Seasonal and upstream rainfall controls on monsoon flood onset in Bangladesh’s Haor wetlands

This study quantifies the lead–lag coupling between monsoon rainfall and floodplain inundation in northeast Bangladesh (Haor region) using JRC Global Surface Water (30 m, 1984-present) and CHIRPS rainfall data. Monthly flood area was computed using two metrics: RAW (all detected water) and SEASONAL (permanent water removed using the JRC Occurrence mask of at least 80%). Annual timing metrics for flood onset and peak were derived from monthly area series using relative (5–30% of annual maximum) and absolute (10–50 square kilometers) thresholds. Pearson correlations between rainfall and inundation area across lags minus three to plus three months show peak correlations at minus two to minus three months (r approximately 0.39–0.45) for both metrics and for local and upstream rainfall, while zero-lag correlations are near zero after deseasonalization. Comparing 1984–2000 to 2001–2024, flood onset has advanced by approximately 1.6–2.2 months, while peak timing has been delayed by about 0.9 months, indicating a lengthened flood season. Findings are robust across thresholds; fixed-area thresholds show consistent behavior but exclude a larger number of early years. Multi-year maximum inundation maps are provided for major flood years (1988, 2004, 2017, and 2020) and fully reproducible Google Earth Engine code. Overall, the findings demonstrate a predictable rainfall-to-flood response that, together with the lengthened flood season, provides a practical 2-to-3-month window for anticipatory early warning, preparedness, and seasonal decision-making in flood-exposed livelihoods.