An ensemble SPARROW framework for assessing phosphorus export risk and supporting adaptive management in the northern Lake Erie basin.

内容摘要

Phosphorus (P) loading data from 95 tributary monitoring stations were assimilated into a Bayesian SPARROW ensemble framework to quantify long-term total phosphorus (TP) delivery to the northern Lake Erie shoreline. Ten covariate-based model configurations, representing alternative agricultural source proxies and land-to-water delivery controls, were calibrated and synthesized using performance-weighted averaging to generate spatially explicit estimates of TP export, delivery, and in-network retention across 865 catchments. The ensemble produced a robust delivered-load benchmark of approximately 1200-1300 t P y-1, providing a defensible baseline for adaptive management and for evaluating progress toward the International Joint Commission's 40% TP reduction target. Source apportionment indicates that agricultural non-point sources dominate basin-scale loading, contributing approximately 82% of total TP load (∼1045 t P y-1), with high-intensity cropland alone accounting for ∼65% (∼818 t P y-1). Urban sources contribute ∼16% (∼205 t P y-1), while natural land cover remains minor at ∼1.4% (∼18 t P y-1). A novel finding is the identification of greenhouse operations as a substantial contributor to TP loading (∼15%, ∼187 t P y-1), particularly within the Essex Region, Grand River, and Long Point Region conservation authorities. Delivered fertilizer losses exceed those from manure applications, and the predicted spatial patterns suggest that Ontario's Nutrient Management Act, together with associated investments in beneficial management practices, has effectively reduced manure-derived P losses. This provides a novel basin-scale evidence of conservation policy effectiveness. Low in-stream attenuation (∼0.1% km-1) indicates limited natural buffering capacity, whereas reservoirs retain approximately 17-23% of incoming TP, highlighting engineered impoundments as complementary mitigation features. Achieving the 40% reduction target will therefore require sustained fertilizer input management, transport-focused BMPs in tile-drained landscapes, and explicit consideration of legacy P pools and temporal lags. Uncertainty surrounding greenhouse, and potentially manure, transport dynamics underscores the need for improved monitoring in identified critical areas. Although the northern Lake Erie basin contains an extensive monitoring network of approximately 300 stations, data fragmentation limits calibration efficiency and contributes to structural uncertainty. Harmonized datasets would improve model constraint and strengthen cross-scale integration with mechanistic watershed models. Overall, the ensemble SPARROW framework provides both a basin-wide TP baseline and a data-driven constraint for improved policy evaluation, adaptive management, and watershed model integration.