Energy partitioning and controlling factors of evapotranspiration in an alpine meadow in the permafrost region of the Qinghai-Tibet Plateau

Abstract

Energy partitioning and evapotranspiration (ET) of alpine meadows in permafrost areas are crucial for water cycle on the Qinghai-Tibet Plateau. However, seasonal (freeze–thaw cycle) variations in energy partitioning and ET and their driving factors must be clarified. Therefore, 4-year energy fluxes [i.e. latent heat (LE) and sensible heat (H)] were observed, and bulk parameters [i.e. surface conductance, decoupling coefficient (Ω), and Priestley–Taylor coefficient (α)] were estimated in an alpine meadow in the Qinghai-Tibet Plateau. Mean daily LE (27.45 ± 23.89 W/m²) and H (32.51 ± 16.72 W/m²) accounted for 31.71% and 50.14% of available energy, respectively. More available energy was allocated to LE during the rainfall period, while 67.54 ± 28.44% was allocated to H during the frozen period. H was half the LE during rainfall period and seven times the LE during frozen period due to low soil water content and vegetation coverage during the frozen season. Mean annual ET was 347.34 ± 8.39 mm/year, close to mean annual precipitation. Low mean daily Ω (0.45 ± 0.23) and α (0.60 ± 0.29) throughout the year suggested that ET in the alpine meadow was limited by water availability. However, ET was constrained by available energy because of sufficient water supply from precipitation during rainfall season. In contrast, large differences between ET and precipitation indicated that soil water was supplied via lateral flow from melting upstream glaciers and snow during the transition season. The results suggest that seasonal variations in bulk parameters should be considered when simulating water and energy fluxes in permafrost regions.

摘要

青藏高原冻土区高寒草甸能量分配及蒸散发影响因子

高寒草甸的能量分配和蒸散发对青藏高原多年冻土区水循环至关重要。然而，能量分配、蒸散发及其驱动因素的季节变化(冻融循环)仍需要明确。因此，本研究在位于青藏高原风火山流域的高寒草甸进行了为期4年的能量通量(包括潜热和感热)观测，并估算了大气边界参数(包括表面导度，解耦系数和Priestley–Taylor系数)。研究结果表明，研究区日均潜热(27.45 ± 23.89 W/m²)和显热(32.51 ± 16.72 W/m²)分别占可利用能量的31.71%和50.14%。在降雨期，更多可利用能量被分配到潜热；而在冻结期，67.54 ± 28.44%的可利用能量分配给显热。显热在降雨期间是潜热的一半，而由于冻结期较低的土壤水分含量及植被盖度，显热在冻结期间是潜热的7倍。研究区年均蒸散发为347.34 ± 8.39 mm/year，接近年均降水量。较低的日均解耦系数(0.45 ± 0.23)和Priestley–Taylor系数(0.60 ± 0.29)表明高寒草甸的蒸散发受水分供应限制。然而，在降雨期由于降水充足，蒸散发受到可利用能量的限制。在过渡期，蒸散发和降水之间存在较大差异，表明在该季节上游冰川和雪的融水通过侧向流动补给到土壤中。本研究的结果表明，在未来模拟多年冻土区水和能量通量时应考虑大气边界参数的季节变化。

INTRODUCTION

Alpine meadows cover more than 30% of the entire Qinghai-Tibet Plateau (QTP) (Lin et al. 2020), known as the Asian Water Tower, which serves as the headwater of many large rivers in Asia (Immerzeel et al. 2010; Pithan 2010; Wang et al. 2022a). Unfortunately, the QTP, with an average altitude of more than 4000 m above sea level (asl), is one of the most susceptible areas to climate change, experiencing severe permafrost degradation, glacier retreat, and lake expansion (Song et al. 2013; Yao et al. 2018; Zhao et al. 2020). Permafrost occupies 52% of the total area of the QTP (Zou et al. 2017), and the seasonal thawing and freezing processes of the active layers alter the energy fluxes between the land and atmosphere, affecting the energy balance of the entire ecosystem (Yao et al. 2019). Additionally, evapotranspiration (ET), which accounts for 2/3 of the total precipitation (P), plays a crucial role in the water cycle and energy balance of terrestrial ecosystems (Fisher et al. 2017; Hu et al. 2023; Lin et al. 2021; Oki and Kanae 2006). Therefore, identifying the variations in the energy and ET in alpine meadows in permafrost areas could provide valuable information on how hydrological processes and energy fluxes respond to climate changes on the QTP.

The interactions between microclimate and vegetation determine the absorption and release of sensible heat (H) and latent heat (LE), which are the dominant components of the energy cycle of the ecosystem (Sun et al. 2019). The Bowen ratio (β) (the ratio of H to LE) can serve as an indicator of the energy balance of the ecosystem and exhibits different seasonal patterns for various vegetation types (Chen et al. 2022). Seasonal changes in soil moisture (SM) and phenological periods can cause notable variations in the β value, ranging from 0.02 to 17.00, in steppe ecosystems over the QTP (Gu et al. 2005). In contrast, it was previously found that a sufficient water supply, high vapor pressure deficit (VPD) and considerable evaporative cooling effects constrained the β value below 1 during both the growing and non-growing seasons in a meadow steppe (Chen et al. 2022). Seasonal freezing and thawing of the active layers alter hydrothermal conditions and vegetation growth in permafrost areas and in turn affect energy partitioning in these areas (Gu et al. 2015). Previous studies revealed significant seasonal variations in the carbon and water fluxes in different land-cover regions and permafrost regions on the QTP (Lin et al. 2019, 2021). Lin et al. (2023) indicated that the water use efficiency could exhibit the opposite response to environmental changes between alpine meadows and steppes. Therefore, quantifying the patterns of energy partitioning for a specific vegetation type, i.e. alpine meadow, could aid in fully understanding the energy cycle over the TP. However, the seasonal patterns of energy partitioning in alpine meadows in permafrost areas remain unclear, limiting our understanding of the seasonal energy distribution in the soil–plant–atmosphere continuum in high-altitude regions.

Theoretically, high LE and low H fluxes are induced by a high atmospheric water demand, water and/or energy availability, and interactions between the surface and atmosphere, indicating higher water loss through ET. Plant transpiration affects ET by regulating stomatal behavior at the leaf scale (Costa et al. 2010; Coupel-Ledru et al. 2016; D’Odorico et al. 2020). However, to explain the variations in ET at the ecosystem scale, radiative and aerodynamic properties, in addition to the canopy stomatal conductance (g_s), must be included (Baldocchi et al. 1991; Green et al. 2020; Jarvis and Mcnaughton 1986). Therefore, the surface conductance (G_s) is recommended to characterize the vapor exchange between the surface and atmosphere at the ecosystem scale (Harris et al. 2004; Kumagai et al. 2004; Marques et al. 2020). G_s is influenced by meteorological factors (e.g. radiation, VPD, and air temperature), hydrological conditions (e.g. precipitation and soil water content), and biological factors (e.g. leaf area index and g_s) (Marques et al. 2020; Zha et al. 2013). Furthermore, accurate estimation of G_s is crucial for simulating ET when using land surface models (De Kauwe et al. 2017; Marques et al. 2020). Other bulk surface parameters, such as the Priestley–Taylor coefficient (α) and decoupling coefficient (Ω), can be utilized to estimate the effects of environmental and biological factors on ET (Wang et al. 2022c; Yue et al. 2022). For example, Ω indicates the coupling between vegetation and the atmosphere; a high Ω value indicates low surface roughness, aerodynamic conductance, and canopy boundary layer conductance levels, which reduce the exchange rate between the vegetation surface and atmosphere (De Kauwe et al. 2017; Jarvis and Mcnaughton 1986). Thus, the bulk surface parameters could serve as better indicators than single environmental factors to characterize water and energy exchange between the land surface and atmosphere (Xu et al. 2022). However, due to the inefficiency of the observed data and changes in climate and vegetation, it is challenging to quantify the relationship between ET and bulk surface parameters in permafrost areas (Ma et al. 2015). Therefore, it is necessary to elucidate the mechanisms of how bulk surface parameters respond to meteorological, hydrological, and biological factors to better understand the driving factors of ET in alpine meadows in permafrost areas on the QTP.

The regulation of energetic and hydrological processes by biological, environmental, and climatic factors in dry regions has remained a critical research topic (Hu et al. 2022; Sun et al. 2019; Xu et al. 2022; Yao et al. 2015). Observing and analyzing the magnitude and seasonal/interannual variations in bulk surface parameters could help us distinguish the impacts of natural changes and human activities on the water cycle and improve the estimation of land surface processes used in hydrometeorological and climate models (De Kauwe et al. 2017; Marques et al. 2020). Previous studies have focused on illustrating the relationship between bulk surface parameters and the energy distribution and water processes between the land surface and atmosphere (Park et al. 2020; Xu et al. 2022). However, in contrast to other ecosystem at lower altitudes, the unique climate and terrain conditions in high-altitude regions, especially in permafrost regions, result in distinct seasonal changes in ecohydrological processes (Wei et al. 2021; Yang et al. 2023). Moreover, different vegetation types should also be considered to separate the spatial variations in energy partitioning over the TP (Li et al. 2022; Lin et al. 2023). However, the coupling mechanism between climate, hydrology, and the underlying surface of the alpine meadow in the hinterland of the QTP has been neglected due to the harsh and carriable climate. The present observations and research are mainly concentrated in the northeastern and southern parts of the QTP, where site-scale observation data are sufficient (Lin et al. 2023; Ma et al. 2015; Qin et al. 2009). Therefore, this study aimed to (i) reveal the energy partition patterns during different hydrothermal periods and (ii) identify the driving factors behind the seasonal changes in ET using bulk surface parameters of the alpine meadow in the permafrost region on the QTP.

MATERIALS AND METHODS

Site description

The study site occurs in the Fenghuo Mountain region (34° 43ʹ 3.8″ N, 92° 53ʹ 19.7″ E, 4754 m a.s.l.), which is located in the hinterlands of the QTP and is characterized by a permafrost-underlain meadow (Fig. 1a). The depth of the active layer ranges from 0.8 to 1.5 m (Wang et al. 2008; Wu et al. 2010). The mean annual air temperature (T_air) and P from 2000 to 2020 were −5.11 ± 0.46°C and 351.99 ± 57.42 mm/year, respectively (Fig. 1b). The mean monthly T_air was above 0°C, and P accounted for more than 85% of the annual P from June to September. The alpine meadow is the most extensive vegetation type on the QTP, with the vegetation coverage exceeding 70% in the study area (Zhang et al. 2015). The dominant plant species include Kobresia pygmaea, K. humilis Sergievskaya, K. capillifolia, K. myosuroides Foiri, K. graminifolia, Carex atrofusca Schkuhr subsp. minor, and Carex scabriostris (Chen et al. 2017).

Meteorological data

In 2012, a meteorological station was established in the study area. Sensors were installed at the 2-m height to measure T_air and relative humidity (HMP 155A, Vaisala, Finland), wind speed (u) and direction (03002, RM Young, USA), and photosynthetic active radiation (LI190SB, Li-Cor, USA). Additionally, the NR01 sensor (Hukseflux, Netherlands) was used to measure the upward longwave radiation (R_ul), upward shortwave radiation (R_us), downward longwave radiation (R_dl), and downward shortwave radiation (R_ds) at the same height. The net radiation was calculated by $Rn=(Rdl+Rds)−(Rul+Rus)$⁠. P was measured with a snow and rain gauge (T-200B, Geonor, Norway). Surface soil heat fluxes (G) were measured with an HFP01SC instrument (Hukseflux, Netherlands). Sensors were installed at four depths (5, 10, 20, and 40 cm) to measure soil temperature (T_soil) (ST01, Hukseflux, Netherlands) and SM (CS616, Campbell Scientific, USA). All meteorological data were recorded by a CR1000 data logger (Campbell, USA) at 30-min intervals.

Flux data

A flux tower was installed in August 2014 and equipped with an open-path infrared gas analyzer (Li-7500A, Li-Cor, USA) and a three-dimensional sonic anemometer (Windmaster Pro, Gill, UK) at a height of 2 m. The fluxes of CO₂, LE, and H were recorded at 10 Hz. To estimate the footprint of the flux measurement, the prediction model of Kljun et al. (2015) was used, shown that 90% of the fluxes occurred within a radius of 200–300 m of the flux tower (Song et al. 2020). The raw data were processed in EddyPro software (Li-Cor, USA), including planar fit coordinate rotation (Wilczak et al. 2001), lag correction of H₂O and CO₂ relative to the vertical wind component, spike count/removal (Mauder et al. 2013), and Webb–Pearman–Leuning density fluctuation correction (Webb et al. 1980). The output included 30-min fluxes (e.g. H, LE, and CO₂) and air properties (e.g. T_air, air pressure, air density, air heat capacity, air molar volume, and relative humidity). Values outside the ranges of −100 to 500 W/m² for H and −50 to 450 W/m² for LE were excluded. Data gaps existed after the above processes, at 21.38% for LE and 17.78% for H. Gap filling was performed using the REddyProc R package, downloaded from https://www.bgc-jena.mpg.de/bgi/index.php/Services/REddyProcWeb, to achieve a continuous dataset.

Bulk surface parameters

G_s was calculated using the Penman–Monteith equation (Monteith and Unsworth 2013), expressed as:

where ρ is the air density (kg/m³), C_p is the specific heat of moist air at a constant pressure (J/(kg K)), VPD is the vapor pressure deficit (kPa), Δ is the slope of the saturation vapor pressure curve relative to the temperature (kPa/°C), γ is the psychrometric constant (kPa/K), β is the Bowen ratio, and G_a is the aerodynamic conductance (m/s) (Table 1). G_a is the reciprocal of the aerodynamic resistance to water vapor transport (r_a), which comprises the aerodynamic resistance of momentum (r_aM) and the quasi-laminar layer resistance (r_b). These r_a components can be obtained follows Tan et al. (2019):

where G_aM and u_* are the aerodynamic conductance for momentum (m/s) and friction velocity (m/s), respectively.

The Ω, ranging from 0 to 1, can be used to evaluate the degree of interaction between the surface and atmosphere and to describe the sensitivity of ET to changes in the surface conductance (Jarvis and McNaughton 1986). ET is generally dominated by the available energy (the available energy is defined as the difference between R_n and G) when Ω approaches 1 (indicating that the surface is completely decoupled from the overhead conditions); in contrast, ET is driven by the bulk surface conductance and atmospheric humidity deficit when Ω approaches 0 (Baldocchi and Xu 2007). The equation of Jarvis and McNaughton (1986) can be used to estimate Ω:

To determine whether the atmospheric demand or surface moisture is the dominant factor influencing ET, the Priestley–Taylor coefficient (α) was calculated according to Priestley and Taylor (1972). α is the ratio of ET to the equilibrium evapotranspiration (ET_eq):

Statistical analysis

To avoid numerical instability when the dominator approaches 0, the local daytime (9:30–16:00) values of the bulk parameters (G_s, G_a, Ω, and α) were used in the calculations in this study. The values on snow days were also excluded (albedo > 0.3, albedo = R_us/R_sd).

Data from 1 October 2014 to 30 September 2016 and 1 October 2018 to 30 September 2020 were used in this study due to instrument malfunctions in 2017 and 2018. We divided the freeze–thaw period according to the multiyear mean T_soil and P. From the long-term records, P during the June to September period accounted for approximately 85% of the annual P. The period with T_soil below 0°C was defined as the frozen period. According to the multiyear mean daily values, the soil began to freeze from approximately 16 October to 15 April of the next year. Then, the freeze–thaw cycle was divided into three periods: the frozen period (16 October–15 April), the rainfall period (1 June–30 September), and the transition period (the rest time).

The independent samples t-test was conducted to analyze the differences in R_n, G, LE, and H between the different seasons. Least squares linear regression analysis between the available energy (R_n − G) and turbulence energy (LE + H) was used to analyze the energy colure of the flux measurement. Correlation analysis was performed to identify the relationship between G_s and VPD, SM, and normalized difference vegetation index (NDVI). The relationship between α and VPD and G_s was also analyzed through correlation analysis.

RESULTS

Variations in the meteorological factors

The mean daily T_air was −3.93 ± 8.15°C and 2.30°C lower than the mean daily T_soil throughout the year (Fig. 2a). The mean daily T_air was −10.77 ± 4.79°C, 4.99 ± 2.83°C, and −1.31 ± 2.76°C during the frozen, rainfall, and transition periods, respectively. Moreover, the mean daily T_soil was −7.96 ± 4.07°C, 6.60 ± 2.47°C, and 0.82 ± 1.43°C during the frozen, rainfall, and transition periods, respectively. The mean daily R_s and VPD were higher during the rainfall period than during the other two periods (Fig. 2b). There were no significant seasonal patterns in wind speed (Fig. 2c). The seasonal changes in SM followed the variations in P (Fig. 2d), with P during the rainy season accounting for approximately 90.31% of the total annual P. The highest NDVI occurred during the rainy season and decreased to nearly 0 during the frozen period (Fig. 2e).

Energy partitioning and energy balance

The diurnal patterns of R_n, G, LE, and H exhibited unimodal curves in every month (Fig. 3a–d). The highest daily R_n, G, LE, and H were 201.78, 16.92, 101.07, and 93.99 W/m², respectively. The mean daily R_n, G, LE, and H were 73.13 ± 46.18, −1.86 ± 6.87, 27.45 ± 23.89, and 32.51 ± 16.72 W/m², respectively (Fig. 3e–h). R_n and LE showed significant seasonal changes, with the values following the order patterns of rainfall period > transition period > frozen period. H did not show significant differences between the three seasons (P > 0.05).

The daily (LE + H) exhibited a high correlation with the daily (R_n − G), with a slope of 0.71 and R² of 0.85 (Fig. 4). The mean daily LE/(R_n − G) and H/(R_n − G) were 0.32 ± 0.19 and 0.50 ± 0.28, respectively (Fig. 5a and b). The mean daily LE/(R_n − G) was the highest during the rainfall period (0.48 ± 0.07), followed by the transition period (0.42 ± 0.11) and the frozen period (0.17 ± 0.14). However, the mean daily H/(R_n − G) was the highest during the frozen period (0.68 ± 0.28) but the lowest during the rainfall period (0.29 ± 0.09). The mean daily H was almost half the LE value during the rainfall period (β = 0.63 ± 0.25) but approximately seven times higher than LE during the frozen period (β = 6.99 ± 5.34, Fig. 5c). During the transition period, β was approximately 1.

Temporal changes in ET

The mean annual ET was 347.34 ± 8.39 mm/year (Fig. 6). The ET was 72.15 ± 6.41, 45.96 ± 4.98, and 229.24 ± 6.38 mm during the transition, frozen, and rainfall periods, respectively. The mean daily ET was 1.88 mm/day during the rainfall season, higher than that during the transition period (1.11 mm/day), and the frozen season (0.25 mm/day). The highest daily ET could reach 3.52 mm/day. The highest monthly ET occurred in July and August, ranging from 63.78 to 72.17 mm/month during the study period.

Bulk surface parameters

The mean daily daytime G_s, Ω, and α were 8.45 ± 6.10 mm/s, 0.45 ± 0.23, and 0.60 ± 0.29, respectively (Fig. 7). The mean daily daytime values of these three parameters exhibited the same seasonal patterns, i.e. rainfall period > transition period > frozen period. During the rainfall season, the mean daily daytime G_s, Ω, and α were 11.44 ± 5.01 mm/s, 0.63 ± 0.12, and 0.82 ± 0.11, respectively, while during the frozen season they were only 4.24 ± 3.79 m/s, 0.23 ± 0.10, and 0.28 ± 0.14, respectively.

The daily daytime G_s decreased with increasing VPD during the frozen (R² = 0.11, P < 0.05), transition (R² = 0.51, P < 0.05), and rainfall (R² = 0.49, P < 0.05) periods (Fig. 8a). However, the daily daytime G_s showed a positive correlation with SM during the rainfall period (Fig. 8b). No significant relationship was observed between G_s and SM during the transition and frozen seasons (P > 0.05). The daily daytime G_s showed a positive correlation with the NDVI during the transition and rainfall periods (Fig. 8c). The daily daytime α decreased with increasing VPD during the transition (R² = 0.22, P < 0.05) and rainfall (R² = 0.17, P < 0.05) seasons, but the correlation was not significant during the frozen season (P > 0.05, Fig. 9a). The daily daytime α was highly correlated with G_s during all periods (P < 0.05, Fig. 9b).

DISCUSSION

Energy balance and partitioning

Linear regression was conducted to assess the closure of the energy balance, using the sum of the daily sensible and latent heat fluxes (LE + H) as the dependent variable and the difference between the daily net radiation and the ground heat flux (R_n − G) as the independent variable. Our results indicated that the daily (LE + H) was lower than the available energy. It was found that the eddy covariance data caused underestimation of the available energy, although the degree of closure in our study was comparable to that in different ecosystems in China (Li et al. 2005) and globally (Wilson et al. 2002). There are several possible reasons for the energy imbalance, e.g. systemic sampling errors in the flux measurements, systematic instrument bias, neglect of other energy sinks, loss of low- and high-frequency turbulences and advection effects (Li et al. 2005; Wilson et al. 2002). Regular instrument maintenance and data quality monitoring, particularly under adverse weather conditions during the non-growing season, are crucial for reducing the data loss resulting from instrument malfunctions and power supply issues, thereby improving the quality of flux observation data and enhancing energy closure in eddy covariance systems (Li et al. 2005). In addition, statistical energy correction methods were also recommended before analyzing water and carbon fluxes (Hu et al. 2018, 2023; Twine et al. 2000).

The mean daily daytime β value in this study occurred within the range of previous studies on the QTP (Ma et al. 2015; Shang et al. 2015; Wang et al. 2022b). However, the β value in our study was higher than that of an alpine wetland in the central QTP region and a grassland in the eastern QTP region (Shang et al. 2015; Wang et al. 2022b), indicating drier conditions at our study site. The higher β and H/(R_n − G) during the frozen season than during the rainfall season indicated that more of the available energy was allocated to LE during the rainfall season, while more of the energy was allocated to H during the frozen period. Similar seasonal patterns in energy partitioning were also observed in the alpine steppe and alpine meadow ecosystems in the northern, central, and eastern parts of the QTP (Gu et al. 2005; Wang et al. 2018; Yao et al. 2008). The vegetation phenology and water availability (including P and SM) were determined as the driving factors of energy partitioning (Gu et al. 2005; Yue et al. 2022). Chen et al. (2022) indicated that the increase in the available water and vegetation development were the main reasons for the increase in LE during the wet season. The onset of plant senescence and decrease in the canopy conductance led to changes in LE during the transition period. A negative H value was observed due to the advection of H, indicating that H still provides energy for vapor exchange during the frozen period (Lei and Yang 2010). However, studies in the arid zone showed different energy distribution patterns to ours (Bian et al. 2003; Chen et al. 2022). In the western QTP region, which is occupied by an arid desert steppe, the extremely low P level (less than 100 mm/year) leads to low LE even during the rainfall season (Bian et al. 2003). As a result, the energy allocated to H is higher than that allocated to LE throughout the year in the arid desert steppe ecosystem. Chen et al. (2022) also found that β in the meadow steppe remained stable over different seasons. Moreover, vegetation cover induces a change in surface characteristics, affecting energy partitioning patterns. For example, energy exchange in arid desert areas (e.g. the Gbo desert steppe) was dominated by H, while LE dominated vegetated areas (e.g. arid desert wetland, Wang et al. 2022b).

Driving factors of ET

G_s, determined by both the boundary layer conductance and aerodynamic conductance, is affected by leaf and canopy properties such as the leaf area, canopy stomatal conductance, structure, and roughness (De Kauwe et al. 2017). Therefore, the seasonal changes in G_s can be explained by the leaf phenology (Krishnan et al. 2012; Marques et al. 2020; Yue et al. 2022). During the transition and rainfall periods, G_s was strongly related to the NDVI, indicating that the morphophysiological state also affected ET (Marques et al. 2020). The larger leaf areas during the rainfall period indicated more stomata for transpiration and more water intercepted by the canopy for evaporation (Dieleman et al. 2012; Hu et al. 2017; Raz-Yaseef et al. 2010). Additionally, VPD imposed a significant negative effect on G_s during the rainfall period. On the one hand, a higher VPD could cause ET reduction by constraining g_s (Medlyn et al. 2017; Rigden and Salvucci 2017); on the other hand, a higher VPD could result in an increase in ET due to the notable vapor pressure difference between air and stomatal/soil pores (Hsu et al. 2021). Moreover, a high VPD is accompanied by a low water availability (low soil water content) for plant growth. Therefore, VPD, SM, and NDVI jointly controlled ET during the transition and rainfall periods at the study site. In contrast, the relationship between G_s and VPD, SM, and NDVI was nonsignificant during the frozen period. The decrease in G_s values during this period was caused by leaf senescence and an insufficient water supply, which prevented excessive water loss from plants and resulted in ET reduction (Marques et al. 2020). In addition, the soil conductance mainly affected G_s, which did not change with VPD variation when vegetation withered (Wang et al. 2022b).

Compared with some forest ecosystems, the mean Ω value was low in our study, indicating that ET in the study area is limited by the water availability (Bracho et al. 2008; De Kauwe et al. 2017; Zhang et al. 2016). This finding is consistent with those of previous studies conducted in other parts of the QTP, where P was low and concentrated throughout the year (Ma et al. 2015; Wang et al. 2022c). However, during the rainfall period, the mean daytime Ω value increased, suggesting that the studied area was energy limited. This phenomenon is not unique to the study area. In addition to maintaining the alpine steppe in the central QTP region during the rainy season, Ma et al. (2015) determined that the available energy exerted the greatest impact on the hydrological cycle. More than 90% of P occurred during this period, providing ample water for plant transpiration and canopy/soil evaporation. Furthermore, frequent convective activities resulted in large radiation fluctuations during the rainfall season, which constrained ET (Wang et al. 2022b). During the frozen period, the mean Ω value sharply decreased, indicating strong control of ET by the surface conductance. Although the solar radiation and VPD declined during this period, the atmospheric demand for ET was also lower. Therefore, the aerodynamic and radiation terms were not the primary limiting factors of ET during the frozen period. Instead, the low water availability supply stemming P and thawing water in the active layers limited ET. Thus, the primary driving factor of ET variation was SM during the frozen period.

Previous studies have considered an α value of 1.26 for surfaces with a low roughness (Hobbins et al. 2001; Priestley and Taylor 1972). However, further research has shown that the α value varies depending on the site (Hu et al. 2018; Liu et al. 2016; Zhang et al. 2017). For example, Liu et al. (2016) found a negative correlation between α and the aridity index, indicating that areas with higher α values tend to be drier. The mean α value in our study also suggests that alpine meadow permafrost with a high α value is water limited. Furthermore, the significant correlation between α and G_s is consistent with previous studies on grasslands, suggesting that G_s significantly affects ET (Krishnan et al. 2012; Ma et al. 2015). The nonlinear relationship approximately agrees with theoretical studies on surface ET. Previous studies found that α is sensitive to G_s when G_s is below a certain threshold, and the threshold varied according to the ecosystem and climate (Ma et al. 2015; Marques et al. 2020; Tan et al. 2019; Zha et al. 2013). Specifically, α was higher in areas controlled by R_n than in areas controlled by SM. Generally, the G_s value reflects the effect of biological factors on ET change, while the ET_eq value reflects the effect of environmental factors on ET variation (Wang et al. 2022c). Research has shown that changes in the actual ET are primarily influenced by changes in P rather than in ET_eq in drought-prone regions (Yang et al. 2007). However, the actual ET variability was mainly explained by ET_eq, and no significant relationship was observed between ET and P at our study site.

CONCLUSIONS

To reveal the seasonal variations in energy partitioning and the driving factors of ET variation, energy fluxes were observed using the eddy covariance technique, and bulk surface parameters were calculated for four complete freeze–thaw cycles in the alpine meadow of the permafrost region on the QTP. During the rainfall season, the available energy was more notably allocated to LE due to the sufficient available water for supporting plant growth and evaporation. In contrast, the mean daily H remained stable across the different seasons and was seven times higher than LE during the frozen season due to the low temperatures and dry weather conditions. The mean annual ET was 347.34 ± 8.39 mm/year, close to the alpine meadow precipitation. Although the low mean annual bulk surface parameters suggested that the alpine meadow was water limited, the available energy was the limiting factor of ET during the rainfall period in such this dry area because of the sufficient water supply from P. However, ET was approximately five times higher than P during the transition period, indicating that a large amount of water should be supplied by lateral flow from melting upstream glaciers and snow during the transition season.

Acknowledgements

We thank Xiaopeng Chen, Tianxu Mao, and Tao Zhang for their help in data collection of this study. We are grateful to the Editors and anonymous referees for providing valuable comments.

Funding

This research was supported by the National Key Research and Development Program of China (2022YFC3201702), National Natural Science Foundation of China (U2240226, 42201146), Sanjiangyuan National Park Joint Research Program of Chinese Academy of Sciences and The People’s Government of Qinghai Province (LHZX-2020-10-3), and Science and Technology Project of Sichuan Province (2022NSFSC1001).

Conflict of interest statement. The authors declare that they have no conflict of interest.

REFERENCES