1 Introduction

Permafrost distribution and its thermal state are very important for cold regions engineering design, construction, and environmental protection (Cheng and Wu, 2007). The area of permafrost distribution in China is the largest in terms of middle-and high-altitude permafrost regions, as the Qinghai-Tibet Plateau (QTP) permafrost is of a higher temperature compared with that in Siberia and the Arctic, which are more sensitive to global climate warming and human activity (Wu et al., 2002; Haeberli and Hohmann, 2008; Li et al., 2008; Ran et al., 2012).

Significant permafrost degradation has occurred and is occurring in most permafrost regions in China (Wang et al., 2000; Wu and Liu, 2004; Wu and Zhang, 2008; Jin et al., 2009; 2011). In the QTP, the increase of mean annual air temperature (MAAT) was about 0.2~0.4 ℃ from the 1970s to the late 1990s (Wang et al., 2000). The decadal averages of MAATs from 1961-2010 rose by 1.3 ℃, with an average increase rate of 0.03 ℃/a (Jin et al., 2011). At Xidatan near Golmud city in north boundary of permafrost in QTP along the Qinghai-Tibet Railway (QTR), the lower limit of permafrost moved upward about 25 m from 1975 through 2002 (Nan et al., 2003). Cheng et al. (2012) reported on the decadal changes in permafrost distribution on the QTP over the past 50 years (1960-2009), demonstrating that the rate of permafrost loss has accelerated since the 1980s, and about one-fifth of the total area of permafrost in the 1960s has degraded. In northeast China, the position of the southern limit of permafrost (SLP) has migrated northward about 20 to 100 km in the 2000s relative to the 1970s, according to the empirically correlated of SLP with MAAT isotherm. The areas of sporadic discontinuous and isolated patchy permafrost have decreased by 90, 000~100, 000 km², or 35%~37% of their total areal extent (Jin et al., 2007).

Many of these studies emphasized the migration of permafrost "boundaries" based on the relationship between climate warming and permafrost "boundaries", which are naturally continuous, inexact representations of permafrost distribution and permafrost degradation (Yang et al., 2010). These are not comprehensive and do not adequately reflect changes in the permafrost thermal state, especially in the interior of permafrost zones. More important is that the response time and affected depths of permafrost to climate warming depend on extents, durations, amplitudes, and rates of climate warming and are closely related to soil types, surface coverage, ice content, groundwater occurrence, geothermal anomalies, and human activities (Cheng and Jin, 2013). The complex physical mechanisms of the interactions between climate change and permafrost are currently very unclear (Jin et al., 2011), and large of degree of uncertainty may exist in their evaluations. The current warming climate might not cause a large area of permafrost to disappear, because the thermal inertia of permafrost may allow the permafrost to persist for a long time (Cheng et al., 2012). Therefore, the role of permafrost "boundaries" change is limited for specific applications, especially in the engineering field.

The primary objective of this paper is to evaluate the thermal state, that is, the stability degradation in the permafrost regions over China during the past 30 years, by means of a thermal stability classification system using mean annual air temperature (MAAT) derived from a new, high-resolution atmospheric reanalysis dataset.

2 Classification system of permafrost stability

Cheng (1984) developed a thermal stability-based permafrost classification system that uses three-dimensional zonation (altitude, latitude, and aridity [or longitude]). The classification definitions are shown in Table 1. Permafrost is classified as extremely stable, stable, sub-stable, transitional, unstable, or extremely unstable types. The stability and design principles for engineering in the types of permafrost are also proposed. Obviously, it is more useful to describe permafrost degradation as related to engineering rather than changes in the permafrost extent.

Three criteria, the mean annual ground temperature (MAGT), the thickness of permafrost, and the mean annual air temperature (MAAT), were used to define the classification system. MAGT is the most direct indicator to determine whether permafrost exists. Although measurement of long-term MAGT or regional MAGT is almost impossible due to the high cost of drilling boreholes, MAGT often been used as a true value to validate and calibrate other indicators, or to monitor permafrost changes at the local scale. MAAT is the most frequently used indicator of permafrost, being easy to measure and having better spatial representativeness. MAAT is therefore used in this paper. However, the correlation between the southern limit of permafrost (SLP) and the MAAT is spatially heterogeneous, especially at the local scale. For example, the MAATs of -1.0, 0, and 1 ℃ in the western, middle, and eastern sections of northeast China, respectively, were used to distinguish between permafrost and seasonal frozen ground (Jin et al., 2007). In the QTP, the MAAT of -2 ℃ usually was used. In this paper, a uniform 0℃-MAAT-isotherm was used to distinguish between permafrost and seasonal frozen ground. Thus, the extremely unstable type was defined as -2 to 0 ℃ of MAAT (Table 1) at the national scale.

Cheng (1984) developed a classification system for the QTP, which is a typical altitudinal permafrost region. We assumed that it also applies to the permafrost regions in northeast China because permafrost is not present in the plains at the same latitude, although there was permafrost named as 'latitude permafrost' in the past (Jin HJ, personal communication, 2011).

3 Datasets 3.1 Data acquisition

A new air temperature dataset was developed by the Data Assimilation and Modeling Center for Tibetan Multi-spheres, Institute of Tibetan Plateau Research, Chinese Academy of Sciences (ITPCAS) (He, 2010), and currently covers the period 1980 to 2011. Its spatial resolution is 0.1° and its temporal resolution is 3 h. The 740 in-situ observations collected from the China Meteorological Administration (CMA) operational meteorological stations were merged into the Princeton meteorological reanalysis data (Sheffield et al., 2006) to produce near-surface air temperatures. The validity of this dataset has been evaluated by comparing it with other reanalysis data; the results showed that this dataset is more accurate with higher spatial and temporal resolution than the Global Land Data Assimilation System (GLDAS), Princeton, and NCEP/NCAR reanalysis data (He, 2010; Chen et al., 2011; Pan and Li, 2011). This dataset offers a new potential for monitoring permafrost changes.

The original ITPCAS near-surface air temperature dataset from 1981 to 2010 with parameters of 3 h and 0.1°×0.1° was collected from the Environmental and Ecological Science Data Center for West China (WestDC) (http://westdc.westgis.ac.cn). The MAATs were derived from the original ITPCAS near-surface air temperatures using simple arithmetic averages.

3.2 Validation of the ITPCAS near-surface air temperature

The accuracy of the ITPCAS near-surface air temperature dataset has been evaluated at daily and hourly levels using high-resolution data collected through the GEWEX Asia Monsoon Experiment-Tibet (GAME-Tibet), the CEOP Asia-Australia Monsoon Project-Tibet (CAMP-Tibet), and the Watershed Airborne Telemetry Experimental Research (WATER) data (Li et al., 2009; He, 2010; Pan and Li, 2011). The results showed that the ITPCAS dataset is more accurate with higher spatial and temporal resolution than the other main reanalysis datasets available currently (He, 2010; Pan and Li, 2011). The hourly validation used four in-situ near-surface air temperature observations in 1998 of CAMP-Tibet, including the D110, D66, MS3608, and Tuotuohe stations. The results showed that the mean bias error of ITPCAS near-surface air temperature is about 1 ℃, which is more accurate than the Princeton near-surface air temperature (He, 2010). A similar conclusion was found by hourly comparison with in-situ near-surface air temperatures from seven WATER sites during the entire year of 2008 (Pan and Li, 2011). The validity of this dataset was also evaluated by an input land surface model to improve the land surface temperature modeling (Chen et al., 2011).

Overall, the accuracy of the ITPCAS MAAT dataset is good enough, although it somewhat underestimates in alpine mountains areas. That can be explained by the scale or representation difference between in-situ MAAT and 0.1°. Nevertheless, the accuracy is acceptable for national-scale research. For fine-scale study, a new model needs to be developed that integrates the topographic conditions (such as aspect, slope), solar radiation, soil moisture, soil properties, surface coverage, ice content, and even human activities. Recently, Bonnaventure et al. (2012) and Gruber (2012) presented a permafrost probability model that integrated some of above factors to improve the mapping resolution of permafrost distribution, especially in mountainous regions.

3.3 Statistical characteristics of MAAT

The statistical characteristics of MAAT in the permafrost regions, which were derived from the official Map of Snow, Ice and Frozen Ground in China (Shi, 1988), are presented in Table 2. Table 2 shows that the decadal averages of MAAT from 1981-2010 in the permafrost region in China rose by 0.91 ℃, with an average increase rate of 0.03 ℃/a. This is in agreement with a similar warming rate found by Jin et al. (2007) and Cheng et al. (2012), indirectly proving the good potential of the ITPCAS MAAT dataset for monitoring climate change in permafrost regions. In the permafrost region of China, about 10.5% of MAATs are greater than 0 ℃, but only 3.1% are greater than 3 ℃. That can be explained by permafrost not being present in some parts of the permafrost regions; this means that inconsistencies in permafrost region boundaries inevitably exist between the official Map of Snow, Ice and Frozen Ground in China and the ITPCAS MAAT dataset.

4 Results and discussion

According to the classifications of permafrost stability in Table 1 and the decadal average MAATs, we produced a permafrost stability type map for the 1980s, 1990s, and 2000s (Figure 1a-c). Figure 1 shows that the spatial pattern of the permafrost distribution in China is consistent with the existing permafrost map. The permafrost boundary in the 1980s (Figure 1a) is more similar to the Map of Snow, Ice and Frozen Ground in China(Shi, 1988), which was edited based on field survey data in the early 1980s. The permafrost boundary in the 2000s (Figure 1c) is more similar to the permafrost map edited by Ran et al. (2012), which represents the permafrost distribution state in the 2000s. This demonstrates the rationality of these permafrost stability type maps.

4.1 Permafrost stability change in the time domain

The areas of permafrost stability type were derived from the permafrost stability maps of the 1980s, 1990s, and 2000s (Table 3). Table 3 shows that the permafrost stability has continuously degraded over the past 30 years. The area of stable type has decreased and the unstable type has increased. Specifically, the stable type has incurred the most serious degradation, with a rate of 0.90% per year, while the areas of unstable type and extremely unstable type have increased with a rate of 0.38% and 0.80% per year, respectively. Overall, the climate change led to the degradation of permafrost stability. The areas of upper-zone and middle-zone altitudinal permafrost in China have decreased with a rate of 0.88% and 0.36% per year, respectively, while that of the lower zone has increased with a rate of 0.56% per year. Additionally, the total area of permafrost has decreased with a rate of about 0.22% per year. This decrease is mainly due to the degradation of the lower zone of permafrost, especially the degradation of the extremely unstable type, about 44.2% of which has changed into seasonal frozen ground.

4.2 Permafrost stability change in the spatial domain

We analyzed the permafrost stability change in the spatial domain from the following two aspects. First, the type variation of permafrost stability changes over the past 30 years was quantitatively evaluated using the transition matrix method, as shown in Tables 4-6. The permafrost stability was degraded overall but it was locally strengthened. For example, for the transitional type, about 47.22% was changed in the 1980s and 1990s; 45.24% of it was degraded into the unstable type but 1.73% of the region's permafrost still changed into the sub-stable type, that is, 1.73% of the transitional type became more stable from the 1980s to the 1990s (Table 4).

During the 1990s to the 2000s, about 63.54% of the region's transitional type of permafrost was changed; 61.42% of it was degraded into the unstable type but 2.12% still changed into the sub-stable type (Table 5). For all of the permafrost stability types during the 1980s to the 2000s, the greatest frequencies or greatest areas of stability changes occurred in the following order: extremely stable type > stable type > extremely unstable type > sub-stable type > unstable type > transitional type. Although the permafrost stability occasionally strengthened in each type, it was not enough to change the overall worsening trend of permafrost degradation (Table 3 and Table 6).

Secondly, we evaluated the spatial distribution of permafrost stability change over the past 30 years. Figure 1d-f shows that the degradation of the permafrost stability occurred over large areas in China, including the Qilian Mountains, a broad area in the southeast part of the QTP, and west and south of the northeast permafrost area. The degradation occurred mainly during the 1980s to 1990s in the northeast permafrost area and the Qilian Mountains, and during the 1990s to 2000s in most areas of the QTP. This is consistent with previous research (Jin et al., 2007; Marchenko et al., 2007; Cheng et al., 2012; Zhao et al., 2012).

One difference from previous studies at the national scale is that we found that degradation occurred not only in the transition regions between permafrost and seasonally frozen ground, but also and more importantly, it occurred in the interior of permafrost regions. From 1980 to 2010 the total degraded areas comprised 101.21×10⁴ km²; the degradation in the transition regions between permafrost and seasonally frozen ground was only 12.68×10⁴ km²or 13% of the total degraded areas. In the interior of permafrost regions, about 87.25×10⁴ km²was degraded by one stability level, about 10.7×10⁴km²was degraded by two stability levels, and 0.21×10⁴ km²was degraded by three stability levels. Also, approximately 3.05×10⁴ km²of seasonal frozen ground was changed to permafrost.

4.3 The implications of permafrost stability degradation

The degradation of permafrost stability will have systemic impacts on engineered infrastructures in the permafrost regions, the water balance, and the global carbon budget. With the degradation of permafrost stability, engineered infrastructures in permafrost zones will likely face a risk of increased deterioration and damage. The degradation of permafrost stability may also shift the global carbon budget. Tarnocai et al. (2009) reported that the northern permafrost region globally contains approximately 1, 672 Pg of organic carbon, making permafrost a very important carbon reservoir at the global scale. With climate warming, the spatial extent of permafrost declines and its stability is degraded, causing rapid carbon release to the atmosphere. Furthermore, the degradation of permafrost stability may also affect the hydrologic cycle in cold regions in two aspects. One is its effect on the interactions between surface water and subsoil water by changing the soil hydraulic conductivity, and the other is the ice-rich permafrost itself contributing to surface water. Overall, permafrost degradation affects the quantity of available water resources and its seasonal and annual distribution.

5 Summary

Based on a thermal stability-based permafrost classification system, the degradation of permafrost stability in China over the past 30 years is evaluated using a new, high-resolution near-surface air temperature reanalysis dataset, i.e., ITPCAS. Results showed that the permafrost extent on the QTP clearly decreased by 22% from 1980 to 2010, that is, a loss of 12.68×10⁴ km². The degradation occurred not only in the transition regions between permafrost and seasonally frozen ground, but also and more importantly, it occurred in the interior of permafrost regions. The degradation of permafrost stability will have systemic impacts on engineered infrastructures in permafrost regions, the water balance, and the global carbon budget.

This evaluation takes a macro view of the degradation of permafrost thermal stability in northeast China from 1980 to 2000 based on the empirical relationship between permafrost stability and MAAT proposed by Cheng (1984). This hopefully will be useful for assessments and designs to maintain the long-term stability of engineered infrastructures in the permafrost regions, and other applications. In the future, more quantitative models, such as the thermal dynamic model (Greshchev, 1982) and the evaluation model of permafrost thermal stability and thawing sensibility (Wu et al., 2002), should be integrated into a physical model to evaluate the thermal stability of permafrost at a finer resolution. Many factors, such as climate, vegetation, soil constitution, ice content, and terrain, should be integrated in a more subtle way.

Acknowledgments:

This study is supported by National Natural Science Foundation of China Projects (No. 41471359), the Youth Innovation Promotion Association of the Chinese Academy of Sciences (No. 2016375), and the Chinese Academy of Sciences Action Plan for West Development Project "Remote Sensing Data Products in the Heihe River Basin: Algorithm Development, Data Products Generation and Application Experiments" (No. KZCX2-XB3-15). We thank the anonymous reviewer for their extremely helpful comments on this paper.