The permafrost zone of the Tibetan Plateau was selected as study area. The Tibetan Plateau has experienced a unique climate due to the complex topography and the influence of the Asian monsoon. Mean annual air temperature is below 0℃. Alpine grassland is the dominant vegetation type, while forest, wetland and crop in the eastern part occupied less area of the permafrost zone (Figure 1). Land cover data used in this study was reclassified from the Chinese land cover in 2005 (Liu et al., 2005) which was provided by Data Center for Resources and Environmental Sciences, Chinese Academy of Sciences (RESDC, http://www.resdc.cn). Internationally, permafrost was classified into different zones based on its permafrost extent estimated in percent area, including continuous permafrost (90%-100%), discontinuous permafrost (50%-90%), sporadic permafrost (10%-50%), and isolated patches of permafrost (0%-10%). According to this classification system, continuous permafrost zone on the plateau is only 300.01 km², 1.83% of total permafrost area. The distribution of the various types of permafrost zone was mapped in Figure 1. The digital data for permafrost boundaries were obtained from the American National Snow and Ice Data Center (NSIDC, http://nsidc.org/data/docs/fgdc/ggd318_map_circumarctic).

Long-term records of satellite-derived LST are currently accessible through a number of remote sensing platforms (Westermann et al., 2011). LST from the Moderate Resolution Imaging Spectroradiometer (MODIS) platform was most widely used. In this study, the LST product (MOD11A2) from 2000 to 2010 was extracted from the MODIS datasets of the Land Processes Distributed Active Archive Center (LP DAAC, https://lpdaac.usgs.gov/ products/modis_products_table/mod11a2). This dataset covering the study area is composed of daily 1-km LST values that are stored on a 1-km sinusoidal grid as the average of clear-sky LST values over an 8-day period. Accuracy of the product has been assessed over a widely distributed set of locations and time periods via several ground-truth and validation efforts. We calculated the mean value of LST for days and nights and then derived the mean annual values.

The NPP product (MOD17A3) from 2000 to 2010, produced by the Numerical Terra dynamic Simulation Group (NTSG)/University of Montana, was downloaded from LP DAAC and processed by MRT software to convert the format and the projection system. The accuracy of the NPP product at the spatial resolution of 1 km was assessed, and uncertainties in the product were established via independent measurements made in a systematic and statistically robust way that represents global conditions (Zhao et al., 2005). These data have been evaluated and used in studies from around the world (Zhao and Running, 2010). For a further instruction of the MODIS NPP on the Tibetan Plateau, a comparison was developed between observed NPP transformed from biomass (Ma et al., 2010) and MODIS NPP for grassland (Figure 2). The regressive coefficient (0.74) indicated a good modelling result for the MODIS products on this Plateau.

The correlation coefficient at the pixel scale was calculated using equation 1. The spatial pattern of the diverse correlation coefficients represents the different influences of LST on NPP; x_i and y_i represent the values of the two variables in the ith year, respectively.

From 2000 to 2010, annual NPP increased significantly (p < 0.01) with increasing LST (Figure 3), and the increase of annual LST accounts for 25% of the increase in NPP (R² = 0.25), with a correlation coefficient of 0.50. The positive correlation between NPP and surface temperature is consistent with that found by Mao et al. (2012) in the permafrost zone of Northeast China. Warming terrestrial surface temperatures accelerated permafrost thawing associated with more soil moisture and more mineralization and decomposition of nutrients, which increased vegetation productivity over a given period (Schuur et al., 2007). Meanwhile, the warming temperature may result in increasing growing season and thus improving the annual vegetation productivity (Zhuang et al., 2003). On the Qinghai-Tibet Plateau, Chen et al. (2013) also confirmed that warming climate enhanced NPP significantly. However, previous studies have mostly discussed the negative role of permafrost degradation in China (e.g., Wang et al., 2000). Our results revealed that warming temperature in the permafrost zone of the Tibetan Plateau played a positive role in vegetation growth and carbon sequestration from 2000 to 2010.

The map of mean annual NPP from 2000 to 2010 (Figure 4) showed a clear increase in NPP from the west to east that explained spatial gradient of NPP in the permafrost zone of the Tibetan Plateau. This result is consistent with estimation using the Terrestrial Ecosystem Model (TEM) (Zhuang et al., 2010), which attributed the west to east increase of NPP to the parallel increase in air temperature and precipitation. The alpine climate limited the growing environment of vegetation, further leading to lower NPP with a mean value of 80.5 gC·m^-2·yr^-1. As a result of differences in growth condition, mean annual NPP from 2000 to 2010 exhibited great variation among all 1 km × 1 km pixels with the largest value of 1398.4 gC·m^-2·yr^-1 and a standard deviation of 97.6 gC·m^-2·yr^-1. Several regions in the discontinuous permafrost zone, such as the area in the western part of the study region (Jin et al., 2009), where land degeneration is evident or the active layer is thin, exhibited low NPP values or had no vegetation. Larger NPP values were observed on the eastern side of the plateau where sporadic permafrost is the dominant type and is likely correlated with relatively high temperatures and soil moisture for alpine meadow and forest (Yang et al., 2008; Fan et al., 2010). Literature revealed that vegetation cover and biomass on the Tibetan Plateau were well correlated with permafrost depth (Wang et al., 2012). Owing to the various extents of permafrost thaw, different hydrothermal conditions contributed to the variations in NPP among various types of permafrost zone.

In an attempt to explicitly understand the correlation between NPP and LST at the scale of 1 km × 1 km, correlations between dynamic NPP and LST from 2000-2010 were detected for all pixels (Figure 5). A positive correlation between NPP and LST was primarily found for the permafrost zone of the Tibetan Plateau. Although positive correlations were found for most of the pixels, 28.84% of the pixels showed negative correlations between NPP and LST. In these cases, the degraded permafrost induced by warming temperature not only accelerated the degeneration, desertification and salinization of the grassland, but also enhanced moisture evapotranspiration and reduced soil moisture by removing the impermeable layer (Yang et al., 2004). Jin et al. (2009) hypothesized that the degradation of permafrost lowers groundwater table and dries the surface soil, which degrades the alpine grassland. The negative correlations of NPP to LST can be attributed to limited moisture availability for vegetation growth induced by permafrost degradation.

Annual mean NPP for different vegetation types from 2000 to 2010 were calculated (Figure 6). Mean annual NPP values were observed as 272.7 gC·m^-2·yr^-1 for forest, 220.1 gC·m^-2·yr^-1 for crop, 149.3 gC·m^-2·yr^-1 for wetland, and 68.6 gC·m^-2·yr^-1 for grassland, respectively. Meanwhile, significant increases (P < 0.05) in NPP were observed for each vegetation type, which can be attributed to more soil moisture, longer growth season and increase in the thickness of active layer induced by warming permafrost (Camill et al., 2001). Along with the warming temperature, forest showed the largest increase in NPP from 2000 to 2010, while grassland showed the smallest.

Correlations between NPP and LST during 2000-2010 were investigated to analyze the impacts of warming temperature on different vegetation types. NPP of the four vegetation types exhibited positive correlations with increasing LST, with a coefficient of 0.70 for forest, 0.50 for grassland, 0.38 for crop, and 0.33 for wetland. Based on the annual variations in NPP and its dependence on LST, we can conclude that forest NPP presented the most obvious response to warming temperature in the permafrost zone. Warming temperature in permafrost zone and induced rising CO₂ concentration contributed to the improvement in forest NPP (Zhou et al., 2013). Widespread permafrost degradation caused by warming climate on the Tibetan Plateau affected the soil moisture content and nutrient supply, thereby affecting ecosystem structure and function. And moisture content of the surface layer decreased with increasing thaw depth (Yang et al., 2013). Alpine vegetation showed notable and complicated responses to warming permafrost in different regions. Therefore, the NPP changes depending on LST in different types of permafrost zones classified with various thawing extents need to be investigated.

Obvious annual variations and differences in NPP were observed among different types of permafrost zone (Figure 7). Mean NPP in the continuous and discontinuous permafrost zones had similar values, which might be correlated with the same low temperatures and low extents of thawing. Vegetation in the sporadic permafrost zone had the largest mean NPP value (114 gC·m^-2·yr^-1), while the zone with isolated patches of permafrost had the smallest value (14 gC·m^-2·yr^-1). Additionally, the sporadic permafrost zone and the isolated patches of permafrost zone had the largest and smallest standard deviations of NPP over the 11-year period, respectively. An optimal combination of water and energy is necessary for vegetation growth. Therefore, the differences in mean NPP in various permafrost zones can be interrelated with different factors limiting vegetation growth. Low temperature was the most likely factor explaining low NPP in continuous and discontinuous permafrost zones, while low moisture availability was the likely factor limiting vegetation growth in the isolated patches of permafrost zone.

A slight increase in NPP for all types of permafrost zone was observed from 2000 to 2010. This increasing trend was primarily caused by warming temperatures and an associated increase in soil moisture and growing season length (Beck and Goetz, 2011). Various patterns and dynamics of NPP demonstrated the different response extents of vegetation to warming climate in the various permafrost zones on the Tibetan Plateau.

At the different types of permafrost zone scale, the relationship between NPP and LST changed from significantly positive in the continuous permafrost zone to relatively weakly positive in the discontinuous, sporadic and isolated patches of permafrost zones, with a correlation coefficient of 0.62, 0.51, 0.52 and 0.41 respectively. NPP in the continuous permafrost zone indicated that it has the greatest sensitivity to the changing LST, while NPP in isolated patches of permafrost zone suggested the least. Similar to the result from Yi et al. (2011), which was based on LST and fractional vegetation cover, various correlation coefficients in our study revealed the different responses and sensitivity of NPP to warming temperature in different types of permafrost zone. This result again demonstrated the positive effect of permafrost degradation on NPP. Increasing temperature, as well as the associated changes to the hydrological regime, directly and indirectly affected the NPP (Natali et al., 2012). This finding also illustrated the conclusion that permafrost degradation played a vital role in NPP variability that was also complicated with factors limiting vegetation growth in different permafrost zones. Consequently, the correlations between LST and NPP in the discontinuous, sporadic, and isolated patches of permafrost zones were not significant.

Permafrost degradation induced by varied energy balance and changed permafrost temperature is an indisputable fact along with the warming climate. Obvious changes in hydrologic condition, soil composition, and microbial activity were further observed with the degradation of permafrost. These changes imposed significant influences on vegetation and ecosystem. However, the effects of permafrost degradation on vegetation had different reports. On the one hand, negative effects of permafrost degradation were detected on the vegetation. It was mainly attributed to loss of soil moisture and nutrient in a way of declining water level, increasing surface runoff due to thawed permafrost (Jin et al., 2009). On the other hand, positive impacts of permafrost degradation on vegetation have been observed in many literatures. Prolonged growing season, increased active soil layer and soil water, raised decomposition velocity of organic substances and increased carbon dioxide levels could be related to the improvement of vegetation growth (Euskirchen et al., 2006; Chen et al., 2012). The varied response of NPP to the warming LST throughout the permafrost zone of the Tibetan Plateau demonstrated this difference. The influences of permafrost degradation on vegetation productivity have been related to topography, vegetation type, soil texture, as well as the time. Therefore, it is essential to develop a long term field monitoring to further detect the responses of vegetation characteristics to altered environment resulted from permafrost degradation in the future. In addition, more data and products, such as the land surface water index (LSWI) from remote sensing will contribute to the related study.