Using ERA-Interim reanalysis dataset to assess the changes of ground surface freezing and thawing condition on the Qinghai–Tibet Plateau

  • Yanhui Qin
  • Tonghua WuEmail author
  • Ren Li
  • Wenjun Yu
  • Tianye Wang
  • Xiaofan Zhu
  • Weihua Wang
  • Guojie Hu
  • Liming Tian
Original Article


It is important to assess the freezing and thawing condition of ground surface for understanding the impacts of frozen ground on surface and subsurface hydrology, the surface energy and moisture balance, ecosystem conservation, and engineering construction on the Qinghai–Tibet Plateau (QTP). However, assessing the changes of ground surface freezing and thawing condition on the QTP still remains a challenge owing to data sparseness and discontinuous observations. The annual ground surface freezing index (GFI) and ground surface thawing index (GTI) could be used to predict changes of the thermal regime of permafrost and can be good indicators of climate change on the QTP, which has important engineering applications. In this study, we first calibrated the reanalysis ground surface temperature (GST) data using the methods of elevation correction on the QTP. After calibration, the quality of reanalysis data has been improved significantly. For the annual time series, the root mean square error decreased from 7.7 to 1.6 °C, the absolute value of mean bias error decreased from 7.5 to 0.0 °C, and the correlation coefficient increased from 0.62 to 0.86. Second, we estimated the annual and seasonal spatial distributions of GST. The spatial distribution of spring and autumn GST closely resembled the annual mean pattern. The long-term mean GFI and GTI from the calibrated reanalysis dataset were 1322.3 and 2027.9 °C/day, respectively. The GFI and GTI were presented as latitude and elevation zonation; it can also be seen that permafrost mostly occurred in the high GFI and low GTI regions. Estimating the GFI and GTI precisely will be utilized to model the permafrost distribution and estimate active layer thickness in the future.


Permafrost Freezing and thawing conditions ERA-Interim reanalysis dataset Qinghai–Tibet Plateau 


Ground surface is the interface for energy and moisture exchange between the ground surface and atmosphere in terms of sensible and latent heat fluxes (Li et al. 2014a, b). The ground surface temperature (GST) is considered as an important upper boundary condition parameter for modeling and predicting the changes of the thermal regime of permafrost (Wu et al. 2013). In permafrost regions, the annual ground surface freezing and thawing index calculated from GST could be used to predict and map permafrost distribution and to estimate active layer thickness (Nelson and Outcalt 1987; Nelson et al. 1997; Zhang 2005b; Zhang et al. 2005; Frauenfeld et al. 2007). These indices could also provide important information on climate variability and be used in seasonally frozen ground regions to identify snow type (Hinkel and Nelson 2003; Wang 2008; Ran et al. 2012, 2015). To simulate the spatial distribution of permafrost and estimate the active layer thickness, it is important to develop a dataset of the GST over Qinghai–Tibet Plateau (QTP). However, reanalysis products could contain uncertainties for various assimilation processes, and these biases would have significant influence on the models driven by these reanalysis data (Smith et al. 2001; Berg et al. 2003). It had been found that topographical correction may be one of the critical steps to assess the uncertainties of reanalysis air temperature, especially in the high mountains and the west of China (Frauenfeld 2005; Zhao et al. 2008; Wang et al. 2015). A simple correction method with a constant lapse rate has been used to significantly reduce elevation-related errors from reanalysis the air temperature dataset (Zhao et al. 2008; Bracegirdle and Marshall 2012; Jones and Lister 2014; Wang et al. 2015). It was necessary to process the elevation correction in the ECMWF soil temperature. The reanalysis soil temperature was more close to the observed after elevation correction in the United States and Europe (Albergel et al. 2015). Until now, there was little work of reanalysis datasets about GST on the region with complex topography. Therefore, it is necessary to estimate and calibrate the reanalysis data before applying the reanalysis dataset to models on the QTP.

Owing to the low latitude and extremely high elevation, the permafrost region on the QTP is unique. The permafrost area on the QTP amounts to approximately 1.35 × 106 km2 (Li and Cheng 1996; King et al. 2006). Based on station observations, remote sensing, and model simulations, previous studies demonstrated a rapid climate warming on the QTP over the past several decades (Glickman and Zenk 2000; Harris 2010; Hu et al. 2012; Wu et al. 2013; You et al. 2013; Niu et al. 2015; IPCC 2013; Ji and Kang 2015). Permafrost is an essential climate variable, which makes mapping and monitoring of its thermal state an important work. Because of the high elevation and complex terrain, QTP plays an important role on the climate patterns over East Asia, especially during the Asian monsoon process and the northern hemisphere atmospheric circulation, and it is also one of the most sensitive regions to recent climatic changes (Yanai and Li 1994; Feng et al. 1998; Wu and Zhang 1998; Elberling et al. 2010; Kang et al. 2010; Yang et al. 2014).

The complex topography, severe weather, and harsh environmental conditions of the QTP make it very difficult to obtain long-term observed meteorological variables (Ma et al. 2008). The spatial distribution of meteorological stations is uneven and mainly located in the eastern and central QTP. In addition, there are a few stations situated in the western and higher elevations over the QTP (>4000 m). However, the sparse and uneven distribution of meteorological stations over the QTP results in some uncertainties in evaluating climate changes (Song et al. 2015). The reanalysis products had widely been used to predict climatic trends and change in many areas (Li and Cheng 1996; Poveda et al. 2006; Wang et al. 2006; Fealy and Sweeney 2007; Jin et al. 2007; Ciccarelli et al. 2008; Stammerjohn et al. 2008). In recent decades, tremendous progress has been made in developing consistent long-term grid datasets (Gibson et al. 1997; Simmons and Gibson 2000; Kistler et al. 2001; Kanamitsu et al. 2002; Onogi et al. 2007; Berrisford et al. 2009). These products provided basic meteorological variables, such as tropospheric pressure heights, humidity, winds, air temperature, GST, precipitations, radiation fluxes, and so on. The comparison of air temperature obtained from observations and reanalysis datasets was reported on both regional and global scales (Frauenfeld 2005; Frauenfeld et al. 2007; Zhao et al. 2008; You et al. 2008, 2010a, 2010b, 2013; Wang and Zeng 2012). These assessments of reanalysis data against observations over the QTP mainly focused on surface pressure, precipitation, air temperature, and heat fluxes, while a few of them referring to the GST.

In this study, the reanalysis data from ERA-Interim and long-term meteorological observation data in GST over the QTP were compared. The calibrated reanalysis temperature series was compared with the raw data to assess the influence of elevation errors in the reanalysis datasets. The relationship between elevation errors and temperature bias was analyzed. In addition, the annual and seasonal reanalysis GST was calibrated by elevation correction. The calibrated reanalysis GST corresponds to the elevation of the model grid cells [the NASA/NGA Shuttle Radar Topography Mission (SRTM) digital elevation map (DEM) V4 with spatial resolution of 90 m]. Finally, the distribution of annual and seasonal GST, and ground surface freezing and thawing index over the QTP were estimated through a dataset of calibrated monthly reanalysis GST. The evaluation of freezing and thawing index would provide inputting data to simulate the active layer thickness and to map permafrost distribution over the QTP.

Data and methods

We used the observed daily GST data (dataset of the entire year) from the National Climate Center of CMA (China Meteorological Administration) ( The GST was measured on a standard yard with snow cover on the surface in the winter and with bare ground in the summer. In addition, data quality control and a homogeneity assessment were checked using the RClimDex software ( A total of 69 meteorological stations with continuous and high-quality records in or around the permafrost regions during the period of 1980–2013, which were all above 2000 m on the QTP, were selected in this study. The spatial distribution of meteorological stations is shown in Fig. 1.
Fig. 1

The locations of selected meteorological stations on the QTP

The ERA-Interim reanalysis data were produced by the ECMWF (Adrian Simmons et al. 2007). The inputting data included observations from aircraft, ocean-buoys, satellite-borne instruments, radiosonde, and other platforms, with a declining number of radiosonde ascents, since the late 1980s. It used 4D variational analysis on a spectral grid with triangular truncation of 255 waves (corresponding to approximately 80 km) and a hybrid vertical coordinate system with 60 levels. The ECMWF global model was used for forward integration in the 4D variational analysis, and the temporal length of the variation window was 12 h. The monthly temperature was obtained since 1979 with a spatial resolution of 0.125° (longitude) × 0.125° (latitude) (Dee et al. 2011). More detailed information about the data can be obtained from the previous literature (Berrisford et al. 2009). The ERA-Interim gridded monthly land surface temperature series dataset from 1980 to 2013 was obtained from the website

The surface elevation was derived from three datasets: (1) elevation of each meteorological station provided by the National Climate Center of the CMA, (2) ERA-Interim reanalysis model topography (available from, and (3) the NASA/NGA SRTM DEM V4 with spatial resolution of 90 m, which can be obtained from the website

To compare the reanalysis data with observation data, the observation data were converted from daily average to monthly average. The reanalysis data represented the value in a grid box centered on the geographic coordinate. It proved difficult to compare the station data with the reanalysis data, because the station data only represented a single site. The simplest method was to compare the value observed at the station with the reanalysis value for the grid box in which the station lied. Another approach was used in which the reanalysis value assigned for comparison with the station value was obtained from the weighted average of the reanalysis values of the four grid boxes whose center lied closest to the station. The average of the four grid boxes was obtained from the inverse distance weighted average (Jiang et al. 2008). You et al. (2013) further considered that the biases between these methods were lower for comparisons on the QTP, which suggested that it had no significant influence on the results. In this study, the simplest method was used to compare the reanalysis data with the observations, as the 69 stations occupied 69 grid boxes, respectively.

The root mean square error (RMSE) and mean bias error (MBE) were used to compare the observations and reanalysis GST. The RMSE and MBE were calculated as follows:
$$\it {\text{RMSE}} = \sqrt {\frac{{\sum\nolimits_{i = 1}^{n} {(T_{i} - t_{i} )^{2} } }}{n}}$$
$$\it \it {\text{MBE}} = \frac{{\sum\nolimits_{i = 1}^{n} {(T_{i} - t_{i} ) } }}{n}$$
where T i and t i were reanalysis (calibrated) and observed seasonal (annual) mean GST, respectively. The monthly temperature was calculated on an annual and seasonal basis, i.e., spring, summer, autumn, and winter.
In this study, the ground surface freezing index (GFI) and ground surface thawing index (GTI) were computed based on the method provided by Frauenfeld et al. (2007) and Wu et al. (2011). The GFI/GTI was determined as the sum of the daily GST below/exceeding 0 °C, which was calculated using the following equations (Frauenfeld et al. 2007). To include the entire progress of freezing and thawing cycling, the GFI was calculated from July 1 to the next June 30, and the GTI was calculated from January 1 to December 30. The GFI and GTI over the QTP were calculated using the following equations (Frauenfeld et al. 2007):
$$ \begin{aligned} {\text{GFI}} = \int_{{{\text{t}}1}}^{{{\text{t}}2}} {\left| T \right|{\text{d}}t\;T < 0\;^\circ {\text{C}}} \hfill \\ = \sum\limits_{i = 1}^{Nf} {\left| {T_{i} } \right|} \;T_{i} < 0\;^\circ {\text{C}} \hfill \\ \end{aligned}$$

Equation (3) was the theoretic equation for calculating the GFI, where GFI denoted the ground surface freezing index. The GFI were integrated from the beginning (t1) to the end (t2) of the cold season. The expression 2 of Eq. (3) was the empirical formula for calculating the GFI, where i = 1, 2, 3, … N f referred to the days with the temperature falling below 0 °C; T i denoted the daily mean temperature.

A similar method was used to calculate the GTI. T i was integrated from the beginning (t1) to the end (t2) of the warm season, which was the sum of daily temperature exceeding 0 °C:
$$ \begin{aligned} GTI = \int_{{{\text{t}}1}}^{{{\text{t}}2}} {\left| T \right|{\text{d}}t\;T < 0\;^\circ {\text{C}}} \hfill \\ = \sum\limits_{i = 1}^{Nt} {\left| {T_{i} } \right|} \;T_{i} < 0\;^\circ {\text{C}} \hfill \\ \end{aligned}$$
where i = 1, 2, 3,… Nt was the number of days with temperature exceeding 0 °C.

The trend of GTI and GFI was conducted by the linear regression statistics, and the P value was also used to determine the significance level. If the P value was small (e.g., less than 0.05), it meant that the correlation was not a coincidence.


The correlations between ERA-Interim reanalysis dataset and observations

The correlations between reanalysis dataset and observations in the seasonal and annual mean GST were shown in Table 1. Overall, there were weak correlations between the annual GST from the ERA-Interim reanalysis dataset and observations. The strongest correlation of the two datasets was in winter, with a correlation of 0.73, and the weakest correlation appeared in spring, with the value of 0.56. As shown in Fig. 2, there was a cold bias between the reanalysis dataset and the observations. The RMSE at four stations from the reanalysis dataset was less than 2 °C (5.8 %), and the RMSE at most of the stations was greater than 2 °C (94.2 %). There were 15 stations with an RMSE greater than 10 °C, accounting for 14.5 % of all of the stations (Fig. 2a). The average value of the RMSE in all the stations was 7.7 °C, whereas the MBE at two stations were within ±2 °C, and the MBE at most of stations were lower than −4 °C. The average value of MBE in all the stations was −7.5 °C (Fig. 2b). Moreover, there were 15 stations whose MBE was less than −10 °C, which accounted for 14.5 % of the total number of meteorological stations.
Table 1

Comparison of annual and seasonal RMSE, MBE, and slope of linear regression between raw and calibrated ERA-Interim reanalysis GST from 1980 to 2013 over the QTP

Reanalysis data


Mean temperature of observations (°C)

Mean temperature of calibrated reanalysis data (°C)

Raw RMSE (°C)

Calibrated RMSE (°C)

Raw MBE (°C)

Calibrated MBE (°C)

Raw slope

Calibrated slope















































Fig. 2

Distribution of annual RMSE (a) and MBE (b) for the ERA-Interim reanalysis GST over the QTP from 1980 to 2013

The spatial distribution of elevation errors between the reanalysis and observations was shown in Fig. 3. There were only two stations whose elevation in reanalysis dataset was lower than that of observations (Fig. 3a). The maximum elevation error at the observation stations was up to 1310.0 m, and the mean value was 560.8 m. The elevation errors over the QTP suggested that larger positive elevation errors can be found in the southeastern part of the QTP, the southern margin of the Qaidam Basin, and the western Kunlun region, whereas the negative elevation errors were distributed over the margin of the QTP, especially in the Yarlung Zangbo River and the Himalayas regions. Therefore, the errors of reanalysis GST could result from the difference between the reanalysis dataset and the observations.
Fig. 3

Spatial distribution of elevation errors, which was calculated by using reanalysis minus observation over the QTP (a observed sites and b SRTM DEM)

The relationship between the elevation error and the mean annual MBE was shown in Fig. 4. The negative correlation can be found for annual series in different elevation errors (Fig. 4a–c). Therefore, the positive elevation errors would correspond to negative MBE. Thus, the negative MBE for 98.5 % of stations had positive elevation errors, which were shown in Figs. 2b and 3a. When the elevation of the 14 stations was between 2000 and 3000 m, the slope of linear regression on an annual basis was −0.50 °C/100 m, and the correlation coefficient was −0.78 (Fig. 4a). However, when the elevation of the 36 stations ranged from 3000 to 4000 m, the slope of linear regression of annual basis was −0.62 °C/100 m, and the correlation coefficient was −0.73 (Fig. 4b). The elevation of 19 stations exceeded 4000 m, the slope of linear regression on annual basis was −0.69 °C/100 m, and the correlation coefficient was −0.93 (Fig. 4c). All the correlation coefficients (P < 0.01) were shown in Fig. 4.
Fig. 4

Correlation between annual MBE and elevation errors (reanalysis minus observation) for the ERA-Interim reanalysis dataset over the QTP from 1980 to 2013 (a 2000 m < elevation ≤ 3000 m, b 3000 m < elevation ≤ 4000 m, and c elevation >4000 m)

The calibration of ERA-Interim GST dataset

According to the relationship between MBE and elevation error, the reanalysis GST was calibrated through elevation corrections. The correlations between the calibrated reanalysis GST and the observations were shown in Fig. 5. After calibration, the fitting lines were much closer to the diagonal lines. The detailed information about the raw and calibrated ERA-Interim reanalysis GST was listed in Table 1.
Fig. 5

Correlations between seasonal and annual GST derived from observations and calibrated reanalysis dataset for each station over the QTP from 1980 to 2013. The dotted line was diagonal line of equality, and R stood for correlation coefficient and P for statistical significance

After calibration, the ERA-Interim reanalysis GST agree well with the observations. For the annual series, the RMSE has decreased from 7.7 to 1.6 °C, and the absolute value of MBE has decreased from −7.5 to 0.0 °C. The correlation coefficient of the annual reanalysis data also has increased from 0.62 to 0.86. The calibrated reanalysis GST were located at the right of the diagonal line in some stations, and to the left of the diagonal line in other stations. The mean annual MBE of calibrated data was close to 0.0 °C.

In addition, the slope of linear regression derived by calibrated reanalysis data was closer to 1 than that derived by the raw reanalysis data. Similar to the annual time series, the seasonal time series of GST have also been improved greatly after calibration. The calibration of GST performed better in winter and autumn than in spring and summer. The RMSE showed a greater improvement after calibration at most stations. After calibration, the annual RMSE has decreased on a large scale. The RMSE at 55 stations was less than 2 °C (Fig. 6). The topographical correction had limited influence on the spring calibrated GST, which may be because the influence factors to the reanalysis GST in the south of QTP was complicated and changed significantly in a reanalysis grid.
Fig. 6

Distribution of seasonal and annual RMSE of raw (the 1st column) and calibrated (the 2nd column) ERA-Interim reanalysis GST on the QTP from 1980 to 2013

Before calibration, the stations, whose RMSE were less than 2 °C in spring, summer, autumn, and winter, occupied 1.4, 1.4, 1.4, and 0 %, respectively. After calibration, the stations, whose RMSE were less than 2 °C in spring, summer, autumn, and winter, accounted for 56.5, 59.4, 68.1, and 52.2 %, the decreased proportion of RMSE in spring, summer, autumn, winter, and annual was 98.6, 97.1, 97.1, 92.8, and 97.1 %, respectively. The increased proportion of MBE in spring, summer, autumn, winter, and annual was all 98.6 %, as shown in Fig. 7. The annual MBE ranged from −12.8 to 1.9 °C before elevation correction, but the calibrated GST ranged between −4.8 and 7.6 °C. A total of 59 stations (85.5 %) had an annual MBE between ±2 °C after calibration.
Fig. 7

Distribution of seasonal and annual MBE of raw (the 1st column) and calibrated (the 2nd column) ERA-Interim reanalysis GST on the QTP from 1980 to 2013

The annual and seasonal GST climatology on the QTP

We examined the calibrated reanalysis annual and seasonal GST climatology over the QTP. The estimated annual GST spanned a wide range over the QTP, from −15 °C in the western Kunlun to 20 °C in the south Tibet. The regions, whose annual GST was below 0 °C, approximately correspond to the permafrost regions. The spring and autumn climatology closely resemble the annual mean pattern (Fig. 8a, c, e).
Fig. 8

Seasonal and annual GST climatology of the QTP from 1980 to 2013 (the climatology was the average of 1980–2013) (a spring, b summer, c autumn, d winter, and e annual)

The winter GST was below −8 °C in the central of the QTP, higher than −8 °C in the margin of the QTP, and was around −6 °C in the desert regions of Qaidam Basin. In summer, the GST rose above 0 °C on the QTP. The ground surface in the central of the QTP warmed up to between 6 and 15 °C, and the ground surface in the margin of the QTP warmed up to 15 °C, and the desert regions in the Qaidam Basin warmed up to approximately 20 °C.

We calculated the GFI and GTI using the calibrated reanalysis monthly GST dataset. The long-term mean GFI from the calibrated reanalysis dataset amounted to 1322.3 °C/day. The climatology of the GFI was shown in Fig. 9a. The long-term climatology of GFI strongly resembles the cold season temperature climatology of the QTP (Figs. 8d, 9a). The mean GST was lower in the western QTP than in the eastern QTP because of the higher elevations in the western region. When the elevation was above 4000 m, the GFI was greater than 1500 °C/day. When the elevation was above 5000 m, the GFI exceeded 2000 °C/day. The maximum value of the GFI exceeded 3000 °C/day, which was located in the western Kunlun area. The Qilian Mountain, located at the northeast QTP, was also one of the high value centers of GFI.
Fig. 9

Climatology of the GFI (a) and GTI (b) from 1980 to 2013 (the climatology was the average of 1980–2013)

Similarly, the long-term climatology of the mean GTI from the calibrated ERA-Interim reanalysis dataset was 2027.9 °C/day. The climatology of GTI was reminiscent of the long-term warm season temperature climatology (Figs. 8a, c, 9b), with high values corresponding to warm areas on the QTP. The GTI also depended on the elevation and latitude, namely, the GTI declining with the rising elevation and the northward latitude. The GTI was the lowest in regions whose elevation was above 5000 m. The climatology of the GTI also corresponded to the climatology of the active layer thickness in the permafrost region. The GFI and GTI were affected by vegetation, snow cover, topography, microclimate environment, as well as by elevation and latitude (Wu et al. 2013; Luo et al. 2014).

The GTI showed a statistically significant positive trend of 3.75 °C/day/year (P < 0.01) for the period 1980–2013 (Fig. 10). However, the GFI did not show a significantly negative trend. There was no significant declining trend of GFI after the 1990s, which was opposite of the global warming hiatus. This result was consistent with the fact that the significant winter warming after the 1990s on the QTP has been occurring (Duan and Xiao 2015). The GFI and GTI were relatively low in 1997, which probably resulted from the heavy snow observed from September to December (Yang et al. 2010; Wu et al. 2013).
Fig. 10

Time series of the GTI and GFI on the QTP from 1980 to 2013. The red bold lines denote the 5-year running average of GST


On the QTP, most of the meteorological stations (50 stations) were lower than 4000 m, while the height of the assimilation model of the reanalysis data was above 4000 m (Frauenfeld 2005; Dee et al. 2011). The elevation differences between each station and the corresponding grid cell were great. Most of meteorological stations were located on the flat ground or in the mountain valleys. The GST integrated the effect of climatic elements, such as air temperature and seasonal snow cover, and interactions with ground surface characteristics, such as vegetation, surface soil texture, and terrain. The observation of the GST was measured on a standard yard of snow cover in winter and bare ground in summer. It was known that the rainy season on QTP usually occurred from May to October, and at the same time, the active layer above the permafrost began melting and vegetation started to grow. In general, the warming or cooling effect of the snow cover on the ground surface was different on a daily, monthly, and annual basis (Zhang 2005a; Qin et al. 2006). On the QTP, the high vegetation coverage lowered the ground surface temperature and reduced the seasonal thaw depth (Zhou et al. 2000; Jin et al. 2008). Snow cover and vegetation were the main factors for the thermal offset of the air temperature and GST.

The climatology derived from the long-term meteorological observations, and the ERA-Interim reanalysis GST were compared. Reanalysis data were lower than the observations; the average RMSE and MBE of all of the stations were 7.7 and −7.5 °C, respectively. The output of reanalysis data represented an average temperature for a grid box, while the observation at meteorological station represented only one point. In addition, the temporal windows between the different data sources were inconsistent (Zhao et al. 2008; Song et al. 2015). The R between the calibrated ERA-Interim GST and the observation in spring was relatively less than that in other seasons and for annual average. The greater MBE occurred in the south and southeast of QTP, which could result from the melting of thicker snow cover in spring in those regions (Zhang et al. 2004).

There were different corrective factors at each elevation gradient (Fig. 4). Previous studies applied a correction method with a constant lapse rate, and the elevation-related errors have been reduced significantly from the reanalysis air temperature dataset (Bracegirdle and Marshall 2012; Jones and Lister 2014; Wang et al. 2015). The raw reanalysis air temperature being cooler than the temperature in the observations was consistent with other studies (Frauenfeld 2005; Zhao et al. 2008; You et al. 2010a, b). Therefore, the elevation correction was necessary if the reanalysis GST was applied as the input data of permafrost distribution models.

Although Guglielmin et al. revealed that a very close relationship existed between the GST and air temperature (Guglielmin 2006), the air temperature was unable to simply replace the GST in permafrost studies. However, the lack of a sufficient and reliable GST dataset was one of the most important limitations for permafrost modeling, so it was significant to develop an exact dataset to map the permafrost distribution over the QTP. Most previous studies focused on the estimation and application of reanalysis air temperature.

There may be some uncertainties using the reanalysis GST to calculate spatial distribution of GFI and GTI on the QTP. The uncertainty test was used the observed GST to validate the results. The GFI and GTI were calculated by the calibrated reanalysis monthly GST, and the CMA observed GST. The RMSE between the calibrated and observed GFI and GTI was shown in Fig. 11. For most of the sites, the RMSE of GFI was less than GTI. The RMSE of GFI and GTI was less than 730.0 °C/day (Fig. 11). The higher RMSE of the GTI can be found in the southern and southeastern region of the QTP, which had a large amount of glaciers. The complex underlying surface may increase the uncertainty on the reanalysis data cells, which was averaged in an area of a grid (0.125° × 0.125°). The central regions of the QTP are underlain by continuous permafrost, which had a relative uniform underlying surface in one reanalysis grid, the calibrated effect of which was better than that of other regions (Fig. 11). Therefore, it is reliable to use the calibrated reanalysis GST to map the permafrost distribution and calculate the active layer thickness on the QTP.
Fig. 11

Distribution of annual RMSE for the calibrated reanalysis GFI (a) and GTI (b) over the QTP from 1980 to 2013

We provided a calibrated monthly reanalysis GST dataset to estimate the climatology of the long-term GFI and GTI from 1980 to 2013. The long-term mean GFI and GTI from the calibrated reanalysis dataset were similar to the observations in the central region of the QTP (Wu et al. 2013). It can be seen that permafrost mostly occurred in the high GFI and low GTI regions (Figs. 1, 9). The mean annual air temperature was often used to determine the southern boundary of the permafrost (Jin et al. 2007; Ran et al. 2012). The isotherms of mean annual air temperature had been determined to agree well with the QTP observation-based permafrost maps (LIGG 1988; Li and Cheng 1996; Zhou et al. 2000; CAREERI 2006; Ran et al. 2012). However, some literature suggested that the air temperature was not an optimal predictor for permafrost distribution because of the effect of snow cover and vegetation.

However, permafrost and permafrost-free areas coexist in the transition elevation from 4000 to 5000 m, according to the map of permafrost distribution released in 1996 (Li and Cheng 1996). The GFI and GTI were important parameters to model permafrost distribution. The results showed that permafrost mostly developed in the high GFI and low GTI regions. The calibrated reanalysis GST reflects recent summer warming on the QTP (Fig. 10). The GTI has been widely used, because there was a close relationship between the GTI and active layer thickness (Romanovsky and Osterkamp 1997; Nelson et al. 1998; Zhang et al. 2005). The statistically significant positive trend of GTI was 3.75 °C/day/year (P < 0.01) for the period 1980 to 2013 (Fig. 10), which indicated that the calibrated reanalysis GST data can be used to assess the changes of the active layer thickness. The GFI and GTI could be reliable for modeling the thermal regime of permafrost and predicting permafrost distribution. The calibration of topographical errors was necessary to evaluate the uncertainties of reanalysis surface temperature, especially applied in the studies on climate changes.


It is of great significance to develop a dataset of the GST over the QTP to simulate permafrost distribution and to estimate the active layer thickness. We calibrated the reanalysis data using the methods of error corrections on the QTP. After calibration, the annual and seasonal time series of the GST had been improved greatly in the RMSE, MBE, and R. For the mean annual series, the RMSE has decreased from 7.7 to 1.6 °C, and the MBE has increased from −7.5 to 0.0 °C. The correlation coefficient has also increased from 0.62 to 0.86. And after calibration, the decreased proportion of RMSE for GST in spring, summer, autumn, winter, and annual was 98.6, 97.1, 97.1, 92.8, and 97.1 %, respectively. The increased proportion of the MBE for GST in spring, summer, autumn, winter, and annual all amounted to 98.6 % after calibration. The calibrated ERA-Interim reanalysis dataset showed that the annual mean GST across the central QTP was below 0 °C. In addition, the 0 °C GST isotherms agreed well with the boundary of permafrost on the QTP. The spatial distribution of spring and autumn GST closely resembled the annual mean pattern. The long-term mean GFI and GTI from the calibrated reanalysis dataset were 1322.3 and 2027.9 °C/day, respectively. It was shown that the calibrated reanalysis GST can provide a higher resolution for monitoring the thermal state of permafrost on the QTP. The statistically significant positive trend of GTI from 1980 to 2013 was consistent with the recent climate changes on the QTP. Therefore, the calibrated reanalysis GST data could be used to reflect the changes of the active layer thickness.



This research is supported by the National Natural Science Foundation of China (41271086; 41271081; 41421061), and the Hundred Talents Program of Chinese Academy of Sciences (51Y251571, 51Y551831). The authors also thank the National Climate Center, ECMWF and NASA for providing the data for this study.


  1. Albergel C, Dutra E, Muñoz-Sabater J, Haiden T, Balsamo G, Beljaars A, Isaksen L, de Rosnay P, Sandu I, Wedi N (2015) Soil temperature at ECMWF: an assessment using ground-based observations. J Geophys Res Atmos 120(4):1361–1373CrossRefGoogle Scholar
  2. Berg AA, Famiglietti JS, Walker JP, Houser PR (2003) Impact of bias correction to reanalysis products on simulations of North American soil moisture and hydrological fluxes. J Geophys Res Atmos (1984–2012) 108(D16):4490. doi: 10.1029/2002JD003334
  3. Berrisford P, Dee D, Fielding K, Fuentes M, Kallberg P, Kobayashi S, Uppala S (2009) The ERA-Interim archive. ERA Rep Ser 1:1–16Google Scholar
  4. Bracegirdle TJ, Marshall GJ (2012) The reliability of Antarctic tropospheric pressure and temperature in the latest global reanalyses. J Clim 25(20):7138–7146CrossRefGoogle Scholar
  5. CAREERI (Cold and Arid Regions Environmental and Engineer Research Institute) (2006) Map of the glaciers, frozen ground and deserts in China. Beijing: Sino Maps Press scale 1:4,000,000 (in Chinese)Google Scholar
  6. Ciccarelli N, Von Hardenberg J, Provenzale A, Ronchi C, Vargiu A, Pelosini R (2008) Climate variability in north-western Italy during the second half of the 20th century. Glob Planet Change 63(2):185–195CrossRefGoogle Scholar
  7. Dee DP, Uppala SM, Simmons AJ, Berrisford P, Poli P, Kobayashi S, Vitart F (2011) The ERA-Interim reanalysis: configuration and performance of the data assimilation system. Q J R Meteorol Soc 137(656):553–597. doi: 10.1002/qj.828 CrossRefGoogle Scholar
  8. Duan A, Xiao Z (2015) Does the climate warming hiatus exist over the Tibetan Plateau? Sci Rep 5. doi: 10.1038/srep13711
  9. Elberling B, Christiansen HH, Hansen BU (2010) High nitrous oxide production from thawing permafrost. Nat Geosci 3(5):332–335CrossRefGoogle Scholar
  10. Fealy R, Sweeney J (2007) Statistical downscaling of precipitation for a selection of sites in Ireland employing a generalised linear modelling approach. Int J Climatol 27(15):2083–2094. doi: 10.1002/joc.1506 CrossRefGoogle Scholar
  11. Feng S, Tang M, Wang D (1998) New evidence for the Qinghai-Xizang (Tibet) Plateau as a pilot region of climatic fluctuation in China. Chin Sci Bull 43(20):1745–1749CrossRefGoogle Scholar
  12. Frauenfeld OW (2005) Climate change and variability using European Centre for Medium-Range Weather Forecasts reanalysis (ERA-40) temperatures on the Tibetan Plateau. J Geophys Res. doi: 10.1029/2004jd005230 Google Scholar
  13. Frauenfeld OW, Zhang T, McCreight JL (2007) Northern Hemisphere freezing/thawing index variations over the twentieth century. Int J Climatol 27(1):47–63. doi: 10.1002/joc.1372 CrossRefGoogle Scholar
  14. Gibson R, Kallberg P, Uppala S, Hernandez A, Nomura A, Serrano E (1997) ERA description. Re-Analysis Project Report Series No. 1, European Centre for Medium-Range Weather Forecasts (ECMWF). Reading, UKGoogle Scholar
  15. Glickman TS, Zenk W (2000) Glossary of meteorology. American Meteorological Society, Boston, Mass, pp 1–638Google Scholar
  16. Guglielmin M (2006) Ground surface temperature, active layer and permafrost monitoring in continental Antarctica. Permafrost Periglac Process 17(2):133–143CrossRefGoogle Scholar
  17. Harris RB (2010) Rangeland degradation on the Qinghai-Tibetan plateau: a review of the evidence of its magnitude and causes. J Arid Environ 74(1):1–12. doi: 10.1016/j.jaridenv.2009.06.01 CrossRefGoogle Scholar
  18. Hinkel KM, Nelson FE (2003) Spatial and temporal patterns of active layer thickness at Circumpolar Active Layer Monitoring (CALM) sites in northern Alaska, 1995–2000. J Geophys Res Atmos. doi: 10.1029/2001jd000927 Google Scholar
  19. Hu G, Dong Z, Lu J, Yan C (2012) Driving forces responsible for aeolian desertification in the source region of the Yangtze River from 1975 to 2005. Environ Earth Sci 66(1):257–263CrossRefGoogle Scholar
  20. IPCC (2013) Climate Change 2013: the physical science basis. In: Stocker TF, Qin D, Plattner GK, Tignor M, Allen SK, Boschung J, Nauels A, Xia Y, Bex V, Midgley PM (eds) Contribution of working group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, UK and New York, NY, USA, pp 1–1535Google Scholar
  21. Ji Z, Kang S (2015) Evaluation of extreme climate events using a regional climate model for China. Int J Climatol 35(6):888–902. doi: 10.1002/joc.4024 CrossRefGoogle Scholar
  22. Jiang F, Hu R, Li Z (2008) Variations and trends of the freezing and thawing index along the Qinghai-Xizang Railway for 1966–2004. J Geog Sci 18(1):3–16. doi: 10.1007/s11442-008-0003-y CrossRefGoogle Scholar
  23. Jin H, Yu Q, Lü L, Guo D, He R, Yu S, Li Y (2007) Degradation of permafrost in the Xing’anling Mountains, Northeastern China. Permafr Periglac Process 18(3):245–258CrossRefGoogle Scholar
  24. Jin H, Wei Z, Wang S, Yu Q, Lv L, Wu Q, Ji Y (2008) Assessment of frozen-ground conditions for engineering geology along the Qinghai-Tibet highway and railway, China. Eng Geol 101(3):96–109CrossRefGoogle Scholar
  25. Jones PD, Lister DH (2014) Antarctic near-surface air temperatures compared with ERA-Interim values since 1979. Int J Climatol 35:1354–1366CrossRefGoogle Scholar
  26. Kanamitsu M, Ebisuzaki W, Woollen J, Yang S-K, Hnilo J, Fiorino M, Potter G (2002) NCEP-DOE AMIP-II Reanalysis (R-2). Bull Am Meteorol Soc 83(11):1631–1643CrossRefGoogle Scholar
  27. Kang S, Xu Y, You Q, Flügel W-A, Pepin N, Yao T (2010) Review of climate and cryospheric change in the Tibetan Plateau. Environ Res Lett 5(1):015101CrossRefGoogle Scholar
  28. King L, Herz T, Hartmann H, Hof R, Jiang T, Ke C, Yi C (2006) The PACE monitoring strategy: a concept for permafrost research in Qinghai-Tibet. Quatern Int 154:149–157. doi: 10.1016/j.quaint.2006.02.017 CrossRefGoogle Scholar
  29. Kistler R, Collins W, Saha S, White G, Woollen J, Kalnay E, Kousky V (2001) The NCEP-NCAR 50-year reanalysis: monthly means CD-ROM and documentation. Bull Am Meteorol Soc 82(2):247–267CrossRefGoogle Scholar
  30. Li SD, Cheng GD (1996) Map of permafrost on the Qinghai-Tibet Plateau Lanzhou: Gansu Culture Press scale 1:3,000,000 (in Chinese)Google Scholar
  31. Li R, Zhao L, Wu T, Ding Y, Xiao Y, Jiao Y, Qin Y, Xin Y, Du E, Liu G (2014a) Investigating soil thermodynamic parameters of the active layer on the northern Qinghai-Tibetan Plateau. Environ Earth Sci 71(2):709–722CrossRefGoogle Scholar
  32. Li R, Zhao L, Wu T, Ding Y, Xiao Y, Hu G, Zou D, Li W, Yu F, Jiao Y, Qin Y (2014b) The impact of surface energy exchange on the thawing process of active layer over the northern Qinghai-Xizang Plateau, China. Environ Earth Sci 72(6):2091–2099CrossRefGoogle Scholar
  33. LIGG (Lanzhou Institute of Glaciology and Geocryology Sciences) (1988) Map of snow, ice and frozen ground in China Beijing: Cartographic Publishing House scale 1:4,000,000 (in Chinese)Google Scholar
  34. Luo D, Jin H, Jin R, Yang X, Lü L (2014) Spatiotemporal variations of climate warming in northern Northeast China as indicated by freezing and thawing indices. Quatern Int 349:187–195CrossRefGoogle Scholar
  35. Ma L, Zhang T, Li Q, Frauenfeld OW, Qin D (2008) Evaluation of ERA-40, NCEP-1, and NCEP-2 reanalysis air temperatures with ground-based measurements in China. J Geophys Res Atmos (1984–2012) 113(D15). doi: 10.1029/2007JD009549
  36. Nelson FE, Outcalt SI (1987) A computational method for prediction and regionaliztion of permafrost. Arct Alp Res 19(3):279–288. doi: 10.2307/1551363 CrossRefGoogle Scholar
  37. Nelson F, Shiklomanov N, Mueller G, Hinkel K, Walker D,Bockheim J (1997) Estimating active-layer thickness over a large region: Kuparuk River basin, Alaska, USA. Arctic and Alpine Research, pp 367–378Google Scholar
  38. Nelson FE, Hinkel KM, Shiklomanov NI, Mueller GR, Miller LL, Walker DA (1998) Active-layer thickness in north central Alaska: Systematic sampling, scale, and spatial autocorrelation. J Geophys Res 103(28):963–973Google Scholar
  39. Niu F, Lin Z, Lu J, Luo J, Wang H (2015) Assessment of terrain susceptibility to thermokarst lake development along the Qinghai-Tibet engineering corridor, China. Environ Earth Sci 73(9):5631–5642CrossRefGoogle Scholar
  40. Onogi K, Tsutsui J, Koide H, Sakamoto M, Kobayashi S, Hatsushika H, Takahashi K (2007) The JRA-25 reanalysis. J Meteorol Soc Jpn 85(3):369–432CrossRefGoogle Scholar
  41. Poveda G, Waylen PR, Pulwarty RS (2006) Annual and inter-annual variability of the present climate in northern South America and southern Mesoamerica. Palaeogeogr Palaeoclimatol Palaeoecol 234(1):3–27. doi: 10.1016/j.palaeo.2005.10.031 CrossRefGoogle Scholar
  42. Qin D, Liu S, Li P (2006) Snow cover distribution, variability, and response to climate change in western China. J Clim 19(9):1820–1833CrossRefGoogle Scholar
  43. Ran Y, Li X, Cheng G, Zhang T, Wu Q, Jin H, Jin R (2012) Distribution of permafrost in China: an overview of existing permafrost maps. Permafrost Periglac Process 23(4):322–333. doi: 10.1002/ppp.1756 CrossRefGoogle Scholar
  44. Ran Y, Li X, Jin R, Guo J (2015) Remote sensing of the mean annual surface temperature and surface frost number for mapping permafrost in China. Arct Antarct Alp Res 47(2):255–265CrossRefGoogle Scholar
  45. Romanovsky V, Osterkamp T (1997) Thawing of the active layer on the coastal plain of the Alaskan Arctic. Permafrost Periglac Process 8(1):1–22CrossRefGoogle Scholar
  46. Simmons AJ, Gibson JK (2000) The ERA-40 Project Plan, ERA-40 Project Report Series No. 1 ECMWF. Shinfield Park. Reading, UK, 63Google Scholar
  47. Simmons A, Uppala S, Dee D, Kobayashi S (2007) ERA-Interim: new ECMWF reanalysis products from 1989 onwards. ECMWF Newsl 110(110):25–35Google Scholar
  48. Smith SR, Legler DM, Verzone KV (2001) Quantifying uncertainties in NCEP reanalyses using high-quality research vessel observations. J Clim 14(20):4062–4072CrossRefGoogle Scholar
  49. Song C, Ke L, Richards KS, Cui Y (2015) Homogenization of surface temperature data in High Mountain Asia through comparison of reanalysis data and station observations. Int J Climatol. doi: 10.1002/joc.4403 Google Scholar
  50. Stammerjohn SE, Martinson DG, Smith RC, Iannuzzi RA (2008) Sea ice in the western Antarctic Peninsula region: spatio-temporal variability from ecological and climate change perspectives. Deep Sea Res Part II 55(18):2041–2058. doi: 10.1016/j.dsr2.2008.04.026 CrossRefGoogle Scholar
  51. Wang XL (2008) Penalized maximal F test for detecting undocumented mean shift without trend change. J Atmos Ocean Technol 25(3):368–384. doi: 10.1175/2007jtecha982.1 CrossRefGoogle Scholar
  52. Wang A, Zeng X (2012) Evaluation of multireanalysis products with in situ observations over the Tibetan Plateau. J Geophys Res. doi: 10.1029/2011jd016553 Google Scholar
  53. Wang S, McGrath R, Semmler T, Sweeney C (2006) Validation of simulated precipitation patterns over Ireland for the period 1961-2000. Int J Climatol 26(2):251–266. doi: 10.1002/joc.1246 CrossRefGoogle Scholar
  54. Wang S, Zhang M, Sun M, Wang B, Huang X, Wang Q, Feng F (2015) Comparison of surface air temperature derived from NCEP/DOE R2, ERA-Interim, and observations in the arid northwestern China: a consideration of altitude errors. Theor Appl Climatol 119(1–2):99–111CrossRefGoogle Scholar
  55. Wu G, Zhang Y (1998) Tibetan Plateau forcing and the timing of the monsoon onset over South Asia and the South China Sea. Mon Weather Rev 126(4):913–927CrossRefGoogle Scholar
  56. Wu T, Wang Q, Zhao L, Batkhishig O, Watanabe M (2011) Observed trends in surface freezing/thawing index over the period 1987-2005 in Mongolia. Cold Reg Sci Technol 69(1):105–111. doi: 10.1016/j.coldregions.2011.07.003 Google Scholar
  57. Wu T, Zhao L, Li R, Wang Q, Xie C, Pang Q (2013) Recent ground surface warming and its effects on permafrost on the central Qinghai-Tibet Plateau. Int J Climatol 33(4):920–930CrossRefGoogle Scholar
  58. Yanai M, Li C (1994) Mechanism of heating and the boundary layer over the Tibetan Plateau. Mon Weather Rev 122(2):305–323CrossRefGoogle Scholar
  59. Yang M, Nelson FE, Shiklomanov NI, Guo D, Wan G (2010) Permafrost degradation and its environmental effects on the Tibetan Plateau: a review of recent research. Earth Sci Rev 103(1):31–44CrossRefGoogle Scholar
  60. Yang K, Wu H, Qin J, Lin C, Tang W, Chen Y (2014) Recent climate changes over the Tibetan Plateau and their impacts on energy and water cycle: a review. Glob Planet Change 112:79–91CrossRefGoogle Scholar
  61. You QL, Kang SC, Aguilar E, Yan YP (2008) Changes in daily climate extremes in the eastern and central Tibetan Plateau during 1961–2005. J Geophys Res Atmos. doi: 10.1029/2007jd009389 Google Scholar
  62. You Q, Kang S, Pepin N, Flügel WA, Yan Y, Behrawan H, Huang J (2010a) Relationship between temperature trend magnitude, elevation and mean temperature in the Tibetan Plateau from homogenized surface stations and reanalysis data. Glob Planet Change 71(1):124–133CrossRefGoogle Scholar
  63. You QL, Kang SC, Pepin N, Flugel WA, Sanchez-Lorenzo A, Yan YP, Zhang YJ (2010b) Climate warming and associated changes in atmospheric circulation in the eastern and central Tibetan Plateau from a homogenized dataset. Glob Planet Change 72(1):11–24. doi: 10.1016/j.gloplacha.2010.04.003 CrossRefGoogle Scholar
  64. You Q, Fraedrich K, Ren G, Pepin N, Kang S (2013) Variability of temperature in the Tibetan Plateau based on homogenized surface stations and reanalysis data. Int J Climatol 33(6):1337–1347. doi: 10.1002/joc.3512 CrossRefGoogle Scholar
  65. Zhang T (2005a) Influence of the seasonal snow cover on the ground thermal regime: an overview. Rev Geophys 43(4). doi: 10.1029/2004RG000157
  66. Zhang T (2005b) Spatial and temporal variability in active layer thickness over the Russian Arctic drainage basin. J Geophys Res. doi: 10.1029/2004jd005642 Google Scholar
  67. Zhang Y, Li T, Wang B (2004) Decadal change of the spring snow depth over the Tibetan Plateau: the associated circulation and influence on the east Asian summer monsoon. J Clim 17(14):2780–2793CrossRefGoogle Scholar
  68. Zhang T, Frauenfeld O, McCreight J, Barry R (2005) Northern Hemisphere EASE-Grid annual freezing and thawing indices, 1901–2002. National Snow and Ice Data Center/World Data Center for Glaciology. Digital media, BoulderGoogle Scholar
  69. Zhao T, Guo W, Fu C (2008) Calibrating and evaluating reanalysis surface temperature error by topographic correction. J Clim 21(6):1440–1446. doi: 10.1175/2007jcli1463.1 CrossRefGoogle Scholar
  70. Zhou YW, Guo DX, Qiu GQ, Cheng GD, Li SD (2000) Geocryology in China. Science Press, Beijing, pp 1–439 (in Chinese) Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2016

Authors and Affiliations

  • Yanhui Qin
    • 1
  • Tonghua Wu
    • 1
    Email author
  • Ren Li
    • 1
  • Wenjun Yu
    • 1
  • Tianye Wang
    • 1
  • Xiaofan Zhu
    • 1
  • Weihua Wang
    • 1
  • Guojie Hu
    • 1
  • Liming Tian
    • 1
  1. 1.Cryosphere Research Station on the Qinghai-Tibet Plateau, State Key Laboratory of Cryosphere Sciences, Cold and Arid Regions Environmental and Engineering Research InstituteChinese Academy of SciencesLanzhouChina

Personalised recommendations