Air pollution in Gaza during the post-october 7 era: a satellite and machine learning assessment

Armed conflicts pose severe environmental challenges, particularly in densely populated and infrastructure-limited regions. The Gaza Strip exemplifies such a context, where the intersection of warfare, urban density, and environmental vulnerability demands urgent scientific inquiry. This study aims to assess the environmental impact of the 2023–2024 war on air quality in the Gaza Strip by examining temporal and spatial changes in key atmospheric pollutants. We use daily observations of five pollutants, nitrogen dioxide (NO₂), sulfur dioxide (SO₂), carbon monoxide (CO), methane (CH₄), and the ultraviolet aerosol index (UVAI), obtained from the Sentinel-5P TROPOspheric monitoring instrument (TROPOMI) satellite and combine these with meteorological data (temperature, humidity, wind speed, and precipitation) to explore their behavior before and during the conflict. Our methodology integrates time-series analysis with statistical and machine learning models, including SARIMAX, Holt-Winters, Random Forest, and XGBoost, to forecast pollutant concentrations based on pre-war conditions and identify deviations post-October 2023. The findings reveal distinct responses to pollutants during the war. UVAI and CO showed sharp and sustained increases linked to widespread combustion and infrastructure damage, while CH₄ concentrations exhibited a steady rise associated with the collapse of waste management. SO₂ displayed episodic spikes, likely tied to fuel depot destruction and generator use, whereas NO₂ trends showed temporary suppression due to mobility restrictions and reduced industrial activity. Our findings demonstrate that traditional forecasting models may require adaptation to conflict-specific conditions, given altered emission sources and rapid pollutant dispersal in a small geographic area like Gaza. Policy implications include the urgent need for conflict-sensitive environmental monitoring systems, the integration of satellite data into humanitarian planning, and the development of adaptive forecasting models that incorporate war-related variables, such as infrastructure damage and displacement patterns.