New Dataset Maps Maize Yields Across Northeast China at 10-Meter Resolution
12 Oct 2026
A 10 m resolution maize yield dataset covering Northeast China from 2019 to 2024 has been released, providing detailed information on maize yield at the field scale. The dataset was developed by a research team led by Prof. LI Qiangzi and associate researcher DU Xin from the Aerospace Information Research Institute of the Chinese Academy of Sciences (AIRCAS) and is described in a recent study published in Earth System Science Data and the dataset is openly available through Zenodo.
Northeast China is one of the major maize-producing regions and an important commodity grain base in China. Reliable information on crop yield is essential for monitoring grain production, assessing agricultural losses and supporting agricultural management. However, existing yield statistics and remote sensing products are generally available at relatively coarse spatial scales, making it difficult to characterize yield differences among individual fields.
To address this challenge, the research team developed a yield estimation framework that combines the WOFOST crop growth model with a Gated Recurrent Unit (GRU) deep learning model. The crop growth model was used to represent the physiological processes of maize yield formation under different environmental and management conditions, while the deep learning model learned yield dynamics from Sentinel-2 vegetation information. Importantly, the framework does not require field-measured yield data for model training, making it suitable for large-area yield mapping where field observations are limited.
The resulting dataset provides annual maize yield estimates at 10 m resolution across Heilongjiang, Jilin, Liaoning and eastern Inner Mongolia for 2019-2024. Independent field observations collected during 2022-2024 showed good agreement with the mapped yields, demonstrating the capability of the framework to capture field-scale yield variability. The dataset also showed regional patterns consistent with agricultural statistics.
The high-resolution maps reveal clear spatial differences in maize yield across Northeast China. Higher yields are mainly concentrated in major agricultural plains, including the Songnen Plain, Liaohe Plain and Sanjiang Plain, while lower yields are more common in northern hilly areas and the semi-arid western region. The maps also capture substantial year-to-year changes in maize production. In particular, the effects of the typhoon and flooding in 2023 were reflected by pronounced yield reductions in parts of southern Heilongjiang and northern Jilin.
The dataset provides a new high-resolution data source for monitoring maize yield and agricultural disasters in Northeast China. It can support applications in precision agriculture, agricultural insurance, disaster assessment, black soil conservation and food security monitoring, and provides a potential approach for large-area crop yield estimation in regions where field observations are scarce.
Maize yield estimation framework for Northeast China. The framework combines the WOFOST crop growth model with a GRU deep learning model to estimate maize yield at 10 m resolution from Sentinel-2 observations without using field-measured yield data for model training. (Image by AIRCAS)
Research News
New Dataset Maps Maize Yields Across Northeast China at 10-Meter Resolution
A 10 m resolution maize yield dataset covering Northeast China from 2019 to 2024 has been released, providing detailed information on maize yield at the field scale. The dataset was developed by a research team led by Prof. LI Qiangzi and associate researcher DU Xin from the Aerospace Information Research Institute of the Chinese Academy of Sciences (AIRCAS) and is described in a recent study published in Earth System Science Data and the dataset is openly available through Zenodo.
Northeast China is one of the major maize-producing regions and an important commodity grain base in China. Reliable information on crop yield is essential for monitoring grain production, assessing agricultural losses and supporting agricultural management. However, existing yield statistics and remote sensing products are generally available at relatively coarse spatial scales, making it difficult to characterize yield differences among individual fields.
To address this challenge, the research team developed a yield estimation framework that combines the WOFOST crop growth model with a Gated Recurrent Unit (GRU) deep learning model. The crop growth model was used to represent the physiological processes of maize yield formation under different environmental and management conditions, while the deep learning model learned yield dynamics from Sentinel-2 vegetation information. Importantly, the framework does not require field-measured yield data for model training, making it suitable for large-area yield mapping where field observations are limited.
The resulting dataset provides annual maize yield estimates at 10 m resolution across Heilongjiang, Jilin, Liaoning and eastern Inner Mongolia for 2019-2024. Independent field observations collected during 2022-2024 showed good agreement with the mapped yields, demonstrating the capability of the framework to capture field-scale yield variability. The dataset also showed regional patterns consistent with agricultural statistics.
The high-resolution maps reveal clear spatial differences in maize yield across Northeast China. Higher yields are mainly concentrated in major agricultural plains, including the Songnen Plain, Liaohe Plain and Sanjiang Plain, while lower yields are more common in northern hilly areas and the semi-arid western region. The maps also capture substantial year-to-year changes in maize production. In particular, the effects of the typhoon and flooding in 2023 were reflected by pronounced yield reductions in parts of southern Heilongjiang and northern Jilin.
The dataset provides a new high-resolution data source for monitoring maize yield and agricultural disasters in Northeast China. It can support applications in precision agriculture, agricultural insurance, disaster assessment, black soil conservation and food security monitoring, and provides a potential approach for large-area crop yield estimation in regions where field observations are scarce.
Maize yield estimation framework for Northeast China. The framework combines the WOFOST crop growth model with a GRU deep learning model to estimate maize yield at 10 m resolution from Sentinel-2 observations without using field-measured yield data for model training. (Image by AIRCAS)