TL;DR
This study investigates the factors driving cotton cultivation patterns in China's Tianshan Mountains region from 2000-2020. By combining novel LESH and GWR models, researchers identified elevation, sunshine, and GDP as key drivers, highlighting the importance of spatial heterogeneity and factor interactions for sustainable cotton farming.
What Problem Does This Paper Address?
Cotton is a vital global crop, and understanding its cultivation patterns is crucial for optimizing production and ensuring sustainability. However, previous research often overlooks the complex interplay of factors and the significant spatial variations (heterogeneity) that influence where and how cotton is grown. This study aims to fill this gap by systematically analyzing the spatiotemporal evolution of cotton cultivation in the Northern Slope of the Tianshan Mountains (NSTM), China, a major cotton-producing region, and identifying the driving forces behind these changes, considering both global and local influences.
Methods
The researchers employed a two-pronged modeling approach:
- Locally Explained Stratified Heterogeneity (LESH) Model: This model, integrated with Shapley Additive Explanations (SHAP), is used to quantify the global influence of individual factors and their interactions on cotton distribution. It addresses the "black-box" problem of traditional spatial heterogeneity models by providing clear contributions of each factor.
- Geographically Weighted Regression (GWR): This method is used to explore local variations in the relationships between explanatory factors and cotton cultivation. GWR allows regression coefficients to vary across geographic locations, providing a more nuanced understanding of how factors influence cotton distribution at a local level.
By combining LESH and GWR, the study offers a comprehensive analysis of both global and local drivers and their interactions.
Datasets
- Landsat-5 TM and Landsat-8 OLI imagery: Used to derive cotton distribution maps from 2000 to 2020.
- High-resolution remote sensing imagery: Including GF-1, GF-2, GeoEye-1, and QuickBird, used for sample collection and validation.
- Ancillary data:
- Topography: ASTER GDEM V2 for elevation (ELE) and derived slope (SLO).
- Climate: ERA5-Land monthly averaged data for temperature (TEM), precipitation (PRE), and wind speed (WS); National Earth System Science Data Center and Science Data Bank for sunshine duration (SD).
- Soil: Resource and Environmental Science Data Platform for soil type (ST).
- Water Resources: Data Sharing and Service Portal for river network (WAT); ERA5-Land for evaporation (EVA) and runoff (RO).
- Socio-economy: Resource and Environmental Science Data Platform for gross domestic product (GDP).
- Administrative boundaries: Xinjiang Uygur Autonomous Region Platform for Common Geospatial Information Services.
- Cultivated land data: Used for masking.
- Statistical yearbooks: Xinjiang Statistical Yearbook and Xinjiang Production and Construction Corps Statistical Yearbook for validation.
No new benchmark datasets were introduced; the study utilized existing publicly available data.
Key Insights and Ideas
- Spatial Heterogeneity is Key: The study strongly emphasizes that cotton cultivation patterns are not uniform and that accounting for spatial variations is crucial for accurate analysis and effective management.
- Interactions Matter: The influence of individual factors is amplified or modified by their interactions with other factors, particularly between topographic, climatic, and socio-economic variables. Ignoring these interactions leads to an incomplete understanding.
- Dynamic Factor Importance: The influence of different drivers changes over time. For instance, sunshine duration's importance is waning, while temperature and GDP are increasing in significance.
- Elevation as a Primary Driver: Elevation plays a dominant role, explaining a significant portion of the spatial heterogeneity in cotton distribution, likely due to its influence on climate and suitability.
- Nonlinear Interactions: The study identified "nonlinear weakening interactions," where the combined effect of two factors is less than the sum of their individual effects, suggesting complex relationships like partial redundancy or mutual inhibition.
Key Results
- Cotton Expansion: The cotton planting area in the NSTM has expanded significantly, with an average annual growth rate of 2.10 × 10³ km²/year from 2000 to 2020.
- Dominant Drivers: The primary drivers of cotton distribution were identified as elevation (ELE), sunshine duration (SD), slope (SLO), temperature (TEM), runoff (RO), and gross domestic product (GDP).
- Elevation's Dominance: ELE explained approximately 40% of the spatial heterogeneity in cotton distribution.
- Shifting Influences: SD's influence declined, SLO remained stable, TEM's importance increased, and GDP showed a progressive upward trend.
- Spatial Patterns: Cotton cultivation is concentrated in the central and western counties of the NSTM, exhibiting high-high and low-low agglomeration patterns. Hot spots for cotton cultivation expanded spatially from 2010 to 2020.
- Model Performance: The LESH model achieved an optimal scale of 17 km for analysis, and the GWR model showed high fitness (adjusted R² from 0.75 to 0.86).
Limitations and Future Work
- Limited Human Factors: The study acknowledges that it observed fewer human factors influencing cotton distribution, such as market supply, policy, technological innovation, and price volatility. These factors can induce significant variations.
- Quantifying Latent Mechanisms: The identified nonlinear interactions suggest underlying mechanisms of mutual inhibition or interdependence that require further investigation for precise guidance in optimizing cotton planting.
- Future Work: The authors suggest that future research should focus on quantifying these latent mechanisms and incorporating a broader range of human factors to provide more comprehensive insights for cotton cultivation optimization.
Summary updated