Land use plays an important role in shaping the Earth’s surface. However, the effects of land use and land cover change (LULCC) on runoff are generally not well represented in Earth system models. For example, deforestation is typically observed to decrease runoff, while the Energy Exascale Earth System Model (E3SM), a model that was recently developed based on the Community Earth System Model, predicted a decrease in runoff after deforestation. This system-wise model bias may lead to unreliable conclusions when the model is used for scenario analysis in climate and environmental studies. We thus constrained the offline E3SM land model (ELM) simulations with FLUXNET-observed evapotranspiration and a new global gridded runoff data product. The constrained ELM successfully matched observed effects of LULCC on runoff. The constrained ELM was then used to study LULCC effects on global flood risk over the past 50 years (1960-2010). Results showed that LULCC increased flood risk in west tropical Africa by about 2 times and in Paraguay River by about 1.5 times, where large deforestation occurred during this period. In contrast, LULCC decreased flood risk in Central Europe by about 50% because of large forest gains during this period. These findings are useful for evaluating flood risks associated with LULCC.