Seasonal Differences in Land Surface Temperature under Different Land Use/Land Cover Types from the Perspective of Different Climate Zones

: The process of urbanization is accelerating, and land surface temperature (LST) is increasing, seriously threatening human health. Therefore, it is crucial to explore the differences in LST of different land use/land cover (LULC) types. Using MOD11A2 and MCD12Q1 data, this study explored the seasonal differences in LST of each LULC type from the perspective of different climate zones. The results showed that the maximum and minimum LSTs during the day were higher than those at night. During the day, the LSTs of urban and built-up and barren lands were higher than those of forests, grasslands, and water bodies; at night, the LSTs of urban and built-up lands decreased but remained high, while barren lands showed a signiﬁcant decrease to LSTs even lower than those of water bodies. In addition, the difference in daytime LST of the LU16 type (barren lands) in different climatic zones was the most obvious and was much higher than that of other LULC types in the middle temperate and south temperate zones, but much lower than those in the middle subtropical and north subtropical zones. This comparison of the LST differences of each LULC type under different climate backgrounds provides an important reference for rational urban planning.


Introduction
In the process of global rapid urbanization, numerous natural features have been transformed into impervious surfaces [1][2][3], resulting in an annual increase in land surface temperature (LST). Extreme high-temperature conditions have frequently occurred in large cities, triggering a series of environmental problems such as the urban heat island effect and extreme climatic events [4][5][6][7], causing serious threats to human physical and mental health [8][9][10]. Therefore, research on the spatiotemporal distribution and mechanism of LST has become a focus of Chinese and foreign scholars [11][12][13].
In LST research, the LST data acquisition method is a high priority. Compared with the air temperature data obtained by meteorological stations, LST is easier to obtain, has larger coverage, and is spatially continuous. With the development of remote sensing technology, LST acquisition methods mainly include remote sensing image inversion and numerical simulation [14][15][16]. The inversion of LST from remote sensing images mainly calculates LST based on the reflectivity of ground objects in the thermal infrared band through the thermal infrared band data in satellite sensors, through single-channel algorithm, split-window algorithm, single-window algorithm, and atmospheric correction methods [17][18][19][20]. At present, there are many remote sensors for LST inversion, such as Landsat series (TM, ETM+, and TIRS) and Advanced Spaceborne Thermal Emission &

Data Sources
This study mainly used MODIS LST product (MOD11A2), MODIS LULC produ (MCD12Q1), digital elevation model (DEM) data, and climate division data. See Table  for details.

Methods
At present, the common thermal infrared remote sensing data mainly include Landsat series (TM, ETM+, and TIRS), Advanced Spaceborne Thermal Emission & Reflection Radiometer (ASTER), Moderate Resolution Imaging Spectroradiometer (MODIS), and Advanced Very High Resolution Radiometer (AVHRR) data [26,[44][45][46][47]. Among them, compared with MODIS data, Landsat series satellites have higher resolution (TM: 120, ETM+: 60, TIRS: 100 m); however, their revisit period is longer (16 d), and only daytime data can be obtained free of charge. ASTER data has a higher resolution (90 m), and the revisit period is also 16 d, but it can obtain day and night data. The Landsat series and ASTER are more suitable for small and medium-scale research; Environment-1 B satellite (HJ-1B) data has a lower resolution (300 m), and the playback period is long (31 d). AVHRR has a lower resolution (1100 m) but a short revisit period, passing the same area twice a day. MODIS data resolution is low (250 m, 500 m, and 1000 m), but it can obtain daily day and night data, which is suitable for large-scale research and is widely used in LST research [48,49]. At present, most studies take a single city as an example [50][51][52][53], which Land 2022, 11, 1122 4 of 17 lacks general laws from a large-scale perspective. Due to the advantages of MODIS data in large-scale research, this paper selected the MOD11A2 data and obtained by the split window algorithm inversion; the error was <1 K in most cases [54][55][56].
This study selected MODIS LST product (MOD11A2) as the LST data of the study area (the error of MOD11A2 was <1 K in most cases [54][55][56]). To study the diurnal and seasonal differences of LST, this paper selected a total of 1288 image data in 2020 to calculate LST, and the MRT tool was used to perform projection and stitching processing, and the missing data were set to null values. Then, ArcGIS 10.4 software (version number: 10.4.1.5686)was used to clip according to the vector boundary of the study area. Finally, the image pixel value of MOD11A2 was converted to degrees Celsius by using ArcGIS 10.4 software (digital numbers [DNs]); the calculation formula is as follows: where DN is the brightness temperature of the MOD11A2 image. A total of 28 images of MODIS 2020 LULC product (MCD12Q1) were selected as the LULC data of the study area (accuracy assessment indicated that the Collection 6 product had an overall accuracy of 73.6% for the primary LCCS layer [57]), and were projected and spliced using MRT tools. Then, resampling to 1 km by ArcGIS 10.4 software and clipping according to the vector boundary of the study area were conducted to obtain the LULC spatial distribution map in China. When calculating the average LST of different LULC types, we first converted the LULC raster images into vector format and then used the zonal statistics tool of ArcGIS 10.4 software for calculations.
In order to further analyze the LST changes, this paper used the raster calculator tool of ArcGIS to calculate the same using the formula: where, LST C represents LST changes, and LST 2020 and LST 2010 are the annual average LSTs in 2020 and 2010, respectively.

LULC Spatial Distribution
From the perspective of the spatial distribution of LULC ( Figure 2), croplands were mainly distributed in the northeastern middle temperate, south temperate, and north subtropical zones. Barren lands were mainly distributed in the south temperate, middle temperate, and northwestern plateau climate zones. Grasslands were mainly distributed in northwestern China. Forests were mainly distributed in the subtropical, tropical, and northeastern temperate zones of southern China. Urban and built-up areas were mainly distributed in the eastern coastal areas. In addition, we calculated the proportion of the area of different LULC types in each climate zone ( Table 2). The results showed that barren land and grasslands were the main land types in the plateau climate zone, accounting for 90.3% of the total area. The middle tropical zone was dominated by woody savannas, savannas, grasslands, and evergreen broadleaf forests, accounting for 80.6% of the total area. In the northern tropics, evergreen broadleaf forests and savannas accounted for 28.4% and 28.5% of the total area, respectively. Woody savannas and savannas accounted for relatively large proportions in the south subtropical and middle subtropical zones, and croplands accounted for the largest proportion in the north subtropical zone, followed by the savannas. The largest proportions of barren land and croplands were in the south temperate zone, accounting for 40.4% and 33.7%, respectively. Grasslands and barren land accounted for the largest proportions in the middle temperate zone (45.6% and 22.2% of the total area, respectively), and woody savannas (66.8%) accounted for the largest proportion in the north temperate zone. subtropical zones, and croplands accounted for the largest proportion in the north subtropical zone, followed by the savannas. The largest proportions of barren land and croplands were in the south temperate zone, accounting for 40.4% and 33.7%, respectively. Grasslands and barren land accounted for the largest proportions in the middle temperate zone (45.6% and 22.2% of the total area, respectively), and woody savannas (66.8%) accounted for the largest proportion in the north temperate zone.

LST Spatial Distribution
The LST calculation results are shown in Figure 3 (validated with metrological station data, the RMSE result was 2.94 • C, and the distribution of meteorological stations is shown in Figure A1). There were obvious differences in the LSTs between the day, night, and seasons in China. Overall, the maximum LSTs during the day were higher than those at night: Annual mean LST: (44. . This was mainly due to the rapid heating of the surface due to the influence of solar radiation during the day; whereas at night, without the effect of solar radiation, the ground dissipated heat outward, so the LST was . This was mainly due to the fact that the maximum daytime LST occurred near the Tarim Basin in the Xinjiang Uygur Autonomous Region, and the solar radiation was strong during the day, causing the temperature to sharply increase. However, the Tarim Basin belongs to the temperate continental climate, which is cold in winter and hot in summer, and the annual and daily temperature ranges are very large. The maximum LST at night occurred in the southeastern coastal area, which belongs to the tropical and subtropical monsoon climate, and is affected by the difference in thermal properties between land and sea. The temperature was the highest in summer and the lowest in winter, and there was little difference between spring and autumn.

Relationship between LULC and LST
As shown in Figure 4 and Table 3, the mean LST of different LULC types also showed significant differences in different seasons and between day and night. For the same LULC type, the annual mean daytime LST was similar to the mean LST in autumn, and the seasonal trend decreased from summer > spring > autumn > winter; e.g., at LU1 it decreased from that in summer (19.47 • C) > spring (16.84 • C) > autumn (14.08 • C) > winter (7.42 • C). The annual mean LST at night was similar to that in spring or autumn, and the seasons showed a trend of summer > spring/autumn > winter; e.g., at LU1 where it decreased from that in summer (9.44 • C) > autumn (6.26 • C) > spring (9.44 • C) > winter (−0.35 • C), which was consistent with the seasonal differences in LST. For different LULCs, there were significant differences in the mean LST of the year, season, day, and night. During the day, the highest LSTs in the year, spring, summer, autumn, and winter occurred in LU14, LU13, LU16, LU13, and LU2 types, respectively, and the minimum LSTs occurred in LU15, LU15, LU15, LU3, and LU3 types, respectively. At night, the highest LSTs in the year, spring, summer, autumn, and winter were in LU2, LU2, LU13, LU2, and LU2 types respectively, and the minimum LSTs were all in the LU15 type. In addition, during the day, urban and built-up, and barren lands had higher LSTs, and forests, grasslands, snow and ice, and water bodies had lower LSTs. The LSTs of urban and built-up lands decreased at night, but remained high in many LULC types, while that in barren land showed a significant decrease (to temperatures even lower than that of water bodies), which was mainly due to the strong solar radiation during the day, and the rapid heating of barren land and built-up areas. At night, with no solar radiation, the built-up area still had a high LST due to human activities and other factors, the specific heat capacity of bare soil was small, and the temperature decreased rapidly. As shown in Figure 4 and Table 3, the mean LST of different LULC types also showed significant differences in different seasons and between day and night. For the same LULC type, the annual mean daytime LST was similar to the mean LST in autumn, and the seasonal trend decreased from summer > spring > autumn > winter; e.g., at LU1 it decreased from that in summer (19.47 °C) > spring (16.84 °C) > autumn (14.08 °C) > winter (7.42 °C). The annual mean LST at night was similar to that in spring or autumn, and the seasons showed a trend of summer > spring/autumn > winter; e.g., at LU1 where it decreased from that in summer (9.44 °C) > autumn (6.26 °C) > spring (9.44 °C) > winter (−0.35 °C), which was consistent with the seasonal differences in LST. For different LULCs, there were significant differences in the mean LST of the year, season, day, and night. During the day, the highest LSTs in the year, spring, summer, autumn, and winter occurred in LU14, LU13, LU16, LU13, and LU2 types, respectively, and the minimum LSTs occurred in LU15, LU15, LU15, LU3, and LU3 types, respectively. At night, the highest LSTs in the year, spring, summer, autumn, and winter were in LU2, LU2, LU13, LU2, and LU2 types respectively, and the minimum LSTs were all in the LU15 type. In addition, during the day, urban and built-up, and barren lands had higher LSTs, and forests, grasslands, snow and ice, and water bodies had lower LSTs. The LSTs of urban and built-up lands decreased at night, but remained high in many LULC types, while that in barren land showed a significant decrease (to temperatures even lower than that of water bodies), which was mainly due to the strong solar radiation during the day, and the rapid heating of barren land and built-up areas. At night, with no solar radiation, the built-up area still had a high LST due to human activities and other factors, the specific heat capacity of bare soil was small, and the temperature decreased rapidly.   To further analyze the LST difference of each LULC type in different climatic zones and ensure that all LULC types were distributed in each climatic zone, this study took five land-use types (LU10, LU12, LU13, LU16, and LU17) as examples to analyze the different LULC types. The LST differences of each LULC type under the climate zone are shown in Figure 5. Overall, the mean LST of the same LULC type (except LU16) under different climate backgrounds (except the plateau climate region) showed a trend of being high in the south and low in the north, which was mainly due to the solar radiation differences in different regions in the north and south. Specifically, in the daytime, the annual mean LSTs of LU10 and LU17 exhibited a pattern of decreasing from that in the middle tropical > north tropical > south subtropical > middle subtropical > north subtropical > south Regardless of day and night, the mean LST of each LULC type was different from that without distinguishing climate zones. When the climatic zones were not distinguished, the mean LST of different LULC types during daytime showed LU13 > LU16 > LU12 > LU10 > LU17, and nighttime showed LU13 > LU17 > LU12 > LU16 > LU10. However, there are significant differences in the statistics of LST differences of LULC types in different climatic zones. Specifically, in the daytime, among these five LULC types (LU10, LU12, LU13, LU16, and LU17), except in the south temperate, middle temperate, and north temperate zones, the remaining climate zones showed that LU12 and LU13 had higher LSTs (in the middle tropical and north temperate zones, LU12 > LU13, the rest were LU13 > LU12), and that LU10, LU16, and LU17 had lower LSTs. The LST of LU16 was the highest in the south temperate and middle temperate zones and the lowest in the middle subtropical and north subtropical zones. At night, LU17 and LU13 had higher LSTs in the plateau climate zone, middle subtropical, north subtropical, southern temperate, and middle temperate zones; whereas LU10, LU12, and LU16 had smaller LSTs. LU13 and LU16 in the middle tropical, north tropical, and southern subtropical zones had larger LSTs (LU16 > LU13), and LU10, LU12, and LU17 had smaller LSTs (LU17 > LU12 > LU10, consistent with the results when no climate zone was distinguished). The LST in the north temperate zone showed a trend of LU17 > LU10 > LU13 > LU12.

Variations in LST for Different LULC Types Changes
In order to study the effect of land use changes on LST, this paper selected LULC (MCD12Q1) and LST (MOD11A2, validated with metrological station data, the RMSE result was 4.14 °C, the distribution of LST and LULC in 2010 was shown in Figure A2) data in 2010 for processing, and obtained LST ( Figure A3) and LULC ( Figure A4) (Still taking LU10, LU12, LU13, LU16 and LU17 as examples) changes in 10 years (2010-2020), and using ArcGIS spatial analysis and zonal statistical tools to calculate the LST changes caused by LULC changes ( Table 4). The results showed that, except for LU12 and LU13

Variations in LST for Different LULC Types Changes
In order to study the effect of land use changes on LST, this paper selected LULC (MCD12Q1) and LST (MOD11A2, validated with metrological station data, the RMSE result was 4.14 • C, the distribution of LST and LULC in 2010 was shown in Figure A2) data in 2010 for processing, and obtained LST ( Figure A3) and LULC ( Figure A4) (Still taking LU10, LU12, LU13, LU16 and LU17 as examples) changes in 10 years (2010-2020), and using ArcGIS spatial analysis and zonal statistical tools to calculate the LST changes caused by LULC changes ( Table 4). The results showed that, except for LU12 and LU13 in the south subtropical zone, the LST showed a decreasing trend in other climate zones even if the types of LULCs did not change. This was because the changes of these five LULC types mainly occur in southern temperate, middle temperate, plateau climate, and northern subtropical zones, which were affected by strong cold air in 2020, resulting in low temperatures (according to the statistics of the China Meteorological Administration, source: http: //www.cma.gov.cn/, access date: 2 July 2022). Meanwhile, affected by the Coronavirus disease (COVID-19) in 2020, LST also showed a decreasing trend [58,59]. When analyzing the response of LST to LULC changes, it was found that, except for the plateau climate zone, the LST changes in the remaining climate zones from LU10, LU12, LU16, and LU17 to LU13 were larger than those from LU13 to LU10, LU12, LU16, and LU17, indicating that even under the influence of cold air, urban and built-up lands still had a certain thermal insulation effect compared with other LULC types. However, the LST changes from LU12, LU13, LU16, and LU17 to LU10 were smaller than those from LU10 to LU12, LU13, LU16, and LU17, indicating that grasslands had a certain cooling effect compared with other LULC types.

Relationship between LULC and LST
Most studies have found that for a city in a certain area, the temperature of vegetation and water bodies is low, and that of buildings is high [60][61][62][63][64]. There are few studies on the impact of different LULC types in different cities on LST [65][66][67], and it has been found that urban parks have different cooling effects under different climate backgrounds [44]. Therefore, this study discussed the difference in the LST of different LULC types under different climatic backgrounds in China and analyzed the warming and cooling effects of each LULC type under different climate backgrounds. For example, we found that the LU16 type had high LST in the south temperate and middle temperate zones, but a low LST in the middle subtropical zone. This finding has important significance as a reference for the rational use of urban land. Meanwhile, we analyzed LST variations for changes in different LULC types and found that even under the influence of cold air, compared with other LULC types, LU10 had lower LST and LU13 had higher LST. In addition, we found that extreme climate had a greater impact on LST.

Seasonal LST Variations of Different LULC Types
LST also shows seasonal differences under different LULCs [68,69]. For example, Yang et al. [27] analyzed the relationship between LCZ and LST and found that the mean LST in barren land was lower than most building types in summer but higher than most building types in winter. To further analyze the seasonal difference of the mean LST of each LULC type under different climate zones, in this study we calculated the seasonal difference of LST of LULC types under different climate zones ( Figure 6). The results showed that, for the same LULC type (except LU16), during the daytime in spring and summer, the mean LST of each LULC type first decreased, then increased, and then decreased again with the increase of the latitude of the climate zones (except the plateau climate zone). However, during the daytime in autumn and winter, and at night in all seasons, the mean LST of each LULC type showed a decreasing trend with the increase of the latitude of the climate zones (except the plateau climate zone), which was consistent with the annual mean LST results. For the same climate zone, there were differences in the mean LST of each LULC type in different seasons. Taking the north temperate zone as an example, in daytime in spring and summer, the mean LST showed a trend of LU12 > LU13 > LU10 > LU17; and in daytime in autumn, the mean LST showed a trend of LU10 > LU12 > LU13 > LU17. In daytime in winter, the mean LST showed a trend of LU17 > LU10 > LU13 > LU12, which was mainly due to the location of the north temperate zone in the northernmost part of China. This area is covered with a large amount of snow and ice in winter, resulting in a decrease in the LST to a lower temperature than that of water bodies. In the context of the frequent occurrence of the global urban heat island effect, the LULC types with lower LSTs in each climatic zone can be identified by comparing the difference in mean LSTs in different LULC types in different climatic zones. Then, the proportions of the areas of these LULC types can be increased to alleviate the LST. This study shows that the proportions of water bodies and grassland areas in each climate zone can be increased to reduce the LST.

Limitations
Based on the LST and LULC data of MODIS, this paper analyzed the mean LST difference of each LULC type under different climate zones and found that the LST of each

Limitations
Based on the LST and LULC data of MODIS, this paper analyzed the mean LST difference of each LULC type under different climate zones and found that the LST of each LULC type was significantly different under different climate zones. There were also differences in LST under the same climate zone, and the LST can be effectively decreased by increasing the proportion of LULC types with lower LSTs. However, the relationship between LULC and LST was complex. This study analyzed the mean LST of different LULC types but did not analyze the landscape pattern of LULC and spatial autocorrelation of LST. In the future, the difference in LST caused by the spatial agglomeration characteristics of different LULC types should be analyzed. In addition, the division of climate zones in this study was mainly based on temperature, and in future studies, factors such as precipitation should be considered, and the climate zones may be further divided into arid and humid regions.

Conclusions
Using MOD11A2 and MCD12Q1 in 2020 as data sources, this study explored the seasonal differences in LST of each LULC type from the perspective of different climate zones in China. The main conclusions were as follows: (1) Croplands were mainly distributed in the northeast middle temperate, south temperate, and north subtropical zones. Barren lands were mainly distributed in the south temperate, middle temperate, and northwest plateau climate zones. Grasslands were mainly distributed in northwest China. Forests were mainly distributed in southern China subtropical, tropical, and northeastern temperate zones. Urban and built-up areas were mainly distributed in eastern coastal areas. (2) Overall, the maximum and minimum LSTs during the day were higher than those at night. LST also showed seasonal differences. During the daytime, the maximum LST decreased from that in summer (63. forests, grasslands, snow and ice, and water bodies had lower LST. At nighttime, LST decreased in urban and built-up lands; however, the LST remained high at nighttime in many LULC types. Barren lands showed a significant decrease in LST at nighttime, i.e., to temperatures even lower than those of water bodies.
These comparisons of the differences in LST of each LULC type under different climate zones and identification of the LULC types with lower LSTs provide an important reference for alleviating the increasing urban heat island effect, temperature improvement, and urban planning.