LNAPL Modeling recovery and endpoint example

LNAPL in the subsurface 

LNAPL Modeling 

recovery and endpoint example 

Randall Charbeneau, P.E. 

Professor of Civil Engineering, University of Texas 

& 

Mark Adamski, P.G. 

Technical Specialist and Environmental Business Manager, BP America

Objective 

1. Describe the API 

LNAPL Liquid 

Distribution and 

Recovery Modeling 

Tool (LDRM). 

2. Demonstrate 

application of the 

modeling tool for a 

petroleum refinery. 

LNAPL in the subsurface

LNAPL Liquid Distribution and Recovery 

Modeling (LDRM) Tool 

Purpose: provide a simple, physically based model to assess 

LNAPL distribution and recovery using conventional LNAPL 

liquid recovery technologies (single and dual-pump wells, 

vacuum-enhanced wells, skimmer wells, and trenches) based 

on LNAPL thickness in a monitoring well 

Chronology: 

• 1999 Release (API #4682) – dual spreadsheet model using Brooks and 

Corey parameters and a simplified relative permeability function (based 

on the Burdine equations) 

• 2003 Release (API #4729) – multiple spreadsheets (stand alone) using 

van Genuchten parameters with different spreadsheets for Burdine and 

Mualem relative permeability models, and for single and two-layer 

representations 

• 2007 Release (API 4760) – single “Windows” application using van 

Genuchten parameters with option of 1-, 2- or 3-layers, relative 

permeability models, units, data representation, and vertical gradient 

through FGS 


Recovery Well Application 

LNAPL Plume may be covered by multiple capture 

regions, each considered separately 


Scenario-Based Model 

Model Domain Consists Only of a “Capture Region” 

Required Site Parameters 

Q w 

R c 

• LNAPL thickness, b n 

• Ground surface elevation, z gs 

• Water table elevation, z wt 

• Elevation of soil facies interface, 

z 12 

, z 23 

Radial Flow to a 

Recovery Well 


Required Soil and Fluid Parameters 

Soil Parameters 

• Porosity, n 

• Hydraulic conductivity, K ws 

• Van Genuchten, α 

• Van Genuchten, N 

• Irreducible Water Saturation, S wr 

Fluid Parameters 

• LNAPL density (ratio), ρ r 

• LNAPL viscosity (ratio), µ r 

• Air/water surface tension, σ aw 

• Air/LNAPL surface tension, σ an 

• LNAPL/water interfacial tension, σ nw 


Soil Heterogeneity 

Up to three soil layers with abrupt vertical facies transition 

K w1 

, n 1 

, α 1 

, N 1 

, S wr1 

z 12 

K w2 

, n 2 

, α 2 

, N 2 

, S wr2 

z 23 

K w3 

, n 3 

, α 3 

, N 3 

, S wr3 


Well Parameters 

• Recovery time, T R 

• Radius of pumping well, R W 

• Radius of capture, R C 

• Radius of Influence, R I 

• Groundwater production rate, Q w 

• Water saturated thickness at well, b w 

• Suction pressure (vacuum-enhanced system), P v 

• Screen length (vadose zone), b a 

• Air radius of capture, R A 

• If Q w 

= 0 and P v 

= 0 then Skimmer Well is assumed 


Application to a Closed Refinery Site Located Near 

the Missouri River 


Application of LLRM 

Recovery well at a industrial site in the Midwestern U.S. 


Cross-Section Layout 


Cross-Section A-A’ 


Site-Specific Data 

• Recovered fluid measurements 

• LNAPL density and viscosity 

• Surface and interfacial tension values 

• Soil core measurements 

• Grain size distribution 

• Capillary pressure curves 

• Hydraulic conductivity 

• LNAPL saturation (Dean Stark method) 

• Site-specific 

• LNAPL recovery rates 

• Groundwater recovery rates 


LNAPL Thickness and Water Table 

Elevation 

LNAPL Thickness (ft) 

20 

18 

16 

14 

12 

10 

8 

6 

4 

2 

LNAPL Thickness (Gray) and 

Water Table Elevation (Blue) 

with Datum at 704.5 ft 

0 

Aug-96 Aug-98 Aug-00 Aug-02 Aug-04 

Date 


Monitoring Well Data 

These data suggest that for this site with LNAPL trapped 

beneath FGZ, monitoring well LNAPL thickness will 

probably be a poor indicator of the amount of LNAPL 

present within the subsurface 

• Application of LDRM with no model calibration 


Fitting Capillary Pressure Curves – General Case 

12.0 

4 - 714' 

4 - 709' 

Capillary Pressure Head (m) 

10.0 

8.0 

6.0 

4.0 

2.0 

Burdine 

a = 0.80 m-1 

N = 3.50 

Swr = 0.30 

Mualem 

a = 0.20 m-1 

N = 5.0 

Swr = 0.55 

4 - 707' 

5 - 708' 

5 - 707' 

5 - 705' 

6 - 709' 

6 - 707' 

6 - 704' 

vG-B 

vG-M 

0.0 

0.0 0.2 0.4 0.6 0.8 1.0 1.2 

Saturation 


Early 1-Layer Model Representation 


Groundwater Pumping and Lowered Water Table 

(Smearing) 


Initial Recovery Estimate for all Wells 

350,000 

Total LNAPL Recovery Volume [gallons] 

300,000 

250,000 

200,000 

150,000 

100,000 

50,000 

Total Modeled NAPL Recovery (All Six Wells - Dual 

Pumping) 

Total Actual NAPL Recovery (All Six Wells - 2 yrs) 

0 

0 1 2 3 4 5 6 

Time [yr] 


LNAPL recovery – 

model predicted and actual (2 years) 

350,000 


300,000 

250,000 

200,000 

150,000 

100,000 

50,000 


Pumping) 


0 

0 1 2 3 4 5 6 

Time [yr] 



model predicted and actual (2.75 years) 

300,000 

Modeled vs. Actual LNAPL Recovery for the Lower Refinery Recovery Well System: 

June 1, 2003 through May 1, 2006 (35 Months) 


250,000 

200,000 

150,000 

100,000 

50,000 

Total Dual PumpModeled NAPL Recovery (All Six Wells) 

Actual Recovery Summary 

0 

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 

Time [yr] 




300,000 


June 1, 2003 through Sep. 14, 2006 (39.5 Months) 


250,000 

200,000 

150,000 

100,000 

50,000 

Total Dual PumpModeled NAPL Recovery (All Six Wells) 

Actual Recovery Summary 

0 

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 

Time [yr] 




350,000 


June 1, 2003 through Aug. 21, 2008 (62.1 Months) 


300,000 

250,000 

200,000 

150,000 

100,000 

50,000 


Pumping) 


0 

0 1 2 3 4 5 6 

Time [yr] 


Why was the model off? (not that it was very far off) 

• Was it that - the model or algorithms are pretty good but 

not that accurate? 

• Was it that – the soil is too variable away from well? 

• Was it that – the LNAPL Impacts vary away from well? 

Nope 


Water level in Missouri River makes constant 

pumping assumption difficult 

Groundwater Pumping Rate 

System Wide 

50 

45 

One layer model 

timeframe 

Actual Pumping Rate 

Average Pumping Rate 

Modelled Pumping Rate 

31.9 gpm 

40 

Groundwater Pumping Rate (GPM) 

35 

30 

25 

20 

15 

26.0 gpm 

23.7 gpm 

14.5 gpm 

30.5 gpm 

15.1 gpm 

21.7 gpm 

10 

5 

0 

Jun-03 Dec-03 Jun-04 Dec-04 Jun-05 Dec-05 Jun-06 Dec-06 Jun-07 Dec-07 Jun-08 


Vertical Distribution Comparison – 1 Layer versus 3 Layer 

740 

735 

730 

LNAPL Saturation & Relative Permeability 

0.0 0.1 0.2 0.3 0.4 

LNAPL thickness 7 ft 

LNAPL Thickness (ft) 

20 

18 

16 

14 

12 

10 

8 

6 

4 

2 

0 

LNAPL Thickness (Gray) and 

Water Table Elevation (Blue) 

with Datum at 704.5 ft 

Aug-96 Aug-98 Aug-00 Aug-02 Aug-04 

LNAPL Saturation & 

Relative Permeability 

0.0 0.2 0.4 

Date 

Elevation [ft] 

725 

720 

715 

710 

• LNAPL thickness 7 ft 

• LNAPL thickness 18 ft 

• 3 layer accounts SCM 

better 

• Layering 

• LNAPL thickness 

• etc 

705 

700 


Models are non-unique solutions 

The better model will correlate to more observations 

• January / February 2007 LNAPL recovery model at the site 

was revisited and updated using the latest understanding and 

newest three layer version of the API LDRM. 

• Site conditions were unconfined and confined aquifer / LNAPL 

conditions versus the 1-Layer model assumed unconfined 

conditions 

• 1 Layer only modeled the sand layer that contained majority of LNAPL. 

• 1 Layer model could not account for large LNAPL thickness observed with 

high water table conditions. 

Most Significant Difference 

for Vertical LNAPL Distribution 

• Small contributions/impacts of LNAPL recovery from the upper fine grained 

soil layer were not accounted for (but not a big deal) 

• 1 Layer model did not illustrate why higher water-table resulted in higher 

recovery rates or larger gauged thicknesses 

• At this site as water-table rose LNAPL thickness observed increased 

• 3 Layer model sets realistic expectation regarding large gauged 

thicknesses versus impacts by accurately accounting for the vertical 

variability of soil characteristics 

• During the 3-layer model prep - accounting for variable historic 

water pumping rates was also identified as necessary for improved 

accuracy 


Most Significant Difference 

for Recovery Performance

Updated site model with new 3 layer model 

(initial 3 layer model from 2/07) 

400,000 

Cumulative LNAPL Recovery at Six Lower Refinery Recovery Wells 

Actual LNAPL Recovery vs. One-Layer and Three-Layer Model Results: 6/19/2003 to 6/22/2008 

350,000 

Cumulative LNAPL Recovery Volume (Gallons) 

300,000 

250,000 

200,000 

150,000 

100,000 

50,000 

0 

One-Layer (2005) 

Three-Layer Model (2007) 

Actual 

0 1 2 3 4 5 6 

Time (Years) 

updated through 5/1/2008 


AK2 

Updated site model with new 3 layer model 

(1.75 yrs of additional recovery) 

400,000 

Cumulative LNAPL Recovery at Six Lower Refinery Recovery Wells 

Actual LNAPL Recovery vs. One-Layer and Three-Layer Model Results: 6/19/2003 to 8/21/2008 

350,000 

Cumulative LNAPL Recovery Volume (Gallons) 

300,000 

250,000 

200,000 

150,000 

100,000 

50,000 

2 Recent Increases in Q w from 

15.1 to 21 and 32 GPM 

One-Layer (2005) 

Three-Layer Model (2007) 

Actual 

0 

0 1 2 3 4 5 6 

Time (Years) 


updated through 5/1/2008

Slide 29 

AK2 

Mark, for this slide if you double click on the graph, the actual data spread sheet and the 1-layer model spread sheet are in here so 

you can update this graph as you like. You can make whaterver version you like, exit out of the excel sheet, make a copy, paste 

special as a metafile and then make a new slide with the picture but the data is then only in one slide to limit file size 

akirkman, 8/28/2008

US Army Corp of Engineers makes constant 

pumping assumption difficult 

Groundwater Pumping Rate 

System Wide 

50 

45 

One layer model 

timeframe 

Actual Pumping Rate 

Average Pumping Rate 

Modelled Pumping Rate 

31.9 gpm 

40 

Groundwater Pumping Rate (GPM) 

35 

30 

25 

20 

15 

26.0 gpm 

23.7 gpm 

14.5 gpm 

Three layer 

model timeframe 

15.1 gpm 

21.7 gpm 

10 

5 

0 

Jun-03 Dec-03 Jun-04 Dec-04 Jun-05 Dec-05 Jun-06 Dec-06 Jun-07 Dec-07 Jun-08 


Recovery Endpoint – When to Stop LNAPL Liquid 

Recovery 

Performance-based endpoint: percent of LNAPL in recovery 

liquid Q n 

/(Q n 

+ Q w 

) 

1.4 

Percent LNAPL in Recovery Liquid 

1.2 

1 

0.8 

0.6 

0.4 

0.2 

0 

Water Production = 9.7 gpm 

0 2 4 6 8 10 12 

Time (years) 


Same Data – Easier to pick endpoint 

Percent LNAPL in Recovery Liquid 

10 

1 

0.1 

0.01 

Time (years) 

0 2 4 6 8 10 12 

0.001 


Cumulative well recovery: All wells since 1988 

(approaching a natural endpoint for this system?) 

Total Cumulative LNAPL and Model-Predicted LNAPL Recovery - Six Remaining Wells 

Cumulative LNAPL Recovery (gallons) 

4,000,000 

3,500,000 

3,000,000 

2,500,000 

2,000,000 

1,500,000 

1,000,000 

500,000 

Total Six Well Recoverable LNAPL 

(3.06MM gallons) 

Currently 93% recovery 


95% of Recoverable LNAPL 

Recovered by 2013 


Cumulative LNAPL Recovery 

3-layer Model-Pedicted LNAPL Recovery 

Cumulative Groundwater Recovery 

Model-Predicted GW recovery 

300,000,000 

250,000,000 

200,000,000 

150,000,000 

100,000,000 

50,000,000 

0 

1988 

1988 

1989 

1990 

1991 

1992 

1993 

1994 

1995 

1996 

1997 

1998 

1999 

2000 

2001 

2002 

2003 

2004 

2005 

2006 

2007 

2008 

2009 

2010 

Cumulative Groundwater Recovery (gallons) 

2011 

2012 

2013 

0 


Key Points 

• API LDRM is a free and fairly simple tool 

• With good site data and an accurate conceptual site model we 

can estimate LNAPL distribution in the subsurface 

• With good site data and an accurate conceptual site model 

AND GOOD JUDGEMENT we can make good predictions for 

LNAPL recovery 

• Combine all the above and with good understanding of LNAPL 

behavior, we can have more productive discussions about 

expectations for LNAPL recovery and ultimately LNAPL 

endpoints 

• Progress of LNAPL recovery Estimates: 

with the old pancake model – within two orders of magnitude 

using the LDRM and site data – within one order of magnitude 

careful use of the LDRM by an experienced user applying GOOD 

JUDGEMENT – within a factor of 2 to 3.

LNAPL Modeling recovery and endpoint example

Create successful ePaper yourself

Delete template?

Save as template?