Urban methane under watch: Sentinel‑5P/TROPOMI and LSTM reveal hotspot dynamics over Tehran (2019–2025)

Rapidly rising atmospheric methane threatens near-term climate goals, yet the behaviour of urban methane over Middle Eastern megacities remains poorly constrained. Using Sentinel-5P/TROPOMI XCH₄ retrievals processed in Google Earth Engine, we assemble a daily record of column methane over Tehran for 2019–2025 and derive seasonal and annual means at the city scale. The annual mean XCH₄ increases from about 1906 ppb in 2019 to nearly 1978 ppb in 2025, corresponding to a trend of roughly 10 ppb yr⁻¹ and indicating a persistent strengthening of the urban methane burden. Gridded annual maps and a district-level aggregation show that the highest mean XCH₄ values occur persistently over the northern and north-eastern districts of Tehran, while elevated enhancements also appear over southern industrial and landfill-influenced areas; these hotspot polygons are quantified in terms of area and mean XCH₄ for each year.​ To link column enhancements with discrete emitters, we compile a multi-year inventory of satellite-detected methane plumes around Tehran, including source locations and annual emission rates (Q, ton yr⁻¹), and examine how plume occurrence and intensity co-vary with the hotspot fields. Building on the daily XCH₄ time series, we train a long short-term memory (LSTM) network to forecast next-day column methane over the city; on the independent test period the model achieves a correlation of about 0.7 with observations, a mean absolute error near 20 ppb, a root-mean-square error around 26 ppb, and a small positive mean bias of ~ 3 ppb. By jointly analysing daily, seasonal and annual XCH₄, district-scale hotspots and plume emissions, and by demonstrating a deep-learning forecast of urban methane, this study provides a comprehensive satellite-based picture of methane dynamics over Tehran and a readily transferable framework for monitoring methane hotspots in rapidly growing cities across the Middle East and developing regions.