Analysis of the effects of heatwave events on continental surfaces from multi-spectral satellite series

Over the past few decades, Tunisia has faced heatwaves, droughts, and severe meteorological fluctuations, significantly impacting its agricultural sector and natural landscapes. This study focuses on the prolonged impact of heatwaves, with an emphasis on forests and pasture areas. To do so, multispectral remote sensing and deep learning techniques were employed to predict time series of NDVI (Normalized Difference Vegetation Index) based on monthly maximum temperatures and monthly precipitation. The identification of heatwave periods relies on the use of predefined thresholds, the 95th and 99th percentiles, calculated through quantile analysis using daily maximum temperature data from 2000 to 2022, collected from 15 stations across the country. On average, it was found that there were 16 heatwave days each year; however, their occurrence varies depending on both time and geographical location. The study then assesses the impact of identified heatwaves on various land cover classes in distinct climatic contexts using MODIS sensor-derived Normalized Difference Vegetation Index to calculate calibrated NDVI and Vegetation Anomaly Index (VAI). This analysis highlights a significant correlation between heatwave events, NDVI, and monthly precipitation. Finally, a Long Short-Term Memory (LSTM) neural network model is developed to predict NDVI time series. The results underscore the importance of the unique characteristics of each land cover type, such as dense forests, open forests, pastures in subhumid bioclimate, and pastures in arid bioclimate, in their ability to respond to extreme climate variations.