Land surface models (LSMs) need to balance a complicated trade-off between computational cost and complexity in order to adequately represent the exchanges of energy, water and matter with the atmosphere and the ocean. Some current generation LSMs use a simplified or composite canopy approach that generates recurrent errors in simulated soil temperature and turbulent fluxes. In response to these issues, a new version of the interactions between soil–biosphere–atmosphere (ISBA) land surface model has recently been developed that explicitly solves the transfer of energy and water from the upper canopy and the forest floor, which is characterized as a litter layer. The multi-energy balance (MEB) version of ISBA is first evaluated for three well-instrumented contrasting local-scale sites, and sensitivity tests are performed to explore the behavior of new model parameters. Second, ISBA-MEB is benchmarked against observations from 42 forested sites from the global micro-meteorological network (FLUXNET) for multiple annual cycles.
It is shown that ISBA-MEB outperforms the composite version of ISBA in improving the representation of soil temperature, ground, sensible and, to a lesser extent, latent heat fluxes. Both versions of ISBA give comparable results in terms of simulated latent heat flux because of the similar formulations of the water uptake and the stomatal resistance. However, MEB produces a better agreement with the observations of sensible heat flux than the previous version of ISBA for 87.5 % of the simulated years across the 42 forested FLUXNET sites. Most of this improvement arises owing to the improved simulation of the ground conduction flux, which is greatly improved using MEB, especially owing to the forest litter parameterization. It is also shown that certain processes are also modeled more realistically (such as the partitioning of evapotranspiration into transpiration and ground evaporation), even if certain statistical performances are neutral. The analyses demonstrate that the shading effect of the vegetation, the explicit treatment of turbulent transfer for the canopy and ground, and the insulating thermal and hydrological effects of the forest floor litter turn out to be essential for simulating the exchange of energy, water and matter across a large range of forest types and climates.
The land surface model (LSM) is one of the key parameterization schemes of
atmospheric models used for numerical weather prediction and climate
simulations. It is used to compute the turbulent fluxes of heat, water and
momentum, along with the radiative fluxes at the surface–atmosphere
interface. In addition, the current generation of LSMs are used to compute
flux exchanges of certain chemical species (such as carbon dioxide and
biogenic organic volatile carbon) and emissions of particles (such as
aerosols from dust or biomass burning). The interactions between
soil–biosphere–atmosphere (ISBA) is part of the SURFace EXternalisée
platform (SURFEX) developed in recent years at Météo-France
In order to remain consistent with the aforementioned developments and to
respond to both current and future user demands, a multi-energy balance (MEB)
approach has been developed
The surface in forested regions beneath the canopy is often covered by
a layer of dead leaves or needles, branches, fruits and other organic
material, which can be characterized as a litter layer. The explicit
inclusion of such a layer is neglected for the most part in LSMs used
in global-scale models, or it is only partly taken into account.
For example,
The goal of the current study is to evaluate the impact of a new
parameterization of the soil–litter–vegetation–atmosphere continuum at the
local scale. In the first part of this study, an in-depth evaluation for
three forest sites in France is carried out. They have been selected in order
to represent a range of forest types and climates. The main goal is to
understand the effect of both the explicit canopy layer and the litter layer
on the available energy partitioning (latent, sensible, ground), soil
temperatures and soil water content. The second part of this study consists
of a benchmark study using 42 sites from the FLUXNET network
This is essential since ISBA is used within the SURFEX platform in various
configurations at resolutions ranging from several kilometers at the regional
scale, such as within the operational mesoscale numerical weather prediction
model AROME
The standard ISBA model uses a single composite soil–vegetation surface
energy budget, which means that the properties of the soil and vegetation are
aggregated within each grid-cell point
The multi-energy balance model, ISBA-MEB, treats up to three fully coupled
distinct surface energy budgets (i.e., the snow surface, the bulk vegetation
canopy and the ground, which is characterized as either a soil surface or
litter layer; see Sect.
Two methods are generally used to represent the effect of a litter layer
within LSMs. The first method consists of adding a specific ground resistance
in order to reduce the soil evaporation due to the presence of a litter
layer. For example,
Energy balance closure is a well-known issue when turbulent fluxes are
computed with the eddy covariance technique. The closure,
Note that there is no specific check here that res
Among the terms in Eq. (
Three well-instrumented forest sites have been used for model evaluation
that cover a range in climate, soils and vegetation characteristics. Their
location is shown in Fig.
The location of the three well-instrumented forested sites in France.
Characteristics of the three French forest sites.
The Barbeau site is located in the Barbeau National Forest, which is
approximately 60 km from Paris (48.4
The Le Bray site is located about 20 km from Bordeaux, France
(44.7
The Puéchabon site is located roughly 35 km northwest of Montpelier in the
Puéchabon State Forest (43.7
The second part of the study uses observations from a subset of the FLUXNET
sites to assess systematically or benchmark the suite of ISBA versions
developed at Météo-France. The FLUXNET database has been used for LSM
evaluation by several widely used LSMs
The forested sites from the FLUXNET network. Selected sites (shown) have a maximum energy imbalance at or below 20 %. The circles indicate the location of sites retained for this study.
All the simulations are performed using SURFEXv8. We use the diffusive soil
(DF) option meaning that the soil heat and mass transfers are solved on a
multi-layer grid
Three simulations were performed for each site in order to assess the impact of the new canopy and litter layers on the simulated fluxes, soil temperature and soil moisture. The models were forced with atmospheric data at half-hourly time steps. In the first simulation, the reference model (i.e., using a single composite vegetation and surface-soil layer) was used and it is hereafter referred to as ISBA. The second simulation with the explicit canopy layer corresponds to the ISBA model using the MEB approach (referred as MEB herein for simplicity). The last simulation was carried out with both the explicit canopy layer and the explicit forest litter layer and it is referred to as MEBL.
The pedotransfer functions of
The main model parameters for each site. Literature indicates that values come from studies cited in the text and estimated means that values were provided by the principal investigators of each site.
Measured litter thickness reported for various sites.
At the Le Bray site, the water table has a significant influence on the seasonal soil wetness. Measurements of its depth were available and allowed for a simple parameterization to be developed. It consists of a strong relaxation towards saturation in soil layers below the observed water table depth. Thus, soil moisture within the saturated zone is very close to the observed values in the saturated or nearly saturated layers, while soil moisture above this zone is permitted to freely evolve.
For each site, simulated turbulent heat fluxes were compared to both the original and the adjusted-observed values assuming that model results should fall inside the area delimited by these two curves. Moreover, since a proper evaluation of each flux component of the energy balance has to be done with a closed energy budget (in order to be consistent with the model which imposes closure by design), the scores are computed with the adjusted flux values.
The ECOCLIMAP land cover and the HWSD soil databases were used to prescribe
most of the needed parameters as was done for the local-scale evaluation for
the three French sites (in the previous section). This is also consistent
with the method used for spatially distributed offline and fully coupled
(with the atmosphere) simulations with ISBA. However, note that soil texture,
canopy height and vegetation type were chosen in agreement with literature
values for each site where available; therefore, they superseded the ECOCLIMAP
values for these parameters. The default thickness of the litter layer is set
to 3 cm based on the sensitivity test results presented in
Sect.
Initial conditions can have a significant impact on both the sensible and the latent heat fluxes, but they were not known; thus, a spin-up period was used for each site. Sites were initialized with saturated soil water content conditions and the first available year was repeated at least 10 times until a predefined convergence criteria was achieved. Only the latent and sensible heat fluxes are evaluated since the ground heat flux, soil temperature and soil water content were not available at each FLUXNET site.
The MEB model has been compared to both observations and the standard ISBA composite vegetation model for several well-instrumented contrasting (in terms of forest type and climate) sites in France, and to a subset of the FLUXNET sites for multiple annual cycles using a benchmarking application. The results of these evaluations and of several sensitivity tests are given in this section.
The total net radiation (
RMSE (root mean square error),
Since MEB explicitly uses the shortwave radiation transmitted through the
canopy for the ground net radiation computation, it can be compared with the
photosynthetically active radiation (PAR) measurements below the canopy
within the Barbeau and Puéchabon forests (Fig.
The modeled (thin dashed line) and observed (thick
dashed line) incoming shortwave radiation transmitted through
the
canopy at
The simulated LW
This is due, in part, to the use of the same values of emissivity for soil,
snow and vegetation in ISBA and MEB. It is also due to the fact that
LW
At Le Bray and Puéchabon, the simulated values of total LE are relatively
close between the three simulations and in fairly good agreement with the
adjusted measurements. Above the forest, most of the evapotranspiration comes
from canopy transpiration (Fig.
The partitioning of latent heat flux for each site and
model option into transpiration (black), ground/litter
evaporation (white) and evaporation from the canopy (gray). Values for Le
Bray (2006), Puechabon (2006) and Barbeau (2013) are shown in panels
For the Mediterranean forest at Puéchabon, most of the net radiation is
converted into sensible heat flux (Fig.
The two other sites have a higher annual evapotranspiration than the
Mediterranean site and a larger ground evaporation, especially for ISBA. At
Le Bray, evapotranspiration is almost the same for the three simulations
(Fig.
At Barbeau, the interpretation of the results is much more complex because of
the deciduous broadleaf nature of the forest. In winter, the LAI is very low
and latent heat flux is dominated by the ground evaporation. In spring, when
net radiation increases but vegetation is not fully developed, MEB
significantly overestimates LE owing to a large ground evaporation
(Fig.
For all of the sites, more significant differences occur for the simulated
The improvement for the two MEB simulations are mainly related to three
processes;
The use of an explicit canopy layer which
intercepts most of the downward solar radiation thereby reducing
the net radiation at the ground surface. This leaves more energy
available for turbulent fluxes in contrast to ISBA which is
directly connected to forest floor and can more easily
propagate energy into the ground by conduction. The use of a lower roughness length ratio (momentum to
heat: The presence of a litter layer limits the penetration of energy into
the ground because of its low thermal diffusivity
which leads to a reduction of the ground heat flux (amplitude) and
thus to more available energy for
A substantial effect of both MEB and MEBL is the ground heat flux reduction.
The explicit representation of the canopy induces a shading effect due to the
leaves, stems and branches, so that less energy reaches the ground. But in
addition, the explicit representation of the litter layer also modifies the
ground heat flux by acting as a buffer for the top soil layer due to its low
thermal diffusivity. It can be seen in Fig.
The total soil water content calculated over the root depth indicated in Table
The monthly diurnal cycle composite at Puechabon. MEBL is in red, MEB in blue, ISBA in green, measurements are indicated by a dashed black line and adjusted measurements are represented using a solid black curve. As a visual aid, the area between the latter two curves is shaded. Ideally, model results fall within this area.
As in Fig.
As in Fig.
The Taylor diagram for ground heat flux,
The ISBA overestimation of the flux amplitude often represents several 10's
of W m
The root mean square error, RMSE, square correlation coefficient,
A good description of the soil thermal characteristics is needed to model
temperatures at different depths, along with a correspondingly good estimate
of the surface-soil heat flux. The soil characteristics (thermal conductivity
and heat capacity) are calculated based on the input soil texture (sand and
clay fractions) and organic matter contained in the soil. The soil
temperature simulation statistics in Table
Monthly average soil temperature (K) diurnal cycle (at 0.04 m
soil depth) composites at Le Bray
The observed and simulated soil water content
time series for an annual cycle are
displayed in Fig.
More significant differences can be seen in terms of the top soil water
content. In particular, for the Le Bray and Barbeau forests, at the end and
beginning of the year, respectively, ISBA and MEB simulate unrealistic drops
in the liquid soil water content (Fig.
There is uncertainty with respect to the definition of several key model
parameters which are not usually available in the observations so that
several sensitivity tests have been undertaken. Three parameters have been
tested; (i) the extinction coefficient for the long-wave transmission through
the canopy
For all three of the sites, the long-wave transmission
coefficient has a very weak influence on each of the simulations for a
reasonable range of values (0.3 to 0.7, 0.5 being the default value which is
based on an value used by
The litter layer thickness values tested range between 0.01 and 0.10 m based
on values from the literature (Table
The
The clumping index was tested since it is the key parameter of the radiative
transfer controlling the transmission and absorption of incoming shortwave
radiation through the canopy
In summary, owing to these sensitivity tests, the default extinction
coefficient for long-wave transmission is retained as a constant value of
0.5. A default constant value of litter thickness is defined as 0.03 m since
it is both not widely observed and it tends to be a threshold for which the
effect of litter on
The comparison of modeling results with field measurements from over 42
FLUXNET sites (Fig.
MEBL and ISBA generally performed well in simulating the
Scatter plots of RMSE,
The sensible heat flux AE (Fig.
The amplitude of the sensible heat flux diurnal cycle is also improved as
suggested by the improved RMSE values (Fig.
The latent heat flux, LE, results between the two models are more similar
than for
As in Fig.
In a general, a very good consistency is found between the analysis in
Sect.
As mentioned in Sect.
Scatter plots of RMSE, for sensible heat flux
Finally, it is of interest to determine if the results are conditioned by
certain key physiographic parameters, notably the LAI. The differences in
RMSE between MEBL and ISBA are shown in Fig.
Box plots of improvement in RMSE between MEBL and
ISBA. Each RMSE value is computed with over a month-long period
that satisfies the energy budget closure condition. The sensible heat
flux (
This study is the second of a set of two papers which describe the introduction of the new multi-energy balance (MEB) approach within the interactions between soil–biosphere–atmosphere (ISBA) model as part of the SURFEX platform. Two new explicit bulk layers have been implemented, one for the vegetation canopy and the other for a litter layer. This paper describes a two-part local-scale offline evaluation of both the bulk canopy scheme (MEB) and the combined bulk canopy–litter layer (MEBL) approaches, and the results are also compared to the standard composite vegetation ISBA model. The model parameterization governing the litter layer is also presented. The evaluation is done by investigating the ability of the models to simulate the fluxes above (sensible and latent heat flux) and below (ground heat flux) different forest canopies and the ability of the different approaches to reproduce the observed soil temperatures and soil water content.
In the first part of this study, an evaluation over three well-instrumented forested sites in France was done using observed forest characteristics, and turbulent, radiative and heat conduction (ground flux) measurements with a particular attention paid to the energy balance closure issue. The mid-latitude forest sites were contrasting in terms of both vegetation type (needleleaf, broadleaf) and climate (Mediterranean, temperate). In the second part of this study, a statistical evaluation was done using the framework of a benchmark platform based upon 42 sites scattered throughout the world from the FLUXNET network.
In terms of the model evaluation for the three French forested sites, the
standard ISBA model was found to underestimate the amplitude of the sensible
heat flux,
In terms of temporal dynamics, the main differences in latent heat flux
between ISBA and MEBL occur during spring for the deciduous forest site where
the litter layer acts to significantly limit soil evaporation, whereas ISBA
and MEB (without explicit litter) overestimate evapotranspiration due to
strong ground evaporation (owing to a relatively low LAI combined with large
incoming radiation and generally low to unstressed soil conditions). And
despite the overall similar total annual evapotranspiration simulated by ISBA
and MEBL, the partitioning between canopy evapotranspiration,
The main conclusions made from analysis of the three well-instrumented sites were found to be consistent with the results of a statistical benchmark analysis over a subset of 42 forested sites from the FLUXNET network. The sensible heat flux RMSE was improved for 87.5 % of these sites with the new parameterizations (MEBL). The selected sites encompassed a wide range of climate and several different forest land-cover types, thus it is a necessary test before implementing MEBL in regional to global-scale applications. The benchmark also showed that the impact of the explicit treatment of the canopy and litter layer was more significant for relatively open canopies (low to medium LAI values), whereas for closed canopies (high LAI), all three of the model approaches simulation results converge. This is not overly surprising since in the limit as a canopy becomes tall and quite dense, the composite scheme resembles a vegetation canopy (the soil contribution becomes significantly less). But again, ISBA tends to simulate considerably more bare-soil evaporation in the peak growing season (maximum LAI) than MEBL, so there is error compensation which is masked to a large extent when looking at the total evapotranspiration.
In terms of prospectives, evaluation of MEB/MEBL is ongoing, and offline
spatially distributed applications within the SIM chain (SAFRAN-ISBA-MODCOU,
The MEB code is a part of the ISBA LSM and is available as
open source via the surface modeling platform called SURFEX, which can be
downloaded at
Forest litter is represented using a single model layer which generally
ranges in thickness from 0.01 to 0.10 m, and in the absence of ancillary
data, the default value is 0.03 m. When this option is active, an additional
layer is added to the soil for the thermal and energy budget computations
with litter-specific thermal properties. This means that the numerical
solution method is identical to that presented in the companion paper by
The energy budget for the snow-free litter layer can be expressed as
The liquid-water content of the litter layer evolves following
The phase change rate,
It is assumed that litter below the canopy is spatially homogeneous so that
it intercepts all of the incoming radiation. Thus, the net radiation
The below-canopy sensible heat flux,
Finally, the ground conduction flux (W m
The water intercepted by the litter layer corresponds to the sum of the rain
passing through the canopy and the drip from the canopy.
The litter thermal conductivity,
These three scores are defined respectively as
The centered root mean square error (cRMSE) difference between MEBL and ISBA is defined as
The authors declare that they have no conflict of interest.
This work is a contribution to the ongoing efforts to improve the SURFEX platform. The authors would like to thank the teams of the three French eddy flux towers involved in this study: Le Bray, Barbeau and Puéchabon, for making their data available, and to all of the contributors to the FLUXNET database. This study was fully funded by a Météo-France grant. Edited by: S. ValckeReviewed by: two anonymous referees