Introduction
Flooding is a natural phenomenon that human societies have accepted as an unavoidable event. However, the occurrence, magnitude, and frequency of floods are influenced by multiple factors that vary depending on the climatic, natural, and geographical conditions of each region. Consequently, the relationship between rainfall and runoff differs significantly from one watershed to another. The study area is located in the northwestern part of Iran, within East Azerbaijan Province. The Sarand watershed is a sub-basin of the Aji Chai River, situated in the northeast of Tabriz. The watershed's outlet is positioned to the west of Khajeh County, while its northern boundary extends toward the city of Varzeqan. Additionally, the Heris County is located to the southeast of this watershed. This research aims to identify and map areas susceptible to flash floods in the Sarand chai watershed. By employing the Modified Flash Flood Potential Index (MFFPI) model and analyzing physiographic parameters, a flood potential map has been developed, and the role of key contributing factors to flood occurrence has been assessed. The findings of this study can be utilized for optimizing water resource management, minimizing flood-related damages, and supporting regional development planning.

2- Research Methodology
For the purpose of this study, the parameters of slope, flow accumulation, soil texture, curvature, land use, and land permeability were selected. Subsequently, the weight of each parameter was applied to its five sub-parameters, and the final score for each layer was calculated within the GIS environment. In the first stage, the integration of the six weighted layers resulted in the generation of a flash flood potential map. In the next step, approximately 200 random points were selected from the watershed using the "Create Random Points" tool in ArcGIS. The values for these points were extracted from all six layers as well as from the flash flood potential map within the GIS environment. The normality of the data was assessed using the Kolmogorov-Smirnov statistical test in SPSS software. Subsequently, correlation and multiple linear regression analyses were conducted to determine the strength of the relationship between dependent and independent variables. In the final stage, based on the results of the statistical analyses, the most influential parameters were identified, and the final flash flood potential map was generated using the significant parameters. In total, 380 points (including flooded and non-flooded areas) were extracted and organized in ArcGIS. Seventy percent of the data was used for model training and the remaining 30 percent for validation. The model’s performance was evaluated using Accuracy, Sensitivity, and Specificity indices.
3- Results
In the first stage of model implementation, the parameters of slope, slope curvature, drainage density, lithology, soil texture, and land use were used. In the second stage, based on the results from the multiple regression analysis, layers with minimal influence were removed. The layers of slope, lithology, and soil texture, which had the greatest impact, were then used to produce the final flash flood map. According to the final map, 36% of the Sarand chai watershed area lies in the very high and high flood risk zones. Based on the spatial distribution of risk zones, areas with very high and high risk are primarily located in riverbeds and low-gradient sedimentary plains perpendicular to the river's axis. In contrast, areas with medium risk are more common in upper terraces, while low and very low-risk areas are mostly found in mountainous and hilly elevations. This pattern highlights the significant role of slope, elevation, and geomorphological characteristics in determining the potential for flash floods in the region.
4- Discussion & Conclusions
The flash flood susceptibility map for the Sarand Chai watershed was generated using physiographic parameters and the Modified Flash Flood Potential Index (MFFPI) model within the ArcMap software environment. The map was classified into five susceptibility zones: very low, low, moderate, high, and very high. To examine the relationships between the variables, Spearman’s correlation coefficient test and multiple linear regression analysis were performed using SPSS software. In this analysis, the physiographic parameters were considered as independent variables, while the flash flood susceptibility layer was treated as the dependent variable. The results of the correlation analysis revealed that slope, lithology, and soil texture had the greatest influence on flash flood occurrence, whereas flow accumulation and curvature had the least impact in this watershed. In the final stage, the three most influential parameters—slope, lithology, and soil texture—were identified and selected. A new flash flood susceptibility map was then produced by applying the respective weights to these layers. The results of the susceptibility zoning map analysis showed that areas with very high, high, moderate, low, and very low flood risk accounted for 19.3%, 16.5%, 25.4%, 24.3%, and 14.3% of the total watershed area, respectively. The assessment of the MFFPI model’s accuracy and performance demonstrated that the model performed well in generating the flood hazard susceptibility map, with coefficients of 87.59% and 88.59% for the training and validation datasets, respectively.