Advancing Water Resources Research and Management

Quantifying nutrient loads contributed by non-point source pollution at the watershed scale is a difficult undertaking at best. Extensive water quality sampling and stream flow gauging during base flow and storm events is normally necessary to get realistic values. The mass sampling approach is usually too expensive for the evaluation of large watersheds. In this research, a methodology is presented for the development of a simplified, inexpensive combination of limited sampling strategies, utilization of a Geographic Information System (GIS) database and mathematical modeling techniques to characterize non-point source loads and assess pollution potential in the sub-basins of large watersheds.

Three different models were developed for the purpose of estimating nutrient concentrations and loads for streams in 29 sub-basins from water quality data collected in eight intensively sampled sub-basins. The models developed were designated as the concentration model, the load model and the loading factor model.

The concentration model appeared to perform better than the load model or the loading factor model, but the loading factor model has some attributes that the author suggests make it worth further investigation. Results from the methodologies presented were compared with measured values and the best estimates of other researchers with favorable results. Average annual sub-basin concentrations and loads along with total watershed loads were estimated for the Arkansas portion of the Illinois River basin.

__________________________

¹Visiting Assistant Professor, Department of Civil Engineering, 4190 Bell Engineering, University of Arkansas, Fayetteville, Arkansas 72701 (501-575-7535)(rdw@engr.uark.edu)

The objective of this study was to develop and test methodologies for estimating non-point source pollution loads in multiple watersheds utilizing a limited sampling strategy, (GIS) databases and computer models. In watershed characterization studies, the majority of costs are associated with sampling and sample analysis. If a methodology can be developed which would reduce the total number of samples collected but still give a reasonably accurate characterization of the watershed, the costs associated with such studies could be significantly reduced.

Sampling of representative basins in a large watershed and extrapolation of the data collected to other basins with similar characteristics may be a valid, economical alternative to intensive sampling of all basins in a given watershed to obtain the same results. If the method reasonably simulates the non-point source loads in a large watershed, significant savings in analytical and labor costs can be achieved in future watershed analysis and characterization.

The goals of the project were as follow:

• Develop a database containing base flow water quality information for all 37 sub-basins of the Illinois River in Arkansas.
• Develop a database containing storm event information on 8 representative sub-basins and limited storm event information from other sub-basins for purposes of comparison to the results of predictions.
• Utilize a GIS database for the study area documenting land use, land cover, geology, soils, topography and other pertinent characteristics of the study area.
• Determine any functional relationships between land use and areal surface water quality.
• Develop and utilize predictive models to estimate pollutant loads in the sub-basins and watershed.
• Determine total loads of selected parameters (Total Phosphorus, Total Nitrogen, and Total Suspended Solids) contributed by the Arkansas portion of the Illinois River at the watershed scale and the individual contributions from the sub-basins.

The water quality and GIS databases were successfully developed and the information obtained was used to determine relationships between spatial characteristics and land use to water quality at the sub-watershed and watershed level. Two predictive models were developed through the use of multiple linear regression and one predictive model developed using trend analysis.

The concentration models and load models were developed in the form:

L or C = |C₀ + C₁(U₁) + C₂(U₂) …… + C_n(U_n)|

where:

L = Load (kg/ha-year)

C = Average Annual Concentration (mg/L)

C₀…C_n are coefficients

U₁…U_n are percentages of land use types and a shape factor equal to stream length divided by the basin area.

The loading factor models were developed in the form:

L = C₁A₁ + C₂A₂ ... + C_nA_n

where:

L = Load (kg/ha-yr)

C₁...C_n are loading factors

A₁...A_n = area of a specific land use (ha) 

For overall performance, the concentration model compared more favorably to measured concentrations and loads in headwaters basins, and estimates from other sources. Figure 1 shows the measured values for total phosphorus loads in kg/ha-yr as a line plot of the actual data. The trend lines from the concentration model, the load model and the loading factor model are shown. The concentration model trend line is shown to fit the actual data set better than the other two models. Similar results were obtained for total nitrogen and total suspended solids. In the author's opinion, the concentration model would be the best method to use in estimating concentrations and loads for the system. The loading factor model has some attractive features in that it is easy to apply and relatively easy to understand. Refinement of information gathering techniques to develop this type of predictive model should be further investigated.

It is apparent to the researcher that more accurate results could have been obtained for all predictive methods if more accurate storm and base flow estimates for both the intensively sampled sub-basins and all other basins could have been obtained. The concentration model probably performed the best because it was the least affected by flow. Both the load model and the loading factor model results could have been affected adversely by marginally accurate flow estimates.

Other factors involved (and beyond the researcher’s control) were the overall variability in multiple portions of the system, unregulated, undocumented land use practices (for example, animal waste land application amounts and timing), uncertainty in the role of groundwater influences upon the stream system and rapidly changing urban/suburban growth patterns. Rainfall events in Northwest Arkansas also tend to be intermittent and irregular in spatial distribution and intensity, making rainfall patterns and storm event simulations difficult to perform.

Though it is suspected that non-point source nutrients are utilized by biota in the stream system, no decay rates were applied to projected storm flow or other non-point source nutrient inputs. Nutrient inputs could have been overestimated by this work if a decay rate is applicable.

In conclusion, considering all the factors involved, the work appears to be as good at watershed characterization as much more sophisticated models. The limited sampling strategy shows potential for cost-effective watershed studies.

Proceedings TOC

home page

Maintainer: AWRA Webserver Team
Copyright © 1999 American Water Resources Association