Assessment of uncertainty in soil test phosphorus using kriging techniques and sequential Gaussian simulation: implications for water quality management in southern Quebec

Abstract

Missisquoi Bay, located in southern Quebec at the north-eastern extremity of Lake Champlain, is subject to eutrophication arising from excess nutrients, predominantly phosphorus (P), contributed by agricultural watersheds. Factors such as land use pattern, management practices, soil properties and geomorphology have an impact on soil P levels. Land patches under different management practices introduce a cyclic pattern, especially when fitting the variogram. Geostatistics procedures were used to model soil test phosphorus (STP) within the 11 km2 Castor Watershed in southern Quebec, Canada. An ordinary kriging (OK) method was used to estimate STP at unsampled points, but this had a smoothing effect, resulting in an underestimation of high values and overestimation of low values. Therefore, a more efficient technique was needed to draw predictions from the conditional probability distribution at the simulation grid nodes. A sequential Gaussian simulation (SGS) was adopted for this purpose, and used to create 50 equal probable realizations. Uncertainty was modelled using the E-type (mean) of the realizations, which ranged from 12.5 to 223 mg P kg–1 soil. The adequate spatial probability pattern for STP is a valuable criterion when seeking to delineate areas of high STP for site-specific best management practices (BMPs).