Capturing spatiotemporal heterogeneity in fertilizer application for better modelling paddy water nitrogen and phosphorus pollution at regional scale

Accurately simulating paddy water nitrogen and phosphorus (N/P) concentration is crucial for identifying key control areas of agricultural non-point source pollution (ANPSP), especially in the paddy-dominated regions. In China, smallholder farming has resulted in significant spatiotemporal heterogeneity in fertilizer application date/ amount among different fields at the regional scale. However, due to the lack of realistic fertilizer application date/amount data at field scale, previous regional scale models have employed uniform fertilization management, neglecting spatiotemporal heterogeneity in fertilizer application date/amount. This has led to significant uncertainties in assessing paddy water N/P pollution at the regional scale. To address this issue, we developed a new model (PWNPRS) for simulating spatiotemporal hotspots of paddy water N/P pollution, based on remote sensing and a first-order kinetic model: (1) we developed a quantitative assessment method for fertilizer application date/amount based on phenological/vegetation information extracted from remote sensing datasets; (2) we modified and improved the first-order kinetic model by incorporating fertilizer application date/amount to predict paddy water N/P concentration at the regional scale. The spatiotemporal fluctuations of fertilizer application date/amount and paddy water N/P pollution were simulated by using the developed model in a paddy-dominated regions in Danyang, China, as a case study for model validation. The models performances and spatiotemporal evaluation indicate that the PWNPRS model is an innovative and promising approach to accurately capture the spatiotemporal pattern of fertilizer application date/amount and paddy water N/P concentration at the regional scale. The PWNPRS model can be extended to other paddy-dominated regions for N/P concentration prediction under data-limited conditions. Our model considers the interactive effects of fertilization date/amount and N/P concentration attenuation rates, allowing it to capture the dynamic spatiotemporal variations in paddy water N/P concentrations at a regional scale. Simultaneously, a robust and accurate model provide a theoretical basis for optimizing agricultural ecological management.