Search
Close this search box.
Search
Close this search box.

Maize Yield Prediction Based On Multi-Modality Remote Sensing And Lstm Models In Nitrogen Management Practice Trials

Abstract:

Genotype x Environment x Management focused field studies are necessary to improve policies related to nitrogen use. Time series data were acquired by multiple sensors on UAVs, providing high resolution remote sensing-based candidate features for end-of-season yield prediction models. The Shapley Additive Explanations (SHAP library) was investigated for feature selection in LSTM models developed to investigate the impact of nitrogen management practices on end-of-season maize yield. Experiments included hyperspectral and LiDAR remote sensing and weather inputs, as well as multiple rates and forms of nitrogen application. Resulting models achieved relative RMSE values of <10%.

Published in: 2022 12th Workshop on Hyperspectral Imaging and Signal Processing: Evolution in Remote Sensing (WHISPERS)