Projecting Soil Salinity in Vietnam's Red River Delta to 2035

6 hours ago

2 min read

Projecting Soil Salinity in Vietnam's Red River Delta to 2035. The authors, including Dung Duc Tran (HAPRI Senior Research Associate), pair hybrid machine learning with Landsat OLI to project soil salinity to 2035. Coastal and estuarine zones face rising salinity as cropland shrinks and aquaculture expands. Huu Duy Nguyen, Dung Duc Tran (HAPRI Senior Research Associate) and colleagues. Illustrative photograph: a Vietnamese farmer crouches on a paddy bund, dipping a handheld meter into water beside patchy, yellowing rice, with a newly dug aquaculture pond, a moored boat and a flat estuary horizon behind.

Projecting Soil Salinity in Vietnam's Red River Delta to 2035

www.hapri.org

Dung Duc Tran — Senior Research Associate

A study co-authored by Dung Duc Tran (HAPRI Senior Research Associate) with colleagues uses hybrid machine learning to project intensifying soil salinity in the coastal and estuarine areas of Vietnam's Red River Delta to 2035, alongside a projected decline in cropland and an expansion of aquaculture. The authors use the delta as a testbed and say the study offers transferable insights for deltas worldwide.

Context

Soil salinization is a major driver of global land degradation, yet monitoring it remains difficult due to its complex spatiotemporal dynamics and climate change. In the Red River Delta, severe salinity undermines agriculture and food security, which the authors say makes the region an ideal testbed for deltaic studies.

Unlike prior research focusing on single time points, this study integrates historical salinity data with climate projections to model long-term dynamics.

*Less cropland, more aquaculture by 2035*

Methods

Using field data, the authors applied hybrid machine learning approaches that pair XGBoost (XGB) with optimization algorithms: Sandpiper Hunting Optimization, Northern Goshawk Optimization, Electromagnetic Search Optimization, Monkey Gorilla Optimization and Grey Wolf Optimization. According to the authors, this addressed spatial heterogeneity and filled missing data. The study had three objectives: simulating soil salinity for 2005–2025, projecting its dynamics to 2035 under climate scenarios, and assessing the impacts on cropland.

Key findings

Of the hybrid models tested, XGB-NGO performed best (R²=0.886), followed by XGB-SHO/ESOA (0.881) and XGB-MGO/GWO (0.876).
The projections indicate intensifying salinity (EC>4 mS/cm) in coastal and estuarine areas due to sea-level rise.
Cropland declines from 2,796.3 km² (86.2%) in 2005 to 2,424.0 km² (74.7%) in 2035 (−13.3%), while aquaculture expands from 68.5 km² (2.1%) to 179.0 km² (5.5%) (+161%).
Salinity-affected cropland declines under RCP 4.5 (18.02 km²) and RCP 8.5 (19.0 km²) by 2035 via land conversion.

Keywords:

Soil Salinity
Red River Delta
Vietnam
Machine Learning
Climate Change
Agriculture

Link:

https://doi.org/10.1016/j.eti.2025.104649

Citation:

Nguyen, H. D., Dang, D. K., Dinh, T. T. H., Ha, M. C., Truong, Q. H., Tran, D. D., Park, E., Bui, Q., & Petrisor, A. (2025). Projecting spatiotemporal dynamics of soil salinity using hybrid machine learning and Landsat OLI: The Red River Delta in Vietnam as a testbed for vulnerable deltas globally. Environmental Technology & Innovation, 40, 104649.