TL;DR
This paper introduces an improved coupled System Dynamics (SD) and Patch-Generating Land Use Simulation (PLUS) model to predict the territorial spatial layout of the Dongting Lake Basin (DLB) in China from 2000 to 2050. By identifying optimal driving factors and simulation steps, the study provides a more accurate and efficient method for understanding and planning spatial development in lake-type basins, crucial for sustainable regional planning.
What Problem Does This Paper Address?
China's approach to territorial spatial planning is evolving from simple land use/land cover analysis to a more comprehensive "territorial spatial layout," considering production, living, and ecological spaces. However, much of this planning is done at administrative division scales, which may not fully capture the complex, non-linear drivers of spatial change, especially in unique geographical areas like lake-type basins. These basins, characterized by radial water systems and dense human settlement around water, have distinct evolutionary patterns influenced by factors beyond administrative boundaries. The Dongting Lake Basin (DLB), a vital ecological and economic region, exemplifies these characteristics, yet its future spatial development and the best methods for predicting it remain underexplored. This paper aims to fill this gap by developing and applying a robust modeling framework specifically tailored for lake-type basins.
Methods
The core of this research is the development of an "improved SD-PLUS coupling model." This approach combines two powerful modeling techniques:
- System Dynamics (SD): Used for quantitative prediction, SD models capture the complex feedback loops and dynamic behaviors of systems. In this study, it's employed to simulate the numerical structure of land use demand based on various subsystems (population, economy, environment, etc.) and their interrelationships.
- Patch-Generating Land Use Simulation (PLUS): This model excels at spatial pattern prediction. It integrates a land-use expansion analysis strategy (LEAS) to determine transfer probabilities based on driving factors and a cellular automaton (CA) with multiple random seeds (CARS) to simulate the spatial distribution of land types.
The novelty lies in the coupling and optimization of these models. Before simulation, the researchers meticulously identified the "best combination of driving factors" and the "best research step size" (5-year vs. 10-year intervals) affecting spatial change. This was achieved by:
- Cramer's V: Used to quantify the explanatory power of 18 potential driving factors (natural and social) on spatial layout changes. Factors with low explanatory power were eliminated.
- Multi-Layer Perceptron Artificial Neural Networks (MLP-ANN): Applied to test the accuracy of different combinations of the remaining driving factors and step sizes. This data-driven approach helps select the most predictive set of factors and the optimal temporal resolution for the simulation.
The final model construction involved feeding the optimized driving factors and step size into the coupled SD-PLUS framework to simulate and predict the territorial spatial layout.
Datasets
The study utilized the China's Multi-Period Land-Use Land-Cover Remote Sensing Monitoring Dataset (CNLUCC) from 2000, 2005, 2010, 2015, and 2020. This dataset provides high-resolution (30m x 30m) land-use information, categorized into 6 primary and 25 secondary types. Socio-economic data at the county level, relevant to the Dongting Lake Basin, were also collected and spatialized.
Key Insights and Ideas
- Lake-Type Basins are Unique: The research strongly emphasizes that lake-type basins have distinct spatial evolution patterns due to their radial water systems and concentrated human settlements, necessitating specialized modeling approaches beyond administrative boundaries.
- Optimizing Drivers and Steps is Crucial: Simply including more driving factors doesn't guarantee better accuracy. Identifying the most influential factors and the optimal temporal resolution (10-year step size in this case) significantly improves simulation accuracy and efficiency.
- SD-PLUS Synergy: The combination of SD for quantitative demand and PLUS for spatial allocation proved highly effective, overcoming the limitations of individual models.
- Future Stability with Trends: While the overall spatial structure of the DLB is predicted to remain relatively stable by 2050, specific trends like ecological space decrease, urban expansion, and farmland reduction are clearly identified, offering actionable insights for planning.
Key Results
- Optimal Model Parameters: The best combination of driving factors was identified as: "elevation, soil types, biodiversity value, distance from the administrative center, annual gross output value of the primary industry, annual gross output value of the secondary industry, annual population distribution, annual sales of consumer goods, annual fixed-asset investment." The optimal research step size was determined to be 10 years.
- High Simulation Accuracy: The improved SD-PLUS model achieved a high simulation accuracy, with a standard Kappa coefficient of 0.9181 for the DLB in 2020, outperforming previous models like Markov-PLUS (0.9074).
- Future Spatial Trends (2020-2050):
- Ecological Space: Decreases slowly initially, then accelerates between 2020-2025, before stabilizing.
- Urban Space: Shows rapid expansion until 2035, then slows down.
- Agricultural Space: Decreases gradually, with a slight increase in 2025.
- Forest Land: Decreases rapidly until 2025, then stabilizes.
- Farmland: Shows a gradual decrease over the long term.
- Three-Type Space Evolution: Ecological space shows a declining trend, urban space expands significantly, and agricultural space experiences a slight decrease.
Limitations and Future Work
- Natural Expansion Scenario: The current study simulates a "natural expansion scenario" without incorporating explicit policy constraints (e.g., urban development boundaries, agricultural land protection, ecological red lines). The authors acknowledge that real-world planning involves such constraints.
- Data Availability: Obtaining precise data for these control lines at provincial, municipal, and county levels is challenging. Future work could integrate these constraints into the PLUS model once data becomes available.
- Scale of Study: The DLB is a large lake-centric basin. Future research could explore different scales of basins (small river basins, large river basins, multi-basin junctions) to compare and generalize findings on territorial spatial layout evolution.
- FoM Interpretation: While the FoM metric was used for accuracy checks, its interpretation, especially at different scales, requires further in-depth study.
Summary updated