Grazing intensity and human activity intensity data sets on the Qinghai‐Tibetan Plateau during 1990–2015

As the ‘third pole’ of the world, the Qinghai‐Tibetan Plateau (QTP) is extremely ecologically sensitive and fragile, while also facing increasing levels of human activity, especially overgrazing. In this study, livestock data from the statistical yearbook were transformed to 1 km grazing intensity raster data by integrating the net primary productivity (NPP) data for 1990, 1995, 2000, 2005, 2010 and 2015. Furthermore, the human activity intensity (HAI) data at 1 km resolution were evaluated by the entropy weight method, which applied eight types of spatial data, including grazing intensity, Night‐Time Light, population density, Gross Domestic Product (GDP) density, the ratio of cultivated land, distance to road, distance to town and the slope of the Normalized Difference Vegetation Index (NDVI). Also, the five‐year interval human activity intensity data on the QTP from 1990 to 2015 were evaluated. By preparing the historical spatial data sets of grazing intensity and human activity intensity, our study will help to explore the influence of human disturbance on the alpine ecosystems on the QTP, as well as provide effective support for the decision‐making of the government aiming to achieve regional ecosystem management and sustainable development.


| INTRODUCTION
With the rapid development of social economy and the increase of global population so far during the 21st century, the intensity and scale of mankind's influence on the natural world have been steadily increasing (Li et al., 2019), which has brought great disturbance and significant pressure on ecosystems at different scales. Many researchers have shown that human demand for natural systems is rapidly accelerating (Jones et al., 2018;Motesharrei et al., 2016), which may undermine the ecosystem's stability and its ecosystem services (Ali et al., 2020;Krausmann et al., 2013;Steffen et al., 2015). Especially for grasslands, overgrazing has caused ecosystem degradation and has altered the community composition of grasslands, thus resulting in the decrease of grassland diversity (Sun et al., 2020a;Wang et al., 2015a). Therefore, it is necessary to evaluate grazing intensity and human activity intensity so that we can correctly evaluate the scale, intensity and temporal changes of these disturbances in order to prevent possible ecological threats.
The assessment method of human activity intensity evolves from statistical analysis methods to quantitative methods (Li et al., 2018a;. The available statistical analysis methods make it difficult to gain a highquality spatial data set of human activity intensity (Xu et al., 2015), while quantitative methods are explored based on multiple indexes of pressure change and state change . For pressure change, human activity intensity is evaluated through weight-based multi-index superposition systems, such as the human footprint index (Karimi & Jones, 2020), karst disturbance index (Tlhapiso & Stephens, 2020) and ecological footprint (Wu, 2020). For state change, human activity intensity is evaluated primarily considering land-use change and ecosystem service change (Rong et al., 2017). Among these, considering the spatial differences of human activity intensity, the weight-based multi-index, such as human footprint index, has been widely applied at different scales around the world (Correa Ayram et al., 2017;Johnson et al., 2017), including nature reserve management (Li et al., 2018a) and human activity influence assessment (Li et al., 2018b). Grazing intensity is mostly evaluated by the statistics of number of livestock, which is summarized according to the statistical yearbook at the county scale and lacks quantitative assessment at raster scale (Ouyang et al., 2016). Overall, spatial data of human activity assessment are still in a constant development process, while spatial data of grazing intensity are still in the exploratory stages.
As the 'roof of the world' and 'third pole of the world', the QTP provides various ecosystem services, including climate regulation (Jin et al., 2005), abundant biodiversity and genetic species (Sun et al., 2012), and abundant water resources, as it is known as the 'Asian water tower' (Pan et al., 2015). In recent decades, the social economy has developed rapidly and human activities have increased significantly, especially since the opening of the Qinghai-Tibet Railway. The impact of human activities on the ecological environment on the QTP has always been a research hotspot in the area of ecosystem research (Li et al., 2018b). Due to its unique and fragile ecosystem, it is significantly sensitive to climate warming and human disturbances as compared to other regions (Pan et al., 2017). It is widely acknowledged that the grassland and ecosystem services are undergoing various types of degradation, such as a decline in the water supply (Pan et al., 2015) and the loss of species and biodiversity (Li et al., 2019;Yang et al., 2018). However, only a few preliminary studies have considered the quantification of human activity intensity and graze intensity, as well as the implications for ecosystem management, on the QTP (Lu et al., 2017;Zhao et al., 2015a). Therefore, it is the need to accurately evaluate human activity intensity and grazing intensity for exploring the effect of human activity on the ecosystem, which thus provides guides for the ecological protection and restoration on the QTP.
Human activity is regarded as an important factor influencing the ecological environment, but only one or more factors have been analysed, so a comprehensive analysis of the QTP is still lacking (Harris, 2010;Zhao et al., 2015a). In order to assess human activity intensity on the QTP, Zhong et al. (2008) first used only the variables of cultivation activity and highway distribution (Zhong et al., 2008). However, in recent decades, the areas of built-up land, railways and expressways on the QTP also have shown increasing trends (Lu et al., 2017). Subsequently, Zhao et al. (2015) used population density, number of villages and road length in order to evaluate the disturbance of human activity on vegetation change on the QTP (Zhao et al., 2015b). Four categories of human influencing factors, including land use/cover, population density, road distribution and grazing density, were selected in order to map human influence intensity on the QTP from 1990 to 2010, and human influence intensity (HII) was overall revealed to be low, but steadily increasing (Li et al., 2018b). Furthermore, in a study by Li et al. (2018), combing population density, land-use intensity, road and railways and grazing intensity together, electricity infrastructure was newly introduced in order to quantify the human footprint in Tibet for 1990 and 2010 (Li et al., 2018a). However, these results were still insufficient and many other factors may also reflect human activity intensity, such as GDP density, cultivation activity and the slope of NDVI (Mrabet et al., 2017;Solen et al., 2018).
In this study, the entropy weight method was applied in order to determine the weight attribution of various human influence factors. Eight human activity factors including Night-Time Light, population density, GDP density, the ratio of cultivated land, distance to road, distance to town, grazing intensity and the slope of NDVI were considered to quantify and map human activity intensity on the QTP. Then, six periods of human activity intensity and grazing intensity data sets from 1990, 1995, 2000, 2005, 2010 and 2015 were prepared. Through the quantitative assessment of human activity intensity and grazing intensity, the results will be conducive to providing a guidance for ecosystem service decision-making on the QTP.

| SITE DESCRIPTIONS
The Qinghai-Tibetan Plateau (QTP) (26°00′-39°47′N, 73°19′-104°47′E) is located in the southwest of China, covering a quarter of China's total land area, including Tibet, Qinghai, southern Xinjiang, western Sichuan, and parts of Gansu and Yunnan provinces. It is the largest and highest plateau in the world, with an average elevation higher than 4,000 m   (Figure 1). The plateau is the source of important rivers in East, Southeast and South Asia, with abundant water resources totalling 546.34 billion m 3 , known as the 'Asian water tower' (Sun et al., 2020b). The typical climate type is a continental plateau climate with strong solar radiation and low temperatures (Pan et al., 2017). The average annual temperature ranges from −6°C to 20°C, and the average annual precipitation varies from 415 mm to 515 mm, primarily occurring during the period from May to September. Due to its spatial differences, the temperature and precipitation significantly differ from the northwest part to the southeast part. In recent decades, with the development of urbanization and increased population growth on the QTP, especially since the opening of the Qinghai-Tibet Railway, the economy has developed rapidly and human activities have increased significantly, including agricultural activities, grazing, mineral exploration, tourism and the construction of ecological projects (Li et al., 2018a). According to statistics, human activity intensity on the QTP increased by 28. 43%-31.45% during 199043%-31.45% during -201043%-31.45% during (Li et al., 2018b). The regional economy is dominated by animal husbandry, and due to the large grassland coverage, the animal husbandry economy has flourished. Tourism is another economic pillar industry, as since 2000, the average annual growth rate of tourism population on the QTP has been 25.31%.

| Grazing intensity data
3.1.1 | The NPP data NPP refers to the total amount of organic matter accumulated by the plant community through photosynthesis per unit of time and space, which reflects the growth status of vegetation. In this study, NPP is used to take account of the spatial heterogeneity of grassland conditions, which can reflect the grazing intensity . In general, the larger the NPP value, the larger the grazing intensity. The NPP data were obtained from | 143 SUN et al.
the monthly NPP data set covering China's terrestrial ecosystems at north of 18°N (1985-2015), published in the Journal of Global Change Data & Discovery (http://www.geodoi.ac.cn). The data covered the period of 1990-2015, and the spatial resolution was 1 km (Chen, 2019).

| Grazing intensity
The grazing intensity data on the QTP for 2000 and 2010 were obtained from China's ecosystem assessment and ecological security database (http://www.ecosy stem.csdb.cn/ index.jsp) (Ouyang et al., 2016). These data were obtained by the number of sheep at county scale according to the statistical yearbook. However, the grazing intensity data for 1990, 1995 and 2015, as well as all data during 1990-2015 at raster scale, were not produced. Thus, we rasterized the countyscale data based on the NPP data and the statistical yearbook from 1990 to 2015 on the QTP.
3.1.3 | Vegetation type map of 1:1,000,000 According to the vegetation map of 1:1,000,000 in China, the grassland area was extracted on the QTP, primarily including alpine meadow, alpine steppe and warm steppe.

| Human activity intensity data
Considering previous studies of human activity intensity (Li et al., 2018a;Li et al., 2018b;Sun et al., 2020b), eight factors were applied to quantify and map human activity intensity on the QTP, including grazing intensity, Night-Time Light, population density, GDP density, the ratio of cultivated land, distance to road, distance to town and the slope of NDVI. Among these, positive indicators included grazing intensity, Night-Time Light, population density, GDP density and the ratio of cultivated land, which indicated that the larger the value, the higher the human activity intensity. Conversely, negative indicators were distance to road, distance to town and the slope of NDVI, indicating that the lower the value, the higher human activity intensity.

| Night-Time Light
Night-Time Light data were regarded as an effective representation of human activity, which represented electricity infrastructure intensity and energy development (Elvidge et al., 2001;Sun et al., 2020b). The Defense Meteorological Satellite Program Operational Linescan System (DMSP/OLS) data were provided by the National Oceanic and Atmospheric Administration (NOAA)'s National Centers for Environmental Information (https://ngdc.noaa.gov/eog/downl oad.html), which covered the years of 1992, 1995, 2000, 2010 and 2013. The DMSP-OLS data from 1990 and 2015 were unavailable and thus were substituted by data from 1992 and 2013, respectively.

| Population density and GDP density
Population density and GDP density were important factors reflecting the interactions between human activity and ecosystems . Population data for 1990for , 1995for , 2000for , 2005for , 2010for , and GDP density for 1995for , 2000for , 2005for , 2010 and 2015, were obtained from the Resources and Environmental Data Cloud Platform, Chinese Academy of Sciences (http://www.resdc.cn/). The GDP data from 1990 were substituted by data from 1995. Specifically, these data represent the spatial distribution grid data for population and GDP per square kilometre.

| Land use and land cover (LUCC)
Land use change is also an important indicator influencing the intensity of human activity (Liu et al., 2019). The LUCC data at a 100 m resolution for 1990, 1995, 2000, 2005 and 2010, as well as the data of 1 km resolution for 2015, were acquired by the Data Center for Resources and Environmental Sciences, Chinese Academy of Sciences (RESDC) (http://www.resdc. cn). The 1 km land-use data in 2015 were resampled to a 100 m resolution.

| Grazing intensity
The Qinghai-Tibet Plateau has the largest alpine grassland ecosystem in the world, with a grassland area of approximately 16.538 × 10 5 km 2 , accounting for 41.88% of China's grassland area (Hao et al., 2020). Therefore, grazing intensity was considered to be a key factor influencing human activity intensity. The rasterized grazing intensity data above for 1990, 1995, 2000, 2005, 2010 and 2015 were used in order to evaluate human activity intensity at the 1 km scale.

| NDVI data
The NDVI data from 1986 to 1997 were derived from the NOAA Global Inventory Monitoring and Modeling System (GIMMS) (https://ecoca st.arc.nasa.gov/). The temporal resolution is twice per month, and the spatial resolution is 1/12 of a degree. Then, the data during the period of 1986-1997 were resampled from 8 km resolution to 1 km resolution. The NDVI data during 1998-2015 were obtained by the vegetation instrument of SPOT-4 and SPOT-5 satellites from National Science and Technology Infrastructure (http://westdc.west-gis.ac.cn/). The source data were firstly performed according to atmospheric correction, radiation correction and geometric correction in order to generate the maximum synthetic data of 10 days. Then the formula (NDVI = 0.004 × DN − 0.1) was used to convert the original raster DN data to the NDVI data within the range of −1 to 1.
The detailed information of the spatial data sources is shown in Table 1.

| Data re-processing
Before processing the data, all of the original data should be preprocessed considering the differences of resolution, projection and spatial range for the different data. In this study, the spatial raster was uniformly resampled to a resolution of 1 km, and the spatial range of human activity intensity was uniformly clipped to the range of the QTP. For grazing intensity, the spatial range was uniformly clipped to the range of the grassland on the QTP, and the coordinate projection systems were uniformly transformed to Krasovsky_1940_ Albers. The two standard latitudes of the projection system were 25°N and 47°N, respectively, and the central meridian was 105°E. Furthermore, according to the correction parameters from previous studies (Chang et al., 2020;Chen et al., 2019;Letu et al., 2010), the desaturation method based on the frequency distribution of the light DN value was used in order to correct the discontinuity and oversaturation of the Night-Time Light data during the period of 1992-2013. The road and town data were used to calculate the distance to road and distance to town by the Euclidean distance in ArcGIS, and the slope values of NDVI during the periods of 1986-1990, 1991-1995, 1996-2000, 2001-2005, 2006-2010 and 2011-2015 were processed by the linear regression analysis with ordinary least squares (OLS). The ratio of cultivated land for 1990, 1995, 2005, 2010 and 2015 was calculated by the land-use data based on the focal statistics in ArcGIS.

| Grazing intensity
Based on the grazing intensity data for 2000 and 2010, the variable of grazing intensity at county scale was rasterized. The unit of county-scale grazing intensity for 2000 and 2010 was the livestock unit (annual average value), and it was assumed that the NPP distribution could be applied in order to relocate the grazing intensity variability within a certain county. Thus, in this study, the uniform county-scale grazing intensity for 2000 and 2010 was rasterized based on the NPP data. The formulas were calculated as follows: where Y is the county-scale grazing intensity; X i was the NPP value at ith raster; and k was the conversion coefficient at each county.
( where Z is the raster-scale grazing intensity at the county; L i was the NPP average value at the ith county; and k i was the conversion coefficient of the ith county. For 1990For , 1995For , 2005, the number of livestock in each county of the QTP was counted according to the statistical yearbook data of each county from 1990 to 2015. The change rates of livestock stock in 1990, 1995 and 2005 relative to 2000 were calculated for each county, and the change rate of livestock stock in 2015 relative to 2010 was calculated for each county as well. Then, by multiplying the countyscale grazing intensity for 2000 and 2010 by the change rate of each county, the grazing intensity in 1990, 1995, 2005 and 2015 was obtained at the county scale. Furthermore, the raster-scale grazing intensity of each county in 1990, 1995, 2005 and 2015 was obtained through the NPP conversion coefficient k. Finally, the grazing intensity of grasslands and the grazing intensity overall on the QTP were quantified and mapped in 1990in , 1995in , 2000in , 2005in , 2010in (Sun et al., 2020a.

| Human activity intensity
Determining weight to factors is an important step towards evaluating the intensity of human activity. Two common methods that are applied in order to determine weights are the objective weighting approach and the subjective weighting approach. Among these, the entropy weight method belongs to the objective weighting method, which can preclude the effect of some subjective factors on the assessment results Wang et al., 2015b). In this study, the entropy weight method was used to determine weight attribution, and the calculation process of the entropy weight method is as follows:

| Standardization of the indicators layer
Standardization can eliminate the influences of different dimensions. For the positive indicator and negative indicator, the range of each indicator is standardized to 0-100 using Equations 4 and 5, respectively.
where, X j is the standardized value of the jth indicator, and X i is the original value; X max and X min are the maximum and the minimum values, respectively.

| Constructing the judgement matrix
If there are p evaluation rasters r ij , and each raster has a q evaluation indicator, then the matric R was constructed as Equation 6: where R is the evaluation matrix, and r ij is the matrix value of ith evaluation raster in the jth evaluation indicator.

| Calculation of entropy
The entropy was calculated as Equations 7 and 8: where e j is the entropy value of indicator j; P ij is the proportion of the value of the ith evaluation raster in the jth evaluation indicator, when P ij = 0, P ij ln P ij = 0; and ln P is the information entropy coefficient.

| Calculation of the entropy weights
The entropy weight is calculated as Equation 9: where W j is the entropy weight of indicator j; e j is the entropy of indicator j; and j is human activity indicator. The weight results are shown in Table 2.
Through the entropy weights allocation above, all human activity factor layers were overlaid to quantify the human activity intensity on the QTP for 1990QTP for , 1995QTP for , 2000QTP for , 2005QTP for , 2010 and 2015. The formula of human activity intensity is calculated as follows: where HAI is human activity intensity; X j is the standardization of indicator j; W j is the entropy weight of indicator j; and j is human activity indicator.
Furthermore, the change rate of human activity intensity from 1990 to 2015 was calculated by the linear regression analysis with ordinary least squares (OLS). The slope is calculated as follows: where Slope is the change trend of HAI; n is the time series; and y i is the value of HAI at year i. When Slope > 0, it indicates an increasing trend of HAI in n years; and when Slope < 0, it indicates a decreasing trend of HAI in the study period. Figure 2 shows the working flowchart of this study.

| Data composition
The grazing intensity data sets included six periods of grazing intensity data on the grasslands of the QTP for 1990, 1995, 2000, 2005, 2010(Sun et al., 2020a. Each period's data were stored in TIFF format and were named by the year for 1990, 1995, 2000, 2005, 2010 and 2015, respectively. These data were saved as a compressed file titled 'Grazing intensity datasets on the grassland of the QTP during 1990-2015.zip', and the total data size was 17.0 MB. The grazing intensity data sets also covered six periods of grazing intensity data on the QTP for 1990, 1995, 2000, 2005, 2010(Sun et al., 2020a. The data were saved as a compressed file as well ('Grazing intensity datasets on the QTP during 1990-2015. zip'). Each data period was stored in TIFF format and named by year, and the total data volume was 17.0 MB. The human activity intensity data sets also had six periods of human activity intensity on the QTP for 1990, 1995, 2000, 2005, 2010(Sun et al., 2020a. Each data period was named by year and saved in TIFF format. These data were saved as a compressed F I G U R E 3 Grazing intensity data sets on the grasslands of the QTP for 1990QTP for , 1995QTP for , 2000QTP for , 2005QTP for , 2010QTP for and 2015 file ('Human activity intensity datasets on the QTP during 1990-2015.zip'), and the total size was 63.6 MB. Finally, the change rate of human activity intensity was provided on the QTP from 1990 to 2015 and was titled by 'slope1990-2015' (Sun et al., 2020a). The data set was stored in TIFF format with a compressed file titled 'The change rate of human activity intensity dataset on the QTP from 1990 to 2015.zip', and the size was 10.7 MB.

F I G U R E 4
Grazing intensity data sets on the QTP for 1990QTP for , 1995QTP for , 2000QTP for , 2005QTP for , 2010QTP for and 2015 | 149 SUN et al.

| Grazing intensity data sets
Based on the NPP data and livestock data from the statistical yearbook, the rasterized grazing intensity data sets on the grasslands of the QTP were presented in Figure 3. Temporally, mean grazing intensity was 8. 12, 7.99, 7.56, 7.86, 10.16 and 9.30, respectively for 1990, 1995, 2000, 2005, 2010 and 2015. It was shown that grazing intensity decreased first and increased from 1990 to 2015, implying that although overgrazing severely threatened biodiversity and changed the F I G U R E 5 Human activity intensity data sets on the QTP for 1990QTP for , 1995QTP for , 2000QTP for , 2005QTP for , 2010QTP for and 2015 structure of grassland plant communities, moderate grazing disturbance may actually contribute to increased grassland diversity. Therefore, we should arrange the time and patterns of grazing reasonably in order to promote the sustainable development of the grasslands. Spatially, grazing intensity showed no significant differences from 1990 to 2015 on the QTP (Figure 3a-f). The high grazing intensity areas were primarily located in the central and eastern areas of the QTP, including the middle and western areas of Tibet, southern and eastern areas of Qinghai, northern areas of Sichuan, and southern areas of Gansu. However, the low grazing intensity mostly occurred in the north and south, which covered the low grassland. Furthermore, the grazing intensity values were set as 0 in the none-grassland area of the QTP, and grazing intensity data sets on the whole area of the QTP were rasterized according to the grassland data sets, as shown in Figure 4a-f. Meanwhile, the spatio-temporal characteristics of grazing intensity on the QTP were similar to the grazing intensity data set on the grasslands of the QTP.

| Human activity intensity data sets
According to the weight assign results, eight human factor layers were overlaid in order to map human activity intensity on the QTP. Human activity intensity data sets were presented in Figure 5. Temporally, mean HAI for 1990HAI for , 1995HAI for , 2000HAI for , 2005HAI for , 2010HAI for and 2015, respectively, indicating that HAI has shown a decreasing trend overall. Spatially, HAI showed no significant variations during 1990-2015 on the QTP (Figure 5a-f). The high-HAI areas were distributed among the central and eastern regions of the QTP, including the middle part of Tibet, southern and eastern areas of Qinghai, northern areas of Sichuan, and southern areas of Gansu, while the low-HAI areas were primarily located in the northwestern area, particular in the unpopulated areas of north Tibet. In order to explore the change trends of HAI from 2000 to 2015, the change rate was calculated on the QTP (Figure 6). HAI in approximately 8% of the study area showed an increasing trend, and significant increases of HAI were primarily distributed in the north and south Tibet, south and east Qinghai and south Gansu ( Figure 6). Only 2.20% of the study remained unchanged on the QTP. Meanwhile, HAI in approximately 65% of the study area showed a significant decrease (p < 0.05), and 24.01% of the study area experienced a nonsignificant decrease, indicating that the ecological environment improved greatly on the QTP during the time period 1990 to 2015. This may be due to the implementation of ecological protection projects since 2000, such as the implementation of the 'Natural Forest Conservation Program' and the 'Grain for Green Program'.

| DISCUSSIONS
With the continued development of social economy and technology, the influence of human activities on the natural environment has been constantly deepening, thus resulting in a series of ecological environmental problems. As an important indicator with which measure the degree of human impact on the natural environment, human activity intensity and its impact on the ecological environment were analysed here. In this study, the spatial data sets of grazing intensity and human activity intensity were mapped and quantified from 1990 to 2015 on the QTP, which may help to explore the spatial characteristics and laws of grazing activity and human activity distribution. Among them, the quantitative evaluation of grazing intensity can provide a support for grassland degradation and play a guiding role in grassland ecosystem management on the QTP. Meanwhile, the evaluation of human activity intensity reflected the relationship changes between human activity and the ecological environment, and it also helped to explore the interactive effects of human-natural factors on the ecosystem, which in turn provided a scientific basis for the formulation of land-use policy and ecological environment construction in future studies. The high-quality spatial data of grazing intensity and human activity intensity would also contribute to distinguishing the effect of human activity and climate change on the earth's ecosystem as a whole, further promoting an integrated understanding of the impact of human activity on the ecosystem. This could provide effective guidance and policy support for regulating human activity and promoting the sustainable development of the world.
In our study, the spatio-temporal characteristics of grazing intensity and HAI were explored in-depth. Compared with previous studies (Duan & Luo, 2019;Li et al., 2018a;Li et al., 2018b;Zhao et al., 2015a), the similarities were F I G U R E 6 The change rate of the human activity intensity data set on the QTP from 1990 to 2015 as follows. Spatially, the southern and eastern areas of the QTP were the gathering regions of human activity, and climate and topography were the main driving forces resulting in the distribution of human activity intensity. These areas have abundant precipitation, favourable temperatures and low altitude, which contributed to the growth of many crops and the survival of humans (Sun et al., 2020b). In this study, higher HAI regions were primarily distributed in the south and east of the QTP, while lower HAI regions were located in the northwest. At previous studies, Zhao et al. (2015) found that the southern and eastern regions of the QTP were deeply disturbed by human activity, including central Tibet, the east of the Qinghai Province, and part of each of the Yunnan and Sichuan Provinces (Zhao et al., 2015a). Li et al. (2018) also concluded that higher HII areas were concentrated in central Tibet, as well as the east and southeast of the QTP (Li et al., 2018b). Furthermore, higher grazing intensity regions were primarily located in the middle of the QTP, while lower value regions were distributed in the south and the north, indicating that the grasslands were the most susceptible regions by grazing disturbance on the QTP. Temporally, the change rate of HAI represented the variation trends from 1990 to 2015 at different regions of the QTP, and 7.91% of HAI increased and 89.50% of HAI decreased, in accordance with previous studies (Duan & Luo, 2019;Li et al., 2018a;Li et al., 2018b). Li et al. (2018) demonstrated that 28.43% of raster-scale HII and 31.45% of county-scale HII on the QTP showed increasing trends from 1990 to 2010 (Li et al., 2018b), indicating that the ecosystem was effectively influenced by human activity. In addition, the selection of human activity factors referred to and included various human factors that have been applied in previous studies (Duan & Luo, 2019;Li et al., 2018b). The differences between these previously mentioned were that the assessment methods and the selection of the human influence factors differed. In this study, we applied the entropy weight method in order to evaluate human activity intensity, and we comprehensively considered the impacts of eight types of human factors on the ecosystem of the QTP. Grazing intensity was also rasterized based on the NPP data and livestock data from the statistical yearbook. The prior analysis indicates the reliability and validity of our methodology and data sets, and taking regional characteristics into consideration, the data sets can also be applied in many other studies and regions as well.
However, some uncertainties and limitations may still exist in this study. In the selection of human activity factors, many other factors may also influence human activity, such as tourism, mineral resources exploitation and ecological project construction. These potential factors may threaten the ecosystems on the QTP, and this may underestimate the actual human activity intensity on the QTP for lack of consideration of all possible human factors. Thus, we should pay comprehensive attention to considering the influences of other factors with regards to human activity on the QTP. Due to the adverse climate and topographical conditions, some areas of the QTP were inaccessible, and thus some remote sensing and socio-economic data are not available from these areas. For example, road and town data at 1:250,000 in 2015 were used in order to obtain the data of distance to road and distance to town, which did not demonstrate any dynamic changes. The night-time light data in 1990 and 2015 were replaced by the same data from 1990 and 2013, respectively, and these circumstances may weaken the credibility and accuracy of the human activity assessment results. Therefore, we should carry out long-term data monitoring in future studies, especially in the remote areas of the QTP.

| DATA AVAILABILIT Y
The grazing intensity and human activity intensity data sets from the QTP during the period of 1990-2015 are all decompressed in TIFF format. All data sets presented in this paper were released on September 23, 2020, and are available at the Science Data Bank (http://www.dx.doi.org/10.11922/ scien cedb.00171, Sun et al., 2020) with the file protection period until September 23, 2021 (one year after initial access). The private link for downloading the data is here: http://www.scidb.cn/api/sdb-perso nal-servi ce/datas et/surl/ emaYVj. These maps that have been produced from data sets help to provide an intuitive description of the availability of each data set and to facilitate the selection of data concerning human-natural interactions and the driving mechanism of the ecosystem. These data can be opened and manipulated by ArcGIS, QGIS, ENVI and ERDAS.

| CONCLUSIONS
Increasing human activities, especially overgrazing, have resulted in the degradation of the ecological environment and the ongoing loss of biodiversity. The grazing intensity and human activity intensity data are scarce but invaluable at high mountainous regions, especially on the QTP. Longterm, high spatial resolution and high-quality grazing and human activity data at raster scale on the QTP are vital for developing a deeper understanding of human disturbance on the ecosystem. In this study, a rasterized grazing intensity data set from 1990 to 2015 was presented combing the NPP data and livestock data obtained from the statistical yearbook. Furthermore, eight types of spatial data were applied in order to evaluate the intensity of human activity and HAI data sets of the QTP at 1 km raster scale were derived for 1990, 1995, 2000, 2005, 2010 and 2015. Compared with previous data sets regarding the QTP, the data sets in this study provide more comprehensive and high-quality information. Therefore, the high-resolution data sets will contribute to promoting a greater scientific understanding of the interactions between humans and the ecosystem on the QTP, and even the earth as a whole, thus facilitating ecosystem management and sustainable development in alpine regions.