31.07.2013 Views

THE FLORIDA STATE UNIVERSITY ARTS AND SCIENCES ...

THE FLORIDA STATE UNIVERSITY ARTS AND SCIENCES ...

THE FLORIDA STATE UNIVERSITY ARTS AND SCIENCES ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

CHAPTER TWO<br />

2. LINEAR REGRESSION<br />

Anderson (1998) established that the denitrification rate could be predicted by using one<br />

of the most important factor in denitrification, i.e, organic carbon (OC %). Otis (2007)<br />

has used organic carbon and WFP to determine denitrification rates in the Wekivia river<br />

basin. Following the approach used by Anderson (1998) and Otis (2007) approach a<br />

hierarchal linear regression analysis is conducted on the dataset. Initially all data are<br />

clumped together and the relationship between the denitrification rate and organic carbon<br />

is investigated, this yields a very weak correlation (r 2 = 0.1). The same relationship is<br />

investigated using data from similar studies. While there is an improvement in the<br />

coefficient of determination, is it still below the targeted coefficient of determination<br />

value of 0.6. This section describes in detail the hierarchal linear regression analysis.<br />

2.1. All Available data<br />

Using all the available literature values the denitrification rate is plotted against organic<br />

carbon (Figure 2.1). While Anderson’s (1998) linear regression is a reasonable estimation<br />

based on the data used by him and helped account for loss of nitrogen due to<br />

denitrification, it may not necessarily be applicable in a universal setting Based on<br />

literature studies an increase in the organic carbon should result in an increase in the<br />

denitrification rate. The results obtained do show that the denitrification rate increases<br />

with an increase in organic carbon content, however the coefficient of determination is<br />

extremely low (r 2 = 0.1).<br />

There are several factors that may account for the low correlation value. It is likely that<br />

this erroneous result may be caused by the conversion of units to a common unit of kg N<br />

ha − 1 d − 1 or from the clumping together of data from different studies, many of which have<br />

used different methods. Some of the methods within the dataset estimate and measure the<br />

potential denitrification rate i.e. the maximum possible rate under ideal conditions; others<br />

are field experiments that report the actual rate of denitrification. Even within the same<br />

32

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!