Does urbanization increase diurnal land surface temperature variation? Evidence and implications

Highlights

• Satellite images were used for landscape changes assessment in Taipei.

• Polices on urban development can quickly affect landscape changes in Taipei.

• Urbanization increases urban diurnal land surface temperature variation.

• Urbanization results in a greater increase in urban heat absorption than in thermal inertia.

Abstract

The diurnal land surface temperature (LST) variation is a primary characteristic of the effects of urbanization. However, no study to date has focused on changes in diurnal LST variation in urban environments. This paper investigates the effects of urbanization on landscape pattern and diurnal LST variation of Taipei City, using MODIS thermal images and SPOT multispectral remote sensing images over the 1994–2010 period. Supervised land-cover classifications were conducted to investigate decadal land-cover changes within the study area. A remote-sensing-based urbanization index was adopted as a quantitative measure of urbanization-induced changes in landscape patterns. Diurnal LST variations were assessed using MODIS Aqua satellite images. We found that the diurnal LST variation increases with the urbanization index, with the adverse effects of urbanization on the diurnal LST variation being more substantial in the earlier stages of urbanization. Increasing diurnal LST variation due to urbanization implies that urbanization is likely to result in a greater increase in urban heat absorption than in thermal inertia. This research provides insightful supporting evidence that, in mitigating UHI effects, measures that can reduce urban heat absorption such as urban parks, community green spaces, green roofs and cool (or permeable) pavements should be given higher priority.

Introduction

Urbanization typically results in built-up environments containing large areas of impervious surfaces with increased thermal inertia. Urbanization impacts stormwater runoff, air quality and carbon concentration (American Forests, 2005), net primary productivity and soil respiration rates (Kaye, Mcculley, & Burke, 2005), and ambient air temperatures (Cheng, Su, Kuo, Hung, & Chiang, 2008). Researchers have developed different landscape metrics for quantitatively evaluating the effects of urbanization including changes in land-cover patterns, impacts on urban ecosystems, and deterioration of urban air quality (O’Neill et al., 1988; Plotnic, Gardner, & O’Neill, 1993; Riitters et al., 1995; Riitters, O’Neill, Wickham, & Jones, 1996). Several studies have also demonstrated that such landscape metrics can be derived using remote sensing images (Deng, Wang, Hong, & Qi, 2009; Griffith, Stehman, Sohl, & Loveland, 2003; Herold, Scepan, & Clarke, 2002). Most of such landscape metrics have been used to characterize changes in land-cover patterns and their effects on environmental and ecological regularities within a region. Frohn and Hao (2006) suggested that spatial resolution should be considered when applying metrics to quantify the spatial patterns of landscapes. Li, Zhou, and Ouyang (2013) showed that LST was negatively correlated with the percentage of green space cover in Beijing; however, the relationship varied depending on spatial resolution. Sun, Wu, Lv, Yao, and Wei (2013) demonstrated that landscape-level metrics could identify different types of urban growth in the city of Guangzhou, China.

Although numerous landscape metrics have been proposed, only a few of such metrics are suited to quantify the degree of urbanization. The most apparent characteristic of urbanization is the increase of built-up land cover. Therefore, the percentage of apartment buildings in total construction housing (Kouichi, 1987), road density and percentage of built-up area (Lee, Ding, Hsu, & Geng, 2004), and the proportion of impervious surfaces (Yuan & Bauer, 2007) were conveniently used as urbanization indices. However, modern urban planning often requires the establishment of large green spaces to alleviate the adverse effects of urbanization such as an increase in the ambient air temperature and deterioration of urban air quality. To reflect the effects of urban green spaces, a remote-sensing-based urbanization index that integrates the normalized difference vegetation index (NDVI) and coverage ratio of built-up land cover was developed and used to assess the degree of urbanization in different Asian cities (Guo, Wang, Cheng, & Shu, 2012; Hung, Chen, & Cheng, 2010).

One of the major concerns related to urbanization is its effect on the urban thermal environment. Atmospheric heat islands are normally measured by in situ air temperature sensors; by contrast, a surface urban heat island (SUHI) refers to the excess warmth of urban areas compared with the non-urbanized surroundings, and it is measured using LST levels observed by thermal infrared remote sensing (Yuan & Bauer, 2007). Voogt and Oke (2003) concluded that NDVI-based measures are favorable for describing the differences between urban and rural surface properties. Oke, Johnson, Steyn, and Watson (1991) addressed the roles of surface geometry and surface thermal properties in the creation of SUHIs and the importance of assessing these parameters in both urban and rural environments.

The intensity of the UHI effect is often measured as the difference between the temperature found in the center of an urban area and the background rural temperature (Atkinson, 2003: Hung, Uchihama, Ochi, & Yasuoka, 2006). The thermal inertia estimated using remote sensing thermal images has also been used in studies related to UHI effects (Chen, Du, & Dong, 2008; Nasipuria, Majumdar, & Mitra, 2006). The estimation of the thermal inertia of an earth surface object is dependent on two factors: the albedo and diurnal LST variation (Nasipuria et al., 2006). The apparent thermal inertia (ATI) is a measure of the proportion of radiation absorbed, divided by the subsequent increase in surface temperature ($\Delta T$):

$$ATI=\frac{NC(1-\alpha )}{\Delta T}$$

where N is the scaling factor, C is a constant determined by the latitude and solar declination angle, and $\alpha$ is the apparent albedo (Nasipuria et al., 2006). Therefore, the diurnal temperature variation in an object is dependent on its absorbed energy [i.e., $NC(1-\alpha )$] and apparent thermal inertia. Objects possessing low thermal inertia and albedo (and, by implication, high absorptance) display considerable diurnal temperature variation. Both thermal inertia and albedo vary depending on the type of land cover, and therefore changes in land cover caused by urbanization have a direct impact on the diurnal temperature variation in an urban environment. Hence, the variation in diurnal LST is a primary characteristic of the effects of urbanization.

In contrast to the UHI effect, few studies have focused on diurnal temperature variations in urban environments (Argüeso, Evans, Fita, & Bormann, 2014; Georgescu, Moustaoui, Mahalov, & Dudhia, 2011; Qiao, Tian, & Xiao, 2013; Wouters, De Ridder, Demuzere, Lauwaet, & van Lipzig, 2013). Patterns of land-cover change, including the composition of vegetation, water, and built-up, are complicated, and characterizing how they are linked with changes in diurnal LST variation is difficult. For example, the instantaneous field of view of most space-borne thermal sensors covers a substantial mix of surface elements. Therefore, the effect of urbanization on diurnal LST variation is scale-dependent and involves a certain degree of uncertainty because of the combined effect of different land-cover types within the spatial scale of interest. In addition, most studies on urban temperature environments have focused on air temperatures or UHI effects (Argüeso et al., 2014, Qiao et al., 2013; Qu, Wan, & Hao, 2014). Nevertheless, it is still unclear how urbanization-induced landscape changes affect diurnal LST variation and what such changes in urban diurnal LST variation imply. To the best of our knowledge, no study to date has focused on changes in diurnal LST variation in urban environments. Therefore, the objectives of the current study were threefold: (1) to quantitatively investigate landscape changes in the study area by using a remote-sensing-based urbanization index; (2) to assess the effect of urbanization on diurnal LST variation and investigate its implications; and (3) to evaluate the effectiveness of the urbanization index in characterizing such effects, taking into account possible uncertainties.