Evaluation and prediction of water conservation of the Yellow river basin in Sichuan Province, China, based on Google Earth Engine and CA-Markov

Zhichong Yang, Xiaoai Dai, Heng Lü, Chao Liu, Ruihua Nie, Min Zhang, Лей Ма, Naiwen Li, Tiegang Liu, Yuxin He, Zhengli Yang, Ge Qu, Weile Li, Youlin WangHeliyon 2023 · DOI

TL;DR

This study quantifies water conservation (WC) in China's Yellow River Basin in Sichuan Province (YRS) from 2001-2020 using Google Earth Engine and predicts future WC under different scenarios with CA-Markov. The findings highlight climate as the primary driver of WC, with land cover playing a significant role, and suggest a shift in ecological management strategies towards improving land use quality and afforestation over simply reducing cropland.

What Problem Does This Paper Address?

The Yellow River Basin (YRB) in China faces severe soil erosion, and its upper reaches in Sichuan Province (YRS) are crucial for downstream water conservation (WC). Despite growing attention and policy interventions aimed at improving WC, there's a significant research gap regarding long-term WC trends in the YRS, largely due to a lack of in-situ data. This study aims to fill this gap by providing a quantitative evaluation of WC, analyzing its spatio-temporal variations, and predicting future trends under various scenarios to inform more effective ecological management.

Methods

The researchers employed a multi-pronged approach:

  • Water Balance Equation: WC was calculated using the fundamental water balance equation: WC = Precipitation (PR) - Evapotranspiration (ET) - Runoff (RO). This calculation was performed annually for the May-July period.
  • Google Earth Engine (GEE): GEE was utilized as a cloud-based platform for processing and analyzing large-scale, multi-source remote sensing data. This allowed for efficient spatio-temporal analysis of WC and land cover (LC) over the study period.
  • CA-Markov Model: This model was used to predict future land cover changes and, consequently, future WC. It integrates the Markov chain for transition probability analysis with Cellular Automata for spatial simulation, enabling the modeling of complex geographical processes.
  • Mann-Kendall Test: This non-parametric statistical test was applied to identify abrupt change points in climate and WC time series, helping to pinpoint significant shifts in trends.

Datasets

The study utilized several publicly available datasets, primarily accessed through Google Earth Engine:

  • Precipitation (PR): CHIRPS (Climate Hazards Group InfraRed Precipitation with Station data) at 0.05° resolution.
  • Evapotranspiration (ET): MOD16A2.006 (MODIS/Terra Net Evapotranspiration) at 500m resolution, monthly data.
  • Runoff (RO): TerraClimate at 0.04° resolution, 8-day data.
  • Land Cover (LC): MCD12Q1.006 (MODIS/Terra+Aqua Land Cover Type) at 500m resolution, yearly data, using the IGBP classification scheme.
  • Validation Data: Land use/land cover (LUCC) data with 30m spatial resolution from the Data Center for Resources and Environmental Sciences, Chinese Academy of Sciences (RESDC), was used to validate the accuracy of MODIS LC data.

Key Insights and Ideas

  • Climate is King, but Land Cover Matters: While precipitation (PR) is identified as the dominant factor driving inter-annual WC variations, land cover (LC) significantly influences WC capacity. This highlights the interconnectedness of climate and land management.
  • Elevation's Role: Higher elevations in the YRS are associated with higher runoff (RO) and consequently lower WC, suggesting that terrain plays a crucial role in shaping water conservation patterns.
  • Unexpected Land Cover Contributions: Counterintuitively, croplands showed a higher WC capacity than grasslands in this region. This challenges the conventional understanding and the rationale behind some existing ecological policies.
  • GFG Policy Nuances: The "Grain for Green" (GFG) policy has led to significant land cover changes, but its long-term impact on WC improvement may gradually decrease, suggesting a need for policy re-evaluation.
  • Transforming Management: The study strongly advocates for a transformation in ecological management, emphasizing improving land use quality and converting grasslands to forests as more effective strategies than simply reducing cropland.

Key Results

  • WC Fluctuations: The total WC in the YRS fluctuated between 1.93 and 6.77 billion m³ from 2001 to 2020, with an average annual increase of 4 billion m³.
  • Dominant Climate Factors: Precipitation (PR) showed the highest randomness (13.53 billion m³), followed by runoff (RO) (10.30 billion m³), and evapotranspiration (ET) (1.46 billion m³). PR's variability directly impacts WC.
  • Abrupt Change Points: Mann-Kendall tests identified abrupt change points in WC and climate factors in 2008, 2012, and 2018.
  • Land Cover WC Capacities: Forests exhibit the highest WC capacity (over 600 mm/km²), followed by grasslands (around 250 mm/km²). Barren land, however, can lead to a WC loss of approximately 300 mm/km².
  • Future WC Projections: Under a "current scenario" (reflecting recent trends), WC in YRS is predicted to exceed 7.5 billion m³ by 2025. However, some counties like Hongyuan and Aba might experience WC losses.
  • CA-Markov Accuracy: The CA-Markov model demonstrated high accuracy, with an overall accuracy of 98.1% and a Kappa coefficient of 0.96 for LC prediction.

Limitations and Future Work

  • Methodological Refinements: The authors acknowledge that the WC evaluation method on GEE could be improved by incorporating more hydrological mechanisms and complex physical models.
  • Data Resolution and Period: The study was limited by the availability of MODIS data (starting from 2001). Extending the study period further back would require addressing data matching challenges. Higher spatial resolution LC data (e.g., Dynamic World) is available but only for recent years.
  • Seasonal WC: While the study focused on the rainy season (May-July) for WC calculation, acknowledging that seasonal WC changes are more reflective of local WC function, future work could explore intra-annual variations.
  • Unified Standards: The paper points out the lack of unified standards for WC evaluation and validation across different studies, which hinders direct comparison and robust assessment.
  • Future Directions: The authors suggest exploring complex physical models on GEE, investigating longer time series with higher resolution data, and developing unified WC evaluation standards. They also emphasize the need for policy adjustments, particularly focusing on converting grasslands to forests and improving land use quality rather than solely reducing cropland.

Summary updated

Abstract

The Yellow River Basin in China has the world's most serious soil erosion problem. The Yellow River Basin in Sichuan Province (YRS), as the upper reaches of the Yellow River, and its water conservation (WC) capacity greatly affects the ecological environment of the downstream basin. In recent years, YRS has received more and more attention, and numerous policies have been developed to improve local WC. However, there is a vacancy in the long-term research of WC in the YRS due to the lack of in-situ data. This study quantitatively evaluated the WC of YRS from 2001 to 2020 through Google Earth Engine (GEE) and analyzed the spatio-temporal variations of WC and land cover (LC). CA-Markov predicted the LC and WC in 2025 under three scenarios to assess the contribution of different scenarios to WC. The WC in YRS fluctuated from 1.93 to 6.77 billion m3. The climate is the dominant factor of WC change, but the effect of LC on WC is also evident. The WC capacity increases with vegetation coverage and height. The WC capacity of forests per km2 exceeds 600 mm, while that of grasslands is about 250 mm, and barren can cause around 300 mm of WC loss. In 2025, the WC in YRS may exceed 7.5 billion m3, but the past ecological management mode should be transformed. Improving the quality of land use and converting grasslands to forests is better than reducing cropland to improve WC.
BibTeX
@inproceedings{yang2023evaluation,
  author = {Zhichong Yang and Xiaoai Dai and Heng Lü and Chao Liu and Ruihua Nie and Min Zhang and Лей Ма and Naiwen Li and Tiegang Liu and Yuxin He and Zhengli Yang and Ge Qu and Weile Li and Youlin Wang},
  title = {{Evaluation and prediction of water conservation of the Yellow river basin in Sichuan Province, China, based on Google Earth Engine and CA-Markov}},
  booktitle = {Heliyon},
  year = {2023},
  doi = {10.1016/j.heliyon.2023.e17903},
}
Figure