Climate variability and recurrent droughts pose increasing irrigation challenges for smallholder maize farmers in southern Africa. This study developed a scalable Earth observation and artificial intelligence (EO–AI) framework combining satellite data, machine learning, and crop water modeling to estimate daily maize actual crop evapotranspiration (ETc) in South Africa’s Vhembe District. Five machine learning models were rigorously validated against benchmark ET c derived from the FAO-56 Penman–Monteith reference evapotranspiration (ET 0 ) method multiplied by locally calibrated crop coefficients (K c ). Random Forest and k-Nearest Neighbors models demonstrated superior performance, with R 2 consistently exceeding 0.99, root mean square error (RMSE) below 0.06 mm/day, and normalized RMSE (NRMSE) less than 2%, outperforming support vector machine, MARS, and XGBoost models. The EO–AI framework effectively captured fine-scale spatial and temporal ET c variability, with daily actual maize ET c at 6.5 mm/day during peak crop sensitivity periods. An operational irrigation decision-support prototype translated these predictions into targeted field-level water-deficit alerts for farmers. This work highlights the value of EO–AI frameworks for delivering high-resolution, daily ET c mapping in fragmented, cloud-prone landscapes, enabling more precise and resilient irrigation strategies for smallholder systems.
Climate variability threatens smallholder maize production in semi-arid Southern Africa, necessitating accurate irrigation management. We developed an Earth Observation–machine learning framework integrating Sentinel-1 SAR, TU Wien retrievals, and meteorological data to generate daily 10 m resolution root-zone soil moisture estimates (0–100 cm) for South Africa’s Vhembe District (2017–2022). Five algorithms—Random Forest (RF), Extreme Gradient Boosting (XGBoost), Support Vector Machine (SVM), k-Nearest Neighbors (KNN), and Multivariate Adaptive Regression Splines (MARS)—were calibrated using ~50,000 observations from two monitoring stations across six depths and five growing seasons. RF and XGBoost achieved highest accuracy (R2 = 0.96–0.97, RMSE < 0.025 cm3/cm3), detecting critical irrigation thresholds (management allowable depletion = 0.23 cm3/cm3, field capacity = 0.35 cm3/cm3) with operational precision (nRMSE < 0.05). Depth-stratified validation revealed strong SAR surface correlations (r = 0.84–0.85 at 10 cm) declining systematically with depth (r < 0.2 below 40 cm), confirming ML models integrate satellite observations at shallow layers with meteorological gap-filling at depth. District mapping showed 79–94% of maize areas required irrigation during dry years (2017–2019, 2021–2022) versus 32% in wet 2020–2021. The framework provides a transferable pathway for precision irrigation in smallholder systems, pending vegetation-corrected retrievals and expanded validation.
South African irrigation schemes face critical challenges of water scarcity, infrastructure deterioration, and limited monitoring capacity, threatening agricultural productivity and food security. This scoping review systematically analyses 59 peer-reviewed publications (2000–2025) on sensor-based and acoustic irrigation monitoring technologies in South Africa, using transformer-based natural language processing (Sentence-BERT embeddings), unsupervised Machine Learning (UMAP dimensionality reduction, HDBSCAN clustering), and geospatial mapping applied to literature retrieved from Web of Science and Scopus. Results show that water quality monitoring (42.4% of studies) and remote sensing (25.4%) dominate the national research landscape, while soil moisture sensing and modelling remain comparatively limited. Notably, no peer-reviewed studies applying acoustic monitoring technologies to irrigation were identified, representing a critical gap despite proven international applications for leak detection (95–98% accuracy), widespread infrastructure aging (over 50% of schemes exceeding 30 years), and reported water losses of 30–60% in poorly managed systems. Reported experimental water savings range from 15% to 30%, yet applications remain largely confined to pilot-scale implementations concentrated within a limited number of Water Management Areas. Persistent adoption barriers include infrastructure unreliability, financial inaccessibility, limited digital literacy, and weak institutional coordination. The review recommends: (i) expanding research coverage across underrepresented regions and Water Management Areas; (ii) strengthening extension support and technical training to enable broader adoption; and (iii) integrating low-cost sensor networks with predictive, data-driven irrigation advisory systems. These priorities aim to support scalable, context-sensitive irrigation modernisation under increasing water scarcity pressures.
Soil contamination in South African agriculture poses escalating threats to food security and ecosystem integrity, yet the geospatial and machine learning evidence base addressing this problem has never been systematically synthesised. This scoping review, conducted within the PRISMA-ScR framework, applied SVM-assisted screening to 2000 retrieved records, yielding a final corpus of 228 eligible studies published from 2003 to 2025. To characterise temporal, thematic, and geographic patterns in the corpus, we applied machine learning-assisted topic modelling (LDA, k = 7), logistic growth modelling, keyword co-occurrence network analysis, and technology–contaminant evidence gap matrices. Remote sensing was the dominant methodology throughout the review period (n = 142; 62.3% of studies), with machine learning rising to the highest adoption rank from approximately 2020 onwards. Logistic modelling estimated a carrying capacity of K = 292.3 (95% CI: 269–324) studies and an inflexion year of 2020.2 (95% CI: 2019.4–2021.1), projecting 90% saturation by 2028. Research effort was highly concentrated in KwaZulu-Natal and the Eastern Cape, while Pesticides/Herbicides and acid mine drainage each comprised only three corpus studies. Deep learning registered zero entries across all cells of both the technology–contaminant and technology–province evidence matrices. Targeted investment in field validation, hyperspectral and deep learning deployment for underrepresented contaminants, and interpretable modelling for regulatory defensibility are identified as priority actions for the next research cycle.