Determining a precise nitrogen fertilizer requirement for maize in a particular field and year has proven to be a challenge due to the complexity of the nitrogen inputs, transformations and outputs in the nitrogen cycle. Remote sensing of maize nitrogen deficiency may be one way to move nitrogen fertilizer applications closer to the specific nitrogen requirement. Six vegetation indices [normalized difference vegetation index (NDVI), green normalized difference vegetation index (GNDVI), red-edge normalized difference vegetation index (RENDVI), triangle greenness index (TGI), normalized area vegetation index (NAVI) and chlorophyll index-green (CIgreen)] were evaluated for their ability to detect nitrogen deficiency and predict grain maize grain yield. Strip trials were established at two locations in Arkansas, USA, with nitrogen rate as the primary treatment. Remote sensing data was collected weekly with an unmanned aerial system (UAS) equipped with a multispectral and thermal sensor. Relationships among index value, nitrogen fertilizer rate and maize growth stage were evaluated. Green NDVI, RENDVI and CIgreen had the strongest relationship with nitrogen fertilizer treatment. Chlorophyll Index-green and GNDVI were the best predictors of maize grain yield early in the growing season when the application of additional nitrogen was still agronomically feasible. However, the logistics of late season nitrogen application must be considered.

Unmanned aerial vehicles (UAVs) equipped with multispectral sensors offer high spatial and temporal resolution imagery for monitoring crop stress at early stages of development. Analysis of UAV-derived data with advanced machine learning models could improve real-time management in agricultural systems, but guidance for this integration is currently limited. Here we compare two deep learning-based strategies for early warning detection of crop stress, using multitemporal imagery throughout the growing season to predict field-scale yield in irrigated rice in eastern Arkansas. Both deep learning strategies showed improvements upon traditional statistical learning approaches including linear regression and gradient boosted decision trees. First, we explicitly accounted for variation across developmental stages using a 3D convolutional neural network (CNN) architecture that captures both spatial and temporal dimensions of UAV images from multiple time points throughout one growing season. 3D-CNNs achieved low prediction error on the test set, with a Root Mean Squared Error (RMSE) of 8.8% of the mean yield. For the second strategy, a 2D-CNN, we considered only spatial relationships among pixels for image features acquired during a single flyover. 2D-CNNs trained on images from a single day were most accurate when images were taken during booting stage or later, with RMSE ranging from 7.4 to 8.2% of the mean yield. A primary benefit of convolutional autoencoder-like models (based on analyses of prediction maps and feature importance) is the spatial denoising effect that corrects yield predictions for individual pixels based on the values of vegetation index and thermal features for nearby pixels. Our results highlight the promise of convolutional autoencoders for UAV-based yield prediction in rice.