ارزيابي دوره احياء پس از خشكسالي در اقليم و پوشش‌هاي مختلف زمين

eva‎luation of drought recovery period in different land covers and climate

مهندسي نقشه برداري

Effective planning and management to address the potential impacts of drought require a precise assessment and analysis of this complex natural phenomenon. Drought significantly affects vegetation growth. Understanding changes in vegetation during drought periods and its recovery after drought is crucial for agricultural management and adaptation to climate change. Researchers have developed various drought indices based on ground weather data as well as diverse remote sensing data. Utilizing machine learning models to integrate remote sensing data for reconstructing a composite drought index can enable more accurate monitoring of drought. In this study, three machine learning methods—Random Forest (RF) based on decision trees, XGBOOST based on gradient boosting, and Deep Learning CNN—were examined to model the Standardized Precipitation Evapotranspiration Index (SPEI) using various remote sensing data products. The SPEI index is utilized to identify drought events, and subsequently, the vegetation recovery period is estimated using the gross primary productivity (GPP) variable. Vegetation and surface temperature data were obtained using the MODIS satellite, precipitation data from the GPM satellite, and soil moisture and evapotranspiration data from GLDAS to compute remote sensing drought indices and model the SPEI for Iran over the period from 2003 to 2022. The best model was employed to create spatial distribution maps of SPEI at the optimal temporal scale. The results indicate that the XGBOOST model performs better than the other algorithms (CNN and RF), achieving a correlation coefficient (R) of 0.75 and a root mean square error (RMSE) of 0.59 when using seven remote sensing drought indices to model the monthly SPEI. When modeling the monthly SPEI using four drought indices in this algorithm, R decreases to 0.69 and RMSE increases to 0.71. The spatial distribution map of the monthly SPEI for Iran, created using the XGBOOST method, provides comprehensive information about drought. In the 20-year monitoring of drought in Iran, the years 2008 and 2021 were identified as the most severe drought years. In the second phase of this study, the vegetation response to drought and the duration of recovery after the end of drought were examined using the SPEI index model and GPP data. The results indicate that the vegetation recovery period after the drought in 2008 ranges from 30 days to 120 days. This recovery duration is influenced by the diverse climates of Iran and various land cover types. In the northern regions of the country, given the humid climate, ample rainfall, and moderate temperatures, even during drought events, there is no significant decline in GPP, resulting in a negligible recovery period. Conversely, in the arid regions of Iran in the Zagros Mountains, drought has been relatively severe, and due to the climate of this area, the recovery period was assessed to be close to 120 days. This extended recovery period is also observed for pastures and agricultural lands.

5

75560