Examining Land Use/Land Cover Change and the Summertime Surface Urban Heat Island Effect in Fast-Growing Greater Hefei, China: Implications for Sustainable Land Development

Yingying Li, Yu Liu, Manjula Ranagalage, Hao Zhang, Rui ZhouISPRS Int. J. Geo Inf 2020 · DOI

TL;DR

This study investigates the relationship between land use/land cover (LULC) changes and the summertime surface urban heat island (SUHI) effect in Greater Hefei, China, between 1995 and 2016. The findings reveal that rapid urban expansion, particularly in the city center, significantly intensifies the SUHI effect, highlighting the need for sustainable land development strategies to mitigate its impacts.

What Problem Does This Paper Address?

Urbanization is a global phenomenon, leading to significant alterations in land surface characteristics and consequently impacting local climates. One of the most prominent consequences is the urban heat island (UHI) effect, where urban areas experience higher temperatures than their rural surroundings. While the UHI effect has been studied extensively, particularly in large coastal cities, there's a need for more detailed analysis in rapidly developing inland cities, especially concerning the surface UHI (SUHI) and its direct link to land use changes. This paper addresses this gap by quantitatively examining the relationship between LULC change and SUHI in Greater Hefei, a fast-growing city in central China, aiming to provide actionable insights for sustainable urban planning and climate adaptation. The study also highlights the limitations of traditional regression models in handling multicollinearity and quantifying the relative importance of various factors, motivating the use of more advanced statistical methods.

Methods

The researchers employed a multi-faceted approach to analyze LULC change and its impact on SUHI.

  • LULC Classification: Landsat TM and OLI/TIRS satellite imagery from 1995, 2002, 2007, and 2016 were used. A supervised maximum likelihood classification method was applied to categorize LULC into five classes: developed land, forest, cropland, water body, and bare land. Commercial 30m LULC maps and field survey data were used for validation.
  • LULC Change Measurement: To understand the spatial patterns of LULC change in relation to urban development, concentric zonal buffers were created at various distances (0-1.5 km, 1.5-3.0 km, ..., >66.0 km) from the city center. This allowed for an analysis of LULC changes along the urban-rural gradient.
  • SUHI Measurement: The surface urban heat island intensity (SUHII) and spatial extent (SUHISE) were calculated using retrieved Land Surface Temperature (LST). SUHII was defined as the LST difference between impervious surfaces and cooling surfaces within a buffer zone. SUHISE was measured by the number of pixels with SUHII ≥ 0. LST retrieval involved a series of steps, including converting digital numbers to brightness temperature and then to surface temperature, accounting for surface emissivity variations using NDVI.
  • Statistical Analysis: To quantify the relationship between LULC change (represented by area proportions of LULC types and population density within buffers) and SUHI indicators (SUHII and SUHISE), a Partial Least Squares Regression (PLSR) model was utilized. PLSR was chosen to address potential multicollinearity among independent variables and to provide a more robust analysis compared to traditional regression methods. Leave-one-out cross-validation was employed to ensure model reliability.

Datasets

  • Landsat Series Imagery: Multitemporal Landsat TM (August 1, 1995; July 11, 2002; August 2, 2007) and OLI/TIRS (July 25, 2016) images were used for LULC classification and LST retrieval.
  • Commercial 30m LULC Maps: These maps (1995, 2002, 2007, 2016) were used as auxiliary data for LULC classification and validation.
  • Local Socioeconomic Data: This included population density data, used as an independent variable in the PLSR models.
  • Urban Planning Documents: These provided context for understanding the study area's development.

Key Insights and Ideas

  • Heterogeneous Urban Growth: The study highlights that urban development in Greater Hefei is not uniform. The city center (downtown Hefei) experienced the most significant increase in developed land, indicating a rapid transition from a provincial capital to a major city.
  • SUHI Mirrors Urbanization: The spatial patterns of the summertime SUHI intensity and extent closely follow the distribution of developed land. This direct correlation underscores the impact of human-induced land surface changes on local thermal environments.
  • Buffer Zones Reveal Gradients: The concentric buffer analysis effectively demonstrated how LULC changes and SUHI effects vary along the urban-rural gradient. Higher SUHI levels were observed closer to the city center, gradually decreasing with distance.
  • PLSR's Utility: The application of PLSR proved crucial in disentangling the complex relationships between multiple LULC variables and SUHI, overcoming the limitations of multicollinearity that plague simpler regression models.
  • Implications for Planning: The findings strongly suggest that current land development patterns, particularly the concentration of development in the city center, are exacerbating the SUHI effect. This calls for a re-evaluation of urban planning strategies to promote more balanced development and incorporate green infrastructure.

Key Results

  • LULC Change: Between 1995 and 2016, cropland decreased by 14.26%, while developed land increased by a remarkable 126.57%. Forest, bare land, and water bodies also showed increases.
  • SUHI Intensity and Extent: Rapid urban expansion led to a significant increase in SUHII levels and SUHISE, particularly within the 6-21 km buffer zones.
  • PLSR Model Performance: The PLSR models explained a substantial portion of the variance: approximately 61.3% for SUHII and 79.8% for SUHISE.
  • Predictor Importance: For SUHII, the relative importance of predictors was: Developed land > Bare land > Pop_density. For SUHISE, the relative importance was similar, with Pop_density slightly surpassing Bare land. Negative coefficients indicated that increased cropland, water body, and forest cover helped mitigate SUHI.

Limitations and Future Work

  • Cloud Contamination and Revisiting Interval: The availability of cloud-free Landsat imagery with a 16-day revisiting interval limited the analysis to daytime SUHI effects. Future work could incorporate more frequent satellite data or other remote sensing platforms.
  • Air Temperature Data: The study relied on LST, not air temperature. Integrating site-specific air temperature data from weather stations, with careful consideration of station relocation biases, could provide a more comprehensive understanding of the UHI effect.
  • Resolution of SUHI Products: The 30m resolution of SUHI products might be insufficient for characterizing thermal divergence in complex urban environments. Exploring thermal sharpening techniques using higher-resolution satellite imagery is suggested.
  • Comparison with CUHI: Direct comparison between SUHI and canopy UHI (CUHI) was not possible. Future research could integrate in-situ measurements and computational fluid dynamics (CFD) to simulate CUHI indicators.
  • Comprehensive Modeling: The authors suggest that future models should integrate both biophysical LULC features and socioeconomic variables (like population density) to better understand the complex interactions driving SUHI.

The study concludes by emphasizing that optimizing land development patterns is key to mitigating the summertime SUHI effect and adapting to climate change, advocating for a shift towards ecological resilience and the integration of green spaces into urban planning.

Summary updated

BibTeX
@article{li2020examining,
  author = {Yingying Li and Yu Liu and Manjula Ranagalage and Hao Zhang and Rui Zhou},
  title = {{Examining Land Use/Land Cover Change and the Summertime Surface Urban Heat Island Effect in Fast-Growing Greater Hefei, China: Implications for Sustainable Land Development}},
  journal = {ISPRS Int. J. Geo Inf.},
  year = {2020},
  doi = {10.3390/ijgi9100568},
}
Figure