Temporal and spatial variations of energy balance closure across FLUXNET research sites

Temporal and spatial variations of energy balance closure across FLUXNET research sites

Agricultural and Forest Meteorology 271 (2019) 12–21 Contents lists available at ScienceDirect Agricultural and Forest Meteorology journal homepage:...

2MB Sizes 0 Downloads 32 Views

Agricultural and Forest Meteorology 271 (2019) 12–21

Contents lists available at ScienceDirect

Agricultural and Forest Meteorology journal homepage: www.elsevier.com/locate/agrformet

Temporal and spatial variations of energy balance closure across FLUXNET research sites Wenhui Cui, Ting Fong May Chui

T



Department of Civil Engineering, Room 6-18A, Haking Wong Building, The University of Hong Kong, Pokfulam, Hong Kong

A R T I C LE I N FO

A B S T R A C T

Keywords: Eddy covariance technique Energy balance closure Carbon dioxide fluxes FLUXNET

There is always a discrepancy between available energy and output energy in the surface energy budget of FLUXNET research sites. Using the daily data retrieved from the FLUXNET database, the energy balance closure (EBC) of around 150 sites covering nine land covers and five Köppen climate zones [i.e. tropical area (A), dry area (B), mild temperate area (C), snow area (D) and polar area (E)] was analyzed. The temporal and spatial variations of EBC in different land covers and climate zones were summarized, and the relationships between EBC and environmental variables were explored. The possible differences in the EBCs of sites with open path (OP) and closed path (CP) gas analyzers were also examined at different precipitation levels in various climate zones. The results showed that EBC was positively related to air temperature (Ta) and vapor pressure deficit (VPD) below certain thresholds. Better EBCs were observed in land covers with stable evaporative fraction (i.e. the ratio of latent heat flux to the sum of latent and sensible heat fluxes). For land covers with seasonal varying evaporative fraction, the larger evaporative fraction in summer corresponded with better EBCs. OP systems resulted in better EBCs at various precipitation levels, and EBCs decreased with increasing precipitation for both OP and CP systems. In addition, the relationship between EBC and CO2 fluxes was different among the different land covers. There was a positive relationship for most land covers but not for savannah, shrub land, and evergreen broadleaf forest. The relationships between EBC and CO2 as well as EBC and other environmental variables are cross-influenced, which could be related to the stomata aperture and metabolism of the vegetation. Overall, this study summarized patterns of EBCs that could be used to correct eddy covariance data and energy balance closure related models. It further enhanced our understanding of the potential link between EBC and vegetation physiology that could facilitate the modeling and prediction of biophysical processes related to water, energy, and carbon fluxes from the leaf to ecosystem levels.

1. Introduction Eddy covariance (EC) systems are currently one of the most reliable methods for measuring carbon, energy, and water fluxes, and they have become a standard technique in the study of surface–atmosphere boundary layer interactions (Majozi et al., 2017). They facilitate the study of the net exchanges between the land surface and atmosphere and their biological and climatological controls (Aubinet et al., 1999; Baldocchi et al., 2001). EC fluxes are being measured in different ecosystems across the world. FLUXNET is a global network of EC towers that aims to understand the mechanisms controlling the exchanges of CO2, water vapor, and energy across a spectrum of landscapes on temporal and spatial scales. It archives the data from the various sites over the world, and provides information for the modeling of carbon and water cycling across regions and the globe. However, energy



imbalance, which is when the available energy (e.g., net radiation minus ground heat flux) is not equal to the sum of output energy (e.g., sensible and latent heat fluxes), exists across all the FLUXNET sites, with an imbalance ratio of around 10–30% (Wilson et al., 2002; Stoy et al., 2013). The primary reasons behind usually include systematic measurements and sampling errors (e.g. discrepancy between the flux footprint and net radiation sensors), instrument bias (e.g. different types of gas analyzers), the loss of low and high frequency for flux measurements, horizontal advection induced by land surface heterogeneity, unaccounted soil and canopy storage at the short time scale, etc. (Wilson et al., 2002; Foken, 2008a; Leuning et al., 2012) Examining or minimizing the EC imbalance is challenging. For example, to address the influence of advective flux divergence, at least several EC towers with sonic anemometers and infrared gas analyzers at different heights are needed to measure the temperature and humidity

Corresponding author. E-mail address: [email protected] (T.F.M. Chui).

https://doi.org/10.1016/j.agrformet.2019.02.026 Received 28 August 2018; Received in revised form 26 January 2019; Accepted 18 February 2019 0168-1923/ © 2019 Elsevier B.V. All rights reserved.

Agricultural and Forest Meteorology 271 (2019) 12–21

W. Cui and T.F.M. Chui

understanding and application of FLUXNET data product and be beneficial to the modeling and prediction of the biophysical processes related to water, energy, and carbon fluxes from the leaf level to the ecosystem level. In this study, it is acknowledged that energy balance is not achieved for the research sites of FLUXNET. The sites are further assumed to be uniformly managed and perfectly flat, extending over an area much larger than the flux footprint; that is, they are ideally suited to the application of the EC technique.

gradients, which is infeasible in most cases because of the significant resources required (Leuning et al., 2012). Even in an experimental setup with multiple towers to capture the horizontal gradients, 10% of the energy budget was still unaccounted for (Kochendorfer and Paw U, 2011). Other than the horizontal advection in the atmosphere, Cui and Chui (2017) explored the lateral exchange of water and energy in the subsurface by using an array of temperature and water-level sensors. However, the subsurface exchange was found to be insufficient to close the energy budget. In addition to experimental approaches, some studies have also tried to solve the energy imbalance issue theoretically. For example, Cheng et al. (2017) proposed a correction method for the EC data in high frequency to remedy the energy imbalance caused by neglecting the smaller eddies in the initial subrange. Furthermore, given the multiple influential factors that are also site-specific, it is always difficult to compare results across sites and to deduce the potential reasons behind energy imbalance. Only long-term comparative studies under differing climatic conditions and over different ecosystems might give conclusive explanations (Haslwanter et al., 2009). Overall, despite the tremendous efforts of the researchers all over the world, the energy imbalance problem is still not resolved. FLUXNET facilitates the study of energy balance closure (EBC) across various ecosystems and climate regions, as demonstrated by some preceding works. For example, Wilson et al. (2002) found the general lack of energy imbalance of around 20% at FLUXNET sites, and the imbalance was prevalent for all kinds of land covers and climate zones. Foken (2008a) emphasized the significant influence of exchange processes on larger scales over the heterogeneous landscape on the energy imbalance. Leuning et al. (2012) explicitly pointed out the problem of analyzing data in half-hour resolution and suggested that the energy imbalance was smaller in daily average. Regarding to the relationship between EBC and environmental variables, Franssen et al. (2010) reported that EBC was a function of atmospheric stability, friction velocity, thermally induced turbulence, and the time of the day, based on an analysis using over 20 sites. Stoy et al. (2013) analyzed the data from 193 sites in FLUXNET and identified the significant relationship between EBC and precipitation, gross primary production, landscape-level heterogeneity in topography and plant type. The potential relationship between CO2 uptake and EBC has also been explored in some previous studies. For example, Schmid (1997) and Churkina et al. (2005) showed that carbon uptake represents a measure of water/energy availability, which could influence the energy balance. Wilson et al. (2002) suggested that at a given value of photosynthetically active radiation, the CO2 uptake was less when the energy imbalance was greater. Barr et al. (2006) attempted to link the energy imbalance and CO2 flux deficits, and adjust mean annual NEE based on energy closure. However, Wilson et al. (2002) and Foken (2008a) did not consider the difference among various land covers; Franssen et al. (2010) and Stoy et al. (2013) used the data in hourly resolution, which was problematic since the influence of storage change could be significant (Foken, 2008a; Leuning et al., 2012). In addition, the vapor pressure deficit (VPD) is closely related to evapotranspiration and could therefore influence the variation of latent heat flux. However, this potential influence was not considered in the above studies during the examination of the relationship between energy imbalance and environmental variables. Moreover, when looking into the potential link between surface energy imbalance and CO2 uptake, the analysis was either based on single vegetation type, or lumping different biomes together. Furthermore, the idea of using EBC to adjust CO2 flux measurements was still controversial (Baldocchi, 2008). Finally, although a body of work exists on the energy imbalance problem, to the best of our knowledge, no study has summarized EBC temporally or spatially at the ecosystem scale across the FLUXNET research sites. The goals of this study were to summarize the temporal and spatial variations of the EBC over different land covers and to explore the potential link between environmental variables, energy fluxes, energy imbalance, and carbon uptake. The results could extend the

2. Data and method 2.1. Energy balance closure ratio A simplified land surface energy balance equation is written as (1)

Rn − G − S = LE + H −2

−2

where Rn is the net radiation (W m ), G is the soil heat flux (W m ), LE is the latent heat used for evapotranspiration, H is the sensible heat exchanged with the air (W m−2), and S is the sum of the storage change and metabolic terms (W m−2). Rn and G are usually measured by a radiometer and soil heat flux plates (or temperature sensors), respectively. The turbulent flux terms (i.e., LE and) are measured from the fluctuations of wind speed (measured using a sonic anemometer), air temperature (measured using a temperature sensor) and the gas density (measured using a gas analyzer). Although S can influence the energy balance, especially for tall vegetation (Meyers and Hollinger, 2004; Franssen et al., 2010), it was unavailable in the FLUXNET database. However, calculating the energy budget at daily resolution could minimize the influence of S (Leuning et al., 2012). Therefore, S was not considered in this study; a more simplified EBC ratio was calculated using the daily data as follows:

EBC =

LE + H Rn − G

(2)

Perfect closure is achieved when the EBC is one. 2.2. Data The EC data were retrieved from the FLUXNET Database (2015). The data quality was checked and gaps in the data were filled using standard procedures applied by the FLUXNET community (Reichstein et al., 2005; Papale et al., 2006; Baldocchi, 2008). As mentioned in Section 2.1, daily data could minimize the influence of S, and thus the daily data were used in this study. The data covered the nine main land covers defined by the International Geosphere-Biosphere Programme (IGBP) classification, which are evergreen needle leaf forest (ENF), evergreen broadleaf forest (EBF), deciduous broadleaf forest (DBF), mixed forest (MF), shrubland (SH), cropland (CROP), savannah (SA), grassland (GRA), and wetland (WET). Among all the sites in the FLUXNET database, only the sites where the four energy components in Eq. (2) (i.e. Rn, G, H, and LE) were available for at least one year were considered. Moreover, calculated EBC values that were beyond two standard deviations from the mean were removed from the dataset. Additionally, as suggested by Aubinet et al. (1999) and as applied in Stoy et al. (2013), turbulent flux measurements with an apparent lack of EBC (i.e., EBC values larger than 2 or less than -1) were removed from the data sets. Based on the above filtering criteria, data from 156 sites have been analyzed. Moreover, to ensure sufficient turbulent transport for the flux terms, additional filtering procedure that excluding data where the daily friction velocity (u*) exceeded the threshold of each site were adopted in this study (Aubinet et al., 1999; Reichstein et al., 2005). The remaining data was used in the following analysis. Although the standard deviation of vertical wind speed could be more representative of turbulent conditions (Jocher et al., 2017), it was not available in the FLUXNET database. The u* threshold was also commonly adopted in 13

Agricultural and Forest Meteorology 271 (2019) 12–21

W. Cui and T.F.M. Chui

flux measurement studies (Barr et al., 2013; McHugh et al., 2017). The u* threshold values were calculated from the ensemble of u* thresholds derived from the moving point detection method (Barr et al., 2013) and the changing point detection method (Papale et al., 2006).

Table 1 Atmospheric stability classes. (Taken from Foken, 2008b).

2.3. Environmental variables In this study, a few fundamental and representative environmental variables for examining their potential relationships with EBC were selected. Air temperature (Ta) has been suggested to have the largest influence on climate models (Groenendijk et al., 2009) and is closely related with the four basic energy components (i.e., Rn, G, H, and LE). VPD is closely related to the evapotranspiration of vegetation. Friction velocity which is associated with shearing stress, can determine atmospheric stabilities, under which the measurements of EC system could be influenced (Aubinet et al., 1999; Foken, 2008b). Therefore, u*, Ta, and VPD were selected as variables. Many previous studies found that moisture availability was also highly related with EBC. However, for most of the sites in FLUXNET, soil water content measurements were not available. Here the precipitation and evaporative fraction [LE/(LE +H)] were chosen as the surrogates for moisture availability. Given that the performance of open path gas analyzer (OP) and closed path gas analyzer (CP) is different during wet conditions (Liu et al., 2006; Haslwanter et al., 2009), the relationship between moisture availability and EBC was separately examined for sites with OP and CP. Among the 156 sites, gas analyzer information (OP or CP) was available for 105 sites through their publications, and sites with both OP and CP were not considered. It should also be noted that among the sites with CP, two sites were equipped with a new model LI 7200 gas analyzer, which overcame the shortcomings of both OP (i.e., LI 7500) and the old models of CP (i.e., LI 6262 and LI 7000). The data of these two sites were excluded when analyzing the difference between OP and CP. The data were categorized into four daily precipitation levels: no rainfall (0 mm), more than 0 and less than or equal to 20 mm, more than 20 mm and less than or equal to 50 mm, and more than 50 mm. In addition, atmospheric stratification can significantly influence EBC. It can be described with the stability parameter ε , which can be calculated as

ε=

Obukhov length L

Stability parameter ε ≈ 1/L

very unstable unstable neutral to slightly unstable neutral to slightly stable stable very stable

−30 −100 −300 5000 250 60

−0.033 −0.01 −0.0033 0.0002 0.004 0.017

(Köppen, 1884; Rubel and Kottek, 2011). It considers both precipitation and temperature, along with vegetation, and provides a simple framework within which climate variability and vegetation can be described in an integrated way. In this study, the Chen and Chen (2013) classification approach was adopted to categorize the 156 sites into the five main climate zones based on the latitude and longitude of each site. The distribution of 156 sites over nine land covers was shown in Fig. 1. 2.5. Two-dimensional kernel density estimation A two-dimensional (2-D) kernel density estimation was adopted to present the relationships between EBC and atmospheric stabilities. Kernel density estimation is a non-parametric approach for estimating probability density functions from a given dataset. Histograms are the simplest non-parametric density estimator, but they suffer from the arbitrary choice of bin size. Kernel density estimators place a smooth kernel, usually a Gaussian window, which was used in this study, with a given bandwidth around measured data points and combine them to estimate the probability distribution of all observations. A linear diffusion algorithm following Botev (2015) was used to choose the optimal kernel bandwidths and avoid over- or under-smoothing. A kernel density estimation function following Botev (2015) was applied using MATLAB (2014) to produce the density distribution. Kernel density estimation was used in this study to visually display rather than statistically interpret the data. Similar to a probability distribution or density function, the sum of all kernel density values equals one. 2.6. Statistical method

z−d L

(3)

Because of the common occurrence of outliers in EC data, this study calculated the median EBCs and the environmental variables under each category (e.g. land covers and climate zones). The Kruskal–Wallis test was also applied to evaluate whether the medians of multiple groups (e.g., different Köppen climate zones) were significantly different from each other (Kruskal and Wallis, 1952). The significant difference test under a 95% confidence level was computed using MATLAB (2014). In addition, to describe the significance of the trends in spatial variation in the five Köppen climate zones and the relationship between EBC and other variables, an analysis of variance (ANOVA) with a Pvalue was applied.

where z is the measurement height (m), d is the displacement height (m), L is the Monin–Obukhov length (Obukhov, 1946; Monin and Obukhov, 1954; Stoy et al., 2013), which can be derived from

L=

Stability levels

−ρCp u*3 Ta kgH

(4) −3

where ρ is the density of air (kg m ), Cp is the specific heat capacity of air (J K-1), k is von Karman’s constant ( = 0.4, Högström, 1996; Foken, 2008b), g is the gravitational acceleration (m s-2), and Ta is the air temperature in Kelvin. ε < 0, ε ≈ 0, and 0 < ε < 1, respectively, correspond to unstable, neutral, and stable stratifications (Foken, 2008b). More specifically, the atmospheric stability can be classified as shown in Table 1. Because the values of z and d were not recorded in the FLUXNET database and the order of magnitude of (z − d) is normally within 10 m, the value of (z − d) is approximated as 1, and ε is thus equal to 1/L.

3. Results 3.1. Overview of sites and EBC The sites were categorized and Table 2 shows the numbers of the sites in the various land covers (the row labelled as “Biomes”) and main Köppen climate zones (the column labelled as “Köppen”). The median values of EBC in each category is also shown, and the land covers are listed in order from high EBC to low EBC. The median values of EBC in each land cover category are significantly different with each other, based on the Kruskal-Wallis test result, with the p-value smaller than 0.001. The better EBCs were observed in SA, SH, and EBF, with values larger than 0.85. WET, MF, and DBF exhibited relatively poorer EBCs,

2.4. Köppen climate classification To explore the characteristics of EBC in different climate regions across the various ecosystems, a widely used vegetation-based climate classification system was applied in this study. This system was created by Wladimir Köppen and consists of five main climate zones: A: tropical area; B: dry area; C: mild temperate area; D: snow area; E: polar area 14

Agricultural and Forest Meteorology 271 (2019) 12–21

W. Cui and T.F.M. Chui

Fig. 1. Locations and land cover types of study sites (FLUXNET 2015 dataset) on a world map of Köppen Climate Classification system. (Adopted from Chen and Chen, 2013).

for OP, which was consistent with the findings of Haslwanter et al. (2009) and Sánchez et al. (2010). Under wet conditions, the quality of original data of OP can be low and thus excluded; the gap is filled by the FLUXNET community using marginal distribution sampling method (Reichstein et al., 2005). The gap-filled data also showed a better EBC with OP than CP under different precipitation levels. In addition, the difference between OP and CP increased from Köppen A to E for the daily precipitation level of 50–100 mm. For example, the discrepancy between OP and CP for Köppen A was 0.014, whereas it was 0.344 between OP and CP for Köppen D. Moreover, even for CP which is supposed to work during the rainfall (Burba et al., 2012), EBCs decreased with the increase in precipitation level, and the difference between different levels of precipitation was larger than that of OP. Finally, the decreasing trend from Köppen A to E shown in Table 3 agrees with the results discussed in Section 3.2.1.

with values less than 0.75. Also, the EBCs of Köppen A and B showed a small difference, but there was an obvious decreasing trend from Köppen A to E, with a linear regression coefficient of 0.967 (not shown in the table). 3.2. Spatial variation of EBC in different Köppen climate zones 3.2.1. Relationship between EBC and Ta and between EBC and VPD In Fig. 2(a), the median EBC, Ta, and VPD values show general decreasing trends from Köppen A to E, except the median VPD value in Köppen B, which was very high but was consistent with the fact that Köppen B are dry areas. The declining trends of EBC and Ta showed higher linear goodness-of-fit coefficients (R-squared values) of 0.967 and 0.945, respectively, as compared to that of VPD at only 0.684 (the linear lines are not shown in the figure). Moreover, Fig. 2(b) and (c) show the relationships between EBC and Ta and between EBC and VPD, respectively. Fig. 2(b) shows a generally positive relationship between EBC and Ta. Fig. 2(c) shows a positive relationship between EBC and VPD when VPD is small (e.g., less than 10 hPa); however, it shows a slight declining trend at higher VPD values.

3.3. Seasonal variation of EBC and its relationship with Ta, VPD and evaporative fraction In Fig. 3, median EBCs in different months display significant seasonal variations for ENF, CRO, DBF, MF, and WET among the nine land covers, with a P-value less than 0.001. For example, the EBC in December for ENF was as low as 0.33, but it was higher than 0.85 in August. Land covers SA, SH, EBF, and GRA showed relatively stable EBCs, particularly SA, which exhibited good EBCs throughout the year, with values around 0.9.

3.2.2. Relationship between EBC and precipitation Table 3 shows the median EBCs under different precipitation levels in the five Köppen climate zones. Generally, the median EBCs decreased with increasing precipitation for both gas analyzer systems (i.e., OP and CP). When there was no rainfall (i.e., 0 mm), better EBCs were observed Table 2 The number of sites and median value of EBC in each category. Biomes Köppen

SA

SH

EBF

GRA

ENF

CRO

DBF

MF

WET

Total

Median EBC

A B C D E Total Median EBC

4 5 4 0 0 13 0.890

0 4 3 3 2 12 0.877

4 0 8 0 0 12 0.854

1 3 18 6 1 29 0.834

0 1 8 29 1 39 0.812

0 1 12 6 0 19 0.808

1 1 8 5 0 15 0.738

0 1 2 2 0 5 0.720

1 0 4 3 4 12 0.701

11 16 67 54 8 156  

0.908 0.905 0.819 0.783 0.705   0.819 

15

Agricultural and Forest Meteorology 271 (2019) 12–21

W. Cui and T.F.M. Chui

Fig. 2. (a) The variation of medians of EBC, Ta, and VPD in different Köppen climate zones; (b) the overall relationship between EBC and Ta in all five climate zones; and (c) the overall relationship between EBC and VPD in all five climate zones.

the sum of turbulent fluxes in most months of SA and the whole year of SH, with LE/(LE+H) smaller than 0.5. This was opposite with EBF, in which the latent heat accounted for a large percentage throughout the year, with LE/((LE+H) larger than 0.6. These three land covers (i.e., SA, SH and EBF) had relatively stable evaporative fractions throughout the year, and were also the ones with the best EBCs (Table 2). For the other land covers, most of them, except ENF, displayed obvious seasonality, with the largest value in June to August, which was similar with the seasonal variation of EBC. ENF also displayed a seasonal variation, but the high values appeared in July to October. Moreover, the positive relationship between evaporative fraction and EBC was not as obvious as that in Ta and VPD [Fig. 4(f)].

Table 3 Median EBC values under various precipitation levels and in different Köppen climate zones with either OP or CP systems. Köppen climate zone

A B C D E

Gas analyzer

OP CP OP OP CP OP CP OP

6 3 10 28 18 17 18 5

Overall

0

Daily precipitation (mm) 0-20

20-50

50-100

0.909 0.876 0.919 0.879 0.875 0.803 0.815 0.797

0.906 0.771 0.889 0.786 0.674 0.731 0.721 0.623

0.850 0.867 0.837 0.737 0.442 0.584 0.546 0.499

0.807 0.821 0.823 0.739 0.445 0.638 0.294 0.275

No. of sites

0.906 0.829 0.915 0.848 0.805 0.763 0.758 0.711

3.4. Potential relationship between CO2 uptake and EBC Three variables associated with CO2 fluxes [i.e. net ecosystem exchange (NEE), ecosystem respiration (RE), gross primary production (GPP)] were also investigated. In Fig. 5, the median values of daily NEE, RE, GPP for twelve months show obvious seasonal variations in most land covers. For most land covers (e.g. ENF, CRO, DBF, MF, and WET), GPP decreased rapidly starting from April, and tended to had the largest CO2 assimilation in June or July. Similarly, RE and NEE displayed the highest values during June or July. Land covers with obvious growing periods, that is, DBF and CRO showed the largest seasonal variations in GPP, RE, NEE. Taking GPP as an example, the magnitude of variations could be up to 11 and 10 g C m−2 d-1 respectively for DBF and CRO. While, the NEE over SA, SH, EBF and GRA displayed relatively small variations, and the difference in the CO2 uptake of these land covers was within 1 g C m−2 d-1 throughout the year. The seasonal variations of RE and GPP over SA and EBF exhibited opposite trend comparing with other land covers, which had the smallest value in June and July and the largest value in January and December. This could be related with the fact that SA and EBF sites are located close to the equator or in the Southern Hemisphere. It should be noted that the seasonality of CO2 uptake (i.e., NEE) is similar with that of EBC. For example, the seasonal variations of both NEE and EBC over SA, SH, EBF and GRA were not as obvious as other land covers. The magnitude of seasonal fluctuations over CRO and DBF was also larger than other land covers for both NEE and EBC. In addition to the similarity between the seasonal variations of NEE and EBC, Fig. 5(d)–(f) more explicitly display the relationships between EBC and the variables related with CO2 fluxes for the nine land covers. EBCs over most land covers (e.g. CRO, DBF, ENF, MF and WET) increased with the increasing GPP, RE and CO2 uptake (i.e., more negative NEE). Taking CRO as an example, EBC at GPP larger than 3 gCm−2d-1 and RE larger than 4gCm−2d-1 was much better than that at smaller values. Also, there was no obvious trend for land covers with

Fig. 3. Seasonal variations of EBCs in different land covers.

Generally, the median values of Ta and VPD of each month displayed obvious seasonal variations for most land covers, as shown in Fig. 4(a) and (b), respectively, except for SA and EBF. The relatively gentle seasonal variation of SA and EBF might be because most of the sites are located in tropical and subtropical areas. Moreover, in addition to SA and EBF, the Ta and VPD values of SH and GRA were higher than those of other land covers in most months [Fig. 4(a) and (b)]. These four land covers (i.e., SA, EBF, SH, and GRA) corresponded with the four best EBCs shown in Fig. 4. The positive relationships between Ta and EBC and between VPD and EBC are respectively shown in Fig. 4(d) and (e). The evaporative fraction [LE/(LE+H)] showed relatively large variations among the different land covers comparing with that in Ta and VPD, as displayed in Fig. 4(c). Firstly, the sensible heat dominated 16

Agricultural and Forest Meteorology 271 (2019) 12–21

W. Cui and T.F.M. Chui

Fig. 4. Seasonal variations of Ta (a), VPD (b), and evaporative fraction (c) and the relationships between EBC and Ta (d), between EBC and VPD (e) and between EBC and evaporative fraction (f).

little seasonal variations (e.g., SA, SH and EBF).

different atmospheric stability levels, from very unstable to very stable as listed in Table 1. Better and more stable EBCs were observed when the friction velocity was larger and 1/L was smaller (more negative). This suggests that unstable atmospheric stratification resulted in better EBCs. During near-neutral conditions when 1/L approaching zero, there were significant fluctuations in EBC [Fig. 6(b)]. Fig. 6(c) shows EBCs for different atmospheric stability levels, in which EBC approached 0.9

3.5. Relationship between EBC and atmospheric stability Fig. 6(a) and (b) respectively show the relationships between EBC and friction velocity and 1/L, based on the kernel density distribution of daily data. Fig. 6(c) displays the variation of median EBCs under

Fig. 5. Seasonal variations of NEE (a), RE (b), and GPP (c) over different land covers and the relationship between EBC and GPP (d), EBC and RE (e), EBC and NEE (f). 17

Agricultural and Forest Meteorology 271 (2019) 12–21

W. Cui and T.F.M. Chui

Fig. 6. (a) 2-D kernel density distribution between EBC and friction velocity (u*); (b) 2-D kernel density distribution between EBC and 1/L; and (c) EBC under different atmospheric stability levels.

Advection and local circulation induced by surface heterogeneity or elevation differences are regarded as the main reasons behind energy imbalance (e.g. Foken, 2008a; Stoy et al., 2013). In this study, the field sites were assumed to be uniformly managed and flat which should be acceptable for most of the sites. Overall, the worst EBCs were observed in the mixed-forest (i.e., MF) and wetland sites (i.e. WET), which reflected the influence of surface heterogeneity on EBC as mixed-forest and wetland are the most heterogeneous land covers. The EBC of mixed forest which is interspersed mixtures or mosaics of the other forest types (e.g. ENF, EBF, DBF) was also worse than that of each single forest type. All the four types of forest (e.g. ENF, EBF, DBF and MF) are dominated by woody vegetation with height exceeding two meters. However, mixed-forest (MF) consisting of different types of trees showed the worst EBC, which indicated the potential influence of vegetation composition on EBC that should be further examined in future studies.

during the unstable condition and sharply declines in the neutral cases. 4. Discussion 4.1. Relevance of common reasons behind energy imbalance to this study The primary reasons behind energy imbalance usually include systematic errors, instrument bias, low and high pass filtering of turbulent fluxes, advection due to surface heterogeneity, and some unaccounted energy (e.g. canopy heat storage) (Twine et al., 2000; Wilson et al., 2002; Foken, 2008a; Stoy et al., 2013). One example of systematic errors is associated with the mismatch between the flux footprint and the source area of sensors measuring net radiation, ground heat flux, etc. (Schmid, 1997). Instrument bias includes the different effects of OP and CP on the turbulent fluxes and the bias in instrumentation measuring net radiation and ground heat flux. High frequency losses can be caused by inadequate sensor frequency response, the separating distance between anemometer and gas analyzer, tube attenuation and time lags of CP, etc. (Leuning and Judd, 1996). Low-frequency losses can be induced by using insufficient time interval to define mean fluxes and thus excluding some relevant low frequencies (Foken et al., 2012). This study analyzed the medians of large data sets, and thus minimized the problems associated with systematic errors and some instrument bias (e.g. the bias related to net radiometers and ground heat flux plates). However, correction algorithms that account for low and high frequency loss are not uniform across the FLUXNET sites and there is a large discrepancy among different correction methods (Aubinet et al., 2000; Wilson et al., 2002). Therefore, it is beyond the scope of this study to discuss the influence of low and high frequency loss on EBC. Regarding the mixed performance of CP and OP, there are no generic explanations that can be applied across different sites. On the one hand, CP seems to provide an alternative for wet conditions under which there is a considerable flux data loss with OP. However, on the other hand, CP requires significantly different corrections for high frequency filtering, time delays, etc. This study compared EBCs of OP and CP under conditions with and without rainfall, for different land covers and climate zones. When there was no rainfall, better EBCs were observed in the sites with OP which demonstrated the better performance of OP. When there was rainfall, better EBCs were also observed in OP than CP at different precipitation levels. It should, however, be noted that OP systematically removed and gapfilled the data during rainfall events. The generally better EBCs observed in OP in this study is similar to the conclusion of Wu et al. (2015) which indicated that the underestimation of gas concentration and turbulent fluxes for CP could lead to more serious energy imbalance. Moreover, the worse EBCs in CP during heavier rainfall implied that the data collected by CP also needs to be carefully checked and corrected during the wet conditions.

4.2. The influence of environmental variables on EBC The influence of Ta and VPD on EBC appears to be positive. However, the increasing trend of EBC turns to be moderate beyond certain values of Ta and VPD. The relationship could be related with two simultaneous but opposing effects. On the one hand, VPD has been widely recognized to drive evaporation and water transport (Kucera, 1954; Zhang et al., 2017), and larger VPDs result in more evapotranspiration and thus higher latent heat flux. Ta is also generally positively related with latent heat flux. On the other hand, high temperature and large VPD can also reduce stomatal conductance and inhibit stomata aperture. At low levels of temperature and VPD, the positive influence might dominate. Therefore, as Ta and VPD increases, the larger latent heat flux is generated which might lead to a larger EBC. However, as Ta or VPD further increases, the constraints on the stomata become stronger and stomata might close to minimize water release. The closure of stomata could limit the increase of latent heat flux, and thus EBC would only increase gently and even become steady. It should be noted that, Ta could simultaneously affect net radiation, ground heat flux, latent and sensible heat fluxes, either directly or indirectly (Cho et al., 2012). The latent heat flux might stop increasing at high temperatures due to the closure of stomata, but the sensible heat flux could still continue to increase. Therefore, the limiting effect in Ta (Fig. 4(d)) is not as obvious as in VPD (Fig. 4(e)). Given the removal of data collected by OP during rainfall and the subsequent gap filling performed by the FLUXNET community, the difference in EBCs among various precipitation levels was small for sites in Köppen A, B and C (Table 3). There was a large difference in cold areas (e.g. Köppen D and E) which could be induced by other reasons (e.g. inaccuracy of instrument in the snow and polar areas). However, a 18

Agricultural and Forest Meteorology 271 (2019) 12–21

W. Cui and T.F.M. Chui

WET, but not for SA, SH and EBF. It was also found that the relationships between EBC and Ta, VPD, evaporative fraction and CO2 fluxes were similar and the influences of environmental variables on EBC could be crossed. EBC displayed positive relationships with Ta and VPD (Figs. 2 and 4) for the various land covers. However, the large seasonal variations of Ta and VPD for SH and GRA did not induce significant seasonal variations of EBC (Fig. 4). Similar situation occurred to the relationship between EBC and evaporative fraction over GRA, which had a very large variation in evaporative fraction but stable EBC. Meantime, the NEEs of GRA and SH were relatively small [Fig. 5(c)] and the seasonal variations were relatively insignificant. As discussed in Section 4.2 regarding the effect of Ta, VPD and moisture availability on EBC, the interactions of the above relationships might be associated with vegetation physiology. The water and energy budgets of the land surface can be affected by vegetation, which controls the passage of water vapor and CO2 through the stomata (Reichstein et al., 2013; Rodrigues et al., 2014; Zhao et al., 2016). First, the biochemical energy storage, which is the net result of photosynthesis, autotrophic respiration, and heterotrophic respiration, could affect the energy balance (Gu et al., 2007). However, it is usually small comparing with other energy components. Gu et al. (2007) pointed out that it was only larger than 10% of the net radiation less than 10% of the measured time. Oke (1988) found that it was about 3.2 Wm−2 per gm−2 h-1 of CO2 assimilation and the largest values of biochemical heat storage were typically within 6 to 16 Wm−2 for crops. The crops in Fig. 5(d) showed that a better EBC was observed for larger CO2 assimilation, which indicated that the biochemical energy storage could be neglected and would not be the main reason behind the potential link between CO2 flux and EBC. Second, transpiration associated with water release is the inevitable byproduct of photosynthesis because stomata are open during photosynthesis (Oke, 1988; Holding and Streich, 2002). Meantime, canopy transpiration is a dominant component of evapotranspiration across a variety of ecosystems (Williams et al., 2004; Eamus et al., 2013). Intense CO2 assimilation would therefore enhance transpiration and the associated latent heat flux. For the land covers (e.g. CRO, DBF, WET in Fig. 5) where EBC increased with the evaporative fraction, the enhanced latent heat flux could lead to a larger EBC. The biomass of GRA sites is usually smaller than other sites, and the metabolism is not as intense. Therefore, the response to the changes of biophysical factors (e.g. Ta, VPD, evaporative fraction) and the EBC variation are not as obvious as those in other land covers. However, this study does not recommend adjusting the CO2 fluxes in proportion to the energy deficit, as advocated by some other researchers such as Twine et al. (2000) and Barr et al. (2006). One reason is that the relationships between EBC and CO2 fluxes related variables are not consistent among different biomes and the trends displayed in Fig. 5(d)–(f) over some land covers are not linear. Furthermore, some reasons behind a bad EBC do not necessarily imply an underestimation of CO2 fluxes. For example, although instrument errors were ignored in this study due to the statistical analysis of large datasets, they should be considered during the analysis of a specific site. The measurements of net radiation and ground heat flux also are not accurate in the snow and polar areas (Oke, 1988; Stiegler et al., 2016), which could cause bad EBC but may not influence the measurement of CO2 fluxes. Moreover, the attenuation of water vapor for CP system tube could significantly influence the EBC of CP system. Therefore, for CP system, it is not appropriate to correct CO2 flux using EBC.

significant difference in EBCs among those of CP was displayed, which indicated that rainfall could affect the performance of CP. It is therefore suggested that data from CP during rainfall should also be carefully checked and corrected. Among the different land covers, the best EBCs were observed over SA, SH and EBF (Table 2), all of which had the most stable evaporative fractions, and had either large latent heat flux or sensible heat flux. The SA sites in this study are located in tropical areas and most of them are covered with short vegetation (e.g. grassland), while the SH sites had little green vegetation according to the satellite images from the FLUXNET site (FLUXNET Database, 2015). The SA and SH sites therefore had large sensible heat flux. On the other hand, EBF sites are located in both tropical and subtropical areas, and they had large biomass of evergreen leaves, which could lead to large latent heat flux. For the other land covers with significantly varying evaporative fractions, EBC increased with the increase in evaporative fraction (i.e., the increase in the percentage of latent heat flux), which indicated the dominant influence of latent heat on EBC. Poorer EBCs were observed for stable and neutral atmospheric conditions (Fig. 6) because the mechanically induced turbulence and friction velocity were reduced. This result was different from the conclusion in Stoy et al. (2013), which stated that EBC during neutral condition was better than that during other conditions. However, neutral stability occurs when wind shear, rather than buoyancy, drives vertical mixing. It can be seen as a condition in which vertical mixing is driven by exogenous forcing rather than buoyancy originating from surfacing heating (Wolf et al., 2008). Under such stability stratification, the errors in determining the integral turbulence characteristics can be very high (Foken, 2008b), and the percentage loss of energy and gas fluxes can also be particularly high (Wolf and Laca, 2007). Moreover, for the stable stratification, the turbulence is dampened strongly by the negative buoyancy (Foken, 2008b), and the advective transport during stable conditions could contribute to the energy imbalance (Leuning et al., 2012). In addition, the stable atmospheric condition is often influenced by internal gravity waves (Zilitinkevich and Calanca, 2000) under which the perturbations of the vertical wind velocity are highly connected with changes of scalar quantities. Thus, the EC flux measurements under this circumstance usually give unrealistic values of the turbulent fluxes (Foken and Wichura, 1996; Foken, 2008b). Therefore, the better EBC values observed during the unstable condition in this study are reasonable, which also agreed with the results of Barr et al. (2006) and Anderson and Wang (2014). 4.3. Potential link between CO2 uptake and EBC Among the reasons behind energy imbalance, instrument bias associated with the separating distance between anemometer and gas analyzer, advection, and the high and low frequency loss could also influence CO2 measurements (Wilson et al., 2002). Although the effects of instrument bias and systematic errors can generally be neglected in this study due to analyzing the medians of large datasets, the influence of advection on EBC and CO2 flux can be relevant. As discussed previously, advection could be the reason behind the serious energy imbalance observed in WET and MF. However, whether CO2 flux is underestimated when there is a serious energy imbalance cannot be concluded for WET and MF, since there were no comparisons of the CO2 flux of these sites. This study has narrowed down the common reasons behind both energy deficit and CO2 underestimation (e.g. excluding the reasons associated with random errors). However, it is still unclear whether advection causes similar impacts on both EBC and CO2 fluxes. To further examine this question, future studies can consider comparing the measured CO2 fluxes and estimated ones from other approaches (e.g. remote sensing method) for WET and MF sites. The relationships between CO2 uptake and EBC were different among the various land covers. For example, there was a positive relationship between CO2 uptake and EBC for CRO, DBF, ENF, MF and

4.4. Possible applications of the summarized EBC from this study The energy balance budget equation is used in different modeling and measuring methods to determine the sensible or latent heat flux from other available energy components. For example, the two-source surface energy balance model is often used to estimate evapotranspiration. It is based on a remote sensing method and it forces a closure amongst the energy budget components (Kustas and Norman, 19

Agricultural and Forest Meteorology 271 (2019) 12–21

W. Cui and T.F.M. Chui

1999; Anderson et al., 2008). EC flux measurements are usually used to validate the estimation by the energy balance model results, but the turbulent fluxes are always underestimated (e.g. the energy deficit displayed across the FLUXNET sites). The summarized spatial and temporal variations of EBC can help to identify the portion of discrepancies between modeling results and EC observations that is due to the underestimation of EC system and the portion that is caused by modeling errors. They can further facilitate the correction and improvement of energy balance closure forced models. Moreover, EBC is a metric used to assess the data quality of EC measurements (Twine et al., 2000; Wilson et al., 2002). It is also applied as a corrector factor in a data product developed by the FLUXNET community for modelers. The turbulent fluxes (i.e. LE and H) are corrected by multiplying the original or gapfilled LE and H by an EBC correction factor. However, the data product is still under testing and development (FLUXNET Processing Pipeline, 2017). Therefore, the summarized results of this study could be used to assess the correction factor for specific vegetation types and climate conditions. Finally, one main challenge is to directly link the observations at plant level to ecosystem level to improve the understanding of the biotic interactions with environmental factors and climate (Reichstein et al., 2013). In this study, the temporal and spatial variations of EBC, as well as their related environmental factors, were summarized under various climate zones and different land covers. The relationships between EBC and CO2 as well as EBC and other environmental variables are cross-influenced, which could be related to the stomata aperture and metabolism of the vegetation. The sensible and latent heat fluxes determine the characteristic leaf temperature, which in turn influences vapor pressure gradient and relative humidity at the leaf surface. These environmental variables govern stomatal resistance, transpiration and the kinetics of CO2 uptake at the leaf level, but they all can exert an influence on the EBC at the ecosystem level. The results of this study therefore could be helpful to the development of large-scale land surface models and the validation of the plant modules of plot-scale models. More specifically, models of evapotranspiration and CO2 flux tended to consider different physical processes and use modeling approaches (e.g. energy balance vs. biogeochemical cycling) (Leuning, 1995; Baldocchi and Wilson, 2001; Anderson et al., 2008). However, water vapour and CO2 fluxes at the leaf surface are jointly controlled by stomata. These fluxes can be well-correlated spatially and temporally, and should be simulated using coupled-models at the landscape scale in the future studies.

temperature, which in turn influences vapor pressure gradient and relative humidity at the leaf surface. These environmental variables govern stomatal resistance, transpiration and the kinetics of CO2 uptake at the leaf level, but they all can exert an influence on the EBC at the ecosystem level. Overall, this study summarized patterns of EBCs that could be used to correct eddy covariance data and energy balance closure related models. It further enhanced our understanding of the potential link between EBC and vegetation physiology that could facilitate the modeling and prediction of biophysical processes related to water, energy, and carbon fluxes from the leaf to ecosystem levels. Acknowledgments We thank all of the researchers for their hard work behind each flux tower, as well as the organizer of the FLUXNET database. This work was supported by the General Research Fund of the Research Grants Council, Hong Kong (Project code: 17202114). References Anderson, R.G., Wang, D., 2014. Energy budget closure observed in paired Eddy Covariance towers with increased and continuous daily turbulence. Agric. Forest Meteorol. 184, 204–209. Anderson, M.C., Norman, J.M., Kustas, W.P., Houborg, R., Starks, P.J., Agam, N., 2008. A thermal-based remote sensing technique for routine mapping of land-surface carbon, water and energy fluxes from field to regional scales. Remote Sens. Environ. 112 (12), 4227–4241. Aubinet, M., Grelle, A., Ibrom, A., Rannik, Ü., Moncrieff, J., Foken, T., Kowalski, A.S., Martin, P.H., Berbigier, P., Bernhofer, C., 1999. Estimates of the annual net carbon and water exchange of forests: the EUROFLUX methodology. Adv. Ecol. Res. 30, 113–175. Aubinet, M., Grelle, A., Ibrom, A., Rannik, Ü., Moncrieff, J., Foken, T., Kowalski, A.S., Martin, P.H., Berbigier, P., Bernhofer, Ch., Clement, R., Elbers, J., Granier, A., Grünwald, T., Morgenstern, K., Pilegaard, K., Rebmann, C., Snijders, W., Valentini, R., Vesala, T., 2000. Estimates of the annual net carbon and water exchange of European forests: the EUROFLUX methodology. Adv. Ecol. Res. 30, 114–175. Baldocchi, D., 2008. ‘Breathing’ of the terrestrial biosphere: lessons learned from a global network of carbon dioxide flux measurement systems. Aust. J. Bot. 56 (1). Baldocchi, D.D., Wilson, K.B., 2001. Modeling CO2 and water vapor exchange of a temperate broadleaved forest across hourly to decadal time scales. Ecol. Modell. 142, 155–184. Baldocchi, D., Falge, E., Gu, L., Olson, R., Hollinger, D., Running, S., Anthoni, P., Bernhofer, C., Davis, K., Evans, R., 2001. FLUXNET: a new tool to study the temporal and spatial variability of ecosystem-scale carbon dioxide, water vapor, and energy flux densities. Bull. Am. Meteorol. Soc. 82, 2415–2434. Barr, A.G., Morgenstern, K., Black, T.A., McCaughey, J.H., Nesic, Z., 2006. Surface energy balance closure by the eddy-covariance method above three boreal forest stands and implications for the measurement of the CO2 flux. Agric. Forest Meteorol. 140, 322–337. Barr, A.G., Richardson, A.D., Hollinger, D.Y., Papale, D., Arain, M.A., Black, T.A., et al., 2013. Use of change-point detection for friction-threshold evaluation in eddy-covariance studies. Agric. For. Meteorol. 171–172, 31–45. Botev Z., 2015. Retrived from https://www.mathworks.com/matlabcentral/ fileexchange/17204-kernel-density-estimation. Burba, G., Schmidt, A., Scott, R.L., Nakai, T., Kathilankal, J., Fratini, G., Hanson, C., Law, B., McDermitt, D.K., Eckles, R., Furtaw, M., Velgersdyk, M., 2012. Calculating CO2 and H2O eddy covariance fluxes from an enclosed gas analyzer using an instantaneous mixing ratio. Glob. Change Biol. 18, 385–399. Chen, D., Chen, H.W., 2013. Using the Köppen classification to quantify climate variation and change: an example for 1901–2010. Environ. Dev. 6, 69–79. Cheng, Y., Sayde, C., Li, Q., Basara, J., Selker, J., Tanner, E., Gentine, P., 2017. Failure of Taylor’s hypothesis in the atmospheric surface layer and its correction for eddycovariance measurements. Geophys. Res. Lett. 44, 4287–4295. Cho, J., Oki, T., Yeh, P.-F., Kim, W., Kanae, S., Otsuki, K., 2012. On the relationship between the Bowen ratio and the near-surface air temperature. Theor. Appl. Climatol. 108 (1–2), 135–145. Churkina, G., Schimel, D., Braswell, B.H., Xiao, X., 2005. Spatial analysis of growing season length control over net ecosystem exchange. Glob. Chang. Biol. 11, 1777–1787. Cui, W., Chui, T.F.M., 2017. Subsurface lateral heat flux within the heterogeneous surface of a subtropical wetland and its potential contribution to energy imbalance. J. Hydrometeorol 18, 3125–3144. Eamus, D., Cleverly, J., Boulain, N., Grant, N., Faux, R., Villalobos-Vega, R., 2013. Carbon and water fluxes in an arid-zone acacia savanna woodland: an analyses of seasonal patterns and responses to rainfall events. Agric. For. Meteorol. 182–183, 225–238. FLUXNET Database, 2015. Retrieved from http://fluxnet.fluxdata.org/data/fluxnet2015dataset/. FLUXNET 2015 Processing Pipeline, May 12, 2017. Retrieved from http://fluxnet. fluxdata.org/wp-content/uploads/sites/3/fluxnet2015_docs_processing_20170512. pdf.

5. Conclusion The temporal and spatial variations of EBC of around 150 FLUXNET sites were summarized using their median values. EBC displayed obvious seasonal patterns for sites of ENF, CRO, DBF, MF and WET, and it generally decreased from Köppen climate zone A to E. Generally, EBC increased with increasing Ta and VPD, but only below certain thresholds. OP systems resulted in better EBCs at various precipitation levels, and EBCs decreased with increasing precipitation for both OP and CP systems. Better EBCs were observed in land covers with stable evaporative fraction [i.e., LE/(LE+H)]. For land covers with seasonal varying evaporative fraction, the larger evaporative fraction in summer corresponded with better EBCs. EBCs during unstable atmospheric stratification was better than that under other conditions, as the errors in determining the integral turbulence characteristics and the advective transport are relatively less comparing with neutral and stable conditions. The relationship between EBC and CO2 fluxes was different among the different land covers. There was a positive relationship for most land covers but not for savannah, shrub land, and evergreen broadleaf forest. The relationships between EBC and CO2 as well as EBC and other environmental variables are cross-influenced, which could be related to the stomata aperture and metabolism of the vegetation. The sensible and latent heat fluxes determine the characteristic leaf 20

Agricultural and Forest Meteorology 271 (2019) 12–21

W. Cui and T.F.M. Chui

McHugh, I.D., Beringer, J., Cunningham, S.C., Baker, P.J., Cavagnaro, T.R., Mac Nally, R., Thompson, R.M., 2017. Interactions between nocturnal turbulent flux, storage and advection at an “ideal” eucalypt woodland site. Biogeosciences 14, 3027–3050. Meyers, T.P., Hollinger, S.E., 2004. An assessment of storage terms in the surface energy balance of maize and soybean. Agric. For. Meteorol. 125, 105–115. Oke, T.R., 1988. Boundary Layer Climates, 2nd edition. Methuen, London. Papale, D., Reichstein, M., Aubinet, M., Canfora, E., Bernhofer, C., Longdoz, B., Kutsch, W., Rambal, S., Valentini, R., Vesala, T., Yakir, D., 2006. Towards a standardized processing of Net Ecosystem Exchange measured with eddy covariance technique: algorithms and uncertainty estimation. Biogeosciences 3, 571–583. Reichstein, M., Falge, E., Baldocchi, D., Papale, D., Aubinet, M., Berbigier, P., et al., 2005. On the separation of net ecosystem exchange into assimilation and ecosystem respiration: review and improved algorithm. Glob. Chang. Biol. 11 (9), 1424–1439. Reichstein, M., Bahn, M., Mahecha, M.D., Kattge, J., Baldocchi, D., 2013. Linking plant and ecosystem functional biogeography. PNAS 111 (38), 13697–13702. Rodrigues, T.R., Vourlitis, G.L., Lobo, F.A., Oliveira, R.G., Nogueira, J., 2014. Seasonal variation in energy balance and canopy conductance for a tropical savanna ecosystem of south central Mato Grosso, Brazil. J. Geophys. Res. 119, 1–13. Rubel, F., Kottek, M., 2011. Comments on:’ The thermal zones of the Earth’ by Wladimir Köppen (1884). Meteorol. Z. 20 (3), 361–365. Sánchez, J.M., Caselles, V., Rubio, E.M., 2010. Analysis of the energy balance closure over a FLUXNET boreal forest in Finland. Hydrol. Earth Syst. Sci. 14 (8), 1487–1497. Schmid, H.P., 1997. Experimental design for flux measurements: matching scales of observations and fluxes. Agric. For. Meteorol. 87, 179–200. Stiegler, C., Johansson, M., Christensen, T.R., Mastepanov, M., Lindroth, A., 2016. Tundra permafrost thaw causes significant shifts in energy partitioning. Tellus B Chem. Phys. Meteorol. 68 (1), 30467. Stoy, P.C., Mauder, M., Foken, T., Marcolla, B., Boegh, E., et al., 2013. A data-driven analysis of energy balance closure across FLUXNET research sites: the role of landscape scale heterogeneity. Agric. Forest Meteorol. 171-172, 137–152. Twine, T.E., Kustas, W.P., Norman, J.M., et al., 2000. Correcting eddy covariance flux underestimates over a grassland. Agric. For. Meteorol. 103, 279–300. Williams, D.G., Cable, W., Hultine, K., Hoedjes, J.C.B., Yepez, E.A., Simonneaux, V., ErRaki, S., Boulet, G., de Bruin, H.A.R., Chehbouni, A., Hartogensis, O.K., Timouk, F., 2004. Components of evapotranspiration in an olive orchard determined by eddy covariance, sap flow and stable isotope techniques. Agric. For. Meteorol. 125, 241–258. Wilson, K., Goldstein, A., Falge, E., Aubinet, M., Baldocchi, D., Berbigier, P., Bernhofer, C., Ceulemans, R., Dolman, H., Field, C., Grelle, A., Ibrom, A., Law, B.E., Kowalski, A., Meyers, T., Moncrieff, J., Monson, R., Oechel, W., Tenhunen, J., Valentini, R., Verma, S., 2002. Energy balance closure at FLUXNET sites. Agric. Forest Meteorol. 113, 223–243. Wolf, A., Laca, E.A., 2007. Cospectral analysis of high frequency signal loss in eddy covariance measurements. Atmos. Chem. Phys. Discuss. 7, 13151–13173. Wolf, A., Saliendra, N., Akshalov, K., Johnson, D.A., Laca, E., 2008. Effects of different eddy covariance correction schemes on energy balance closure and comparisons with the modified Bowen Ratio system. Agric. For. Meteorol. 148, 942–952. Wu, J.B., Zhou, X.Y., Wang, A.Z., Yuan, F.H., 2015. Comparative measurements of water vapor fluxes over a tall forest using open- and closed- path eddy covariance system. Atmos. Meas. Tech. 8, 4123–4131. Zhang, D., Du, Q., Zhang, Z., Jiao, X., Song, X., Li, J., 2017. Vapour pressure deficit control in relation to water transport and water productivity in greenhouse tomato production during summer. Sci. Rep. 7 srep43461. Zhao, W., Liu, B., Chang, X., Yang, Q., Yang, Y., Liu, Z., Cleverly, J., Eamus, D., 2016. Evapotranspiration partitioning, stomata conductance, and components of the water balance: a special case of a desert ecosystem in China. J. Hydrol. 538, 374–386. Zilitinkevich, S.S., Calanca, P., 2000. An extended similarity theory for the stably stratified atmospheric surface layer. Quat. J. R. Meteorol. Soc. 126, 1913–1992.

Foken, T., 2008a. The energy balance closure problem: an overview. Ecol. Appl. 15 (6), 1351–1367. Foken, T., 2008b. Micrometerorolgoy. Springer-Verlag, Berlin Heidelberg ISBN: 978-3540-74665-2. Foken, T., Wichura, B., 1996. Tools for quality assessment of surface-based flux measurements. Agric. For. Meteorol. 78, 83–105. Foken, T., Leuning, R., Oncley, S.R., Mauder, M., Aubient, M., 2012. “Chapter 4: corrections and data quality control”. Eddy Covariance: A Pratical Guide to Measurement and Data Analysis. Springer Dordrecht Heidelberg, London New York. Franssen, H.J.H., Stockli, R., Lehner, I., Rotenberg, E., Seneviratne, S.I., 2010. Energy balance closure of eddy-covariance data: a multisite analysis for European FLUXNET stations. Agric. For. Meteorol. 150, 1553–1567. Groenendijk, M., van der Molen, M.K., Dolman, A.J., 2009. Seasonal variation in ecosystem parameters derived from FLUXNET data. Biogeosci. Discuss. 6, 2863–2912. Gu, L.H., Meyers, T., Pallardy, S.G., Hanson, P.J., Yang, B., Heuer, M., Hosman, K.P., Liu, Q., Riggs, J.S., Sluss, D., Wullschleger, S.D., 2007. Influences of biomass heat and biochemical energy storages on the land surface fluxes and radiative temperature. JGR Atmospheres 112. Haslwanter, A., Hammerle, A., Wohlfahrt, G., 2009. Open-path vs. Closed-path eddy covariance measurements of the net ecosystem carbon dioxide and water vapor exchange: a long-term perspective. Agric. For. Meteorol. 149, 291–302. Högström, U., 1996. Review of some basic characteristics of the atmospheric surface layer. Boundary Layer Meteorol. 78, 215–246. Holding, D.R., Streich, A.M., 2002. Plant Growth Processes: Transpiration, Photosynthesis, and Respiration. Retrieved from. https://extension.unl.edu/ statewide/antelope/Plant%20Growth%20Processes.pdf. Jocher, G., Marshall, J., Nilsson, M.B., Linder, S., Simon, G.D., Hornlund, T., Lundmark, T., Nasholm, T., Lofvenius, M.O., Tarvainen, L., Wallin, G., Peichl, M., 2017. Impact of canopy decoupling and subcanopy advection on the annual carbon balance of a boreal scots pine forest as derived from eddy covariance. J. Geophys. Res. Biogeosci. 123. Kochendorfer, J., Paw U, K.T., 2011. Field estimates of scalar advection across a canopy edge. Agric. For. Meteorol. 151, 585–594. Köppen, Wladimir, 1884. Translated by Volken, E.; Brönnimann, S. "Die Wärmezonen der Erde, nach der Dauer der heissen, gemässigten und kalten Zeit und nach der Wirkung der Wärme auf die organische Welt betrachtet" [The thermal zones of the earth according to the duration of hot, moderate and cold periods and to the impact of heat on the organic world)]. Meteorol. Z. 20 (3), 351–360 (published 2011). Kruskal, W.H., Wallis, W.A., 1952. Use of ranks in one-criterion variance analysis. J. Am. Stat. Assoc. 47 (260), 583–621. Kucera, C.L., 1954. Some relationships of evaporation rate to vapor pressure deficit and low wind velocity. Ecology 35, 71–75. Kustas, W.P., Norman, J.M., 1999. Evaluation of soil and vegetation heat flux predictions using a simple two-source model with radiometric temperatures for partial canopy cover. Agric. For. Meteorol. 94 (1), 13–29. Leuning, R., 1995. A critical appraisal of a combined stomatal-photosynthesis model for C3 plants. Plant Cell Environ. 18, 339–355. Leuning, R., Judd, M.J., 1996. The relative merits of open- and closed path analysers for measurements of eddy fluxes. Glob. Chang. Biol. 2, 241–254. Leuning, R., van Gorsel, E., Massman, W.J., Issac, P.R., 2012. Reflections on the surface energy imbalance problem. Agric. Forest Meteorol. 156, 65–74. Liu, H., Randerson, J.T., Lindfors, J., Massman, W.J., Foken, T., 2006. Consequences of incomplete surface energy balance closure for CO2 fluxes from open-path CO2/H2O infrared gas analyzers. Boundary-Layer Meteorol. 120, 65–85. Majozi, N.P., Mannaerts, C.M., Ramoelo, A., Mathieu, R., Nickless, A., Verhoef, W., 2017. Analysing surface energy balance closure and partitioning over a semi-arid savanna FLUXNET site in Skukuza, Kruger National Park, South Africa. Hydrol. Earth Syst. Sci. Discuss. 21, 3401–3415. MATLAB, 2014. The MathWorks, Inc., Natick, Massachusetts, United States.

21