The study revealed that grassland vegetation tends to recover in the arid area of northwest China

On June 1, the team of Professor Zheng Yi of the School of Environmental Science and Engineering of Southern University of Science and Technology published the latest research results in Geophysical Research Letters. The research team innovatively applied deep learning methods to grassland evolution prediction and revealed a new mechanism of global greening in inland arid areas.

Research diagram Courtesy of SUSTech

Since the 80s of last century, the global land vegetation has shown a trend of “greening”, and the carbon dioxide fertilization effect, the change of hydrothermal conditions and atmospheric nitrogen deposition are considered to be the main driving mechanisms. In recent years, grassland vegetation in the arid area of northwest China has also tended to recover, but its mechanism lacks systematic and quantitative analysis.

In this study, the research team used Landsat-7 ETM+ images acquired from the growing season from 2001 to 2015 as the data source for remote sensing interpretation of grassland cover. Using the driving data provided by HEIFLOW ecohydrological model and RIEMS regional climate model, a convolutional long-term short-term memory neural network (ConvLSTM) deep learning model was constructed to realize the spatiotemporal dynamic prediction of grassland cover in the lower Gobi region of Heihe River Basin, the second largest inland river basin in China.

The results show that with the support of remote sensing big data, the ConvLSTM deep learning model can accurately depict the evolution characteristics of grassland coverage in the study area with a spatial resolution of 1 km, and can also effectively predict the extreme evolution scenario of the mutual transformation of bare land and grassland. Existing process-driven or data-driven models are difficult to achieve this level of accuracy.

Between 2001 and 2015, the total area of grassland in the study area increased from 568 square kilometers to 741 square kilometers, an increase of about 30%. Attribution analysis using deep learning models found that 62% of vegetation changes were attributed to ecological flow management implemented in the basin since 2000, 32% were due to transboundary impacts of natural hydrological changes in the middle and upper reaches of the basin, and only 23% were local climate change factors in the study area. Both ecological flow management and transboundary impacts work through the restoration of groundwater in the study area.

This study quantitatively reveals the mechanism of grassland restoration in inland arid areas with ecological flow management and transboundary hydrological impacts, expands the understanding of global greening, and also demonstrates the great potential of artificial intelligence based on big data in ecohydrology research, which is of great significance for vegetation restoration and water resources management in arid areas.

Li Siqi, a 2021 doctoral student in the School of Environmental Science and Engineering of SUSTech, is the first author of the paper, Zheng Yi is the corresponding author, and SUSTech is the first unit of the paper. (Source: China Science News, Diao Wenhui) 

Related paper information:

Source link

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top button