Influence of Land Use and Land Cover Change on Land surface temperature

: The rapid urbanization process has recently led to significant land use and land cover (LULC) changes, thereby affecting the climate and the environment. The purpose of this study is to analyze the LULC changes in Hefei City, Anhui Province, and their relationship with land surface temperature (LST). To achieve this goal, multitemporal Landsat data were used to monitor the LULC and LST between 2005 and 2015. The study also used correlation analysis to analyze the relationship between LST, LULC, and other spectral indices (NDVI, NDBI, and NDWI). The results show that the built-up land has expanded significantly, transforming from 488.26 km 2 in 2005 to 575.64 km 2 in 2015. It further shows that the mean LST in Hefei city has increased from 284.0 K in 2005 to 285.86 K in 2015. The results also indicate that there is a positive correlation between LST and NDVI and NDBI, while there is a negative correlation between LST and NDWI. This means that urban expansion and reduced water bodies will lead to an increase in LST.


Introduction
With the development of the economy, the urban population has grown rapidly, which has promoted rapid urbanization. Land use and land cover (LULC) patterns have undergone tremendous changes, as well as changes in various biophysical climatic conditions, especially urban land surface temperature (LST) [1,2]. The conversion of LULC (such as wetlands, vegetation, and agricultural areas) to impervious land can greatly affect LST [3]. LST is regularly measured from satellite sensors with medium spatial scale and high temporal resolution (such as Landsat). Generally, the LST data derived from the satellite's thermal infrared (TIR) band is a key variable to understand the impact of changes in LULC caused by urbanization [4]. Spectral indices from remote sensing data can usually provide a comprehensive understanding of the relationship between LST and LULC conditions [5,6]. The most common satellite-derived indicators for estimating the temporal and spatial changes of LST are the normalized difference vegetation index (NDVI), the normalized difference built-up index (NDBI), and the normalized difference water index (NDWI) [7]. Previous studies have analyzed the different relationships between LULC, LST, NDVI, NDBI, and NDWI [7][8][9][10][11][12][13]. These results are mainly attributable to the growth and expansion of cities brought about by urbanization and socioeconomic development, which affect land use and regional climate change.
This study aims to monitor and analyze the spatiotemporal trends of LULC changes and establish their relationship with the LST changes of Hefei city, China. More specifically, the study seeks to: (a) map and analyze the various changes in the LULC pattern of Hefei city from 2005 to 2015; (b) study the distribution of LST, NDVI, NDBI, and NDWI of the Hefei city; (c) analyze the relationship of LST and indices (NDVI, NDBI, and NDWI).

Study Area
Hefei city is the largest and capital city of Anhui Province, China, which comprises four urban districts, one countylevel city, and four counties. In this study, four urban districts were chosen as the study area, which covers an area of 1308 km 2 and is situated between 116°51'-117°26'E and 31°38'-32°4'N. (Figure 1).

Research Methodology
The methodological flow chart illustrated in Figure 2 summarizes the several procedures used in this study.

LULC Classification
The maximum likelihood classification (MLC) method is one of the most widely used LULC classification methods due to its high classification accuracy [14,15]. In this study, the MLC method was used to divide the LULC in 2005, 2010, and 2015 into four categories, including built-up land, agricultural land, forest land, and water bodies.

Accuracy Assessment
The standard thematic accuracy assessment was adopted using overall accuracy (OA), and Kappa hat (KH) coefficient based on error matrix between classified LULC data and ground reference information data [16]. For thematic accuracy assessment, the number of samples was estimated based on the multinomial distribution [16,17], and sample points were allocated for thematic accuracy assessment using a stratified random sampling technique.

NDVI, NDBI, and NDWI Estimation
The NDVI, NDBI, and NDWI are the most used indicators in environmental studies, which were applied to extracting vegetation conditions, impervious surfaces, and water bodies from remotely sensed data, respectively. These indices were estimated using Equation (1) to Equation (3) [7].

LST Retrieval
The TIR bands from Landsat 5/7/8 were used to retrieve LST through a single channel (SC) algorithm. The process to retrieve LST is discussed below.
(1) Brightness temperature calculation DNs of the TIR band of Landsat images were converted to spectral radiance using the radiance scaling factors provided in the metadata file. Then the brightness temperature (BT), which is the effective temperature viewed by the satellite under an assumption of unity emissivity, can be converted from spectral radiance with the following formula [1]: where, � is the spectral radiance (W/ (m 2 ·sr·μm)), � is the radiance multiplicative scaling factor, � is the radiance additive scaling factor, ��� is the pixel value in DN, is BT in Kelvin (K), K 1 , and K 2 are the thermal conversion constants.
(2) Emissivity extraction In this study, the NDVI threshold method [18] were adopted to estimate the emissivity of different land surfaces as follows: where, ε is the band emissivity, ε v and ε s are the emissivities of vegetation and soil, respectively, P v is the proportion of vegetation, NDVI v and NDVI s are the NDVI for a fully vegetated pixel and a soil one, respectively, and F' is a geometrical factor ranging between 0 and 1.
(3) Water vapor content extraction MODIS Reprojection Tool and MRTSwath tool were used for reprojection and conversion of MODIS product to DN values of water vapor content. To retrieve the real value of water vapor content, the DN values were divided by the scaling factor (scaling factor = 1000).
(4) Land surface temperature derivation The SC algorithm provided by Jiménez-Muñoz and Sobrino [19] was chosen for LST retrieval from Landsat data using the following Equation (7): where, T s is the land surface temperature, ε is the surface emissivity, T sen is the at-sensor BT; b γ = c 2 /λ; and φ 1 , φ 2 , and φ 3 are the atmospheric functions which can be obtained as a function of the total atmospheric water vapor content (w) [20][21][22].

LULC Classification and Change Detection
The classified LULC maps of Hefei city are presented in Figure 3 and quantified in Table 1   The result reveals that the built-up land has expanded from 488.26 km 2 (37.31%) in 2005 to 575.64 km 2 (43.99%) in 2015 as the most one among the four LULC types in Hefei city. However, the agricultural land and forest land decreased from 692.84 km 2 (52.95%) and 21.88 km 2 (1.67%) in 2005 to 625.01 km 2 (47.76%) and 8.14 km 2 (0.62%) in 2015, respectively. Meanwhile, the water bodies remained basically unchanged. The gradual increase in built-up land and decrease in agricultural land can be attributed to urban growth. It can be seen from the LULC classification and change detection that the urbanization process of Hefei may be an important factor in the transition from the natural surface to the built-up land.

Accuracy Assessment of LULC Classification
In this study, 443 sample points based on the multinomial distribution with the desired precision of 5% and a level of confidence of 85% were applied to access thematic accuracy assessment. Table 2 shows the error matrix and accuracy of LULC classification in 2005, 2010, and 2015. The OA and KH of the three maps were above 87%, and 0.8, which signifies a reliable LULC classification [23].

LST Distribution
The The development of urban land has led to an increase in land surface temperature in Hefei, which is consistent with previous studies [24].

NDVI, NDBI, and NDWI Distribution and Their Relationship with LST
The distribution of NDVI of Hefei city in 2005, 2010, and 2015, is presented in Figure 5. To examine the relationship between NDVI, NDBI, NDWI, and LST, the correlation was calculated with the Band Collection Statistics tool in ArcGIS and the results are displayed in Table 3. It shows a positive correlation between the values of LST and NDVI and NDBI, and a negative correlation between LST and NDWI. The results indicate that lower LST values corresponded to lower NDVI (e.g., water bodies), while higher LST values corresponded to higher NDVI (e.g., agricultural land). Furthermore, the results also manifest that lower LST values corresponded to lower NDBI (e.g., water bodies), while higher LST values corresponded to higher NDBI (e.g., built-up land). In contrast, the results imply that lower LST values corresponded to higher NDWI (e.g., water bodies), while higher LST values corresponded to lower NDWI (e.g., built-up land). The positive correlation between LST and NDVI, the positive correlation between LST and NDBI, and the negative correlation between LST and NDWI conform to earlier studies. Yuan Chi et al. [8] analyzed relationships among LST and NDVI in the Yellow River Delta, China, the positive correlations were observed in all seasons of 2016-2017. Subhanil Guha et al. [11] focuses on the study of land surface temperature with NDBI using Landsat 8 OLI and TIRS data in Florence and Naples city, and LST performs a strong correlation with NDBI (positive). Bayes Ahmed et al. [25] analyzed LULC and LST in Dhaka, Bangladesh, and the results of multiple correlation and regression analyses indicate that LST presents a negative correlation with NDWI.

Conclusions
The multi-temporal Landsat satellite data was used to analyze the spatiotemporal impact of LULC changes on LST in Hefei city from 2005 to 2015 in this study. The LULC change analysis indicates a rapid urban growth in Hefei city with a considerable built-up land increase from 488.26 km2 in 2005 to 575.64 km2 in 2015. The LST analysis result revealed that the mean LST increased from 284.0 K in 2005 to 285.86 K in 2015. The correlation between LST and land-use indices was also analyzed in this study. The study suggests a positive relationship between LST and NDBI while establishing a negative relationship between LST and NDWI during the different periods. This implies that higher LST is experienced along with a decline in water bodies and an increase in built-up land. The findings of this study suggest that LULC changes in Hefei city have substantially influenced LST. Consequently, the mitigation plan to reduce land surface temperature due to urbanization should be prepared by the corresponding agencies.