A CNN-RNN Framework for Crop Yield Prediction.

Area of Science:

• Agricultural Science
• Data Science
• Machine Learning

Background:

• Crop yield prediction is complex, influenced by genetics, environment, and management.
• Accurate yield forecasting is crucial for food security and agricultural planning.

Purpose of the Study:

• To develop and evaluate a novel deep learning framework for enhanced crop yield prediction.
• To compare the performance of the proposed model against established machine learning methods.

Main Methods:

• A hybrid deep learning model integrating Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs) was developed.
• The CNN-RNN model was trained and tested using historical environmental and management data for corn and soybean yields across the US Corn Belt.
• Performance was benchmarked against Random Forest (RF), Deep Fully Connected Neural Networks (DFNN), and LASSO regression.

Main Results:

• The CNN-RNN model achieved superior performance, with a root-mean-square-error (RMSE) of 9% and 8% for soybean and corn yields, respectively.
• The proposed model significantly outperformed all other tested methods in forecasting crop yields.
• The model demonstrated effective generalization to new environments without substantial accuracy loss.

Conclusions:

• The CNN-RNN framework offers a powerful approach for crop yield prediction, adept at capturing temporal dependencies and genetic improvements without explicit genotype data.
• This model's ability to generalize and its interpretability via backpropagation provide valuable insights into yield-influencing factors.
• The study highlights the potential of deep learning for advancing agricultural forecasting and decision-making.