A Multitemporal and Multilevel Land Surface Temperature Regional Attribute Change Analysis in Henan, China, Using MODIS Imagery

Zongze Zhao, Bingke Sun, Gang Cheng, Cheng Wang, Na Yang, Hongtao Wang, Xiaojie TangSustainability 2022 · DOI

TL;DR

This paper introduces a novel "Multitemporal Multilevel" (MTML) method to analyze land surface temperature (LST) changes in Henan, China, using MODIS data. By examining LST across different time periods and temperature levels, the study reveals complex relationships between temperature, vegetation, and built-up areas, highlighting regional variations in heat island effects.

What Problem Does This Paper Address?

Urbanization significantly impacts land surface temperature (LST), leading to phenomena like the urban heat island (UHI) effect. While satellite remote sensing provides valuable LST data, traditional analysis methods often struggle to capture the nuanced, hierarchical changes in temperature across different regions and over time. Existing approaches can miss how temperature patterns evolve across various temperature thresholds and their correlation with land cover types. This paper aims to address this by developing a method that can analyze these complex, multi-layered temperature dynamics.

Methods

The core of this research is the Multitemporal Multilevel (MTML) method. It involves several key steps:

  1. Data Acquisition and Preprocessing: MODIS LST data from five specific dates between 2006 and 2021 were acquired. To ensure comparability, data were selected under similar climatic conditions, and the influence of climate was accounted for.
  2. Threshold Superposition Analysis: LST data were processed using threshold superposition to create binary grids representing different temperature levels. This allows for the identification of connected regions at specific temperature ranges.
  3. Connectivity Analysis: A contraction-based connectivity analysis was applied to these binary grids to identify and delineate "connected regions" of LST at various temperature levels.
  4. Tree Structure Construction: A hierarchical tree structure was built to represent the inclusion relationships of these connected regions across different temperature levels. Nodes in this tree represent connected regions, allowing for an analysis of how temperature centers evolve and are nested.
  5. Attribute Information Calculation: For each connected region and temperature level, attributes such as average LST, NDVI (Normalized Difference Vegetation Index), and NDBI (Normalized Difference Built-up Index) were calculated.
  6. Statistical and Correlation Analysis: The study then statistically analyzed these attributes to understand changes over time and correlations between LST and land cover indices (NDVI, NDBI) at different temperature levels and across different region sizes. Correlation coefficients (CC_TV for LST-NDVI, CC_TB for LST-NDBI) were computed.

Datasets

  • MODIS LST (MOD11_L2) products: Surface temperature data from the Terra satellite.
  • MODIS calibrated radiation products (MOD021KM): Used for generating NDVI and NDBI.
  • Dates: 15 May 2006, 1 May 2010, 7 May 2014, 29 April 2017, and 8 May 2021.
  • Study Area: Henan Province, China.
  • Ancillary Data: Weather data from a weather data website was used for climate condition analysis.

Key Insights and Ideas

  • Hierarchical Temperature Analysis is Crucial: Simply looking at average LST isn't enough. The MTML method reveals that temperature patterns have a hierarchical structure, with "temperature centers" nested within broader temperature regions, which is vital for understanding UHI dynamics.
  • Complex LST-Land Cover Relationships: While generally, higher temperatures correlate with more built-up areas (positive NDBI correlation) and less vegetation (negative NDVI correlation), this relationship weakens at higher temperature levels. This suggests that at extreme heat, other factors might dominate or the influence of vegetation/built-up areas becomes less distinct.
  • Regional Hotspots Evolve: High-temperature areas are not static. The study identifies persistent hot spots in the northern, central, and western parts of Henan, while eastern and southern regions tend to be cooler. The distribution and intensity of these hot spots change over time, reflecting urbanization and land-use shifts.
  • Size Matters for Attribute Stability: The statistical attributes of connected regions are more stable and reliable for trend analysis when the regions are larger. Smaller regions are more susceptible to pixel-level variations.

Key Results

  • Negative Correlation with NDVI, Positive with NDBI: Generally, LST shows a negative correlation with NDVI and a positive correlation with NDBI.
  • Decreasing Correlation at Higher Temperatures: As temperature levels increase, the correlation between LST and surface feature indices (NDVI, NDBI) decreases.
  • Regional Temperature Distribution: Concentrated high-temperature areas are predominantly found in the northern, central, and western regions of Henan Province, with lower temperatures in the east and south.
  • Temporal Trends: The study observed shifts in LST, NDVI, and NDBI over the years, indicating dynamic changes in land use and urban development impacting temperature. For instance, a decrease in NDVI and an increase in NDBI between 2010 and 2014 correlated with rising LST.
  • City-Specific Patterns: Analysis of seven cities revealed that Luoyang, Sanmenxia, and Xinxiang tend to have more high-temperature connected regions, while Anyang, Zhengzhou, and Kaifeng have fewer. Hebi consistently shows the fewest.

Limitations and Future Work

  • Data Resolution: While MODIS provides broad coverage, its 1 km resolution might limit the analysis of very fine-scale urban features.
  • Climate Influence: While efforts were made to account for climatic conditions, residual influences might still exist.
  • Future Work: The authors suggest that further research could explore the influence of different land-use types on LST in more detail and potentially apply the MTML method to other regions to compare findings. The study also hints that the weakening influence of vegetation and building types on LST at higher temperatures warrants further investigation.

Summary updated

Abstract

Temperature is an important aspect of land–atmosphere studies and plays a key role in urban environmental change. With the continuous development of satellite remote sensing sensors, remote sensing technology has become an important means of obtaining large-scale land surface temperature (LST) data. LST can be calculated from the thermal infrared band data of remote sensing images to analyze changes in temperature and determine its relationship with the surface type. In this study, a multitemporal multilevel (MTML) method for analyzing remotely sensed LST data is presented that analyzes attribute changes and correlations of remotely sensed LST data in different periods and at different temperature levels. First, the LST data were obtained under the same climatic conditions at different times, and the influence of climatic conditions on the LST data was excluded. Threshold superposition analysis was then performed on the temperature data to generate temperature-connected regions of different levels, and a tree structure was constructed. Each node in the tree structure represented a connected region. Finally, the attribute information of different connected regions at different levels was calculated, and the attribute changes and correlations between different times and levels were analyzed. In this study, five MODIS LST datasets from 15 May 2006, 1 May 2010, 7 May 2014, 29 April 2017, and 8 May 2021 in Henan Province of China were obtained, and MTML analysis was carried out. The experimental results showed that a negative correlation exists between temperature and the vegetation index, while a positive correlation exists between temperature and the built-up index. However, with an increase in the temperature level, the correlation between temperature and the surface feature type index decreased. In addition, there were more concentrated high-temperature areas in the northern, central, and western regions of Henan Province and lower temperatures in the eastern and southern regions.
BibTeX
@inproceedings{zhao2022multitemporal,
  author = {Zongze Zhao and Bingke Sun and Gang Cheng and Cheng Wang and Na Yang and Hongtao Wang and Xiaojie Tang},
  title = {{A Multitemporal and Multilevel Land Surface Temperature Regional Attribute Change Analysis in Henan, China, Using MODIS Imagery}},
  booktitle = {Sustainability},
  year = {2022},
  doi = {10.3390/su141610071},
}
Figure