The Influence of Horizontal Thermal Advection on Near-Surface Energy Budget Closure over the Zoige Alpine Wetland, China

Abstract

Near-surface energy budget closure has been a trending topic in land surface processes research, especially on the underlying surfaces of heterogeneous wetlands. In this investigation, the horizontal thermal advection caused by thermal inhomogeneity over the alpine wetland is calculated based on the eddy covariance data observed at the Flower Lake observation field and WRF modelling data over the Zoige alpine wetland, China. The contribution of horizontal thermal advection to the near-surface energy closure is analysed. The results show that the mean horizontal heat advection of the Zoige wetland is 20.2 W·m⁻², and the maximum value reached 55.0 W·m⁻² in the summer of 2017. After introducing thermal advection into the near-surface energy balance equation, the near-surface energy closure ratio increased from 72.3% to 81.0%.

1. Introduction

As the main underlying surface type of the Earth’s atmosphere, land directly affects atmospheric circulation and regional climate change through the exchange of momentum, heat, and mass between the land and atmosphere [1]. In terms of energy exchange, the surface energy balance (SEB) is an essential cornerstone of any theoretical description of the Earth’s climate system [2]. A large number of land surface process observation experiments have been conducted in the past half-decade [3,4,5,6,7], and 10.0–30.0% non-closed energy was found between the near-surface turbulent flux and the effective energy, which has been heavily covered in the past 25 years. To further explore this uncertainty, a concerted field experiment was carried out in the United States in 2000 (EBEX-2000) [4]. The experiment was conducted on irrigated farmland with a uniform underlying surface. It was found that the proportion of non-closed energy accounted for 10.0% of the effective energy. Some researchers speculate this is caused by secondary circulations on the landscape scale beyond the actual flux footprint [8,9]. However, those experiments were based on many assumptions. When the SEB is stripped of its assumptions, terms such as advection arise, which are rarely studied experimentally [4,10,11,12,13].

The knowledge of near-surface energy balance has notably accumulated in recent years. It is generally regarded that the causes of non-closure can be divided into the following three categories: data error, thermal storage [14], and the impact of a nonideal situation [15,16]. The ideal situation refers to the observation of the Eddy Covariance System that needs to satisfy the following three conditions [17]: ① Temporal stationarity: the storage items from the surface to the observation height can be ignored during the observation period; ② Regional homogeneity: there is no horizontal advection flux and horizontal turbulent flux; ③ The vertical average velocity is zero: no vertical advection flux. However, in situ observations, most of the atmospheric underlying surfaces cannot fully satisfy the assumption of spatial homogeneity. Therefore, the energy transport caused by advection cannot be ignored in actual observations [18]. Kochendorfer proposed a conceptual model to address horizontal heat advection in the land surface energy budget balance [19]. The study found that energy budget closure near the leading edge improved by more than 20% with the inclusion of the horizontal and vertical advection of latent and sensible heat. Based on this model, Harder studied sensible and latent heat advection on the surface of patchy snow and found that horizontal heat advection on sensible and latent heat had the same magnitude, accounting for 31.0% and 33.0% of the net snowmelt energy, respectively [20]. Morrison quantified the magnitude of horizontal thermal advection under a nonuniform underlying surface in a dry lake bed desert environment. It was found that the thermal advection can reach approximately 100.0 W·m⁻² [13]. The influence of horizontal thermal advection on the energy exchange between land and atmospheres is fully affirmed by the above experiments. Besides, the effects of large eddies on flux transport are important factors accounting for the non-closure problem, especially in heterogeneous surfaces [21,22,23].

However, there are few observations on horizontal energy advection over alpine wetlands. The Zoige alpine wetland is located in the source area of the Yellow River in the northeastern Qinghai-Tibetan Plateau. It is the most important water conservation area in the upper reaches of the Yellow River. Due to the location between monsoon and non-monsoon regions, it is highly sensitive to climate change. Affected by global warming, the area of wetlands has decreased by more than 598.6 km² in recent 40 years [24]. In recent years, a few observation experiments have been conducted in wetlands. Lu investigated the characteristics of water-heat exchange between wetlands and the atmosphere. The experiment found an energy non-closure for more than 30% [25]. The Zoige wetland consists of meadows and water bodies. There are obvious differences in specific heat capacity, albedo, vegetation, and moisture between meadows and water bodies. The heterogeneity may cause considerable energy advection [26,27]. This horizontal energy advection is inherently not captured by single-tower observation [28]. Thus, it is of scientific significance to evaluate the impact of horizontal thermal advection on the interpretation of the energy budget closure over the Zoige wetland.

This study focuses on the near-surface energy budget closure over a heterogeneous lake–land junction of the Zoige alpine wetland. The influence of horizontal energy transfer caused by underlying surface heterogeneity on the energy budget closure is quantified to improve the understanding of the energy budget closure of the Alpine wetland.

2. Materials and Methods

2.1. Observation Site

The observational experiment was conducted at the Zoige Plateau Wetland Ecosystem Research Station, Chinese Academy of Sciences (hereinafter referred to as “the Flower-lake Observation Field” (FOF)). The FOF is located north of the Zoige alpine wetland (33.9°N, 102.8°E), which is east of the Qinghai-Tibet Plateau, as shown in Figure 1.

The FOF belongs to the humid monsoon climate zone in the alpine plateau, and its average elevation is approximately 3500.0 m. The annual mean temperature was 2.2 °C in 2017, and the annual precipitation was 570.9 mm. The FOF is situated at a typical transition zone from the alpine meadow to the shallow water body, and the highest temperature gradient at the lake–land boundary could reach 20.0 °C·km⁻¹. There are plains and mountains in the northern area with a range of 10.0 km from the observation site. The elevation difference between the plain and the mountainous is 1000.0 m approximately. The water body is 0.0–8.0 km to the south of the FOF and usually enters the frozen period around December and thaws around April of the next year.

2.2. Instrument and Data

The main observation equipment deployed in this study is the eddy covariance micrometeorological system (Figure 1e). It is equipped with an ultrasonic anemometer (CSAT3, Campbell Scientific, Inc., Logan, UT, USA) to measure the instantaneous orthogonal wind speed; an open path CO₂/H₂O gas analyser (li-7500RS, Li-Cor Biosciences, Lincoln, NE, USA) to measure the water vapour and CO₂ concentrations. Additional instruments include a CR3000 datalogger; four-component net radiation sensors (CNR4, Kipp&Zonen B.V., Delft, The Netherlands); self-calibrating soil heat flux plates (HFP01SC); soil moisture sensors (CS616, Campbell Scientific, Inc., USA); 109 temperature sensors (Campbell Scientific, Inc., USA) and SI-111 surface infrared temperature sensor (Campbell Scientific, Inc., USA). The ultrasonic anemometer and the gas analyser were mounted at a height of 2.4 m above the land surface, the radiometer was collected at a height of 1.6 m, and the soil heat flux plate was buried at a depth of 10.0 cm below the land surface.

The data used in this study were collected at the Flower-lake Observation Field (FOF) from 11 July 2017, to 26 July 2017, and the preliminary QA/QC data were implemented. The data can be divided into three parts: eddy covariance data, radiation observation data, and soil hydrothermal observation data. Eddy covariance data include latent heat flux (LE), sensible heat flux (H), horizontal wind speed, and air temperature at 30 min intervals. The radiation observation data are net radiation data (Rn) at 30 min intervals. Soil water and heat observation data include soil heat flux, soil temperature, soil humidity, and water temperature. Since the soil heat flux sensor is buried at a depth of 10.0 cm under the land surface, the soil heat flux is corrected to the surface by calculating the soil heat storage [29].

For the temperature gradient data, the Weather Research and Forecasting model (WRF, version 4.3) was employed because it has sufficient capability to provide temperature gradient data with high temporal and spatial resolution. Community Land Model (CLM, Version 4) is coupled with WRF to provide surface physical processes. The triple-nested downscaling method was applied to simulate local-scale surface temperature variation. A domain with a 0.5-km horizontal resolution is nested inside parent domains. For the model’s initial and boundary conditions, the NCEP FNL (Final) Operational Global Analysis data (6-h) on a 0.25° grid was used. The comparison between simulation results and observation data is shown in Figure 2a,b. The correlation coefficient between simulated temperature data and observed data is 0.84. It should be noted that Beijing Standard Time (BST) is used in this study, while the local time of the Zoige wetland is one hour later than BST.

2.3. Energy Closure Ratio and Components

With considering horizontal thermal advection, the energy imbalance or closure ratio in Zoige wetland is expressed as:

$$Res=\left(R_n-G_0-H_A-L{E}_A\right)-\left(H+LE\right)$$

(1)

$$CR=\frac{{{\displaystyle \sum }}_{t1}^{t_2}\left(H+LE\right)}{{{\displaystyle \sum }}_{t_1}^{t_2}\left(R_n-G_0-H_A-L{E}_A\right)}$$

(2)

where Res (W·m⁻²) is the energy balance residual on near-surface energy exchange; CR is near-surface energy closure in the observation period; $R_n$ (W·m⁻²) is the net radiation; $G_0$ (W·m⁻²) is the soil heat flux corrected to the land surface; $H$ (W·m⁻²) is the sensible heat flux; and $LE$ (W·m⁻²) is the latent heat flux. $H_A$ (W·m⁻²) is sensible heat advection; $L{E}_A$ (W·m⁻²) is latent heat advection. $t_1$ and $t_2$ is the start and ending times of observational experiment, respectively. Ordinary least squares (OLS) were also used to assess surface energy closure.

Due to the soil heat flux plate being buried under 10 cm of soil, the heat stored in the upper soil should be considered into soil heat flux (G), thus [29]:

$$G_0=G_{10}+{{\displaystyle \int }}_0^{0.1}\frac{\partial {\rho }_s{c}_{s}T\left(z\right)}{\partial t}dz$$

(3)

where $G_0$ (W·m⁻²) is the soil heat flux corrected to the land surface; ${\rho }_s$ (kg·m³) is the density of soil; $c_s$ (J·(kg·°C)⁻¹) is the specific heat capacity; $T\left(z\right)$ is soil temperature profile in the depth of 10 cm and 0 cm. In the actual calculation, $\frac{\partial T\left(z\right)}{\partial t}$ is differentiated as $\frac{{\overline{T}}_i-{\overline{T}}_{i-1}}{\Delta t}$, where ${\overline{T}}_i$ is soil $T$ in $i$ time step; ${\overline{T}}_{i-1}$ is soil 30 min before ${\overline{T}}_i$; $\Delta t$ = 30 min.

Soil volume specific heat capacity in Equation (3) is calculated as follows:

$${\rho }_s{C}_s={\rho }_{dry}{C}_{dry}\left(1-{\theta }_{sat}\right)+{\rho }_w{C}_w{\theta }_w$$

(4)

where ${\rho }_s{C}_s$ is volume specific heat capacity of wet soil (J·m⁻³·K⁻¹); ${\rho }_{dry}{C}_{dry}\left(1-{\theta }_{sat}\right)$ is volume specific heat capacity of dry soil (J·m⁻³·K⁻¹), which is 1.27 × 10⁶ J·m⁻³·K⁻¹ (Tanaka et al., 2001); ${\rho }_w{C}_w$ is the volume specific heat capacity of liquid water (J·m⁻³·K⁻¹); ${\theta }_w$ is volumetric water content.

Plant photosynthesis [30] and air heat storage [31] were also calculated. However, it is found that these two terms are less than 5.0 W·m⁻². Considering the effect of instruments error, terms less than detection limit (5 W·m⁻²) were ignored. Thus, the effect of plant photosynthesis and air heat storage would not be considered in this research.

2.4. Near-Surface Horizontal Thermal Advection

To quantify the horizontal advection, the scalar net source/sink rate can be described by the scalar equation of the two-dimensional Reynolds budget [32]:

$$\overline{S}=\frac{\partial \overline{c}}{\partial \overline{t}}+\overline{u}\frac{\partial \overline{c}}{\partial \overline{x}}+\overline{w}\frac{\partial \overline{c}}{\partial \overline{z}}+\frac{\partial }{\partial \overline{x}}\overline{u^{\prime }{c}^{\prime }}+\frac{\partial }{\partial \overline{z}}\overline{w^{\prime }{c}^{\prime }}$$

(5)

where $\overline{S}$ (scalar unit s⁻¹) is scalar net source/sink rate; $\overline{c}$ is the average density of a scalar variable within half an hour; t is time; $u$ is horizontal wind speed; $w$ is vertical wind speed; $u^{\prime }$ is the pulsation of horizontal wind speed; and $w^{\prime }$ is the pulsation of vertical wind speed. For simplicity, the volume could be restricted to a two-dimensional box along the x and z axes, and x axe represents the downwind direction. The first term of formula ($\frac{\partial \overline{c}}{\partial \overline{t}}$) is the scalar storage, e.g., when $c$ is air temperature $T$, this term indicates heat storage ($\frac{\partial \overline{T}}{\partial \overline{t}})$; the second term $\left(\overline{u}\frac{\partial \overline{c}}{\partial \overline{x}}\right)$ is the average horizontal advection; the third term $\left(\overline{w}\frac{\partial \overline{c}}{\partial \overline{z}}\right)$ is the average vertical advection; the fourth term $\frac{\partial }{\partial \overline{x}}\overline{u^{\prime }{c}^{\prime }}$ is horizontal turbulent flux; the fifth term is $\frac{\partial }{\partial \overline{z}}\overline{w^{\prime }{c}^{\prime }}$ vertical turbulent flux. When $c$ is air temperature $T$, this term indicates sensible heat flux $\frac{\partial }{\partial \overline{z}}\overline{w^{\prime }{T}^{\prime }}$.

The second and fourth terms of the above equation reflect the nature of the horizontal heat exchange in the heterogeneous underlying surface [33]. This horizontal heat transportation could be conducted by both turbulence and advection. However, in actual observations, the horizontal turbulent transfer $\frac{\partial }{\partial \overline{x}}\overline{u^{\prime }{c}^{\prime }}$ is neglected due to its small magnitude; therefore, integrating Equation 5 with respect to the profile depth ($z_m$) and substituting sensible heat (${\rho }_a{C}_{p}T$) into scalar $c$, the conservation equation in heat flux form can be further derived [16,18]:

$$F_{z_m}=\underset{F_{st}}{\underbrace{{\rho }_a{C}_p{{\displaystyle \int }}_0^{z_m}\frac{\partial \overline{T}}{\partial t}dz}}+\underset{F_{HA}}{\underbrace{{\rho }_a{C}_p{{\displaystyle \int }}_0^{z_m}\overline{u}\frac{\partial \overline{T}}{\partial x}dz}}+\underset{F_{VA}}{\underbrace{{\rho }_a{C}_p{{\displaystyle \int }}_0^{z_m}\overline{w}\frac{\partial \overline{T}}{\partial z}dz}}+\underset{H}{\underbrace{{\rho }_a{C}_p\overline{w^{\prime }{T}^{\prime }}{|}_{z_m}}}$$

(6)

where $H$ (W·m⁻²) is turbulent sensible heat flux; $F_{VA}$ (W·m⁻²) is vertical heat advection; $F_{HA}$(W·m⁻²) is horizontal heat advection; $F_{st}$(W·m⁻²) is air heat storage; $F_{z_m}$ (W·m⁻²) is heat net source/sink flux; ${\rho }_a$ (kg·m⁻³) is the air density; $C_p$(J·(kg·°C)⁻¹) is the heat capacity of air at constant pressure; $\overline{T}$ is the average temperature within half an hour; $T^{\prime }$ is the pulsation of temperature; $z_m$ (m) is the profile depth of interest. In addition, soil heat storage and other terms of heat storage should also be considered in $F_{st}$. Thus, the horizontal advection of sensible heat and latent heat (substituting latent heat ${\rho }_a{L}_{v}q$ into scalar $c$) can be expressed as follows [18,19]:

$$H_A=F_{HA}={\rho }_a{C}_p{{\displaystyle \int }}_0^{z_m}\left(\overline{u}\frac{\partial \overline{T}}{\partial x}+\overline{v}\frac{\partial \overline{T}}{\partial y}\right)dz$$

(7)

and:

$$L{E}_A={\rho }_a{L}_v{{\displaystyle \int }}_0^{z_m}\left(\overline{u}\frac{\partial \overline{q}}{\partial x}+\overline{v}\frac{\partial \overline{q}}{\partial y}\right)dz$$

(8)

where $H_A$ (W·m⁻²) is horizontal sensible heat advection; $L{E}_A$ is horizontal latent heat advection. $L_v$ (2.835 × 10⁶ J·kg⁻¹) is the latent heat of vaporization. $q$ (kg·kg⁻¹) is specific humidity. $\overline{u}$ (m·s⁻¹) is the average downwind speed and $\overline{v}$ (m·s⁻¹) is the average crosswind speed. $\frac{\partial \overline{T}}{\partial x}$, $\frac{\partial \overline{T}}{\partial y}$ (°C·km⁻¹) and $\frac{\partial \overline{q}}{\partial x}$, $\frac{\partial \overline{q}}{\partial y}$ (kg·(km·kg)⁻¹) is horizontal temperature and specific humidity gradient at $z_m$; It was calculated from surface temperature and specific humidity field from WRF and observation data. Data on wind speed, air density, and specific heat capacity were collected from FOF. Equations (7) and (8) show that horizontal thermal advection is determined by surface wind and temperature gradient and specific humidity. A positive value (cold/dry advection) indicates a decrease in the energy budget within a cubic air unit, while a negative value (warm/wet advection) means an increase.

2.5. Flux Footprint and the Configuration of Surface Temperature

Footprint analysis was adopted to determine the source area of the energy fluxes measured with the eddy covariance system. There are many factors that influence the spatial extent and size of source area including surface wind speed and direction, atmospheric stability, surface roughness length and measurement hight [34]. A 2-dimensional footprint model [35] was used to estimate the footprint for each time step of flux measurements around the EC tower. The footprint is a transfer function f between 0 and 1:

$$F_c\left(0,0,z_m\right)=\underset{S}{{\displaystyle \int }}f\left(x,y\right) Q_u\left(x,y\right)dxdy$$

(9)

where $f\left(x,y\right)$ is the flux footprint; $F_c$ is a flux density at the observation site; $Q_u\left(x,y\right)$ is a source or sink of passive scalars at the surface and S denotes the integration domain. $f\left(x,y\right)$ could be comprehended as the weight of the influence on the observation point from the flux budget at point (x, y). Figure 3 depicts the configuration of surface temperature and footprint during the observation period. Surface temperature is derived from landsat satellite data with a resolution of 30 m. Footprint result is calculated from daytime flux data from 11 July to 26 July. The result shows that more than 80% of the flux source was obtained from an upwind distance of 100 m over a temperature heterogeneous surface in the daytime of summer 2017.

In landscape scale, the temperature difference between the northwest and southeast of the observation site is obvious. The northwest is a high-temperature area with a maximum temperature of 25.1 °C, and the southeast is a low-temperature area with a minimum temperature of 12.4 °C. Combined with the wind speed frequency data, the dominant wind direction of the FOF is a southeast wind, and the half-hour average friction speed is between 0.0 and 1.0 m·s⁻¹. This indicates a continuous flow of cooler air from the lake surface to the observation site and thus results in horizontal heat outflow. In addition, the underlying surface of the observation field is grassland-wetland-lake from north to south. The characteristics of energy exchange on lake surface are obviously different from those on grassland [34,36,37,38,39,40]. Therefore, under the support of surface wind, the different energy exchange patterns of the lake surface would affect the wetland observation site thus resulting in obvious CR underestimates.

2.6. Advection: Mean Velocity, Temperature, and Moisture Gradient

The calculation of Equations (6) and (7) require the support of surface wind, temperature, and moisture profile both in time series and space gradient. Those profiles are present in Figure 4 with the expectation to build a stronger understanding of the input variables. As the definition, when a wide range of temperatures (moisture) and surface wind is observed, advection may provide a strong source of heat transport. Characteristics of wind speed, temperature, and moisture are necessary for the interpretation of H_A and LE_A.

Figure 4a describes variations of wind velocity from 11 July to 25 July; Figure 4a* present a 15-day average velocity in each hour of one day. The surface wind has an obvious diurnal variation with a range of 0.4 m·s⁻¹ to 9.2 m·s⁻¹. From night to afternoon, mean wind velocity is moderate at 1.5 m·s⁻¹. However, from 16:00 to sunset, mean velocity has a rapid elevation to 4.5 m·s⁻¹ indicating a strong heat transport appears at sunset under a high-temperature gradient condition. Figure 4b,b* shows variations of temperature and gradient and its mean characteristic. To straight reflect the influence of temperature gradient on horizontal thermal advection, the component of the temperature gradient is calculated in the direction of wind speed (red line). Surface temperature gradient (Effective Grad T) has the same intraday variation characteristic with temperature (T), increasing at sunrise, reaching maximum at 15:00, and decreasing rapidly at sunset. The maximum mean temperature is 18.9 °C and the max temperature gradient reaches 1.8 °C·hm⁻¹ in the afternoon. Compared with surface wind, surface temperature has a moderate diurnal variation, assigning a stable signal to sensible heat advection. Figure 4c,c* depicts variations of specific humidity and is the mean characteristic. Different from the correlation of T and grad T, diurnal variation of specific humidity has the opposite relationship between humidity gradients. Specific humidity is stable in the daytime at 8–10 g·kg⁻¹, and decreases at night. As humidity rises in the morning, the humidity gradient increases. However, the humidity gradient is 1.0% range of humidity, while the temperature gradient is 10.0% range of temperature. Thus, the heat transferred by temperature would be stronger than moisture.

3. Results

3.1. Diurnal Variation of Horizontal Thermal Advection in Zoige Wetland

Horizontal thermal advection (H_A + LE_A) has regular diurnal variations because of the difference in the heating effect of solar radiation on the inhomogeneous underlying surface. However, unstable surface wind under different weather conditions results in unstable diurnal variations both in H_A and LE_A. To determine the basic features of H_A and LE_A, 15 consecutive typical sunny days are selected to reflect the magnitude and frequency characteristics over the Zoige wetland. Time series of H_A and LE_A are present here (Figure 5). The sign convention used corresponds with the definition in Section 2.3, where positive values represent local cooling, and negative values represent local heating.

Figure 5a,b describes the time series variation of H_A and LE_A. The horizontal advection showed two patterns during 11–19 July and 20–25 July, with different diurnal variation stability. During 11–19 July, both H_A and LE_A have unstable diurnal variations with dispersed maximum centers. This is caused by unstable surface wind. Figure 6 presents surface wind direction and velocity in those two periods. Though both of the two periods have the same dominant wind direction, surface wind in the first period (a) has a more dispersed distribution than the second period (b). Thus, period I represents the pattern in an unstable surface wind condition, and period II represent a stable southeast wind condition. In the condition of stable southeast wind, advection shows continuous diurnal variations, with a maximum of H_A (178.1 W·m⁻²) in the evening and a maximum of LE_A (39.1 W·m⁻²) in the morning. Thus, both surface wind and temperature (moisture) gradient dominates the variation of H_A and LE_A but surface wind determines the frequency of advection, and temperature (moisture) gradient determines the magnitude.

Figure 7 shows mean diurnal variations of H_A and LE_A from 11 July to 25 July. All data are statistically averaged through the corresponding time of each day to obtain the characteristics of intraday variation. Figure 7 indicates both H_A and LE_A have unimodal diurnal variation characteristics. H_A increases continuously after sunrise, reaching the maximum value of average flux (52.2 W·m⁻²) at 17:30. Superimposing the southwest winds brought by the valley winds, the wetland has obvious lake-land winds in the later afternoon. This leads to a large and stable H_A. Under strong wind conditions, the maximum value of horizontal heat advection can reach 178.1 W·m⁻². After sunset, H_A decreases rapidly with the descent of the horizontal temperature gradient. The ensemble average of H_A is 17.8 W·m⁻² which means there is cold advection affecting the observation site overall. Different from the characteristic of H_A, LE_A reaches a maximum in the morning with an average value of 8.0 W·m⁻² and ensemble average of LE_A is 2.4 W·m⁻² during the observation period. LE_A is much smaller than H_A (17.8 W·m⁻²). This is caused by the small magnitude of the moisture gradient. FOF located over the wetland’s underlying surface which has sufficient vapor support and a high average temperature. The higher the temperature, the stronger the air’s capacity to dissolve water vapor. Thus, specific humidity over the wetland surface is probably higher than that over the lake surface and results in a small moisture gradient in the daytime. As a result, the energy transfer of water vapor advection is not as strong as that of temperature advection.

3.2. Surface Energy Budget of Zoige Wetland in Summer

To reflect the contribution of thermal advection (H_A and LE_A) to the surface energy budget over the Zoige wetland, characteristics of energy exchange in summer are presented in this section.

Figure 8 depicted the diurnal ensemble average variation of components referred to in Section 2.6. Zoige wetland is located at 30° of mid-latitudes with an altitude of 3700 m. The maximum net radiation (R_n) of average diurnal variation here is 754.0 W·m⁻² with a daytime length of 12.5 h. Water body accounts for more than 30.0% of landcover of the observation field. Sufficient support of water vapor determines latent heat flux (LE) as the main form of energy exchange between land and atmosphere in wetlands. The maximum mean latent heat flux is 333.3 W·m⁻² and the aggregate latent heat exchange accounts for 57.5% of net radiation. Besides, high water coverage causes huge soil heat storage and soil heat flux. Revised soil heat flux (G₀, Equation (3)) has a maximum of 225.1 W·m⁻² accounting for 30.0% of max R_n in the daytime. However, the underlying surface would transfer heat to the atmosphere continuously at night. The average diurnal net energy exchange of G₀ is 9.1 W·m⁻² accounting for 4.5% of R_n. Sensible heat flux is not as strong as that over grassland or desert underlying surface since most of the surface energy is transferred by LE to the atmosphere over wetlands. The maximum of mean sensible heat flux is 72.2 W·m⁻² and the aggregate latent heat exchange accounts for 11.6% of net radiation. As mentioned in Sect. 1, there is an obvious energy non-closure in Zoige wetland due to the heterogeneous of the underlying surface. The maximum of mean energy balance residual (Res, Equation (1)) is 152.1 W·m⁻² and the daily average of Res is 32.1 W·m⁻² accounting for 16.2% of the daily average R_n. The non-closure ratio would be 26.3% if neglect the effect of horizontal thermal advection. The remaining energy residual may be related to vertical advection, large eddies, and local circulation over wetlands. It is also related to the mismatching of instrument footprint.

Horizontal thermal advection (H_A + LE_A) has the same magnitude with sensible heat flux but the phase is different over the Zoige wetland. The daily average value of thermal advection is 20.2 W·m⁻², while that of H is 23.0 W·m⁻² which is 10.2% of R_n. The diurnal variation of advection and H is unimodal. The peak of thermal advection lags behind that of H. Specifically, thermal advection is usually lower than H during the daytime but higher than H after 16:30. Thermal advection at night is not obvious. Latent heat flux dominates the surface energy budget of wetlands. The daily average value of latent heat flux is 114.3 W·m⁻², which is five times the magnitude of thermal advection. The maximum of mean thermal advection is 55.0 W·m⁻² (appears at 17:30) and that of LE is 332.6 W·m⁻² (appears at 14:00). Overall, thermal advection has the same magnitude with sensible heat flux and accounts for 10.2% of net radiation.

3.3. The Contribution of Horizontal Thermal Advection to Energy Budget Closure

To quantify the influence on the near-surface energy closure, horizontal thermal advection (H_A + LE_A) is introduced into the energy budget for analysis. Figure 9 depicts the near-surface energy closure ratio of FOF with and without considering thermal advection. The ordinate represents the turbulent flux (H + LE) and thermal advection (H_A + LE_A), and the abscissa is the surface available energy (SAE = R_n − G₀ − H_A − LE_A). The red solid line is the ordinary least squares (OLS) fitting line, and the slope represents the energy closure ratio. Different dot colours correspond to different dot densities.

Considering the impact of horizontal thermal advection, the energy closure ratio derived from OLS increases from 64.0% to 72.0% and the closure ratio (CR, Equation (2)) rises from 72.3% to 81.0%. The contribution of thermal advection to the near-surface energy closure is 8.7%. It means that approximately one-tenth of net radiation energy is transferred horizontally by horizontal thermal advection instead of participating in local land–air energy exchange. Combined with the wind direction and temperature field data, it can be found that cold near-surface advection occurs at this time. Therefore, it is speculated that the surface energy closure is lower when cold near-surface advection is arriving, and the influence is approximately one-tenth of surface available energy in the Zoige wetland. Overall, neglecting horizontal cold advection results in a significant underestimation of the closure ratio under heterogeneous underlying surfaces.

4. Conclusions

With the consideration of soil heat storage, horizontal thermal advection was introduced into the near-surface energy balance equation to quantify its contribution to the surface energy budget closure over the Zoige alpine wetland. The results show as follows:

(1)

Horizontal sensible heat advection is stronger than horizontal latent heat advection. The ensemble average of H_A is 17.8 W·m⁻² while that of LE_A is 2.4 W·m⁻² during the observation period.

(2)

The magnitude of the horizontal thermal advection is approximately equal to that of sensible heat flux at Zoige alpine wetland during strong wind conditions. The diurnal variation was unimodal with a mean of 20.2 W·m⁻² and a maximum value of 55.0 W·m⁻².

(3)

After considering the contribution of thermal advection, the energy closure ratio of the Zoige alpine wetland increased from 72.3% to 81.0% in the summer. It means approximately one-tenth of net radiation energy is transferred horizontally by thermal advection in Zoige alpine wetland during a cold advection.

Due to the heterogeneous underlying surface of the Zoige alpine wetland, heat would be transported horizontally by surface wind. Therefore, horizontal heat advection, as an expenditure of available energy budget terms, plays an important role in energy closure over heterogeneous underlying surfaces. Compared with the thermal advection in a dry lake bed desert environment, the thermal advection of wetlands is stronger under strong wind conditions. This is due to the obvious surface heterogeneity on wetland boundaries. If the location of the flux tower is changed deeper over the wetland, the decrease in temperature gradient and wind speed will bring lower sensible heat advection. Besides, the appearance of stable airflow which is dominated by mesoscale circulation also greatly affects the characteristics of thermal advection. Nevertheless, secondary circulations on the landscape scale also have a profound effect on the energy budget over heterogeneous underlying surfaces. It may interact with advection and results in energy transfer potentially. This could also be estimated by large eddy simulations over wetlands in the future.