Using machine learning models, remote sensing, and GIS to investigate the effects of changing climates and land uses on flood probability

[Display omitted] •The effect of climate change on flood risk in the future using machine learning models.•The effect of land use change on flood risk in the future using machine learning models.•Develop susceptibility model to delineate flood probability zones.•Results of both methods were compared and their efficiency was assessed. The purpose of this study is to investigate the effects of climate and land use changes on flood susceptibility areas in the Tajan watershed, Iran. To do this, land use changes over the next 20 years (2019–2040) were predicted from land use changes of the past 29 years (1990–2019) using the land change modeler (LCM) method. Future climate change was projected for the next 20 years (2020–2040) based on climate data from 1990 to 2015 using Lars-WG software and two scenarios, RCP2.6 and RCP8.5. Twelve factors that influence flooding and 262 locations of past floods were used to model the spatial pattern of flood susceptibility in the watershed. A random forest (RF) model and a Bayesian generalized linear model (GLMbayes) were used to predict areas susceptible to flooding. The results showed that elevation (21.55), distance from river (15.28), land use (11.1), slope (10.58), and rainfall (6.8) are the most important factors affecting flooding in this basin. The factors were modified according to land use changes and climate changes and the models were revised. The land use and climate forecasting in this region indicate that land use change, like decreased forest cover (−77.19 km2) and reduced rangeland (−218.83 km2) near rivers and downstream, can be expected and rainfall is projected to increase (under from both scenarios). These changes would result in increased probabilities of flooding in the downstream portion of the watershed and near the sea. The area-under-the-curve evaluation of the models indicates that the RF model more accurately predicted flood probability (0.91) than did the GLMbayes model (0.847).