Effect of a coupled soil water–plant gas exchange on forest energy fluxes: Simulations with the coupled vegetation–boundary layer model HIRVAC

Page 1

R

Eev

BT

a

A

R

R

1

A

P

K

A

F

S

W

S

0d

e c o l o g i c a l m o d e l l i n g 2 1 4 ( 2 0 0 8 ) 75–82

avai lab le at www.sc iencedi rec t .com

journa l homepage: www.e lsev ier .com/ locate /eco lmodel

eview

ffect of a coupled soil water–plant gas exchange on forestnergy fluxes: Simulations with the coupledegetation–boundary layer model HIRVAC

jorn Fischer ∗, Valeri Goldberg, Christian Bernhoferechnische Universitat Dresden, Institute for Hydrology and Meteorology, Department of Meteorology, 01062 Dresden, Germany

r t i c l e i n f o

rticle history:

eceived 11 January 2007

eceived in revised form

5 February 2008

ccepted 28 February 2008

ublished on line 22 April 2008

eywords:

tmospheric boundary layer model

orest energy fluxes

oil moisture

ater stress

tomatal reaction

a b s t r a c t

Long-living plant communities such as forests reduce their transpiration by closing and

opening the leaf stomata as a common strategy to save water in dry periods. Meteorological

models including vegetation should consider this mechanism to simulate realistic water

transport from the plant to the atmosphere. Results of the German network project VERTIKO

showed that commonly used meso-models such as Lokalmodell (German Weather Service)

often overestimate evapotranspiration of vegetated surfaces during dry periods. This is,

among other things, due to the insufficient plant-specific coupling between the soil water

content and the physiological reactions of leaf stomata in the implemented SVAT modules.

This study presents an approach to describe the above-mentioned coupling mechanism by

upgrading the coupled vegetation boundary layer model HIRVAC. A stomatal reaction on soil

moisture change, which is a part of HIRVAC, is parameterised in the included mechanistic

photosynthesis model for C3 plants (PSN6).

In the new HIRVAC version several parameters of the PSN6 model were scaled by a power

function of the matrix potential to consider the stomatal reaction to changes in soil water

content. This leads to an adaptation of the additional humidity source term in the basic

equation of HIRVAC. As a result the humidity profiles in the canopy air, the latent heat flux

and the energy balance of each canopy model layer are changed.

The new parameterisation in HIRVAC was applied for the VERTIKO special observation

period in May and June 2003 for vegetation parameters adapted to the Tharandter Wald

Anchor Station (experimental site of the Department of Meteorology, TU Dresden).

The HIRVAC modification leads to a realistic decrease in latent heat flux for dry soil con-

ditions. Without coupling, latent heat flux increases continuously due to an increase in

the atmospheric driving parameters of vapour pressure deficit (vpd) and temperature in

the canopy. Despite some differences during the night the simulated and measured sen-

sible heat flux agree very well, especially under dry soil conditions, while the correlation

between measured and simulated latent heat flux is only moderate. The best agreement

between simulated and m

temperature under moder

∗ Corresponding author. Tel.: +49 351 463 31341; fax: +49 351 463 31302.E-mail address: [email protected] (B. Fischer).

304-3800/$ – see front matter © 2008 Elsevier B.V. All rights reserved.oi:10.1016/j.ecolmodel.2008.02.037

easured latent heat flux was reached for the vpd and crown air

ate soil moisture conditions.

© 2008 Elsevier B.V. All rights reserved.

Page 2

76 e c o l o g i c a l m o d e l l i n g 2 1 4 ( 2 0 0 8 ) 75–82

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 762. Methods. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

2.1. Model description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 762.2. New parameterisation of the model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

3. Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 783.1. Model settings and input data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 783.2. Model results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

4. Conclusions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

1. Introduction

Latent and sensible heat fluxes are important variables inmeteorology, hydrology and ecology, as they interact withthe local climate by mass and energy exchange between theground and the atmosphere. To reduce uncertainty in predic-tions of surface energy exchange or to fill data gaps a betterunderstanding of ecosystem processes controlling energy bal-ance components is required (Falge et al., 2005).

Plants usually experience a fluctuating water supply dur-ing their life cycle due to continuously changing atmosphericfactors. Even in areas where annual rainfall is high, unevendistribution often exposes plants to periodic soil drying. Plantswill meet various soil water deficiencies of varying severities,frequencies, and durations during their growth stage (Lianget al., 2002). The increase in hydraulic resistance in the soil-root interface (dry period) (Barataud et al., 1995) reduces watertransport to roots and then to leaves and makes a fast recoveryof the soil water potential impossible under severe droughtconditions (Sellin, 1998). The low water content leads to thestomata closing, whereby the transpiration is strongly reduced(Jones, 1998; Gottschalck et al., 2001). This is an importantstrategy to protect trees against loss of water during dryness.

The results of the German network projects VERTIKO(Bernhofer and Kostner, 2006) and Eva grips (Mengelkamp etal., 2006) showed that commonly used meso-models such asLokalmodell (German Weather Service, Heret et al., 2006) orREMO (Jacob and Podzun, 1997) often overestimate evapo-transpiration (ET) of vegetated surfaces during dry periods.In the projects RICE and RILPS (Henderson-Sellers, 1996) thecoupling between the soil and atmosphere in SVAT mod-els was analysed. As a result, the very large differences inbasis of the transpiration were detected (Mahfouf et al., 1996).This is, among other things, probably due to the insufficientplant-specific coupling between the soil water content and thephysiological reactions of leaf stomata in the implementedSVAT modules (e.g. for global circulation model ECHAM5(Roeckner et al., 2003) for global climate models CLASS(Verseghy, 1991), CSIRO (Kowalczyk et al., 1991), BUCKET(Manabe, 1969); for meso-scale model LAPS (Mihailovic andRajkovic, 1991) or for ecological model BIOME2 (Haxeltine et

results, for example, in the precipitation forecast being lesscertain. Therefore, meteorological models should include amechanism to couple the soil water with a physiologicallybased plant control of ET for a more realistic simulation ofwater transport from the rooting zone in the soil to the atmo-sphere.

This study presents an approach to describe the above-mentioned coupling mechanism. The coupled vegetationboundary layer model HIRVAC using a dependence of stomatalreaction on soil moisture change in the included gas exchangemodel PSN6 (Falge et al., 1996).

2. Methods

2.1. Model description

HIRVAC (High Resolution Vegetation Atmosphere Coupler) isa 1.5 dimensional planetary boundary layer model (Mix etal., 1994). The basic version was developed in the 1980s, atthe Humboldt University of Berlin (HUB, Mix, 1991). By Mixet al. (1994), Ziemann (1998) and Goldberg (1999), the modelwas extended by a multilayer vegetation and soil modules,adjusted and tested. Goldberg and Bernhofer (2001) appliedthe model to current problems of soil–vegetation–atmosphereinteractions. HIRVAC was coupled with PSN6 (Falge et al.,1996). This is a mechanistic leaf photosynthesis model for C3plants developed by Farquhar et al. (1980).

The model HIRVAC has a resolution of 120 layers betweenthe lower (soil surface) and upper model boundary (typical2 km above the ground) whereas the layer distance increaseswith a geometric progression.

Vegetation is considered by extension of the basic modelequations for momentum, temperature and humidity withadditional source terms (terms with the factor “j”) which aresolved numerically for each canopy model level (Eqs. (1)–(4)).In the model, a one-and-a-half turbulence closure is appliedby using the TKE equation, the mixing length formulation byLajhtman and Zilitinkevic, as well as Kolmogorov’s expression

al., 1996)). This leads to a continuous increase in ET causedby an increase in the atmospheric saturation deficit. Conse-quently this leads to an overestimation of water transportfrom the rooting zone in the soil to the atmosphere which

of the turbulent-transfer coefficient (see also Mix et al., 1994).

∂vx

∂t= f (vy − vgy) + ∂

∂zK

∂vx

∂z− j

{nwcdLADvx

√(vx

2 + vy2)}

(1)

Page 3

n g

wogKhwcsT

ata

avhsHbs

fidrn

F0

e c o l o g i c a l m o d e l l i

∂vy

∂t= f (vx − vgx) + ∂

∂zK

∂vy

∂z− j

{nwcdLADvy

√(vx

2 + vy2)}

(2)

∂�

∂t= ∂

∂zKT

∂�

∂z+ (1 − jnw)

1�cp

∂Blw

∂z+ j

{LAD

rb(Tw − T)

}(3)

∂q

∂t= ∂

∂zKq

∂q

∂z+ j

{LAD

rb + rs(qw − q)

}(4)

here z is the vertical coordinate; vx, vy are the componentsf horizontal wind speed; vgx, vgy are the components ofeostrophic wind speed; � is the potential temperature; K,

T, Kq, are the turbulent-transfer coefficients for momentum,eat, and moisture; f is the coriolis parameter, Blw is the long-ave atmospheric radiation; nw = 0 . . . 1 is the crown cover;

d = 0.1. . .0.3 is the drag coefficient; LAD is the leaf area den-ity; rb, rs are the leaf boundary layer and stomatal resistance;

w, T are the temperature of the vegetation surface and thembient air; qw is the specific saturation humidity at Tw; q ishe specific humidity of the ambient air; j = 1 inside and j = 0bove the canopy.

These terms are included in the first calculation levelsbove the surface (max. 60 layers) and are parameterised withegetation parameters (crown cover, leaf area density, canopyeight, drag coefficient) dependent on the type and verticaltructure of the vegetation. The high vertical resolution ofIRVAC in the first decametres above the ground (60 layersetween 0 and 30 m) permits a very detailed differentiation oftructured vegetation (Fig. 1).

Similar approaches for the coupling of vegetation can beound in the literature, (e.g. Groß, 1993). The additional terms

n the temperature and moisture equations (Eqs. (3) and (4))epend on the plant-specific leaf boundary layer and stomatalesistance and were simulated in HIRVAC using the mecha-istic photosynthesis module PSN6 (Falge et al., 1996). The

ig. 1 – Scheme of the coupled soil–vegetation–boundary layer m.9 m deep soil compartment.

2 1 4 ( 2 0 0 8 ) 75–82 77

stomatal resistance is derived from the stomatal conductancegs (5).

gS = Fma

Fmacrit× gfac × 100 × (NP + fvc × RD) × RH

C+ gmin (5)

Fma is a reduction parameter at the field capacity depen-dent on volumetric soil moisture, Fmacrit is the correspondingparameter, gfac (dimensionless) reflects apparent sensitivity ofstomata to changes in NP, RH, and C. NP is the net CO2 fixationrate, fvc is a parameter between 0.5 and 1, RD is the rate of darkrespiration in the light, RH is relative humidity, C is the CO2

partial pressure at the leaf surface and gmin is the cuticularconductance with closed stomata.

The data exchange between PSN6 and the boundary layermodel at each calculation time step and for each model layerwithin the canopy realises a permanent coupling betweenatmosphere and vegetation (left side of Fig. 1, see also Goldbergand Bernhofer, 2001). At present there are PSN6 parametersfor spruce (Picea abies), pine (Pinus sylvestris), beech (Fagussylvatica), rape (Brassica), cereal crop (Triticale) and grass.Radiation transfer in HIRVAC is calculated according to Beer’slaw and optionally uses a random function to include sun-fleck dynamics. To include precipitation, interception and soilmoisture distribution multilayer models for interception andsoil water (based on the hydrological model BROOK90) wereapplied (Federer, 1995; Baums et al., 2005).

The model was tested and the results were compared withresults from measurements and SVAT models for differentobservation periods (Falge et al., 2005).

2.2. New parameterisation of the model

In the new HIRVAC version several parameters of the modelPSN6 were scaled by a power function of soil moisture (matrix

odel HIRVAC for an example canopy (30 m spruce) and a

Page 4

l i n g 2 1 4 ( 2 0 0 8 ) 75–82

Fig. 2 – Aerial view of the Anchor Station Thread, fluxtower in the spruce forest (A), additional climate station in

of the year (DOY) 213 and 230) combined with higher cloud

78 e c o l o g i c a l m o d e l

potential) to consider the stomatal reaction to a change in soilwater content (according to the STANDFLUX stand model ofthe University of Bayreuth, Falge et al., 1997; Falge, personalcommunication):

gfac = gfac,c101.294(�−�FC) (6)

fvc = fvc,c101.294(�−�FC) (7)

c = cc101.294(�−�FC) (8)

where � and �FC (<0) are the actual matrix potential and thematrix potential in the effective root zone at field capacity(unit: mega Pascal). gfac is the linear factor of the empiricalmodel to describe stomatal conductance dependent on netcarbon dioxide (CO2) fixation rate, rate of dark respirationassumed to continue in the light, relative humidity, CO2 partialpressure at leaf surface and minimal conductance if stomataare closed. fvc is the carboxylase capacity and c is the electrontransport capacity during the photosynthesis process (Falge etal., 1996). The additional index “c” refers to the constant inputparameters before modification.

The negative sign of the matrix potential Eqs. (6)–(8) lead toa reduction of gfac, fvc and c and, applied in the PSN6 model,to a decrease in the stomatal conductance (or an increase instomatal resistance). Therefore the amount of the additionalhumidity source term in Eq. (4) of HIRVAC is reduced. As aresult the humidity profiles (q = q(z)) in the canopy air, thelatent heat flux and the energy balance of each canopy modellayer are changed.

3. Results and discussion

3.1. Model settings and input data

The model was already calibrated und verified in sev-eral studies before and has proven its applicability tosoil–biosphere–atmosphere interactions (see Baums et al.,2005; Goldberg and Bernhofer, 2001). The new parameterisa-tion of PSN6 was tested in HIRVAC for a sunny summer day(30.5.2003) using vegetation parameters for the pine stand ofKerigk forest (experimental site of the German Weather Ser-vice’s Lindenberg meteorological observatory) and for a drysummer period (29th July–18th August 2003, days 210–230 ofthe year) using vegetation parameters for the spruce stand atthe experimental site of the Department of Meteorology of TUDresden in the Tharandter Wald forest (Fig. 2). Regarding themain wind direction at the experimental site is the forest aquite homogeneous area (in which large forest resources arelocated).

The pine stand (Lindenberg) is 52 years old with a heightof 14 m, a leaf area index of 3 and a crown cover of 80%. Thespruce stand (Tharandter Wald) is approximately 125 years oldwith a height of 28 m, a leaf area index of 6.5 and a crown coverof 70%.

These stations were one part of the VERTIKO networkproject (2001–2004, Bernhofer and Kostner, 2006) which aimedto determine (measurement and model) vertical turbulentheat and trace gas fluxes over typical Central European land

an open area (B) and the main wind direction.

uses. As a result a data bank with a consistent time series ofall interesting meteorological quantities was provided.

The HIRVAC model was driven by measured solar andatmospheric radiation, precipitation and soil temperature.The time step was 10 min and the integration time was 48 h(first simulated day was used to minimise initial and boundaryeffects for the following day of interest). The sensible (H) andlatent heat flux (LE) were calculated using flux gradient rela-tionship (Eqs. (9) and (10)) (after Oke, 1987), and additionally LEwas calculated for the Lindenberg site using a bulk approach(Eq. (11)) (Bailey and Davies, 1981) and from the direct outputof PSN6 (transpiration of single leafs scaled up to the canopy)for the Tharandt site.

HFRGS = −�cPKT�ref − �crown

zref − zcrown(9)

LEFRGS = −�LKqqref − qcrown

zref − zcrown(10)

LEBULK = �Lqs,crown − qref

rs + rb(11)

where zref, �ref, qref are the height, potential temperature andspecific humidity of the reference layer above the canopy,zcrown, �crown, qcrown are the respective quantities for thecrown maximum area, qs,crown is the specific saturationhumidity at maximum biomass concentration, and rs and rb

are the stomatal and boundary layer resistance of the samecanopy level as for qs,crown.

Fig. 3 gives an overview of the input data used for the HIR-VAC simulations in the summer period of 2003. This periodwas characterised by mostly fair weather with several shortrain events (on 1st and 18th July with the corresponding days

cover and lower radiation input. The intensity and duration ofprecipitation were not high enough (maximum 2 mm, excepton 18th August at 10 mm) to interrupt the long dry periodsignificantly.

Page 5

e c o l o g i c a l m o d e l l i n g

Fig. 3 – Input data (solar radiation (I(s) and precipitation (P))from the Anchor station Tharandter Wald for the period2

3

Ift

cu

mFcT

Fas

9th of July–18th of August 2003, DOY 210–230.

.2. Model results and discussion

n the first case the new parameterisation was applied for aair and sunny day (30th May 2003) at the Lindenberg site toest the principal functionality of the modified approach.

The results clearly illustrate the strong influence of thehanged soil moisture regime on the turbulent heat fluxessing the modified calculation of the PSN6 parameters.

Beginning with the second simulation for moderate soil

oisture conditions (� around −0.13 MPa for sandy loam,

ig. 4d) the latent heat flux (Fig. 4a and b) decreases, espe-ially in the afternoon, while the sensible heat flux increases.he measured values for the latent heat flux (eddy covari-

ig. 4 – Daily course (30th of May 2003) of latent heat flux (LE), capproach (Bulk) (b) and of sensible heat flux (H) (c) above the cantand near Lindenberg (height 14 m, leaf area index 3, crown cov

2 1 4 ( 2 0 0 8 ) 75–82 79

ance) are reached with drier soil conditions (simulation 3),especially if the bulk approach was used. The best agreementbetween the simulated and measured sensible heat fluxes(eddy covariance) (Fig. 4c) could be found for dry and verydry soil conditions (simulations 3 and 4) during the day hourswhile the night values are underestimated in the model. Ithas to be mentioned that the “measured” latent heat flux isthe residuum of the other energy balance components. So thesimulated LE values (assuming energy balance closure in themodel HIRVAC) should be not far from the measurements. Onthe day in the example and for very dry soil conditions thedifferences between measured and simulated LE are relativelysmall if the coupled soil moisture–stomatal reaction approachis used. Possible reasons for deviations are an inexact parame-terisation of the stand structure as well as differences betweensimulated and measured radiation balance components. Thelatent heat flux calculated with the bulk approach shows amore distinct diurnal reaction under dry soil conditions, pos-sibly due to the straighter coupling with the atmospheric vpdand the PSN6 output rs and hence to the modified parame-terisation of gfac, fvc and c (Eqs. (1)–(3)). The afternoon courseof LE using the bulk approach seems not to be realistic com-pared with measurements, probably caused by an insufficientconsideration of turbulent diffusion between the model layersin that case. The simulation using the original version leadsto an overestimation of LE and an underestimation of H com-pared with measurements (Fig. 4, legend index 0). In this case

the results are comparable with the simulated fluxes for wetsoil conditions (Fig. 4, legend 1), and there are no appreciabledifferences in the simulated results for different soil moistureranges (results not depicted here).

lculated with flux gradient relationship (FGRS) (a) and bulkopy for different matrix water potential (� )(d) of a pineer 80%).

Page 6

80 e c o l o g i c a l m o d e l l i n g 2 1 4 ( 2 0 0 8 ) 75–82

Fig. 5 – Measured (solid line) and modelled (with FGRS-fluxgradient relation ship), (dashed line) sensible heat flux (H),and measured precipitation (P) of the investigation period.

Fig. 6 – Measured (residuum of energy balance), (solid line)and modelled (direct output from PSN6), (dashed line) latentheat flux (LE), and measured precipitation (P) of the

Fig. 7 – Daily measured (black) and modelled (sparse)evapotranspiration (ET).

Fig. 8 – Comparison between measured (x-axis) and

ues are slightly underestimated by the model perhaps dueto outliers in the measurements. Both measurements andsimulations reflect the increasing water stress. Although themeteorological conditions do not change LE is reduced from

Fig. 9 – Comparison between measured (residuum of

investigation period.

In the second case the new parameterisation was appliedfor a longer dry period in summer 2003. This summer wasone of the warmest and driest of the last 100 years in CentralEurope and was combined with a significant increase in fatal-ities (several thousands in Germany, Luterbacher et al., 2004;Schar et al., 2004).

The selected period from 29th July to 18th August 2003 wascharacterised by rainless weather only interrupted by shortrainfall events with temporarily higher cloudiness.

Fig. 5 shows the daily course of sensible heat flux in theinvestigated period. For a better orientation regarding themeteorological conditions the precipitation was also plotted.The modelled and measured heat fluxes agree quite well forthe diurnal variation as well as for the minimum and maxi-mum. On sunny days the maximum values range between 300and 500 W/m2, on rainy days they are about 200 W/m2. Thelargest differences are found during the night with 50 W/m2

(underestimation of measurements by the model).Figs. 6–10 summarise the most important results of this

investigation.

In Fig. 6 the daily course of latent heat flux is shown. The

diurnal variation of simulated and measured LE values agreesadequately, but several deviations are observed. In the periodfrom 7th to 18th August (DOY 219–230) the measured val-

modelled (y-axis) sensible heat flux (H), (FGRS-flux gradientrelation ship), (half-hour means for irradiation > 0).

energy balance), (x-axis) and modelled (y-axis) latent heatflux (LE), (direct output from PSN6), (half-hour means forirradiation > 0).

Page 7

e c o l o g i c a l m o d e l l i n g

Fig. 10 – Difference of the latent heat flux (LE) betweenmodel runs with moist and dry soil (solid line), comparedwa

22ld

tvsa1li1

flsadiomtwtLeslts

rrmiPtwtb1

ith modelled (moist (dashed line) and dry soil (fat dots))nd measured (triangle) soil water content (�).

00 to 100 W/m2. Only after the rain event on 18th August (DOY30) the latent heat flux increases for a short time. In the fol-owing days ET remains small because the rain water is wellistributed in the extremely dry soil.

Fig. 7 shows the corresponding daily sum of ET of the inves-igation period. At the beginning of this period ET reachedalues between 2 and 3 mm which suggests a normal waterupply from the soil. In the following two weeks, the measurednd simulated ET decreases continuously to amounts betweenand 1.5 mm/day. The permanently high insolation and the

ack of precipitation led to a continuous drying-out of the root-ng zone which is interrupted shortly after the rain event on8th August (DOY 230) where ET temporarily increases.

In Figs. 8 and 9 measured and simulated turbulent heatuxes are compared neglecting the night hours. The mea-urements and model results of H agree well (Fig. 8) withrise in the linear regression of 0.965 and a coefficient of

etermination of 0.63. For the latent heat flux (Fig. 9) a risen the regression line of 1.13 was found. The bad coefficientf determination of 0.24 hints at more uncertainties in theeasured and simulated LE (e.g. the higher fault liability of

he measuring system used and the problematic turbulentater transport in the model boundary layer) in contrast to

he sensible heat flux. The systematic overestimation of lowE and an underestimation of high LE by the model could bexplained by the time lag between the simulated and mea-ured daytime turbulence regime. The modelled boundaryayer turbulent regime reacts faster to the solar irradiationhan the eddies actually measured by the eddy correlationystem.

Fig. 10 illustrates the differences in LE between two modeluns with different soil moisture regimes as the essentialesult of this study. First the model runs with constant soil

oisture (without water stress) and second with an increas-ngly dry soil (the corresponding water content also in Fig. 10).ositive difference values mean a decrease in evaporation forhe dry soil. The figure shows that for the two model runs

ith different soil moisture, differences in LE increase during

he investigated dry period. The differences determined rangeetween 0 W/m2 (29th July; DOY 210) and about 150 W/m2 on7th August (DOY 229).

2 1 4 ( 2 0 0 8 ) 75–82 81

The course of these differences can be explained by thereaction of the vegetation over the day. The increasingly lowerwater content of the soil leads to the stomata closing and thusto a decrease in the transpiration. In combination with highsolar radiation (Fig. 4) and high values of vpd (not depictedhere) this results in an increasing water stress in the periodfrom 8th to 14th August (DOY 220–226) which is describedby the model comparison of the two soil moisture regimes(constantly wet and increasingly dry).

The soil water contents measured show that the modelledvalues agree well over a large range of the period. The devia-tions in the soil water content (29th July and 18th August; DOY210 and 230) are based on the layer view of the soil model.Because of the precipitation there is a fast increase in thesoil moisture, which, however, quickly decreases again dueto evaporation and outflow into other layers. During the mod-elling bypass rivers by the soil surface into lower soil layerswere not considered.

4. Conclusions

Several SVAT models (e.g. SIB2, Sellers et al., 1996; NOAA-LSM, Mitchell, 2002) consider the effect of soil water on ET butthey are usually not able to simulate feedback effects betweenthe vegetated surface and the atmospheric boundary layer(ABL) because of the lack in vertical resolution and a physicaldescription of the ABL.

The results of this study clearly show that the upgradedmodel HIRVAC is able to simulate the coupling between soilmoisture and evapotranspiration in a drying period. Themodified parameterisation of the stomatal resistance in thesubmodel PSN6 reflects the link between soil moisture, atmo-spheric humidity and ET via stomatal control fairly well. Thissimple modification leads to a noticeable improvement inthe simulated canopy air moisture regime. It allows us tomimic the measurement of turbulent fluxes without usingnew “tuning” parameters. In drier soil the reduction in ETranges between 30 and 150 W/m2 in the Tharandter Wald for-est. This leads to a new partitioning of modelled water budgetcomponents of forested catchments and has possible conse-quences for runoff simulations.

The results of this study identify several problems whichmust be examined in further investigations. One problem isthe nocturnal offset of LE of the forest stand using FGRS (fluxgradient relationships) for calculation. Here the values do notfall below 50 W/m2 at night, which might be due to the inac-curate simulation of nocturnal thermal stability. Improvingthis deficiency in model physics is only feasible with a betterparameterisation of the nocturnal turbulence regime insidethe canopy in HIRVAC.

There is a strong link between water and carbon exchangevia the plant stomata. The resulting effect of the new HIRVACparameterisation on the carbon budget will be investigated.The closing of the plant stomata causes a decrease in

CO2 assimilation. The effects of respiration and assimilationshould be visible in measurements of carbon fluxes fromforests and must be used for a better simulation of the carbonbudget in HIRVAC.

Page 8

l i n g

r

Verseghy, D.L., 1991. CLASS—a Canadian land surface scheme for

82 e c o l o g i c a l m o d e l

Acknowledgements

This study is a part of the VERTIKO research project fundedby the German Federal Ministry for Education and Research(BMBF) with grant number 07ATF37. The authors also thankDr. Eva Falge from the Max-Planck Institute for Chemistry inMainz for her assistance during the implementation of thephotosynthesis model and Lindenberg Meteorological Obser-vatory for meteorological and soil data.

e f e r e n c e s

Bailey, W.G., Davies, J.A., 1981. The effect of uncertainty inaerodynamic resistance on evaporation estimates from thecombination model. Bound.-Layer Meteorol. 20, 187–199.

Barataud, F., Moyne, C., Breda, N., Granier, A., 1995. Soil waterdynamics in an oak stand. 2. A model of the soil-root networkcompared with experimental data. Plant Soil 172, 29–43.

Baums, A.-B., Goldberg, V., Bernhofer, C., 2005. Upgrading thecoupled vegetation boundary layer model HIRVAC by new soilwater and interception modules. Met. Z. 14 (2), 211–218.

Bernhofer, Ch., Kostner, B. (Eds.), 2006. VERTIKO: ProjektSummary, 6. Tharandter Klimaprotokolle Dresden Bd., p. 35.

Falge, E., Graber, W., Siegwolf, R., Tenhunen, J.D., 1996. A model ofthe gas exchange response of Picea abies to habitatconditions. Trees 10, 277–287.

Falge, E., Ryel, R.J., Alsheimer, M., Tenhunen, J.D., 1997. Effects ofstand structure and physiology on forest gas exchange: asimulation study for Norway spruce. Trees 11, 436–448.

Falge, E., Reth, S., Bruggemann, B., Butterbach-Bahl, K., Goldberg,V., Oltchev, A., Schaaf, S., Spindler, G., Stiller, B., Queck, R.,Kostner, B., Bernhofer, C., 2005. Comparison of surface energyexchange models with eddy fluxdata in forest and grasslandecosystems of Germany. Ecol. Model. 188, 174–216.

Farquhar, G.D., von Caemmerer, S., Berry, J.A., 1980. Abiochemical model of photosynthetic CO2 assimilation inleaves of C3 species. Planta 149, 78–90.

Federer, C.A., 1995. BROOK90: a simulation model forevaporation, soil water, and stream flow, Version 3.1.computerfreeware and documentation. USDA Forest Service,PO Box 640, Durham, NH 03824.

Goldberg, V., 1999. Zur Regionalisierung des Klimas in denHochlagen des Osterzgebirges unter Berucksichtigung desEinflusses von Waldern. Tharandter Klimaprotokolle, Band 2,194.

Goldberg, V., Bernhofer, Ch., 2001. Quantifying the couplingdegree between land surface and the atmospheric boundarylayer with the coupled vegetation-atmosphere model HIRVAC.Annal. Geophys. 19, 581–587.

Gottschalck, J.C., Gillies, R.R., Carlson, T.N., 2001. The simulationof canopy transpiration under doubled CO2: the evidence andimpact of feedbacks on transpiration in two 1Dsoil–vegetation–atmosphere-transfer models. Agric. For.Meteorol. 106, 1–21.

Groß, G., 1993. Numerical Simulations of Canopy Flows. Springer,Berlin, p. 167.

Haxeltine, A., Prentice, I.C., Creswell, I.D., 1996. A coupled carbonand water flux model to predict vegetation structure. J. Veg.Sci. 7, 651–666.

Henderson-Sellers, A., 1996. Soil moisture: A Critical Focus for

Global Change Studies. Global Planetary Change 13, 3–9.

Heret, C., Tittebrand, A., Berger, F.H., 2006. Latent heat fluxessimulated with a non-hydrostatic weather forecast modelusing actual surface properties from measurements andremote sensing. Bound.-Layer Meteorol. 121, 175–194.

2 1 4 ( 2 0 0 8 ) 75–82

Jacob, D., Podzun, R., 1997. Sensitivity studies with the regionalclimate model REMO. Meteorol. Atmos. Phys. 63, 119–129.

Jones, H.G., 1998. Stomatal control of photosynthesis andtranspiration. J. Exp. Botany 49, 387–396.

Kowalczyk, E.A., Garratt, J.R., Krummell, P.B., 1991. A soil-canopyscheme for use in a numerical model of the atmosphere—1Dstand-alone model. CSIRO Division of Atmospheric ResearchTech Pap 23.

Liang, Z., Zhang, F., Shao, M., Zhang, J., 2002. The relations ofstomatal conductance, water consumption, groth rate to leafwater potential during soil drying and rewatering cycle ofwheat (Triticum aestivum). Bot. Bull. Acad. Sin. 43, 187–192.

Luterbacher, J., Dietrich, D., Xoplaki, E., Grosjean, M., Wanner, H.,2004. European seasonal and annual temperature variability,trends, and extremes since 1500. Science 303 (5663),1499–1503.

Mahfouf, J.-F., Ciret, C., Ducharne, A., Irrannejad, P., Noilhan, J.,Shao, Y., Thornton, P., Xue, Y., Yang, Z.-L., 1996. Analysis oftranspiration results from the RICE and PILPS Workshop.Global Planetary Change 13, 73–88.

Manabe, S., 1969. Climate and the ocean circulation: 1, Theatmospheric circulation and the hydrology of the Earth’ssurface. Mon. Weather Rev. 97, 739–805.

Mengelkamp, H.-T., Beyrich, F., Heinemann, G., Ament, F., Bange,J., Berger, F.H., Bosenberg, J., Foken, T., Hennemuth, B., Heret,C., Huneke, S., Johnsen, K.-P., Kerschgens, M., Kohsiek, W.,Leps, J.-P., Liebethal, C., Lohse, H., Mauder, M., Meijninger,W.M.L., Raasch, S., Simmer, C., Spieß, T., Tittebrand, A.,Uhlenbrook, S., Zittel, P., 2006. Evaporation over aheterogeneous land surface: The EVA GRIPS Project. Bull. Am.Meteorol. Soc. 87, 775–786.

Mihailovic, D.T., Rajkovic, B., 1991. A coupled soil moisture andsurface temperature prediction model. J. Appl. Meteorol. 30,812–822.

Mitchell, K., 2002. The Community NOAH Land Surface model(LSM). User’s Guide Public Release Version 2.5.2., NCEP/EMC,24 pp. (ftp://ftp.emc.ncep.noaa.gov/mmb/gcp/ldas/noahlsm/,06/05/22).

Mix, W., 1991. Grenzschichtmodelle fur die Umweltuberwachung.Meteorol. Rdsch. 44, 43–55.

Mix, W., Goldberg, V., Bernhardt, K.-H., 1994. Numericalexperiments with different approaches under large-areaforest canopy conditions. Meteorol. Zeitschrift (N.F.) 3 (4),187–192.

Oke, T.R., 1987. Boundary Layer Climates, second ed. Routledge,London.

Roeckner, E., Bauml, G., Bonaventura, L., Brokopf, R., Esch, M.,Giorgetta, M., Hagemann, S., Kirchner, I., Kornblueh, L.,Manzini, E., Rhodin, A., Schlese, U., Schulzweida, U.,Tompkins, A., 2003. The atmospheric general circulationmodel ECHAM5. Part I: model description, Tech. Rep. 349,Max-Planck-Inst. fur Meteorol., Hamburg, Germany.

Schar, C., Vidale, P.L., Luthi, D., Frei, C., Haberli, C., Liniger, M.A.,Appenzeller, C., 2004. The role of increasing temperaturevariability in European summer heatwaves. Nature 427 (6972),332–336.

Sellers, P.J., Randall, D.A., Collatz, G.J., Berry, J.A., Field, C.B.,Dazlich, D.A., Zhang, C., Collelo, G.D., Bounoua, L., 1996. Arevised land surface parameterization (SiB2) for atmosphericGCMs. Part I: model formulation. J. Climate 9, 676–705.

Sellin, A., 1998. The dependence of water potential in shoots ofPicea abies on air and soil water status. Ann. Geophys. 16,470–476, Springer-Verlag.

GCMS: I soil model. Int. J. Climatol. 11, 111–133.Ziemann, A., 1998. Numerical simulation of meteorological

quantities in and above forest canopies. Meteorol. Z. 7,120–128.