Geometry and adjacency effects in urban land surface temperature retrieval from high-spatial-resolution thermal infrared images
Graphical abstract
Introduction
Land surface temperature (LST) is a physical variable belonging to the set of Essential Climate Variables (ECVs) required by the United Nations Framework Convention on Climate Change (UNFCCC) and the World Climate Organization (WCO) to have an impact on grand societal challenges, including the United Nations Sustainable Development Goals (SDGs) (Bojinski et al., 2014; UN, 2021). With the increase in buildings and impervious surfaces, urbanization has changed the original natural surface coverage (Weng, 2012; Zhu et al., 2019). Urban surfaces with built-up areas are more effective in absorbing energy than horizontal surfaces owing to the complex urban geometry, which reduces the direct reflection to the atmosphere and increases multiple-scattering (Offerle et al., 2003). Urban geometry includes building height, building density, height/width ratio and sky view factor (SVF) (Brown et al., 2001; Grimmond, 2007; Sharmin et al., 2017). These urban characteristics alter the natural surface energy and radiation balances such that the urban land surface temperature (ULST) is relatively high. The phenomenon of urban heat islands (UHIs) then arises, with a significant temperature difference between urban and rural areas (Manoli et al., 2019; Oke, 1982; Voogt and Oke, 2003). The ULST is a critical parameter for studying the UHI effect (Lai et al., 2018; Weng, 2009). With the development of thermal infrared (TIR) remote sensing techniques, the ULST obtained from remotely sensed data provides a synoptic view and an effective measure of the UHI, by observing the temperature differences between urban surfaces and the surrounding rural regions (Bektaş Balçik, 2014; Weng, 2009; Zhou et al., 2018).
Current LST retrieval algorithms always assume that the pixel is flat, uniform, and spatially isothermal (Li et al., 2013). This assumption is used to ensure that the LST is invariant among different bands and does not cause much error for most natural surfaces. However, urban pixels are much more complicated and often consist of various components with different temperatures and emissivities (Voogt, 1995; Zhan et al., 2013). Owing to the three-dimensional structure inside and around the pixel, the multiple-scattering radiance cannot be ignored in the modeling of pixel thermal emissions. Yang et al. (2015b) analyzed the geometry effect on the ULST and specifically demonstrated that such an effect can cause a temperature difference of approximately 2 K, and this difference is increased with the decrease in SVF. The directional anisotropy of the ULST caused by different geometries was analyzed, and the results showed that the effective anisotropy should be considered in ULST retrieval (Lagouarde et al., 2012; Voogt and Oke, 1998; Zhan et al., 2013). The heterogeneity of the urban surface causes a considerable number of mixed pixels owing to the generally coarse spatial resolution of the TIR remote sensing data. The methods for separating the component temperatures through directional TIR data in vegetation and soil areas have been studied; however, these methods are usually only valid for vegetated surfaces (Jia et al., 2003; Ren et al., 2013). Currently, ULST retrieval algorithms that consider the geometric effects and material characteristics are rare, primarily because of the complexity of geometry inside the urban area, which requires further study.
Directional simulations of surface temperatures in urban areas have been studied (Lagouarde et al., 2012; Lagouarde et al., 2010; Wang et al., 2018a; Wang et al., 2020; Wang et al., 2018b) however, these simulation models require many input parameters and are usually difficult to apply when retrieving large-area urban surface temperatures. Hence, the thermal radiative transfer process considering the geometric characteristics within urban areas and its application in remote sensing retrieval are critical for ULST retrieval studies. Some studies started from the process of thermal radiative transmission and used single-band TIR data or split window algorithms to retrieve the ULST (Yang et al., 2015b; Yang et al., 2016). Currently, there is no ULST retrieval algorithm for the complex underlying surface of an urban area that is suitable for three or more TIR channels. However, more infrared band information is helpful in improving the accuracy of LST retrieval (Li et al., 2013; Nie et al., 2020). The temperature emissivity separation (TES) algorithm with stability and good accuracy is widely used in LST retrieval, for example, the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) LST products (AST_08) (Gillespie et al., 1998; Sabol et al., 2009). However, as we previously mentioned, the traditional TES algorithm regards the urban surface as a flat surface, similar to the other LST retrieval algorithms. Yang et al. (2016) compared the ASTER LST products with the ULST with the geometry effect and found that the conventional TES is not suitable for ULST retrieval, because the empirical relationship between the minimum spectral emissivity and the maximum minimum differences was not established for three-dimensional urban surfaces. Thus, it is necessary to establish a new relationship for high-density urban areas.
In this study, the TIR data from the Visible and Infrared Multiple-Spectral imager (VIMS) on the Chinese GaoFen-5 (GF-5) satellite were used to investigate the impact of the SVF on the ground temperature and top of atmosphere (TOA) brightness temperature and the ULST retrieval algorithm development. The GF-5 satellite was launched on May 9, 2018 and contains four TIR bands (8.01–8.39 μm, 8.42–8.83 μm, 10.30–11.30 μm, and 11.40–12.50 μm) with a spatial resolution of 40 m, much finer than ASTER (90 m), Landsat 8 (100 m) and MODIS (1 km). Several papers have reported LST retrieval from VIMS images using the split-window algorithm, the TES algorithm, and a hybrid algorithm combining the two (Chen et al., 2019; Ren et al., 2018; Yang et al., 2018; Ye et al., 2017), but all of them focused on the flat surface, and therefore cannot be directly applied to estimate the ULST. Nevertheless, these studies provided us with some references to develop the specified ULST algorithm. Because the TES algorithm is available for at least three TIR bands and has been used in some TIR sensors, the primary aim of this study is to refine the TES algorithm for urban pixels based on a new thermal radiative transfer model and definitions.
Section snippets
Methodology
An ULST retrieval methodology considering the geometry and adjacency effects is proposed and presented in this section, and the technical flowchart of the methodology is shown in Fig. 1, including four steps:
Step 1: Urban radiative transfer physical model; This is the core and foundation of this study. The urban canopy multiple-scattering thermal radiative transfer (UCM-RT) model is proposed based on the thermal radiative transfer model for the flat surface. The detail will be presented in
The impact of urban SVF on ground temperature and TOA brightness temperature
Based on Eqs. (10), (13), (14), the effective canopy emissivity (), the effective urban canopy temperature (), and the brightness temperature at TOA (BT) were simulated from the variables in the current databases. The impact of urban SVFin and SVFadj on the ground temperature and TOA brightness temperature were first investigated using a simulated dataset.
Study region and data
The core urban region of Beijing, China was selected as the study area for retrieving the ULST using the new UTES algorithm. The study region has a warm and semi-humid continental monsoon climate, with an average annual temperature of 13.5 °C and an average annual precipitation of 511.1 mm. This region is primarily characterized by buildings, roads, vegetated land, and bare land. Land use and land cover data (e.g., asphalt surface, concrete surface, and vegetation) with 1 m resolution were
Application of the high spatial resolution thermal infrared image on ULST retrieval
The large-area pixels in the large-scale remote sensing images smooth out the three-dimensional structure of the pixels and the influence of the thermal radiation of the adjacent pixels. The effect of the adjacent pixels is usually neglected when the pixel size is large enough, for example, 300 m for HJ-1B and 1 km for MODIS (Yang et al., 2015a). However, for high spatial resolution images with strong heterogeneity, the influence of thermal radiation between pixels cannot be ignored (Duan et
Conclusion
An urban canopy multiple-scattering thermal radiative transfer (UCM-RT) model for urban pixels was proposed in this study. The UCM-RT model considers (1) the thermal emissions of the urban pixel surface and the multiple scattering in the urban canopy, (2) the ambient radiance scattered from the surrounding adjacent pixels and their scattering in the urban canopy, and (3) the downward atmospheric radiance reflected by urban pixels. The influence of geometric effects on ground temperature and TOA
Credit author statement
All authors contributed to the discussion and revision of the manuscript.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgements
This work was supported in part by a grant from the National Natural Science Foundation of China (No. 41771369), the Open Research Fund of the National Earth Observation Data Center (No. NODAOP2020001), National Key Research and Development Program (No. 2017YFB0503905-05), National High-Resolution Earth Observation Project of China (No. 11-Y20A32-9001-15/17), and Beijing Nova Program (No. Z171100001117079). The authors thank the China National Space Administration (https://www.cheosgrid.org.cn/
References (66)
- et al.
The ASTER spectral library version 2.0
Remote Sens. Environ.
(2009) - et al.
Multi-level monitoring of three-dimensional building changes for megacities: trajectory, morphology, and landscape
ISPRS J. Photogramm. Remote Sens.
(2020) - et al.
Long-term accuracy assessment of land surface temperatures derived from the advanced along-track scanning radiometer
Remote Sens. Environ.
(2012) - et al.
Validation of collection 6 MODIS land surface temperature product using in situ measurements
Remote Sens. Environ.
(2019) - et al.
Influence of adjacency effect on high-spatial-resolution thermal infrared imagery: implication for radiative transfer simulation and land surface temperature retrieval
Remote Sens. Environ.
(2020) - et al.
Diurnal temperature cycle as observed by thermal infrared and microwave radiometers
Remote Sens. Environ.
(2015) - et al.
Modelling daytime thermal infrared directional anisotropy over Toulouse city centre
Remote Sens. Environ.
(2010) - et al.
Experimental characterization and modelling of the nighttime directional anisotropy of thermal infrared measurements over an urban area: case study of Toulouse (France)
Remote Sens. Environ.
(2012) - et al.
Does quality control matter? Surface urban heat island intensity variations estimated by satellite-derived land surface temperature products
ISPRS J. Photogramm. Remote Sens.
(2018) - et al.
Satellite-derived land surface temperature: current status and perspectives
Remote Sens. Environ.
(2013)