Simulating rice pesticide runoff at basin level via an integrated RICEWQ-AnnAGNPS modeling system

Thiobencarb is a widely applied herbicide in Northern California rice fields. The offsite movement of thiobencarb residues from upstream rice fields has potential to threaten non-targeted aquatic species in the downstream waterbodies. Ecohydrological models are often employed to evaluate the fate and transport of pesticide at watershed scale. However, it is hard to correctly reflect pesticide fate in both water and paddy sediment phase, because of the relative complicated hydrological and water quality conditions (ponded condition, irrigation/drainage management) of rice fields. Models designed specific for rice pesticide simulation are often at a field scale, and pesticide transport to channel and in-stream system is lacking. Furthermore, for a complicated basin including rice farming, , an integrated modeling system is needed not only to take loadings from rice fields, but also from other landuses, for example, orchards. To fill these research gaps, we integrated field level RICEWQ model with watershed level AnnAGNPS model, with RICEWQ simulating rice farming and AnnAGNPS simulating other landuses as well as stream processes. The objective of this study is to demonstrate the application of this new integrated modeling system in the Colusa River basin, where rice fields are dominant crops, but also containing many other landuses. Detailed irrigation information was collected for non-rice crops to better reflect the runoff volumes in summer season, when rainfall is usually not available in this typical Mediterranean climate region. We calibrated/validated the integrated modeling system using data collected at Colusa Basin Drain station 5 (CBD5). Furthermore, to increase the simulation efficiency for each iteration during model calibration, the newly developed AnnAGNPS Subsetting Tool for Independent Runs (ASTIR) was used to extract information from a large AnnAGNPS watershed dataset, and then invoke multiple instances of AnnAGNPS to run concurrently on the subsets.