collected in a database compatible with the Geographic 

Information System from the referencing of the 

enterprises to enable spatial analysis. 

Such results of evaluation of responses obtained by 

the use of the questionnaires were fed into a database in 

Access and subsequently inserted into the environment 

of ESRI ArcGIS 9 which interface controls and 

handling are similar to others standards software of 

engineering, facilitating data manipulation and vectors. 

The structuring of the data followed the proposal 

already included in the work Martinetti (2006), Sacute 

(2006) and Gonçalves Neto (2006) which related the 

type, the relationship of the work, time and intensity of 


68 


activity, but was kept the same geographic information 

system. 

The selection of companies whose impacts would be 

considered in this proposal was based on the criteria 

proposed by Lollo (2006, p. 8) described as follows: 

“(1) can be identified neighborhood impacts occurred 

by the time of deployment of business or during the 

execution of their activities; (2) companies should have 

characteristics that allow their classification as 

supermarkets, and (3) firms should be installed in urban 

areas”. Figure 1 shows the companies selected on this 

criterion. 

The survey of the surroundings of each company on 

the map of the city was through sketches prepared 

during field visits. The visits to the site of each 

company had as its initial objective, to answer some 

questions that employees of these establishments did not 

know or did not respond, but was also serving to 

provide a more reliable location of each company, with 

the realization of these sketches. 

The simplicity of spatial representation of 

neighborhood impacts (defined area of influence 

surrounding each company considered) significantly 

facilitated the structuring of the data model in GIS. 

Thus, taking up the company as spatial reference data, 

the data structure in ArcGIS was composed of three 

fundamental layers: “Streets”, “Points” and “Buffers”. 

The spatial representation of the companies was 

done by using points, based in sketches made in field. 

The category “Companies” (with model data object 

type) absorbed the information about the spatial location 

of each company (represented by points referenced to 

the network of streets), and a set of data (as data table) 

with all information obtained in the surveys, which 

allowed the evaluation of neighborhood impacts. 

The category that has been the basis used for 

mapping the location of firms and evaluation of their 

areas of influence was named “Network Logradouros” 

(with cadastral data model) and corresponds to the 

spatial representation of the axes of public areas (roads 

centers or centerlines) with details of their name and 

direction of traffic flow. 

The database relating to this category was already 

available in the laboratory of the Center for GIS (NGeo 

- UFSCar), which is used for other academic work and 

its upgrade has performed steadily. 

The category designated as buffers (thematic data 

model) has the spatial representation of areas of 

influence of the companies studied. A circular buffer 

was the choice to standardize the treatment given to the 

spatial area of influence of all businesses and consider 

that in this case, the intensity of impacts can be 

considered uniform in all directions. 

Fig. 1 Companies selected. 



69 

The location of each company was represented on 

the map using coordinates of the UTM system which 

indicates the central point of the frontage portion 

which belongs to the company. 

After the conclusion of this Step, the sum of the 

weights of the issues data were transported to an 

Access base and after to ArcGIS. Thus, the goal was 

reached by the proposed matrix method: analyze not 

just isolated to quantify the impact of a single 

company, but also enable its comparison with another 

company and also compare the results with the set of 

all companies with respect to the variable of urban 

infrastructure. 

Regarding the classification of neighborhood 

impacts, Lollo (2006) proposed the separation of 

variables directly related to the impact on urban 

infrastructure to be evaluated in the array: high 

density population, realty valorization, ventilation 

and lighting, cultural heritage and urban 

transformations, traffic generation, demand for public 

transport and urban and community facilities. 

The next phase (impact evaluation) corresponded 

to the treatment in GIS, of the information that 

coming from the previous steps, so as to allow three 

types of analysis: (i) from impact isolated 

(interpretation of the causes and consequences of 

each impact considered, for each of the areas 

assessed); (ii) in groups of impacts, according to the 

classification already given (checking for all areas 

assessed, the existence of a group that is more 

significant impacts), and (iii) the combination of 

several related impacts (assess the full extent of 

impacts generated by each area, to classify the areas 

according to the total impact generated by each one 

and see if there are portions of the urban area more 

weaker on this type of proposal occupancy). 

ANALYSIS AND RESULTS 

The work done by Lollo & Gonçalves Neto (2006), 

showed that urban impacts may have a direct 

relationship to the size of the building, but in the case 

of supermarkets and even more specifically on the 

issue of urban impacts that reality exists, but not 

shown be the largest contributor to the intensity of 

impacts, partly due to the impossibility of assessing 

post-deployment of temporary impacts depending on 

the size of the building. 

The data obtained showed the prevalence of some 

types of impact, especially those related to real estate 

appreciation, the high density and traffic flow. 

Table 3 shows the results obtained through the 

use of questionnaires. The companies were identified 

by symbols and each column represents a type of 

impact associated with putting the building in the 

urban system. 

This table was inserted into the database after the 

processing of the results obtained in the field in 

numbers and in the GIS environment, allowing the 

Table 3. Urban impacts identified in the area 

ID 

Question 

Super 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 

M-01 × × × × × × × × × × × × × 

G-01 × × × × × × × × × × × × × × × × × 

G-02 × × × × × × × × × × × × × × × 

G-03 × × × × × × × × × × × × × × 

G-04 × × × × × × × × × × × × × × × × 

M-02 × × × × × × × × × × × × × 

M-03 × × × × × × × × × × × × × × 

P-01 × × × × × × × × × × × × × × × × 

G-05 × × × × × × × × × × 

M-04 × × × × × × × × × × × × × × × × 

G-06 × × × × × × × × × × × × × 

M-05 × × × × × × × × × × × × × 

M-06 × × × × × × × × × × × 

G-07 × × × × × × × × × × × × × × × × 

G-08 × × × × × × × × × × × × × × × × × 

M-07 × × × × × × × × × × × × 

ID SUPER (Supermarket ID) – P: Small Companies; M: Medium Companies; G: Large Companies; Questions: 1 to 10 – Real Estate 

Valuations; 11 to 14 – Urban Landscapes; 15 to 19 – Public Heritages; 20 to 24 – Traffic Volumes; 25 to 31 – Urban Facilities. 



70 

graphical representation and spatial impacts. It is 

emphasized here a curious fact with respect to the 

results by the field survey. Observing Table 3, it is 

clear that the largest contribution to the amount of 

work has been grouped into two main cores of the 

components of analysis: the real estate valuation and 

traffic volumes. The first component of this 

observation is directly related to the shift of people to 

nearby supermarkets. 

In some areas in the municipality, the supermarket 

was the factor responsible for the attraction of 

people, given the existence of some as the Dotto 

supermarket that exists in the city more than 40 

years, this season which saw the promotion of 

technological growth of the country, promoted by the 

then president, Juscelino Kubitschek, and his slogan 

of “grow 50 years in 5”. 

With the coming of multinationals into the 

country, the Brazilian citizen could live further away 

from the workplace, therefore the second component 

of the observation results, the acquisition of a 

material as a vehicle has become easier, moreover, 

the urban development grown creating more 

pathways to link the neighborhoods to the 

commercial centers. 

Therefore, there was a swelling population which 

had the opportunity to reside in areas far beyond the 

commercial centers evidenced by Figs 2 and 3. 

Fig. 2 Representation in "pie charts" of the relationship between the groups “real estate valuation” and the sum total of variables. 



71 

Fig. 3 Representation in “pie charts” of the relationship between the groups “traffic volume” and the sum total of variables. 

About the influence areas of selected supermarkets, 

it was observed that in several cases, there is the 

interpenetration or overlapping of the lines (Fig. 4). 

According Chasco Yrigoyen & Uceta (1998) apud 

Sharma et al. (2006), “influence area or commercial 

area of a municipality or a commercial equipment is the 

area that has a strong trade dependence on the place of 

study” (p. 112). 

The main occurrences of this grouping of influence 

areas appeared frequently in the central part of the 

municipality (which has different uses of urban land) 

and near to the Cidade Jardim neighborhood which has 

a residential character. 

This overlap of buffers demonstrates the intention of 

different companies reaching geographically the same 

customer. This can be seen graphically by the existence 

of intersections when there is conflict in different 

spheres of influence. 

In these regions confirmed that the urban impacts 

cannot be treated separately because, depending on the 

proximity of the objects of study, the variables of each 

company began to interfere with the behavior of others. 

The Master Plan of the city of São Carlos does not 

refer to possible interferences grouped in the period 

before the implementation of the venture, dealing with 

the variables for the study of urban isolation without 

admitting the existence of integrating factors for the use 

of urban facilities. 

This is a problem that occurs in the most of the laws 

about the subject, because the municipal government 

has not been able to produce some material about NIS. 

Moreover the NIS does not have a definite pattern of 

federal legislation. 

We must consider the efforts made by municipal 

authorities to tread a path of urban sprawl without a 

backing of the Federation, but the results showed a real 

need to reach a standard that considers the variables of 

integration of urban occupation, which must be added 

the mitigation measures in the period prior to 

installation of new developments. 



72 

Fig. 4 Influence areas of supermarkets. 

Fig. 5 Representation in proportional symbols of the relationship between the variables “supplier vehicles” and the sum of variables. 



73 

CONCLUSIONS 

The use of standardized questionnaire on field 

surveys, attached to the information provided by the 

enterprises and added the field visits, allowed the 

identification of impacts on the study. 

The ventures consolidated prior to the drafting of 

the City Statute were those who obtained the highest 

measurements of impacts because the surrounding 

urban growth in the region, unlike the supermarkets 

built or acquired after the Federal Law, which have 

found a developed urban structure and legal 

obligations to be met. 

The urban sprawl around the supermarkets was 

aggravated by the excessive number of cars 

purchased by local residents in recent years and 

consequently increase of the flow of vehicles on 

roads which were not scaled enough to support such 

a volume of traffic. 

This new condition prevents the delivery of 

products to the supermarket is done effectively, 

which contributes directly to the final assessment of 

results, therefore, the variables related to the flow of 

vehicles were the largest contributors in the final 

results. 

Figure 5 shows the quantification of the variable 

related to the logistical vehicles coming from 

suppliers. 

The Master Plan of the city of São Carlos has a 

pretty advanced legislation regarding the NIS, but 

still treats the components of analysis alone neither a 

standard method for the quantification of the impacts 

(of any type), because is backed by the Statute of the 

City, which also lacks a standard for determining the 

variables needed for the preparation of any NIR in 

the municipality. 

Finally, the NIS is still evolving and needs to be 

deepened in order to provide conditions to develop 

methods that consider the interdependence of 

variables and its integration with the urban 

environment. 

The geographic information system in this case, 

was the best option to generate relationships between 

variables and their spatial representation effectively. 

Clearer legislation, setting standards, coupled with 

a GIS which relates to the integration of urban 

elements contribute to the improvement in decisionmaking 

by the government. 

REFERENCES 

Brasil. Law 10.257, July 10 th (2001) Regulates arts. 182 and 183 

of the Constitution set out the general guidelines of urban 

policy and other matters. Official Gazette, 10 July 2001. 

Falcoski, L.A.N. (1997) Dimensões Morfológicas de 

Desempenho: Instrumentos Urbanísticos de Planejamento e 

Desenho Urbano. Doctoral Dissertation submitted to the 

Graduate Program in Environmental Structures College of 

Architecture and Urbanism, University of São Paulo, USP, 

São Paulo, Brazil, 1997. 

Gonçalves Neto, P.S. & Röhm, S.A. (2006) Avaliação de 

Impactos na Infra-estrutura Urbana Decorrentes da 

Implantação de Empresas do Pólo de Alta Tecnologia de São 

Carlos. 

Gonçalves Neto, P.S. (2006) Estudo de impacto de vizinhança 

por meio do planejamento por desempenho na Região Central 

de São Carlos. Trabalho de Graduação Integrado, UFSCar, 

São Carlos. 

Gregori, I.C. (2004) Mesa Temática: Cidades e Gestão Urbana, 

O Estatuto da Cidade: perspectivas e desafios dos novos 

instrumentos de reforma urbana. Universidade de Santa Cruz 

do Sul – RS. 

IBGE – Brazilian Institue of Geography and Statistics (2009) 

Cidades@: informations from data base of IBGE, 

Rio de Janeiro, 2009, Available in http://www.ibge.gov.br/ 

cidadesat/default2.php Accessed on 14 August 2009. 

Leão-Aguiar, L., Campos, L.R., Moraes, L.R.S. & Borja, P.C. 

(2005) Estudo de Impacto de Vizinhança: Instrumento de 

Gestão Pública para a Cidade de Salvador. 

Lindau, L.A., Diógenes, M.C. & Pinto, A.B. (2009) 

Quantificação dos impactos de pólos geradores de tráfego. 

Available in: http://www.producao.ufrgs.br/arquivos/ 

disciplinas/412_impactos_polo_gerador_versao_lindau_rev.do 

c Accessaed in 10 September 2009. 

Lollo, J.A. (2006) Utilização de Sistema de Informações 

Geográficas em Estudo de Impacto de Vizinhança: o caso do 

Pólo Tecnológico de São Carlos. 

Lollo, J.A. & Röhm, S.A. (2005) Proposta de Matriz para 

Levantamento e Avaliação de Impactos de Vizinhança. 

Martinetti, T.H. & Röhm, S.A. (2006) Avaliação de Impactos de 

Vizinhança no Meio Físico usando Sistema de Informações 

Geográficas – o caso de Pólo de Alta Tecnologia de São 

Carlos (SP). 

Molina, V. (2008) Efeito do Porte do Município em Impactos de 

Vizinhança Gerados por Supermercados. Research Plan 

forwarded to PPGEU/UFSCar, Universidade Federal de São 

Carlos, São Carlos, 2008, 3–8. 

Moreira, A.C.M.L. (1999) Parâmetros para elaboração do 

relatório de impacto de vizinhança. Magazine of the Graduate 

Program in Architecture and Urbanism of FAU/USP, 7, p. 

107–118. 

Sacute, B.C.A. & Röhm, S.A. (2006) Avaliação de Impactos 

Urbanísticos Decorentes da Implantação de Empresas do Polo 

de Alta Tecnologia de São Carlos. 

Silva, L.R., Kneib, E.C. & Silva, P.C.M. (2006) Proposta 

metodológica para definição da área de influência de pólos 

geradores de viagens considerando características próprias e 

aspectos dinâmicos de seu entorno. Article Published in the J. 

Civil Engng, 27, 111–126. 

Unifra (2009) Manual de Normas para Apresentação de 

Trabalhos Científicos. Available in http://www.unifra.br/ 

cursos/economia/downloads/Normas_UNIFRA_versao1.pdf 

Accessed in 23 December 2009. 


Lins, Cruz, Vieira, C. Neto and Miranda 

74 

J U E E 



ISSN 1982-3932 

doi: 10.4090/juee.2010.v4n2.074080 




DETERMINING INDICATORS OF URBAN HOUSEHOLD 

WATER CONSUMPTION THROUGH MULTIVARIATE 

STATISTICAL TECHNIQUE 

Gledsneli M.L. Lins 1 *, Walter S. Cruz 2 , Zedna M.C.L.Vieira 3 , Francisco A.C. Neto 4 and 

Érico A.A. Miranda 5 

1 Natural Resources Graduation Program, Federal University of Campina Grande, Brazil 

2 Department of Civil Engineering, Federal University of Campina Grande, Brazil 

3 

CAPES/National Postdoctoral Program Scholar, Federal University of Campina Grande, Brazil 

4 Civil Engineering Undergraduate Student, Federal University of Campina Grande, Brazil 

5 Department of Economy, Federal University of Campina Grande, Brazil 

Received 09 September 2010; received in revised form 30 November 2010; accepted 15 December 2010 

Abstract: 

Keywords: 

Water has a decisive influence on populations’ life quality – specifically in areas like 

urban supply, drainage, and effluents treatment – due to its sound impact over public 

health. Water rational use constitutes the greatest challenge faced by water demand 

management, mainly with regard to urban household water consumption. This makes it 

important to develop researches to assist water managers and public policy-makers in 

planning and formulating water demand measures which may allow urban water 

rational use to be met. This work utilized the multivariate techniques Factor Analysis 

and Multiple Linear Regression Analysis – in order to determine the participation level 

of socioeconomic and climatic variables in monthly urban household consumption 

changes – applying them to two districts of Campina Grande city (State of Paraíba, 

Brazil). The districts were chosen based on socioeconomic criterion (income level) so 

as to evaluate their water consumer’s behavior. A 9-year monthly data series (from year 

2000 up to 2008) was utilized, comprising family income, water tariff, and quantity of 

household connections (economies) – as socioeconomic variables – and average 

temperature and precipitation, as climatic variables. For both the selected districts of 

Campina Grande city, the obtained results point out the variables “water tariff” and 

“family income” as indicators of these district’s household consumption. 

Household water consumption, factor analysis, linear regression, urban water 

demand, water efficient use 


* Correspondence to: Gledsneli M. de L. Lins, Tel.: +55 83 3310 1157; Fax: + 55 83 3310 1388 

E-mail: glins16@yahoo.com.br 

Journal of Urban and Environmental Engineering (JUEE), v.4, n.2, p. 47-80, 2010


75 

INTRODUCTION 

The access to potable water is one of the most serious 

problems that urban centres are facing today all over 

the world. Albeit Brazil is privileged in terms of 

freshwater resources – it detains about 14% of the 

world’s water availability – this richness is very 

unequally distributed among its regions: Amazonia 

(North region) concentrates 70%; Central-West 

region, 15%; South and Southeast regions, 6% each 

one; and Northeast region, just 3% of these resources. 

In relation to the main consumptive water uses, 

Brazilian water resources are shared by irrigation 

(63%), human supply (18%), and industry (5%) 

(ANA, 2007). 

Whether these resources are utilized rationally, i.e., 

with economic efficiency, social equity and 

environmental sustainability, in the long run they 

could become a competitive advantage which would 

put Brazil among those countries high on the human 

development index (Lanna, 2008). Nevertheless, 

Brazil’s population and economic growth occurred 

during the 20 th century led to a predatory exploration 

of its natural resources in general, and, particularly, of 

its water resources (ANA, 2007). 

Thus, along with the problems resulting from water 

scarcity in the Northeast region and water pollution in 

the Southeast region, Brazilian’s large urban centres – 

like São Paulo e Recife – as well as medium size ones 

– like Campina Grande, State of Paraíba – have faced 

water supply crises which subjected population to 

tight rationing measures (Tomaz, 2001). This situation 

makes evident the need for integrated water resources 

management, and, more emphatically, the need for 

urban water demand management measures. 

Water demand management (WDM) involves 

actions and sound methods to push the community in 

the direction of an appropriate use of water thus 

reducing water consumption by the final user and, at 

the same time, incrementing new consumption habits 

that would bring no impairment to comfort and 

hygienic necessities as provided by existing systems 

(USEPA, 1998). Then, water demand management 

goes beyond consumption management: rather than a 

question of organizing consumption data and of 

building graphics, WDM insists on studying these 

data and on guaranteeing the system feedback (Lins & 

Ribeiro, 2007). 

The need for urban water demand management 

measures, which includes public policies for 

stimulating household water consumption efficiency, 

exceedingly justifies the development of researches 

that can indicate paths to address this objective. 

Within this context, this work aims to analyze 

possible relationships between socioeconomic 

variables, climatic factors, and household water 

consumption for two districts of Campina Grande, 

Paraíba, Brazil, based on the multivariate statistical 

techniques Factor Analysis and Multiple Linear 

Regression Analysis and considering water 

potentiality and availability, so as to assist water 

managers and public policy-makers in planning and 

regulating water consumption. 

STUDY CASE 

Study area characterization 

Campina Grande is the second largest city in the State 

of Paraíba, with a population of over 380 000 

inhabitants (IBGE, 2010), which is divided along 49 

districts. It is an important educational, industrial and 

technological centre situated in a semiarid region in 

northeastern Brazil. Located in the highest portion of 

the Borborema Plateau’s oriental cliffs, with altitudes 

ranging from 500 to 600 m (average altitude of 

551 m), the city presents a semiarid equatorial 

climate, average temperatures around 25°C, and an 

average annual rainfall of 730 mm (PMCG, 2002). 

Situated within the limits of the Paraíba River basin, 

Campina Grande, however, lies closely to the basin 

boundaries and, consequently, is not served by the 

main river. Figure 1 shows the city’s localization. 

Fig.1 Maps: Brazil and Paraíba (Campina Grande city). 

Water supply has become a historical problem. In 

the period from 1997 to 1999, the severe drought that 

fell upon the northeastern region of Brazil aggravated 

the critical water storage of the Epitácio Pessoa 

Reservoir that supplies the city and brought about a 

very serious crisis in water supply, affecting Campina 

Grande and other five cities in Borborema’s region 

and subjecting 500 000 people to tight rationing 

measures (54% less water in its final phase) between 

November 1998 and April 2000 (Rêgo et al., 2000; 

Galvão et al., 2001). 

In 2001 and 2002, during the dry season, new 

rationing measures had to be adopted motivated by the 

very low storage levels in the reservoir; in 2003, due 



76 

to the same reason, a system of consumption quotas 

was adopted, for a few months, for all consumers; 

reverting this situation, a considerable increase inflow 

into the reservoir, registered in January 2004, allowed 

the accumulation of the reservoir’s maxim storage 

volume (411 686 287 m 3 ) and avoided a new – and 

already planned – water rationing (Vieira et al., 2005). 

An analysis of the many factors that contributed to 

these crises – the city’s geographic location, the lack 

of proper water resources management, the adoption 

of rationing policies only at times of impending crisis, 

among others – indicates the necessity for effective 

water resources management, particularly considering 

the demand side management and its potential for 

inducing rational consumption, in order to avoid the 

occurrence of new crises in the city’s water supply 

(Vieira, 2002). 

Nevertheless, it is important to emphasize that, in 

spite of these crises’ seriousness and except for the 

rationing measures, no other water demand 

management initiatives have been taken in Campina 

Grande. Besides some lack of political will, the need 

for knowledge about household consumption 

characteristics and relationships – which could guide 

public policies formulation – appears as one of the 

factors that have made it difficult to implement urban 

water demand management in Campina Grande. 

As an example of the very few existing works that 

address this city’s necessity, Lins et al. (2008), 

utilized Factor Analysis and Simple Linear Regression 

Analysis (based on data relative to the year 2000) to 

identify the possible interrelations between the 

variable “population per age group” and “household 

water consumption” in Campina Grande city; the 

obtained results indicate that, for the predominant age 

group (0 to 49 years old), the city’s average household 

water consumption (103 L/inhabitant·day) is inferior 

to the present Brazilian average consumption. 

Based on socioeconomic criterion (family income 

level), two districts of Campina Grande city were 

chosen as study area: 

• Dinamérica district, comprising an area of 

1.33 km 2 , a population of 3626 inhabitants 

(90.7% of which are included within the age 

group “0 to 49 years old”) and an average family 

income of US$1,612, which characterizes it as a 

medium income district; and 

• Mirante district, comprising an area of 0.52 km 2 , 

a population of 1056 inhabitants (88% of which 

within the age group “0 to 49 years old”) and an 

average family income of US$6,741, which 

characterizes it as a high income district (IBGE, 

2008–2009 & SEPLAN/CG, 2000). Figure 2 

shows these districts’ location in Campina 

Grande city’s map. 

Fig.2 Map of Campina Grande city, indicating Dinamérica and Mirante districts location. 



77 

METHODOLOGY 

Database 

The following types of monthly data – relating to the 

selected districts – were used in this work, all of them 

referring to the period from 2000 to 2008 (9-year data 

series): socioeconomic (average family income, water 

tariffs applied by the local water supply company, and 

economies, i.e., the number of household connections to 

the public water supply system), climatic (average 

temperature and precipitation), and household water 

consumption. These data were made available by 

SEPLAN/CG (Campina Grande’s Municipal Secretariat 

of Planning), EMBRAPA (Brazilian Enterprise for 

Agricultural Research), and CAGEPA (Water and 

Sewage Company of Paraíba, the local water supply 

company). 

Statistical Techniques 

This work utilized two multivariate statistical 

techniques – Factor Analysis and Multiple Linear 

Regression Analysis – which were made viable by the 

Statistical Package for the Social Sciences – SPSS. 

At first, a Factor Analysis (FA) was carried out 

considering as variables, for both Dinamérica and 

Mirante districts: monthly average temperature, monthly 

average precipitation, economies, monthly water tariffs, 

and monthly family income. 

Factor Analysis is a multivariate statistical analysis 

technique which changes the original variables set into a 

smaller one composed by factors; the order in which 

these factors are obtained corresponds to their 

importance with respect to the quantity of variance of 

the original variables set that each of them can explain. 

This means that the first factor explains the greatest 

quantity of variance of the data set; the second factor 

explains the greatest possible quantity of remainder 

variance; and so on. As this work deals with a 5- 

variable set, the maximum number of factors to be 

extracted is five, i.e., equal to the number of original 

variables. Here, the factoring (extraction of factors) 

method utilized was Principal Component Analysis (PCA). 

On the other hand, as the quantity of variance 

explained by each factor is represented by the quotient 

between the factor’s eigenvalue and the number of 

variables (or the total number of factors), it is 

considered, in general, that only those factors with 

eigenvalues greater than or equal to 1 will take part in 

the final solution. In other words, factors’ eigenvalues 

which are less than 1 are eliminated from the final 

solution (Cruz, 1983). 

Generally, the factors thus extracted are not easily 

interpreted. In order to facilitate their interpretation, it is 

realized a rotation of the coordinate axes which 

graphically represent the factors. According to the 

rotation method adopted – orthogonal or oblique – the 

variables’ coordinates, also called “factors loading,” 

represent the projection of the variables’ variance in this 

new coordinate system. One of the most utilized 

methods is the so-called Varimax rotation which aims to 

maximize the variance projection of each variable on the 

factors. This was the rotation method used in this work. 

The other multivariate analysis technique utilized 

was Multiple Linear Regression Analysis. This 

technique shows the linear combination between a 

variable, called dependent variable, and a set of 

variables called independent or explicative variables 

(Matos, 2000). 

This work adopted the significance level of 5% and, 

for both districts, considered the “household 

consumption” (calculated, for each district, as the 

quotient between the total district water consumption 

and the number of economies) as dependent variable; 

the variables initially suggested as independent 

variables were those which have presented the greatest 

loadings on Factor 1, the most important factor. Based 

on this set of variables thus defined, a Multiple Linear 

Regression Analysis was performed, and, through the 

“stepwise method” (i.e., a method executed step-bystep, 

which includes each independent variable 

according to its contribution for increasing the 

determination coefficient of the linear regression 

equation), the final linear regression equation was 

obtained. 

RESULTS FOR DINAMÉRICA DISTRICT 

Factor Analysis Application 

Table 1 shows the first obtained results for Dinamérica 

district, indicating the initial eigenvalues and the 

percentages of variance for each one of the five factors. 

Based on the criterion of eigenvalues greater than 1, 

two factors were extracted (Factor 1 and Factor 2) 

which explain nearly 76% of the original data set 

variance. 

Following the methodological steps described above, 

the Varimax method application resulted as shown in 

Table 2. In this case, the socioeconomics variables 

(Tariff, Economies and Income) represent a high 

loading on Factor 1, while the climatic variables 

(average temperature T, and average precipitation P) 

represent a high loading on Factor 2. 

Table 1. Total variance explained 

Factor Eigenvalue Variance (%) 

Cumulative 

(%) 

1 2.473 49.469 49.469 

2 1.338 26.768 76.238 

3 0.655 13.109 89.347 

4 0.481 9.629 98.976 

5 5.12.10 -2 1.024 100.000 

Extraction Method: Principal Component Analysis. 



78 

Table 2. Rotated matrix 

Variables Factor 1 Factor 2 

T 6.882 × 10 -2 0.831 

P -2.458 × 10 -2 - 0.825 

Tariff 0.975 9.846 × 10 -2 

Economies 0.860 6.362 × 10 -2 

Income 0.855 -3.907 × 10 -2 

Rotation Method: Varimax rotation. 

Multiple Linear Regression Analysis Application 

The Multiple Linear Regression Analysis between 

“household consumption” and the explicative variables 

extracted from the Factor Analysis (Factor 1) presented 

the final regression equation as expressed by Eq. (1): 

Y = 1.933 + 0.131 X 2 – 8.246×10 -4 X 3 (1) 

where Y is the household consumption (m 3 /month); X 2 is 

the water supply company’s tariff; and X 3 is the monthly 

average family income. 

This final regression equation’s determination 

coefficient R 2 is equal to 0.896, which means that 89.6% 

of the dependent variable Y’s variance are explained by 

the independent variables X 2 (Tariff) and X 3 (Income). 

Table 3 shows the Analysis of Variance (ANOVA) 

for Eq.(1). 

Based on Table 3 analysis, one can verify that 

statistics F presents a value equal to 447. Admitting a 

given level of significance, this F’s value must be 

compared to a value obtained from a “Table of F’s 

values,” which considers the regression and residual 

number of degrees of freedom (df). Only if the F’s value 

presented in Table 3 is greater than any F’s values 

tabulated, for the considered significance level and 

degrees of freedom, it will be possible to say that Eq.(1) 

is statistically significant at that adopted significance level. 

It is important to emphasize that SPSS’ ANOVA 

straightly indicates the value of the smallest significance 

level (the sixth column of Table 3) from which Eq. (1) 

is statistically significant. In other words, Table 3 (at its 

fifth column) indicates the F’s value which is greater 

than any tabulated F for significance levels greater than 

that one presented at its sixth column. 

Since the adopted significance level (5%, i.e., 0.05) 

is greater than the smallest significance level (0.00) 

showed by Table 3, it is possible to affirm that F = 447 

Table 3 – ANOVA 

Model 

Sum of Mean 

df 

squares square 

Regression 21,81 2 10,9 

Residual 2,53 104 ,024 

Total 24,34 106 

df = degrees of freedom; F = Fisher’s statistics; 

Sig = Significance level. 

F 

Sig 

447 .00 

Table 4. Dinamérica district: coefficients of linear regression 

equation model 

Model B t Sig 

Constant 

Tariff 

Income 

1.933 

.131 

-8.246×10 -4 39.746 

27.699 

-15.579 

.000 

.000 

.000 

B = Unstandardized coefficient; t = t-Student; 


is greater than the tabulated value of F for the 5% – 

significance level, and, consequently, the obtained final 

regression equation is statistically significant at this 

adopted significance level. 

Table 4 indicates the constant part, the partial 

regression coefficients, as well as other information 

which are pertinent to the statistical quality analysis of 

the independent variables in Eq. (1). 

As Table 4 shows, the values assumed by statistics t 

are greater than those tabulated for any significance 

level which is greater than 0.000 (result presented by 

SPSS at the fourth column of Table 4). Thus, the partial 

regression coefficients are statistically significant at the 

adopted significance level (5%). 

Therefore, for Dinamérica district, Eq. (1) can be 

considered as a linear regression equation model which 

associates household consumption to the independent 

variables Tariff and Income. 

RESULTS FOR MIRANTE DISTRICT 

Factor Analysis Application 

Following the methodological steps, Table 5 shows the 

first results for Mirante district, i.e., the initial 

eigenvalues and the percentages of variance for each 

one of the five factors. Again, the criterion of 

eigenvalues greater than 1 implied in the extraction of 

two factors, which explain about 79% of the original 

data set variance. 

The rotated matrix (Varimax Method), shown by 

Table 6, presents a result which is similar to that 

obtained for Dinamérica district, i.e., the 

socioeconomic variables (Tariff, Economies and 

Income) presented a high loading on Factor 1, while 

the same occurred with the climatic variable (T and 

P) in relation to Factor 2. 

Table 5. Total variance explained 

Factor Eigenvalue Variance (%) 

Cumulative 

(%) 

1 2.614 52.285 52.285 

2 1.336 26.717 79.002 

3 0.640 12.800 91.802 

4 0.325 6.491 98.293 

5 8.53×10 -2 1.707 100.000 

Extraction Method: Principal Component Analysis. 



79 

Table 6. Rotated matrix. 

Variables Factor 1 Factor 2 

T 5.491×10 -2 0.836 

P - 5.394×10 -2 - 0.817 

Tariff 0.941 0.175 

Economies 0.944 4.598×10 -2 

Income 0.876 -1.992×10 -2 

Rotation Method: Varimax rotation. 

Multiple Linear Regression Analysis Application 

The linear regression analysis was performed by 

considering the district’s household consumption and its 

possible relations to the explicative variables extracted 

from the Factor Analysis (Factor 1). The following final 

regression equation was obtained: 

Y = 5.970 + 0.104X 2 – 6.13×10 -4 X 3 (2) 

where Y is the household consumption (m 3 /month); X 2 is 

the water supply company’s tariff; and X 3 is the monthly 

average family income. 

Equation (2) has a determination coefficient R 2 

equal to 0.912, which means that 91.2% of the 

dependent variable Y’s variance are explained by the 

independent variables X 2 (Tariff) and X 3 (Income). 

Table 7 presents the results of the Analysis of 

Variance (ANOVA). 

The value of statistics F is equal to 537, which is 

greater than any F’s values tabulated for the 5% – 

significance level (in accordance with the explanation 

already given for Table 3 analysis). Consequently, this 

means that Eq. (2) is statistically significant at this 

adopted significance level. 

Table 8 shows the constant part, the partial 

regression coefficients, and other information related to 

the statistical quality analysis of the independent 

variables in Eq. (2). 

Table 7. ANOVA. 

Model 

Regression 

Residual 

Total 

Sum of 

squares 

117.35 

11.36 

128.72 

df 

2 

104 

106 

Mean 

square 

58.68 

.109 

df = degrees of freedom; F = Fisher’s statistics; 


F 

Sig. 

537 .000 

Table 8. Mirante district: coefficients of linear regression equation 

model. 

Model B t Sig 

Constant 

Tariff 

Income 

5.970 

.104 

-6.134×10 -4 55.059 

32.443 

-26.716 

.000 

.000 

.000 

B = Unstandardized coefficient; t = t-Student; 


As statistics t values are greater than those tabulated 

for any significance level greater than 0.000, the partial 

regression coefficients are statistically significant. 

Therefore, for Mirante district, Eq. (2) can be 

considered as a linear regression equation model which 

associates household consumption to the variables 

Tariff and Income. 

CONCLUSION 

Household water consumption is a complex function of 

several factors, amongst which are socioeconomic and 

climate ones. For the study case considered in this work, 

and specifically for the two selected districts, it was 

verified that the variables “average temperature” and 

“average precipitation” did not take part in the final 

solution, due to the fact that none of them have 

presented a high loading on Factor 1, the most important 

factor that has been indicated by Factor Analysis 

application. 

Each obtained linear regression equation, Eq. (1) and 

Eq. (2), respectively for Dinamérica and Mirante 

districts, can be considered as a linear regression model 

which associates household consumption to water tariffs 

and family income. 

Then, the obtained results point out the 

socioeconomics variables (water tariffs and family 

income) as indicators of urban household consumption 

for these Campina Grande city’s districts, so that they 

can be utilized to assist decision-makers and to afford a 

more effective management of this consumption. 

The authors suggest that other investigations must be 

realized, including different districts and/or sectors of 

Campina Grande city, so as to point out linear 

regression models which can incorporate any variability 

with regard to this city’s household water consumption, 

as well as to support urban water demand public 

policies formulation. 

Acknowledgements The third author gratefully 

acknowledges the scholarship provided by the Brazilian 

Agency CAPES, through the National Postdoctoral 

Program (PNPD). 

REFERENCES 

ANA – Brazilian National Water Agency (2007) Relatório nacional 

sobre o gerenciamento da água no Brasil- GEO Brasil Recursos 

Hídricos Available in: http://www.ana.gov.br. Accessed in: 15 

May 2010. 

Cruz, W.S. (1983) Métodos estatísticos multivariados aplicados à 

análise de transportes. MSc Thesis, Universidade Federal da 

Paraíba, Campus II, Campina Grande, Paraíba, Brasil. 

Galvão, C.O., Rêgo, J.C., Ribeiro, M.M.R. & Albuquerque, J.P.T. 

(2001) Sustainability characterization and modeling of water 

supply management practices. IAHS Publ. 268, 81–88. 

IBGE – Instituto Brasileiro de Geografia e Estatística (2010). Dados 

da cidade de Campina Grande – PB. Available in: 

http://www.ibge.org.br/cidades. Accessed in: 20 May 2010. 



80 

IBGE – Instituto Brasileiro de Geografia e Estatística (2008-2009) 

Pesquisa de orçamentos domiciliares. Available in: 

http://www.ibge.org.br/estadosat/. Accessed in: 20 May 2010. 

Lanna, A.E.L. (2008) A economia dos recursos hídricos: os desafios 

da alocação eficiente de um recurso (cada vez mais) escasso. 

Estudos Avançados, 22(2), 113–130. doi: 10.1590/S0103- 

40142008000200008. 

Lins, G.M.L. & Ribeiro, M.M.R. (2007) Gestão da demanda de água 

em centros urbanos do semiárido nordestino. Proc. XVII 

Seminário Brasileiro de Recursos Hídricos, ABRH, São Paulo, 

SP, Brasil. 

Lins, G.M.L., Cruz, W.S. & Miranda, E.A.A (2008) O uso de 

técnicas estatísticas multivariadas na determinação de indicadores 

de consumo doméstico urbano de água. Proc. IX Simpósio de 

Recursos Hídricos do Nordeste, ABRH, Salvador, Brazil, 25–28 

Nov. 

Matos, O.C. (2000) Econometria Básica: teoria e aplicações. São 

Paulo: Atlas. 

PMCG – Prefeitura Municipal de Campina Grande (2002) Perfil do 

município de Campina Grande. Available in: 

http://www.pmcg.pb.gov.br. Accessed in: 27 May 2010. 

Rêgo, J.C., Ribeiro, M.M.R. & Albuquerque, J.P.T. (2000) Uma 

análise da crise de 1998–2000 no abastecimento d’água de 

Campina Grande-PB. Proc. V Simpósio de Recursos Hídricos do 

Nordeste, ABRH, Natal, RN, Brasil. 

SEPLAN/CG – Secretaria de Planejamento de Campina Grande 

(2000) Dados dos bairros – geral/ In: Perfil do município de 

Campina Grande. Available in: http://www.pmcg.pb.gov.br. 

Accessed in: 27 May, 2010. 

Tomaz, P. (2001) Economia de água para empresas e residência: 

Um estudo atualizado sobre o uso racional da água. São Paulo: 

Editora Navegar, 112p. 

USEPA – U.S. Environmental Protection Agency (1998) Water 

Conservation Plan Guidelines. Available in: 

http://www.epa.gov/own. Accessed in: 12 December 2000. 

Vieira, Z.M.C.L. (2002) Análise de conflitos na seleção de 

alternativas de gerenciamento da demanda urbana de água. MSc 

Thesis, Federal University of Campina Grande, 132p. 

Vieira, Z.M.C.L., Braga, C.F.C. & Ribeiro, M.M.R. (2005) Conflict 

analysis as a decision support tool in urban water demand 

management. IAHS Publ. 293, 65–72. 


Martínez and Poleto 

81 

J U E E 



ISSN 1982-3932 

doi: 10.4090/juee.2010.v4n2.081104 




COMPARATIVE ANALYSIS OF EROSION MODELING 

TECHNIQUES IN A BASIN OF VENEZUELA 

Adriana M. Márquez ∗ and Edilberto Guevara-Pérez 

Centre of Environmental and Hydrologic Research, Carabobo University, Venezuela 

Received 20 May 2010; received in revised form 20 October 2010; accepted 16 December 2010 

Abstract: 

Keywords: 

This paper investigates a comparative analysis of erosion modeling techniques 

based on deterministic models, both empirically-based and process-based 

models. The empirical modeling is based on statistical and artificial intelligence 

techniques. In the first, two types of statistical regression model structures are 

investigated, a linear multiple regression model structure and a nonlinear 

multiple regression model structure. In the second, tools such as artificial neural 

networks (ANN) and fuzzy inference system (FIS) are used. The physical 

process-based modeling involves the calibration, validation and testing of the 

models components: WEPP, EUROSEM and CIHAM-UC. The input and output 

variables of models were collected during rainy and dry (irrigation) seasons in 

Chirgua river basin, Venezuela for two years (2008–2009). Ninety-seven rainfall 

storms and 300 irrigation events were measured. Satisfactory fit was found in the 

techniques investigated, R 2 close to 0.7. 

Sediments; hydrological models; soil erosion model; regression model; 

hydrology; runoff; artificial neural networks; fuzzy inference systems 


∗ Correspondence to: Adriana Márquez. Telefax: 0058-414-4165856. 

E-mail: ammarquez@uc.edu.ve 


Márquez and Guevara-Pérez 

82 

INTRODUCTION 

The conventional techniques of erosion modeling are 

empirically and process-based. The empirical models 

include the Universal Soil Loss Equation (USLE) 

(Wischmeier & Smith, 1958), and its latter 

developments, Revised Soil Loss Equation (RUSLE) 

and Modified Universal Soil Loss Equation (MUSLE) 

(Renard et al., 1997a, 1997b). The process-based 

models are represented, e.g., by the WEPP (Water 

Erosion Prediction Project) (Flanagan et al., 2001), 

DWEPP (Dynamic Water Erosion Prediction Project) 

(Bulygina et al., 2007) and EUROSEM (European Soil 

Erosion Model) (Morgan et al., 1998). There have been 

a large number of studies to calibrate and validate the 

runoff and erosion components of process-based models 

(e.g., Flanagan et al., 1995; Klik et al., 1995; Savabi et 

al., 1996; Zhang et al., 1996; Nearing et al., 1998; 

Zeleke, 1999; Ranieri et al., 1999; Santoro et al., 2002; 

Bulygin et al., 2002; Laflen et al., 2004; Silva et al., 

2007; Santos et al., 2003). Many erosion models require 

a large amount of data for calibration and validation as 

well as high performance, meaning, the complex 

algorithms, execution times that eventually cause 

problems, and a programming language necessary to 

minimize the time compute. Additionally, the ANN and 

FIS have been proposed as efficient tools for modeling 

and hydrology forecasting (e.g., Bishop, 1994; Jain & 

Ormsbee, 2002; Panigrahi & Mujundar, 2000). In this 

work, process-based models, statistical regression, 

ANN’s and FIS’s are used to model the furrow erosion 

process. 

It is generally assumed that furrow erosion processes 

are similar to rill erosion processes that occur under 

rainfall. There are similarities in the processes, but there 

are also differences in the conditions (Trout & Niebling, 

1993). Most rainfall erosion occurs during a few highly 

erosive events, while furrow erosion occurs at low-tomoderate 

rates during several irrigations with controlled 

water application. Most irrigation furrows are on slopes 

of less than 3%, while most rill erosion research is 

carried out on slopes greater than 3%. However, in spite 

of the controlled inflows and relatively low slopes, in 

some areas with highly erodible soils, there is 

significant furrow irrigation causing erosion damage 

(Koluvec et al., 1993; Trout, 1999). The present study 

had two main objectives: (1) calibrate, validate and test 

models for the estimation of furrow erosion processes 

and (2) compare experimentally observed furrow 

erosion processes with simulated results by different 

methods. 

MODEL DESCRIPTION 

Water Erosion Prediction Project 

WEPP model was developed by the United States 

Department of Agriculture – Agricultural Research 

Service (USDA-ARS). This model is based on steadystate 

equation assuming a uniform downslope overland 

flow; therefore, variables are expressed on a total width 

or total area basis (Foster & Meyer, 1975; Foster, 1982; 

Foster & Lane, 1987; Foster, 1990). WEPP 

distinguishes erosion processes involved in interrill and 

rill erosion. The basic relationship is given by: 

∂ ( CA) 

∂( 

CQ 

+ 

) = S 

(1) 

∂t 

∂x 

where C is the concentration of sediment in the flow 

(kg/m 3 ), A the cross-sectional area of flow (m 2 ), Q the 

flow discharge (m 3 /s), t the time (s), x the distance 

downslope (m), the term ∂ ( CQ) 

/ ∂x 

the rate of change 

of sediment with the distance, ∂ ( CA) 

/ ∂t 

the storage 

rate of sediment within the flow depth, S the source/sink 

term for sediment (kg/s/m 2 ) is given by: 

S = D I 

+ D R 

(2) 

By assuming quasi-steady sediment movement, Eq. 

(1) is reduced to: 

∂ ( CQ) 

= D I 

+ D R 

∂x 

( wTc 

− CQ) , CQ ≥ wTc 

(3) 

⎛ CQ ⎞ 

D 

⎜1 

⎟ 

c − , CQ ≤ wTc 

= ⎝ wT 

D 

c ⎠ 

R 

(4) 

0.5V 

Q 

s 

where D I is the interrill sediment delivery to the rill 

(kg/s/m 2 ), D R the flow erosion rate (kg/s/m 2 ), T c the flow 

transport capacity in the rill (kg/s/m), w the rill width, V s 

the particle fall velocity (m/s), Q the flow discharge 

(m 3 /s), and D c the flow soil detachment capacity 

(kg/s/m 2 ) is computed as: 

D 

c 

Kr 

( τ − τc 

), 

0 

= , 

τ ≥ τ 

c 

τ ≤ τ 

c 

(5) 

τ = ρ W 

gSR 

(6) 

where K r is the rill erodibility (s/m), τ the flow shear 

stress acting on the soil (Pa), τ c the critical shear stress 

(Pa), ρ w the water density (kg/m 3 ), g the acceleration 

due to gravity (m/s 2 ), S slope (m/m), R hydraulic radius 

(m). Rills are assumed to be rectangular with widths that 

depend on flow rate. 

The simplified form of the equation of Yalin, (1963) 

to estimate the sediment transport capacity has been 

adapted by Foster & Meyer, (1975) as: 

T 

b 

c 

= aτ 

(7) 



83 

where T c is the sediment transport capacity, a and b are 

coefficients obtained by empirical fitting. 

Dynamic Water Erosion Prediction Project 

Bulygina et al. (2007) developed a dynamic version of 

the WEPP model, which quantifies the dynamics of 

sediments within a storm event. This model is based on 

Eqs (1)–(6). Many equations of sediment transport were 

developed for stream flow, and later on applied to 

shallow overland flow and channel flow. According to 

Alonso et al. (1981) and Foster & Meyer, (1972), the 

Yalin equation applies better for shallow flows 

associated with upland erosion. In DWEPP, the Yalin 

flow transport capacity equation is given by: 

⎡ 1 ⎤ 

T c = 0.635Gs 

gd ρwτδ⎢1 

− ln( 1+ β) ⎥⎦ (8) 

⎣ β 

δ = 

β = 

Y 

Ycr 

0, 

−1, 

Y < Y 

cr 

Y ≥ Y 

−0.4 

0. 5 

2 

s cr 

cr 

(9) 

.45G Y δ 

(10) 

τ 

Y = (11) 

ρ ( G −1 gd 

w s ) 

where G s is the particle specific gravity, d the particle 

diameter (m), δ and β are calculated by Eqs (8) and (9), 

respectively, Y is the dimensionless shear stress from 

the Shields diagram and Y cr the dimensionless critical 

shear stress from the Shields diagram. 

European Soil Erosion Model 

The European Soil Erosion Model (EUROSEM) was 

developed by Morgan et al. (1998). It is based on the 

Eqs (1) and (2), which simulate sediment transport, 

erosion and deposition over the land surface by rill and 

interill processes in single storms for both, individual 

fields and small catchments. Model output for one 

single storm includes total runoff, total soil loss, 

hydrographs and sediment graphs. The soil detachment 

by runoff is modeled in terms of a generalized erosiondeposition 

theory proposed by Smith et al. (1995). 

According to it, D R is the flow erosion rate (kg/s/m 2 ), 

(positive for erosion and negative for deposition); the 

flow transport capacity concentration (T c ) represents the 

sediment concentration at which the flow erosion rate 

and corresponding deposition rate are in balance. A 

general equation for flow erosion rate is expressed in 

terms of the settling velocity and flow transport capacity 

concentration, as follows: 

D 

R 

= V ( T − C) 

(12) 

s 

c 

The runoff capacity to transport detached soil 

particles is expressed in terms of the concentration, T c 

(kg/m 3 ) is modeled as a function of unit stream power 

using the relationship of Govers (1985): 

b 

c 

= a( ω− ωc 

(13) 

T ) 

ω = VS 

(14) 

where ω c is the critical value of unit stream power, ω 

the unit stream power (m/s), V s the particle fall velocity 

(m/s), a and b are experimentally derived coefficients 

depending on particle size, S the slope (%) and V the 

mean flow velocity (m/s). 

Erosion model of CIHAM-UC 

At the present, an erosion model is being developed at 

the Centre of Environmental and Hydrologic Research, 

based on Eqs (1)–(3) and Eq. (15) of Simons et al. 

(1981), which is based on power relationships that 

estimate sediment transport based on the flow depth h 

and velocity V. 

These power relationships were developed from a 

computer solution of the bedload transport equation of 

Meyer-Peter & Müller, (1948). Guevara & Márquez, 

(2009) propose it to estimate the sediment transport 

capacity in rill, and can be expressed as follows: 

b c 

T 

c 

= ah V 

(15) 

where T c is the flow transport capacity (kg/s/m), a, b 

and c are parameters estimated by empirical adjustment, 

V the flow velocity (m/s) and h the flow depth (m). 

The flow erosion rate D R (kg/s/m 2 ) is described in 

terms of the generalized theory of erosion-deposition 

modifying proposal by Smith et al. (1995). This 

condition can be expressed as: 

DR = T w − f CV 

(16) 

c 

where V s is the particle fall velocity (m/s), w the furrow 

width (m), f s =A ec /A rill, similarity factor, A ec experimental 

cylinder area from hydrometer test method (D422) 

(American Society for Testing and Materials, 2007). 

Linear Multiple Regression Model 

The Linear Multiple Regression Model (LMRM) used is 

based on the structure given by: 

s 

D P P P P (17) 

R 

= β1 t 

+ β2 

t −1 

+ β3 

t −2 

+ ... + β12 

t−6 

where D R is the flow erosion rate (mg/s/m 2 ) at time t, β s 

represents the regression coefficients to be determined; 

P s represents the rainfalls correspond to defined time 

s 



84 

intervals within a single storm; and t index representing 

time. 

Nonlinear Multiple Regression Model 

The Non-Linear Multiple Regression Model (NLMRM) 

used is based on the structure given by: 

n n 

n 

n 

D 

R 

= β1 P t + β2P 

t−1 

+ β3P 

t−2 

+ ... + β12P 

t−6 

(18) 

where n is the order of polynomial regression. The rest 

of the variables are explained earlier. 

Artificial Neural Networks 

Neural networks are composed of simple elements 

operating in parallel, inspired by biological nervous 

systems. As in nature, the connections between 

elements largely determine the network function. A 

neural network can be trained to perform a particular 

function by adjusting the values of the connections 

(weights) between elements. Typically, neural networks 

are adjusted, or trained, so that a particular input leads 

to a specific target output. Figure 1 illustrates such a 

situation. The network is adjusted, based on a 

comparison of the output and the target, until the 

network output matches the target (Demuth et al., 

2009). 

Target 

In this network, each element of the input vector p is 

connected to each neuron input through the weight 

matrix w. By considering a neuron with a single R- 

element input vector, p, the individual element inputs 

p 1 , p 2 ,…p R , are multiplied by weights w 1,1 , w 1,2 , ... w 1,R 

and the weighted values are fed to the summing 

junction, as shown in Fig. 2. This sum is Wp, the dot 

product of the (single row) matrix W and the vector p. 

The neuron has a bias b, which is summed with the 

weighted inputs to form the net input n. This sum, n, is 

the argument of the transfer function f defined as 

follows: 

n 

= , 

w1 ,1 p1 

+ w1,2 

p2 

+ ... + w1 

R pR 

+ b (19) 

This expression can be written as: 

n ( i) 

= Wp + b 

(20) 

where, the ith neuron has a summing that gathers its 

weighted inputs and bias to form its own scalar output n 

(i). The various n(i) taken together form an S-element 

net input vector n. 

The sigmoid transfer function is commonly used in 

backpropagation networks, in part because it is 

differentiable. As shown in Fig. 3, the sigmoid transfer 

function takes the input, which can have any value 

between plus and minus infinity, and squashes the 

output into the range 0 to 1. 

Input 

Adjust 

Weights 

Neural 

Network 

Output 

Compare 

a 

+1 

Fig. 1 Scheme Input/Output to ANN. 

0 

n 

Neuron Model 

One or more of the neurons can be combined in a layer, 

and a particular network could contain one or more such 

layers. A one-layer network with R input elements and S 

neurons is shown in Fig. 2 as follows: 

p 1 

p 2 

p R 

Inputs 

w 1,1 

w 1,R 

Layer of Neurons 

Σ 

1 

Σ 

1 

Σ 

1 

b 1 

b 2 

b s 

Fig. 2 Single layer network architecture of neurons. 

n 1 

n 2 

n s 

f 

f 

f 

a = f (Wp + b) 

a 1 

a 2 

a s 

-1 

a = logsig(n) 

Fig. 3 Log-Sigmoid transfer function. 

a 

+1 

-1 

a = purelin(n) 

Fig. 4 Lineal transfer function. 

Occasionally, the Linear Transfer Function Purelin is 

used in backpropagation networks. The linear transfer 

function is illustrated in Fig. 4. 

The symbol in the square to the right of each transfer 

function graph shown represents the associated transfer 

function. These icons replace the general f in boxes of 

network diagrams to show the particular transfer 

function being used. 

n 



85 

Feedforward Network 

Feedforward networks often have one or more hidden 

layers of sigmoid neurons followed by an output layer 

of linear neurons. Multiple layers of neurons with 

nonlinear transfer functions allow the network to learn 

nonlinear and linear relationships between input and 

output vectors. The linear output layer lets the network 

produce values outside the range –1 to +1. To describe 

networks having multiple layers, the notation must be 

extended. Specifically, it needs to make a distinction 

between weight matrices that are connected to inputs 

and weight matrices that are connected between layers. 

A network can have several layers and it uses layer 

weight (LW) matrices as well as input weight (IW) 

matrices. The network shown in Fig. 5 has R inputs, S 1 

neurons in the first layer, and S 2 neurons in the second 

layer. A constant input 1 is fed to the bias for each 

neuron. Note that the outputs of each intermediate layer 

are the inputs to the following layer. Thus layer 2 can be 

analyzed as a one-layer network with S 1 inputs, S 2 

neurons, and an S 2 ×S 1 weight matrix W 2 . The input to 

layer 2 is a 1 ; the output is a 2 . Now that all the vectors 

and matrices of layer 2 have been identified, it can be 

treated as a single-layer network on its own. This 

approach can be taken with any layer of the network. 

Fuzzy Logic 

Input 

Inputs Terms 

Fig. 6 Diagram of a fuzzy inference system. 

Fuzzy logic has two different meanings. In a narrow 

sense, fuzzy logic is a logical system, which is an 

extension of multivalued logic. However, in a wider 

sense fuzzy logic (FL) is almost synonymous with the 

theory of fuzzy sets, a theory which relates to classes of 

objects with unsharp boundaries, in which membership 

is a matter of degree. The point of fuzzy logic is to map 

an input space to an output space, and the primary 

mechanism for doing this is a list of if-then statements 

called rules. All rules are evaluated in parallel, and the 

order of the rules is unimportant. The general 

description of a fuzzy system is shown in Fig. 6 

(MATLAB, 2010). 

Fuzzy Inference System 

Rules 

Output 

Outputs Terms 

Input 

p 

Rx1 

1 

R 

I 

W 

S 1 x 

R 

b 1 

S 1 x 

Hidden Layer 

+ 

n 

1 

S 1 x1 

S 1 

a 1 

S 1 x1 

Output 

Layer 

S 2 x1 

a 2 =purelin (LW 2, 1 

a 1 +b 2 ) 

Fig. 5 Archictecture of a feedforward network. 

1 

L 

W 

S 2 x 

S 1 

b 

+ 

n 

2 

S 2 x 

1 

a 2 =y 

S 2 x1 

S 2 

Fuzzy inference is the process of formulating the 

mapping from a given input to an output using fuzzy 

logic. The process of fuzzy inference involves three 

components: (1) membership functions, (2) logical 

operations and (3) If-Then rules. (1) A membership 

function (MF) is a curve that defines how each point in 

the input space is mapped to a membership value (or 

degree of membership) between 0 and 1. Membership 

functions commonly used are as follows: the piece-wise 

linear functions, the Gaussian distribution function, the 

sigmoid curve, quadratic and cubic polynomial curves, 

(2) the logical operations, one particular 

correspondence between two-valued and multivalued 

logical operations can be defined by the fuzzy 

intersection or conjunction (AND), fuzzy union or 

disjunction (OR), and fuzzy complement (NOT). (3) If- 

Then rules are used to formulate the conditional 

statements that comprise fuzzy logic. 

Table 1. Description of study area 

Subbasin 

Geographic Location Area Land Use 

Zone 

Agriculture Avian 

North Coordinate West Coordinate (ha) 

Domestic % Other 

% 

% 

High 

I 10º 13′ 55″ 10º 15′ 00″ 68º 12′ 10″ 68º 11′ 05″ 244.38 95.30 2.70 2 0 

II 10º 13′ 00″ 10º 14′ 00″ 68º 11′ 10″ 68º 12′ 00″ 209.21 87.99 4.46 6.13 1.42 

Medium III 10º 10′ 10″ 10º 11′ 50″ 68º 11′ 10″ 68º 10′ 20″ 320.64 93.04 0.74 6.22 0 

IV 10º 11′ 50″ 10º 12′ 20″ 68º 11′ 30″ 68º 10′ 30″ 162.58 77.78 5.56 16.66 0 

Low V 10º 12′ 25″ 10º 13′ 10″ 68º 11′ 10″ 68º 11′ 50″ 131.74 83.33 5.56 16.67 0 

VI 10º 12′ 18″ 10º 15′ 45″ 68º 11′ 67″ 68º 15′ 07″ 250.64 94.04 0.84 7 0 

Percentage Average 88.58 3.31 9.11 0.24 

Total 1319.19 1169 44 120 3.16 



86 

Two types of fuzzy inference systems can be 

implemented: Mamdani-type and Sugeno-type. 

These two types of inference systems vary somewhat 

in the way outputs are determined. In this 

investigation Sugeno-type system is used, which can 

be used to model any inference system in which the 

output membership functions are either linear or 

constant. 

Fuzzy Modeling Scenario 

To apply fuzzy inference to a system for which a 

collection of input/output data models, a model 

structure based on characteristics of variables in a 

system does not necessarily have been 

predetermined. In some modeling situations, the 

membership functions cannot be discerned from 

looking at data. Rather than choosing the parameters 

associated with a given membership function 

arbitrarily, these parameters could be chosen so as to 

tailor the membership functions to the input/output 

data in order to account for these types of variations 

in the data values. In such cases, neuro-adaptive 

learning techniques incorporated in the ANFIS 

command of MATLAB can be used. The acronym 

ANFIS derives its name from adaptive neuro-fuzzy 

inference system. Using a given input/output data 

set, the toolbox function ANFIS constructs a fuzzy 

inference system (FIS) whose membership function 

parameters are tuned (adjusted) using either a 

backpropagation algorithm alone or in combination 

with a least squares type of method. This adjustment 

allows to fuzzy systems to learn from the data they 

are modeling. 

FIS structure and parameter adjustment A 

network-type structure similar to that of a neural 

network, which maps inputs through input 

membership functions and associated parameters, 

and then through output membership functions and 

associated parameters to outputs, can be used to 

interpret the input/output map. The parameters 

associated with the membership functions changes 

through the learning process. The computation of 

these parameters (or their adjustment) is facilitated 

by a gradient vector. This gradient vector provides a 

measure of how well the fuzzy inference system is 

modeling the input/output data for a given set of 

parameters. When the gradient vector is obtained, 

any of several optimization routines can be applied 

in order to adjust the parameters to reduce some 

error measure. This error measure is usually defined 

by the sum of the squared difference between actual 

and desired outputs. ANFIS uses either back 

propagation or a combination of least squares 

estimation and backpropagation for membership 

function parameter estimation. 

MATERIALS AND METHODS 

Measurements were made in Chirgua River Basin, 

located in the north central region of Venezuela during 

2008–2009. Each subbasin was divided into zones, 

which were classified as follows: high: I and II, 

medium: III, low: IV, V and VI. Each zone is mainly 

used for agricultural purposes. Information on the 

location, area and land use is shown in Table 1. Five 

fields were selected with the following slopes in tillage 

orientation: 0.008 ± 0.0055 m/m; 0.01 ± 0.00197 m/m; 

0.015 ± 0.0006 m/m; 0.025 ± 0.0033 m/m and 0.13 ± 

0.0156 m/m, whose soils range from silty sand to silty 

clay. Two types of crops usually are grown in rotation, 

according to the season: dry (potato: Solanum 

Tuberosum) and rainy (corn: Zea Mays); whose 

development cycles have the following time periods: 12 

and 18 weeks, respectively. 

The irrigation events for the whole cycle lasted as 

follows: two hours for the first ten weeks and one hour 

for the last two weeks. A sprinkler irrigation system was 

used to provide water to crops. The inflow rate that was 

applied to furrows varied between 0.4 and 0.6 L/s. A 

plot of 64 wheel-compacted furrows on each field was 

split into eigth, 8-furrow blocks. The tests were carried 

out in three furrows with a width and length as varied as 

follows: 0.3 and 0.35 m; 100 and 200 m, respectively. 

Furrows were divided into four equal-length sections 

(¼, ½, ¾, and at the end of the furrow), and 

measurement stations were established at the 

downstream end of each section. Flows were measured 

using a steel plate with a 60º V-notch weir, by applying 

the volumetric method. The number of measurements 

carried out during the dry and rainy seasons were 300 

and 162, respectively, which included: a) Soil physical 

properties: undisturbed samples were captured to 

determine through laboratory analysis saturated 

hydraulic conductivity, specific gravity, shear stress, 

moisture content, Atterberg limits and particle size 

(hydrometer and sieve analysis). Infiltration rates were 

measured with a single-ring infiltrometer (diameter 30 

cm), which was driven up to 5 cm deep into the soil. 

(Table 2 and Fig. 7), b) Flow physical properties: flow 

rates and sediment concentrations were measured at 

time intervals ranging from 20 to 120 min at a time, 

after the arrival of water to each measuring station. 

Sediment concentration was determined by applying the 

2 540 B method (American Public Health Association, 

1995). 120 irrigation events were measured in five 

different agricultural fields (5 fields × 24 events / field = 

120 irrigation events). During the rainy season the 

following events were measured: 73 for 2 hours, 24 for 

an hour and 20 for half an hour, with a precipitation 

gauge attached to an electronic storage unit, which has 

following geographic location: °W 68º 10’ 50’’and 

°N 10º 11’ 45’’. Total data were divided into three 



87 

groups: 60% for calibration, 20% for validation and 

20% for testing. 

Models data 

Table 2 shows soil physical properties. Soil is 

composed of soil particles, where approximately less 

than 3% represents < 2.00 mm, 46.1% represents < 

0.074 mm (fine fraction) and 50.9% represents >0.074 

mm (finest fraction: silt and clay). Saturated hydraulic 

conductivity is close to 1 mm/h in most areas, which is 

classified as low to lower (Terzaghi & Peck, 1967). The 

friction angle varies from 29.8 to 32.7º. The cohesion 

varies from 0.1 to 0.43 kgf/cm 2 . The initial and final 

moisture contents of the soil vary from 7.1 to 21.8%; 

from 26.1 to 54%, respectively. The porosity varies 

from 0.4 to 0.5 cm 3 /cm 3 . The liquid limit, plastic limit 

and plasticity index range respectively as follows: 26.2 

and 33.2%; 23 and 28%; 4.3 and 5.2%. Specific gravity 

of solids varies from 2.5 to 2.6. Soil type varies from 

organic silt (OL) to plasticity low silty sand (SM) (e.g., 

Lambe & Whitman, 1972; Juarez & Rico, 1991) (Table 

2). 

Infiltration rates for silty soil are highly variable. 

Figure 7(a) shows the cumulative infiltration versus 

time for 582 infiltrometer tests over one period, which 

last 2.28 h each. Infiltration depths at the end of this 

time have a mean of 22.39 mm and a standard deviation 

of 21.02 mm. Infiltration depths greater than 22.39 mm 

are measured during the rainy season, where rainfall 

events occur less frequently than those of irrigation, 

therefore the storage capacity of the soil is higher than 

those available during the irrigation period. Figure 7(b) 

shows the histogram for the cumulative infiltration at 

t = 2.28 h, together with the corresponding probability 

density functions. Exponential and Weibull functions 

result in the best fit. 

Table 2. Soil physical properties 

Soil physical 

Zone 

Properties I II III IV V VI 

µ σ CV µ σ CV µ σ CV µ σ CV µ σ CV µ σ CV 

Texture 

PRetained Percentage on Sieve 

>9.51mm 

(Sieve 3/8″) 3.9 4.4 112.9 0.0 0.0 - 1.2 1.1 93.2 1.2 1.5 126.5 0.0 0.0 - 0.3 0.5 146.9 

9.51 – 4.76 mm 

(Sieve # 4) 2.6 1.9 70.1 4.0 4.5 110.6 1.3 0.8 63.3 1.4 1.3 94.1 0.9 0.3 38.2 2.2 2.1 96.8 

4.76 – 2.00 mm 

(Sieve # 10) 4.0 2.5 61.8 5.5 5.3 95.3 2.7 1.1 39.0 2.7 1.5 55.9 3.7 2.8 75.9 0,7 1.3 188.2 

2.00 – 0.84 mm 

(Sieve # 20) 6.7 3.6 54.0 10.9 7.8 71.7 5.7 2.1 36.5 5.6 2.0 34.9 8.3 4.1 49.2 7.2 3.5 48.2 

0.84 – 0.42 mm 

(Sieve # 40) 10.9 5.2 47.3 14.5 6.4 44.2 8.7 2.8 32.1 8,3 1.8 22.1 13.8 3.9 28.7 10.4 3.7 36.1 

0.42 – 0,149 mm 

(Sieve # 100) 22.6 5.2 23.0 31.2 9.8 31.6 15.7 2.9 18.4 18.8 3.1 16.4 19.0 3.3 17.5 32.5 8.0 24.7 

0.149 – 0.074 mm 

(Sieve #200) 14.4 6.3 43.8 28.3 12,6 44.6 13.0 3.4 26.1 15.1 2.5 16.6 15.7 5.1 32.2 28.5 11.5 40.2 

< 0.074 mm 

Plate 34.9 9.3 26.6 5.9 3.5 60.3 51.7 9.5 18.4 48.2 8.6 17.9 38.7 7.8 20.2 18.3 12.0 65.8 

Total Percentage Average 

Sand (%) 46.4 4.2 9.1 30.7 5.6 18.4 46.1 1.6 3.5 46.2 2.1 4.6 46.1 1.6 3.5 46.1 1.6 3.5 

Silt (%) 49.5 2.7 5.5 36.7 11.7 31.9 46.9 2.3 4.9 46.9 2.3 4.9 49.5 2.7 5.5 46.9 2.3 4.9 

Clay (%) 4.1 2.0 48.4 19.9 8.2 41.3 7.0 2.4 33.9 6.9 2.5 36.2 6.6 1.8 27.6 7.0 2.4 33.9 

Hydraulic Properties 

K s ¹ (mm/h) 0.4 0.1 33.3 24.0 22.7 94.3 0.9 0.6 60.6 1.4 0.6 41.8 0.9 0.6 60.6 0.9 0.6 60.6 

Structural Properties 

Friction Angle (º) 29.9 3.9 12.9 29.8 1.1 3.8 29.8 1.1 3.8 32.7 1.1 3.4 31.8 1.9 5.9 32.7 1.1 3.4 

Cohesion 

(kgf/cm 2 ) 0.3 0.0 15.4 0.1 0.0 60.0 0.1 0.0 60.0 0.3 0.0 6.7 0.43 0.03 7.0 0.3 0,0 6.7 

Phase Relations 

Initial moisture 

content (%) 21.8 11.5 52.6 7.1 3.5 49.2 18.9 11.8 62.5 15.0 7.1 47.2 20.7 10.6 51.3 18.9 11.8 62.5 

Final moisture 

content (%) 26.1 11.5 44.0 51.0 11.7 22.9 40.5 7.7 19.0 54.0 12.6 23.4 40.5 7.7 19.0 40.5 7.7 19.0 

Porosity 

(cm 3 /cm 3 ) 0.4 0.1 13.5 0.4 0.1 11.6 0.4 0.1 14.6 0.4 0.1 18.9 0.5 0.1 16.7 0.4 0.1 14.6 

Liquid limit (%) 26.2 2.5 9.4 30.7 4.1 13.3 32.4 4.4 13.4 29.6 2.0 6.8 33.2 4.2 12.8 32.4 4.4 13.4 

Plastic limit (%) 23.1 5.0 21.6 26.2 3.7 14.0 27.1 2.9 10.6 25.2 3.4 13.6 28.3 3.9 13.9 27.1 2.9 10.6 

Plasticity index 

(%) 5.1 2.2 43.8 4.5 2.5 54.3 5.2 4.0 76.0 4.3 2.6 60.0 4.9 1.2 25.0 5.2 4.0 76.0 

Specific gravity 2.6 0.1 3.9 2.5 0.1 4.8 2.5 0.1 4.3 2.6 0.2 7.7 2.7 0.1 3.7 2.5 0.1 4.3 

¹ Satured hydraulic conductivity, µ: mean, σ: standard deviation, CV: coefficient of variation in percentage (%) 



88 

Fig. 7 Cumulative infiltration depth: (a) temporal variation, and (b) frequency histogram and predicted Weibull probability 

density function. 

Discharge, Q (L/s) 

Time (h) 

Fig. 8 Measured hydrographs for different slope furrows during each irrigation. 

During the dry season, runoff is stabilized after the 

first half hour when the irrigation is applied. The time 

for flow stabilization is approximately one hour; which 

occurs faster in the furrows where the slope was equal 

to 0.8% (Fig. 8). Flow rate at steady state varies from 

0.3 to 0.5 L/s, decreasing from 0.05 to 0.1 L/s after the 

seventh irrigation. The effect of the coverage radio of 

the sprinkler on flow varying between the different 

furrows is low within an irrigation event. 

A comparison of the sediment load during irrigation 

and rain events of one and two hours is shown in Fig. 9. 

The maximum sediment load observed during the 

irrigation and rainfall events of one hour varies 

approximately from 100 to 400 mg/s, from 600 to 

100 mg/s, respectively (Fig. 9(a)). The maximum 



89 

sediment load observed for the two-hour event varies 

from 500 to 2000 mg/s for irrigation and from 2500 to 

7000 mg/s for rain (Fig. 9(b)). Therefore, there is 

significant variation in sediment load between the 

events of rain and irrigation, where the first is 5 to 10 

times more erosive than the latter. 

A comparison of rainfall events in both one and two 

hours is shown in Fig. 10, which shows that the rainfall 

for these events ranges as follows: from 20 to 30 mm, 

and from 25 to 40 mm, respectively (Figs 10(a) and 

(b)). 

Fig. 9 Measured sediment load during rainfall and irrigation events. 

Cumulative Precipitation, P (mm) 

Time (h) 

Fig. 10 Precipitation depth. 



90 

RESULTS 

Modeling of furrow erosion based on physical 

processes 

Process of particle detachment 

Table 3 shows the parameters of Eq. (5), used to 

estimate the capacity of soil particle detachment in 

furrows. In general, the following statements can be 

made about the parameters K c and τc: (a) K c varies from 

1.31E–06 to 1.525E–06 s m –1 ; (b) τc varies from 0.704 

to 1.489 Pa; (c) estimate standard error of the 

parameters is low in all cases. 

Table 4 shows fit statistics of Eq. (5) to the 

observations obtained from field testing in the furrows 

of different slopes. The results are summarized as 

follows: R 2 is equal to 0.682, R 2 adj. is equal to 0.681, its 

reduction is not significant in relation to R 2 . Mallows C p 

decreases slightly compared to the number of 

independent variables in Eq. (5). Durbin-Watson 

statistic is equal to 0.66. By comparing the errors during 

the calibration and validation stages, with emphasis on 

the Average Absolute Relative Error (AARE), and the 

Average Relative Error (ARE) the following results are 

found: AARE varies from 47.87 to 57.9, ARE varies 

from –20.52 to –33.74. In general, the errors do not vary 

significantly between the stages of calibration and 

validation. The rest of errors can be seen in Table 4. 

Process of sediment transport 

Table 5 shows the parameters of Eqs 7, 13 and 15 for 

estimating the sediment transport capacity in furrows. In 

general, the following statements can be made about the 

parameters K t , b and c: Coefficient of transport (K t ): 

(a) the units differ between Eqs 7, 13 and 15, (b) the 

ranges found for Eqs 7, 13 and 15 vary as follows: 

9.929E–08 and 1.526E–07; 7.634 and 34.583; 0.421 and 

0.892, respectively, (c) estimate standard error of 

parameters is moderately low in all cases. Parameter b: 

(a) the ranges found for Eqs 7, 13 and 15 vary as 

follows: 1.519 and 1.705; 0.398 and 1.299; 2.087 and 

2.228; respectively, (b) estimate standard error of 

parameters is low in all cases. Parameter c: (a) the 

range found for Eq. (15) varies from 1.097 to 1.193; (b) 

estimate standard error of parameters is low in all cases. 

Table 6 shows fit statistics of Eqs 7, 13 and 15 to 

the observations obtained from field testing in the 

furrows. R 2 : varies from 0.72 to 0.88. R 2 adj.: its 

reduction is not significant in relation to R 2 . Mallows 

C p : decreases slightly compared to the number of 

independent variables in each equation. Durbin-Watson 

statistic: values found to Eqs 7, 13 and 15 are equal to 

0.719, 1.09 and 0.74, respectively. By comparing the 

errors during the calibration and validation stages with 

emphasis on the Average Absolute Relative Error 

(AARE) and the Average Relative Error (ARE) the 

following results in Eqs 7, 13 and 15 are found: AARE 

varies from 56.319 to 63.081; from 12.125 to 13.482; 

from 29.585 to 29.894. ARE varies from –13.365 to 

–1.831; from –1.831 to 1.974; from –16.597 to –18.547. 

In general, the errors do not vary significantly between 

calibration and validation stages. The rest of errors can 

be seen in Table 6. 

Combination of furrow erosion processes 

The representation of furrow erosion processes for 

following slopes: 0.8%, 1%, 1.5% and 2.5% is shown in 

Fig. 11, using the Eqs 7, 13 and 15 for estimating the 

transport capacity of sediments (Figs 11a, 11c, 11e and 

11h, and Eqs 4 and 12 to estimate the net detachment 

capacity (Figs 11b, 11d, 11f and 11g). The cycle under 

irrigation has been divided into three stages; each stage 

represents 1/3 of the total duration for 12 weeks, 

obtaining data average values for each phase. 

Table 3. Regression coefficients of particle detachment capacity model 

Eq. Variable Parameter Unit Average Standard Error Minimum Limit Maximum Limit 

5 D c K r s/m 0.000 001 418 5.43 497E–8 0.00 000 131 163 0.000 0015 255 

(kg/s/m 2 ) τ c Pa 1.09 707 0.199 697 0.704 172 1.48 997 

Table 4. Performance criteria of particle detachment capacity model 

Eq. p n R 2 (R 2 ) adj. C p SEE d MSSE AAE AARE AE ARE n AARE ARE 

(a) During Calibration 

(b) During Validation 

5 1 320 0.682 0.681 0 2.45E–6 0.66 6.04E–12 2.08E–06 47.87 –7.72E–13 –20.52 84 57.9 –33.74 

Eq.: equation; p: number of independent variables in the model; n: number of observed data; R 2 : determination coefficient; (R 2 ) adj. : adjusted 

determination coefficient; C p : coefficient of Mallows; SEE: Standard error of estimate; d: statistic of Durbin-Watson; MSSE: mean of sum of 

squared error; AAE: average absolute error; AARE: average absolute relative error (%); AE: average error; ARE: average relative error (%). 



91 

Fig. 11 Comparisons of simulated furrow erosion processes with observed data. 



92 

Table 5. Regression coefficients of sediment transport capacity model 

Eq. Variable Parameter Unit Average Standard Error Minimum Limit Maximum Limit 

7 T c K t kg 1–b m –(1–b) s –(1–2b) 1.2598E–7 1.35 793E–8 9.92 906E–8 1.52 669E–7 

kg/m/ s b 1.61 238 0.0 473 371 1.51 934 1.70 542 

13 T c K t kg m –(b+1) s –(1–b) 21.1093 6.843 7.634 34.583 

b 0.8489 0.2287 0.3985 1.2993 

kg/m/s ω c m s –1 –0.05 414 0.0221 –0.0978 –0.0104 

15 T c K t kg m (–1–b–c) s –(1+b) 0.657 032 0.120 287 0.421 274 0.89 279 

b 2.15 807 0.0 359 035 2.0877 2.22 844 

kg/m/ s c 1.14 551 0.0 244 874 1.09 751 1.1935 

Table 6. Performance criteria of sediment transport capacity model 

Eq. p n R 2 R 2 adj C p SEE d MSSE AAE AARE AE ARE n AARE ARE 



7 1 437 0.73 0.732 0 1.102E–06 0.719 1.22E–12 9.16E–07 63.08 4.24E–08 –31.831 100 56.31 –13.36 

13 1 320 0.72 0.720 0 0.34 1.09 0.115 0.287 12.12 0.0028 –1.831 96 13.48 1.974 

15 2 700 0.88 0.882 1 1.644E–06 0.741 2.70E–12 1.08E–06 29.58 –5.75E–08 –16.597 165 29.89 –18.54 

Comparison between the transport capacities of 

sediments observed and estimated by Eqs 7, 13 and 15 

shows in Figs 11a, 11c, 11e and 11h. The solid line 

represents the observed sediment load in furrows, while 

the bars represent estimates by Eqs 7, 13 and 15, where 

the following observations can be made: (a) Eq. (13) 

overestimated the values of the observed transport 

capacity, but the rest of equations underestimated them 

and (b) the observed sediment load in stages from 1 to 3 

varies as follows: from 1000 to 2500 mg/s, from 2000 to 

500 mg/s; less than 500 mg/s. 

Values estimated by Eq. (7) in stages from 1 to 3 

vary as follows: from 500 to 1000 mg/s, from 0 to 500 

mg/s, less than 500 mg/s. Values estimated by Eq. (13) 

in stages from 1 to 3 vary as follows: from 1000 to 2500 

mg/s, from 2000 to 500 mg/s, less than 500 mg/s. 

Estimates from Eq. (15) in steps 1 through 3 vary as 

follows: from 1500 to 3500 mg/s, from 500 to 2000 

mg/s, from 500 to 1000 mg/s, (c) the difference between 

observed and estimated values is significant in the first 

stage of the cycle and decreases towards the second and 

third stages. 

According to the theory of models based on physical 

processes, when the sediment transport capacity is 

exceeded by the flow sediment load occurs the 

deposition process (Foster and Meyer, 1975, Morgan et 

al., 1998). 

Deposition process estimated by Eqs 4 and 12 is 

shown in Figs 11(b), 11(d), 11(f) and 11(g), where the 

following results determine: (a) most times Eq. (12) 

estimates higher values than Eq. (4), (b) the estimated 

values by Eq. (4) in the stages from 1 to 3 vary as 

follows: from 2 to 6 kg/m 2 , from 0 to 4 kg/m 2 , from 0 to 

2 kg/m 2 , respectively. Estimates from Eq. (12) in stages 

from 1 to 3 vary as follows: from 6 to 12 kg/m 2 , from 2 

to 6 kg/m 2 , from 2 to 4 kg/m 2 , (c) the difference 

between observed and estimated values is significant 

during the first stage of the cycle and decreases towards 

the second and third stages. 

Modeling furrow erosion based on regressions 

Table 7 shows the parameter values in the furrow 

erosion models based on linear regressions, Eqs 17(a) 

and (b), as well as nonlinear regressions, Eqs 18(a)–(e). 

In general, the following statements can be made in 

relation to the parameters β 1 to β 6 : in Eq. (17a) the 

parameters vary from –4.267 to 4.382, in Eq. (17b) they 

vary from –7.5 to 9.14, in Eq. (18a) they vary from – 

0.031 and 0.2697, in Eq. (18b) they vary from –0.014 to 

0.093, in Eq. (18c) they vary from –0.629 to 1.938, in 

Eq. (18d) they vary from –0.046 to 1.012. 

Table 8 shows the fitted statistics of erosion models 

based on regression from the observations obtained 

when the field was examined in furrows, in the 

following slopes: 0.8%, 1%, 1.5%, and 2.5%. R 2 : varies 

from 0.90 to 0.53, and it decreases when the polynomial 

power is increased. R 2 adj its reduction is not significant 

in relation to R 2 . Mallows C p : decreases slightly 

compared to the number of independent variables in 

each equation. Durbin-Watson Statistics: varies 

between 0.7 and 1.68. By comparing the errors during 

the calibration and validation stages, with emphasis on 

the average absolute relative error (AARE), and the 

average relative error (ARE) the following results are 

found: AARE varies from 6.21 to 12.57, ARE varies 

from –0.71 to 6.78. In general, (a) errors tended to be 

lower in the calibration stage with respect to validation 

and (b) errors increase when the polynomial power is 

increased. The rest of errors can be seen in Table 8. 

Figure 12 shows a comparison of the estimated 

erosion results using Eqs 17(a), 17(b), 18(a), 18(b), 

18(c) and 18(d) to the observations under irrigation in 

the furrows with following slopes: 0.8%, 1%, 1.5% and 

2.5%. Additionally, it includes a comparison with the 

physical process model represented by Eq. (16), and the 

artificial intelligence technique based on neural 

networks, which will be widely discussed in the next 

section. 



93 

Table 7. Regression coefficients of erosion model based on LMR and NLMR 

Eq. Dependent 

variable 

Independent 

variable 

Parameter Unit Average Standard 

Error 

Minimum 

Limit 

Maximum 

Limit 

17(a) D Pt β mg m –2 s –1 mm –1 1.398583862 1.442760413 –4.267676785 1.47050906 

- 

LMRM 

r 

mg m –2 s –1 Pt-1 

1 

β 2 mg m –2 s –1 mm –1 0.833406797 1.784933149 –2.716135469 4.382949064 

n=1 Pt-2 β 3 mg m –2 s –1 mm –1 0.731227333 1.392642416 –2.03820026 3.500654927 

D=1H Pt-3 β 4 mg m –2 s –1 mm –1 0.183129882 1.814769332 –3.425745022 3.792004785 

Pt-4 β 5 mg m –2 s –1 mm –1 0.737812187 1.250395401 –1.748741065 3.224365439 

Pt-5 β 6 mg m –2 s –1 mm –1 0.559520883 0.731816989 –0.895780305 2.014822071 

17(b) D r Pt β 1 mg m –2 s –1 mm –1 0.794050629 1.4177439 –3.585879266 1.997778008 

- 

LMRM mg m –2 s –1 Pt-1 β mg m –2 s –1 mm –1 2.195990758 2.69409942 –7.5012253 3.109243783 

- 

n=1 Pt-2 

2 

β 3 mg m –2 s –1 mm –1 0.780383734 3.769700299 –6.642929634 8.203697103 

D=2H Pt-3 β 4 mg m –2 s –1 mm –1 2.177019018 1.357605748 –0.496385258 4.850423294 

Pt-4 β 5 mg m –2 s –1 mm –1 4.541647609 2.338266456 –0.062878815 9.146174033 

Pt-5 β 6 mg m –2 s –1 mm –1 0.099997046 0.333760338 –0.557245543 0.757239634 

18(a) D r Pt β 1 mg m –2 s –1 mm –1 0.119031048 0.073909116 –0.031707895 0.269769991 

NLMR 

M mg m –2 s –1 Pt-1 β 2 mg m –2 s –1 mm –1 0.055276888 0.084390798 –0.116839631 0.227393406 

n=2 Pt-2 β 3 mg m –2 s –1 mm –1 0.066816629 0.061319061 –0.058244676 0.191877933 

D=1H Pt-3 β 4 mg m –2 s –1 mm –1 0.050906233 0.084917164 –0.122283818 0.224096283 

Pt-4 β 5 mg m –2 s –1 mm –1 0.075651656 0.055759098 –0.038070006 0.189373318 

Pt-5 β 6 mg m –2 s –1 mm –1 0.05458908 0.025644464 0.002286744 0.106891417 

18(b) D r Pt β 1 mg m –2 s –1 mm –1 0.065622334 0.013487094 0.038182579 0.093062088 

NLMR 


n=3 Pt-2 β 3 mg m –2 s –1 mm –1 0.002274855 0.007430699 –0.012843047 0.017392756 

D=1H Pt-3 β 4 mg m –2 s –1 mm –1 0.01600253 0.014611824 –0.013725509 0.045730568 

Pt-4 β 5 mg m –2 s –1 mm –1 0.008392722 0.009444179 –0.010821644 0.027607088 

Pt-5 β 6 mg m –2 s –1 mm –1 0.006471389 0.003122043 0.000119533 0.012823245 

18 (c) D r Pt β 1 mg m –2 s –1 mm –1 0.565467014 0.170384074 0.227999229 0.902934798 

NLMR 


n=2 Pt-2 β 3 mg m –2 s –1 mm –1 1.110633143 0.418014423 0.282701296 1.938564991 

D=2H Pt-3 β 4 mg m –2 s –1 mm –1 0.334619992 0.145250564 0.046932389 0.622307595 

Pt-4 β 5 mg m –2 s –1 mm –1 –1.53E–01 0.240668154 –0.629188284 0.324160779 

Pt-5 β 6 mg m –2 s –1 mm –1 0.042533197 0.020393795 0.002140637 0.082925756 

18(d) D r Pt β 1 mg m –2 s –1 mm –1 0.305101494 0.068845599 0.168692499 0.44151049 

NLMR 


n=3 Pt-2 β 3 mg m –2 s –1 mm –1 0.691301164 0.161871969 0,370571984 1.012030344 

D=2H Pt-3 β 4 mg m –2 s –1 mm –1 0.08363529 0.049106013 –0.013662171 0.18093275 

Pt-4 β 5 mg m –2 s –1 mm –1 –0.3610 0.0819 –0.5233 –0.1988 

Pt-5 β 6 mg m –2 s –1 mm –1 0.01057652 0.003464163 0.003712711 0.017440329 

Table 8. Performance criteria of erosion model based on LMR and NLMR 

Eq. Order D p N R 2 (R 2 ) adj Cp SEE d MSSE AAE AARE AE ARE n AARE ARE 



17(a) 1 1 6 90 0.900 0.894 5 2.85 0.146 8.179 2.121 

17(b) 1 2 6 264 0.895 0.893 5 4.22 0.160 17.874 3.179 

18(a) 2 1 6 37 0.717 0.672 5 0.78 0.181 0.609 0.593 7.184 0.0008 –0.71 8 9.09 6.78 

18(b) 3 1 6 39 0.705 0.660 5 0.84 0.708 0.909 0.708 8.486 0.0024 –0.78 6 6.21 –2.31 

18(c) 2 2 6 122 0.725 0.713 5 1.50 0.945 2.252 1.183 10.060 0.0685 –0.45 10 6.83 4.06 

18(d) 3 2 6 118 0.538 0.518 5 1.90 1.680 3.626 1.430 11.859 0.2065 0.90 14 12.57 3.31 



94 

Erosion, Dr (kg / m 2 ) 

Irrigation Stage 

Fig. 12 Comparison of furrow simulated erosion by regression models with observed data. 

The cycle under irrigation has been divided into 

three stages; each stage represents 1/3 of the total 

duration for 12 weeks, obtaining an averaged data for 

each phase. In Fig. 12, solid lines represent the models 

by which a satisfactory approximation to the furrow 

erosion observed data is obtained, and the bars represent 

the rest of models, where the estimated values differ 

significantly from the observed data. In general, Eq. 

(17b) achieves the best approximation with respect to 

the rest of the linear and nonlinear models. 

Modeling of furrow erosion based on artificial 

intelligence techniques 

Table 9 shows the fitted statistics to sigmoidal function 

using the Artificial Neural Network (ANN) technique to 

the observations of rainfall-erosion process for the onehour 

and two-hour events, both with different amounts 

of data. Statistics in the stages of training, validation 

and testing vary for the networks mentioned before as 

follows: R²: For ANN1 in each stage is equal to 0.686; 

0.897 and 0.876; for ANN2 in each stage is equal to 

0.39; 0.615 and 0.49. 

For ANN3 in each stage is equal to 0.758; 0.831 and 

0.786. By comparing the errors during the calibration 

and validation stages, with emphasis on the average 

absolute error (AAE), the following results are found: 

AAE varies from 0.1 to 0.14. In general, (a) the error 

does not vary significantly between stages of 

calibration, validation and testing, (b) the error tends to 

be lower as the variability in the data is lower, which 

promotes a better network learning with respect to the 

observations made. 



95 

(a) 

D=1 h 

N=144 

(b) D=1 h 

N=144 

(c) D=1 h 

N=144 

(d) D=2h 

N=438 

(e) D= 2h 

N=438 

(f) D=2h 

N=438 

Simulated Erosion (Dr) mg /m 2 /s 

(g) D=2 h 

N=219 

(h) D=2h 

N=219 

(i)D= 2h 

N=219 

Observed Erosion (DR) mg/m 2 /s 

Observed Data 

Training Best Linear Fit 

Validation Best Linear Fit 

Test Best Linear Fit 

Fig. 13 Comparison of simulated erosion by Artificial Neural Network with observed 1 h and 2h rainfall events. 

Comparison of estimated values by the application of 

ANN’s, to the observed data of furrow erosion under 

rainfall and irrigation events of one and two hours, 

during the stages of training or calibration, validation 

and testing, is shown in Fig. 13. Vertical axis represents 

the simulated erosion (D R sim), and the horizontal axis 

represents the erosion observed (D R obs.). Each pair of 

coordinates is represented by a dot. 

Lines in Figs 13(a)–(i) represent the ratio of 1:1, 

where the approximation of points to the lines is a 

measure of how well the correlation between the set of 

erosion estimated values using the ANN technique, and 



96 

the obtained values by field tests in furrows in different 

slopes is. Figures 13a–13c and Figs 13g–13i show the 

approximate points randomly around the 1:1 line, 

indicating that the network has successfully learned the 

pattern shown by observations during rainfall events. 

Figures 13d–13f show that there is a fairly satisfactory 

approximation of the points to the 1:1 line, which 

explains that there is some variability in the 

observations that the network was not able to simulate. 

Fuzzy Logic 

Modeling furrow erosion using a Fuzzy Inference 

System 

Table 10 shows the two membership functions 

coefficients for the two input variables to each Fuzzy 

Inference System (FIS). Characterized FIS’s are three: 

FIS 1 corresponds to natural rainfall events that last an 

hour, FIS 2 corresponds to the natural rainfall events 

that last two hours, and FIS 3 corresponds to irrigation 

events that last two hours. Input 1 is represented by 

measurements of natural and simulated rainfall 

(irrigation) in those events that last an hour every 10 

minutes, and two hours every 20 minutes. Input 2 is 

represented by the infiltration values range as shown in 

Fig. 7. 

As an example, FIS 1 is described, represented by 

input membership functions of trapezoidal type. The 

trapezoidal functions for the rainfall representation have 

the following parameters: trapmf1: 1.827; 2.974; 4.557 

and 5.928. Trapmf2: 4.677; 5.705; 7.562 and 8.709. The 

trapezoidal functions for the infiltration representation 

have the following parameters: trapmf1: –0.508; 0.0293, 

0.859; 1.381. Trapmf2: 3.637; 4.676; 7.46 and 9.147. 

The linear output surface has the following parameters: 

2.081 for the rainfall; –2.675 for infiltration and 0.0487 

for the constant term. The rest of FIS’s can be seen in 

Table 10. Table 11 shows the fitted statistics of erosion 

modeling under rainfall and irrigation events through 

the various FIS’s from all the observations obtained in 

field testing in furrows of different slopes. By comparing 

the Average Absolute Errors during the calibration, 

validation and testing stages, the following results are found: 

FIS 1: 1.108; 1.5147 and 1.489; FIS 2: 2.403; 2.5977 and 

2.676; FIS 3: 1.4; 2.396 and 1.565, respectively. In general, 

errors do not vary significantly between the stages of the 

calibration, validation and testing. 

Table 9. Performance criteria of erosion model based on ANN 

Model N D AAE R 2 

(a) During calibration/training 

ANN 1: 6-20-6 144 1 0.13962 0.686 

ANN 2: 6-100-6 438 2 0.1750 0.390 

ANN 3: 6-100-6 219 2 0.1283 0.758 

(b) During validation 

ANN 1: 6-20-6 144 1 0.146 0.897 

ANN 2: 6-100-6 438 2 0.166 0.615 

ANN 3: 6-100-6 219 2 0.103 0.831 

(c) During testing 

ANN 1: 6-20-6 144 1 0.146 0.876 

ANN 2: 6-100-6 438 2 0.166 0.490 

ANN 3: 6-100-6 219 2 0.1039 0.786 

N: number of observed data; D: irrigation and rainfall events duration; AAE: average absolute error; R 2 : determination coefficient; ANN: 

artificial neural network 

Table 10. Coefficients of membership function from various FIS 

Model Inputs Membership Function Parameter- 1 Parameter- 2 Parameter- 3 Parameter- 4 

Input 1 Trapmf-1 1.827 2.974 4.557 5.928 

FIS 1 

Input 1 Trapmf-2 4.677 5.705 7.562 8.709 

Input 2 Trapmf-1 –0.508 0.0293 0.859 1.381 

Input 2 Trapmf-2 3.637 4.676 7.460 9.147 

Output linear 2.801 –2.675 0.048 

Input 1 Trapmf-1 –0.9727 0.7139 3.089 5.373 

FIS 2 

Input 1 Trapmf-2 3.637 4.676 7.46 9.147 

Input 2 Trapmf-1 –0.5081 0.0293 0.859 1.381 

Input 2 Trapmf-2 3.637 4.676 7.46 9.147 

Output Linear 5.13 0.1209 –0.179 

Input 1 Trapmf-1 –5.071 –1.601 4.226 7.563 

FIS 3 

Input 1 Trapmf-2 5.466 7.789 12.28 15.75 

Input 2 Trapmf-1 –0.501 0.021 0.859 1.381 

Input 2 Trapmf-2 0.979 1.398 2.179 2.717 

Output Linear 2.298 2.073 4.518 



97 

Table 11. Performance criteria of erosion model based on FIS 

Model N D AAE 

(a) During calibration/training 

FIS 1 (trapmf) 88 1 1.108 

FIS 2 (trapmf) 262 2 2.007 

FIS 3 (trapmf) 180 2 1.400 

(b) During validation 

FIS 1 (trapmf) 28 1 1.5147 

FIS 2 (trapmf) 88 2 2.1407 

FIS 3 (trapmf) 60 2 2.396 

(c) During testing 

FIS 1 (trapmf) 28 1 1.489 

FIS 2 (trapmf) 88 2 2.367 

FIS 3 (trapmf) 60 2 1.565 

N: number of observed data; D: irrigation and rainfall events duration; AAE: average absolute error; R 2 : determination coefficient; ANN: 

artificial neural network 

(a) 

Erosion DR (mg m ‐2 s ‐1 ) 

(b) 

(c) 

FIS Data Points * Training Data Points o Validation Data Points + Test Data Points 

Fig. 14 Comparison of simulated erosion by Fuzzy Inference System with observed 2h rainfall events. 



98 

A comparative example of the estimated values by 

applying the FIS to the observed data of furrow erosion 

under rainfall events that last two hours, during the 

stages of training, validation and testing; using a sample 

of 35 data, is shown in Fig. 14. Total number of data in 

each stage can be seen in Table 11. Vertical axis 

represents the simulated erosion (D R sim) and the 

observed (D R obs). The horizontal axis represents a 

number assigned to each value; each pair of coordinates 

is represented by a dot. Figure 14(a) shows the results 

in the training stage of the FIS, which shows that the 

data points (circles) were satisfactorily approximate to 

the FIS (asterisks). Figures 14(b)–(c) show a good 

approximation of the points found in the FIS. 

DISCUSSION OF RESULTS 

Comparison of furrow erosion processes simulated 

with experimental data 

Process of particles detachment As shown in the 

results section, the range of values of concentrated flow 

erodibility K c and critical shear stress τ c , obtained from 

the fit of the Eq. (5) varied between 1.31E–06 and 

1.525E–06 s/m, 0.704 and 1.489 Pa, respectively. 

Knapen et al. (2007) conducted a literature review to 

compile the values of K c and τ c reported for different 

soils and tillage conditions. Studies included: (a) 

concentrated flow experiments on field plots, and (b) 

flume experiments in the laboratory. Characteristics of 

the experimental methods used are summarized in 

Tables 12 and 13. Values ranges have been determined 

empirically for K c and τ c parameters from the excess 

shear stress linear equation (Eq. 5), including the size of 

the sample by using field and laboratory experiments, 

they varied as follows: 0.000 001 and 0.1 m/s (n = 151), 

0.01 and 20 Pa (n = 161), 0.00 001 and 0.7 s/m 

(n = 185), 0.03 and 60 Pa (n = 220), respectively. By 

comparing these value ranges with those reported on 

Table 3 for Eq. (5), it was found: (a) K c values fell 

within the range for the field experiments and were 

lower than those found from laboratory experiments, (b) 

τ c values were totally included within the reported 

ranges for the field and laboratory experiments. 

From the above, comparisons show that the K c value 

range obtained in this study was significantly different 

from the one reported by Knapen et al. (2007), while a 

lower variation for τ c value range was found. There are 

several reasons for these differences, two of the main 

are (1) differences in experimental conditions in which 

the data were collected, and (2) the variation of soil 

types and environmental conditions. By comparing the 

experimental and environmental conditions of the 

research described in the methods section with outlined 

in Tables 12 and 13, some similarities were found alike 

as those in field studies conducted by Bjorneberg et al. 

(1999). With regard to laboratory experiments, cases 

with which to make a comparison were not found. 

The value ranges found for K c and τ c by Bjorneberg 

et al. (1999) varied as follows: from 0.0003 to 0.006 

m/s and from 1.2 to 1.8 Pa, respectively. Despite the 

similarities in the experimental conditions shown in 

Table 12, the value range of K c was significantly 

different. Causes of the differences appear to be due to 

environmental conditions, mainly in both climate and 

agricultural practices, because the experiments of 

Bjorneberg et al. (1999) were made in beans and corn 

fields, where the wetting and drying sequences, 

consolidation and residues could be contributing factors 

to variability of K c and τ c values. 

In general, the Eq. (5) statistics show a satisfactory 

fit between most of the equation and the observed data, 

for the following reasons: (1) coefficient of 

determination R 2 was estimated at 0.7 (Ramírez et al., 

2004), (2) Durbin-Watson coefficient presented 

evidence that there is a slight randomness within 

consecutive residues, (3) Average Absolute Relative 

Error and Average Relative Error were moderately low. 

Process of sediment transport The ranges of the 

parameters K t and b reported in the results section for 

models of sediment transport capacity represented by 

Eqs 7, 13 and 15 varied as follows: 9.929E–08 and 

1.52 669E–07; 7.634 and 34.583; 0.4212 and 0.892, 

respectively. The value ranges reported for field tests by 

Finkner et al. (1989) and Trout (1999) for Eq. (7) 

(Table 14), differ significantly from those indicated in 

Table 5. As for Eq. (13), the ω c critical unit current 

power value was negative and slightly different from 

zero, which differs from that proposed by Yang, (1973), 

which is equal to 0.002 m/s. 

Despite the differences, it can be observed that in 

both cases ω c tended to zero. According to theory, the 

ω c value should be positive. The results of negative ω c 

and its low standard error, along with the high 

correlation make it difficult to explain the physical 

meaning of the estimated parameter. However, the 

estimated negative value range of ω c was not 

significantly different from zero. 



99 

Table 12. Characteristics of field plot in which concentrated flow erosion was measured 

Country 

Soil Slope Q inflow I rainfall τ Rill plot dim. 

(No.) (%) (l min –1 ) (mm/h) (Pa) (m × m) 

Surface Condition 

Source 

Iran 1 n.a. 132–1693 n.a. 

2.2– 

I, varying vegetation 

15 × 0.3 

13.2 

cover 

Adelpour et al. (2004) 

USA 1 

0.5– 

Length.:110– I, residue removed 

30–40 n.a. n.a. 

1.33 

256 

and tilled 

Bjorneberg et al. (1999) 

Brazil 1 n.a. n.a. n.a. n.a. 9 × 0.5 

I, residue removed 


Braida & Cassol (1996) 

Brazil 1 6.7 0–50 74 n.a. n.a. n.a. Cantalice et al. (2005) 

USA 2 3–15 96–768 n.d. 

4.0– 


10 × 0.75 

37.3 


Franti et al. (1985; 1999) 

Brazil 1 n.a. 0 60 n.a. 9 × 0.5 



Giasson & Cassol, (1996) 

USA 30 4–13 7–35 62 n.a. 9 × 0.46 



Gilley et al. (1993) 

USA 4 0.5–3 n.a. n.a. 1–36 30.5×0.91 I, no cover Hanson (1989;1990a; 1990b) 

USA 1 1–3 4–17.10 n.a. 

12– 

55 

29×1.8 I, no cover Hanson & Cook (1999) 

USA 2 5–11 n.d. n.a. n.a. 10.7×3 

I, varing canopy 

cover 

Hussein & Laflen (1982) 

USA 2 4 11–189 64 

0.7– 

I, varying tillage 

4×0.2 

14 

practices 

King et al. (1995) 

USA 2 3–15 n.a. n.a. n.a. n.a. n.a. Laflen (1987) 

USA 56 2–13 0.1–0.6 63 0–22 9–11×0.5–3 

I, residue removed Laflen et al. (1991), Elliot et 


al. (1989) 

USA 1 3–6 8–38 n.a. 2–10 5.5×2 



Mamo & Bubenzer (2001b) 

Canada 3 12–14 n.a. 25–30 n.a. 10×0.8 

I, vegetation free, 

seedbed conditions 

Merz & Bryan (1993) 

USA 1 4–6 8–38 51 

1.3– 

I, tilled and all 

68.6×6.1 

6.1 

residue buried 

Morrison et al. (1994) 

USA 2 3–5 8–53 64 0–6 6.1×0.76 



Norton & Brown, (1992) 

USA 1 27 16–23 n.a. 

24– 

I, vegetation clipped 

6×0.3 

192 

to various heights 

Prosser et al. (1995) 

Australia 1 1–12.7 1.7–8 n.a. n.a. 20×1 

I, vegetation clipped 

to various heights 

Prosser (1996) 

Brazil 1 10 12–120 65 

2.5– 

19 

6×0.2 n.a. Reichert et al. (2001) 

USA 1 2–31 8–60 60 2–7 4.6×0.3 


and plants clipped 

West et al. (1992) 

USA 1 

0.52– 


15–46 n.a. n.a. 204–256 

1.33 


Trout et al. (1999) 

n.a.: not available. Soils: number of soil types tested. Slope: slope of soil surface. Q inflow : simulated concentrated flow discharge. I rainfall : 

simulated rainfall intensities. τ: Range of applied flow shear stresses. Rill plot dimensions: length×width. Surface condition: (S: smoothened, 

I:irregular) 

With respect to Eq. (15), comparisons can not be 

made with the values proposed by Simons et al. (1981) 

because the values reported in literature correspond to 

flow in rivers. 

In general, fitted statistics in Eqs (7), (13) and (15) 

indicated a satisfactory approximation between observed 

and estimated values, for the following reasons: (a) 

coefficient of determination varied between 0.72 and 

0.88, according to Ramirez et al. (2004), (b) adjusted R 2 

coefficient decreased slightly in relation to R 2 , (c) 

Durbin-Watson coefficient presented evidence that there 

is a slight randomness between consecutive residues, (3) 

Average Absolute Relative Error and Average Relative 

Error were moderately low. 

Combination of furrow erosion processes It can be 

observed from Figure 11 that the sediment transport 

capacity values obtained by Eqs (7) and (13) were lower 

than the observed sediment load values in the stream for 

different slopes of furrows in the three stages within the 

cycle under irrigation, which indicates the occurrence of 

concentrated flow and deposition process. As for Eq. 

(15), estimated values of the transport capacity were 

very close to the sediment load current, therefore, 

deposition processes occur at low rates. According to 

theory, the highest proportion of detached sediments is 

transported out of the furrow (e.g., Nearing et al., 1989, 

Morgan, 1998; Bulygin et al., 2002). 



100 

Table 13. Characteristics of laboratory flume experiments in which concentrated flow erosion was measured 

Country 

Soil Slope Q inflow I rainfall 

Flume dim. 

(N°) (%) (l min –1 τ (Pa) 

) (mm/h) 

(m × m × m) 

Source 

Remark 

Italy 6 10–40 2.4–18 / 1–12 1.5 × 0.2 Ciampallini & Torri (1998) a 

Australia 1 2–7 2.5–40 / n.a. 1.8 × 0.6 × 0.2 Crouch & Novruzi (1989) b, e 

Mexico 1 0 n.a. / 1.1–38.2 4.9 × 0.3 × 2.5 Ghebreiyessus et al. (1994) 

Belgium 2 5–21 3.3–60 / 1–24 4 × 0.4 × 0.45 Giménez & Govers (2002) 

Belgium 2 9–21 6.7–60 / 1–16 2 × 0.10 × 0.09 Giménez & Govers (2002) 

Belgium 2 1.5–7 0.7–10.4 / 0,1–1,3 6 × 0.12 Govers (1985) b, e 

Belgium 1 20–30 18–114 / 4.4–22.4 2 × 0.1 × 0.09 Gyssels et al. (2006) 

Canada 5 n.a. n.a. / n.a. 9.1 × 0.15 Kamphuis & Hall (1983) b, e 

USA 5 0.2 n.a. / n.a. 18.3 × 0,77 × 0,46 Laflen & Beasley (1960) a, e 

USA 7 0.2 n.a. / n.a. 22 × 0.4 × 0.76 Lyle & Smerdon (1965) a, e 

USA 1 3–5 7.6–37.9 / n.a. 4 × 0.2 × 0.05 Mamo & Bubenzer (2001a) 

Canada 1 14 / 34 n.a. 10 × 0.8 × 0,2 Merz & Bryan (1993) b, c, e 

USA 8 n.a. 6.1–30.3 / 0–3.2 1.84 × 0.1 × 0,19 Moody et al. (2005) e 

Belgium 1 10–35 5.6–11.5 / 1.6–5.7 2 × 0.1 × 0.09 Nachtergaele & Poesen (2002) 

USA 1 n.a. n.a. / 0.1–1.3 18.3 × 0.3 × 0.38 Partheniades (1965) e 

S-Africa 12 2–20 0.08–0.32 / 0,2–4.9 0.5 × 0.05 × 0.13 Rapp (1998) 

Israel 3 2–20 0.08–0.32 / 0.2–4.9 0.5 × 0.05 × 0.13 Rapp (1998) 

Belgium 1 2.6–14 n.a. / n.a. 2 × 0.4 Rauws (1987) 

Belgium 1 3–33 n.a. 20 n.a. 2.5 × 0.6 Rauws & Govers (1988) b, c, e 

USA 1 2–20 0.04–0.2 / 0.4–4.8 0.5 × 0.05 × 0.12 Shainberg et al. (1994) b, e 

Israel 3 5 0.04–0.32 / 0.8–1.8 0.5 × 0.05 × 0.12 Shainberg et al. (1996) 

Australia 16 5–30 0.1–1.8 100 n.a. 3 × 0.8 × 0.15 Sheridan et al. (2000a,b) d 

Canada 1 9 n.a. 35 n.a. 15 m long. Slattery & Bryan (1992) 

USA 11 0–1 n.a. / n.a. 18.3 × 0.77 × 0.46 Smerdon & Beasley (1959) 

USA 5 0.1–0.2 n.a. 0–127 n.a. 22 × 0.8 Smerdon (1964) 

Italy 4 1–31 n.a. 15–110 0–3.3 2 × 0.5 × 0.1 Torri et al. (1987) 

USA 1 7 2–4 / 1.9–3.9 2.73 × 0.46 × 0.88 Van Klavern & McCool (1998 

China 1 10,5–21.2 2.5–6.5 / 1–10 5 × 0.3 Zhang et al. (2005) a 

USA 5 1.5–5 3.8–15.1 / 0.5–2.5 6.4 × 0.15 × 0.05 Zhu et al. (1995) 

USA 5 0–5.5 n.a. / 0.5–2.5 6.4 × 0.15 × 0.05 Zhu et al. (2001) 

n.a.: not available. /:not applicable. Soils: number of soil types tested. Slope: slope of soil surface. Q inflow : simulated concentrated flow 

discharge. I rainfall : simulated rainfall intensities. τ: Range of applied flow shear stresses. Flume dimensions: length×width×depth. Remarks: 

(a) values for K c and/or τ c deduced from graphs, (b) values calculated from critical shear velocity, (c) values for poorly cohesive or noncohesive 

soils, (d) tests on coal mine soils and overburdens, (e) excess shear stress equation was not used to deduce K c and/or τ c . 

Table 14. Models parameters to estimate sediment transport capacity of flow in furrow and characteristics of field tests 

Tc= Kt(τ) b 

Source 

T c 

Slope Q I 

K 

(kg/m/s) 

t τ(Pa) b 

rainfall Rill plot Surface 

(%) (L / min) (mm/ h) dim. (m×m) condition 

Finkner 0.0006–0.2366 0,015–0,05 1–6 1.5 2–20 n.a. 80–83 10, 50, 100 n.a. 

et al. 

m long 

(1989) 

Trout 

(1999) 

0.002–0.035 0,017– 

0,065 

n.a. 2 1.3 6–18 n.a. 204 m long residue 

removed 


0.001–0.02 0,015–0,15 n.a. 4 0.5 6–45.6 n.a. 256 m long residue 

removed 


Source T c (ppm) C 

ω c 

(m/s) 

b 

Tc = c[(ω–ωc)/Vs] b 

Slope 

(%) 

Q inflow 

(L/min) 

Irainfall 

(mm/h) 

Rill plot 

dim. 

(m×m) 

Surface 

condition 

Loch 41.000–87.000 n.a. 0.002 n.a. 4 15–107 95 0.4×1.5–22.5 removed 

(1984) 


n.a.: not available. Remark: K t : (a) obtained by empirical fit, (b) obtained by equation of Yalin (1963). 

Remark 

b 

a 

a 

Remark 

b 

Trout (1996) found that at the end of the furrows of 

bean crops, erosion increases from the beginning to the 

end of the cycle from 1 to 5 Mg/ha, while the process 

was reversed in the corn crop due to the reduction of 

erosion from 2 to 0.1 Mg/ha at the end of the cycle. In 

this study, the variability in the rate of erosion along the 

furrows was not significant, based on the uniformity of 

both water supplies through the sprinklers as well as 

frequency of wetting in each furrow. The erosion rate, 

from the beginning to the end of the cycle, varied from 

0.6 to 0.1 Mg/ha (Fig. 12). This range is lower than that 

found by Trout (1996) for bean and corn fields. 



101 

Trout (1999) reported sediment transport rates in 

furrows of slope at 1.3% and 5.2% for corn and bean, 

that varied from the beginning to the end of the cycle as 

follows: from 20 000 to 4000 mg·s –1 m –1 , from 25 000 to 

5000 mg·s –1 m –1 , respectively. These ranges were higher 

than those found in this study, which varied from 1428 

to 10 000 mg·s –1 m –1 (mg of sediment per meter of 

wetted perimeter per second). 

Furrow erosion models based on regressions 

Estimated erosion values by the multiple linear models 

were closer to the field observations than those which 

were obtained by models based on nonlinear 

polynomials. Linear model showed that the maximum 

loads of sediments during events of natural and artificial 

rain were successfully estimated. However, there was a 

lack of adjustment for values at the beginning and at the 

end of such events, which coincides with the results of 

Jain et al. (2004), Kumar (2001) and Srinivasulu et al. 

(2006) (Fig. 12). Findings in polynomial-based models 

indicate that the estimation process of the observed 

variable tends to decrease as the power increases. 

Modeling of furrow erosion using artificial 

intelligence techniques 

Artificial Neural Networks Estimated erosion values 

by artificial neural networks became successfully closer 

to the observations during the stages of calibration, 

validation, and testing, in terms of the coefficient of 

determination and the Average Absolute Error. The 

ANN estimate improved as the variability of data for 

training and testing of the network was lower, which 

helped the network to learn the pattern of data. 

Fuzzy Inference System Estimated erosion values by 

FIS’s became successfully closer to the observations 

during the stages of calibration, validation and testing, 

in terms of the Average Absolute Error. The pattern in 

the input data was better represented by using the 

trapezoidal function both for rainfall, as well as for 

infiltration. The system output was represented by a 

multiple linear equation. Estimated erosion resulted 

closer to observations in the stages of training, 

validation and testing (Fig. 14). Estimation by ANN 

improved as the variability of data for training and 

testing of the network was lower, which helped the 

network to learn the pattern of data. The application of 

FIS’s has been used successfully for modeling of 

reservoir operation by Panigrahi (2000), which validates 

its use for simulating hydrological processes. 

Comparison of models to estimate rill erosion The 

differences between the dynamic versions of the WEPP 

and EUROSEM models were significant in the 

magnitudes of the events; however, both of them 

estimated the erosion processes, namely, sediment 

transport and deposition (Fig. 11). As for CIHAM-UC 

model was different from WEPP and EUROSEM 

models because they produced smaller estimates of 

deposition process. 

The comparison between the different models 

indicates the following order on the quality of the 

estimation in furrow erosion: Artificial intelligence 

techniques, models based on physical processes, linear 

regression, and finally, the non-linear regression (Fig. 

12). 

CONCLUSIONS 

It is possible to make estimates of the physical 

processes of erosion in furrows with a moderate to high 

accuracy, although it is believed that the estimation can 

be increased and improved by making adjustments of 

the models for each slope of furrow. 

Models based on linear regressions achieved better 

estimates at the observed values than the nonlinear 

models based on polynomials. This should be 

emphasized in training with new data to better estimate 

the values at the beginning and end of natural and 

artificial rain events. 

Modeling of furrow erosion using artificial 

intelligence techniques resulted in a satisfactory 

approximation of the estimated values to those 

observed, improving accuracy through the application 

of ANN’s with respect to the FIS’s. The estimation 

quality improves as variability in the model data is low. 

The comparison between the different models 

indicate the following order on the quality of the 

estimation of furrow erosion: artificial intelligence 

techniques, models based on physical processes, linear 

regression and finally, the non-linear regression 

Acknowledgment The research was developed in 

Centro de Investigaciones Hidrológicas y Ambientales 

de la Universidad de Carabobo (CIHAM-UC) with the 

financial support of CDCH-UC and FONACIT. 

REFERENCES 

AMERICAN PUBLIC HEALTH ASSOCIATION (1995) Standard 

Methods for The Examination of Water and Wastewater. 

American Public Health Association, United States of America, 

Washington, DC, 19th Edition. 2–53. 

AMERICAN SOCIETY FOR TESTING AND MATERIALS 

(2007) Standard Test Methods for particle-size analysis of soils. 

American Society for Testing and Materials International, United 

States of America, West Conshohocken. (Reaproved 2007), 1–8. 

Adelpour, A.A., Soufi, M. & Behnia, A.K. (2004) Channel erosion 

thresholds for different land uses assessed by concentrated 

overland flow on a silty loam. Proc. International Soil 

Conservation Organisation Conference. Brisbane, Australia. 

Alonso C.V., Neibling, W.H. & Foster, G.R. (1981) Estimating 

sediment transport capacity in watershed modeling. Trans. ASAE 

24(5), 1211–1220. 

Bjorneberg, D.L., Trout, R.E., Sojka, R.E. & Aase, J.K. (1999) 

Evaluating WEPP predicted infiltration, runoff and soil erosion 

for furrow irrigation. Trans. ASAE 42(6), 1733–1741. 



102 

Braida, J.A. & Cassol, E.A. (1996) Rill and interrill erodibility of a 

paleudult soil. Rev. Bras. Ciênc. Solo 20(1), 127–134. 

Bulygina, N.S., Nearing, M.A., Stone, J.J. & Nichols, M.H. (2007) 

DWEPP: a dynamic soil erosion model based on WEPP source 

terms. Earth Surf. Processes Landf. 32(7), 998–1012. doi: 

10.1002/esp.1467. 

Bulygin, S.Y., Nearing, M.A. & Achasov, A.B. (2002) Parameters of 

interrill erodibility in the WEPP model. Eurasian Soil Sci. 35(11), 

1237–1242. 

Bishop, C.M. (1994) Neural networks and their applications. Rev. 

Sci.Instrum. 65(6), 1803–1832. 

Cantalice, J.R.B., Cassol, E.A., Reichert, J.M. & Borges, A.L.D. 

(2005) Flow hydraulics and sediment transport in rills of a sandy 

clay loam soil. Rev. Bras. Ciênc. Solo 29(4), 597–607. (in 

portuguese). doi: 10.1590/S0100-06832005000400012. 

Ciampallini, R. & Torri, D. (1998) Detachment of soil particles by 

shallow flow: sampling methodology and observations. Catena 

32(1), 37–53. doi:10.1016/S0341-8162(97)00050-7. 

Crouch, R.J. & Novruzi, T. (1989) Threshold conditions for rill 

initiation on a vertisol, Gunnedah, N.S.W., Australia. Catena 

16(1), 101–110. doi:10.1016/0341-8162(89)90007-6. 

Demuth, H., Beale, M., Hagan, M. (2009) Neural Network Toolbox. 

User Guide. MATLAB. 

Elliot, W.J., Liebenow, A.M., Laflen, J.M. & Kohl, K.D. (1989) A 

compendium of soil erodibility data from WEPP cropland soil 

field erodibility experiments 1987–1988. NSERL Rpt. No. 3. 

Ohio State University and Natural Soil Erosion Research 

Laboratory, Agricultural Research Service, U.S. Department of 

Agriculture, W. Lafayette, Indiana. 

Flanagan, D.C., Ascough, J.C., Nearing, M.A. & Laflen, J.M. (2001) 

The Water Erosion Prediction Project model. In: Harmon, R.S. & 

Doe, W.W. (eds) Landscape Erosion and Evolution Modelling, 

Kluwer: New York, 145–199. 

Flanagan, D.C. & Nearing, M.A. eds. (1995) USDA–Water Erosion 

Prediction Project (WEPP), Hillslope Profile and Watershed 

Model Documentation Technical Documentation, NSERL Report 

10. USDA-ARS National Soil Erosion Research Laboratory: 

West Lafayette, IN. 

Finkner, S.C., Nearing M.A., Foster G.R. & Gilley J.E. (1989) A 

simplified equation for modeling sediment transport capacity. 

Trans. ASAE 32(5), 1545–1550. 

Foster, G.R. & Meyer, L.D. (1972) Transport of particles by shallow 

flow. Trans. ASAE 15(1), 99–102. 

Foster, G.R. & Meyer, L.D. (1975) Mathematical simulation of 

upland erosion using fundamental erosion mechanics. Proc. 

Sediment Yield Workshop. U.S. Sedimentation Laboratory, 

Oxford, MI, 190–201. 

Foster, G.R. (1982) Modeling the erosion process. In: Hahn, C.T. 

(Ed.). Hydrologic Modeling of Small Watersheds, 295–380. 

Foster, G.R., Huggins, L.F. & Meyer, L.D. (1984) A laboratory 

study of rill hydraulics: shear stress relationships. Trans. ASAE 

27(3), 797–804. 

Foster, G.R. & Lane, L.J. (1987) User Requirements: USDA-WEPP 

(Draft 6.3) National Soil Erosion Research Laboratory, West 

Lafayette, Indiana. 

Foster, G.R. (1990) Major developments in prediction of soil erosion 

by water. In: Lal, R. & Pierce, F.J. (eds.) Soil Management for 

Sustainability Soil and Water Conservation Society, Ames, Iowa. 

Franti, F.G., Laflen, J.M. & Watson, D.A. (1985) Soil erodibility 

and critical shear under concentrated flow. ASAE Paper No. 85- 

2033 

Franti, T.G., Laflen, J.M. & Watson, D.A. (1999) Predicting soil 

detachment from high discharge concentrated flow. Trans. ASAE 

42(2), 329–335. 

Ghebreiyessus, Y.T., Gantzer, C.J., Alberts, E.E. & Lentz, R. (1994) 

Soil erosion by concentrated flow: shear stress and bulk density. 

Trans. ASAE 37(6), 1791–1797. 

Giménez, R. & Govers, G. (2002) Flow detachment by concentrated 

flow on smooth and irregular beds. Soil Sci. Soc. Am. J. 66(5), 

1475–1483. doi:10.2136/sssaj2002.1475. 

Gyssels, G., Poesen, J., Van Dessel, W., Knapen, A. & Debaets, S. 

(2006) Effects of cereal roots on detachment rates of single and 

doubledrilled topsoils during concentrated flow. Eur. J. Soil Sci. 

57(3), 381–391. doi: 10.1111/j.1365-2389.2005.00749.x. 

Giasson, E. & Cassol, E.A. (1996) Rill erosion related to inflow 

rates and amounts of incorporated wheat straw in a sandy clay 

loam Paleudult soil. Rev. Bras. Ciênc. Solo 20(1), 117–125. 

Gilley, J.E., Elliot, W.J., Laflen, J.M. & Simanton, S.R. (1993) 

Critical shear stress and critical flow rates for initiation of rilling. 

J. Hydrol. 142(2), 251–271. doi:10.1016/0022-1694(93)90013-Y. 

Govers, G. (1985) Selectivity and transport capacity of thin flow in 

relation to rill erosion. Catena 12(1), 35–49. doi:10.1016/S0341- 

8162(85)80003-5. 

Guevara, E. & Márquez, A. (2009) Modelación de la erosión y 

transporte de sedimentos en un campo agrícola bajo riego. 

Tecnología y Ciencias del Agua (Submitted). 

Hanson, G.J. (1989) An in-situ erodibility testing device. American 

Society of Agricultural Engineers - Canadian Society of 

Agricultural Engineers. Paper No. 89–2151. 

Hanson, G.J. (1990a) Surface erodibility of earthen channels at high 

stresses. Part II-Developing an in situ testing device. Trans. ASAE 

33(1), 132–137. 

Hanson, G.J. (1990b) Surface erodibility of earthen channels at high 

stresses part I. Open channel testing. Trans. ASAE 33(1), 127– 

131. 

Hanson, G.J., Cook, K.R. & Simon, A. (1999) Determinig erosion 

resistance of cohesive materials. American Society of Civil 

Engineers, Proc. International Water Resources Engineering 

Conference. Seattle, Washington, USA. 

Hussein, M.H. & Laflen, J.M. (1982) Effects of crop canopy and 

residue on rill and interrill soil erosion. Trans. ASAE 25(5), 1310– 

1315. 

Jain, A. & Ormsbee, L.E. (2002) Evaluation of short-term water 

demand forecast modeling techniques: conventional methods 

versus AI. J. American Water Works Assoc. 94(7), 64–72. 

Jain, A., Sudheer, K.P. & Srinivasulu, S. (2004) Identification of 

physical processes inherent in artificial neural network rainfall 

runoff models. Hydrol. Processes 118(3), 571–581. doi: 

10.1002/hyp.5502. 

Juarez, E. & Rico, A. (1991) Mecánica de Suelos. Ed. Limusa. 

Mexico. 

Kamphuis, J.W. & Hall, K.R. (1983) Cohesive material erosion by 

unidirectional current. J. Hydr. Engrg. 110(3), 370–370. doi: 

10.1061/(ASCE)0733-9429(1984)110:3(370). 

Knapen, A., Poesen, J., Govers, G., Gyssels, G. & Nachtergaele, J. 

(2007) Resistance of soils to concentrated 

flow erosion: A review. Earth-Science Rev. 80(2), 75–109. 

doi:10.1016/j.earscirev.2006.08.001. 

King, K.W., Flanagan, D.C., Norton, L.D. & Laflen, J.M. (1995) 

Rill erodibility parameters influenced by long-term management 

practices. Trans. ASAE 38(1), 159–164. 

Koluvek, P.K., Tanji, K.K. & Trout, T.J. (1993) Overview of soil 

erosion from irrigation. J. Irrig. Drainage Engr. ASCE 119(6), 

929–946. doi: 10.1061/(ASCE)0733-9437(1993)119:6(929). 

Klik, A., Savabi, M.R., Norton, L.D. & Baumer, O. (1995) 

Application of WEPP hillslope model on Austria. Proc. Annual 

Conference of the American Water Resources Association 

(AWRA). Houston, Texas, 313–322. 

Kumar, A. & Minocha, K. (2001) Rainfall runoff modeling using 

artificial neural networks. J. Hydrol. Eng. 6(2), 176–177. doi: 

10.1061/(ASCE)1084-0699(2001)6:2(176). 

Laflen, J.M. & Beasley, R.P. (1960) Effects of compaction on 

critical tractive forces in cohesive soils. University of Missouri. 

Experiment Station Research Bulletin, 749. 

Laflen, J.M. (1987) Effect of tillage systems on concentrated flow 

erosion. In: Sentis, I.P. (Ed.), Soil Conservation and Productivity. 



103 

Proc. 40 International Conference on Soil Conservation, Maracay, 

República Bolivariana de Venezuela. 

Laflen, J.M., Elliot, W.J., Simanton, J.R., Holzhey, C.S. & Kohl, 

K.D. (1991) WEPP: Soil erodibility experiments for rangeland 

and cropland soils. J. Soil Water Conserv. 46(1), 39–44. doi: 

10.2489/jswc.46.1.39. 

Laflen, J.M., Flanagan, D.C. & Engel, B.A. (2004) Soil erosion and 

sediment yield prediction accuracy using WEPP. 

J. Am. Water Res. Assoc. 40(2), 289–297. doi: 10.1111/j.1752- 

1688.2004.tb01029.x. 

Lyle, W.M. & Smerdon, E.T. (1965) Relation of compaction and 

other soil properties to erosion resistance of soils. Trans. ASAE 

8(4), 419–422. 

Loch, R.J. (1984) Field rainfall simulator studies on two caly spils of 

the Darling Downs, Queensland, III, An evaluation of currents 

methods of deriving soil erodibilities (K factors), Aust. J. Soil 

Res. 22(4), 401–412. 

Mamo, M. & Bubenzer, G.D. (2001a) Detachment rate, soil 

erodibility and soil strength as influenced by living plant roots 

part I: laboratory study. Trans. ASAE 44(5), 1167–1174. 

Mamo, M. & Bubenzer, G.D. (2001b) Detachment rate, soil 

erodibility and soil strength as influenced by living plant roots 

part II: field study. Trans. ASAE 44(5), 1175–1181. 

MATLAB. (2010) Fuzzy Logic Toolbox. User’s Guide. The 

Mathworks Inc. 

Merz, W. & Bryan, R.B. (1993) Critical threshold conditions for rill 

initiation on sandy loam Brunisols: laboratory and field 

experiments in southern Ontario, Canada. Geoderma 57(4), 357– 

385. doi: 10.1016/0016-7061(93)90050-U. 

Meyer-Peter, E. & Müller R. (1948) Formulas for bed-load transport. 

Proc. II Meeting IARH, Stockholm, 39–64. 

Moody, J.A., Smith, J.D. & Ragan, B.W. (2005) Critical shear stress 

for erosion of cohesive soils subjected to temperatures 

typical of wildfire. J. Geophys. Res. 110, F01004. doi: 

10.1029/2004JF000141. 

Morrison, J.E., Richardson, C.W., Laflen, J.M. & Elliott, W.J. 

(1994) Rill erosion of a Vertisol with extended time since tillage. 

Trans. ASAE 37(4), 1187–1196. 

Morgan, R.P.C., Quinton, J.N., Smith, R.E., Govers G., Poesen, J.W. 

A., Auerswald, K., Chisci, G., Torri, D. & Styczen, M.E. (1998) 

The European Soil Erosion Model (EUROSEM): A dynamic 

approach for predicting sediment transport from fields and small 

catchments. Earth Surf. Processes Landf. 23(6), 527–544. doi: 

10.1002/(SICI)1096-9837(199806)23:6. 

Nachtergaele, J. & Poesen, J. (2002) Spatial and temporal variations 

in resistance of loess-derived soils to ephemeral gully erosion. 

European J. Soil Sci. 53(3), 449–463. doi: 10.1046/j.1365- 

2389.2002.00443.x. 

Nearing M.A., Foster G.R., Lane L.J. & Finkner S.C. (1989) A 

process-based soil erosion model for USDA–Water Erosion 

Prediction Project technology. Trans. ASAE 32(5), 1587–1593. 

Nearing, M.A., Bulygin, S.Y. & Kotova, M.M. (1998) Primary 

verification and adaptation of the WEPP model for Ukrainian 

conditions: problems, possible solutions, and perspectives. 

Pochvovedenie 31(2), 96–99. 

Norton, L.D. & Brown, L.C., (1992) Time-effect on water erosion 

for ridge tillage. Trans. ASAE 35(2), 473–478. 1992. 

Lambe, T.W. & Whitman, R.V. (1972) Mecanica de Suelos. Ed. 

Limusa. Mexico. 

Panigrahi, D.P. & Mujumdar, P.P. (2000) Reservoir operation 

modelling with Fuzzy Logic. Water Res. Manag. 14(2) 89–109. 

doi: 10.1023/A:1008170632582. 

Partheniades, E. (1965) Erosion and deposition of cohesive soils. 

Journal of the Hydraulics Division. Proc. American Society of 

Civil Engineers, 105–139. 

Prosser, I.P., Dietrich, W.E. & Stevenson, J. (1995) Flow resistance 

and sediment transport by concentrated overland flow in a 

grassland valley. Geomorphol. 13(1), 73–86. doi:10.1016/0169- 

555x(95)00020-6. 

Prosser, I.P. (1996) Thresholds of channel initiation in historical and 

Holocene times. Adv. Hills. Processes 2(7), 687–708. 

Rapp, I. (1988) Effects of soil properties and experimental 

conditions on the rill erodibilities of selected soils. Ph. D. Thesis, 

Faculty of Biological and Agricultural Sciences, University of 

Pretoria, South Africa. 

Ramirez, H. & De La Vara, R. (2004) Análisis y diseño de 

experimentos. Mc Graw Hill. México. 

Rauws, G. (1987) The initiation of rills on plane beds of noncohesive 

sediments. In: Catena Supplement, 8, 107–118. 

Rauws, G. & Govers, G. (1988) Hydraulic and soil mechanic aspects 

of rill generation on agricultural soils. J. Soil Sci. 39(2), 111–124. 

Ranieri, S.B.L, Sparovek, G., Demaria, I.C. & Flanagan, D.C. 

(1999) Erosion rate estimation using USLE and WEPP on a 

Brazilian watershed. Proc. International Soil Conservation 

Organization Conference. West Lafayette, IN. 

Renard, K.G., Foster, G.R., Weesies, G.A, McCool D.K. & Yoder, 

D.C. (1997a) Predicting Rainfall Erosion Losses: a Guide to 

Conservation Planning with the Revised Universal Soil Loss 

Equation (RUSLE). USDA Agricultural Handbook 703. US 

Government Printing Office: Washington, DC. 

Renard, K.G., Foster, G.R., Weesies, G.A. & Porter J.P. (1997b) 

RUSLE – Revised Universal Soil Loss Equation. J. Soil Water 

Conserv. 49(3), 213–220. 

Reichert, J.M., Schäfer, M.J., Cassol, E.A. & Norton, L.D. (2001) 

Interrill and rill erosion on a tropical sandy loam soil affected by 

tillage and consolidation. In: Stott, D.E., Mohtar, R.H. & 

Steinhardt, G.C. (eds.), Sustaining the Global Farm. Proc. 10th 

International Soil Conservation Organization Meeting, May 24– 

29, 1999, Purdue University and the USDA-ARS Soil Erosion 

Research Laboratory, USA. 

Santoro, V.C., Amore, E., Modica, C. & Nearing, M.A. (2002) 

Application of two erosion models to a large Sicilian basin. Proc. 

Int. Congress of European Soc. for Soil Conservation, Valencia. 

Santos, C.A.G., Srinivasan, V.S., Suzuki, K. & Watanabe, M. (2003) 

Application of an optimization technique to a physically based 

erosion model. Hydrol. Processes 47, 989–1003, doi: 

10.1002/hyp.1176. 

Savabi, M.R., Klik, A., Grulich, K., Mitchell, J.K. & Nearing, M.A. 

(1996) Application of WEPP and GIS on small watersheds in 

USA and Austria. Proc. HydroGIS 96: Application of Geographic 

Information Systems in Hydrology and Water Resources 

Management. IAHS Publication 235. 

Simons, D. B., Li, R.M. & Fullerton, L. (1981) Theoretically derived 

sediment transport equations for Pima County, Arizona. Prepared 

for Pima County DOT and Flood Control District, Tucson, 

Arizona, Colorado. 

Shainberg, I., Laflen, J.M., Bradford, J.M. & Norton, L.D. (1994) 

Hydraulic flow and water quality characteristics in rill 

erosion. Soil Sci. Soc. Am. J. 58(4), 1007–1012. doi: 

10.2136/sssaj1994.03615995005800040002x. 

Shainberg, I., Goldstein, D. & Levy, G.J. (1996) Rill erosion 

dependence on soil water content, aging, and 

temperature. Soil Sci. Soc. Am. J. 60(3), 916–922. 

doi:10.2136/sssaj1996.03615995006000030034x. 

Sheridan, G.J., So, H.B., Loch, R.J., Pocknee, C. & Walker, C.M. 

(2000a) Use of laboratory-scale rill and interrill erodibility 

measurements for the prediction of hillslope-scale erosion on 

rehabilitated coal mine soils and overburdens. Aust. J. Soil Res. 

38(2), 285–297. doi:10.1071/SR99039. 

Sheridan, G.J., So, H.B., Loch, R.J. & Walker, C.M. (2000b) 

Estimation of erosion model erodibility parameters from media 

properties. Australian J. Soil Res. 38(2), 256–284. doi: 

10.1071/SR99041. 

Silva, R.M., Santos, C.A.G. & Silva, L.P. (2007) Evaluation soil loss 

in Guaraira basin by GIS and remote sensing based model. J. 

Urban and Environ. Eng. 1(2), 44–52, doi: 

10.4090/juee.2007.v1n2.044052. 



104 

Slattery, M.C. & Bryan, R.B. (1992) Hydraulic conditions for rill 

incision under simulated rainfall: a laboratory experiment. Earth 

Surf. Processes Landf. 17(1), 127–146. doi: 

10.1002/esp.3290170203. 

Smerdon, E.T. (1964) Effect of rainfall on critical tractive forces in 

channels with shallow flow. Trans. ASAE, Paper No. 63-700. 

Smerdon, E.T. & Beasley, R.P. (1959) The tractive force theory 

applied to stability of open channels in cohesive soils. 

Agricultural Experiment Station University of Missouri Research 

Bulletin 715. 

Smith, R. E., Goodrich, D. & Quinton, J. N. (1995) Dynamic 

distributed simulation of watershed erosion: the KINEROS2 and 

EUROSEM models. J. Soil Water Conserv. 50(5), 517–520. 

Terzaghi, K. & Peck, R.B. (1967) Soil Mechanics in Engineering 

Practice, 2a. ed. John Wiley and Sons, New York. 

Torri, D. (1987) A theoretical study of soil detachability. In: Catena 

Supplement, 10, 15–20. 

Trout, T.J. & Neibling, W.H. (1993) Erosion and sedimentation 

processes on irrigated fields. J. Irrig. Drainage Engr. ASCE 

119(6), 947–963. doi: 10.1061/(ASCE)0733- 

9437(1993)119:6(947). 

Trout. T.J. (1996) Furrow erosion and sedimentation: on field 

distribution. Trans. ASAE 39(5), 1717–1723. 1996 

Trout, T.J. (1999) Sediment transport in irrigation furrows. In: 

Sustaining the Global Farm: Selected Papers from the 10th Intl. 

Soil Conservation Organization Meeting, 710–716. In: Stott, D.E., 

Mothar, R.H. & Steinhardt, G.C. (eds.) West Lafayette, Ind.: 

Purdue University and USDA-ARS National Soil Erosion 

Research Laboratory. 

Van Klaveren, R.W. & MccooL, D.K. (1998) Erodibility and critical 

shear of a previously frozen soil. Trans. ASAE 41(5), 1315–1321. 

West, L.T., Miller, W.P., Bruce, R.R., Langdale, G.W., Laflen, J.M. 

& Thomas, A.W. (1992) Cropping system and consolidation 

effects on rill erosion in the Georgia Piedmont. 

Soil Sci. Soc. Am. J. 56(4), 1238–1243. doi: 

10.2136/sssaj1992.03615995005600040038x. 

Wischmeier W.H. & Smith D.D. (1958) Rainfall energy and its 

relationship to soil loss. Trans. Am. Geophys. Union 39(2), 285– 

291. 

Yalin Y.S. (1963) An expression for bed-load transportation. 

Journal of Hydraulics Division ASCE 89(HY3), 221–250. 

Zeleke, G. (1999) Application and adaptation of WEPP to the 

traditional farming system of the Ethiopian highlands. Proc. 

International Soil Conservation Organization Conference. West 

Lafayette, Indiana. 

Zhang, X.C., Nearing, M.A., Risse, L.M. & McGregor, K.C. (1996) 

Evaluation of runoff and soil loss predictions using natural runoff 

plot data. Trans. ASAE 39(3), 855–863. 

Zhang, X.C., Li, Z.B. & Ding, W.F. (2005) Validation of WEPP 

sediment feedback relationships using spatially distributed rill 

erosion data. Soil Sci. Soc. Am. J. 69(5), 1140–1147. 

doi:10.2136/sssaj2004.0309. 

Zhu, J.C., Gantzer, C.J., Peyton, R.L., Alberts, E.E. & Anderson, 

S.H. (1995) Simulated small-channel bed scour and head cut 

erosion rates compared. Soil Sci. Soc. Am. J. 59(1), 211–218. 

doi:10.2136/sssaj1995.03615995005900010032x. 

Zhu, J.C., Gantzer, C.J., Anderson, S.H., Peyton, R.L. & Alberts, 

E.E. (2001) Comparison of concentrated flow-detachment 

equations for low shear stress. Soil Till. Res. 61(3), 203–212. 

doi:10.1016/S0167-1987(01)00207-0.

clinical waste handling and obstacles in malaysia - Centro de ...

You also want an ePaper? Increase the reach of your titles

Delete template?

Save as template?