STUDY SITES

This study was conducted on the opposite slopes (north-facing and south-facing slopes) of a U-shaped valley at the peak of the Sergyemla Mountains (29°36′N, 94°36′E) in southeast Tibet, China. Abies georgei var. smithii and Juniperus saltuaria are the dominant tree species of timberlines on north-facing and south-facing slopes, respectively. The elevation of the juniper timberline is 100–200 m higher than that of the fir. Along three altitudinal transects (S1–S4 on the south-facing slope; N1–N5 and N6–N8 on the north-facing slope; Fig. 1), 12 long-term plots (50 × 50 m for forests, and 10 × 10 m for shrublands and grasslands) were selected to represent altitudinal vegetation changes from subalpine and timberline forests (tree height >4 m and canopy density >40%) to alpine shrublands and grasslands (Table 1 and Fig. 1). In each of the 12 plots with different plant stature and coverage, the hourly recorded data logger and associated sensors (HOBO weather micro-station, Onset Inc., U.S.A.) were installed to measure soil temperature and moisture. Moreover, the simultaneous measurements of air temperature, relative humidity, rainfall, and snow depth, etc. were obtained from four automatic weather stations located at the sites for both upper limits of fir forest (N2) and juniper forest (S2) and their nearby shrubland (N3) and grassland (S3). The monitoring started in August 2005 for most sites (N1–N4 and S1–S4), and in June 2007 for other sites (N5–N8). In addition, we measured elevation, aspect, stand variables of canopy height and leaf area index (LAI), and soil bulk density for each of the 12 plots in August 2007 (Table 1).

FIGURE 1

Location and vegetation type description of study sites across (a) north-facing fir and (b) south-facing juniper timberline ecotones of (c) a U-shaped valley in the Sergyema Mountains.

TABLE 1

Stand and soil variables in 12 plots along north-facing and south-facing slopes of a valley in the Sergyema Mountains, southeast Tibet of China.

MEASUREMENTS OF STAND AND SOIL VARIABLES

To quantify the effects of vegetation and soil variables on spatial variations in soil temperature and moisture, we measured two major stand variables of LAI and canopy height, as well as soil bulk density for each of the 12 plots. In each forest plot, tree height and diameter at breast height (DBH) were measured for all trees of over 3 cm in DBH. Leaf biomass of trees was calculated from measurements of DBH and tree height using species-specific allometric regressions that were developed in our previous work (Luo et al., 2002), while the undergrowth leaf biomass of shrubs and herbs were measured by harvesting three quadrats (4 × 4 m). For alpine shrublands and grasslands, species-specific leaf biomass of a plot was measured by harvesting three quadrats (4 × 4 m for tall shrubs, 2 × 2 m for low shrubs, and 0.5 × 0.5 m for grasses). We collected leaf samples in different leaf age classes of trees, shrubs and grasses, and measured the specific leaf area (SLA, the ratio of fresh leaf area to dry mass) for each leaf sample. The species-specific LAI was calculated as the product of species-mean SLA and leaf biomass. Forest LAI was the sum of the tree LAI and the undergrowth LAI. Mean canopy height of forests and tall shrublands was calculated from the average DBH (trees) or stem basal diameter (tall shrubs) using the regression of individual tree height against DBH or stem basal diameter, while the canopy height of low shrublands and grasslands was directly measured by a tape. The soil bulk density (5–20 cm in depth) for each of the 12 plots was measured on soil samples extracted with a circular soil cutter (Table 1).

MEASUREMENTS OF SOIL TEMPERATURE AND MOISTURE

Within each plot, soil temperatures and volumetric moisture contents at soil depths of 5 cm (limited to sites N2–N3 and S2–S3), 20 cm, and 40 cm (or 60 cm at N2–N3 and S2–S3) were measured by using a single HOBO micro-station (Onset Inc., U.S.A.). The depth-specific smart sensors of soil temperature (S-TMB-M002, with accuracy ±0.2 °C) and soil moisture (S-SAM-M003, with accuracy ±3%) were connected to the HOBO logger by 2-meter cables. The logger recorded data at an interval of one hour. We downloaded the hourly recorded data from the logger to a computer at six-month intervals.

MEASUREMENTS OF AIR TEMPERATURE, RELATIVE HUMIDITY, RAINFALL, AND SNOW DEPTH

We further established four automatic weather stations at the sites for both upper limits of the north-facing fir forest (N2) and the south-facing juniper forest (S2) and their nearby low shrubland (N3) and grassland (S3) (Fig. 1). Air temperature and relative humidity at a height of 3 m above the ground were simultaneously measured across forested (N2, S2) and non-forested (N3, S3) sites, while the observation of rainfall and snow depth was limited to the two non-forested sites (N3, S3). The combined air temperature and relative humidity probe with vented radiation shield (MP101A, Rotronic Inc., U.S.A.) and the tipping-bucket rain gauge (7852M-AB, Wittich & Visser, Netherlands), as well as the ultrasonic snow depth sensor (260–700, NovaLynx Corp., U.S.A.) were connected to the data logger (HL20, Jauntering Inc., Taiwan). Also, the HL20 logger recorded data at an interval of one hour, and the hourly recorded data were downloaded to a computer at six-month intervals.

CALCULATIONS OF CLIMATIC PARAMETERS

Daily mean air and soil temperatures and soil moisture were the averages of hourly records within a day. Daily air and soil temperature amplitudes were the difference between the hourly records of maximum and minimum temperatures within a day. Hourly vapor pressure deficit was calculated from hourly air temperature and relative humidity (Sheng et al., 2003), and daily mean vapor pressure deficit was the average of the hourly vapor pressure deficits between 8:00 and 20:00 within a day. In this study, our analysis was based on the four-year (from 17 August 2005 to 10 September 2009 for sites N1–N4 and S1–S4) or two-year (from 22 June 2007 to 10 September 2009 for sites N5–N8) averages of daily measurements. Unfortunately, the entire winter and early growing season (2007–2009) data were lost at the grassland site on the north-facing slope (N8) because of power failure. Using the averaged data of daily mean air and soil temperatures, we calculated mean air temperatures for the year (ATa), growing season (ATg), January (AT₁), and July (AT₇), and mean soil temperatures for the year (STa), growing season (STg), January (ST₁), and July (ST₇), as well as the air (ATS) and soil (STS) temperature sums above 5 °C (the sum of daily values) and the growing season length (GSL, see definition below). The slope-specific annual precipitation (AP) was calculated as the average for the four-year sum of logger records, in which approximately 85% of annual precipitation was recorded during May and September. Additionally, we calculated mean soil moistures for the year (SMa), growing season (SMg), January (SM₁), and July (SM₇).

Thomson and Moncrieff (1982) indicated that an air temperature threshold of 2.8 °C was a good predictor of bud burst of Douglas fir in Canada. In this study, we adopted that air temperature threshold to define the growing season length (GSL) of shoot growth. The GSL equaled the growing period between both dates at which the daily mean air temperature first exceed 2.8 °C in spring and first dropped below 2.8 °C in autumn. There is evidence that bud burst and the start of new shoot elongation are mainly dependent on air temperature but not soil temperature (Domisch et al., 2002). According to our phenological observation, the bud burst of Salix oritrepha (a deciduous shrub) outside the forest stand occurred on 15–20 April on the south-facing slope but on 10–15 May on the north-facing slope, which were generally consistent with the predicted beginning dates of the growing season at two non-forested sites of S3 (15 April) and N3 (8 May). In the study area, the snow cover persisted from late November to late April on the south-facing plots but from late November to late May on the north-facing plots.

At high altitudes, the soil warming in spring lagged by over a week behind the increased air temperature because of seasonal snow cover and soil thawing (Gehrig-Fasel et al., 2008). Accordingly, the belowground root growth can start later than the aboveground shoot growth. In a previous worldwide study of high-altitude treeline soil temperature, Körner and Paulsen (2004) defined the belowground growing period by using a −10 cm soil temperature threshold of 3.2 °C. In this study, we did not measure the −10 cm soil temperature. At both species' timberlines, most of the fine roots of trees were distributed in the soils at the 10–30 cm depths. We assumed that the threshold of 3.2 °C could be used to calculate the growing season mean soil temperature based on the daily mean soil temperature at −20 cm. By using the threshold of 3.2 °C to define the beginning and ending of the belowground growing season, the estimated seasonal mean soil temperature at −20 cm was 0.2–1.0 °C lower than those at −5 cm across forested (0.3 °C for N2 and 0.5 °C for S2) and non-forested (1.0 °C for N3 and 0.2 °C for S3) sites. Given the non-linear decrease of soil temperature with depth in soil, the growing season mean soil temperature at −20 cm would be much closer to the −10 cm soil temperature than to the −5 cm soil temperature.

STATISTICAL ANALYSIS

The partial correlation coefficients of multiple linear regression were used for assessing the relative importance of altitude, aspect, soil, and vegetation variables to the spatial variability of soil temperature and moisture. The independent variables included elevation, aspect southness (because aspect is a circular variable, we transformed it to “southness” [cos(aspect − 180°)]), soil and vegetation variables (LAI, canopy height, and soil bulk density). The effect of soil moisture on soil temperature was also considered in the partial correlation analysis.

A simple linear model (y  =  a + bx) was used for examining the relationships of stand variables (canopy height and LAI) to the growing season and July mean soil temperatures (all variables were normalized by the natural logarithmic transformation). Based on observed data across the 12 plots, multiple linear regression models relating seasonal mean soil temperatures to selected significant independent variables were developed.

All statistical analysis was performed using SPSS 15.0 for Windows, and all significant differences were at p < 0.05.

SEASONAL VARIABILITY OF DAILY MEAN SOIL TEMPERATURE AND MOISTURE AND DAILY TEMPERATURE AMPLITUDE

The difference of spring soil warming dates between north-facing (Figs. 2a–2b; N2–N3) and south-facing (Figs. 2c–2d; S2–S3) slopes was up to 20–30 days, which was associated with the much deeper snow depth on the north-facing slope (Figs. 2b, 2d). During snow cover period, daily mean soil temperature across depths and sites was much higher than the mean air temperature (Fig. 2).

FIGURE 2

The annual course of daily mean air and soil temperatures and snow depth across the north-facing (a, N2) fir timberline and (b, N3) low shrubland; and the south-facing (c, S2) juniper timberline and (d, S3) grassland. Snow depth was recorded only at sites N3 and S3.

In summer, the −5 cm soil temperature amplitude at the north-facing fir timberline (N2, <1 °C; Fig. 3a) was much lower than that at the north-facing low shrubland (N3, 2–5 °C; Fig. 3b) and the south-facing juniper timberline and grassland (S2–S3, 2–6 °C; Figs. 3c–3d). In contrast, similar air temperature amplitudes were found at both timberlines (Figs. 3a, 3c). During snow cover period, the −5 cm soil temperature amplitude was generally close to 0 °C (Figs. 3a–3c) except for that of the south-facing grassland where the higher −5 cm soil temperature amplitude (1–6 °C) did not have a seasonal trend (S3; Fig. 3d).

FIGURE 3

The annual course of daily air and soil temperature amplitudes across the north-facing (a, N2) fir timberline and (b, N3) low shrubland; and the south-facing (c, S2) juniper timberline and (d, S3) grassland.

The daily mean soil moisture varied greatly with seasons, aspects, and vegetation types, compared to the relatively unchanged vapor pressure deficit (Fig. 4). At 20 cm depth where most of fine roots of trees were distributed, the daily mean soil moisture was typically >20% during the growing season (Fig. 4).

FIGURE 4

The annual course of daily mean soil moisture, daily precipitation and daily mean daytime vapor pressure deficit (VPD) across the north-facing (a, N2) fir timberline and (b, N3) low shrubland; and the south-facing (c, S2) juniper timberline and (d, S3) grassland. Precipitation was only recorded at sites N3 and S3.

SPATIAL VARIABILITY OF SEASONAL MEAN SOIL TEMPERATURE AND MOISTURE

The north-facing and south-facing slopes had similar annual precipitation (Table 2). On non-forested sites, mean air temperatures for the year, growing season, January, and July were 1.0–1.9 K higher at the south-facing grassland (S3 at 4441 m) than at the north-facing low shrubland (N3 at 4390 m) (Table 2). The growing season length and the air temperature sum above 5 °C were 25 days longer and 361 degree-days higher at the south-facing grassland than at the north-facing low shrubland. In contrast, the mean air temperatures of both timberline sites (N2 at 4320 m and S2 at 4425 m) differed little (0–0.8 K; Table 2).

TABLE 2

Seasonal mean air temperature and annual precipitation across the north-facing fir timberline (N2, 4320 m) and low shrubland (N3, 4390 m), and the south-facing juniper timberline (S2, 4425 m) and grassland (S3, 4441 m) of a valley in the Sergyema Mountains during 2005–2009. The climatic variables were calculated for mean air temperatures for the year (ATa), growing season (ATg), July (AT₇) and January (AT₁), the growing season length (GSL), the air temperature sum above 5 °C (ATS), and annual precipitation (AP).

On average, seasonal mean soil temperatures at −20 cm across altitudinal vegetation types were generally 1.1–2.2 K higher on the south-facing slope than on the north-facing slope (Table 3). Accordingly, the average soil temperature sum above 5 °C was 531 degree-days higher at the south-facing sites (1241 ± 257 degree-days) than at the north-facing sites (710 ± 202 degree-days). Between both timberline sites (S2 vs. N2 and N6), the differences of the −20 cm mean soil temperatures were 1.0–1.1 K for the year, 0.8–1.0 K for growing season, 2.0–2.2 K for July, and 0.6–1.0 K for January. The −20 cm mean soil temperatures for shrublands and grasslands above both timberlines were generally 0.5–1.8 K higher than those of forest sites (at lower altitudes) (Table 3).

TABLE 3

Seasonal mean soil temperature at −20 cm for 12 study sites across north-facing and south-facing timberline ecotones of a valley in the Sergyema Mountains during 2005–2009. The soil temperature variables were calculated for mean soil temperatures for the year (STa), growing season (STg), July (ST₇) and January (ST₁), and the soil temperature sum above 5 °C (STS).

Partial correlation analysis of the pooled data from all 12 plots indicated that canopy height, LAI, and aspect southness were the main factors affecting the spatial variability of seasonal mean soil temperatures (Table 4). Canopy height and LAI, which were higher on the north-facing slope and decreased with altitude (p < 0.01; from Table 1), were negatively correlated with the growing season and July mean soil temperatures (Fig. 5) even though the mean soil temperatures varied little with altitude (p > 0.10; from Table 3). A combination of aspect southness and canopy height/LAI explained 56–77% of the variations in the −20 cm mean soil temperatures for the year, growing season, and July across the 12 plots (Table 5).

FIGURE 5

Relationships of mean soil temperatures for growing season (ST_g) and July (ST₇) to (a, c) canopy height and (b, d) LAI across 12 plots with altitudinal vegetation changes from forests to shrublands and grasslands in the Sergyemla Mountains. All variables were in natural logarithmic transformation (original data found in Tables 1 and 3). The trend lines were for the data pooled from the 12 plots.

TABLE 4

Coefficients of partial correlations for linear relationships of −20 cm mean soil temperatures for the year (STa, °C), growing season (STg, °C ), January (ST₁, °C) and July (ST₇, °C) to stand variables (canopy height, m; LAI, m²m⁻²) and environmental variables (altitude, m; southness, calculated as cos(aspect − 180°); soil moisture, %) across 12 plots with altitudinal vegetation changes from forests to shrublands and grasslands in the Sergyema Mountains. Because canopy height and leaf area index (LAI) were highly correlated, their correlations with soil temperature were separately analyzed.

TABLE 5

Multiple linear regression models relating seasonal mean soil temperatures at −20 cm (for the year, STa; growing season, STg; July, ST₇) to aspect southness (calculated as cos(aspect − 180°)) and canopy height (H, m) or leaf area index (LAI, m²m⁻²) across 12 plots with altitudinal vegetation changes from forests to shrublands and grasslands in the Sergyema Mountains.

In contrast, seasonal mean soil moistures for the year, growing season, July, and January did not correlate with altitude, aspect, or stand and soil variables (canopy height, LAI, soil bulk density) across the 12 plots (p > 0.10, data not shown).

CONTROLS OF SNOW AND PLANT COVER ON TEMPORAL VARIABILITY OF ALPINE SOIL TEMPERATURE

In winter, daily mean soil temperatures across depths and sites were generally stable (Fig. 2). In spring, the soil warming date was 20–30 days later on the north-facing slope than on the south-facing slope (Fig. 2). At high altitudes or latitudes, snow cover generally plays a key role in determining winter soil temperature and spring soil warming date (Mellander et al., 2005; Wieser and Tausz, 2007). The insulating effect of snow cover prevents the ground from cooling, and the soils under thick snow cover are typically characterized by an even soil temperature regime (Wundram et al., 2010). In boreal forests, deep snow cover needs substantial energy to melt and generally results in a delayed soil warming date in spring (Mellander et al., 2005; Sutinen et al., 2009). Hence, soil temperature is decoupled from air temperature when the ground is covered with insulating snow.

During the snow-free period, vegetation cover exerts a great influence on soil temperature by changing its relationship with air temperature. Gehrig-Fasel et al. (2008) indicated that daily mean air temperature of previous 1 and 2 days can predict well the daily mean soil temperature during the growing season at treelines in Switzerland. However, our observations demonstrated that the air-soil temperature hysteresis during the growing season (June–September) varied significantly with vegetation types (Fig. 2). The −5 cm soil temperature had the highest correlation with the air temperature of the previous 10–21 days at the north-facing sites (N2 and N3) but of the current day or previous day at the south-facing sites (S2 and S3) (from Fig. 2, data not shown). High LAI and large moss/litter accumulation insulate the soils, resulting in the air-soil temperature decoupling during the growing season (Balisky and Burton, 1995; Bond-Lamberty et al., 2005a).

RELATIVE IMPORTANCE OF ALTITUDE, ASPECT, VEGETATION, AND SOIL VARIABLES TO THE SPATIAL VARIABILITY OF SEASONAL MEAN SOIL TEMPERATURE AND MOISTURE

Under timberline canopies on the opposite slopes of a valley, the exponential decrease in sunlight penetration through the closed canopy can create a similar microclimate, especially seasonal mean air temperature (Table 2). The growing season mean soil temperatures at −20 cm for both timberlines (6.0–7.0 °C) fall within the −10 cm threshold temperature of 6.7 ± 0.8 °C found in global climatic treelines (Körner and Paulsen, 2004). However, the growing-season mean soil temperature was 0.8–1.0 K higher at the south-facing juniper timberline than at the north-facing fir timberline (Table 3), which was directly attributed to the contrasting snow depths and/or LAIs between north-facing and south-facing slopes (Figs. 2b, 2d; Table 1). This is in contrast to the previous reports in European treelines where there is a weak effect of aspect on seasonal mean soil temperature (Körner and Paulsen, 2004; Treml and Banaš, 2008). At high altitudes, slope aspect can affect the amount and daily cycle of solar radiation received at different times of the year, and then local microclimatic differences in snow cover duration and soil temperature and moisture become important factors (Rosenberg et al., 1983; Wieser and Tausz, 2007). In general, snow depth and soil thawing significantly influence seasonal mean soil temperature and soil warming date in boreal forests (Mellander et al., 2005).

Our data further indicated that aspect southness and vegetation variables (canopy height and LAI) rather than altitude were the main factors affecting the spatial variability of seasonal mean soil temperatures across both timberline ecotones (Table 4 and Fig. 5). It has been suggested that there is a remarkable influence of plant stature and canopy on seasonal mean soil temperature at high altitudes (Ballard, 1972; Munn et al., 1978; Körner and Paulsen, 2004; Bader et al., 2007). While global radiation may not change significantly with altitude due to the frequency of cloud cover at high altitudes (Körner, 2003), the significant decrease of LAI and canopy height with increasing altitude at the timberline ecotones enhances solar radiation reaching the ground surface (Balisky and Burton, 1995; Redding et al., 2003; Bond-Lamberty et al., 2005a). Additionally, in contrast to upright tall trees, dwarf plants above alpine timberlines can increase soil temperature by enhancing aerodynamic resistance (Green et al., 1984). LAI and canopy height were negatively correlated with seasonal mean soil temperatures across altitudinal vegetation types even though the mean soil temperatures varied little with altitude.

In contrast to the seasonal mean soil temperature, the seasonal mean soil moisture varied insignificantly with altitude, aspect, and stand and soil variables. The daily mean soil moisture across vegetation types was generally more than 20% during the growing season (Fig. 4). Within a plot, daily mean soil moisture was poorly correlated with daily mean soil temperature (negative or positive). It has been suggested that soil moisture has no significant effect on soil temperature in temperate and boreal forests. This is partly because of the lesser sensitivity of thermal diffusivity to soil moisture in a small soil heat flux (Morecroft et al., 1998) and/or the buffer effect of high LAI on soil temperature variability (Bond-Lamberty et al., 2005b).

EFFECTS OF SOIL TEMPERATURE AND MOISTURE ON LOCAL TIMBERLINE DISTRIBUTION PATTERNS

In the Sergyemla Mountains, there is a lack of explanation for the unique distribution pattern of fir and juniper timberlines that dominate at different altitudes on the opposite slopes of a valley. Shi et al. (2008) indicated that the winter air-temperature extreme is pronounced at the valley bottom juniper site where fir is absent. However, we found that the daily air temperature amplitude was identical between fir and juniper timberlines (Figs. 3a, 3c). Our data further indicated that the −5 cm soil temperature amplitude in the growing season was much lower and stable at the north-facing fir timberline (Fig. 3a), compared to the south-facing juniper timberline (Fig. 3c) and the two nearby non-forested sites (Figs. 3b, 3d). The seed-regenerated fir species (including seedlings, saplings, and trees) was excluded from the south-facing juniper timberline and nearby non-forested sites. More interestingly, some root-sprouting seedlings and adult trees of juniper have been observed at the edge of the north-facing fir timberline. Compared to the relatively widespread root-sprouting juniper seedlings, the seed-based fir seedlings (rooting depth <10 cm) were limited to a stable soil temperature environment during the growing season. Since seedling establishment is important to the position of alpine treeline (Smith et al., 2003; Holtmeier and Broll, 2005 ), such a biological difference in forest regeneration adapted to the growing-season soil temperature extreme suggests an explanation for the distribution pattern of fir and juniper timberlines on opposite slopes of a valley in the Sergyemla Mountains. In response to future climate warming, the position of the fir timberline could remain steady because of the difficulty in seedling establishment, while the juniper timberline might advance and even extend to the north-facing slope consistent with the present observation that some seedlings and adult trees of juniper did scatter at the edge of the north-facing fir timberline.

It has been suggested that adaptations of plants to local climate lead to a LAI that is closely coupled with soil water and nitrogen availability (Woodward et al., 1995; Hirose et al., 1997). Given the similar plant stature and canopy cover such as the two non-forested sites nearby both timberlines (N3 vs. S3, Table 1), daily mean soil moisture at −20 cm during the growing season on the north-facing slope was almost double that on the south-facing slope (Figs. 4b, 4d). Between the fir and juniper timberlines, there were large differences in mean tree height and LAI (10–12 m and 6–8 m² m⁻² for the fir trees, 6–8 m and 4–5 m² m⁻² for juniper trees; Table 1), leading to the similar soil moisture contents at −20 cm (Figs. 4a, 4c). Across ecosystems, stand LAI and productivity are negatively correlated with foliar carbon isotope ratio (δ¹³C) that is mainly determined by soil water availability (Luo et al., 2009). The carbon isotope data by Li (2007) indicate that foliar δ¹³C is >3‰ higher in the south-facing juniper trees (−24.80 ± 1.05‰, n  =  48) than in the north-facing fir trees (−27.91 ± 1.08‰, n  =  130). This implies that the juniper is a relatively drought-resistant and low productivity species, and the north-facing fir forests with high LAI and productivity might not maintain the long-term soil water balance on the south-facing slope.