Work

Harnessing multi-source data and machine learning for enhanced rice yield estimation

Year: 2025

Type: article

Abstract: Accurate rice yield prediction is crucial for ensuring the food security of a region. This study analyses ways of rice yield prediction using machine learning-based modeling techniques and various soil and climate attributes. The Rice yield trend is analyzed for the state of Meghalaya from the period of 2000 to 2018. This study mainly aims to find out the trend of rice yield and investigate how remote sensing, climatic variables, and soil parameters affect rice yield. The valuable insights into vegetation dynamics are captured using NDVI (Normalized Difference Vegetation Index) data. Climate data, such as temperature, precipitation, and wind components, were also incorporated to reason for the effects of weather on rice yield. The prediction accuracy of models is further enhanced by including soil properties like texture, bulk density, land surface temperature, and pH. We have collected historic spatiotemporal data from seventy paddy fields based on the spatial distribution of the crop. The Mann-Kendall (MK) test was employed to evaluate the trend. Additionally, three different recurrent neural network architectures namely Bi-directional Long Short Term Memory (Bi-LSTM), LSTM, and GRU were implemented to build an estimation model. The results of MK test suggest a significant increase in the yearly rice yield. The Bi-LSTM network appeared as the superior model with a Mean Square Error (MSE) of 0.0027 and a Mean Absolute Error (MAE) of 0.02407. Furthermore, the study underscores the influence of temperature, rainfall, and soil parameters in shaping rice yield. The proposed framework demonstrates high predictive accuracy, surpassing conventional regression-based approaches. Integrating multi-source data enables us to identify the key factors driving yield variability, offering actionable insights for optimizing yield strategies and decision-making processes.

Source: Computers and Electronics in Agriculture

DOI: 10.1016/j.compag.2025.110633

Authors: Puja Saha, Amitabha Nath

Institution: North Eastern Hill University

Language: English

Topic: Smart Agriculture and AI

Subfield: Plant Science

Field: Agricultural and Biological Sciences

Domain: Life Sciences

SDGs: Zero hunger, Climate action

Open Access status: hybrid

Estimated APC paid: $3,870