Forecasting groundwater irrigation risk under salinity-sodicity stress using hydrochemical-climate data fusion: A tabular, spatial, and ordinal learning framework for environmental management.

Groundwater salinity and sodicity threaten irrigation sustainability in semi-arid agricultural regions, yet most assessments remain static. This study develops a hydrochemical-climate framework for next-year irrigation-risk screening in Telangana, India, using post-monsoon groundwater records from 2018-2020 enriched with ERA5-Land precipitation and soil-moisture context. The chronological design includes 361 matched pairs for model development between 2018 and 2019 and 356 matched pairs for target-year assessment between 2019 and 2020. In the reported experimental configuration, 71 target-year pairs (20%) were used during adaptation/calibration and the remaining 285 were used for final evaluation; consequently, the reported performance is target-adapted rather than a pure zero-label holdout. Across eight tabular, spatial, ordinal and ensemble candidates, SpatialGNN provided the best retrospective decision-oriented compromise on this 285-sample subset, achieving 63.2% accuracy, 56.1% macro-F1, 63.5% weighted-F1 and 50.0% recall for T3-Restricted wells. CORAL achieved slightly higher T3 recall, whereas several tree and ensemble models achieved slightly higher accuracy, illustrating the trade-off between aggregate fit and high-risk detection. SHAP attribution was dominated by EC and EC-derived salinity-boundary features, with ionic indicators and monsoon precipitation providing additional context. SAR +10%/+20% sensitivity scenarios produced negligible T3-probability changes and no hard-class transitions. The framework is therefore best interpreted as a transparent risk-screening and monitoring-prioritisation workflow under short monitoring records and target-year adaptation.

Authors

Institutions

Publication Details

Journal
PubMed
Published
2026-11-12
DOI
https://doi.org/10.1016/j.jenvman.2026.131311
Primary Topic
Engineering and Agricultural Innovations
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Forecasting groundwater irrigation risk under salinity-sodicity stress using hydrochemical-climate data fusion: A tabular, spatial, and ordinal learning framework for environmental management.

Mohsen Asghari Ilani, Mike Banad, Saba Moftakhar Tehran, Ashkan Kavei et al.
PubMed
Engineering and Agricultural Innovations
article

Forecasting groundwater irrigation risk under salinity-sodicity stress using hydrochemical-climate data fusion: A tabular, spatial, and ordinal learning framework for environmental management.

Mohsen Asghari Ilani, Mike Banad, Saba Moftakhar Tehran, Ashkan Kavei, Eskandari Hamidreza
article en

Abstract

Groundwater salinity and sodicity threaten irrigation sustainability in semi-arid agricultural regions, yet most assessments remain static. This study develops a hydrochemical-climate framework for next-year irrigation-risk screening in Telangana, India, using post-monsoon groundwater records from 2018-2020 enriched with ERA5-Land precipitation and soil-moisture context. The chronological design includes 361 matched pairs for model development between 2018 and 2019 and 356 matched pairs for target-year assessment between 2019 and 2020. In the reported experimental configuration, 71 target-year pairs (20%) were used during adaptation/calibration and the remaining 285 were used for final evaluation; consequently, the reported performance is target-adapted rather than a pure zero-label holdout. Across eight tabular, spatial, ordinal and ensemble candidates, SpatialGNN provided the best retrospective decision-oriented compromise on this 285-sample subset, achieving 63.2% accuracy, 56.1% macro-F1, 63.5% weighted-F1 and 50.0% recall for T3-Restricted wells. CORAL achieved slightly higher T3 recall, whereas several tree and ensemble models achieved slightly higher accuracy, illustrating the trade-off between aggregate fit and high-risk detection. SHAP attribution was dominated by EC and EC-derived salinity-boundary features, with ionic indicators and monsoon precipitation providing additional context. SAR +10%/+20% sensitivity scenarios produced negligible T3-probability changes and no hard-class transitions. The framework is therefore best interpreted as a transparent risk-screening and monitoring-prioritisation workflow under short monitoring records and target-year adaptation.

PubMedVol. 420
University of Kashan (IR), Islamic Azad University, Science and Research Branch (IR), Swansea University (GB), University of Oklahoma (US)
Engineering and Agricultural Innovations
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon