Impacts of meteorological variables and machine learning algorithms on rice yield prediction in Korea.

As crop productivity is greatly influenced by weather conditions, many attempts have been made to estimate crop yields using meteorological data and have achieved great progress with the development of machine learning. However, most yield prediction models are developed based on observational data, and the utilization of climate model output in yield prediction has been addressed in very few studies. In this study, we estimate rice yields in South Korea using the meteorological variables provided by ERA5 reanalysis data (ERA-O) and its dynamically downscaled data (ERA-DS). After ERA-O and ERA-DS are validated against observations (OBS), two different machine learning models, Support Vector Machine (SVM) and Long Short-Term Memory (LSTM), are trained with different combinations of eight meteorological variables (mean temperature, maximum temperature, minimum temperature, precipitation, diurnal temperature range, solar irradiance, mean wind speed, and relative humidity) obtained from OBS, ERA-O, and ERA-DS at weekly and monthly timescales from May to September. Regardless of the model type and the source of the input data, training a model with weekly datasets leads to better yield estimates compared to monthly datasets. LSTM generally outperforms SVM, especially when the model is trained with ERA-DS data at a weekly timescale. The best yield estimates are produced by the LSTM model trained with all eight variables at a weekly timescale. Altogether this study shows the significance of high spatial and temporal resolution of input meteorological data in yield prediction, which can also serve to substantiate the added value of dynamical downscaling.

Random Forests for Global and Regional Crop Yield Predictions.

Accurate predictions of crop yield are critical for developing effective agricultural and food policies at the regional and global scales. We evaluated a machine-learning method, Random Forests (RF), for its ability to predict crop yield responses to climate and biophysical variables at global and regional scales in wheat, maize, and potato in comparison with multiple linear regressions (MLR) serving as a benchmark. We used crop yield data from various sources and regions for model training and testing: 1) gridded global wheat grain yield, 2) maize grain yield from US counties over thirty years, and 3) potato tuber and maize silage yield from the northeastern seaboard region. RF was found highly capable of predicting crop yields and outperformed MLR benchmarks in all performance statistics that were compared. For example, the root mean square errors (RMSE) ranged between 6 and 14% of the average observed yield with RF models in all test cases whereas these values ranged from 14% to 49% for MLR models. Our results show that RF is an effective and versatile machine-learning method for crop yield predictions at regional and global scales for its high accuracy and precision, ease of use, and utility in data analysis. RF may result in a loss of accuracy when predicting the extreme ends or responses beyond the boundaries of the training data.