The north slope of Tianshan Mountains, lying between 42°55’N-46°13’N and 83°25’E- 88°58’E, is located in the northern part of the middle Tianshan Mountains and the southern margin of the Junggar Basin. It is adjacent to the Tianshan Mountains area to the south and the Gurbantunggut Desert to the north. The alluvial flat central principle distributes the broad oasis agricultural area and 11 counties and cities (Figure 1). The study area is under a typical temperate continental climate. The terrain is high in the southeast and low in the northwest, and the landscape types from south to north are the medium-alpine mountain belt, foreland mountain belt, oasis, oasis-desert transitional zone and the desert (Luo et al., 2003). There are obvious interactive responses between vegetation cover types and climate change (Sun et al., 2014). Due to the central alluvial plain being suitable for agricultural development, especially since the 21st century, the cultivated land area has been continuously increasing and the urbanization degree has been continuously enhanced, the north slope of Tianshan Mountains has become one of the most conspicuous areas affected by human activities in Xinjiang (Yin et al., 2020).

The data used in this article mainly include MOD11C3 ground temperature data, Land Use (LU) data, Normalized Differential Vegetation Index (NDVI) data, DMSP/OLS and NPP-VIIRS nighttime light (NTL) Data and Population Distribution (PD) data, etc. (Table 1). MOD11C3 was utilized to explore the characteristics of spatio-temporal differentiation of ground temperature. LU, NTL and PD data were applied to characterize HAI and analyze its spatio-temporal distribution characteristics. NDVI data incarnate the changes in vegetation coverage on the underlying surface.

From 2000 to 2018, the intensity of human activities in the north slope of Tianshan Mountains was low and showed a gradual upward trend, with an average annual HAI of 0.11 and a rate of change of 0.0024·a^-1 (Figure 2). Between 2001-2003, the population of the study area grew slightly, while the construction land and cultivated land area were in the expansion stage. Under the joint influence of the three factors, the HAI in the study area showed a slight increase, and then entered a platform period lasting nearly 10 years. Then, the population grew rapidly from 2012 to 2015, and the expansion of the construction land and cultivated land area resulted in significant aggrandizement in HAI. It can be seen that HAI is predominantly affected by population and construction land, with a lag period of 1-2 years, while the impact of other land types is relatively small (Figure 3b), which also incarnates the one-sidedness of use LU only in characterizing HAI and the rationality of PD in explaining HAI changes.

The spatial differentiation of HAI in the study area is high in the middle and low in the north and south (Figure 4a). The southern mountainous area, northern desert area and low-mountain steppe zone are areas with low human activity and extremely low intensity, accounting for 76.2% of the total area. The cultivated land is predominantly medium-low intensity, and the high-intensity and medium-high intensity areas are concentrated in the urban center as the HAI in Urumqi, Shihezi and Karamay are all above 0.73. From the perspective of spatial distribution, the spatial distribution of HAI in counties and cities con-forms to the circle structure characteristics of urban development (Figure 4b), which shows that HAI and its change rate gradually attenuate from the central city to the outer edge of the city and the surrounding cultivated land, grassland and desert, with Urumqi as the most typical. However, the high-value areas of HAI variability outside the central urban area are all rural residential areas scattered in cultivated land. Judging from the interannual variation of the area proportion of HAI levels, the area proportion of extremely-low intensity levels (unused land) of human activities has dropped significantly from 49.48% in 2000 to 6.93% in 2018, and the areas of the remaining levels are escalating, with the medium-low intensity level (primarily cultivated land, accounting for 41.96%) being the most pronounced. It serves to show that the scope of human activities in the study area is constantly expanding and the scope of radiation influence on the surrounding areas is continually enhancing, and the variation of HAI follows the law of distance attenuation.

From 2000 to 2018, the HAI on the north slope of Tianshan Mountains has been relatively low, with an annual change rate of only 0.0024·a^-1, showing a slight and slow growth trend. Compared with previous research results based on different scales, the growth rate is very small, such as 0.038·a^-1 in the country (Xu et al., 2016) and 0.061·a^-1 in Hainan province (Liu et al., 2020b), which reflects from the side that the relatively lagging socio-economic development and the slow growth of the study area are due to the historical reasons and restrictions on location, geography, etc. The positive growth area of HAI in the study area accounts for 99.33%, and it is rapidly attenuating from the central urban area to the periphery. HAI around scattered rural settlements increased obviously, while other land types have not changed much, which is chiefly due to the centralization of population, the augment of urban-industrial investment (Zhong et al., 2008), the expansion of residential buildings, industrial and mining land and road construction (Figure 3b), and the continuous transformation of marginal land to construction land (Figures 1 and 4b), which embodies the expansion and development trend of urban scale.

During the study period, the annual average LST of the north slope of Tianshan Mountains was 7.18℃, and the coldest and hottest years were 2003 (6.03℃) and 2015 (8.21℃) respectively. The four seasons of LST in 2006 and 2008 were all higher than the multi-year average. Compared with 2000, the LST in the study area increased by 0.82 ℃ in 2019, with an average annual change rate of 0.02 ℃•a^-1, showing a salient fluctuating upward trend (Figure 5a) and was much stronger than the global warming (0.0086±0.006 ℃·a^-1) (Yun, 2019). The spring, summer, and winter have shown a warming trend in the past 20 years, with positive anomaly years being the majority. The significant warming in spring (0.068 ℃•a^-1) made an absolute contribution to the overall warming of the study area (Figure 5a). However, the negative anomaly of LST in spring was larger, and extremely low temperatures occurred frequently in many areas (Mao et al., 2016). LST was relatively stable in summer but fluctuated greatly in winter, and the warming in winter showed an upswing trend over the years.

The LST in the study area was lower in the south and higher in the north, with prominent spatial differentiation. According to the LST statistics of the longitudinal in the central part of the study area (Figure 6b), the LST in the southern mountainous area decreased conspicuously with the increase in altitude, and the correlation coefficient reached -0.84, showing a significant negative correlation, which intuitively reflected the vertical zonality of LST. The LST in the strip-shaped urban area in the piedmont alluvial plain is signally higher than that of the natural vegetation and arable land on both sides, up to 15.23 ℃, and there is an obvious “heat island effect”, mainly due to the artificial heat emission from industry, traffic and life, the physical characteristics of small specific heat capacity, high heat storage and low loss of urban impervious surface, as well as the “insulation” effect of air pollutants. The desert area on the southern margin of the Junggar Basin has a relatively high LST due to its own small specific heat capacity and lack of vegetation evapotranspiration, humidification, and shading. It serves to show that the elevation and underlying surface break the gradual warming characteristic of LST in the study area from south to north along the latitudinal direction, and human activities, as a momentous “force”, play a crucial role in the process of ground temperature change by changing land-use patterns.

LST on the north slope of Tianshan Mountains increased greatly in the past 20 years, and the rate of change was also low in the south and high in the north, with pronounced spatial differences (Figure 6c). The area of the heating zone is about 4.6 times that of the cooling zone, and the desert areas with bare surfaces and sparse vegetation, as well as the urban and surrounding areas predominantly showed warming. The desert area on the southern margin of the Junggar Basin is the region with the most conspicuous warming, with a change rate of as high as 0.09 ℃·a^-1. The densely populated urban and rural construction land is mostly impervious surface. Due to its higher heat absorption rate and smaller specific heat capacity, the surface heats up rapidly, combined with the thermal insulation effect of production and domestic waste heat emissions and air pollutants, it has become a regional heat island and the dominant temperature rise Area. For example, the surrounding countries and cities of Urumqi, the central urban areas of Hutubi and Manas-Shihezi present obvious hot areas, and the range of LST changes in some areas exceed 0.1℃·a^-1. However, there is a conspicuous cold zone in the Kuytun-Usu area, the statistics found that the NDVI of Kuytun City has risen from 0.29 to 0.62, and a good deal of natural grasslands in the city have been converted into cultivated land. Although it was still vegetation under the manual intervention of wasteland reclamation and agricultural planting, the growth was more vigorous and the coverage increased substantially. The accompanying decrease in surface albedo and increase in soil moisture should be one of the reasons for the cooling of the area, which also confirms that the changes in the physical and biological properties of the underlying surface caused by vegetation coverage and human activities are intimately connected to the changes of LST. It is particularly noteworthy that some high-altitude mountains in the south of Usu and Fukang city present obvious warming trends, and the rise of LST will aggravate the melting of glaciers. Meanwhile, the increase of surrounding bare rock, the aggravation of heat absorption and the reduction of the reflectivity of the underlying surface will lead to the rise of LST (Huang et al., 2016). Just as Wang et al. (2017) concluded that the change rate of glacial reserves in Glacier No.1 at the head of the Urumqi River was as high as -291.82 mm·a^-1 from 1959 to 2010, which is a shred of good evidence for the conspicuous rise of LST in the region. The mounting risk of accelerated glacier melting will have a profound impact on the ecological environment, industrial and agricultural development and social economy in the study area (Zhang et al., 2019).

This article centralized the principal human activity areas in the central part of the study area (Figure 1) to explore the spatio-temporal relationship between HAI and LST, as the southern mountainous and northern desert areas are inaccessible. Spatial correlation analysis showed that HAI and LST were significantly positively correlated, showing a strong distribution in the east and a weak in the west (Figure 7). The area of positive correlation accounted for 94.6%, primarily cultivated land and construction land, of which 73.7% was a moderately strong and strong positive correlation, with Urumqi (0.78) and Wujiaqu (0.84) being the most prominent. The HAI in Fukang City and the central part of Changji City was noteworthily enhanced (Figure 4), appearing a strong positive correlation. HAI and LST generally show a strong positive correlation in the central and northern regions due to the combined effects of the large increase of cultivated land, the expansion of internal rural settlements, and the substantial upswing of urban construction land in the center of each county, and the increase of NTL. However, in areas with a strong positive correlation, except for the extension of unused land in the north of Changji and the small expansion of rural settlements in Wujiaqu, the other land-use types did not get marked variation (Figure 7b). The increase in LST should be positively correlated with the slight growth in population (Figure 7c) and the enhancement in NTL brightness (Figure 7d). In the south large grassland area, the population has not changed much over the years, but the area of internal cultivated land and rural settlements around water body has expanded, NTL has generally enhanced, and HAI and LST are also positively correlated. In addition, cultivated land expanded obviously in the south of Urumqi, the rural residential area and urban construction land increased, the population and NTL also augmented greatly, and the correlation between HAI and LST showed a strengthening trend.

There are few areas with a negative correlation between HAI and LST, which are only located in small areas of cultivated land and grassland around Usu and Kuytun. Under the condition of slight enhancement of HAI, LST declines prominently, presenting a conspicuous negative correlation (Figure 7). Combined with the manifestations of human activities in different land-use types, the comparison found that NDVI in this region has been continuously escalating over the years, and the cooling effect of vegetation is the dominant factor leading to its negative correlation. Take Kuytun City as an example (Figure 8), 42.78% of the sparse grassland area has been transformed into cultivated land with higher coverage (Figures 8e and 8f) for nearly 20 years, although the land type has changed, it has only changed from natural vegetation to artificial vegetation, its essence has not changed. Moreover, the NTL intensity is slightly enhanced when the population is basically unchanged, so the area HAI is only slightly increased. The NDVI of the region escalated from 0.30 to 0.62 (Figures 8g and 8h), the LST reduced by 4.66 ℃, and the NDVI (0.73) of cultivated land was 0.25 higher than that of grassland in the same year, while the LST was 0.48 ℃ lower than that of grassland. It serves to show that the shading, heat absorption and humidification effects of corps have gradually become prominent, thus saliently reducing LST. In general, human activities based on land reclamation and agricultural planting make crops under artificial control grow more vigorously and have higher coverage than natural vegetation, and cooperating with agricultural activities such as irrigation has intensified the cooling effect. By analogy, human greening and maintenance activities associated with vegetation, such as agriculture and forestry planting and cultivation, urban greening and afforestation, can pronouncedly inhibit the rise of LST.

In conclusion, the HAI and LST in the human activity area on the north slope of the Tianshan Mountains are strikingly correlated and show an escalating trend. Their spatial differentiation is comprehensively affected by the human activity ranges, manifestations, land-use changes and other factors, of which human activities lead to changes in underlying surface properties such as the continuous growth of population, the expansion of construction land to cultivated land, grassland and reclamation are the principal reasons for LST changes in the study area, and the change of PD and NTL can support a rational explanation of LST’s spatio-temporal changes.

For the definition and characterization of HAI, currently, it is mostly based on the single factor of land use (Xu et al., 2016; Han et al., 2017; Li et al., 2018) or combined with regional economic data, landscape types, population density and other data (Liu et al., 2018).Scholars such as Liu et al. (2020b) and Xu et al. (2016) proposed the method of the equivalent factor of construction land area to represent HAI; Han et al. (2017) calculated HAI comprehensively based on the influence and weight of different land-use types; Rong et al. (2017) obtained HAI through the artificial influence parameters of regional landscape types; Xue (2021) reflected the change of HAI with population density. Whether the changes in land-use type/landscape type or population density can only incarnate a certain part of the influence of human activities and cannot reveal the panoramic image of human activities. Among the many existing HAI studies, the multi-stage data set of human activity intensity in the Yangtze River Economic Belt (Li et al., 2018) developed on the basis of China’s national-scale land use data is represented by adopting the comprehensive evaluation scheme of ecosystem human disturbance index (hereinafter referred to as the comprehensive evaluation scheme). To compare the rationality and applicability of different schemes, the HAI of the north slope of Tianshan Mountains was calculated according to the comprehensive evaluation scheme and normalized separately with the results of this study (Figure 9). The area of grassland (39.8%), unused land (33.7%) and cultivated land (20.7%) in the study area are much larger than the construction land (2.5%). Through comparative analysis, impute to the intense influence of the area proportion of each land-use type on the comprehensive evaluation scheme (Zhao et al., 2015), under the condition that the assignment level of each category is certain, cultivated land and grassland areas with a low level of assignment but a large area have a high HAI and low discrimination (Figure 9b), while the HAI for construction land with the most concentrated and intense human activities is low. While the results of this paper can not only reflect the macroscopic distribution pattern of HAI but also refine the spatial differences of HAI within different land types. As the author of the data set said, “The comprehensive evaluation scheme fails to fully reflect the multi-faceted disturbance of the ecosystem by humans, and is a conservative estimate of HAI” (Li et al., 2018). It is extremely widespread in arid, semi-arid areas and vast underdeveloped areas that the territory is vast and sparsely populated and the area of construction land is relatively small. In contrast, this study combined LU data with PD data and NTL data to consider the dynamics, status, process and results of ecological environment disturbance more thoroughly. The results in towns and surrounding areas were more objective, and the comparison of the disturbance degree of human activities among different land types was more reasonable. Therefore, the HAI representation system constructed in this study is relatively more universal.

At present, the characterization of HAI by NTL data is still in the preliminary stage of exploration, although many scholars have constructive comments and expectations on it (Wang et al., 2012; Deng et al., 2018). Scholars such as Duan et al. (2020), Sanderson et al. (2002) and Venter et al. (2016b) have introduced NTL data into the study of human activity intensity, but most of them only use DMSP/OLS data. The characterized human activity intensity has time discontinuity and data inefficiency, which cannot reflect the current situation and continuous change trend of regional human activity intensity, due to the absence of DMSP/OLS data after 2013. This article attempts to combine the continuous NTL data, PD data and LU data to characterize HAI, which can be used as a reference and useful supplement for the characterization of human activity intensity in the future. Of course, methods to better characterize the intensity of human activity based on NTL data need to be explored in future research.

The results of this study indicated that HAI and LST were positively correlated in the human activity area on the north slope of Tianshan Mountains and had an escalating trend, and the changes in land use, population distribution and nighttime light could support a rational explanation for the spatio-temporal changes of LST, which was consistent with the previous conclusions (Wang, 2008; Deng et al., 2018; Chu, 2019), and human activities did have a pronounced impact on LST. However, the types and manifestations of human activities are diverse, and their effects on LST are also different. As the most intuitive embodiment of human activities, land use patterns can be used to explore the effects of different types of human activities on LST.

The spatial partition statistical method was used to quantitatively analyze the LST changes in the land-use change areas of human activity areas in each stage (Figure 10). This article only counts the six types of transformation with the largest area, on account of the complexity of land class conversion, and the large uncertainty of LST change that exists in the converted land class with a small area. When human activities manifest as the expansion of construction land, the increase of the impervious surface area of construction land converted from cultivated land and grassland leads to the enhancement of the heat island effect (Dong, 2012; Zeng, 2015), and the average temperature rise of LST is the highest in the first stage, reaching 0.36℃•5a^-1. The phenomenon of cooling (-0.11℃•5a^-1) in the area converted to construction land in the second stage is caused by the conversion of large areas of deserts, bare land and other unused lands to construction land. Hence one can see that the transformation of different land types by human beings will have different effects on the LST for the expansion of construction land. When human activities manifest as vegetation destruction (Cui et al., 1998), resulting in the increase of secondary unused land, vegetation degradation, etc., natural vegetation such as forest and grassland is destroyed, and the land is abandoned and converted to sparse grassland or even unused land so that the conversion to unused land and grassland areas generally present a warming state. The newly-added unused land in the study area is primarily distributed in the northern desert area, and the loss of bare surface water content and specific heat capacity due to vegetation degradation and land abandonment leads to an average LST warming of 0.27 ℃•5a^-1. The most significant warming is in the third stage (0.58 ℃•5a^-1), which is 13.49 times the global warming. When human activities manifested as wasteland reclamation and vegetation restoration, the accompanying reduction in deserts and bare land and the growth in vegetation coverage have a conspicuous cooling effect (Xia et al., 2010; Lu et al., 2017). The LST of newly added cultivated land and forest land in this area was cut down by 0.15 ℃•5a^-1 and 0.14 ℃•5a^-1 respectively, attribute to the increase of vegetation coverage and moisture content. The new water area is cooled due to a larger specific heat capacity and has a better regional temperature regulation ability. In the regions where the land type remains the same, the LST also showed an upswing trend, and the temperature in the construction land was the most prominent. The maximum impact of population on LST was 0.005 ℃•5a^-1·thousand people^-1, and the LST rose by 0.18 ℃ at most with the upswing of normalized NTL intensity by 0.01. When the population size and density grow and the brightness of nighttime lights upswing, it means that the intensity of land development and utilization aggrandizes, the social and economic activity upgrades, and the intensification level of human production and living and energy consumption elevates, which directly leads to the salient rise of LST (Guo et al., 2019; Wang et al., 2019; Hu et al., 2021; Shen et al., 2021). Certainly, the types and manifestations of human activities have anisotropic and cross-permeable effects on LST, and the quantification of the effects of different scales and forms of human activities on LST needs to be further explored in prospective research.