4. Discussion

4.1. Similar air temperature vs. different ground temperatures

There were no significant differences in daily air temperatures between the two sites, presumably due to their geographic proximity and mixing of air in the study area (Table 1). In contrast, differences in ground temperatures were significant (Figures
6b and 7). The annual mean difference in Ts or Tns between sites was over 1.3 °C, and the ground at the sunny slope was significantly warmer (1-2 °C) than the shady slope.
There was little diurnal variation in Ts at the shady slope compared to the sunny slope (Figure 12). The daily amplitude of Ts was usually ~5 °C at the shady slope, with a maximum value of ~10 °C in the warm season (Figures 12f-i), and a minimum value of ~2 °C in the cold season (Figures 12j-l). In contrast, the daily amplitude of Ts was usually 10-15 °C at sunny slope, with a maximum close to 20 °C in the warm season and a minimum of ~5 °C in the cold season. The lower Ts throughout the day at the shady slope in the cold season indicate that there is little available heat to offset surface losses at night. As a result the surface offset (ΔTs=Ts-Ta) was significantly lower at the shady slope than that at the sunny slope (Figure 13).
The surface offset (ΔTs) is a function of the surface energy balance and is controlled by factors such as vegetation, snow cover, moisture availability, and topography (Eaton et al. 2001; Beltrami and Kellman 2003). The mean surface offset in 2016-19 was 3.54±0.07 °C at the sunny slope and 1.91±0.12 °C at the shady slope. A ~1.5 °C difference in ΔTs further highlights the warmer ground surface conditions at the sunny slope.
The results indicate that permafrost degradation should consider differences in surface temperature caused by microenvironment, and particularly the slope aspect in mountainous areas. In this study slope aspect was associated with a difference in annual mean surface temperature of about 1-2 °C, and similar studies have reported that the difference in mean ground surface temperature reached 4.6 °C at Tianshan Mountain, China (Cheng 2003) and 3-4 °C in northern Mongolia (Munkhjargal et al. 2020). The data will help inform realistic boundary conditions for modelling permafrost degradation in mountain permafrost areas.

4.2 Seasonal dynamics of water and thermal at surface

Our observations of relations between Ta and Ts were similar to studies at flat sites in Beiluhe basin, QTP (Lin et al. 2019). With the increase in air temperature in the middle of January, the ground began to warm, but with a lag of about one week. Tsneared 0 °C in the middle of April, after which there was a rapid increase in soil moisture induced by downward thawing. Ts reached the maximum value in July and August, at which time the sunny slope was much warmer than the shady slope (Figure 12g and h). The soil moisture content fluctuated drastically with the arrival of monsoon rains in May to October. The higher Ts at the sunny slope may increase the surface evapotranspiration, resulting in lower soil moisture content at the sunny slope than at the shady slope. The higher soil moisture during the entire thawing period was the main control on the energy budget. Though measurements of soil heat flux (G) were attempted in this study, the instruments failed, but G has been shown to be correlated with Rn in Beiluhe Basin (You et al. 2017). The Rn continued to increase and reached their maxima in July-August when the solar altitude is the highest (Figure 14), which was synchronized with variations in Ts. Rnin November-January maintained low values with a ~350 W·m-2 of peak value, less than half the maximum value in May-August (Figure 10). When the surface soil froze and vegetation withered in early October, soil moisture content dropped rapidly. The variation in UR and ULR showed strong seasonality, but the albedo did not increase rapidly in winter. The freeze-thaw cycle of the wet surface soil at the shady slope may trigger a drastic energy change in permafrost regions, which is different from the alternations correlated with the onset of the summer. The results are helpful for understanding the dynamic process of surface water thermal energy in mountain permafrost regions.

4.3 Effect of slope aspect on regional permafrost environment

The observational results show the air temperatures were similar at both sites (Figure 6a). However, the difference of slope direction resulted in different ground temperatures during the freezing and thawing periods at both sites. It is generally recognized that net radiation (Rn) on sunny slopes is greater than on shady slopes. However, in this study Rnat shady slope was slightly higher (~13 W·m-2) than on the sunny slope. The shortwave downward radiation (DR) at the sunny slope was much greater than at the shady slope. Higher albedo at the sunny slope resulted in higher upward shortwave radiation (UR).
The differences caused by the slope direction were apparent in soil moisture measurements. The sunny slope was much warmer, increasing the potential evaporation and resulting in drier soils (Figure 8). The complex interactions between Ta, Ts, and surface soil water retention and heat exchange between the land surface and the lower atmosphere.
Other partitions in the energy balance (e.g. sensible heat flux (H), 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 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 at 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 at the shady slope was slightly higher (~13 W/m2) than the sunny slope, the ground surface temperature was lower due to the high soil heat capacity.
The presence of permafrost supports the development of alpine meadow ecosystems (Jin et al. 2009; Wang et al. 2009), and is strongly related to heat insulation resulting from hydrothermal conditions in the active layer. Soil moisture content decreases with increasing active layer thickness for the slope in the study. Permafrost warming and degradation could potentially lead to a reduction in soil moisture and soil nutrient content, resulting in vegetation degradation and possible desertification over the QTP (Xue et al. 2009).

4.4 Effect of slope aspect onsoil texture and SOM

A high frequency of freeze-thaw cycles is effective for mechanical weathering near-surface bedrock or coarse-grained soil, leading to spalling of relatively small rock fragments (Matsuoka 1994), though lack of moisture may considerably reduce the number of effective freeze-thaw cycles (e.g. Prick 2003). The freeze-thaw cycle frequency is proportional to the mechanical weathering rate of surface materials. The soil texture at surface of sunny slope was finer with greater than 50% silt and clay combined (Figure 4), 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 at the sunny slope due to the high frequency of freeze-thaw cycles (Figure 4a).
Alluvial material coupled with the low temperatures and low decomposition rates in permafrost regions favors the accumulation of SOM (Dorfer et al. 2013). Vegetation conditions and pedogenesis are usually regarded as the dominant controlling factors for SOM accumulation (Jobbagy and Jackson 2000). In the present study, SOM accumulation was much greater at the shady slope than that at the sunny slope, in contrast with the degree of vegetation coverage. We hypothesize that pedogenesis had an important effect on the SOM distribution since the major distinction between the two sites is the different slope aspect, and the vegetation coverage was relatively low at both sites (sunny slope: 16.7%; shady slope: 7.9%). Higher SOM content is expected in the upper soils at the two sites (Figure 5) because the fallen litter on the ground surface and the turnover of fine roots are considered to be the dominant SOM inputs to natural ecosystems (Zollinger et al. 2013). The observations of SOM were similar to those at Eboling Mountain, northeastern QTP (Mu et al. 2016). SOM content increased with depth in the upper permafrost (3-5 m) at the two sites, which may be because ground ice in this depth range contributes to SOM stabilization (Jorgenson and Shur, 2009; Grosse et al. 2011). High SOM is associated with high proportions of fine soil particles as they tend to stabilize and retain more organic matter (Figure 4). Moreover, the humid soil at the shady slope promotes microbial activities that are beneficial to SOM formation (Mu et al. 2016). Thus, greater SOM accumulated at the shady slope, particularly in the top portion of the active layer.

4.5 Effect of factors on the variations within sites

There are clear differences in ground temperature (Ts, Tps, and Tg) between the two sites, and in maximum seasonal thaw depth, which differed by ~1.0 m in 2016-19. These differences are related to the effects of local factors in the study areas. The slope aspect is undoubtedly one of the key factors (Lin et al. 2019). In addition, vegetation can strongly control the depth of seasonal thawing in mountainous areas (Wang et al. 2009; Shiklomanov et al. 2010; Shiklomanov and Nelson 2013; Chang et al. 2014a; Wang et al. 2017; Keuper et al 2017). The greater leaf area of vegetation at the sunny slope can reflect and blocks solar radiation, reducing the amount of net radiation to the ground (Fisher et al. 2016). This can have an important effect on surface temperature variations (Chang et al. 2012). The well-vegetated surface is covered with an insulative organic layer, which has a net cooling effect on the ground surface (Cannone and Guglielmin 2009) and slows the rate of change in the active layer temperature (Fisher et al. 2016). In the alpine meadow regions on QTP vegetation can also slow down the response of permafrost to climate warming through the greater water retention capacity of its root zone (Wang et al. 2009). Variation in vegetation cover will also indirectly affect the active layer thermal regime by affecting soil properties and snow cover (Chang et al. 2014b; Fisher et al. 2016). Though the plant heights (<15 cm) and coverage (8-16%) are relatively low, variations in vegetation conditions still likely impact the thermal regime of the active layer and near-surface permafrost at the study sites (Lin et al. 2019).
Seasonal thaw depths are also related to the soil thermal properties, which depend on soil density, porosity, and texture (Pang et al. 2011; Li et al. 2018). Variations in soil moisture caused by the infiltration of summer precipitation may reduce the seasonal amplitude of soil temperatures, decrease mean annual soil temperatures, and thin active layers (Luo et al. 2020). For the shady slope, a cooling effect may thus be caused by liquid water that accumulated and persisted within the active layer.