Advances in Agriculture and Agricultural Sciences

ISSN 2756-326X

Advances in Agriculture and Agricultural Sciences | Vol. 12, No. 10, October 2026 | pp. 533–540

DOI: 10.46882/2026.AAAS.121003

Research Article

Title: Multi-Spectral Satellite Data Fusion and Deep Learning Predict Regional Maize Yield Gaps Six Weeks Before Harvest Across Smallholder Landscapes

Names of Authors: Tesfaye, B. G.¹, van der Berg, H. M.², and Zhang, Q. L.³

Authors’ Affiliations: ¹Department of Plant Sciences, Haramaya University, Dire Dawa, Ethiopia. ²Laboratory of Geo-Information Science and Remote Sensing, Wageningen University & Research, Wageningen, Netherlands. ³Institute of Remote Sensing and Digital Earth, Chinese Academy of Sciences, Beijing, China.

Abstract: Fragmented field sizes and persistent cloud cover hinder accurate regional crop yield forecasting across rainfed smallholder farming systems. Sentinel-2 optical imagery was fused with Sentinel-1 synthetic aperture radar backscatter and gridded meteorological data inside a spatiotemporal convolutional long short-term memory network. The model was trained and validated against 1,450 ground-truthed smallholder maize plots spanning three administrative zones. Multi-sensor fusion overcame cloud-cover gaps during peak vegetative growth, predicting district-level grain yields six weeks prior to physiological maturity with a root-mean-square error of 0.34 t ha⁻¹ (R² = 0.89). Spatial residual mapping pinpointed soil nitrogen depletion and delayed planting as primary drivers of localized yield gaps. Radar-optical deep learning pipelines strengthen early food security warning systems and targeted agronomic extension.

Keywords: Crop yield forecasting; Sentinel data fusion; Deep learning; Smallholder agriculture; Synthetic aperture radar; Yield gap analysis.

Manuscript Timeline: Received August 23, 2026; Revised September 28, 2026; Accepted October 06, 2026; Published October 13, 2026.

Citation: Tesfaye, B. G., van der Berg, H. M., and Zhang, Q. L. (2026). Multi-Spectral Satellite Data Fusion and Deep Learning Predict Regional Maize Yield Gaps Six Weeks Before Harvest Across Smallholder Landscapes. Advances in Agriculture and Agricultural Sciences, 12(10), pp. 533–540.