Crop Yield Prediction Using Satellite Imagery, Weather Data, and Machine Learning: A Multi-Modal Framework for Precision Agriculture

Shravanakumari H J, Shinty P K

Abstract: Accurate, early prediction of crop yield is one of the more stubborn problems in agricultural planning, food security policy, and commodity markets alike. Farmers, insurers, and governments all want the same thing at different scales: a reliable read on how much a field or a region is likely to produce, well before harvest. This paper presents a multi-modal machine learning framework that fuses multispectral satellite imagery with ground-level weather observations to forecast crop yield at the field level. Vegetation indices such as NDVI and EVI are extracted from Sentinel-2 and Landsat-8 imagery across the growing season and combined with temperature, rainfall, and soil-moisture time series. A hybrid architecture- a convolutional feature extractor for the imagery paired with a recurrent network for the temporal weather signal- is used to learn joint representations that a downstream regression head converts into yield estimates. On a multi-season wheat and maize dataset spanning three agro-climatic zones, the proposed model reaches an R? of 0.87 and reduces RMSE by roughly 18% compared to weather-only and imagery-only baselines. The results suggest that the two data sources are genuinely complementary rather than redundant, and that fusing them is worth the added engineering effort. We also discuss where the approach struggles - cloud cover, smallholder plot sizes, and limited ground-truth labels chief among them- and outline directions for making the system more field-ready.

Keywords: crop yield prediction, remote sensing, satellite imagery, NDVI, weather data fusion, precision agriculture, convolutional neural network, LSTM, machine learning

How to Cite?: Shravanakumari H J, Shinty P K, "Crop Yield Prediction Using Satellite Imagery, Weather Data, and Machine Learning: A Multi-Modal Framework for Precision Agriculture", Volume 15 Issue 8, August 2026, International Journal of Science and Research (IJSR), Pages: 1630-1632, https://www.ijsr.net/getabstract.php?paperid=SR26820130954, DOI: https://dx.doi.org/10.21275/SR26820130954

Download Citation: APA | MLA | BibTeX | EndNote | RefMan