Sand mining's hidden footprint: Multi-signal landscape change around extraction sites on the Ghaghat, Jamuna, Jamuneshwari, Karatoya and Talma rivers of Bangladesh

River sand mining is reshaping riverine landscapes in Bangladesh, but its signature is difficult to separate from channel migration, seasonal water-level change, sediment movement and wider land-use pressure. This study presents a distance-based, multi-indicator geospatial assessment of change around five sand-extraction sites on the Ghaghat, Jamuna, Jamuneshwari, Karatoya and Talma rivers in northern Bangladesh. Change was measured in concentric rings of 0–0.5, 0.5–1, 1–2 and 2–5 km around each extraction point, verified by UAV survey, so that the outer ring serves as an internal reference for the near-site zones. The 2 km reporting distance was derived from a household survey of 1400 respondents (Cochran sample size; 1344 valid answers), in which the median perceived reach of mining effects was 2 km and 92.3% (Wilson 95% CI 90.7–93.6%) placed it at or within 2 km, with no significant difference between rivers (Kruskal–Wallis H = 8.63, p = 0.071; χ 2 = 10.81, p = 0.545). Residents also dated the onset of extraction to 2022–2025, so the comparison spans a pre-extraction baseline and one to four years of extraction. Sentinel-2 Level-2 A median composites for January–April 2016 and 2026 (87 cloud-screened scenes) were processed in Google Earth Engine. A Random Forest classifier mapped five land-use/land-cover classes for each site and year with an independent accuracy assessment, and 2016–2026 transition matrices traced where lost agricultural land went. NDVI was compared on stable land only, and every NDVI-decline pixel was attributed, using MNDWI water frequency and the Bare Soil Index, to inundation, bare-soil increase on dry land or another dry-land cause. Land–water transition was mapped from persistent MNDWI water, screened for river-stage sensitivity against the per-scene water record and CHIRPS rainfall, and converted into bankline migration distances and rates. The results show distance-dependent change that differs by river and by duration of extraction. On the Jamuna the channel migrated onto the extraction site, converting 98% of the 0–0.5 km ring to water and producing stage-robust land-to-water change of 4.0 km 2 and mean bank retreat of 60 m per year within 2 km, while the reach 2–5 km away accreted and the study area as a whole gained 2.6 km 2 of land. On the Ghaghat, sand and bare soil increased most in the 0.5–1 km ring (+6.4 percentage points) and 62% of the NDVI decline within 0.5 km occurred on dry land where bare surfaces expanded. On the Ghaghat and Jamuneshwari about two-thirds of the apparent land–water change was water-level dependent rather than erosional, and on the Karatoya and Talma, where extraction was less than a year old, agricultural land lost within 2 km went to built-up land and tree cover in the same proportions at every distance. Wetter dry-season conditions in 2026 raised NDVI on stable land at four of the five sites, so mean greenness alone would have masked the localised bare-surface expansion near the extraction points. The study shows a ring-based, multi-indicator design with stage screening and pixel-level NDVI attribution can separate near-site change from background river dynamics and identify which rivers, and after how long, show a mining-consistent footprint.