The response of seasonal soil freeze depth to climate change has
repercussions for the surface energy and water balance, ecosystems, the
carbon cycle, and soil nutrient exchange. Despite its importance, the
response of soil freeze depth to climate change is largely unknown. This
study employs the Stefan solution and observations from 845 meteorological
stations to investigate the response of variations in soil freeze depth to
climate change across China. Observations include daily air temperatures,
daily soil temperatures at various depths, mean monthly gridded air
temperatures, and the normalized difference vegetation index. Results show
that soil freeze depth decreased significantly at a rate of
Combining multiple land and ocean surface temperature datasets, the global
mean air temperature increased 0.85
Due to global climate warming, significant efforts have been devoted to permafrost research, such as permafrost variations on the hemispheric-scale permafrost temperature changes (Wu and Zhang, 2008; Romanovsky et al., 2010; Guglielmin and Cannone, 2012; Streletskiy et al., 2014; Wu et al., 2015), permafrost degradation (Jorgenson et al., 2006; Ravanel et al., 2010; Sannel and Kuhry, 2011; Streletskiy et al., 2015a; Park et al., 2016), hydrological processes in permafrost regions (Hu et al., 2009; Wang et al., 2009; Park et al., 2013; Streletskiy et al., 2015b; Ford and Frauenfeld, 2016), feedbacks to climate change (Schuur et al., 2008; Park et al., 2015; Abbott et al., 2016), and other aspects. The increasing thickness of the active layer has been indicated by many observations in permafrost regions at high latitudes and altitudes (Brown et al., 2000; Frauenfeld et al., 2004; Zhang et al., 2005; Fyodorov-Davydov et al., 2008; Smith et al., 2010; Wu et al., 2010; Zhao et al., 2010; Callaghan et al., 2011; X. Li et al., 2012; Liu et al., 2014a, b; Stocker et al., 2014). Less research has focused on SFG areas (Zhang et al., 2003; Frauenfeld et al., 2004; Frauenfeld and Zhang, 2011; Wang et al., 2015), although the near-surface soil freeze–thaw status has been investigated using satellite passive microwave remote sensing (Zhang and Armstrong, 2001; Zhang et al., 2003, 2004; Li et al., 2008; Jin et al., 2015). Peng et al. (2016b) analyzed the response of soil freeze–thaw states to climate change across China, based on observational data. While Peng et al. (2016b) investigated the area extent changes of different soil freeze–thaw states, here we instead focus on seasonal SFD. Regional-scale SFD can be an important indicator of climate change and frozen ground condition in cold regions. Further, SFG is closely related with human activities, because most populated areas are located on SFG.
Shiklomanov (2012) similarly pointed out that SFG has not received much attention despite its vast area extent and importance, mainly due to a lack of long-term observational time series to document changes. Evaluating climatic and environmental changes on SFG requires comprehensive spatial assessments of available soil temperature records (Shiklomanov, 2012). To date, no comprehensive investigation of SFD in relation to climate change has been conducted in China despite the prevalence of SFG in this part of the world. Therefore, using long-term observational data, the goals and unique contributions of this study are (1) to estimate the spatial and temporal variations of seasonal SFD across China, (2) to quantify the potential forcing factors of SFD including climatic and environmental factors, and (3) to establish how SFD variability responds to climate change in China.
Several datasets are used including daily air and ground surface temperature, daily soil temperature at 0–320 cm depth, mean monthly gridded air temperature, and daily snow depth. In addition, we incorporate a 1 km resolution digital elevation model (DEM) and normalized difference vegetation index (NDVI) data. All datasets are described in detail below.
Mean daily air temperature and ground surface temperature data are collected
from the China Meteorological Administration (CMA) for a total of 839
meteorological stations (Fig. 1) available four times daily at 02:00,
08:00, 14:00, and 20:00 (
The observational station distribution across China, including the 839 stations with air and ground surface temperatures (green symbols), 845 soil temperature stations (red symbols), and elevation. The blue solid lines represent the main rivers.
Daily soil temperature data are available for 845 sites across China (Fig. 1) from the CMA, measured at the depths of 0.00, 0.05, 0.1, 0.15, 0.2, 0.4, 0.5, 0.8, 1.6, and 3.2 m. The temporal record varies for these stations, with some observations dating back to the late 1950s, and some only to the 1970s. Some station records end in the 1990s, while others are available through 2006 (Wang et al., 2015). Soil temperature is used to calculate the soil freeze depth; we combine the potential maximum soil seasonal freeze depth in permafrost regions, and the maximum soil freeze depth in SFG. The number of stations with both daily air temperature and soil temperature observations is 729.
Mean monthly gridded air temperature was used to analyze soil freeze
depth at the regional scale across China. We obtained the University of
Delaware's 1900–2014 terrestrial air temperature gridded monthly time
series (
Considering the complex terrain across China and the impacts of elevation on
air temperature, we also used the global 30 arcsec elevation dataset
(GTOPO30;
We obtained daily mean snow depth data for 672 sites across China (Che et al., 2008). The period of record at these locations varies, with some stations dating back to the late 1950s and some only to the 1970s. Some station records end around the 1990s while others are available through 2005. The snow depth was used to assess its influence on soil freeze depth. We calculate the annual maximum snow depth (SND) from the daily data for 1 July–30 June and match those snow depth stations with the soil temperature stations. If there are missing data in the spring, autumn, and winter season of one station, these station data will not be used.
The NDVI dataset used in this study is produced by the Global Inventory Modeling and Mapping Studies (GIMMS) team, available for 1982–2006. It is derived from NOAA AVHRR data, available at 15-day temporal resolution and an 8 km spatial resolution (Tourre et al., 2008). These data were used to assess the influence of vegetation on soil freeze depth. We extracted the NDVI values corresponding to the stations' latitude and longitude coordinates.
Missing data often present a potential problem for analyzing and averaging time series. Therefore, if fewer than 5 days were missing in a given month, filling in missing daily air temperatures was based on highly correlated neighboring sites using linear regression. Missing daily mean ground surface temperatures were estimated through linear regression with the daily mean air temperature at the same station. Based on the daily air temperature, we also calculate the mean monthly air temperature and mean annual air temperature (MAAT). The interpolated results are strongly correlated with observations, as indicated by regression coefficients larger than 0.95.
To improve the original 0.5
Linear least-squares regression between soil freeze depth and annual freezing index based on observational sites. The black solid line represents the linear regression.
The freezing and thawing index can also be an important indicator to assess the
variations in frozen ground (Zhang et al., 1997; Nelson, 2003; Frauenfeld et
al., 2007). There are two primary types of freezing and thawing indices: the
surface freezing and thawing index, calculated from ground surface temperatures,
and the air freezing and thawing index, computed from air temperatures. To
calculate the freezing index, we sum all temperatures below 0
Comparison of the simulated and observed SFD for all stations. The
black solid line is the
Spatial distribution and variability of SFD at the observing
stations.
Various methods are available to calculate the soil freeze depth. For
example, it can be estimated directly from soil temperature, from physical
and statistical models, and based on the Stefan solution. In this study, we
use the Stefan solution to estimate soil freeze depth, which is determined
using Eq. (3):
From the 1 km scale
A number of climatic and environmental variables including MAAT, mean annual ground surface temperature (MAGST), freezing index, thawing index, SND, and NDVI are selected to investigate the potential drivers of the observed long-term SFD changes across China. We use Pearson correlations to analyze the association between these variables and SFD and employ a 95 % significance level to assess the statistical significance for all analyses.
Figure 4 shows the spatial variability and trends of SFD at every location. The highest SFD was mainly located in northeastern and northwestern China and the Tibetan Plateau. In contrast, the lowest SFD was found in the south of China. Locations with SFD greater than 0.4 m are found north of the Yellow River. In the northwest of China, locations with SFD less than 0.8 m are found in the Taklimakan Desert, and some sites with SFD greater than 2.0 m are located in the Altai, Tianshan, and Pamir mountains.
On the Tibetan Plateau, most sites have a SFD greater than 2.4 m. There is
an increase in SFD with increasing latitude and elevation. The significant
SFD changes are between
The standard deviation of SFD at each site across China.
Figure 5 shows the standard deviation of SFD at each site across China. It
varies from 0.00 to 0.27 m. The standard deviation of SFD is generally less
than 0.03 m south of 35
Based on the sites'
1951–2012 SFD anomalies with respect to the 1971–2000 mean (red solid line) based on up to 839 stations across China as depicted in Fig. 1. Included also is the 1 standard deviation range (gray shading), the linear trend from 1967 to 2012 (blue dashed line), and the 7-year smoothing (green line). The inset shows the number of stations contributing to the time series.
Spatial pattern of multi-year mean SFD during 1950–2009 across China.
Based on the 1 km resolution
Spatial variability of SFD anomaly for the decades of the 1950s, 1960s, 1970s, 1980s, 1990s, and 2000s, with respect to the 1950–2009 mean across China.
Figure 9 represents the SFD trend across China from 1950 to 2009. The gray
region represents areas where the SFD trends are not statistically
significant, but they are statistically significant in all other
regions. In general, the SFD decreased significantly over northern China,
except in two small areas. The SFD trend ranges between 0.0 and
Overall, the spatial variability indicates that SFD changes with latitude and elevation at the regional scale across China. As is expected from climate warming, a statistically significant decreasing trend in SFD is evident across China from 1950 to 2009.
SFD trends across China from 1950 to 2009. The gray regions indicate non-significant SFD changes, while trends in all other regions are statistically significant.
To explore the possible variables leading to the documented changes in SFD,
we analyze potentially important factors for soil freeze dynamics: latitude,
altitude, MAAT, MAGST, freezing index including FI
To explore the spatial variability of SFD, we classify the meteorological
stations as either eastern or western based on 110
The relationship between SFD, latitude, and elevation in the east and west of China.
Temperature – including MAGST and MAAT – at the 839 station locations
exhibits a statistically significant increase over the 1951–2013 period of
0.019 and 0.013
SFD time series and trend (black) and the potential forcing
variables:
Soil freeze usually begins in autumn or winter, with temperatures less than
0
The thawing index is used to assess the accumulated positive degree days
during the warm season (Frauenfeld and Zhang, 2011). There are no obvious TI
changes at the station locations until approximately 1985. TI increases
during 1985–2008, followed by a decrease until 2013. From 1951 to 2013,
TI
Correlation between SFD and SND. The variables are standardized to range from 0 to 1.
Figure 12 shows the correlation between SFD and SND. However, the weak
negative correlation between SFD and SND of
As suggested by Shiklomanov (2012), environmental factors likely also affect
SFD. The surface can be affected directly by climate forcing, while the
subsurface effects are more complex. The subsurface soil only indirectly
receives a climatic signal, which is furthermore altered by site-specific
soil processes (e.g., thermal conductivity and analogous soil properties).
Vegetation is a likely environmental factor that influences the soil freeze
depth (Shiklomanov, 2012). Thus, we investigate vegetation using NDVI (Peng
et al., 2013) and find it is significantly correlated with SFD at
Correlation between SFD and mean annual NDVI.
Time series of SFD changes in different climate zones:
Soil freeze–thaw depth changes involve a series of interactions, such as energy exchanges, soil moisture exchanges, and gas exchanges between the atmospheric and terrestrial system. Therefore, variations of soil freeze–thaw most likely have an important effect on geomorphic, hydrological, and biological processes. Similarly, soil freeze–thaw depth changes also have destabilizing effects on engineering structures, such as on improperly constructed infrastructure (Smith and Burgess, 1999; Stendel and Christensen, 2002). The release of additional greenhouse gases to the atmosphere also occurs (Michaelson et al., 1996; Mu et al., 2015). In this paper, we use the Stefan method to calculate SFD, analyze the spatial SFD variability and trends, and quantify the potential driving factors affecting SFD.
SFD variability is susceptible to climate warming and environmental change and is affected by variables including air temperature, ground surface temperature, freezing and thawing index, and vegetation. Many examples of permafrost degradation have been reported, such as deeper the active layer thickness, reduced freeze time duration, and shifts in the timing of thawing and freezing in seasonally frozen ground regions (Henry, 2008; Callaghan et al., 2011; Stocker et al., 2014; Wang et al., 2015). Negative correlations are found here between SFD and temperature (including MAAT and MAGST) because of solar radiation heating the ground and energy transfer into the soil, ultimately increasing the soil temperature. Thus, increasing temperature is found to be the main factor influencing SFD variability in China, as in previous work focusing only on the Tibetan Plateau (Zhao et al., 2004).
The freezing and thawing indices represent the accumulated negative and positive degree days in the cold and warm seasons, respectively (Wu et al., 2011). The positive and negative correlation between SFD and FI and TI was statistically significant, consistent with previous results in other regions (Frauenfeld and Zhang, 2011). Due to the maximum soil freeze depth occurring in the cold season and SFD being affected by temperature, the positive correlation between SFD and FI is reasonable. Although TI is the accumulated temperature in the warm season, it takes some time to transfer the energy into the deeper ground. The energy flux into the soil reduces with increasing soil depth. Therefore, if all the conditions are the same, a larger TI can precondition the ground by increasing the energy in the deeper soil, which can subsequently delay soil freezing. Thus TI is a potential indicator of SFD, indirectly affecting soil temperature (Frauenfeld and Zhang, 2011).
Snow depth can have an effect on soil temperature, which would affect the active layer thickness and seasonal SFD variability. Numerical modeling studies have shown that snow depth does impact SFD (Zhang and Stamnes, 1998; Ling and Zhang, 2003; Park et al., 2015). Park et al. (2015) indicated that both increasing SND and snow structure (e.g., snow density) changes were favorable to soil warming, resulting in active layer thickness decreasing in northern regions as previously found by Frauenfeld et al. (2004). Snow cover insulates the ground during the cold season (Zhang, 2005). Interestingly, in our study we did not find a relationship between SND and SFD. This could be due to the spatial heterogeneity of snow across China. According previous research, snow depth, snow water equivalent, and snow densities are smallest on the Tibetan Plateau compared to other parts of China (Ma and Qin, 2012). Compared with other regions, multi-year average snow depth in general is low in China, especially on the Tibetan Plateau and the east-central mountain regions of China (Zhong et al., 2014), and may therefore have only limited insulating effects. This could lead to the lack of a relationship between SFD and SND across China and motivates future investigation.
A negative correlation between SFD and vegetation, as quantified by NDVI, is found. Vegetation change has a significant influence on the climate system mostly through changes to the surface radiative energy budget, which can be affected the SFD. Based on previous research, vegetation varies in different land cover types and responds to climate change via different physical mechanisms (Snyder et al., 2004), e.g., changes in the surface albedo (e.g., bare ground versus vegetation cover), vegetation transpiration, and shading effects (Kelley et al., 2004; Snyder et al., 2004; Swann et al., 2010; Chang et al., 2012; Zhang et al., 2012). In the cold season, less/decreased vegetation will be more easily snow covered, thus increasing the albedo considerably. Increasing albedo results in less net radiation at the land surface, as more incoming solar radiation is reflected from the surface. Then, the surface air temperature will decrease considerably due to less energy absorbed at the surface. For the colder land surface, the sensible heat flux is reduced. Further, the vegetation decrease results in reducing evapotranspiration, which decreases the latent heat flux (Snyder et al., 2004). Compared to increased vegetation cover, less vegetation causes a large annual-average increase in the surface albedo with the largest changes in the winter and spring seasons, which reduces the amount of net radiation at the surface, making the surface colder and resulting in SFD increases. Conversely, vegetation increases could lead to decreasing SFD. The vegetation's effect on transpiration is primarily important in summer, while SFD primary occurs in winter and spring (Snyder et al., 2004).
The inverse relationship between NDVI and SFD is in agreement with results from many previous studies that similarly found a vegetation increase, or a greening trend, in different regions during recent decades (Peng et al., 2011; Piao et al., 2011; Zhang et al., 2013; Zhu et al., 2016). Because climate change controls the spatial distribution of vegetation, most studies report vegetation increases as impacted by temperature and precipitation increases (Bao et al., 2015; Huang et al., 2016). Similarly, Fig. 11 shows that rising temperature results in a SFD decrease. The negative relationship between SFD and NDVI indicates the effect of vegetation on SFD as well as their inverse relationship.
SFD is affected by many factors, including the climatic and environmental variables considered in this study. However, SFD changes in different regions are also potentially influenced by many other local environmental variables or large-scale teleconnections (Frauenfeld and Zhang, 2011). Thus, it remains difficult to fully account for the spatial variations of SFD at the regional scale.
Our results indicate significant changes of SFD across China. To address the
spatial pattern of SFD changes, we divide the study area into five different
zones, including tropical monsoon (TPM), subtropical monsoon (SM), temperate
monsoon (TM), temperate continental (TC), and Qinghai–Tibetan alpine (QTA)
climate zones, which are categorized by temperature, precipitation, and
other parameters. Results indicate that the 30-year (1971–2000) average SFD
in the SM, TM, TC, and QTA climate zones are 2.8
In this study, we conducted a comprehensive regional-scale investigation of SFD across China. A significant climate indicator, SFD is influenced by many variables including climatic and environmental factors. These factors are often integrated to affect SFD (Lachenbruch and Marshall, 1986; Brown et al., 2000; Frauenfeld et al., 2004). Our results can be summarized as follows.
The spatial distribution of SFD variability is influenced by latitude and elevation across China. High-latitude and high-altitude sites are characterized by large SFD. In contrast, smaller SFD values are generally observed for lower latitude and lower elevation regions.
Of the total 839 sites, we find that the SFD decreased significantly, at
On the regional scale, the 1950–2009 spatial variation of SFD ranges
between 0.0 and 4.5 m across China, with most areas exhibiting significant
decreases between less than 0.0 and
The dataset of mean daily air, ground surface temperature, soil temperature, and snow depth from the China Meteorological Administration is not available for public use.
It can be accessed only by scientific researchers in China through the submission of an application. The website is
The authors declare that they have no conflict of interest.
This study was funded by the National Natural Science Foundation of China (grant nos. 91325202, 41601063, 41671516), the National Key Scientific Research Program of China (grant no. 2013CBA01802), and the Fundamental Research Funds for the Central Universities (lzujbky-2015-217). We acknowledge computing resources and time at the Supercomputing Center of Cold and Arid Region Environment and Engineering Research Institute of Chinese Academy of Sciences.Edited by: A. Kääb Reviewed by: E. Jafarov and one anonymous referee