Evaluation of Drought Monitoring and Prediction Methods for Dryland Croplands in Northeast China

Against the backdrop of intensifying climate change, the increasing risk of droughts, floods, and other climate-related hazards affecting crop production poses a growing threat to regional agricultural production and ecological environments. Northeast China is an important grain-producing region in China. Accurately assessing meteorological drought conditions in dryland croplands and improving short-term prediction capabilities are of great significance for agricultural risk management and food security. This study focuses on dryland croplands in Northeast China and investigates SPEI reconstruction and one-month-ahead prediction driven by environmental variables. The performance of different machine learning and deep learning methods is compared, and the effect of incorporating root-zone soil moisture on model prediction performance is further evaluated. Considering multiple environmental factors, precipitation, soil temperature, land surface temperature, potential evapotranspiration, and soil moisture at different soil depths are selected as model input variables. The Standardized Precipitation Evapotranspiration Index (SPEI) at accumulation timescales of 1, 3, and 6 months (SPEI-1, SPEI-3, and SPEI-6) is used to characterize meteorological drought conditions at different timescales and serves as the target variable for the drought monitoring and prediction models. Model performance is evaluated using the coefficient of determination (R2), root mean square error (RMSE), and mean absolute error (MAE). The results show that, for meteorological drought monitoring, the SPEI reconstruction model based on the Convolutional Neural Network–Long Short-Term Memory (CNN-LSTM) performs best, with R2 values of 0.82, 0.81, and 0.81 for SPEI-1, SPEI-3, and SPEI-6, respectively. Drought frequency in dryland croplands in Northeast China decreases with increasing drought severity, with high-frequency drought areas mainly concentrated in the central, central–western, and western parts of the Songnen Plain. For one-month-ahead prediction, the LSTM model achieves relatively good performance. Although incorporating root-zone soil moisture improves prediction accuracy to some extent, the overall improvement is relatively limited. Prediction performance varies across drought severity categories, and prediction errors generally increase with increasing drought severity, indicating that the model’s ability to predict extreme drought conditions requires further improvement. This study provides a methodological reference for meteorological drought monitoring and short-term prediction in dryland croplands in Northeast China.

Authors

Institutions

Publication Details

Journal
Agriculture
Published
2026-11-16
DOI
https://doi.org/10.3390/agriculture16222518
Primary Topic
Hydrology and Drought Analysis
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Evaluation of Drought Monitoring and Prediction Methods for Dryland Croplands in Northeast China

Hanyu Ju, Xiaoyang Liu, Yang Luo, Qingjia Li et al.
3 links
Agriculture
Hydrology and Drought Analysis
article

Evaluation of Drought Monitoring and Prediction Methods for Dryland Croplands in Northeast China

Hanyu Ju, Xiaoyang Liu, Yang Luo, Qingjia Li, Lei Li
article en

Abstract

Against the backdrop of intensifying climate change, the increasing risk of droughts, floods, and other climate-related hazards affecting crop production poses a growing threat to regional agricultural production and ecological environments. Northeast China is an important grain-producing region in China. Accurately assessing meteorological drought conditions in dryland croplands and improving short-term prediction capabilities are of great significance for agricultural risk management and food security. This study focuses on dryland croplands in Northeast China and investigates SPEI reconstruction and one-month-ahead prediction driven by environmental variables. The performance of different machine learning and deep learning methods is compared, and the effect of incorporating root-zone soil moisture on model prediction performance is further evaluated. Considering multiple environmental factors, precipitation, soil temperature, land surface temperature, potential evapotranspiration, and soil moisture at different soil depths are selected as model input variables. The Standardized Precipitation Evapotranspiration Index (SPEI) at accumulation timescales of 1, 3, and 6 months (SPEI-1, SPEI-3, and SPEI-6) is used to characterize meteorological drought conditions at different timescales and serves as the target variable for the drought monitoring and prediction models. Model performance is evaluated using the coefficient of determination (R2), root mean square error (RMSE), and mean absolute error (MAE). The results show that, for meteorological drought monitoring, the SPEI reconstruction model based on the Convolutional Neural Network–Long Short-Term Memory (CNN-LSTM) performs best, with R2 values of 0.82, 0.81, and 0.81 for SPEI-1, SPEI-3, and SPEI-6, respectively. Drought frequency in dryland croplands in Northeast China decreases with increasing drought severity, with high-frequency drought areas mainly concentrated in the central, central–western, and western parts of the Songnen Plain. For one-month-ahead prediction, the LSTM model achieves relatively good performance. Although incorporating root-zone soil moisture improves prediction accuracy to some extent, the overall improvement is relatively limited. Prediction performance varies across drought severity categories, and prediction errors generally increase with increasing drought severity, indicating that the model’s ability to predict extreme drought conditions requires further improvement. This study provides a methodological reference for meteorological drought monitoring and short-term prediction in dryland croplands in Northeast China.

AgricultureVol. 16(22)
Changchun University of Science and Technology (CN), Chinese Academy of Sciences (CN), Northeast Institute of Geography and Agroecology (CN)
National Natural Science Foundation of China
Climate action, Zero hunger, Life on land
Hydrology and Drought Analysis
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon