This content has been downloaded from IOPscience. Please scroll down to see the full text.

Download details:

IP Address: 138.63.210.14

This content was downloaded on 13/09/2016 at 15:42

Please note that terms and conditions apply.

You may also be interested in:

The European climate under a 2C global warming

Robert Vautard, Andreas Gobiet, Stefan Sobolowski et al.

Disentangling climatic and anthropogenic controls on global terrestrial evapotranspiration trends

Jiafu Mao, Wenting Fu, Xiaoying Shi et al.

Impacts of land use and land cover change on regional climate: a case study in the agro-pastoral

transitional zone of China

Qian Cao, Deyong Yu, Matei Georgescu et al.

Errors and uncertainties introduced by a regional climate model in climate impact assessments:

example of crop yield simulations in West Africa

Johanna Ramarohetra, Benjamin Pohl and Benjamin Sultan

Introduction of a simple-model-based land surface dataset for Europe

Rene Orth and Sonia I Seneviratne

Adapting observationally based metrics of biogeophysical feedbacks from land cover/land use change

to climate modeling

Liang Chen and Paul A Dirmeyer

Is land surface processes representation a possible weak link in current Regional Climate

Models?

View the table of contents for this issue, or go to the journal homepage for more

2016 Environ. Res. Lett. 11 074027

(http://iopscience.iop.org/1748-9326/11/7/074027)

Home Search Collections Journals About Contact us My IOPscience

Environ. Res. Lett. 11 (2016) 074027 doi:10.1088/1748-9326/11/7/074027

LETTER

Is land surface processes representation a possible weak link incurrent Regional Climate Models?

Edouard LDavin1, EricMaisonnave2 and Sonia I Seneviratne1

1 Institute for Atmospheric andClimate Science, ETHZurich, Zurich, Switzerland2 CERFACS,UMR5318-CNRS, Toulouse, France

E-mail: edouard.davin@env.ethz.ch

Keywords: regional climatemodel, land surfacemodel, EURO-CORDEX, land processes, COSMO-CLM2

Supplementarymaterial for this article is available online

AbstractThe representation of land surface processes andfluxes in climatemodels critically affects thesimulation of near-surface climate over land.Herewe present an evaluation of COSMO-CLM2, amodel which couples theCOSMO-CLMRegional ClimateModel to theCommunity LandModel(CLM4.0). CLM4.0 provides amore detailed representation of land processes compared to the nativeland surface scheme inCOSMO-CLM.We performhistorical reanalysis-driven simulations overEuropewithCOSMO-CLM2 following the EURO-CORDEX intercomparison protocol.We thenevaluate simulations performedwithCOSMO-CLM2, the standardCOSMO-CLMandother EURO-CORDEXRCMs against various observational datasets of temperature, precipitation and surfacefluxes. Overall, the results indicate that COSMO-CLM2 outperforms both the standardCOSMO-CLMand the other EURO-CORDEXmodels in simulating sensible, latent and surface radiativefluxesas well as 2-meter temperature across different seasons and regions. The performance improvement isparticularly strong for turbulent fluxes and for dailymaximum temperatures andmoremodest fordailyminimum temperature, suggesting that land surface processes affect daytime evenmore thannighttime conditions. COSMO-CLM2 also alleviates a long-standing issue of overestimation ofinterannual summer temperature variability present inmost EURO-CORDEXRCMs. Finally, weshow that several factors contribute to these improvements, including the representation ofevapotranspiration, radiativefluxes and ground heatflux.Overall, these results demonstrate that landprocesses represent a key area of development to tackle current deficiencies in RCMs.

1. Introduction

The processes occurring at the interface between theland and the atmosphere are involved in climatefeedback mechanisms at various spatial and tem-poral scales and have a direct influence on humansand ecosystems living at this interface (Arnethet al 2010, Seneviratne et al 2010). Conversely, theland surface is continuously transformed by humanactivities through land management and land coverchanges thus exerting a direct anthropogenic forcingon climate. The representation of land surfaceprocesses is therefore a crucial component of climatemodels. Land SurfaceModels (LSMs) used in climatesimulations have been gradually improved over the

last four decades to include more complex and morephysically-based parametrizations (van den Hurket al 2011, Clark et al 2015). While there are sparseindications that the historical development of LSMshas improved the simulated land surface fluxes inoffline mode (mainly illustrated by the better perfor-mance of last generation LSMs compared to the firstgeneration ‘bucket’ model (Best et al 2015, Chenet al 1997)), there is still a lack of formal evidence thatthis has translated into overall better climate modelperformance. This calls for more systematic evalua-tions of LSMs in coupled mode to examine therelationship between the realism of land surfaceprocesses representation and overall climate modelperformance.

OPEN ACCESS

RECEIVED

31March 2016

REVISED

4 June 2016

ACCEPTED FOR PUBLICATION

27 June 2016

PUBLISHED

20 July 2016

Original content from thisworkmay be used underthe terms of the CreativeCommonsAttribution 3.0licence.

Any further distribution ofthis workmustmaintainattribution to theauthor(s) and the title ofthework, journal citationandDOI.

© 2016 IOPPublishing Ltd

Regional Climate Models (RCMs) offer an idealtestbed for investigating this issue. Firstly, becauseRCMs operate in a more constrained ‘space’ asopposed to free-running global models, a direct ‘day-to-day/year-to-year’ comparison with observations ispossible and meaningful. Indeed, using reanalysis aslateral boundary conditions to drive RCMs ensures aconsistency between the simulated and observedsynoptic conditions, while thermodynamic feedbacks(importantly those involving land processes) are stillallowed to respond within this dynamically con-strained system (Giorgi 2006). Secondly, RCMs aretypically run at finer resolution than global models,thus reducing the gap between the resolved scale andthe scale at which land processes actually operate.Thirdly, there is indeed scope for significantly improv-ing land processes representation in current RCMssince they tend to include relatively simple LSMs notreflecting the most recent advances in land surfacemodelling (Davin et al 2011). The heritage of RCMs,which are often based on existing or pre-existingweather forecast models, can at least partly explain thelarger weight given to atmospheric compared to landprocesses development in these models. Against thisbackground, it is legitimate to ask whether the long-standing systematic biases which have been reportedin successive generations of RCM intercomparisons,in particular in the case of Europe (Hagemannet al 2004, Jacob et al 2007, Christensen et al 2007,Kotlarski et al 2014), could be in part due to land pro-cesses representation.

In this study, we evaluate RCM simulations per-formed in the framework of the international inter-comparison project EURO-CORDEX. For one of theEURO-CORDEX models, COSMO-CLM, we addi-tionally perform a simulation in which the standardland surface scheme is replaced by a more advancedLSM. By doing so, the assessed differences between thetwo COSMO-CLM experiments highlight the role ofland process representation. These differences areassessed in the context of the EURO-CORDEX multi-model spread, thus indicating the extent to which landprocesses can impact model performance comparedto other RCMaspects.

2.Methods

2.1. EURO-CORDEXRCMsWe evaluate reanalysis-driven RCM simulations per-formed as part of the EURO-CORDEX project anddownloaded from the Earth System Grid Federation(ESGF) archive. The nine RCMs considered (table 1)provided simulations at 50 km (0.44 degree on arotated grid) spatial resolution on a common analysisdomain encompassing Europe in its entirety. The6-hourly ERA-Interim reanalysis (Dee et al 2011) isused in all models to prescribe lateral boundaryconditions and sea surface temperatures. The longestperiod common to all models (1990-2008) is used forthe analyses.

2.2. COSMO-CLM2

In addition to the aforementioned official EURO-CORDEX simulations we also analyse a simulationperformedwithCOSMO-CLM2. COSMO-CLM2 is analternative configuration of COSMO-CLM featuring adifferent LSM.

COSMO-CLM is a non-hydrostatic regionalatmospheric model jointly developed by the COn-sortium for Small-scale MOdelling (COSMO) and theClimate Limited-area Modelling Community (CLM-Community) and is one of the participating EURO-CORDEX RCMs (table 1). In its standard configura-tion, COSMO-CLM includes TERRA_ML as its LSM.In COSMO-CLM2, however, TERRA_ML is replacedby the more complex Community Land Model(CLM). Earlier versions of COSMO-CLM2 wherebased on CLM3.5 coupled as a sub-routine toCOSMO-CLM (Davin et al 2011, Davin and Senevir-atne 2012). Here we use the more recent versionCLM4.0 (Oleson et al 2010, Lawrence et al 2011) cou-pled to COSMO-CLM via the OASIS3-MCT coupler(Valcke et al 2013).

The main conceptual differences between TER-RA_ML and CLM4.0 concern both biogeophysicaland hydrological processes. Unlike in TERRA_ML, anexplicit canopy layer is considered in CLM4.0, result-ing in specific vegetation temperature and fluxes. Thelinkage between transpiration and photosynthesis isconsidered in CLM4.0 while an empirical formulation

Table 1.EURO-CORDEXRCMs used.

Model Institution LSM

ALADIN5.2 HMS ISBA (Noilhan and Planton 1989,Douville et al 2000)HIRHAM5 DMI (Hagemann 2002)WRF3.3.1 IPSL-INERIS NOAH (Ek et al 2003)RACMO2 KNMI HTESSEL (Balsamo et al 2009)HadRM3P MOHC MOSES (Cox et al 1999)RCA4 SMHI (Samuelsson et al 2006)REMO2009 MPI-CSC (Hagemann 2002, Rechid et al 2009)RegCM4.3 ICTP BATS (Dickinson 1984)COSMO-CLM4.8.17 CLM-Community TERRA_ML (Doms et al 2011)COSMO-CLM2 ETHZurich CLM4.0 (Oleson et al 2010, Lawrence et al 2011)

2

Environ. Res. Lett. 11 (2016) 074027

of stomatal conductance is used in TERRA_ML. Sub-grid scale surface heterogeneity is ignored in TER-RA_ML and is represented using a tile approach inCLM4.0. CLM4.0 additionally considers groundwaterand calculates runoff taking into account sub-gridscale topographic heterogeneity using a TOPMODEL-based approach. A more complete description ofCLM4.0 and its input datasets is provided in Olesonet al (2010). In the present study, CLM4.0 is used in itsbiogeophysics-only configuration without carbon andnitrogen dynamics. The only modification we inclu-ded to CLM4.0 concerns two hydrological parametersinfluencing surface and subsurface runoff (i.e., expo-nential decay factor influencing the saturation excesscomponent of surface runoff and maximum subsur-face drainage). Namely, we reverted to the values usedin CLM3.5 for these parameters (Lawrence et al 2011)since preliminary tests indicated slightly more realisticevapotranspiration rates over Europe for this para-meter choice.

The simulation performed with COSMO-CLM2

follows the EURO-CORDEX protocol and is a sistersimulation of the one performed with COSMO-CLM(table 1). That is, the same model version and modelparameter set are used for the atmospheric comp-onent, the only difference being the LSM used. Indoing so, comparing the COSMO-CLMandCOSMO-CLM2 simulations strictly isolates the effect of landprocesses representation, all else being identical.

2.3. Evaluation datasetsThe various reference products used for evaluation aredescribed in table 2. The selected products covertemperature, precipitation and surface heat and radia-tion fluxes. When possible, different products areconsidered for a given variable in order to account foruncertainties in observation-based datasets. All theproducts are used at a monthly resolution and wereregridded, using bilinear interpolation, to a commonhalf-degree regular grid for comparison with theEURO-CORDEX models. For products not covering

the full 1990-2008 EURO-CORDEX common analysisperiod a shorter time period is used instead.

3. Results

3.1.Overallmodel performanceIn this section, we evaluate overall RCM performanceusing synthetic scores applied over the entire Eur-opean continent. Model performance with respect totemperature, precipitation and surface fluxes isassessed for each of the EURO-CORDEX RCMs(figure 1). As in Davin and Seneviratne (2012), a root-mean-square error (RMSE) is calculated at all grid cellsbased on monthly values over a multi-year period(period depending on reference product, see table 2)thus integrating both temporal and spatial perfor-mance. When possible, different reference productsare used for a given variable, because the choice ofreference product is likely to have amajor influence onthe inferred scores beyond all other methodologicalchoices (Schwalm et al 2013).

Considering first surface fluxes, the coupling withCLM4.0 dramatically improves the performance ofthe standard COSMO-CLM (figure 1(a)). COSMO-CLM2 outperforms not only COSMO-CLM but alsomost other EURO-CORDEX RCMs for both radiativeand turbulent fluxes. In some cases (e.g. for evapo-transpiration) COSMO-CLM2 also outperforms theEURO-CORDEX multi-model mean (MMM). Thisresult highlights the added value of using a moreadvanced LSMcompared to the simpler schemes com-monly used in current RCMs. In line with Best et al(2015), we also note that most models typically havelarger errors for sensible heat flux than for latentheatflux.

As a consequence of the better representation ofsurface fluxes, 2-meter temperature is also bettersimulated in COSMO-CLM2 (figure 1(b)), which out-performs the standard COSMO-CLM as well as mostother RCMs. The improvement is particularly

Table 2.Reference gridded datasets used for evaluation. The time period does not refer to themaximumcoverage but to the time period used in the analysismaximizing the overlapwith the EURO-CORDEXmodels.

Dataset Variables Resolution Time period Reference

CRUTS3.22 2-m temperature 0.5 × 0.5 1990-2008 (Harris et al 2014)precipitation

cloud cover

E-OBS v11 2-m temperature 0.5 × 0.5 1990-2008 (Haylock et al 2008)precipitation

GPCP2.2 precipitation 2.5 × 2.5 1990-2008 (Huffman et al 2009)FLUXNETMTE latent heat 0.5 × 0.5 1990-2008 (Jung et al 2011)

sensible heat

LandFlux-EVAL latent heat 1 × 1 1990-2005 (Mueller et al 2013)SRB3.0 shortwave radiation 1 × 1 1990-2007 (Zhang et al 2015)

longwave radiation

CERES shortwave radiation 1 × 1 2001-2008 (Rutan et al 2015)longwave radiation

3

Environ. Res. Lett. 11 (2016) 074027

substantial for maximum temperature (monthly aver-age of daily maximum temperature) indicating thatthe representation of land processes influences day-timemore than nighttime conditions.While the Diur-nal Temperature Range (DTR) is also improved inCOSMO-CLM2, it is interesting to note that absoluteerrors are typically higher for DTR than for othermetrics (e.g. Tmax) indicating that the representationof the diurnal cycle is a critical remaining deficiency incurrent RCMs.

No improvement, however, is seen for precipita-tion which is even slightly degraded (figure 1(c)). Inview of the multi-model spread this degradationremains relatively minor as COSMO-CLM andCOSMO-CLM2 still cluster together in terms of rank-ing. This degradation for precipitation might seemcounterintuitive given that both surface temperatureand surface fluxes are generally improved in themodel. In this respect, we note that atmospheric para-meters in the model have been tuned in the context of

TERRA_ML and not of CLM4.0. We therefore expectthat a retuning would be necessary to obtain optimalperformances in particular in terms of precipitation.This is, however, not the scope of this study as a differ-ent atmospheric setup would make the attribution ofassessed differences to land processesmore difficult.

For comparison purpose, we also performed thesame multi-variate ranking procedure including inaddition an earlier version of COSMO-CLM2 eval-uated in Davin and Seneviratne (2012) and based onCLM3.5 instead of CLM4.0 (figure S1). The couplingwith CLM3.5 already improves overall model perfor-mance compared to COSMO-CLM, while switchingto CLM4.0 provides further improvements comparedto CLM3.5 in line with global offline evaluation results(Lawrence et al 2011).

The only difference between COSMO-CLM2 andCOSMO-CLM being the LSM used, the overall betterperformance of COSMO-CLM2 can be attributed tothe representation of land processes. This

Figure 1.RMSE-scores (colour) andmodel ranking (numbers) integrating both spatial and temporalmodel performance. RMSEs arecalculated across all land grid points over Europe (-10W30E; 36 N70N) based onmonthly values overmultiple years.MMM:multi-modelmean of EURO-CORDEX excluding COSMO-CLM2.

4

Environ. Res. Lett. 11 (2016) 074027

Figure 2. 2-metre temperature bias (modelminus E-OBS) for COSMO-CLM,COSMO-CLM2 and the EURO-CORDEXMulti-ModelMean (MMM). Indicated in red in the first panel are the two regions used for analysis and defined as in the PRUDENCE project(Christensen et al 2007), the southern region being a combination of two of the original PRUDENCEdomains.

Figure 3. Same asfigure 2 for net shortwave radiation and using SRB as reference product.

5

Environ. Res. Lett. 11 (2016) 074027

interpretation is further supported by the more realis-tic surface fluxes simulated in COSMO-CLM2 asshown in this section. In the next section, we examinemore specifically the nature of themodel biases to bet-ter characterize themechanisms at play.

3.2.Origin ofmodel biasesThe sign and magnitude of model biases is generallynot uniform across Europe and in particular a North-South contrast is visible for temperature (figure 2),radiation (figure 3) and other variables (Supplemen-tary figures). For this reason, we focus our analyses ontwo regions representative of Southern and NorthernEurope (boundaries displayed in figure 2) in additionto the bias maps provided in figures 2, 3 and in theSupplementary Information.

Over Northern Europe, most RCMs under-estimate temperature in particular in spring and sum-mer (figure 4(a)), a tendency mostly affecting daytimetemperatures (figures S2–4). This bias reflects a sys-tematic underestimation of surface shortwave radia-tion (figure 3; figure S6) in turn linked to a tendency ofthe RCMs to overestimate cloud cover (figure S8). Wenote an inter-model correlation between summertemperature and net shortwave radiation of 0.7 overScandinavia, thus confirming that the cause of temp-erature biases in this region is essentially of radiativeorigin.While atmospheric processes are obviously cri-tical for cloud cover biases, land processes may alsoplay a role by indirectly modulating cloud formationthrough energy partitioning at the surface (Davinet al 2011). All the EURO-CORDEX models in factoverestimate summer evaporative fraction over thenorthern half of Europe (figure S11). COSMO-CLM2

alleviates this problem compared to COSMO-CLMwhich implies reduced water input to the atmospherewith beneficial effects on simulated cloud cover andnet shortwave radiation (figure 3), as previously foundin earlier model versions (Davin et al 2011, Davin andSeneviratne 2012).

In contrast, warm biases dominate in summerover Southern Europe (figure 2; figure 4(b)) in con-junction with an overestimation of interannual sum-mer temperature variability (figure 5(c)). Thispersistent deficiency has been reported in previousRCM intercomparisons and attributed to excessivesummer drying (Hagemann et al 2004, Christensenet al 2007, Hirschi et al 2007, Vautard et al 2013,Kotlarski et al 2014). Most EURO-CORDEX RCMsindeed strongly underestimate summer evapo-transpiration over the Mediterranean region(figure 5(a)) and the magnitude of this under-estimation correlates well with temperature biasesacross models (figure 5(c)). Both the evapotranspira-tion bias and the resulting temperature bias are largelyalleviated in COSMO-CLM2 compared to COSMO-CLM and other RCMs confirming that land processesrepresentation plays a major role in this long-standingdeficiency. One possible factor in this improvement isthat CLM4.0, unlike the other models consideredhere, includes a representation of groundwater whichcan limit the excessive summer drying. Anotherhypothesis is that COSMO-CLM, as most otherRCMs, generally overestimates evapotranspirationwhen water is not limited (this is for instance the casein the spring for Southern Europe but this happensalso more generally over Northern Europe as men-tioned previously), thus leading to depleted water con-ditions later in summer. COSMO-CLM2 exhibits amore conservative water use behaviour in the springletting more water available for transpiration in thefollowing summer. This might have a physical cause(e.g. higher aerodynamic resistance) or a physiologicalcause linked to the explicit link between transpirationand photosynthesis represented inCLM4.0.

Another aspect playing a role in seasonal temper-ature variations over southern Europe is the repre-sentation of ground heat flux (GHF). Most modelstend to overestimate the annual temperature range,with too low temperatures in winter and the opposite

Figure 4. Seasonal cycle of 2-metre temperature bias (modelminus E-OBS) for (a) Scandinavia and (b) Southern Europe (combiningtheMediterranean and Iberian Peninsula as defined in the PRUDENCEproject).

6

Environ. Res. Lett. 11 (2016) 074027

in summer (figure 4(b)). This problem is notablyreduced inCOSMO-CLM2, which strikingly also exhi-bits a larger GHF annual amplitude (figure 5(b)). A lar-ger amplitude means that more energy is stored intothe ground during summer and subsequently moreenergy can be released to the atmosphere the followingwinter. This results in a dampened annual cycle oftemperature as the additional energy stored into theground cannot be used to warm near-surface air insummer but can be released in winter and limits thenthe winter cooling. The deeper bottom boundary con-dition (42 m) for thermal calculations in CLM4.0compared to other models (usually not more than 10meters) results in a larger soil volume and heat storagecapacity that can explain the larger annual range inGHF. Supporting this interpretation of the importantrole of GHF, we also find a relatively good inter-modelrelationship between the simulated annual temper-ature range and GHF (figure 5(d)). In other words,models with low GHF annual amplitude tend to over-estimatemore the annual temperature range. Previousstudies already highlighted the importance of placing

the bottom boundary condition for thermal calcula-tions much deeper than 10 metre to adequately repre-sent GHF and soil temperature dynamics overseasonal and decadal time scales (Smerdon and Stie-glitz 2006,MacDougall et al 2008, 2010).

4. Conclusions

Despite decades of improvement, RCMs still sufferfrom large systematic biases. In the case of Europe,these biases have been exposed in successive genera-tions of model intercomparisons (Hagemannet al 2004, Jacob et al 2007, Christensen et al 2007,Kotlarski et al 2014). Here we argue that one of themost promising way forward for reducing these biasesis to tackle deficiencies in modelled land-atmosphereprocesses. Based on an evaluation of reanalysis-drivenRCM simulations from the EURO-CORDEX multi-model ensemble, we show that land processes play acentral role in many long-standing issues affectingRCM performance. By coupling the COSMO-CLM

Figure 5.Mean seasonal cycle of (a) latent heatflux (LH) and (b) ground heatflux (GHF) over Southern Europe. Inter-modelrelationships (c) between summer temperature interannual variability and summer latent heatflux and (d) between annualtemperature range andGHF annual amplitude (d) for Southern Europe. Reference datasets shown in black are LandFlux-EVAL andFLUXNETMTE (a) and E-OBS and FLUXNETMTE (c). GHF is not provided on the ESGF archive and is therefore calculated as theresidual of the surface energy balance implicitly assuming thatmodels conserve energy at the land-atmosphere interface.

7

Environ. Res. Lett. 11 (2016) 074027

RCM to a state-of-the-art LSM we furthermore showthat the model performance in simulating surfacefluxes and climate can be dramatically improved, tothe extent that this coupled system outperforms mostother EURO-CORDEXRCMs for a range of simulatedvariables.

In general, temperature biases over Northern Eur-ope are radiation-driven and land processes are foundto play a role through an indirect mechanism invol-ving turbulent energy partitioning at the surface with asubsequent effect on cloud cover and radiation. Incontrast, temperature biases over Southern Europeinvolve more direct couplings (1) between evapo-transpiration and surface temperature and (2)between GHF and surface temperature. The latteraspect, which has been underappreciated so far, isfound to be important for simulating a realistic ampl-itude of the temperature annual cycle. This calls for alarge-scale synthesis of GHF measurements whichwould enable to better constrain ground heat dynam-ics in themodels. Finally, this study illustrates the ben-efit of taking an extended approach to modelevaluation that includes the full surface energy balanceperspective to help understand the origin of modeldeficiencies and guide futuremodel development.

Acknowledgments

We acknowledge funding from the Swiss NationalScience Foundation (SNSF) through the CarboCount-CH Sinergia Project (grant CRSII2 136273) and fromthe European Union’s Horizon 2020 research andinnovation programme through the CRESCENDOproject (grant agreement No 641816). The computingtime was provided by the Swiss National Supercom-puting Centre (CSCS). We also thank Urs Beyerle forsupport with downloading and storage of EURO-CORDEX data and RichardWartenburger andMartinHirschi for their help with downloading and proces-sing some of the observation-based data. We aregrateful to Wim Thiery, Paul Dirmeyer and ananonymous reviewer for providing useful commentson themanuscript.

References

ArnethA et al 2010Nat. Geosci. 3 525–32BalsamoG,Viterbo P, Beljaars A, van denHurk B,HirschiM,

Betts AK and Scipal K 2009 J. Hydrometeorol. 10 623–43BestM J et al 2015 J. Hydrometeorol. 16 1425–42ChenT et al 1997 J. Clim. 10 1194–215Christensen JH, Carter TR, RummukainenMandAmanatidis G

2007Clim. Change 81 1–6 (Suppl. 1)ClarkMP et al 2015Water Resour. Res. 51 5929–56Cox PM, Betts RA, BuntonCB, Essery R LH, Rowntree PR and

Smith J 1999Clim. Dyn. 15 183–203

Davin E L and Seneviratne S I 2012Biogeosciences 9 1695–707Davin E L, Stoeckli R, Jaeger E B, Levis S and Seneviratne S I 2011

Clim.Dyn. 37 1889–907DeeDP et al 2011Quart. J. Roy.Met. Soc. 137 553–97 (656, part a)DickinsonRE 1984GeophysicalMonograph 29,Maurice Ewing vol 5,

Climate Processes andClimate Sensitivity ed JHansen andTTakahashi (Washington,D.C: AmericanGeophysicalUnion) 58–72

DomsG, Förstner J, Heise E,HerzogH J, RaschendorferM,SchrodinR, Reinhardt T andVogel G 2011Adescription ofthe nonhydrostatic regionalmodel LM: II. PhysicalparameterizationTechnical Report COSMO (http://cosmo-model.org/content/model/documentation/core/cosmoPhysParamtr.pdf)

DouvilleH, Planton S, Royer J, StephensonD, Tyteca S, Kergoat L,Lafont S andBetts R 2000 J. Geophys. Res. 105 14841–61

EkM,Mitchell K, Lin Y, Rogers E, Grunmann P, KorenV,GaynoG andTarpley J 2003 J. Geophys. Res. 108 8851

Giorgi F 2006 J. Phys. IV 139 101–18Hagemann S 2002An improved land surface parameter dataset for

global and regional climatemodelsReport No. 336MPI forMeteorology

Hagemann S,Machenhauer B, Jones R, ChristensenO,DequeM,JacobD andVidale P 2004Clim.Dyn. 23 547–67

Harris I, Jones PD,OsbornT J and ListerDH2014 Int. J. Climatol.34 623–42

HaylockMR,HofstraN, TankAMGK,Klok E J, Jones PD andNewM2008 J. Geophys. Res. 113D20119

HirschiM, Seneviratne S I,Hagemann S and Schaer C 2007J. Geophys. Res. 112D22109

HuffmanG J, Adler R F, BolvinDT andGuG2009Geophys. Res.Lett. 36 L17808

JacobD et al 2007Clim. Change 81 31–52 (Suppl. 1)JungM et al 2011 J. Geophys. Res. 116G00J07Kotlarski S et al 2014Geosci.Model Dev. 7 1297–333LawrenceDM et al 2011 J. Adv.Model. Earth Syst. 3M03001MacDougall AH, BeltramiH, Fidel Gonzalez-Rouco J,

StevensMB andBourlon E 2010 J. Geophys. Res. 115D12109MacDougall AH,Gonzalez-Rouco J F, StevensMB andBeltramiH

2008Geophys. Res. Lett. 35 L13702Mueller B et al 2013Hydrol. Earth Syst. Sci. 17 3707–20Noilhan J and Planton S 1989Mon.Weather Rev. 117 536–49OlesonKW et al 2010Technical description of version 4.0 of the

Community LandModel (CLM)NCARTechnical NoteNCAR/TN-478+STR, National Center for AtmosphericResearch, Boulder, CO, p 257

RechidD, Raddatz T J and JacobD2009Theor. Appl. Climatol. 95245–55

RutanDA,Kato S, DoellingDR, Rose FG,Nguyen LT,Caldwell T E and LoebNG2015 J. Atmos. Ocean. Technol. 321121–43

Samuelsson P, Gollvik S andUllerstig A 2006The land-surfacescheme of the RossbyCentre regional atmospheric climatemodel (RCA3)Rep.Met. 122, 25 SMHI

SchwalmCR,HuntinzgerDN,Michalak AM, Fisher J B,Kimball J S,Mueller B, ZhangK andZhang Y 2013Environ.Res. Lett. 8 024028

Seneviratne S I, Corti T,Davin E L,HirschiM, Jaeger E B, Lehner I,Orlowsky B andTeuling A J 2010Earth-Sci. Rev. 99 125–61

Smerdon J E and StieglitzM2006Geophys. Res. Lett. 33 L14402Valcke S, Craig T andCoquart L 2013OASIS3-MCTUserGuide,

OASIS3-MCT2.0 Technical Report TR/CMGC/13/17CERFACS/CNRS SUCURANo1875Toulouse, France

van denHurk B, BestM,Dirmeyer P, PitmanA, Polcher J andSantanello J 2011Bull. Amer.Meteorol. Soc. 92 1593–600

Vautard R et al 2013Clim.Dyn. 41 2555–75Zhang T, Stackhouse PW Jr., Gupta S K, Cox S J andMikovitz J C

2015 J. Quant. Spectrosc. Radiat. Transf. 150 134–47 (SI)

8

Environ. Res. Lett. 11 (2016) 074027