Integrating Sentinel-1 SAR and Sentinel-2 data to estimate soil moisture for irrigation scheduling in maize fields, Bronkhorstspruit, South Africa

A timely irrigation schedule is critical for optimising maize production, especially amid increasing droughts and climate change challenges. Maize, a staple crop in many countries, requires careful water management to maintain yields and promote sustainable agriculture. This study focused on Bronkhorstspruit, South Africa, a region with numerous commercial irrigated farms cultivating various crops year-round. It aimed to contribute to an efficient irrigation scheduling approach for maize by accurately estimating soil moisture by integrating Sentinel-1 synthetic aperture radar (SAR) and Sentinel-2 optical data. The vegetation and soil moisture indices (e.g., MSI and NDVI), which represent crop health and moisture conditions, were analysed using Sentinel-2 images. Concurrently, polarizations (VV and VH) sensitive to soil moisture changes were obtained using Sentinel-1 SAR data. The study utilised an image fusion technique at the pixel level to integrate the SAR and optical data. A fully connected deep neural network (DNN) model was developed, using 80% data for training and 20% for validation, to capture complex nonlinear relationships and optimise the prediction of soil moisture. The results demonstrated strong correlations between soil moisture, leaf area index (LAI), and chlorophyll content, particularly during the V8 (eighth leaf) and R2 (blister kernel) stages. The integration of the datasets led to improved soil moisture prediction accuracy compared to using optical data alone, with relative root mean square error (rRMSE) values decreasing from 14.29% to 13.04% at the V8 stage and from 6.59% to 6.18% at the R2 stage. Polarization configurations influenced accuracy, with VH generally performing better than VV. Lower predictive performance was observed at both the early (V3) and late (R5) growth stages, due to shifts in canopy structure and plant physiology. Nonetheless, this integrated remote sensing approach offers a valuable tool for precision irrigation scheduling in maize fields, helping to optimise water use while sustaining crop yields. These findings support the adoption of smart agriculture technologies in the Global South and highlight the importance of satellite-based soil moisture estimation in advancing adaptive irrigation strategies that address agricultural challenges.