Comparison of Actual and Reference Evapotranspiration Between Seasonally Frozen and Permafrost Soils on the Tibetan Plateau

Abstract

A comparison of evapotranspiration between seasonally frozen and permafrost soils has important theoretical value for studying land surface processes and ecological environmental evolution on the Tibetan Plateau. In this work, the actual (ET_a) and reference (ET₀) evapotranspiration and crop coefficient (K_c) were calculated via eddy covariance data and meteorological gradient data from sites in the Naqu Prefecture and Tanggula Mountains. The variations, differences, and factors influencing the ET_a and ET₀ were analysed. The results revealed that at the two sites in 2008, the annual total ET_a values were 493.53 and 585.17 mm, which accounted for 83.58% and 144.39% of the total annual rainfall, respectively. The ET_a at the Naqu site was affected mainly by the Tibetan Plateau monsoon (TPM), whereas the ET_a at the Tanggula site was strongly affected by both the TPM and the freezing–thawing processes of the permafrost. The annual total ET₀ values at the two sites were 819.95 and 673.15 mm, respectively. The monthly total ET₀ at the Naqu site was greater than that at the Tanggula site. The ET_a and ET₀ values at the two sites were low in winter–spring, high in summer–autumn, and concentrated from May to October. When snow was present, the ET_a values at the Naqu site were relatively high, and the ET₀ values at both sites were very small and even negative at the Naqu site. The ET_a and ET₀ values at the two sites were significantly positively correlated with the net radiation (R_n), surface temperature (T₀), air temperature (T_a), water vapour pressure (e) and soil water content (smc), and negatively correlated with the wind speed (ws). The correlation between the ET_a and the T₀ at the Naqu site was the most significant, and the coefficient of partial correlation was 0.812; meanwhile, the correlation between the ET_a and the smc at the Tanggula site was the most significant, and the coefficient of partial correlation was 0.791. The R_n at the Naqu and Tanggula sites both had greater impacts on the ET₀.

1. Introduction

Evapotranspiration (ET) is a crucial component of the process of water and energy budgeting and is a climatic factor of equal importance to precipitation [1,2]. According to Chahine [3], 65% of rainfall over land comes from evaporation from land areas. Accurately determining the ET is crucial for research in agricultural and ecological environments and hydrology, e.g., managing water resources [4], optimising efficient water-saving irrigation in arid regions [5], reducing wastewater discharge into the environment [6], and determining the impact process and feedback mechanism of climate warming on water resources [7].

The Tibetan Plateau (TP) is known as the “Asian water tower” because of its abundant water. The ET process on the TP governs the variations between soil moisture and atmospheric humidity, directly influencing the regional circulation patterns of the Asian monsoon system [8,9], which is a crucial component of the global energy and water cycles [10]. Many scholars conducted research on the ET at various locations and underlying surface types on the TP in recent years [11,12,13,14,15,16,17,18]. However, most of these studies focused on features at a single site or the whole TP, and few studies have compared the ET between seasonally frozen and permafrost soils across the TP. It is currently unclear whether differences in the ET and its drivers exist between permafrost and seasonally frozen soils [19].

The area of the TP is approximately 257 × 10⁴ km² [20]. Owing to the relatively high elevation and harsh and cold climatic conditions, 96% of the TP area is frozen soil, of which approximately 56% is seasonally frozen soil and approximately 40% is permafrost soil [21]. Therefore, seasonally frozen and permafrost soils are important components of the TP ecosystem. There are differences in the hydrothermal properties and physical processes between regions with seasonally frozen soil and permafrost [22,23,24], and these differences cause different ET processes. A previous study revealed that the ET values of ground surfaces that freeze and unfreeze greatly differ [25]. The ET processes over the TP are closely linked to the climate and hydrology, as well as other biological, physical, and chemical processes. These processes lead to climate change and atmospheric circulation at various scales, from local to global, and then provide direct or indirect feedback to the climate system [26,27,28]. Hence, comparing the ET between seasonally frozen and permafrost soils across the TP is extremely important for predicting climate change, understanding the ecological environment and conducting sustainable development research.

As technology has developed, the eddy covariance (EC) technique provides a direct method for measuring actual evapotranspiration (ET_a) that is widely used worldwide, but these devices are expensive [29]. The reference evapotranspiration (ET₀) is an important component of irrigated agriculture, is widely used to estimate crop water requirements, and can be obtained with only routine meteorological observation data [30]. The crop coefficient (K_c) can be used to determine the ET_a without equipment for direct measurements [30].

In this work, EC data and meteorological gradient data from the Naqu (located in the seasonally frozen soil region) and Tanggula (located in the permafrost soil region) sites are used, and the ET_a, ET₀, and K_c at the two sites were estimated. The variations, differences, and factors influencing the ET_a and ET₀ between the seasonally frozen and permafrost soils across the TP were analysed, with a focus on the impacts of the Tibetan Plateau monsoon (TPM) and soil freezing–thawing processes. We aimed to provide a reference for studying ecological environmental protection and land surface processes on the TP. We hope to provide an accurate ground verification reference for satellite observations and model simulations.

2. Sites and Data

2.1. Sites

In this work, two monitoring sites were chosen: the Naqu site, which is in the Naqu Valley region of the TP, and the Tanggula site, which is located at the Tanggula Pass on the TP (see Figure 1).

The Naqu site is near Niaoqu village, Seni District, Naqu city, Tibet Autonomous Region (91.90°E, 31.37°N), and the elevation is 4509 m above sea level (a.s.l.). This site is located within a plateau subfrigid semihumid climate zone. The ground at the site is flat terrain with expansive surrounding topography and is characterised predominantly by alpine grasslands where the vegetation height is less than 20 cm.

The Tanggula site is located on the gentle slope of the Tanggula Mountain Pass near the Qinghai–Tibet Highway (91.94°E, 33.06°N), located in a plateau frigid semihumid climate zone, and the elevation is 5100 m a.s.l. The ground at the site had flat terrain with expansive surrounding topography, covered by alpine grasslands where the vegetation height is less than 10 cm.

2.2. Data

In this work, the two sites both contained an EC system that was used to calculate the ET_a and a gradient meteorological tower that was used to calculate the ET₀. The study period was the entire year of 2008. In this research, all the times were reported in Beijing time (BJT).

At the Naqu site, the EC system was positioned 20 m above the ground on a tower and composed of an LI-7500 open-path infrared gas analyser (LI-COR Inc., Lincoln, NE, USA), which measures H₂O and CO₂ densities and virtual temperature, and a DAT-600 ultrasonic anemometer (Kaijo Corporation, Tokyo, Japan), which measures three-dimensional wind speed with a sampling frequency of 10 Hz. The data were collected via a CR5000 data logger (Campbell Scientific, Logan, UT, USA). The 10 m gradient meteorological tower was composed of four-component radiation (1.5 m), air temperature and relative humidity (1 and 8.2 m), wind speed (1, 5, and 10 m), wind direction (10 m), air pressure (0.5 m), precipitation (1 m), snow depth (3 m), soil moisture content (4 and 20 cm), soil temperature (0, 4, 10, 20, and 40 cm), and soil heat flux (10 and 20 cm) measurements.

At the Tanggula site, the EC system was located 3 m above the ground on a tripod and was composed of an LI-7500 open-path infrared gas analyser, which measures H₂O and CO₂ densities and virtual temperature, and a CSAT3 ultrasonic anemometer–thermometer (Campbell Scientific, Logan, UT, USA), which measures three-dimensional wind speed with a sampling frequency of 10 Hz. The data logger was also a CR5000 data collector. The 10 m gradient meteorological tower was composed of four-component radiation (2 m), air temperature and relative humidity and wind speed (2, 5, and 10 m), wind direction (10 m), precipitation (1 m), snow depth (2 m), soil temperature (2, 5, 10, 20, and 40 cm), and soil moisture content and soil heat flux (5, 10, and 20 cm) measurements.

3. Methods

3.1. Calculating ET_a

The ET_a was calculated via the EC method as follows:

$$E{T}_a={\displaystyle \frac{\lambda E\cdot s}{L_v\cdot \rho }}$$

(1)

$$L_{\mathrm{v}}=2.5\times {10}^6-2323\cdot T$$

(2)

where ET_a represents the daily actual evapotranspiration (mm/day), λE∙s represents the daily latent heat flux (W/(m²∙day)), ρ represents the water density (kg/m³), L_v represents the latent heat of water (J/kg), and T represents the air temperature (°C).

The EC method has a solid theoretical basis and wide-ranging applications and is recognised as having a high accuracy. In this study, Edire software version 7.1 (Clement R., University of Edinburgh, Edinburgh, UK) was used to process the EC data. Before the calculation of the data, quality control analysis of the EC data was carried out, which included a quality inspection of the raw EC data, spike detection and removal, and time lag correction. Moreover, the estimation of the latent heat flux (λE) necessitated a series of corrections, including sonic temperature correction, coordinate rotations, spectral correction in the high/low-frequency response, and the Webb–Pearman–Leuning correction [31].

For the missing EC observation data, ET_a was calculated via the aerodynamic method to estimate λE using the 10-metre meteorological tower data as follows [32]:

$$LE=0.622\cdot Lv\cdot \rho \cdot k^2\cdot {\displaystyle \frac{[e_a(z_2)-e_a(z_1)]\cdot [u(z_1)-u(z_2)]}{P\cdot {[\ln (\frac{z_2-d}{z_1-d})]}^2}}\cdot {({\Phi }_M\cdot {\Phi }_V)}^{-1}$$

(3)

where ρ is the air density (kg/m³), k = 0.4 is the von Karman constant, e_a is the actual vapour pressure (hPa), u is the wind velocity (m/s), z₁ = 2 m, z₂ = 10 m, P is the air pressure (hPa), d is the zero displacement height (m), Φ_M is the stability function for momentum, and Φ_V is the stability function for water vapour transport.

3.2. Calculating ET₀

The FAO56 Penman–Monteith equation is a widely used method for calculating ET₀ and was proposed by the Food and Agriculture Organization (FAO) of the United Nations; this equation can use general meteorological data to calculate ET₀. The reference surface is defined as “a hypothetical reference crop with an assumed crop height of 0.12 m, a fixed surface resistance of 70 s/m, and an albedo of 0.23” [33].

The ET₀ was estimated via the FAO56 Penman–Monteith equation as follows [33]:

$$E{T}_0={\displaystyle \frac{0.408\Delta (Rn-G)+\gamma \frac{900}{T+273}{u}_2(e_s-e_a)}{\Delta +\gamma (1+0.34u_2)}}$$

(4)

where ET₀ is the daily reference evapotranspiration (mm/day), Δ is the slope of the vapour pressure curve (kPa/°C), R_n is the daily net radiation flux at the crop surface (MJ/(m²∙day)), G is the daily soil heat flux (MJ/(m²∙day)), γ is the psychrometric constant (kPa/°C), T is the mean daily air temperature at a height of 2 m (°C), u₂ is the wind speed at a height of 2 m (m/s), e_s is the saturation vapour pressure (kPa), and e_a is the actual vapour pressure (kPa).

In this study, Δ is given by

$$\Delta ={\displaystyle \frac{4098[0.6108\exp (\frac{17.27T}{T+237.3})]}{{(T+237.3)}^2}}$$

(5)

The R_n was calculated via four-component radiation data as follows:

R_n = S_d − S_u + L_d − L_u

(6)

where S_u is the daily upwards shortwave radiation (MJ/(m²∙day)), S_d is the daily downwards shortwave radiation (MJ/(m²∙day)), L_u is the daily upwards longwave radiation (MJ/(m²∙day)), and L_d is the daily downwards longwave radiation (MJ/(m²∙day)).

The G was calculated according to the following equation [34]:

$$\mathrm{G}\approx \overline{\mathrm{C}}\left(0.01\times {\displaystyle \frac{\partial {\mathrm{T}}_{0\mathrm{cm}}}{\partial \mathrm{t}}}+0.06\times {\displaystyle \frac{\partial {\mathrm{T}}_{5\mathrm{cm}}}{\partial \mathrm{t}}}+0.03\times {\displaystyle \frac{\partial {\mathrm{T}}_{10\mathrm{cm}}}{\partial \mathrm{t}}}\right)+{\mathrm{G}}_{10\mathrm{cm}}$$

(7)

where G represents the daily soil heat flux (MJ/(m²∙day)); $\overline{\mathrm{C}}$ represents the average volume heat capacity (J/(m³∙K)); T_0cm, T_5cm, and T_10cm represent the soil temperatures (K) at 0 cm, 5 cm, and 10 cm, respectively; and G_10cm represents the soil heat flux (W/m²) at a depth of 10 cm.

γ is given by

$$\gamma =0.665\times {10}^{-3}P$$

(8)

e_s is given by

$$e_s={\displaystyle \frac{e^0(T_{\max })+e^0(T_{\min })}{2}}$$

(9)

$$e^0(T)=0.6108\exp ({\displaystyle \frac{17.27T}{T+237.3}})$$

(10)

where e⁰(T) is the saturation vapour pressure (kPa) at air temperature T (°C).

e_a is given by

$$e_a={\displaystyle \frac{e^0(T_{\max })\frac{R{H}_{\min }}{100}+e^0(T_{\min })\frac{R{H}_{\max }}{100}}{2}}$$

(11)

where RH_max is the maximum relative humidity (%), and RH_min is the minimum relative humidity (%).

3.3. Calculating Kc

Kc is the ratio of ET_a to ET₀ [30]:

$$K_c=E{T}_a/E{T}_0$$

(12)

4. Results

4.1. Variations in Meteorological Elements

The Naqu site is located in the seasonally frozen soil region. In January and February, the shallow soil was frozen, and the maximum depth of the freezing soil layer was approximately 1.5 m [31], below which it was in a thawing state. At the beginning of March, the soil slowly thawed from bottom to top, and at the end of March, the soil thawed rapidly from top to bottom, which started a two-way thawing process until it completely thawed. The soil completely thawed from April to October (Figure 2a). In early November, the soil began to freeze from the top layer down to the bottom layer, and then the depth of the freezing soil layer gradually reached the maximum.

The Tanggula site is located in the permafrost soil region. From January to March, the soil was completely frozen. At the end of April, the active layer (AL) began to thaw from the surface, after which it gradually thawed, and the maximum thawing depth of the AL was greater than 3.0 m in September [31], below which it was permafrost soil. In early October, the AL began freezing upwards from the permafrost table. Approximately 10 additional days later, two-way freezing processes started, with the surface rapidly freezing downwards. In November and December, the soil was completely frozen again (Figure 2b). The daily average soil temperatures at the Tanggula site in August and September and at the Naqu site in July–September both exhibited pronounced declines due to heavy rainfall events during both periods. The large amount of cloud cover in the sky during the rainfall events reduced the R_n, which resulted in a significant decrease in the daily average soil temperature.

Figure 3a shows the variation in the daily mean air temperature (T_a) at 10 m at the Naqu and Tanggula sites in 2008. From April to September, the monthly average T_a at the Naqu site was positive, and it was negative in the other months. From June to September, the monthly average T_a at the Tanggula site was positive, and it was negative in the other months. The annual average T_a values at the Naqu and Tanggula sites were −1.34 and −5.80 °C, respectively. The annual instantaneous maximum T_a values were 20.13 and 14.16 °C, respectively, both of which occurred in July. The annual instantaneous minimum T_a values were −36.32 and −30.32 °C, respectively, which occurred after heavy snowfall in early February.

Figure 3b shows the variation in the daily average wind speed (ws) at the Naqu site at 1 m and at the Tanggula site at 2 m in 2008. The statistical results revealed that the monthly average ws of the two sites was less than 5.0 m/s, except in January and April. The annual average ws values at the Naqu and Tanggula sites were 4.73 and 4.69 m/s, respectively, and the annual maximum instantaneous ws values were 36.83 and 21.18 m/s, respectively, which occurred in January and February, respectively.

The snow at the two sites was concentrated from January to March and from October to December (Figure 4). The statistical results revealed that the numbers of days with snow at the Naqu and Tanggula sites in 2008 were 114 and 135 d, respectively; the maximum number of days with snow occurred in November; and the daily maximum snow depth at the Naqu site was lower than that at the Tanggula site.

A weighing rain gauge was used at both the Naqu and Tanggula sites (T200B, Geonor Inc., Akershus, Norway), and the rainfall was counted as the total rainfall (including snowfall). As shown in Figure 5, the rainfall at the two sites was concentrated during the TPM from May to October. The annual rainfall amount at the Naqu site was 590.50 mm, and the rainfall amount from May to October was 568.10 mm, which accounted for 96.21% of the annual rainfall amount. The annual rainfall amount at the Tanggula site was 405.27 mm, and the rainfall amount from May to October was 350.77 mm, which was 86.55% of the annual rainfall amount. This occurred because, on average, the Tibetan Plateau summer monsoon occurred in May, and the whole plateau was wet, “warm”, and rainy. The Tibetan Plateau winter monsoon was initiated in October on average, and the whole plateau was dry and cold and had less precipitation [35,36]. Using the relative coefficient of rainfall [37] combined with the rainfall characteristics of the two sites, it was determined that the rainy season at the Naqu site in 2008 started on 12 May and ended on 30 September. During the rainy season, the total rainfall amount was 512.00 mm, which was 86.71% of the total annual rainfall amount. At the Tanggula site, the rainy season started on 19 May and ended on 25 September. During the rainy season, the total rainfall amount was 286.27 mm, which was 70.64% of the total annual rainfall amount. The rainy season at the Naqu site started earlier and ended later than that at the Tanggula site. The rainfall at the Naqu site was greater than that at the Tanggula site.

4.2. Variations in ET_a

In 2008, the variations in the ET_a at the Naqu and Tanggula sites were similar, and both were concentrated from May to October (Figure 5). The annual total ET_a values at the Naqu and Tanggula sites were 493.53 and 585.17 mm, respectively, which accounted for 83.58% and 144.39% of the annual total rainfall, respectively, indicating that the capacity for evaporation at these two sites was very high. There were other sources of water for the ET_a at Tanggula in addition to the rainfall caused by the TPM. Some studies have also reported that the annual total ET_a is greater than the annual total rainfall at the Tuotuo River and on the northeastern TP [38,39].

ET_a/rainfall is an important parameter for describing the water balance, which reflects the efficiency of the ecosystem utilisation of precipitation. The annual ET_a/rainfall at the Naqu site was 0.84, which is similar to the value of alpine meadows (0.88) [40]. The annual ET_a/rainfall at the Tanggula site was 1.44, which is close to the value of the degraded grassland in Inner Mongolia (1.4) [41]. This finding indicates that the grassland at the Tanggula site degraded, which reduced its ability to conserve sources of water.

The monthly total ET_a at the Tanggula site was greater from January to September, and that at the Naqu site was greater from October to December (see

Table 1. Monthly total amounts (mm/month) of ET_a and ET₀ and monthly average Kc of the Naqu and Tanggula sites in 2008.

	Naqu			Tanggula
	ETa	ET0	Kc	ETa	ET0	Kc
January	1.89	52.11	0.04	11.68	33.62	0.35
February	3.90	38.51	0.10	23.01	27.11	0.85
March	3.65	61.12	0.06	17.26	47.64	0.36
April	2.67	94.92	0.03	21.24	66.06	0.32
May	41.16	95.97	0.43	62.11	77.02	0.81
June	92.48	103.05	0.90	93.35	92.59	1.01
July	98.08	100.59	0.98	106.55	95.95	1.11
August	95.73	97.33	0.98	107.71	90.19	1.19
September	68.09	78.96	0.86	78.25	65.44	1.20
October	38.94	59.76	0.65	36.63	36.69	1.00
November	16.28	9.60	1.70	8.79	20.44	0.43
December	30.66	28.03	1.09	18.60	20.76	0.90

) because both sites experienced deep snow from October to December. During this period, the T_a values at the Tanggula site were very low, all less than 0 °C, which resulted in minimal evaporation and sublimation from the snow surface. Moreover, the T_a at the Naqu site reached 7 °C in the daytime, which resulted in a high rate of evaporation, and sublimation from the snow surface led to a high level of ET_a in Naqu. The maximum monthly total ET_a at the Naqu site occurred in July, and that at the Tanggula site occurred in August. At this time, the R_n and T_a increased, which was also the time when the rainfall was the most concentrated. At the same time, the surface vegetation was in the peak growing season, transpiration was strong, and ET_a reached a maximum. The minimum monthly total ET_a of the two sites occurred in January and November. At this time, the R_n and T_a decreased, the amount of rainfall was very small, the surface vegetation dried out, and the ET_a was very small.

4.3. Variations in ET₀

The ET₀ is the hypothetical evapotranspiration rate of the reference crop canopy [33]. The new definition of ET₀ avoids many problems with the old definition, including connecting the ET₀ with the reference grass, but the new definition maintains continuity with the old definition. The new parameter does not need special regional calibration or wind functions. The ET₀ can be calculated via general meteorological data. This new parameter has high practical application value and accuracy. Recently, it has been widely applied worldwide [42,43,44,45,46,47,48].

As shown in Figure 6, the ET₀ values at the two sites were small in spring but much greater than the ET_a values. Subsequently, the ET₀ gradually increased, reached a maximum in summer, gradually decreased in autumn, and reached its lowest value in winter. When there was snow cover at both sites, ET₀ became very small, and there were even negative values at the Naqu site, which was due to the low R_n and T_a values when there was snow cover.

In 2008, the annual total ET₀ values at the Naqu and Tanggula sites were 819.95 mm and 673.15 mm, respectively. The monthly total ET₀ at the Naqu site was greater than that at the Tanggula site (see Table 1), which was related to the method of calculating the ET₀, and the equation was established from the theories of energy balance and aerodynamics. The equation uses e, R_n, ws, and the air-drying force under certain temperature conditions to determine ET₀ [33]. e, R_n, and ws at the Naqu site were greater than those at the Tanggula site.

4.4. Variations in Kc

It is very important to determine the Kc value of an area. Some locations have only general meteorological data and no EC data. Therefore, we first calculated the ET₀ via the FAO56 Penman–Monteith method and then calculated the ET_a via the Kc values in the same or similar areas.

The daily Kc values at the Naqu and Tanggula sites varied between −3 and 3 in 2008 (Figure 7). The Kc values at the Tanggula site were greater than those at the Naqu site. In the period without snow cover, the Kc values for the two sites were between 0 and 2, which were greater in the rainy season and lower in the dry season. During the rainy season, the average Kc value at the Naqu site was 0.90, and that for the Tanggula site was 1.14. In the dry season, the average Kc value at the Naqu site was 0.18, and that at the Tanggula site was 0.37. In the period with snow cover, the changes in the Kc value at the Naqu and Tanggula sites had no obvious trend, and the average Kc values were 0.51 and 0.68, respectively. The annual average Kc values at the two sites were 0.60 and 0.87, respectively.

5. Discussion

The daily mean T_a of the two sites exhibited obvious seasonal variations with a parabolic shape, and it was affected mainly by solar radiation. The monthly average, monthly maximum, monthly minimum, and monthly range T_a values at the Naqu site were greater than those at the Tanggula site. The altitude and type of underlying surface of the frozen soil were the main factors influencing the T_a at the two sites. The higher altitude and permafrost underlying the surface at the Tanggula site resulted in a lower T_a. With increasing altitude, the annual average T_a decreased, and the variation in the average daily T_a range became more obvious [49]. In February and November to December, the T_a sharply decreased, which was due to the heavy snowfall at the two sites during this period. Snow on a surface led to an increase in the surface albedo, resulting in a decrease in R_n and a sharply decreased T_a [50]. In winter, the warming effect of snow has little effect on the seasonal thawing depth but has an obvious effect on reducing the seasonal freezing depth [51]. Therefore, the impacts of snow at the Naqu and Tanggula sites were different. From November to December, the T_a at the Tanggula site was lower than 0 °C, the snow-thawing speed was slow, and the change in the snow depth curve was also gradual. However, the T_a at the Naqu site in the daytime reached 7 °C, the snow-thawing speed was fast, and the slope of the snow depth curve changed steeply.

The instantaneous maximum ws at the Naqu site each month was greater than that at the Tanggula site. The variations in the ws at the two sites were similar. The ws values at the Naqu site showed a pronounced diurnal variation in winter, and at the Tanggula site, there was a pronounced variation in winter and spring. Specifically, the ws gradually increased in the morning; reached its maximum in the afternoon; and even exceeded 20 m/s because solar radiation warmed the surface during the day, which increased the atmospheric turbulence activity near the surface. Then, the ws gradually decreased and maintained a low ws at night. The diurnal variation in the ws in the other seasons was relatively stable.

The variations in the ET_a and ET₀ at the two sites were similar; both were small in winter–spring and large in summer–autumn. The ET_a values at the Naqu and Tanggula sites were concentrated from May to October. From January to April, the ET_a values at the two sites were very small and even had small negative values because there was little rainfall; in addition, the T_a was low, the snow was not as thick as that from November to December, and condensation occurred. During this period, the ET_a at the Tanggula site was greater than that at the Naqu site because of the deeper snow cover, the longer duration of snowmelt, the number of snow covers, and the snow depth at the Tanggula site were greater than those at the Naqu site. When the seasonally frozen soil at the Naqu site began to thaw at the end of March and the AL at the Tanggula site began to thaw at the end of April, the λE at these two sites increased, and the ET_a increased accordingly. However, because the rainy season did not occur at either site and the surface and atmosphere were relatively dry, the change in ET_a was not as drastic as that in the rainy season. After the rainy season, the ET_a at both sites increased rapidly and was strongly affected by rainfall. After each increase in rainfall, the ET_a peaked. During this period, the ET_a at the Naqu site was 392.87 mm, which was 79.60% of the total annual ET_a. The ET_a at the Tanggula site was 408.00 mm, which was 69.72% of the total annual ET_a. After the conclusion of the rainy season, the ET_a at both sites decreased rapidly. Therefore, the ET_a at both sites was obviously affected by the rainfall caused by the TPM.

However, in terms of the ET_a at the Tanggula site, in addition to the influence of the TPM, the freezing–thawing processes of the AL also had a great impact. The AL at the Tanggula site began to thaw at the end of April. The surface soil of the AL thawed earlier than the bottom soil, which was still completely frozen. Therefore, the surface soil water content (smc) gradually increased, but the range in which the bottom soil changed was small. The AL then gradually thawed downwards. With increasing thawing depth in the AL, the water storage capacity gradually increased. In September, the thawing depth of the AL reached a maximum, where it exceeded 3 m. During the thawing process of the AL, the infiltrating water was blocked by the permafrost and accumulated in the shallow soil. The smc of each layer in the AL was basically saturated; thus, the soil had sufficient water for evaporation. The water content of the AL during the thawing period was one component of the water source of the ET_a. Therefore, the ET_a at the Naqu site was affected mainly by the TPM, whereas the ET_a at the Tanggula site was strongly affected by both the TPM and the freezing–thawing processes of the permafrost.

Meteorological conditions and surface conditions both affect the ET_a and ET₀. Meteorological conditions include the radiation, T_a, air humidity, rainfall, ws, etc. The surface conditions include the surface temperature (T₀), smc, vegetation status, and snow cover [33]. Here, the rainfall and vegetation conditions were not considered, and the snow weather was excluded. Only the relationships between the daily data of the ET_a and the R_n, T_a, ws, water vapour pressure (e), T₀, and smc at the Naqu and Tanggula sites that lacked snow were analysed. To avoid the influence of multicollinearity between the various factors, a partial correlation analysis was adopted. The analytical results (see

Table 2. Coefficient of the partial correlation between the ET_a and ET₀ and the meteorological elements at the Naqu and Tanggula sites in 2008.

	Naqu		Tanggula
	ETa	ET0	ETa	ET0
Net radiation (Rn)	0.702 **	0.826 **	0.782 **	0.874 **
Air temperature (Ta)	0.795 **	0.664 **	0.748 **	0.674 **
Wind speed (ws)	−0.518 **	−0.055	−0.314 **	−0.263 **
Water vapour pressure (e)	0.787 **	0.322 **	0.719 **	0.323 **
Surface temperature (T0)	0.812 **	0.654 **	0.770 **	0.705 **
Soil water content (smc)	0.764**	0.341 **	0.791 **	0.490 **

Note: ** indicates an extremely significant correlation (p < 0.01).

) revealed that the ET_a and ET₀ both exhibited significant positive correlations with the other meteorological elements, except the ws. The correlation between the ET_a and the T₀ at the Naqu site was the most significant, whereas the correlation between the ET_a and the smc at the Tanggula site was the most significant. The correlation between the ET₀ and the R_n at both sites was the most significant, which was related to the method of calculating the ET₀. This finding is related to the conditions of the underlying surfaces of the two sites. Compared with the Tanggula site, the Naqu site is located in the seasonally frozen soil region, with a greater T₀ and lush vegetation. The increase in the T₀ directly affected the soil evaporation and plant transpiration. However, the Tanggula site is located in the permafrost soil region, with a lower T₀ and sparse vegetation. The increase in the T₀ was partly used for soil melting and heat dissipation, so the impact of the increase in the T₀ on the ET_a was not as sensitive as that at the Naqu site. This also resulted in a stronger relationship between the ET_a and the smc at the Tanggula site.

There are two ways in which the ws influences evaporation. On the one hand, due to wind energy, turbulent diffusion strengthens, so water vapour diffusion on the evaporation surface strengthens and enhances evaporation. On the other hand, the wind can strengthen advection. If this advection is cold and wet, it inhibits evaporation [52]. The annual average T_a values at the Naqu and Tanggula sites were very low at −1.34 and −5.80 °C, respectively. The cold and wet advection caused by the glacial winds on the plateau was very strong; thus, ws was negatively correlated with the ET_a.

The ET₀ was relative to a certain reference plane and could not represent the ET_a. Owing to different vegetation or crop types and reference plane conditions, the ET_a and ET₀ greatly differed. The comparison between the ET_a and ET₀ at the two sites is shown in Figure 8. The ET₀ values of the two sites were approximately 25% and 9% higher than the ET_a values, respectively. This occurred because for the Naqu site located in the seasonally frozen soil region, compared with the Tanggula site, the ET₀ itself was greater because of the greater T_a and R_n, and the ET_a was significantly lower than the ET₀ due to water limitations. Similar results were found in the seasonally frozen soil region in Northeast China, especially in dry years [53]. However, the Tanggula site located in the permafrost soil region, the ET₀ itself was relatively small because of the lower T_a. At the same time, the surface soil thawed in summer, which formed an AL with a relatively sufficient water supply. The vegetation had more available water. Therefore, the ET_a was close to the ET₀. Other researchers reported that in the Arctic and northeastern TP, especially in wetland and tundra regions, the ET_a was also close to the ET₀ [54,55].

The Kc value can be used to estimate the ET_a. The monthly average Kc values were greater at the Tanggula site from January to October and greater at the Naqu site in November and December (see

Table 3. The monthly average, monthly maximum, and monthly minimum Kc values at the Naqu and Tanggula sites in 2008.

	Naqu			Tanggula
	Monthly Average	Monthly Maximum	Monthly Minimum	Monthly Average	Monthly Maximum	Monthly Minimum
January	0.04	0.87	−0.06	0.35	1.99	−0.09
February	0.10	1.13	−0.75	0.85	2.63	0.20
March	0.06	1.06	−0.20	0.36	1.62	0.09
April	0.03	0.47	−0.24	0.32	1.33	0.06
May	0.43	0.98	−0.15	0.81	1.63	0.11
June	0.90	1.17	0.54	1.01	2.00	0.52
July	0.98	1.23	0.46	1.11	1.65	0.74
August	0.98	1.20	0.73	1.19	1.84	0.76
September	0.86	1.15	0.53	1.20	1.55	0.56
October	0.65	1.08	0.02	1.00	1.70	−0.32
November	1.70	3.11	−2.94	0.43	1.20	−1.82
December	1.09	2.76	0.11	0.90	1.30	0.47

). Under the given radiation conditions, the ET_a and ET₀ were equal when the water supply was sufficient, which defined it as a humid environment [56]. During the rainy season, the monthly average Kc value at the Tanggula site from June to September was slightly larger than 1.0, and that at the Naqu site in July and August was also close to 1.0. The Kc value was greater than 1.0; that is, the ET_a was greater than the ET₀, so it should be in a humid environment. When the Kc value was greater than 1.0, it indicates that there was almost always precipitation at both sites or continuous precipitation had just occurred. That is, there was sufficient water vapour in the air for evaporation.

6. Conclusions

In this work, the ET_a, ET₀, and Kc were calculated by using EC data and meteorological gradient data from the Naqu and Tanggula sites in 2008. The variations, differences, and factors influencing the ET_a and ET₀ were analysed. In conclusion, an in-depth and detailed analysis of the changes in and differences in evapotranspiration was performed with observation data from different frozen soil regions on the TP. This work not only deepened the understanding of the process of the water–heat coupling in different frozen soil regions but also provided high-precision and accurate calibration and validation data for satellite observations and model simulations.

In our next work, we will use remote sensing models and satellite remote sensing data to analyse the changes in evapotranspiration at different spatial and temporal scales. In the future work, we will downscale the driving data to the resolution of the Moderate Resolution Imaging Spectroradiometer (MODIS) or Landsat data and extract physical quantities, such as the surface temperature, from the MODIS or Landsat data. This will provide the driving data required for the remote sensing model through remote sensing data for pattern calculation and analysis.