Determining the Influence of Land Use Change and Soil Heterogeneities on Discharge, Sediment and Phosphorus

Abstract

Southern Quebec’s Missisquoi Bay, a freshwater body in the northeastern portion of Lake Champlain is threatened by algal blooms arising from excess nutrient inputs contributed by agricultural watersheds which have their outlets in the bay. A version of the Soil and Water Assessment Tool (SWAT) model, calibrated in a previous study to estimate annual runoff, sediment and total phosphorus (TP) fluxes from the Castor subwatershed into the Pike River watershed, which, in turn, flows into the Missisquoi Bay, used static landscapes and single land uses to arrive at its predictions. However, in reality, farmers do rotate crops. Therefore, the present study’s objective was to quantify the impact of soil heterogeneities on land use change patterns in the Castor subwatershed from 1999 to 2011. Data from a 24-point soil survey within the Castor subwatershed were partitioned and regionalized into 5, 10, 15, 20 and 24 heterogeneous regions or configurations. Using the standard soils map (with mean properties) employed in several prior studies in the subwatershed, a sixth configuration termed “Reference,†was also developed. All 6 configurations were factorially combined with either 1999 or 2011 land use data to yield 12 different versions of the SWAT model and quantify the heterogeneities and uncertainty of soil properties on land use change. For hydrology, it was discovered that there were no marked differences in the predictions, which was attributable to the use the SCS-CN subroutine which masks the physical properties of soil parameters within the same hydrologic group. We evaluated all the models for two periods i.e. 1991-1999 and 2000-2007. All the 1999 land use SWAT configurations underestimated runoff, sediment and TP whereas all the 2011 land use SWAT set ups gave higher and more accurate values. For both land use periods, the 5 Region models both showed higher and more accurate estimates, than those set ups with a greater number of regions, but were similar in accuracy to the Reference model set-ups. Since the 5-region configurations showed the highest within-zone heterogeneity, it can be concluded that having many regions (many sampling points as regions) does not necessarily increase SWAT’s prediction accuracy.

Keywords: Curve Number, total phosphorus, partitioning, land use