1 Introduction

Permafrost results from interactions between the land and the atmosphere, so its spatial distribution is mainly determined by climatic conditions (Ferrians and Hobson 1973; Harris 1981; Cheng and Dramis 1992; Riseborough et al. 2008). However, the influences of local environmental factors, such as slope and aspect, vegetation, snow cover, and soil moisture, may outweigh the climatic background, resulting in heterogeneous permafrost conditions such as temperature, active layer, and ground ice content at a local scale (Brown 1973; Williams and Smith 1989; Camill and Clark 1998; Camill 2000; Cheng 2004; Heggem et al. 2006; Pastick et al. 2014; Guo and Sun 2015; Lin et al. 2019; Luo et al. 2019). Local conditions may influence the thermal regime of permafrost by controlling incoming solar radiation, heat conduction, and convection processes and ground ice conditions (Cheng 2003).

The Qinghai-Tibet Plateau (QTP) is the highest plateau on Earth with a cold climate, and has developed the largest high-elevation permafrost area in the world (Lin et al. 2019; Zhou et al. 2000). The occurrence of high-elevation permafrost is mainly controlled by climate, topography, and surface conditions (Cheng 1983; Harris 1986; Gorbunov 1988; Cheng and Dramis 1992). At a global scale, latitude and atmospheric circulation generally control the distribution of permafrost, while local factors such as topography and surface conditions strongly regulate site-scale ground thermal conditions (Zhang et al. 2000).

The QTP has a complex and diverse topography, so the differences in slope aspect are significant. With its low latitude, high elevation, and clear air (Xu and Chen 2006), the absorption or reflection of solar radiation of various slope aspects differs greatly (Wang et al. 2016), significantly changing permafrost conditions (Gorbunov 1978; King 1986).

The influences of slope aspects on permafrost have been reported in several areas. For example, in southeast Yukon, Canada, vegetation growth and active layer thickness differed at four sites with similar altitude and geological conditions but different slope aspects (Price 1971). At the Tianshan Mountains, China, the average annual ground temperature difference between sunny and shady slopes at the same elevation can reach 4.6℃ (Cheng 2003). The difference between the active layer thickness on north- and south-facing slopes in the Kunlun Mountains, China, was ~1.0 m, and permafrost temperatures differed by 0.5 ℃ (Lin et al. 2015; Luo et al. 2019). The permafrost thickness was about 70 m on the southwest slope of the Fenghuo Mountain on the QTP, but about 120–145 m on the northeast slope. In a discontinuous permafrost region of northern Mongolia, the mean ground surface temperature (MGST) difference between shady and sunny slopes was about 3–4°C (Munkhjargal et al. 2020). Slope aspect also affects the stability of subgrade engineering, causing the asymmetrical subsidence along linear infrastructure in permafrost regions (Hu et al. 2002; Lai et al. 2004; Chou et al. 2008; Niu et al. 2015; Zhang et al. 2017).

Although some studies have reported on differences in ground temperature and thaw subsidence on opposing slopes, there is no in-depth explanation of the mechanisms responsible for the differences with supporting field data. Recently, there has been an increasing occurrence of thaw slumps in the hilly mountains with north-facing slopes on the QTP (Luo et al. 2022), but the mechanisms involved were unclear. This study examined differences in ground temperatures, moisture content, radiation, and ground ice content at two gently sloped sites with opposing aspects in a mountainous permafrost area on the QTP. The objective of the study was to elucidate the climate-permafrost relationship in this mountainous permafrost area, and provide information that could help understand the north-facing distribution of thaw slumps.

2 Study Area and Sites

In this section, we provide a detailed overview of the topography, geological conditions, climate, vegetation, and permafrost conditions of the study sites.

2.1 Study Area

The QTP lies in western China (Fig. 1) and the altitude in most areas is over 4,000 m. The common natural conditions are high and cold periglacial environment. Data recorded by the Wudaoliang weather station in central QTP indicated that the annual mean air temperature was between -6.6 and -3.7°C from 1957 to 2019, with an average value of -5.1°C (Zhou et al. 2022). Annual mean precipitation ranged from 136 to 480 mm, while the annual mean potential evaporation was much greater than rainfall, at about 1,588 to 1,626 mm (Lin et al. 2019). About 10% of the annual precipitation falls as snow.

Fig. 1
figure 1

Location of the study area. a Outline of the Qinghai-Tibet Plateau (QTP) and location of the study sites; b aerial view of the study area; c two sloping sites and the surrounding ground conditions. SM swamp meadow, AM alpine meadow, AG alpine grassland, SG sparse grassland, AD alpine desert, MT mountain. The two sites are at about the same elevation

Perennial herbs are widespread in most areas on the QTP. The main vegetation types include the alpine grassland and alpine meadow, accounting for over 40% of the area (Yin et al. 2017). The vegetation communities are simple, and dominant plant species include Kobresia pygmaea, Carex moorcroftii, Stipa purpurea, and Littledalea racemose, among others (Lin et al. 2019). Strong wind erosion on the plateau results in a fragile ecological environment (Li et al. 1996).

The plateau has undulating terrain and most area is covered by fine to gravelly sands on surface. The high elevation and low latitude permafrost lies beneath the surface (Zhou et al. 2000). Affected by climate warming, thermokarst landforms, including thaw slumps, thermokarst lakes, and surface subsidence, are widespread on the plateau and have been eroding permafrost to degradation (Lin et al. 2010; Lin et al. 2011). Permafrost in the interior of the plateau is continuous, and has high volumetric ground ice content with a mean value of ~17% (Lin et al. 2020; Fan et al. 2021). Active layer thickness ranges from 2–3 m and the annual mean ground temperature is -2 to 0°C (Lin et al. 2020). Sediment texture ranges from clay to sandy gravel, which overlies weathered mudstones and sandstones (Yin et al. 2017).

2.2 Site Description

The study examined two sloping sites with opposing aspects, which are located at both sides of the Gu Mountain (Fig. 1). One site was on a south-facing sunny slope (34.8367°N, 92.9206°E; Fig. 2a) and the other is on a north-facing shady slope, approximately 300 m from a typical north-facing thaw slump (34.8486°N, 92.9268°E; Fig. 2b). The slope angle is about 7.5° on the sunny slope, and about 8.1° on the shady slope. The elevation is 4,634 m a.s.l. at the sunny site and 4,638 m a.s.l. at the shady site. The dominant plant species on the sunny slope are Stipa purpurea and Kobresia parva, and the mean vegetation coverage was approximately 16.7%. On the shady slope the dominant plant species are Androsace tapete Maxim. and Carex moorcroftii, and the mean coverage is ~7.9%.

Fig. 2
figure 2

The sediment profile and topography of the two sites, surveyed in July 2016. GMC gravimetric moisture content, EIC excess ice content, PT permafrost table. U23004 is a temperature sensor, UX120-006M is a 4-channel analog logger, and CRN4 is a net radiometer

The sediment profile to 5.0 m depth is also presented in Fig. 2, which includes information on excess ice content (EIC), gravimetric moisture content (GMC), and permafrost table depth (PT). The ground surface at both sites is typically covered by sandy silt or gravel. Below this layer lie coarse-grained deposits that are rich in flaky debris. The stratigraphy was similar up to 5 m depth at the two sites, but the ice content differed (Lin et al. 2020). Based on the information of boreholes on the two slopes, the active layer thickness was greater on the sunny slope than the shady slope. On the sunny slope, 0.5 m below the ice-poor permafrost table (2.5 m) was the ice-rich permafrost; while for the shady slope, near the upper permafrost table (1.5 m) was the ice-rich permafrost. The ground ice was shallower and higher on the shady slopes, and deeper on the sunny slopes (Fig. 2).

Soil texture information with depth is presented in Fig. 3. The near-surface soil at the sunny slope was ~41% silt, 14% clay, and greater than 50% silt and clay combined. Most of the near-surface samples from the shady slope were sand dominated (63%), with only 40% silt and clay combined samples. The soil at the sunny site was generally >60% silt at the depth of 3–4 m, and included high excess ice content in this frost-susceptible layer.

Fig. 3
figure 3

Piper ternary diagrams of soil texture (a) at the sunny slope and (b) at the shady slope. These data were from a borehole at each site

3 Field and Laboratory Methods

This study examined the differences in surface heat exchange between two sloping sites. For this purpose, the data including air temperature, ground temperature, soil moisture, and radiation were collected through a mixed field approach involving borehole drilling and sensor installation. We also surveyed the slope degree and vegetation coverages. This section presents information on measurement accuracy and data processing as well.

3.1 Measurement of Temperature

Two temperature sensors (HOBO, U23-004) were installed at 2 m height and ~1–2 cm depth to record the air temperature (Ta) and the ground surface temperature (Ts) at each of the two sites. The built-in sensor was protected in a solar radiation shield and an external sensor was buried in soil. The accuracy is ±0.21°C in a -40 to 100°C environment. We began to collect temperature data from September 2016, and the frequency was once per 30 minutes.

Nine boreholes were drilled to 2.5 m and formed a rectangular grid with 2.5 m apart at each site (Fig. 2). One temperature sensor (HOBO, TMC20-HD) was installed at each borehole to measure ground temperature at 0.1 m (Tns, near-surface temperature) and 2.5 m (Ttop, top temperature of permafrost). The two sensors were fixed to the outer wall of a polyethylene aluminium composite tube placed in each borehole. The holes were filled with dry sand and packed with a long rod in order to improve contact between the sensors and surrounding ground. The ground temperature has a strong spatial heterogeneity, so we used the multiple measurements at each site to improve accuracy. The field installation was conducted in July–August 2016 and the data collection (with HOBO UX120-006M 4-channel analog logger) also began in September 2016. The measurements accuracy is ±0.21°C under the ambient temperature from -20 to 70°C, and data were recorded every 4 hours.

3.2 Measurement of Soil Moisture Content

Four soil moisture sensors (HOBO, S-SMD-M005) were inserted directly into the soil profile at the depths of 25, 50, 100, and 150 cm to measure layered soil moisture content. The monitoring of volumetric moisture content (m3/m3) began in September 2016 and the data were recorded every 4 hours by using a HOBO H21-002 micro station. The measurement accuracy of the sensors is ±0.031 m3/m3 (±3.1%) from 0–50°C for mineral soil up to 8 ds/m and ±0.020 m3/m3 (±2%) with soil specific calibration. The records for the unfrozen soil water content in freezing period are invalid because the operating temperature exceeded the ambient temperature requirement of the sensor.

3.3 Measurement of Solar Radiation

We used the CNR4 net radiometer (Kipp&Zonen, Delft-The Netherlands) to measure the incoming shortwave and longwave radiation versus surface-reflected shortwave and outgoing longwave radiation. The instrument was installed at the height of 1.8 m at each site. The CNR4 net radiometer consists of a pyrgeometer pair, one facing upward, the other facing downward, and a pyranometer pair in a similar configuration. All sensors were calibrated individually for optimal accuracy.

The spectral range (50% points) of shortwave measurements is 300 to 2,800 nm and 4,500 to 42,000 nm in the longwave spectral range (50% points). The sensitivity of the sensors is 5 to 20 µV/W/m2 and the temperature dependence of sensitivity (−10 to +40ºC) is less than 4%. The instruments can operate under a temperature of −40 to +80°C and 0–100% RH. The instruments were factory calibrated. Two CR1000 data loggers (Campbell Scientific, Edmonton, AB, Canada) were used to sample at 30-minute intervals, and data at each site were stored as 1-hour averages. Data collection began in September 2016. Following data collection, obviously erroneous measurements were removed and gaps were filled by interpolation.

3.4 Site Investigations

The topography was described in detail and the position was recorded by GPS (Trimble GeoXT6000). The multispectral camera (ADC, Tetracam Inc., Chatsworth, CA, USA) was used to determine the vegetation coverage, and the averaged value of multiple images represents the vegetation coverage. The plant species were also identified. The slope angle was measured with a level (Topcon, DL-101C). The depth of seasonal thawing (thickness of active layer) was approximated at each borehole during the drilling in July–August 2016 by monitoring the core temperature, and it is probably an overestimate because the drilling process likely increased the core temperature.

3.5 Data Processing

For the period from September 2016 to August 2021, most of the data was successfully recorded (>90%). Physical damage or sensor performance problem caused by water or mechanical impact affects data quality, and may produce erroneous data. To correct for data errors, we manually deleted marked deviations from prevailing patterns, and compared the data measured at same site/depth.

Annual mean temperatures, including Ta, Ts, and Ttop, were calculated using the daily values for a whole year (365 days) from 1 September to 31 August. Surface offset (ΔTs) and thermal offset (ΔTg) were calculated as follows (Lachenbruch et al. 1988).

$$\Delta T_{{\text{s}}} = {\text{ T}}_{{\text{s}}} {-}{\text{T}}_{{\text{a}}}$$
(1)
$$\Delta T_{{\text{g}}} = {\text{T}}_{{{\text{top}}}} {-}{\text{T}}$$
(2)

Annual accumulated thawing degree days (TDD) and freezing degree days (FDD) were calculated as outlined in Lin et al. (2019). The ratios of air and ground surface FDD and TDD are (Lunardini 1978):

$$n_{t} = {\text{TDD}}_{{\text{s}}} /{\text{TDD}}_{{\text{a}}}$$
(3)
$$n_{f} = {\text{FDD}}_{{\text{s}}} /{\text{FDD}}_{{\text{a}}}$$
(4)

where nt and nf refer to the thawing and freezing n-factors, and subscripts a and s are air and ground surface, respectively.

Net shortwave radiation (Rs, W·m-2), net longwave radiation (Rl, W·m-2), and net radiation (Rn, W·m-2) can be computed using four components (Eqs. 57):

$${\text{R}}_{{\text{s}}} = {\text{DR}} - {\text{ UR}}$$
(5)
$${\text{R}}_{{\text{l}}} \; = \;{\text{DLR}}\; - \;{\text{ULR}}$$
(6)
$${\text{R}}_{{\text{n}}} = {\text{ DR }} - {\text{ UR }} + {\text{ DLR }} - {\text{ ULR}}$$
(7)

where DR and UR are downward and upward shortwave radiation, respectively; DLR and ULR are downward and upward longwave radiation, respectively. The four parameters were measured at both sites using the CNR4 net radiometer.

4 Results

Our results reflect distinctions between heat exchanges in ground surface of the sunny and shady slopes, incorporating the air temperature, ground surface temperature, active layer thickness, moisture, radiation, and albedo.

4.1 Equivalent Air Temperature (Ta) of the Two Sites

The results presented in this article were from 1 September 2016 to 31 August 2021. The annual mean air temperature and derived values are presented in Table 1. There were no significant differences in air temperature (Ta) between the sunny and shady slopes. The annual mean air temperatures (\(\overline{T }\) a) for 2016–2017 and 2017–2018 were − 2.9 and − 2.6°C at the sunny slope site and − 2.6 and − 2.4°C at the shady slope site. The differences between the sites in the years were 0.3 and 0.2°C. Although the annual mean air temperature in 2018–2019 was more than 1°C lower than the previous two years (− 4.0°C at the sunny slope and − 3.8°C at the shady slope), the air temperature difference between the two sites was still only 0.2°C. The record in 2020–2021 also showed that the annual mean Ta was very close at the two sites.

Table 1 Air temperature (Ta), freezing and thawing indices, and air frost number at the two sites in the Beilu River Basin, Qinghai-Tibet Plateau, for September 2016–March 2021

There was little difference in annual thawing degree days (TDD) and freezing degree days (FDD) at the two sites, particularly in the thawing season where the difference was ~20 degree days (Table 1). The difference in calculated air frost number (F) was only 0.01 (0.62 and 0.63), indicating that the air temperatures at the two sites were very similar. The mean annual air temperature in the warm season (Tws, June to September) was 5.0 ± 0.2°C at the sunny site and 5.1 ± 0.2°C at the shady slope. Meanwhile, the mean annual air temperature in the warmest month (Twm, either July or August) was 7.2 ± 0.2°C at the sunny slope and 7.3 ± 0.2°C at the shady slope. The differences in mean air temperatures between the sites during the cold season and the coldest month were slightly greater, ranging from 0.3 to 0.5°C. Given that the sensor accuracy is ±0.21°C, these modest seasonal differences in air temperature also indicate similar conditions at the two sites. The negligible difference in air temperatures between the two sites is likely because of the close proximity of the two sites and efficient air mixing within the study area.

4.2 Differences in Ground Temperature (Ts and Ttop) between the Two Sites

Daily variations in Ts are presented in Fig. 4a. Compared to air temperature, Ts values were significantly different between the sites during the monitoring period. The mean annual Ts were 0.3 ± 0.4°C at the sunny slope site and −1.0 ± 0.4°C at the shady slope site, with a difference of ~1.3°C. During the warm season, the temperature difference was the greatest between the two sites. The mean Ts in the warmest July month was 11.0 ± 0.2°C at the sunny slope site and 9.0 ± 0.2°C at the shady slope site. For the whole warm season (June–September), the mean Ts was 8.5 ± 0.2°C at the sunny site and 6.8 ± 0.2°C at the shady slope site, with a difference of ~1.7°C. The temperature difference on the warmest days was especially great. For example, on 19 July 2017, the daily mean Ts was 18.0°C at the sunny site and 13.1°C at the shady slope site. Similar or larger temperature differences also occurred on 2 August 2018 and 15 August 2019, with surface temperature differences of 5.5°C and 5.1°C, respectively.

Fig. 4
figure 4

Daily variations in ground temperature at two depths from 1 September 2016 to 31 August 2021. a Ground surface temperature (Ts) and b top temperature of permafrost (Ttop). Max, min, and mean values in the figures refer to the maximum, minimum, and mean values recorded by nine sensors at a certain time per site. Sunny/shady-max maximum value at the sunny/shady slope site, Sunny/shady-min minimum value at the sunny/shady slope site, and Sunny/shady-mean mean value at the sunny/shady slope site from the nine boreholes. The series were plotted from daily data, but the symbols are marked for every five days for clarity

In contrast to the warm season, the difference in Ts between the shady and sunny slopes in the cold season (October–May) was relatively small. Mean Ts in January was −10.8 ± 0.6°C at the sunny slope site and −11.0 ± 1.0°C at the shady slope site. The mean Ts over the cold season was −3.9 ± 0.6°C at the sunny site and −4.8 ± 0.6°C at the shady slope site, with a difference of < 1°C.

Daily variations of top temperature of permafrost at 2.5 m depth (Ttop) are shown in Fig. 4b. The difference in Ttop between the two sites was significant up to 2.5 m depth. The daily mean Ttop fluctuated around 0°C at the sunny slope site, but was stably below 0°C at the shady slope site. The daily mean Ttop at the sunny slope site was also always higher than that at the shady slope site, with differences between 0.4–3.4°C, and a mean value of 1.4°C. The annual mean Ttop values at the sunny slope site in 2016–2021 were 0.2, 0.3, and −0.1°C, and −1.2, −1.2, and −1.5°C at the shady slope, respectively. The annual temperature difference was ~1.4°C.

4.3 Depth of Seasonal Thawing

The ground surface begins to thaw when Ts rises above 0°C with the increase of Ta, and the thaw depth usually reaches its annual maximum at the end of August on the QTP. The maximum seasonal thaw depth at both sites in 2016 was estimated during the drilling in July–August, and the observed results were 2.7 ± 0.1 m from the nine boreholes at the sunny slope site and 1.7 ± 0.1 m at the shady slope site (Fig. 2). A nearly 1.0 m difference of the maximum depth of seasonal thaw between the two sites is likely to result from the large difference in Ts between these sites. In 2016–2021, the mean annual Ttop measured at 2.5 m depth was above 0°C (~0.1°C) at the sunny slope site, indicating that the maximum depth of seasonal thaw was >2.5 m. At the shady slope site, the mean Ttop values and accompanying maximum and minimum values all remained below 0°C (~−1.3°C), indicating that the maximum depth of seasonal thaw was <2.5 m. The difference in the depth of seasonal thawing between the two sites is related to the difference in thawing period duration. The thawing period was about 30–40 days shorter at the shady slope than that at the sunny slope in 2016–2021, resulting in a shallower depth of seasonal thawing at the site.

4.4 Soil Moisture Content

Daily variations in soil moisture content at four different depths within the depth of seasonal thaw (0.25, 0.5, 1.0, and 1.5 m) are presented in Fig. 5. The difference in soil moisture content between the two sites was very significant, with the ground of the shady slope always wetter than of the sunny slope. The difference was maintained despite frequent summer precipitation events. The near-surface ground on the shady slope remained relatively moist during the thawing periods. The mean soil moisture content at the 0.25 m depth on the shady slope was 0.374 ± 0.003 m3/m3 during the thawing periods, and was 0.250 ± 0.002 m3/m3 on the sunny slope (Fig. 5a). The difference in moisture content between the two sites occurred up to the 1.5 m depth (Fig. 5d).

Fig. 5
figure 5

Daily variations in volumetric soil moisture content within the active layer from 1 September 2016 to 31 August 2021. a 0.25 m depth; b 0.50 m depth; c 1.0 m depth, and d 1.5 m depth

4.5 Radiation

The four components of the radiation budget at the two sites all showed variation due to seasonal variations in solar altitude (Fig. 6). The seasonal variation in shortwave downward radiation (DR) was evident at both sites (Fig. 6a). The maximum value of daily mean DR reached ~400 W·m-2 in May–July, and gradually decreased to the minimum value of ~100 W·m-2 in mid-December. The variation in daily mean DR during the cold season was low because of clear and dry atmospheric conditions, but was high in the warm season, likely due to days with precipitation and high cloud cover. The monthly mean, seasonal mean, and annual mean DRs are listed in Table 2. The daily mean DR on the sunny slope was ~3% (~6.1 W·m-2) higher than that at the shady slope and more than 5% higher in December and January.

Fig. 6
figure 6

Temporal dynamics of the four daily radiation components. a DR, downward shortwave radiation; b UR, upward shortwave radiation; c DLR, downward longwave radiation; and d ULR, upward longwave radiation.

Table 2 Mean monthly, cold and warm periods, and annual mean energy balance components (W·m-2) at two sites in January–December

In contrast to DR, although shortwave upward radiation (UR) exhibited relatively little seasonal variation, there were many peaks caused by snowfall that increased the surface albedo, especially in March–May (Fig. 6b). The daily mean UR on the sunny slope was much higher than that on the shady slope, and during the cold period, the monthly mean UR was 40% (~12.0 W·m-2) higher (Table 2). The daily mean DR on the sunny slope was higher than that on the shady slope, and the daily mean UR was much higher. As a result, the net shortwave radiation (Rs) was lower on the sunny slope than on the shady slope. The difference in daily mean Rs between the two sites was ~8 W·m-2 during the warm season and ~5 W·m-2 during the cold season.

The daily mean downward longwave radiation (DLR) showed a similar seasonal pattern as DR, but variations in DLR were smaller and less scattered over the year. The DLR lagged behind DR by a few weeks (Fig. 6c) due to the thermal inertia of the Earth system. Daily mean DLR reached a maximum of ~330 W·m-2, then gradually decreased to the minimum value of ~140 W·m-2. Daily mean DLR on the sunny slope was about 1% (~2.7 W·m-2) lower than on the shady slope. Seasonal fluctuation in upward longwave radiation (ULR) was well correlated with DLR. The daily mean ULR was higher on the sunny slope in most months, except in March–May of each year (Fig. 6d). The resulting net longwave radiation (Rl) on the sunny slope was always ~11.7 W·m-2 lower than on the shady slope during the warm season and ~9 W·m-2 lower during the cold season.

The magnitudes of variation in net radiation (Rn) for both sites were different in each month. However, the daily mean Rn in most months was lower on the sunny slope (Fig. 7, Table 2). The mean daily Rn for the whole study period (n = 969 days) was 12.5 ± 0.65 W·m-2 lower on the sunny slope than on the shady slope. This may be related to differences in surface albedo and surface moisture conditions at the sites. In the warm season, the maximum daily Rn was >850 W·m-2, and <350 W·m-2 in the cold season (Fig. 7).

Fig. 7
figure 7

Half-hour variations in net radiation (Rn) on the 15th of each month, 2017

4.6 Surface Albedo

Variations in daily mean surface albedo were consistent with variations in UR (Fig. 8), and similar at the two sites. But the daily mean surface albedo was slightly higher at the sunny slope (n = 1,578, 0.189 ± 0.003) than at the shady slope (n =1,610, 0.173 ± 0.003). As a result, most of the monthly mean albedo values and the annual mean albedo were higher on the sunny slope (Table 2). The surface albedo at the two sites fluctuated greatly over the year, with snowfall in winter causing a sharp increase up to a maximum of 0.99. The surface albedo was not significantly higher over the whole cold season (0.190 ± 0.005 and 0.168 ± 0.005) than the warm season (0.188 ± 0.002 and 0.181 ± 0.003), indicating that the snowfall did not persist on the ground for long at the sites. This was confirmed by field observations in winter (10–15 January 2017, 9–22 December 2018, and 16–25 December 2019).

Fig. 8
figure 8

Temporal variation in daily mean albedo over the study period. The lines are the 5-day moving average.

5 Discussion

In this section, we explore the disparities in local factors between the shady and sunny slopes and analyze how local microenvironmental influences affect permafrost conditions and soils across different slopes. Lastly, we summarize the impact of various slope elements on the development of thaw slumps.

5.1 Similar Air Temperature versus Different Ground Temperatures

There were no significant differences in daily air temperatures between the two sites, presumably due to their geographical proximity (Table 1). In contrast, differences in ground temperatures were significant (Fig. 4). Difference in mean annual ground surface temperature Ts between the sites was over 1.3°C, and the ground of the sunny slope was significantly warmer (1–2°C).

There was little diurnal variation in Ts at the shady site compared to that at the sunny site (Fig. 9). The daily amplitude of Ts change was usually ~5°C on the shady slope, with a maximum value of ~10°C in the warm season, and a minimum value of ~2°C in the cold season. In contrast, the daily amplitude of Ts change was usually 10–15°C on the sunny slope, with a maximum close to 20°C in the warm season and a minimum of ~5°C in the cold season (Fig. 9). The maximum recorded surface temperature at the shady site was ~18°C in 2017 even during warm weather in mid-August, compared to a maximum of 28°C at the sunny site (Fig. 9h). There was little vegetation cover at both sites, and both had similar mineral soils, with some variation in sand and silt content. Variation in these conditions between the sites is unlikely to be the primary driver of the surface temperature difference. The thermal difference is likely due to the site-specific energy balance components and ground surface soil moisture conditions. The soil at the shady slope site was very wet in the warm season, so the high soil heat capacity moderated temperature changes. As a result, Ts at night at both sites were similar, but increased rapidly at the sunny slope site in the day (Figs. 9f–9j). During the cold season, surface moisture freezes and the higher ice content at the shady slope site also results in a higher soil heat capacity.

Fig. 9
figure 9

Half-hour variations in ground surface temperature (Ts) on the 15th of each month, 2017

The lower Ts throughout the day at the shady slope site in the cold season indicate that there is little available heat to offset surface heat losses at night. As a result, the surface offset (\(\Delta {T}_{s}=\overline{{T }_{s}}-\overline{{T }_{a}}\)) was significantly lower at the shady slope site (Fig. 10). Surface offset (ΔTs) is an indicator of the surface energy balance; it is controlled by a combination of local factors, for example, topography, snow cover, moisture availability, and vegetation (Eaton et al. 2001; Beltrami and Kellman 2003). The mean surface offset in 2016–2021 was 3.5 ± 0.1°C on the sunny slope and 1.9 ± 0.1°C on the shady slope. A ~1.5°C difference in ΔTs further highlights the warmer ground surface conditions at the sunny site.

Fig. 10
figure 10

Monthly variations in ΔTs over 2016–2021 at the two sites

The results indicate that the slope aspect-induced changes in the local environment are substantial, especially in the high-altitude mountainous permafrost zone. In this study, the difference in annual mean surface temperature between the slopes was about 1–2°C. In other regions of the QTP, such as Chalaping (the source area of the Yellow River, northeastern QTP) and the source area of the Yangtze River, the difference in annual mean surface temperature between slopes is up to 1.2°C and 2.5°C (Luo et al. 2018; Luo et al. 2020). Similar studies have reported mean surface temperature differences of 4.6°C in the Tianshan Mountains of China (Cheng 2003) and 3–4°C in northern Mongolia (Munkhjargal et al. 2020). Thus, the question of why more thaw slumps developed on shady slopes when near-surface temperatures were higher on sunny slopes needs further answer.

5.2 Seasonal Dynamics of Water and Thermal Conditions at Ground Surface

Our observations of the relationship between Ta and Ts were similar to studies at flat sites in the Beilu River Basin, QTP (Lin et al. 2019). With the increase in air temperature in mid-January, the ground began to warm, but with a lag of about one week. Ts approximated 0°C in mid-April, after which there was a rapid increase in soil moisture content induced by downward thawing. Ts reached the annual maximum value in July and August, at which time the sunny slope was much warmer than the shady slope (Fig. 9). Soil moisture content fluctuated with the arrival of monsoon rains in May to October. The higher Ts on the sunny slope may increase the surface evapotranspiration, resulting in lower soil moisture content on the sunny slope. The higher soil moisture content during the entire thawing period was the main control on the energy budget. Although measurements of soil heat flux were attempted in this study, the instruments failed, but soil heat flux was closely related with Rn in the Beilu River Basin (You et al. 2017). The Rn continued to increase and reached its maxima in July–August with the highest solar altitude (Fig. 11), synchronizing with the variations in Ts. The Rn in November–January maintained low values, with a peak value of ~350 W·m-2, less than one-half the maximum in May–August (Fig. 7). In early October, Soil moisture content dropped rapidly with the surface soil frozen and vegetation withered. The variations in UR and ULR showed strong seasonality, but the ground surface albedo did not increase rapidly in winter. The freeze-thaw cycles of the wet surface soil on the shady slope may trigger a large and rapid energy change in permafrost regions, which is different from that related to the onset of the summer. The results are helpful for understanding the dynamic processes of surface hydrothermal exchange in alpine/plateau permafrost regions.

Fig. 11
figure 11

Temporal variation in daily mean net radiation (Rn) over the study period. The lines are the 5-day moving average.

5.3 Effect of Slope Aspect on Regional Permafrost Condition

The observation results show that the air temperatures were similar at the two sites (Table 1). But the difference of slope aspect resulted in different ground temperatures during the freezing and thawing periods at both sites. The lower Ts at the shady slope resulted in a reduced difference between air and ground surface temperatures (ΔTs). The Tps sensor at 2.5 m depth was within the active layer at the sunny slope site but was below the permafrost table at the shady slope site (Fig. 4b). It is generally recognized that net radiation (Rn) on sunny slopes is greater than on shady slopes. However, in this study, the Rn at the shady slope site was slightly higher (~14 W·m-2) than that at the sunny slope site. The DR on the sunny slope was much greater. Higher albedo on the sunny slope resulted in higher upward shortwave radiation (UR).

The differences caused by the slope aspect were apparent in soil moisture measurements. The sunny slope was much warmer, increasing the potential evaporation and resulting in drier soils (Fig. 5). This phenomenon arises because slope aspect significantly influences the intricate interplay among Ta, Ts, surface soil water retention, and heat exchange between the lower atmosphere and the land surface.

Other partitions in the energy balance (for example, sensible heat flux and latent heat flux (LE)) were not calculated, but a dramatic increase in LE with the rapid increase in soil moisture in thawing period was previously observed on the QTP (You et al. 2017). Studies have shown that the LE was the main portion of energy budget in summer (Wang et al. 2005; Chen et al. 2009; Ma et al. 2009). The lower surface temperature of the shady slope reduces heat input to the boundary layer and decreases the temperature lapse rate in the lower atmosphere (Otterman 1974). Although the Rn on the shady slope was slightly higher (~14 W·m-2) than the sunny slope, the ground surface temperature was lower due to the high soil heat capacity.

5.4 Effect of Slope Aspect on Soil Texture

The difference in aspect results in a significant difference in ground temperature and moisture content, and affects the frequency and intensity of freeze-thaw cycles near ground surface. There is a rapid drop in frequency with depth such that below 5–10 cm only the annual cycle occurs at high latitudes (French 2018). However, significantly more freeze-thaw cycles occur in soils in mid-latitude alpine and subalpine regions. For example, at a depth of 1.5 cm in the Japanese Alps, an annual average of 88 freeze-thaw cycles have been measured over a 5-year period (Matsuoka 1990). In this study, the number of freeze-thaw cycles in the 5 cm depth was significantly higher on the sunny slope than on the shady slope. In 2016–2019 the total number of freeze-thaw cycles on the sunny slope reached 615 times, while only 185 times on the shady slope. Daily freeze-thaw cycles are evident at the surface on the sunny slope in February–May and October in 2016–2017 (Fig. 9).

A high frequency of freeze-thaw cycles is effective for mechanical weathering of near-surface bedrock or coarse-grained soil, leading to spalling of relatively small rock fragments, although lack of moisture may considerably reduce the number of effective freeze-thaw cycles (Matsuoka 1994; Prick 2003). The freeze-thaw cycle frequency is proportional to the mechanical weathering rate of surface materials. Each freeze-thaw cycle may lead to denudation of the surface material because the structure is damaged and strength is reduced. The daily amplitude of Ts was greater at the sunny slope site, thus the mechanical weathering rate of the surface material is likely greater on the sunny slope. The soil texture at the surface of the sunny slope was finer with greater than 50% silt and clay combined (Fig. 3), which may be due to a higher freeze-thaw frequency and weathering rate. In addition, soil freezing is accompanied by the migration of moisture to the freezing front (Everett 1961), further driving the movement of fine soil particles. When the frozen soil is thawing, fine particles also move as moisture seeps downslope. As a result, each freeze-thaw cycle will cause soil redistribution that affects soil texture. In general, fine particles may move downslope more on the sunny slope due to the high frequency of freeze-thaw cycles (Fig. 3).

5.5 Effect of Slope Aspect on Thaw Slumps

The higher summer air temperature and precipitation have been confirmed to be the main triggers for thaw slumps (Kokelj et al. 2015; Lewkowicz and Way 2019; Ward Jones et al. 2019; Luo et al. 2022). However, on the QTP, rainfall and mean annual temperature are roughly equal across gentle slopes with different aspects, even with extremely high summer temperatures on sunny slopes, but different scales of thaw slump developed on shady slopes, which can be attributed to changes in the local environment due to slope aspect effects.

One of the most important factors in the development of thaw slumps is ground ice. The presence of different slope aspects changes the permafrost conditions, especially in soil moisture. Moisture affects the formation of ground ice, and areas with ice-rich permafrost were usually developed with sufficient moisture, such as at lake shores (Kokelj and Burn 2010), in wetland environments (Morse et al 2009), and in wet areas of drained lake basins (Mackay and Burn 2011). Different slope aspects lead to differences in moisture and ground ice content. The accumulation of ground ice showed a high positive correlation with moisture in the active layer and a weak correlation with soil texture (O’Neill 2011). In this study, the moisture content of the active layer was significantly higher on the shady slope than on the sunny slope, and more ground ice also accumulated on the shady slope, which is the same as the results of drilling by Lin et al (2019). Ishikawa et al (2010) also found in Mongolia that permafrost developed on the north-facing slope and rarely on the south-facing slope and flat grassland. The accumulation of ground ice on the north-facing slope due to slope aspect was the foundation for the initiation of thaw slumps.

There are no long time series data to confirm the differences in temperature warming rates between the north and south slopes. However, there were clear differences in ground temperature (Ts and Ttop) between the two sites, and in maximum depth of seasonal thaw, which differed by ~1.0 m in 2016–2021. Seasonal thaw depths are related to the soil thermal properties, which depend on soil density, porosity, and texture (Pang et al. 2011; Li et al. 2018). The active layer of the shady slope is generally shallow, thus the ice-rich layer of the shady slope melts first and thus induces thaw slump under the same warming background.

From the perspective of energy balance, the DR, UR, and ULR were all greater on the sunny slope than on the shady slope, while the DLR was smaller. The combined effect made the Rn of the sunny slope smaller, and the annual Rn difference reached 12.5±0.65 W·m-2, which is in contrary to the usual understanding. In the cold season, both slope sites were dominated by sensible heat exchange transmission; but the shady slope was dominated by latent heat exchange in the warm season. Thus, the difference between sensible and latent heat fluxes was smaller on the sunny slope. The larger latent heat on shady slopes in summer leads to heat accumulation and promotes the melting of ground ice (Fan et al. 2022).

Therefore, the slope aspect changes the thermal regime of the permafrost, the distribution and content of ground ice, and the thickness of the active layer of permafrost, creating a north direction that is more prone to the development of thaw slumps.

6 Conclusion

Based on a detailed examination of soil moisture, near-surface temperatures, and radiation over five years, the near-surface permafrost conditions were studied at two sloping sites with differing aspects on the QTP to understand why the slope aspect affects the distributions of thaw slumps. The results showed that sunny and shady slope sites had similar air temperatures and precipitation, but the difference in slope aspect resulted in a significant difference in annual mean ground temperature of 1–2°C. Higher ground surface temperatures at the sunny slope site increased the difference between air and ground temperatures (ΔTs), and caused deeper active layer thickness of ~1.0 m. Soil moisture content increased rapidly during the thawing of the active layer, but the higher ground surface temperature increased evaporation on the sunny slope, making it much drier than the shady slope. The slope aspect changes radiation, resulting in high downward shortwave radiation (DR) on sunny slopes, but high net radiation on shady slopes due to ground surface albedo. Nonetheless, the ground temperatures on the sunny slope were higher, and the shady slope had smaller active layer thickness and more abundant ground ice than the sunny slope. The data improved knowledge on the ground surface boundary conditions in mountainous permafrost regions, and indicate that sunny slopes may be >1–2°C warmer than shady slopes when the permafrost distribution is similar. However, shady slopes may be more prone to develop thaw slumps due to ground ice, active layer moisture, and radiation.