A step toward considering the return period in flood spatial modeling

In recent years, there has been an increasing interest in spatial modeling, and flood hazard prediction is a major area of interest within the field of hydrology. It is necessary to consider return periods for identifying the flood hazard zones. In hydraulic modeling such as HEC-RAS, this is usually done, but in spatial modeling by machine learning (ML) models, this has not been taken into account so far. This study seeks to obtain data that will help to address this research gap. The Sentinel-1 Radar images have been used for identifying the flooded locations in different return periods. An embedded feature selection algorithm (i.e., recursive feature elimination random forest; RFE-RF) was used in the current research for key feature selection. Then, three ML models of neural networks using model averaging, classification and regression tree, and support vector machine were employed. The flood hazard prediction demonstrated a great performance for all the applied models (i.e., accuracy and precision > 90%, Kappa > 88%). Sensitivity analysis disclosed that the variables of elevation and distance from stream are in the first importance order, the variables of precipitation, slope, and land use are in the second importance order, and other variables are in the third importance order in all return periods. The modeling results indicated that among man-made land uses the irrigated area between 17.7 and 31.4%, dry farming from 0.5 to 2.4%, and residential areas between 8.3 and 25.1% are exposed to high and very high flood hazard areas. The current findings add to a growing body of literature on the spatial modeling of floods.