A Bayesian hierarchical spatially explicit modelling framework to examine phosphorus export between contrasting flow regimes

Abstract

We examine the ability of a SPARROW-based model (SPAtially Referenced Regression On Watershed attributes) to assess regional P export coefficients that can assist with evaluation of nutrient mitigation projects and support adaptive watershed management. Limitations in number of tributary monitoring stations were overcome by assembling multi-agency water quality data from provincial, municipal, citizen science, and academic programs. We introduced a Bayesian hierarchical framework designed to guide parameter estimation from tributary nutrient loading in southern Georgian Bay drainage basin during contrasting flow regimes, such as dry and wet years. Agriculture was identified as a major non-point P source representing between 30 and 48% of delivered P loading. Our source apportionment predicted TP loss rates from croplands that exceeded those from forested areas by 320% during dry years and by 360% during wet years, while low intensity agricultural areas (hay and pasture) exceeded P export from forests by a mere 20% and 30%, respectively. Our study identified urban runoff as another significant non-point nutrient source displaying the highest variability between dry and wet years. In particular, owing to the extensive urbanization in the Lake Simcoe watershed, urban runoff contributed nearly half of delivered P loading from tributaries into the lake. The nutrient loading management plan for Lake Simcoe calls for a reduction in P loading by ∼40% from a long-term average of 72 t P y−1 in 2002–2007 to 44 t P y−1 by 2045. Our analysis emphasizes the importance of mitigating urban non-point sources together with efforts to control agricultural runoff.