Document Type : کاربردی
Authors
1
Professor of Climatology, Department of physical Geography, Faculty of Social Sciences, University of Mohaghegh Ardabili
2
Ph.D. Student of Climatology, Department of Physical Geography, Faculty of Social Sciences, University of Mohaghegh Ardabili, Ardabil, Iran
Abstract
In this study, first, daily discharge data of Lai, Nir, and Yamchi stations were collected from the Regional Water Company of Ardabil Province during the years 2014 to 2021. Then, effective environmental parameters including land slope, distance from the waterway, land use, and soil moisture were plotted in ArcGIS and Google Earth Engine. Also, spectral indices NDWI, AWEI, WRI, and LSWI were extracted from Landsat 8 images. After standardizing the variables, a regression random forest model was trained with 100 decision trees. 70% of the data was used for training and 30% for testing the model. The performance of the model was evaluated with statistical indices R² and MSE equal to 0.9353 and 0.000210, respectively. Finally, flood-prone areas were identified. For validation purposes, the April 2017 flood event was analyzed as a case study. The results showed that the northern part of the basin has the highest potential for flooding due to its high elevation, steep slope, 35% soil moisture, and high AWEI and WRI values. In contrast, the central and southern areas showed a lower probability of flooding due to their gentler slope and lower soil moisture. The random forest model confirmed the accuracy of this model, and its performance was assessed as acceptable with an ROC curve and an AUC value of 0.616. Radar data analysis also showed that the signal reflectance in the northern areas changed significantly before and after the flood, indicating the concentration of water resources in this part of the basin.
Keywords
Main Subjects
)