The accurate prediction of crop yields is crucial for enhancing agricultural efficiency and ensuring food security. This study assesses the performance of the CNN-LSTM-Attention model in predicting the yields of maize, rice, and soybeans in Northeast China and compares its effectiveness with traditional models such as RF, XGBoost, and CNN. Utilizing multi-source data from 2014 to 2020, which include vegetation indices, environmental variables, and photosynthetically active parameters, our research examines the model’s capacity to capture essential spatial and temporal variations. The CNN-LSTM-Attention model integrates Convolutional Neural Networks, Long Short-Term Memory, and an attention mechanism to effectively process complex datasets and manage non-linear relationships within agricultural data. Notably, the study explores the potential of using kNDVI for predicting yields of multiple crops, highlighting its effectiveness. Our findings demonstrate that advanced deep-learning models significantly enhance yield prediction accuracy over traditional methods. We advocate for the incorporation of sophisticated deep-learning technologies in agricultural practices, which can substantially improve yield prediction accuracy and food production strategies.

Measuring maize grain moisture content (GMC) variability at maturity provides an essential piece of information for the formulation of maize harvesting sequences and the applications of precision agriculture. Canopy chlorophyll content (CCC) is an important parameter that describes crop growth, photosynthetic rate, health, and senescence. The main goal of this study was to estimate maize GMC at maturity through CCC retrieved from multi-spectral UAV images using a PROSAIL model inversion and compare its performance with GMC estimation through simple vegetation indices (VIs) approaches. This study was conducted in two separate maize fields of 50.3 and 56 ha located in Hailun County, Heilongjiang Province, China. Each of the fields was cultivated with two maize varieties. One field was used as reference data for constructing the model, and the other field was applied to validate. The leaf chlorophyll content (LCC) and leaf area index (LAI) of maize were collected at three critical stages of crop growth, and meanwhile, the GMC of maize at maturity was also obtained. During the collection of field data, a UAV flight campaign was performed to obtain multi-spectral images from two fields at three main crop growth stages. In order to calibrate and evaluate the PROSAIL model for obtaining maize CCC, crop canopy spectral reflectance was simulated using crop-specific parameters. In addition, various VIs were computed from multi-spectral images to estimate maize GMC at maturity and compare the results with CCC estimations. When the CCC-retrieved results were compared to measured data, the R² value was 0.704, the RMSE was 34.58 μg/cm², and the MAE was 26.27 μg/cm². The estimation accuracy of the maize GMC based on the normalized red edge index (NDRE) was demonstrated to be the greatest among the selected VIs in both fields, with R² values of 0.6 and 0.619, respectively. Although the VIs of UAV inversion GMC accuracy are lower than those of CCC, their rapid acquisition, high spatial and temporal resolution, suitability for empirical models, and capture of growth differences within the field are still helpful techniques for field-scale crop monitoring. We found that maize varieties are the main reason for the maturity variation of maize under the same geographical and environmental conditions. The method described in this article enables precision agriculture based on UAV remote sensing by giving growers a spatial reference for crop maturity at the field scale.