Multi-modal Data Fusion and Deep Ensemble Learning for Accurate Crop Yield Prediction
Akshay Dagadu Yewle, Laman Mirzayeva, Oktay Karakuş · 2025
Read Multi-modal Data Fusion and Deep Ensemble Learning for Accurate Crop Yield Prediction by Akshay Dagadu Yewle, Laman Mirzayeva, Oktay Karakuş (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2502-06062.txt.
Introduction
Forecasting harvests from orbit gets a new model in a 2025 paper by Akshay Dagadu Yewle, Laman Mirzayeva, and Oktay Karakuş. Their system, RicEns-Net, blends radar and optical satellite views with weather data to predict crop yields more precisely.
Field data came from a 2023 open science challenge. From over one hundred candidate inputs, the team kept fifteen features drawn from five data types, which kept the model lean and sharp. The design is a deep ensemble: several learning methods vote together, each covering the others' weak spots. On tests, the model reached a mean error of about 341 kilograms per hectare. The authors present it as a practical tool for farm planning.
Catalog record
| CREATOR | Akshay Dagadu Yewle, Laman Mirzayeva, Oktay Karakuş |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 71 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2502-06062
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Multi-modal Data Fusion and Deep Ensemble Learning for Accurate Crop Yield Prediction Akshay Dagadu Yewle, Laman Mirzayeva, Oktay Karakuş Abstract This study introduces RicEns-Net , a novel Deep Ensemble model designed to predict crop yields by integrating diverse data sources through multimodal data fusion techniques. The research focuses specifically on the use of synthetic aperture radar (SAR), optical remote sensing data from Sentinel 1, 2, and 3 satellites, and meteorological measurements such as surface temperature and rainfall. The initial field data for the study were acquired through Ernst & Young’s (EY) Open Science Challenge 2023. The primary objective is to enhance the precision of crop yield prediction by developing a machine-learning framework capable of handling complex environmental data. A comprehensive data engineering process was employed to select the most informative features from over 100 potential predictors, reducing the set to 15 features from 5 distinct modalities. This step mitigates the “curse of dimensionality" and enhances model performance. The RicEns-Net architecture combines multiple machine learning algorithms in a deep ensemble framework, integrating the strengths of each technique to improve predictive accuracy. Experimental results demonstrate that RicEns-Net achieves a mean absolute error (MAE) of 341 kg/Ha (roughly corresponds to 5-6% of the lowest average yield in the region), significantly exceeding the performance of previous state-of-the-art models, including those developed during the EY challenge. 1 Introduction This paper is grounded in the purpose and drive behind one of the Sustainable Development Goals (SDG) outlined by the United Nations—a comprehensive set of 17 objectives to be achieved by 2030 [ 1 ] . These goals collectively embody humanity’s pursuit of a sustainable future for both the planet and its inhabitants. Serving as a global framework, the 17 SDGs guide international endeavours to address the challenges posed by climate change while harmonising human ambitions for prosperity and improved quality of life. Central to this vision is the assurance of food security [ 2 ] , particularly for a significant portion of the global population residing in environmentally vulnerable areas susceptible to the impacts of climate and weather fluctuations. Rice is one of the most important staple foods globally, feeding more than half of the world’s population. It is cultivated on over 160 million hectares, producing around 500 million metric tons annually, with the majority of production concentrated in Asia. The region accounts for nearly 90% of global rice output, with countries such as China, India, and Indonesia leading in production. Besides being a dietary staple, rice also supports the livelihoods of millions of farmers, playing a vital role in rural economies. However, rice cultivation is highly resource-intensive, requiring significant amounts of water and labour, and is particularly vulnerable to climate change. Rising temperatures, changing precipitation patterns, and the increasing frequency of extreme weather events threaten rice yields globally, posing challenges to food security in many regions. This study underscores the critical importance of rice crops and Vietnam, focusing on their global and regional significance. Rice, often referred to as a “Gift of God” for its nutritional value, is a cornerstone of food security and public health worldwide. Despite its critical role, less than 8% of global rice production enters international trade [ 3 ] , emphasizing its largely localized consumption patterns, as noted by the Food and Agriculture Organisation Corporate Statistical Database (FAOSTAT). However, Vietnam stands as an exception, being one of the world’s leading rice exporters. The Mekong Delta, in particular, is central to the nation’s status as a global rice powerhouse, with its fertile paddy fields supporting diverse rice varieties. In Vietnam, rice holds unparalleled cultural, economic, and ecological importance. It serves as a dietary staple, forming the foundation of traditional cuisine, while also being integral to the livelihoods of millions of rural farmers. The sector is a driver of economic development, supported by government initiatives aimed at boosting productivity and sustainability. However, the ecological and climatic demands of rice cultivation present significant challenges. As a water-intensive crop, rice is highly sensitive to temperature fluctuations, requiring an optimal range of 25–35°C. Brief exposure to extreme heat or untimely rainfall during the reproductive stage can result in sterility or yield loss. This vulnerability, coupled with the long growing periods, makes rice cultivation particularly susceptible to climate change, underscoring the urgency of sustainable practices in Vietnam and globally. This paper introduces RicEns-Net , a novel ensemble learning model designed for accurate crop yield prediction. By integrating five diverse remote sensing data sources—Sentinel-1, Sentinel-2, Sentinel-3, NASA’s Goddard Earth Sciences (GES) Data and Information Services Center (DISC), and field measurements—the proposed framework leverages sophisticated data engineering techniques to enhance predictive performance. The model effectively combines synthetic aperture radar (SAR), multispectral imaging (MSI), meteorological parameters, and ground-truth observations, ensuring a robust and comprehensive approach to crop yield estimation. The inclusion of multi-modal data sources helps mitigate uncertainties associated with individual datasets, improving both spatial and temporal resolution in predictions. The remainder of this paper is organized as follows: Section 2 reviews related work, analysing previous findings and establishing the foundation for the study’s objectives and contributions. Section 3 details the materials and methods, including descriptions of data sources, preprocessing techniques, and feature engineering approaches. Section 4 presents the experimental results derived from the proposed model, followed by Section 5 , which interprets these findings and discusses their implications. Finally, Section 6 summarizes the study, outlines its limitations, and highlights potential directions for future research. 2 Related Works Advancements have influenced the evolution of crop yield monitoring in sensor technology and data analysis methodologies. While early studies highlighted rainfall as the primary determinant of crop yield, a key shift occurred in 1968 with the recognition of soil moisture as a more reliable predictor by Baier and Robertson [ 4 ] . Their work leverages spectral data to estimate crop yield based on vegetation health indicators. Over time, numerous vegetation indices (VIs) have been developed to assess vegetative conditions and physiological characteristics of crops. These VIs, including the Normalized Difference Vegetation Index (NDVI), Leaf Area Index (LAI) [ 5 ] , and Transformed Soil Adjusted Vegetation Index (TSAVI) [ 6 ] , play a crucial role in crop prediction models. Advancements in hyperspectral imaging have enabled the capture of fine-grained spectral data, facilitating the development of biochemical indices for quantifying plant constituents. Rice crop yield estimation relies on understanding the crop’s growth stages and environmental factors. Water level in paddy fields, rather than direct precipitation, is crucial for irrigated rice fields. Accumulating temperature is more important than temperature at certain times, as it affects the crop’s development stages. These factors are integrated into crop models to predict yield accurately. In the early 2000s, research surged leveraging imaging and machine learning technologies for crop yield prediction. Studies introduced novel methodologies, such as artificial neural networks (ANN) and SVR, to analyse remote sensing data and historical yield records. These approaches demonstrated more precise results compared to traditional models and prepared for more precise and scalable methods for estimating crop yields. Uno et al. [ 7 ] analyse hyperspectral images of corn plots in Canada using statistical and ANN approaches, demonstrating the potential of ANNs in predicting yield with higher accuracy compared to conventional models. Li et al. [ 8 ] introduce a