Prediction of spatial variability of phosphorous over the ST-Esprit watershed

Spatial data analysis tools for predicting the variability of non-point source pollutants minimize the time, effort and cost involved in extensive and exhaustive real field data measurements. In this study, exploratory data analysis, fitting of semivariogram models, and kriging techniques of geostatistics were used to develop the spatial variability map of soil phosphorous saturation (P sat) percentage over the St-Espirit watershed (2610 ha), located in Quebec, Canada. The P sat measured values for the 281 geo referenced land parcel units (LPU) within the watershed were interpreted and analyzed using the ArcGIS® tool. The geostatistical extension module of ArcGIS® was used for exploratory data analysis, semivariogram model fitting, and development of a P sat prediction map using the ordinary kriging technique. Using these geostatistical procedures and adjustment of lag sizes and lag intervals representing the data sets, it was estimated that the spherical semivariogram model fitted well to represent the P sat variability with residual sum square (RSS) of 0.0003 and coefficient of determination (R 2) of 0.98. Further, the developed model was used to predict the P sat variability over the St. Esprit watershed using the 1605 geo-referenced LPU locations. The generated spatial variability map was geo-spatially processed with the natural drainage network and land use feature classes of the watershed to ascertain the phosphorous loading and locate vulnerable LPUs for phosphorous management. It was observed that the P sat levels were higher at the up stream locations and near the drainage channels than the locations close to watershed outlet. Also, the land pockets with more than 60% agricultural land use resulted in supra-optimal P sat values (10% > P sat 20%). It was also revealed that, around 23.5% of the watersheds cropped area has reached these threshold levels which necessitate judicious P input management.