Controls on lag time in Philippine catchments identified using rainfall–runoff modelling and a generalized additive model (GAM)

Abstract

Understanding the controls upon lag time, can improve river and flood management decision-making. This study investigates the relative importance of catchment characteristics in explaining lag time variability across the Philippines. Numerically simulated 5-year return period lag times for 291 catchments were analysed using a generalized additive model (GAM) to capture non-linear relationships with location, geology, climate, topography, and land use. The 5-year return period is representative of moderate flood response, as lag time varies little across return periods. Correlation analysis and recursive feature elimination guided variable selection, while bootstrapping assessed model stability and uncertainty. Ten significant controls on lag time were identified, with relief ratio, land cover index, and catchment area most influential. The GAM achieved an R² of 0.77 and explained 84% of deviance. Land cover emerged as the only anthropogenically modifiable control, highlighting a key management lever. National hydrological observations are needed to further support model calibration.