A simple model for predicting agronomy floods in rice fields in Bicol, Philippines

Climate change is expected to intensify the impacts of flood events on agricultural production, particularly in flood-prone regions like the Philippines, where rice farming is heavily affected by frequent typhoons. Flood forecasting and early warning systems can aid in mitigating these risks; however, the insufficient coverage of hydrometric monitoring stations and limited computational resources can be barriers for developing countries. Remote sensing technology offers a promising solution to bridge these gaps, providing critical hydrometric data and enabling more accessible flood prediction models. Leveraging high-spatial resolution, remote sensing-based flood extent data specifically developed for rice fields, we explore the possibility of predicting agronomical flood extent in the Bicol region of the Philippines using a series of simple logistic regression models with different lookback windows. The model predictors only included rainfall at two spatial scales and flow accumulation. The best-performed model, with three-day lookback window, captured 65% of variation in flooding among events. However, the best model did not predict well the variation in flooding within basins, nor did it account for the heterogeneity in the response of flooding to rainfall among basins. We suggested several avenues for improving the model, including incorporating basin characteristics and additional predictors for better capture variation in flooding within and among basins.