Recently, Wang Shudong, a researcher at the State Key Laboratory of Remote Sensing Science of the Institute of Space and Space Information Innovation, Chinese Academy of Sciences, led the eco hydrological remote sensing team to make progress in the field of irrigation water estimation. The team proposed a national scale irrigation water consumption estimation model based on machine learning and remote sensing observation data, and revealed the change trend and economic impact of China's irrigation water use under the future climate change scenario based on this model.

Irrigation agriculture accounts for approximately 20% of the global arable land area and contributes over 40% of food production. China has the world's largest irrigated farmland, accounting for more than half of China's arable land area. It can be said that irrigation water is quite important for crop growth and yield, especially in areas with water scarcity, where frequent droughts and extreme temperatures may exacerbate its impact. Therefore, accurately estimating changes in irrigation water usage is increasingly important for formulating optimal water resource allocation policies.

The existing methods for estimating irrigation water use are constrained by data availability and model structure, and their applicability is poor at the national scale and in future climate change scenarios. Based on this, the team developed a new model based on machine learning, which integrates a series of high-precision hydrological satellite remote sensing products (precipitation, evapotranspiration, soil moisture, and snow water equivalent), meteorological driving factors, economic statistical data, and numerical model simulations to estimate irrigation water at the national scale under a data-driven framework. The new model has shown high accuracy in estimating irrigation water use: compared with the bulletin values of the Ministry of Water Resources of 339 prefecture level cities in China, the determination coefficient ranges from 0.86 to 0.91, and the root mean square error ranges from 0.261 to 0.361 cubic kilometers per year. Through independent observation verification at 11 farmland sites, the study found a significant correlation between simulated data and field observation data, with an accuracy rate of over 90%.

This study further considers a range of climate and socio-economic scenarios, combining the established machine learning framework with four advanced Earth system models to provide trends and related costs of irrigation water usage in China over the next 70 years. Research shows that based on different greenhouse gas emission scenarios, approximately 60% of provinces will see an increase in irrigation water usage over the next 70 years, especially in the northwest and northern regions. Research has found that compared to data from the 1980s to 2010, irrigation water usage in China is expected to increase by a maximum of 17.1% by 2050, and the increase in irrigation water usage will result in an annual cost increase of up to 3.91 billion US dollars; By 2100, it is expected to increase by up to 34.8%, with an annual cost increase of up to $6.5 billion. This highlights the urgency of sustainable utilization and management of water resources.

This study proposes effective methods for estimating current and future irrigation water usage, which is expected to provide recommendations for agricultural water policies. On March 9th, the research findings were published in the journal Earth's Future, titled "Uncovering Current and Future Variations of Irrigation Water Use Across China Using Machine Learning". This study was conducted in collaboration with the Chinese Academy of Aeronautics and Astronautics, the Chinese Academy of Meteorological Sciences, and the University of Pennsylvania in the United States. The State Key Laboratory of Remote Sensing Science is the first completion unit.

Paper link:论文链接