A Comprehensive Remote Sensing Approach for Assessing Forest Degradation and Environmental Stress in Amedi District, Kurdistan Region of Iraq

Forest ecosystems in Amedi district face stress from conflict, fires, climate variability, and unsustainable land use. This study uses remote sensing and geospatial techniques to assess forest degradation from 2000 to 2024. Modified Soil-Adjusted Vegetation Index (MSAVI2), Normalized Difference Vegetation Index (NDVI), Land Surface Temperature (LST), and Normalized Difference Moisture Index (NDMI) data are used to measure changes in forest land (FL) and non-forested land (NFL), NDMI, and LST. The results obtained from NDVI data reveal inadequate regeneration rates to counterbalance the total forest loss, while non-forest regions have significantly increased from 5.8% (160.8 km²) to 19.7% (546.6 km²). NDVI-based FL dropped from 90.3% (2505.3 km²) to 73.3% (2032.2 km²), and MSAVI2-based FL declined from 86.3% (2393.3 km²) to 74.3% (2062.3 km²). Conversely, NDVI also indicated an increase in degraded land cover, especially during 2013–2015 and 2020–2024. The LST class (<35°C) declined from 849 km² to 520 km², while the High (45–50°C) and Extreme (>50°C) LST zones exhibited a marked expansion. Mann-Kendall tests revealed significant temporal changes in NDVI, MSAVI2, NDMI, and LST (Z > ±1.96). Sen’s slope quantified trends; Pearson's correlation confirmed strong vegetation/climate stress links (r > 0.7, p < 0.05). The Mann-Kendall trend test confirmed a significant upward LST trend (τ = 0.312, p = 0.034), with a Sen’s slope of 14.662 km² per unit time. Trend detection using the non-parametric Mann-Kendall test validated significant changes across variables without assuming normality. The ARIMA (1,0,0) model forecasts a continued decline through the projection horizon. Forecast intervals widen at a 95% confidence level due to uncertainty in a near-unit root process. The forecast error measured by MAPE is about 5.0%, showing moderate reliability. ARIMA models further revealed robust forecasting capability for environmental indicators, validating model assumptions with stationary residuals and minimal error. For NDVI-derived forest area, a negative trend was identified (Sen’s slope = –22 km²/unit time, p < 0.0001), confirming ongoing deforestation. Pearson correlation analysis showed strong associations between vegetation indices and temperature/moisture variables (r > 0.7, p < 0.05). Although robust, these relationships may be influenced by confounding factors such as climate change, forest degradation, forest fires, armed conflict, and broader land use and land cover changes. Findings emphasize conservation strategies to combat deforestation, warming trends, and safeguard ecosystem resilience. Continued decline through the projection horizon. Forecast intervals widen at a 95% confidence level due to uncertainty in a near-unit root process. The forecast error measured by MAPE is