Introduction
The comparison of modeled and measured energy and matter fluxes in
a heterogeneous landscape is still a challenge. The fluxes measured
with the eddy covariance technique are related to all surfaces on the
upwind side of the measurements, and the influence of each surface on
the measured data is given by the footprint function. The comparison
of these flux measurements with one-dimensional models can be done over
a homogeneous surface easily or over a heterogeneous surface if the
model is parameterized for all surface characteristics and averaged
according to the percentage of each surface in the footprint
tile approach;. The footprint varies with
changing wind velocity and atmospheric stratification for each
half-hourly-derived flux, and consequently the tile approach must also be
modified every half hour. This approach was proposed by
and successfully improved by and
for two separated surfaces. The tile approach is
based on the application of the eddy covariance method with the
correction of the storage term, which ignores any advection. The more
appropriate approach would use the generalized eddy covariance method
, which includes the advection of matter and
fluxes. Unfortunately, matter advection could only be analyzed
experimentally under very idealized conditions up to now. Experimental
advection studies with an acceptable number of measurement points have
failed . Furthermore, the resolution of
available 3-D models is too large
(25–50 m) for the small-scale heterogeneities on the clearing
(5–10 m). For our case, an experimental advection experiment cannot
be realized and only an LES process study could be applied for
a single 1 h case . Therefore, the one-dimensional
model with a tile approach according to the footprint of each
eddy covariance measurement is the only realistic approach.
A further problem is the so-called unclosed energy balance. The energy
balance with turbulence measurements eddy covariance
method; is not closed by an amount of up to 30 %
according to the overview paper by , while the models
close the energy balance by definition. The main reason for this is
most likely the presence of fluxes caused by larger turbulence
structures like secondary circulations . Measurement
data must be corrected for energy balance closure before comparison
with model outputs can be made. The missing flux will be distributed
between the sensible and the latent heat flux according to the Bowen
ratio or the buoyancy flux
. Carbon and other trace gas fluxes have not
been corrected before the present study.
Simulating the turbulent transfer for heterogeneous landscapes
utilizing a one-dimensional SVAT model (soil–vegetation–atmosphere
transfer) represents a multifaceted challenge. For forest, coherent
structures are a typical phenomenon of turbulent exchange
, and they can be measured with the
eddy covariance technique. However, coherent structures modify the
turbulent transfer of energy and matter in a manner that complicates
capturing them with models. Higher-order closure models are able to
overcome this problem . In this study the
Advanced Canopy–Atmosphere–Soil Algorithm ACASA;
has been utilized with a third-order closure scheme
, and a more exact resolution of the turbulent
structure for forest is provided. Thus, the ACASA model is much better
suited to handling coherent structures and counter gradients than
classical SVAT models with a first-order closure
. provide a summary of the
results of all model studies made in the last nearly 20 years at the
Waldstein–Weidenbrunnen site, without any discussion of specific
problems.
This paper is based on the earlier study by ,
in which the ACASA model was insufficiently applied for the clearing and
compared with eddy covariance data. As a first guess, the tile
approach for the clearing was derived from the relative percentages of
the individual land cover classes of the whole clearing without any
selection according to the footprint. Furthermore, the measured fluxes
were corrected for the energy balance closure. The present paper
extends and completes the first paper in the following directions:
(i) For the clearing, a more precise concept of the comparison of flux
measurements with 1-D modeling using a tile approach according to the
half-hourly footprints. (ii) for the forest, a detailed discussion of
the two different energy balance correction methods is provided. The
inclusion of both surfaces in this study is necessary because flux
studies often use MODIS (Moderate-Resolution Imaging
Spectroradiometer) data with a grid size of 500 or 250 m. Since
the windthrow in 2007, the satellite sees about 50 % forest and
50 % clearing in the relevant grid.
Material and methods
Waldstein–Weidenbrunnen site
The experimental data for the initialization of the model and the
evaluation of its outputs were collected during the third intensive
observation period (IOP3) of the EGER (ExchanGE processes in
mountainous Regions) project . The IOP3 campaign took
place from 13 June to 26 July 2011. The EGER project mainly aimed at
capturing the relevant exchange processes within the
soil–vegetation–atmosphere framework and their interactions at
different scales, and the work in IOP3 thereby focused on the role of
surface heterogeneities in atmospheric exchange and chemistry as well
as biogeochemistry.
The experimental site (50∘8′ N,
11∘52′ E, about 775 m a.s.l.) is part of the
Bayreuth Centre of Ecology and Environmental Research (BayCEER) and
lies in the upper part of the Lehstenbach catchment
(Fig. a). It is situated between two
hilltops: to the southwest the Großer Waldstein
(879 m a.s.l.) and the Bergkopf (857 m a.s.l.) to the
northeast. The Lehstenbach catchment is part of the slopes of the
Waldstein in the northwestern part of the Fichtelgebirge, a low mountain range in
northeastern Bavaria, Germany and is mostly densely forested.
The measurements of IOP3 were carried out in a spruce forest next to
the FLUXNET site Waldstein–Weidenbrunnen (DE-Bay) and in a nearby
clearing with heterogeneous low vegetation located to the south of the
FLUXNET site (Fig. b). The clearing was created by
a windthrow on 18 January 2007 that was induced by the European
windstorm Kyrill . The vegetation of the measurement
site is heterogeneous and there is a slope of 3∘ from the
pine forest to the clearing (from north to south). A more detailed
description of the surrounding topography is provided by
.
(a) Location of the BayCEER research site at the
Waldstein hillsides in the Fichtelgebirge
modified from, published with the kind
permission of Springer, Berlin,
Heidelberg, 2004. (b) Aerial picture of
the patchy landscape showing meteorological towers and masts as well as
the measuring points that are important for this study. MT: main
tower, TT: turbulence tower, TM: turbulence mast; modified from
and published with the kind permission of
Springer, Berlin, Heidelberg, 2017, and
Google Earth, 2015.
![]()
The forest consists mainly of Norway spruce (Picea abies)
with a stand height of about 27 m and a leaf area index (LAI) of
4.8 m2m-2 . The stand age is circa
60 years (estimate from 2013) and the forest structure is
characterized by an open trunk space and a dense crown
space. Most needles are located between 0.5 zhc-1
and 0.9 zhc-1, where z is the measurement height
and hc the stand height of 27 m. In the ground area,
the understory comprises two-thirds crinkled hair grass
(Deschampsia flexuosa) and moss (together LAI ≤0.5 m2m-2), and one-third is characterized by blueberry
(Vaccinium myrtillus) and young Norway spruce (Picea abies, together plant area index (PAI) of 3.5 m2m-2). For more details,
see . The information regarding the overstory trees is
summarized in Table . The vegetation of the
clearing is a conglomeration of young spruce trees, blueberry
(Vaccinium myrtillus), and different grasses
(e.g., Calamagrostis, Agrostis, Poaceae). However, deadwood
and bare soil are also elements of the clearing, and together
they make up one-third of the ground cover of this area. The clearing
also contains approximately 2 % young deciduous trees: alder
(Alnus), maple (Acer), beech (Fagus), and
sorbus (Sorbus), with a greater canopy height than the
remaining species. Due to the heterogeneous vegetation at the
clearing, no distinct mean canopy height can be estimated for the
clearing. Table provides an overview of the
species present and their mean canopy height.
Vegetation at the forest and the Köhlerloh clearing with their main characteristics according to .
Species
Ground
Height
PAI
MSCa
cover
in
in
in
in %
m
m2m-2
molm-2s-1
Forest
Spruce (Picea abies)
100
27
5.6±2.1
0.001
Clearing
Crinkled hair grass (Deschampsia flexuosa)
21.7
0.17±0.05
2.65±1.08
0.127
Young spruce (Picea abies)
21.4
1.21±0.50
8.67±2.29
0.054
Blueberry (Vaccinium myrtillus)
15.9
0.27±0.10
3.46±1.05
0.050
Reed, bent, and true grasses (Calamagrostis, Agrostis, Poaceae)
9.0
0.42±0.11
3.43±1.07
0.049
Rush and sedge family (Juncaceae, Cyperaceae)
3.1
0.74±0.13
1.77±0.60
0.002
Other herbaceousb
1.6
/
/
0.002
Moss
0.9
/
/
/
Dead grass, bare soil
7.2
/
/
/
Deadwood
18.8
/
/
/
Small ditches
0.2
/
/
/
a MSC stands for minimal stomata conductance and is a model parameter utilized in ACASA.b This includes Digitalis purpurea (common foxglove), Epilobium angustifolium (willow herb), and Urtica dioica (stinging nettle).
Experimental setup and data
During IOP3, high-frequency turbulence measurements were conducted at
different measurement towers (Fig. b). For this study,
the data from the towers in the forest (MT, TT) and the
turbulence mast (TM) at the clearing have been especially utilized. The towers
and masts were generally equipped with 3-D sonic anemometers to
enable the collection of information about wind components (u, v, w)
and the sonic temperature (TS). High-frequency gas
analyzers for carbon dioxide (cCO2) and water vapor (q)
were installed in conjunction with sonic anemometers. This allowed the
turbulent exchange for forest and clearing to be investigated with the
devices as summarized in Table , and this has been
utilized for comparison with simulations. Raw flux data (20 Hz) have
been processed with the TK3 software . The flux data
have been additionally classified by their quality according to the
classification scheme by . For comparison with model
results, only data with quality flags of 6 or better have been
utilized. Furthermore, the flux measurements have been energy balance
corrected with the methods described in Sect. .
The ACASA model needs half-hourly meteorological input values as well
as mean values of soil temperature and soil moisture for
initialization. These input values (see Table ) are
provided by standard measurements at the Pflanzgarten
(Fig. b) and measurements at the main tower (MT) for
the spruce forest as and at the turbulence mast (TM) for the
clearing. Table provides an overview of the
meteorological input data for the two simulated sites and the
corresponding instrumentation.
Three golden day periods (GDPs) were selected within IOP3 because of
fair weather conditions and predominantly good performance of the
measurement devices. The first GDP was from 26 to 29 June 2011
(corresponding to the 177th to 180th day of the year, DOY). On the last
3 days of this period, clear-sky conditions prevailed until
14:00 as
did moderate westerly (26 and 27 June) or easterly (28 and 29 June) winds. The second GDP occurred from 4 to 8 July 2011 (DOY:
185 to 189), with the best weather conditions on 6 and 7 July and
partly overcast sky on the rest of the days. The wind in this period
was weak and blew from the west (4, 7, and 8 July) or east (5 and 6 July). Minor
amounts of precipitation were measured on 4 and 8 July. For the last
GDP from 14 to 17 July 2011 (DOY: 195 to 198) data are missing for the
clearing at 5.5 m.
Meteorological instrumentation for the determination of the
sensible and the latent heat flux as well as the net ecosystem
exchange (NEE) for forest and clearing at different measurement
towers in distinct heights, using the eddy covariance technique
(with the frequency of 20 Hz applied as indicated), and
additionally, the most important input parameter of the model
(for this purpose the mean CO2 concentrations have been
averaged from the turbulence data). Documentation of the complete
data set is available in .
Measurement site
Parameter
Height
Device
Pflanzgarten
Precipitation
1 m
OMC 212, Adolf Thies GmbH & Co. KG
Pressure
2 m
AB 60, Ammonit Gesellschaft fürMesstechnik mbH
Weidenbrunnen,
Wind vector and sonic temperature (20 Hz)
32.5 m
USA-1, METEK GmbH
main tower (MT)
Humidity (20 Hz)
32 m
LI-7000b, LI-COR Biosciences
Temperature and relative humidity
31 m
Frankenberger ventilated psychrometer, Theodor Friedrichs & Co
Wind velocity
31 m
Cup anemometer,Theodor Friedrichs & Co
Upwelling and downwelling shortwave radiation
29.5 m
CM 14, Kipp & Zonen
Upwelling and downwelling longwave radiation
29.5 m
CG 2, Kipp & Zonen
Soil heat flux and temperature profile
-0.08 m
HFT, Campbell Sci. Inc.
Weidenbrunnen,
CO2 concentration (20 Hz)
36 m
LI-7500c, LI-COR Biosciences
turbulence tower (TT)
Wind vector (20 Hz)
36 m
USA-1, METEK
Köhlerloh,
Wind vector and sonic temperature (20 Hz)
2.25 m/5.5 m
CSAT3, Campbell Sci. Inc.
turbulence mast (TM)
Humidity (20 Hz)
2.25 m/5.5 m
LI-7000b/LI-7500c,LI-COR Biosciencesa
CO2 concentration (20 Hz)
2.25 m/5.5 m
LI-7000b/LI-7500c,LI-COR Biosciencesa
Upwelling and downwelling shortwave radiation
2.0 m
CNR4, Kipp & Zonen
Upwelling and downwelling longwave radiation
2.0 m
CNR4, Kipp & Zonen
Soil heat flux and temperature profile
-0.15 m
HP3, Rimco
a LI-7000 at 2.25 m and LI-7500 at 5.5 m.b Closed-path gas analyzer, c open-path gas analyzer.
The ACASA model
ACASA was utilized to simulate the turbulent transfer
of heat, water vapor, and CO2 for spruce forest and
clearing. ACASA is a multilayer SVAT model developed at the
University of California, Davis. The main feature of ACASA is that the
turbulent transfer within and above the canopy is calculated by
a diabatic, third-order closure method, which is based on the
theoretical work of . The multilayer
structure of ACASA is represented by 20 evenly distributed atmospheric
layers, which reach from trunk space to twice the canopy height, and
by 15 soil layers. For the calculation of leaf, stem, and soil surface
temperatures, the fourth-order polynomial scheme of is
incorporated. Utilizing this polynomial allows for the calculation of the
temperatures of these components without making substantial errors in
the case of significant deviations from the ambient temperature. The
model was considerably updated by the University of California, Davis
and is still in use
. For this study, the ACASA model was adapted from
a version of April 2013.
Direct and diffuse radiation can be absorbed, transmitted, or
reflected by the canopy, whereby these processes are dependent on the
leaf and branch distribution of the plants. For this purpose, the
aboveground biomass is distributed in 10 different leaf angle classes
. In order to enable a preferably exact allocation
of the energy to sensible and latent heat fluxes, a leaf energy
balance equation has to be solved. Plant physiological feedback to
micro-environmental conditions is incorporated by the approaches of
and to the Ball–Berry stomatal
conductance combined with photosynthetic rates calculated after an
equation of , whereby coupling follows the standards of
. ACASA also includes canopy heat storage, which is
mainly important for tall vegetation, and canopy interception of
precipitation. For thermal and hydrological aspects the soil model is
based on the one-dimensional diffusion equation as
incorporated in MAPS Mesoscale Analysis and Prediction
System;.
About 10 years ago the University of Bayreuth started to work with
ACASA. The first issue was to use a sensitivity analysis to check
whether the model could be applied for a Central European spruce
forest . We also found some model-specific
problems, and the model was again updated by the University of
California, Davis. Using the updated model, a first study for the
Waldstein–Weidenbrunnen site was published . This
study showed that the ACASA model is more accurate (in comparison to
a K-approach model) when coherent structures dominate at night, which
has been investigated in many experiments and summarized by
. The model was changed by for the
spruce-specific parameterizations in three parts. The first change
relates to the soil respiration calculations and had already been
examined by for spruce. The second change was made
in the calculation of photosynthesis according to
because temperatures less than 10 ∘C are very common at the
experimental site. The third adjustment relates to the day respiration
for leaves because quantified the day respiration
as about one-half of the night respiration rate. Furthermore,
determined several parameters of the plants of
the clearing see Table 16.1 of.
Energy balance correction
The eddy covariance data on sensible heat, latent heat, and carbon
dioxide flux have been energy balance corrected for comparison
with the simulations.
For the correction of the energy fluxes, the residual (Res)
arises from the following equation:
Res=QS-H-LE-G-ΔS,
where QS denotes net radiation, H sensible heat flux,
LE latent heat flux, G the ground flux heat, and ΔS the heat storage term of the aboveground biomass, which was determined according to the investigations
of and . Here, the heat storage
term was only estimated for the spruce forest because of its larger
biomass compared to the vegetation in the clearing. The heat storage
was calculated with the following relation:
ΔS=mfcf(Ta(t)-Ta(t-Δt))Δt,
with mf the aboveground biomass of the forest,
cf the specific heat capacity of wood (cf=1702.8 Jkg-1K-1), and Ta the air
temperature in the spruce canopy. In this study, two different
methods for energy balance correction have been applied to the
measurements of IOP3. The first correction method is introduced by
and preserves the Bowen ratio. This method is usually
utilized for the correction of heat fluxes under the assumption of
measuring errors, but requires their scalar similarity. Henceforth,
this method is abbreviated as EBC-Bo. The second correction method
accounts for near-surface secondary circulations, which transport more
sensible heat. Therefore, the alternative correction method (EBC-HB)
proposed by utilizes the buoyancy flux
ratio. Here, for the typical range of the Bowen ratio, more residual
energy is ascribed to the sensible heat flux, as was done in several
studies . The difference to the EBC-Bo correction is
especially pronounced for low Bowen ratios. At night, the
applicability of the corrections is limited due to the occurrence of
minor or negative heat fluxes.
The discrepancy between measured and simulated NEE can be an effect of
the unclosed energy balance on the CO2 fluxes. Therefore, we
propose a new correction method on the basis of the good scalar
similarity between the humidity and the carbon dioxide concentration
mainly in the first part of the day (for scalar similarity see
), while the similarity between the temperature and
other scalars often fails . The measured NEE is
corrected by referring to the energy-balance-corrected latent heat
flux, which is introduced in . For this purpose,
the ratio k between the corrected and the uncorrected latent heat
fluxes is calculated for both correction methods:
kBo=LEBoLEandkHB=LEHBLE.
The measured NEE value was then multiplied by the respective
correction factor kBo or kHB. This correction
method is proposed for the NEE for the first time and the
applicability of this approach is therefore evaluated. The correction
is not very large. Therefore, until shown otherwise, the energy
balance correction can be neglected for scalar fluxes.
Footprint model
Footprint models are currently a commonly used tool for the
identification of the source area of flux measurements
. For the footprint calculations of this study,
a Lagrangian footprint model is applied , whereby a spectral method for the nonlinear flux
averaging of surface characteristics (friction, roughness length)
according to is employed
because the linear averaging of the roughness length (input parameter
of footprint models) is inaccurate. The sensitivity of the Lagrangian
footprint model to the turbulence statistics was tested by
for the Waldstein–Weidenbrunnen site. A footprint
climatology and footprint studies, including the effect of the
footprint on the data quality, are available from
. On average, more than 80 % of the target
area is forest. Only for southerly winds and stable stratification
does the Köhlerloh clearing have a significant influence on the
fluxes measured at the main tower (MT).
In a study by the land use for the
Waldstein–Weidenbrunnen site was determined and mapped, and this was
tested for footprint applications. Due to the distinct small-scale
heterogeneities, the land cover map applied here has a high resolution
(4.5×4.5 m2). This grid size is in agreement with the
size of the typical heterogeneities and the recommendations for
footprint analyses for low measuring heights . The
calculated footprints for 30 min time periods have been used to
generate tile approaches of the model.
Conclusions
The ACASA model, which was originally developed for tall vegetation,
can also be used with high accuracy for low vegetation, as was
demonstrated for the clearing, if the plant-specific parameters are
appropriately implemented in the model. Therefore, it is applicable to
consider ACASA in tile approaches for heterogeneous areas for tall and short vegetation or clearings.
The footprint-averaged tile approach did not produce significantly
better results than the tile approach for the whole clearing, as was
done by . The footprint model is probably not
accurate enough in the location of the effect levels of the footprint,
considering the small-scale heterogeneities of the clearing in
comparison to the size of the footprint area. This issue has already
been shown in a comparison of different footprint models by
. For larger heterogeneities, this approach may be
more appropriate. Furthermore, the clearing area contained randomly
distributed very wet parts, which could not be classified in the
land use characteristics of the tile approaches.
The small underestimation by the model of fluxes above the forest is
probably a local overestimation of the fluxes at the turbulence tower
(TT), which is adjacent to an area of the forest where higher fluxes are
possible . Large-eddy simulation studies have
shown that at a distance of 10 times the canopy height from the
forest edge, higher fluxes are possible . Such
a highly local effect is impossible to model in ACASA.
Assuming that the ACASA model is well parameterized and the available
energy is accurately distributed to the sensible and the latent heat
flux, a good agreement has been found with the energy-balance-corrected measurements. The correction with the buoyancy flux leads to
better results, but this depends on the Bowen ratio; i.e., for
Bo>1 it is better than for Bo<1. This result
supports earlier findings . The idea
behind the buoyancy correction that buoyancy is the reason for
secondary circulations is only partly true. Possible errors in the
eddy covariance method cannot be inferred from the improved
applicability of the Bowen ratio correction for low Bowen ratios
(Bo<1). However, a recent investigation by
showed a potential phase shift between the vertical wind component and
the scalars, which might bias the determination of turbulent
fluxes. This should be an objective of a further study with a data set
containing more data for Bo<1. The correction of the NEE is
probably useful, but the effect is not very significant for Bo>1; however, it might be more applicable for accumulated fluxes.