Gaining prediction accuracy in land use modeling by integrating modeled hydrologic variables

Abstract Land use and hydrology are interdependent, so that land use modeling could benefit from hydrologic modeling. This study aims to integrate spatial predictions of hydrologic variables as provided by a hydrologic model into a land use model for a rapidly urbanizing catchment in India. The benefits of this integration are quantified by comparing predictions of a land use model that uses a basic set of explanatory variables to a land use model that additionally uses the modeled hydrologic variables. Our results indicate that the integration of the hydrologic variables improved the model accuracy indicated by overall accuracy (+3 and + 4 percentage points (pp)), class specific user and producer accuracies (up to +8 pp) and figure of merit (+4 and + 5.3 pp) when compared with land use classifications at two points in time. Moreover, the land use patterns show that the integration of the hydrologic variables helped to avoid allocation errors.