Flood analysis using satellite imagery and machine learning within Google Earth Engine: A catchment-based study in Northern Iran

Knowledge about the location and extent of flooded areas in large catchments with different rainfall- runoff response in each sub-catchment is of key importance for planning flood management strategies. Haraz catchment with an area of more than 4000 square kilometers is located in the north of Iran and is frequently affected by floods. The lack of reliable spatiotemporal information on flood occurrence has been the main limiting factor for assessment of flood hazard and risk in this catchment.

The current availability of satellite remote sensing sensors with high spatial and temporal resolution is highly valuable for detailed analysis of individual flood occurrence across various scales. In this study, we develop a machine learning approach using data from various remote sensing sensors including Landsat, Planet and Sentinel-2 to detect flood events in different tributary areas within the Haraz catchment which have occurred between 2015 and 2021. The random forest algorithm implemented in Google Earth Engine was used for image classification before and after flood events. The areas of each landcover type inundated by flood waters were calculated for the single flood events and the binary flood masks were overlaid on the study area. The results have revealed that seven flood events could be detected, whereas the two events in April 2015 and April 2019 had led to the largest areas of inundation because of the nature of these floods as riverine flood. Moreover, we have found that two parts of the river network – one in middle part of Norroud subcatchment adjacent to Baladeh City and another one in the area of the catchment outlet - have the largest potential for flood risk because of the frequency of inundation and the high vulnerability of built-up areas that occupy the floodplain. Thus, the findings of this study form the basis for a better understanding of the characteristics for recent flood hazard and risk in Haraz catchment.