Source-linked AI summary
Understanding the drivers of sustainable land expansion using a patch-generating land use simulation (PLUS) model: A case study in Wuhan, China
Xun Liang, Qingfeng Guan, Keith C. Clarke, Shishi Liu, Bingyu Wang, Yao Yao
TL;DR
Prior CA research emphasized technical modeling procedures, while limited attention addressed the drivers of individual land-type transitions and detailed patch evolution across multiple land-use types. The study presents a rule-mining framework and patch-growth CA approach, achieving higher simulation accuracy and more similar landscape metrics than other tested models.
Problem
Prior modeling research emphasized technical procedures, with limited attention to drivers of individual land-type transitions and detailed patch evolution across multiple land-use types.
Method
The study presents a rule-mining framework based on land expansion analysis and a CA approach designed to simulate patch growth across multiple land-use types.
Results
FOM increased from 0.1895 to 0.2642, while the model produced more similar landscape metrics than the other models tested.
Takeaways & Limitations
The approach was developed and applied to inform land-use policy and land-management decisions.
Takeaways & Limitations
Raster-based land-use data constrain simulation because they provide less information than the data needed to represent landscape evolution and land-transition drivers.
Abstract
from arXiv · showhide
Cellular Automata (CA) are widely used to model the dynamics within complex land use and land cover (LULC) systems. Past CA model research has focused on improving the technical modeling procedures, and only a few studies have sought to improve our understanding of the nonlinear relationships that underlie LULC change. Many CA models lack the ability to simulate the detailed patch evolution of multiple land use types. This study introduces a patch-generating land use simulation (PLUS) model that integrates a land expansion analysis strategy and a CA model based on multi-type random patch seeds. These were used to understand the drivers of land expansion and to investigate the landscape dynamics in Wuhan, China. The proposed model achieved a higher simulation accuracy and more similar landscape pattern metrics to the true landscape than other CA models tested. The land expansion analysis strategy also uncovered some underlying transition rules, such as that grassland is most likely to be found where it is not strongly impacted by human activities, and that deciduous forest areas tend to grow adjacent to arterial roads. We also projected the structure of land use under different optimizing scenarios for 2035 by combining the proposed model with multi-objective programming. The results indicate that the proposed model can help policymakers to manage future land use dynamics and so to realize more sustainable land use patterns for future development. Software for PLUS has been made available at https://github.com/HPSCIL/Patch-generating_Land_Use_Simulation_Model
1. Introduction
The introduction identifies two persistent gaps in land-use CA models: limited insight into change drivers and inadequate simulation of multiple land-use patch evolution. It motivates a PLUS approach that connects land expansion drivers with raster-based patch-growth simulation for planning and decision-making.
- Most CA research has prioritized technical procedures, calibration, and rules over conceptual understanding of the underlying causes of LULC change.
- Existing CA models are weak at revealing land-use change drivers and cannot capture the spatiotemporal evolution of patches across multiple land-use types.
- Transition analysis examines changed cells between two land-use dates and their neighboring states or contributing-factor layers, but becomes difficult as transition combinations proliferate.With K land-use types, the number of transitions is K^2 − K, increasing computational complexity and reducing flexibility.
- Pattern-analysis CA models avoid exponentially increasing transition combinations through competition among land-use types, but cannot reveal how driving factors cause change over a specific interval.Their one-map distribution rules do not capture land-use change rules from class transitions.
- The resulting modeling need is especially important because natural land-use parcels can be small, irregular, and dispersed, limiting vector-based CA applications.Vector-based land-use data are also more difficult to obtain than raster-based data, constraining large-scale application.
- The study develops a raster-based PLUS model to reveal drivers and their differing contributions while improving simulation of multiple land-use patch growth.The introduction links detailed patch simulation with plan-specific decision-making and landscape-pattern emulation.
2. Method
The PLUS model combines LEAS rule mining, CARS patch-generating cellular automata, and multiobjective programming to simulate and optimize land-use structures. LEAS simplifies transition-rule analysis by focusing on expanding destination types, while CARS integrates land-use demand with local competition and multi-type patch growth.
- 2. Method: The PLUS framework combines LEAS for rule mining, CARS for patch-generating simulation, and MOP for scenario-based land-use optimization.The framework is organized around two PLUS modules and a multiobjective programming component for determining land-use structures under different scenarios.
- 2.1.1. Land Expansion Analysis Strategy: Random forest classification converts land-use expansion into binary classification tasks linking each destination type with multiple driving factors.Changed cells are sampled from two land-use dates, labeled by whether they represent expansion of the target type, and analyzed separately by land-use type.
- 2.1.1. Land Expansion Analysis Strategy: LEAS analyzes growing patches by destination land-use type while ignoring source types, reducing the number of transition types that must be examined.This approach obtains rules for all land-use types and provides temporal information about changes during a specified interval.
- 2.2. CA Model based on Multi-type Random Patch Seeds: CARS uses multi-type random seeds and integrates global land-use demands with local competition through a self-adaptive coefficient.The coefficient responds to the gap between current land amounts and future demands, while neighborhood effects contribute to local growth probabilities.
- 2.2.2. Multi-type Random Patch Seeds based on a Descending Threshold: Threshold descent and random patch seeding allow land-use patches to grow dynamically while restricting spontaneous growth according to transition rules and growth probabilities.The decreasing threshold favors cells with higher overall probabilities, and the resulting patches can develop under spatiotemporal constraints.
- 2.3. Generating Sustainable Land Use Structure with MOP: MOP defines objective functions and constraints to identify sustainable land-use structures while accommodating ecological, economic, and planning considerations.The study uses objectives including ecological capacity and ecological service value, alongside scenario-specific optimization such as economic development.
3. Study Area and Data Sources
The study examines Wuhan, a rapidly growing megacity with ecologically valuable land, using 2003 and 2013 spatial land-use data and multiple environmental and socioeconomic drivers.
- Study Area: Built-up land expanded from 4.19×10^4 ha in 1988 to 49.39×10^4 ha in 2011, encroaching on surrounding ecologically valuable lands.
- Study Area: Wuhan covers 8,494.41 km2 across 11 districts and contains substantial water bodies, hilly forests, and deciduous broadleaf vegetation.
- Data Sources: The study used land-use data classified from Landsat imagery in 2003 and 2013 at 30 m × 30 m resolution across 2931×2931 grid cells.
- Data Sources: Seven land-use types were mapped: built-up areas, cropland, two forest types, grassland, water bodies, and bare land.
- Data Sources: Raster maps represented climatic, environmental, and socioeconomic driving factors aligned with the land-use data extent.
4. Model Implementation and Results
The PLUS model was calibrated with land-expansion samples and neighborhood effects, then evaluated against observed 2013 patterns and alternative CA models. It also identified land-growth relationships and projected 2035 scenarios.
- Model Implementation: The PLUS model used 5% random samples, 14 predictor variables, and 50-tree random forests to generate growth-probability maps for each land-use type.
- Simulation Accuracy: FOM = 0.2642 was achieved by PLUS, whose simulated pattern was most similar to the observed 2013 landscape and had small, scattered errors.
- Simulation Accuracy: FOM increased from 0.1310 to 0.2642 with LEAS and from 0.2514 to 0.2642 with multi-type random patch seeds.
- Simulation Accuracy: PLUS had FOM 0.2642 versus 0.1895 for FLUS and matched 7 of 15 landscape metrics most closely, more than the other tested models.
- Expansion Drivers: Grassland growth was associated with areas far from administrative centers, whereas deciduous forest growth was concentrated adjacent to arterial roads in suburbs.
- 2035 Scenarios: Under 2035 scenarios, BS and ED featured rapid urban expansion and cropland reduction, while EP retained the most ecological land and the least urban land.
- 2035 Scenarios: Around Tangxun Lake, EP best preserved deciduous forest patches and enlarged the forest corridor, whereas ED preserved more cropland but fragmented forest.
5. Discussion
The discussion emphasizes PLUS’s ability to represent time-specific, multi-type land-use change and support policy-oriented scenario optimization. Its simulations were more accurate and landscape-similar than tested alternatives.
- Model Contributions: The PLUS CA module combines multi-type patch generation with rule mining to describe land-use change properties for specific time intervals.
- Model Contributions: LEAS provides quantitative information about how multiple driving factors influence expansion of different land-use types.
- Model Contributions: Because driving forces may change over time, PLUS derives temporally informed transition rules that are more flexible than prior distribution rules.
- Validation: FOM increased from 0.1895 to 0.2642, and PLUS produced more similar landscape metrics than the other tested models.
- Scenario Optimization: Multi-objective programming generated optimized land-use patterns under alternative policies, including economic, ecological-service, and ecological-capacity objectives.
- Scenario Optimization: The SD scenario was designed to maximize economic benefits, ecological service value, and ecological capacity simultaneously.
- Policy Implications: The resulting land-use structure was presented as support for policymakers determining future management objectives and land policies.
6. Conclusion
The PLUS model combines LEAS with a multi-type random-patch CA to simulate complex land-use patches, reveal expansion drivers, and support scenario-based planning. In Wuhan, it produced more accurate and realistic landscape patterns than tested alternatives and offered policy guidance for future land-use management.
- 6. Conclusion: The study introduced PLUS as a framework combining LEAS rule mining with CARS to simulate multi-type land-use patch growth at fine resolution.LEAS provides growth probabilities, while CARS uses multi-type random patch seeds to generate realistic landscape patterns.
- 6. Conclusion: PLUS was calibrated with Wuhan land-use data from 2003–2013 and achieved higher simulation accuracy and more similar landscape patterns than the tested models.The comparison evaluated both simulation accuracy and similarity to observed landscape patterns.
- 6. Conclusion: LEAS revealed transition effects that previous analysis methods did not identify, including distinct relationships between human activity, grassland growth, and deciduous forest expansion.Grassland was most likely to grow where human activity was weaker, whereas new deciduous forests were most likely to grow along arterial roads in suburbs.
- 6. Conclusion: Coupling PLUS with multi-objective programming produced land-use structures and associated economic and ecological benefits under different 2035 scenarios.The sustainable-development scenario was proposed as a baseline for assessing whether the study region followed a sustainable pathway.
- 6. Conclusion: The resulting approach provides substantive guidance for managing future land-use patterns under different development objectives and policies.The authors position the model as a tool for understanding land-expansion mechanisms and supporting decision-making.