A global study analyzed 40 river deltas across five continents to assess land subsidence and its drivers. The research identified that 42,000 km² of delta area and 10.2 million people are currently located below sea level.
Methodology
The study utilized Sentinel-1A/B C-band satellite Synthetic Aperture Radar (SAR) data from September 2016 to May 2023 to generate high-resolution maps of surface deformation. The multi-temporal wavelet-based InSAR (WabInSAR) algorithm was employed, processing approximately 55,000 interferometric pairs. Vertical Land Movement (VLM) rates were estimated and validated against 122 independent Global Navigation Satellite Systems (GNSS) stations, showing an RMSE of 1.2-1.4 mm/yr. The observed deformation rates were corrected for Glacial Isostatic Adjustment (GIA) using a probabilistic model, revealing that GIA contributions vary significantly, obscuring contemporary subsidence in some high-latitude regions.
Anthropogenic Drivers of Subsidence
Three major anthropogenic pressures were analyzed for their relationship to residual land sinking (after GIA correction):
- Groundwater Storage Change: Derived from GRACE/GRACE-FO satellite observations (2002-present), accounting for annual and semi-annual variations using a stochastic-seasonal model. Negative trends, largely due to groundwater extraction, showed a modest cubic correlation (R=0.6) with subsidence rates.
- Sediment Flux Alteration: Quantified as the percentage change between pristine and disturbed sediment fluxes. 62% of the deltas exhibited reduced sediment supply due to human activities, but this showed a poor correlation (R<0.4) with subsidence rates.
- Urban Expansion: Assessed using a global urban land fraction dataset (2000-2020). All deltas showed urban growth, with some exhibiting significant increases. A moderate nonlinear inverse correlation (R=0.38-0.51) was found between urban land conversion and land sinking.
Identifying Key Drivers
A Random Forest (RF) machine learning model was implemented to account for the complex, nonlinear interactions between these drivers and subsidence rates. The RF model significantly outperformed a multilinear regression, achieving an R² of 0.81. Feature importance analysis identified groundwater storage change as the primary driver (average importance ~0.64), followed by urban expansion (~0.23), and sediment flux alteration (~0.13). Local Interpretable Model-agnostic Explanations (LIME) further provided delta-specific insights into these contributions.
Sea Level Rise Analysis
The study also assessed historical (20th century) and present-day (early 21st century) Sea Level Rise (SLR) rates. Historical relative SLR, based on tide gauge data, showed a median trend of 2.9 mm/yr. Present-day absolute SLR, derived from multi-mission satellite altimetry, indicated that 45% of the deltas experience regional SLR rates exceeding the global average of approximately 4 mm/yr.