The effect of land surface changes on Eemian climate

Transcription

1 Clim Dyn (2007) 29: DOI /s x The effect of lnd surfce chnges on Eemin climte Guy Schurgers Æ Uwe Mikoljewicz Æ Mtthis Gröger Æ Ernst Mier-Reimer Æ Miren Vizcíno Æ Arne Winguth Received: 20 Ferury 2006 / Accepted: 29 Jnury 2007 / Pulished online: 9 Mrch 2007 Ó Springer-Verlg 2007 Astrct Trnsient experiments for the Eemin ( 113 ky BP) were performed with complex, coupled erth system model, including tmosphere, ocen, terrestril iosphere nd mrine iogeochemistry. In order to investigte the effect of lnd surfce prmeters (ckground ledo, vegettion nd tree frction nd roughness length) on the simulted chnges during the Eemin, simultions with interctive coupling etween climte nd vegettion were compred with dditionl experiments in which these feedcks were suppressed. The experiments show tht the influence of lnd surfce on climte is minly cused y chnges in the ledo. For the northern hemisphere high ltitudes, lnd surfce ledo is chnged prtilly due to the direct ledo effect of the conversion of grsses into forest, ut the indirect effect of forests on snow ledo ppers to e the mjor fctor influencing the totl sorption of solr rdition. The Western Shr region G. Schurgers (&) U. Mikoljewicz M. Gröger E. Mier-Reimer M. Vizcíno Mx Plnck Institute for Meteorology, Bundesstrsse 53, Hmurg, Germny e-mil: guy.schurgers@nteko.lu.se Present Address: G. Schurgers Deprtment of Physicl Geogrphy nd Ecosystems Anlysis, Lund University, Sölvegtn 12, Lund, Sweden Present Address: M. Vizcíno Deprtment of Geogrphy, University of Cliforni, 531 McCone Hll, Berkeley, CA , USA A. Winguth Deprtment of Atmospheric nd Ocenic Sciences, Center for Climtic Reserch, 1225 W. Dyton St, Mdison, WI 53706, USA experiences lrge chnges in lnd surfce ledo due to the ppernce of vegettion etween nd 120 ky BP. These locl lnd surfce ledo chnges cn e s much s 20%, therey ffecting the locl s well s the glol energy lnce. On glol scle, ltent het loss over lnd increses more thn 10% for ky BP compred to present-dy. 1 Introduction The effect of lnd surfce chnges on climte hs long een excluded from climte model experiments. Growing wreness of the importnce of lnd surfce for climte resulted from numer of studies with chnges in lnd surfce conditions: first to study the sensitivity of model to these chnges, lter to quntify the importnce of relistic lnd surfce chnges for climte. Mny of the studies were performed with pleoclimtic settings, ecuse of the remrkle chnges for the lnd surfce tht hve een recorded for the pst (e.g. s reported y Prentice nd We III 1998). The overview elow riefly illustrtes the richness of types of previous studies, nd tries to group them y the pproch used for studying the lnd surfce effects. Mny studies were performed considering the importnce of tig nd tundr in the high ltitudes, strting with experiments prescriing ledo chnges y Ottermn et l. (1984). In lter studies, deforesttion ws usully prescried for certin regions (e.g. Bonn et l. 1992; Thoms nd Rowntree 1992; Douville nd Royer 1997). Experiments were performed s well for the Mid-Holocene (6 ky BP) prescriing forest expnsion in the north (e.g. Foley et l. 1994; TEMPO 1996), s ws reported from proxy

2 358 G. Schurgers et l.: The effect of lnd surfce chnges on Eemin climte dt for this epoch. These studies usully focus on the enhncement of the climte effect of prescried insoltion chnges due to iogeophysicl feedck from the lnd surfce: expnsion of forests cuses decrese of the lnd surfce ledo, n increse of the mount of sored solr rdition nd therey n increse in temperture. On the one hnd, this increse in temperture cuses decrese of the totl snow covered re nd shortening of the durtion of snow cover on the ground, therey enhncing the ledo effect. On the other hnd, the increse in temperture fvours the growth of forests, which cuses positive feedck s well. Lter, pleoclimte reconstructions of lke levels nd vegettion from proxy dt were used to drive generl circultion models (Coe nd Bonn 1997; Crowley nd Bum 1997; Diffenugh nd Slon 2002). This is step in the direction of relistic modelling of the pst nd the effect of lnd surfce chnges cn e estimted from the difference etween these experiments nd experiments with presentdy lnd surfce, however feedcks etween lnd surfce nd climte re not simulted interctively. To consider these feedcks, simulted vegettion chnges hve een included in pleoclimte simultions nd nlysis (TEMPO 1996). The feedck etween tmosphere nd lnd surfce ws studied in the Shr desert region with n equilirium vegettion model itertively coupled to dynmic tmosphere model for the Mid- Holocene y Clussen (1997), Clussen nd Gyler (1997), Texier et l. (1997), Clussen (1998) nd de Nolet-Ducoudré et l. (2000). The greening of the Shr desert ws reported in mny studies, nd is closely linked to chnges in the surfce ledo. A decrese of surfce ledo nd increse of vegettion cover cuses severl ssocited effects tht ply role in greening of the Shr desert. A decrese of surfce ledo nd thus n increse in the sorption of shortwve rdition cuses n increse in sensile nd ltent het loss to the tmosphere. This dditionl energy in the tmosphere is lnced y n increse in scending ir nd in convection (Chrney 1975; Chrney et l. 1975, 1976). The increse of moist sttic energy of the tmosphere ner the surfce enhnces the circultion pttern, which promotes the trnsport of moisture into the North Africn continent (Elthir 1996; Elthir nd Gong 1996; Clussen 1997; Brconnot et l. 1999). Besides these processes, the presence of vegettion might ply n importnt role in soil moisture ccessiility for evportion nd recycling of precipittion (De Ridder 1998). All these processes result in enhnced precipittion, which is the key limiting fctor for plnt growth in the Shr region. These processes differ sustntilly in the sptil scle of explntion: some focus more on the locl surfce effects, wheres especilly for the moisture trnsport regionl effects re tken into ccount. Lter, the coupling ws performed with more sophisticted models for the vegettion nd lnd surfce. Complex coupled tmosphere vegettion models (or tmosphere ocen vegettion models) were used, which simulte the stte of vegettion directly, nd were usully pplied for time slices of interesting epochs. The Lst Glcil Mximum (21 ky BP) ws studied y Levis et l. (1999) nd Crucifix et l. (2005), the Mid-Holocene ws studied y Gllimore et l. (2005). Some of the studies mentioned ove took the role of decresed CO 2 concentrtion into ccount, e.g. for the Lst Glcil Mximum (Crowley nd Bum 1997; Levis et l. 2000; Crucifix et l. 2005). Longer trnsient simultion of pleoclimtic time periods with erth system models re up to now only performed using intermedite complexity models (EMICs, Clussen et l. 2002), since the use of fully coupled tmosphere ocen generl circultion models is limited y the computtionl resources. Clussen et l. (1999) nd Brovkin et l. (2002) used the CLIMBER-2 model to study the Holocene (from 9 ky BP to present), Clov et l. (2005) used the sme model, now including ice sheets, to study the Eemin nd the glcil inception ( 100 ky BP). The period etween nd 115 ky BP ws studied s well with the MoBidiC model y Crucifix nd Loutre (2002). Performing longer trnsient simultions with tmosphere ocen GCMs hs so fr only een done y ccelerting the insoltion forcing, e.g. s ws done y Lorenz nd Lohmnn (2004). For this study we performed trnsient simultions with complex erth system model, insted of n intermedite complexity model. The erth system model consists of generl circultion models for tmosphere nd ocen, nd models for the terrestril iosphere nd ocen iogeochemistry. The use of generl circultion models enles more comprehensive representtion of wether nd climte thn given y EMICs. By using rther corse resolution, comined with periodiclly-synchronous coupling technique (Susen nd Voss 1996), we were le to perform experiments with unccelerted s well s ccelerted insoltion forcing for longer trnsient periods. The im of this pper is threefold: (1) to descrie the prmetriztion of the iogeophysicl lnd surfce from the stte of the vegettion, (2) to investigte the effect of insoltion chnges on climte, vegettion nd lnd surfce feedcks nd (3) to investigte the underlying mechnisms of these lnd surfce feedcks, nd the importnce of certin lnd surfce prmeters for these. The insoltion effects nd the underlying feedck mechnisms re studied for insoltion chnges during the lst interglcil (Eemin). We chose this time period ecuse it exhiits chnges in the oritl forcing comprle to the Holocene, ut due to lrger eccentricity of the erth s orit, chnges in climte s well s chnges in vegettion nd its feed-

3 G. Schurgers et l.: The effect of lnd surfce chnges on Eemin climte 359 cks re likely to e somewht lrger, especilly during summer. This llows us to study the potentil non-linerities in the climte vegettion system over wider rnge of insoltion thn would hve een possile for the Holocene. In this pper we will focus on the iogeophysicl effects of the vegettion on climte, the iogeochemistry is discussed in Schurgers et l. (2006). 2 Method A coupled erth system model, consisting of tmosphere nd ocen generl circultion models, n ocen iogeochemistry model nd dynmic glol vegettion model, ws used to study the effect of chnges in the lnd surfce on the climte of the Eemin. The coupled tmosphere ocen generl circultion model ECHAM3-LSG, s used in Mikoljewicz nd Voss (2000) nd Voss nd Mikoljewicz (2001), ws improved nd expnded with the dynmic glol vegettion model LPJ (Sitch et l. 2003) nd the ocen iogeochemistry model HAMOCC (Mier-Reimer 1993). The erth system model used here is the sme s in Winguth et l. (2005) nd Mikoljewicz et l. (2007), ut the ice sheets re fixed t their present-dy stte. The CO 2 concentrtion is clculted prognosticlly from the fluxes etween tmosphere, terrestril iosphere nd mrine iogeochemistry. The Lund Potsdm Jen dynmic glol vegettion model (LPJ, Sitch et l. 2003) descries the distriution of plnt functionl types (PFTs) over the erth, llowing more thn one PFT in grid cell. For ech PFT in grid cell the fluxes nd pools of cron re simulted sed on n verge individul of this type. Per PFT, four living iomss pools nd three litter pools re distinguished. Two soil cron pools common to ll PFTs re used per grid cell. Coupling etween the tmosphere nd the terrestril iosphere is performed yerly with monthly dt. The lnd surfce scheme of the ECHAM3 tmospheric generl circultion model (Roeckner et l. 1992) is lrgely unchnged, ut the required input fields re clculted from the simulted vegettion. The vegettion model uses climte prmeters from the tmosphere for its clcultions, the min prmeters re temperture, soil moisture content nd rdition. The lnd surfce prmeters for the feedck to the tmosphere include prmeters influencing oth the rditive (ledo, tree frction) nd the turulent fluxes (vegettion frction, roughness length). A description of the coupling with these prmeters is given elow. By vrying these prmeters etween ctive coupling nd prescried vlues, the mgnitude of their effects on the simulted climte chnge cn e determined. Due to the yerly coupling etween tmosphere nd lnd surfce, lg of 1 yer emerges etween the chnges in climte nd the lnd surfce response to these. However, ecuse the time scles of interest for the lnd surfce processes re much longer for this study, this will not hve sustntil influence on the lnd surfce effects 2.1 Lnd surfce prmetriztion The ledo of the lnd surfce hs lrge impct on the energy lnce. It is determined y the type of vegettion cover, the visiility nd colour of the re soil nd the presence nd smoothness of snow cover. For the clcultion of the ckground ledo (ledo of the surfce without snow cover), the grid cell re is divided over severl su-grid prts: prt tht is covered y lef-covered vegettion, prt tht is covered y lefless vegettion nd prt tht is covered y re soil. The grid-cell ledo is weighted verge of the ledo vlues ssigned to these comprtments: ¼ c v veg;l þðc v;mx c v Þ veg;nl þð1 c v;mx Þ soil : ð1þ In this eqution, c v is the lef-covered vegettion frction, c v,mx is the mximum vegettion frction nd veg,l, veg,nl nd soil re the ledoes for lef-covered vegettion, lefless vegettion nd soil. The lef-covered vegettion ledo is constructed from the ledoes of the plnt functionl types. Ech plnt functionl type ws ssigned n ledo vlue (see Tle 1), nd the vlues re verged ccording to the frction of (lef-covered) vegetted re they cover: veg;l ¼ P P i ðu ic v;i i Þ i ðu ic v;i Þ ; ð2þ in which u i is the phenology stte (etween 0 nd 1, see Sitch et l. 2003), c v,i is the mximum cover frction nd i is the ledo of the plnt functionl type i. The ledo vlues i in Tle 1 were derived from Clussen (1994), Tle 1 Aledo () nd roughness length (z 0 ) ssigned to the plnt functionl types PFT Description i z 0,i 1 Tropicl rod-leved evergreen Tropicl rod-leved ringreen Temperte needle-leved evergreen Temperte rod-leved evergreen Temperte rod-leved summergreen Borel needle-leved evergreen Borel needle-leved summergreen Borel rod-leved summergreen Temperte herceous Tropicl herceous

4 360 G. Schurgers et l.: The effect of lnd surfce chnges on Eemin climte with ledoes for 17 iomes, nd Clussen et l. (1994), with ledoes for 13 simple surfce types. The vlues re in greement with ledo vlues given in Wilson nd Henderson-Sellers (1985), Hgemnn et l. (1999) nd Milly nd Shmkin (2002). A constnt ledo for non-lefcovered re of ll plnt functionl types veg,nl = 0.16 ws used. The re soil ledo is dependent on the soil colour (which is given s input) nd on the soil wter content (Wilson nd Henderson-Sellers 1985). It is clculted s soil ¼ w j;wet þð1 wþ j;dry ; ð3þ in which w is the reltive moisture content of the soil, nd j,wet nd j,dry re the ledoes for wet nd dry soil of soil colour j. The soil colour mp tht is used s input, s well s the ledo vlues for drk, medium nd light soil, were derived from Wilson nd Henderson-Sellers (1985). In ECHAM3, the totl surfce ledo is clculted from ckground ledo nd snow cover ledo. For the clcultion of the snow cover ledo, the tree frction (or forest frction) from LPJ is used. It is clculted s the sum of mximum cover frctions c v,i for ll tree PFTs. Snow cover ledo for forested res lies etween 0.3 nd 0.4, snow cover ledo for unforested res lies etween 0.4 nd 0.8, dependent on surfce temperture (Kukl nd Roinson 1980; DKRZ 1993). The surfce roughness is defined s the height ove the ground tht is otined when the logrithmic wind profile from well ove the cnopy is extrpolted to the point where the wind speed would hve een 0, therey ignoring chnges in the profile next to the surfce. It is used in ECHAM3 to clculte drg coefficients for the fluxes for wter, energy nd momentum (DKRZ 1993). Surfce roughness is constructed from the orogrphy roughness nd the vegettion roughness length (Clussen et l. 1994): qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi z 0 ¼ z 2 0;oro þ z2 0;veg: ð4þ The vegettion roughness is determined y the mount of vegettion nd the type of vegettion. The roughness lengths per plnt functionl type z 0,i were derived from Clussen (1994) nd Clussen et l. (1994), nd re shown in Tle 1. The vegettion roughness length is clculted from the verged drg coefficients (Clussen 1991; Clussen et l. 1994), which uses lending height z : 1 ln 2 z z 0;veg ¼ X i c v;i ln 2 z z 0;i þ 1 c v;mx : ð5þ ln 2 z z 0;soil The lst prt of the eqution clcultes the contriution of re soil to the vegettion roughness length. Soil roughness z 0,soil = m nd the lending height z = 100 m, s tken from Clussen et l. (1994). The orogrphy roughness z 0,oro (Eq. 4) is given s input, nd ws derived from the originl ECHAM3 roughness length field. The mount of vegettion influences evportion. ECHAM3 uses the vegettion frction (the frction of grid cell covered with vegettion) to clculte evportion (DKRZ 1993). This frction is the sum of ll individul PFT covers, corrected with the phenology stte of ech PFT. Vegettion frction, s well s surfce ckground ledo, fetures sesonl cycle, sed on the phenologicl chnges of the vegettion. 2.2 Experiments An experiment ws crried out in which insoltion ws prescried for the time period 113 ky BP (15,000 yers) ccording to Berger (1978). A 1,000-yer spinup run ws performed for this experiment, strting from present-dy conditions nd running with n insoltion forcing ccording to 129 ky BP. In ddition, control run of 10,000 yers with present-dy insoltion ws crried out. Besides these two long runs, set of experiments for the sme period were performed with n ccelerted insoltion forcing, in order to study the influence of the lnd surfce in detil. These experiments were performed with either n interctive or prescried (present-dy) lnd surfce, s well s two prtilly interctive experiments: one in which the lnd surfce ws only interctive for tree frction, nd one in which the lnd surfce ws interctively coupled except for the lnd surfce ckground ledo. In this set of ccelerted experiments, insoltion chnges were ccelerted with fctor 8, so tht the model ws integrted for 1,875 yers. Two dditionl control runs were performed with present-dy insoltion (one with completely interctive lnd surfce, one with completely prescried lnd surfce), with the sme prmeter settings s in the ccelerted experiments. An interctive lnd surfce is denoted with (+), fixed lnd surfce prmeters with ( ) for the experiments, ccelerted experiments (s well s the control runs to these experiments) re preceded y A. The experiments re summrized in Tle 2. Coupling etween the tmosphere nd the other components ws done with periodiclly-synchronous coupling technique (Susen nd Voss 1996). In contrst to the originl version, where the fluxes were kept constnt during the period without interctive clcultion of the tmospheric GCM, n energy lnce model is pplied here to get n pproprite feedck on se surfce temperture nd se ice. Detils re given in Mikoljewicz et l. (2007). For the ccelerted experiments, the mximum length of the synchronous period nd the cutoff vlue for the het flux nomly, which determine the length

5 G. Schurgers et l.: The effect of lnd surfce chnges on Eemin climte 361 Tle 2 Overview of the experiments CTRL + INS + ACTRL + ACTRL AINS + AINS AINS + A AINS F + Control run with present-dy insoltion (10,000 yers) Insoltion experiment (15,000 yers) Control run with present-dy insoltion, with interctive lnd surfce Control run with present-dy insoltion, with prescried (present-dy) lnd surfce Insoltion experiment, with interctive lnd surfce (ccelerted) Insoltion experiment, with prescried lnd surfce (ccelerted) Insoltion experiment, with interctive lnd surfce, ut with prescried ckground ledo (ccelerted) Insoltion experiment, with prescried lnd surfce, ut with interctive tree frction (ccelerted) of the synchronous coupling period (Mikoljewicz et l. 2007), were reduced compred to the unccelerted experiments. This does not ffect the verge climte of the model, ut it might hve n effect on the vriility. For the experiments with prescried lnd surfce, the vegettion model ws still included for nlysis of the vegettion chnges, ut chnges in the vegettion were not llowed to ffect the tmosphere model. The tmosphere nd vegettion model run on T21 grid (roughly ), the ocen nd ocen iogeochemistry on n Arkw E-grid (effectively ). 3 Results 3.1 Lnd surfce prmetriztion The climtologicl lnd surfce prmeters from the control run (CTRL + ) will e compred with the originl ECHAM3 input (s used for the ACTRL nd AINS experiments). The ECHAM3 input fields were minly derived from stellite mesurements, depicting the sitution s influenced y mn. The vegettion nd lnd surfce s modelled from the dynmic vegettion model do not tke humn influence (e.g. deforesttion, lnd use) into ccount. It descries present-potentil vegettion. The simulted vegettion pttern for present-potentil conditions ws descried in Schurgers et l. (2006), nd shows in generl good greement with wht is considered s present-potentil vegettion. The Amzon region is dominted y Svnn due to underestimted precipittion rtes in the tmosphere model, nd the high ltitudes in Europe nd Asi re too cold in the control run compred to oservtions, resulting in n nomlous southwrd shift of the vegettion zones. Vegettion cover, s well s surfce ckground ledo is simulted with n nnul cycle. Figure 1 compres the yerly verge surfce conditions from the control run (vegettion nd tree cover frction nd ckground ledo) with the yerly vlues of the originl ECHAM3 prmetriztion. For oth vegettion nd tree cover frction, model results differ sustntilly from the originl input. For the vegettion cover in Fig. 1, it should e noted tht the originl ECHAM3 model does not consider yerly cycle for the vegettion cover, wheres the newly coupled version does, nd tht vegettion cover is corrected with dependence on soil moisture content in the ECHAM3 prmetriztion, which ws tken into ccount for Fig. 1. The simulted vegettion cover shown here is the mximum cover tht is otined throughout the yer. Lrge differences occur due to the presence of humn influence in the ECHAM3 prmetriztion, which ws not simulted with the model. Although vegettion nd tree cover differ sustntilly, ckground ledo chnges only modertely for most regions with the newly introduced prmetriztion (Fig. 1c). Lrge differences occur in the highest ltitudes, where the originl prmetriztion shows high ledo vlues, wheres the modelled lnd surfce shows prticulrly low ledoes, due to the presence of drk soils. The originl stellite-derived prmetriztion could e ised here due to snow cover influence, which should not e tken into ccount in the ckground ledo, nd the modelled vlues could e too low, ecuse they re sed on re soils, wheres vegettion exists here, which is not simulted y the model. For the rest of the erth, the ledoes mtch quite well. Over tropicl forest, the modelled ledoes re slightly lower thn from the originl prmetriztion, the temperte nd orel regions show good greement, s well s the desert regions in North Afric nd Southern Asi. The control runs with nd without interctive lnd surfce vry only slightly in climte. Glol surfce ir temperture is K for oth control runs (ACTRL nd ACTRL + ), with stndrd devition of 0.27 K for the experiment with fixed lnd surfce (ACTRL ) nd stndrd devition of 0.18 K for the experiment with interctive lnd surfce (ACTRL + ). Locl chnges in nnul temperture etween the control run with interctive lnd surfce (ACTRL + ) nd the control run with fixed lnd surfce (ACTRL ) rech from 2 K for Centrl Asi nd Northwest Afric to +3 K for Ari, minly cused y chnges in the ckground ledo (Fig. 1c). Reltively

6 362 G. Schurgers et l.: The effect of lnd surfce chnges on Eemin climte Fig. 1 Lnd surfce prmeters in ECHAM3 (left) nd verge of the control run (CTRL + ) with the new prmetriztion (right). Vegettion cover (in ECHAM3 corrected for soil wetness, for the CTRL + experiment verge growing seson vegettion cover is shown), tree cover, c surfce ckground ledo (for the CTRL + experiment the yerly verge is shown) c lrge differences re simulted ner Antrctic, however these differences re smller thn two stndrd devitions of the control run. The effect of the newly introduced lnd surfce coupling on precipittion is reltively smll. The min exception is the monsoon over Indi nd Southest Asi, which expnds slightly northwrd, therey extending the lnd surfce re tht is ffected. A slight decrese in precipittion is simulted over Northern Afric. 3.2 Climte chnge The totl nnul mount of incoming solr rdition t the top of the tmosphere is only slightly higher under Eemin conditions thn under present-dy conditions, with smll chnge over the period from to 113 ky BP ( PW more thn present, Fig. 2). The lrge chnges tht occur in the climte over this period re cused y chnges in the sptil nd temporl distriution of the rdition (Fig. 2, c). In the following we discuss nomlies of selected time slices s well s time series from the insoltion experiments to the respective control run. For the erly Eemin, the coupled experiment (INS + ), s well s the ccelerted coupled experiment (AINS + ), simulte higher (more thn 1 K) nnul men surfce tempertures on the northern hemisphere. From pproximtely 125 ky BP till 115 ky BP grdul cooling of the northern hemisphere erth surfce of nerly 1.5 K is simulted (Fig. 3). The southern hemisphere shows very moderte wrming trend during the first 5,000 yers of the simultion. During the lst 5,000 yers of the insoltion experiments moderte cooling trend is ovious. The vriility here is much higher thn in the northern hemisphere for oth the insoltion runs nd the control run (stndrd devitions for the control run CTRL + re 0.15 K for the northern hemisphere nd 0.42 K for the southern hemisphere). This high century scle vriility is cused y vritions in the convection in the Weddell Se nd Ross

7 G. Schurgers et l.: The effect of lnd surfce chnges on Eemin climte 363 insoltion (PW) time (ky B.P.) 118 temperture nomly (K) time (ky B.P.) 118 INS + AINS + AINS - Fig. 3 Averge surfce ir temperture nomlies (K) for the fully coupled insoltion experiment (INS + ), nd the ccelerted insoltion experiments with interctive lnd surfce (AINS + ) nd with fixed lnd surfce (AINS ) for the northern (full line) nd southern (dshed line) hemisphere to the climte of the respective control runs (CTRL +, ACTRL + nd ACTRL ). Shown re 0.8 ky running mens c Fig. 2 Totl incoming solr rdition t the top of the tmosphere over the period 113 ky BP. Annul men totl incoming rdition (in red, with present-dy vlue in lue). Anomlies of the ltitudinl distriution of the nnul men solr rdition (W m 2 ) compred to present-dy. c Time evolution of nomlies of the sesonl cycle of incoming solr rdition (W m 2 ) compred to present-dy Se (see Mikoljewicz et l. 2007). For the ccelerted experiments, the vriility seems to e higher in the running mens, especilly in the southern hemisphere, which is result of verging over fewer model yers due to the ccelertion. The higher surfce tempertures shown in the northern hemisphere for the first hlf of the experiment re minly tking plce during summer, in consistence with the insoltion forcing (Fig. 4). Annul men surfce temperture nomlies in the insoltion experiment INS + (compred to the control run CTRL + ) re lrger over the lnd surfce thn over the ocen, with strong positive nomly for the entire northern hemisphere. The northern hemisphere monsoon regions (North Afric, Southest Asi) re remrkle exceptions, with cooling for the period ky BP of up to 5 K during summer (Fig. 4) nd up to 4 K in the nnul men (Fig. 5). The monsoon circultion differs here considerly from the control run due to n enhncement of rising ir over the lnd surfce nd n enhncement of sinking ir over the ocen compred to present, resulting in huge precipittion increse for Northwest Afric (Fig. 5). The negtive temperture nomly nd positive precipittion nomly for the eginning of the Eemin compred to the control run re closely relted: incresed cloud cover (Fig. 4) decreses the incoming rdition nd incresed evportion cuses n dditionl ltent het loss. To isolte the trnsient insoltion signl, the empiricl orthogonl functions (EOFs, Preisendorfer 1988) from the INS + experiment were clculted for surfce temperture nd precipittion. Hundred yer men nomlies to the control run CTRL + served s input for the nlysis. The first EOFs of temperture nd precipittion explin ech more thn 40% of the totl vrince in the dt set. The first principl component time series (Fig. 6) of oth temperture nd precipittion show cler insoltion signl, following the decrese of incoming rdition in the northern hemisphere minly cused y chnges in the erth s precession. The mtching EOF pttern for surfce tempertures shows the northern hemisphere wrming with mximl mplitude close to the Arctic ocen nd the strong cooling in the monsoon res of Northwest Afric nd Southest Asi. The first EOF of precipittion descries the dipole pttern over the tropicl North Atlntic with decrese of precipittion in the west nd enhnced precipittion over the Western Shr. The second EOF of precipittion (explining 8.3% of the totl vrince) lrgely descries the non-linerity in the precipittion response. Over the Shr it slightly mplifies the precipittion dur-

8 364 G. Schurgers et l.: The effect of lnd surfce chnges on Eemin climte Fig. 4 Surfce ir temperture nomly (K) nd cloud cover nomly ( ) for summer months (JJA) of ky BP (left) nd ky BP (right) of the coupled experiments INS + reltive to the climte of the control run CTRL + ing the erly Eemin, nd it strongly reduces the effect of the first EOF round 121 ky BP. This ehviour ecomes ovious from the corresponding principl component time series. The other EOFs cpture only nturl vriility of the model, nd do not show ny insoltion signl. For ky BP, the mplitude of the temperture nomlies (reltive to the control run) re smller (Fig. 4). During summer, lower tempertures compred to the control run were simulted over the northern hemisphere lnd surfce up to 4 K, with slightly higher tempertures nd slightly higher cloud cover (Fig. 4) over the Shel zone, which is the opposite effect of tht discussed for ky BP. The CO 2 concentrtion increses during the coupled experiment INS +, from round 270 ppm for the eginning ( ky BP) to round 290 ppm round ky BP. This increse is minly relted to decrese of terrestril cron storge (Schurgers et l. 2006). For the ccelerted experiments, the increse in CO 2 concentrtion is smller, nd the chnges re delyed compred to the unccelerted experiment. However, the cron storge in the ccelerted experiments cnnot e interpreted directly, s the ccelertion cused the insoltion chnges to hppen on time scles comprle to those of deep ocen ventiltion. 3.3 Lnd surfce chnges The min fetures of the chnges in climte in the unccelerted experiment re simulted for the ccelerted experiments with nd without lnd surfce coupling s well, however the mgnitudes of the simulted effects differ. In generl, there is much more similrity etween the two experiments with interctive lnd surfce (INS + nd AINS + ) thn etween the two ccelerted experiment (AINS + nd AINS ). The ltter two experiments show remrkle differences, oth in simulted vegettion nd in simulted climte (Fig. 5). The temperture pttern s shown for the INS + experiment in summer for the period ky BP (Fig. 4) domintes the yerly verge for the two ccelerted experiments s well, ut oth the wrming in the high ltitudes of the northern hemisphere nd the cooling in the monsoon re compred to the control run, s descried ove, re lrger for the experiment with interctive lnd surfce (AINS + ) thn for the experiment with prescried lnd surfce (AINS, Fig. 5). A similr mplifiction, oth of the mgnitude nd of the sptil extent, cn e oserved for precipittion (Fig. 5). Chnges in the surfce ledo re considered to e the min cuse of these mplifictions. Figure 5c shows the ledo chnges compred to the respective control runs. For the AINS experiment, ckground ledo ws prescried, so chnges in surfce ledo re only cused y chnges in the snow cover nd chnges in se ice cover. Over lnd, minor chnges in surfce ledo cn e oserved for the high ltitudes. However, if the chnges in vegettion cover re llowed to influence the ledo, s in the interctive experiment (AINS + ), lrge res with

9 G. Schurgers et l.: The effect of lnd surfce chnges on Eemin climte 365 Fig ky BP nnul nomlies of surfce ir temperture (K), precipittion (mm yer 1 ), c surfce ledo ( ) nd d verge vegettion cover ( ), for the experiment with interctive lnd surfce (INS + CTRL +, left) nd the ccelerted experiments with interctive lnd surfce (AINS + ACTRL +, middle) nd fixed lnd surfce (AINS ACTRL, right) decresed ledo compred to the control run (ACTRL + ) cn e oserved for the high ltitudes, s well s for Northwest Afric. These chnges re relted to chnges in the vegettion cover (Fig. 5d), which themselves re often mplified y the coupling s well. Positive feedcks etween vegettion nd climte cuse this mplifiction of the chnges due to chnges in insoltion in the high ltitudes of the northern hemisphere nd in the monsoon res over North Afric nd Southest Asi. These locl chnges in the lnd ledo, nd to lesser extent chnges in the vegettion cover, cuse chnges in the energy lnce of the erth, not only in the res tht re directly suject to these chnges, ut in other prts of the erth s well, s discussed elow.

10 366 G. Schurgers et l.: The effect of lnd surfce chnges on Eemin climte Fig. 6 Prinicpl components for surfce temperture (left, K) nd precipittion (right, mm yer 1 ), the corresponding EOF ptterns. The nlysis ws performed with 100-yer verges. The climte pttern cn e otined y the multipliction of the vlue of the principl component time series nd the respective EOF pttern, summed over ll EOFs temperture time (ky B.P.) precipittion time (ky B.P.) pc1 (44.2%) projection of control run on EOF1 pc1 (49.3%) pc2 (8.3%) projection of control run on EOF1 projection of control run on EOF High ltitude orel forest nd tundr Over the high ltitudes in the northern hemisphere, orel forests cover lrge res up to the Arctic Ocen in the erly Eemin (Fig. 5d), extending further north thn the orel forests in the control simultions. Over the time spn of the trnsient experiment, they show grdul decrese in cover north of 60 N (Fig. 7), which cn e interpreted s southwrd retret of the treeline for most of the northern hemisphere regions. Until 120 ky BP, temperte forests re present s well in higher ltitudes, ut they show sustntil southwrd retret fterwrds. Grsses sty constnt or increse slightly for some orel res, occupying the re tht is ndoned y orel trees. For ll insoltion experiments, the lnd surfce etween 60 nd 90 N is simulted to e wrmer thn for the control runs etween nd 122 ky BP (Fig. 7c), cusing these chnges in vegettion. For the experiment with fixed lnd surfce (AINS ), mximum difference etween insoltion run nd control run of 1.5 K is simulted, which is due to n increse in incoming shortwve rdition, nd which is only very slightly enhnced y snowcover-induced ledo decrese (Fig. 7). This effect is enhnced in the experiments with interctive lnd surfce (INS + nd AINS + ), ecuse chnges in tree frction (Fig. 7) nd ckground ledo cuse decrese in surfce ledo compred to the control runs (CTRL + nd ACTRL +, Fig. 7). A positive temperture nomly of more thn 3 K ws simulted for the erly Eemin. From 118 ky BP onwrds, the opposite effect cn e oserved, with enhnced ledo increse nd surfce temperture decrese due to lnd surfce chnges. Thus vegettion chnges t high northern ltitudes enhnce the simulted climte chnges y lmost fctor of 2. Two

11 G. Schurgers et l.: The effect of lnd surfce chnges on Eemin climte 367 tree frction (-) ledo (-) ner-surfce ir temperture (K) CTRL + ACTRL explntions re importnt for the decrese of surfce ledo for the period 120 ky BP: (1) the ledo for forest is lower thn for grsses, nd (2) the ledo for snow-covered forest is lower thn for snow-covered tundr, due to more irregulr covering of the surfce. The dditionl experiment AINS F +, in which ckground ledo is prescried ccording to the present-dy sitution (Fig. 1), ut tree frction is used interctively, cn help seprting the two effects. The surfce ledo is clculted from the ckground ledo nd the snow ledo in the tmosphere model, nd tree frction is used to clculte the ltter. In the AINS F + experiment, surfce ledo chnges re cused only y chnges in the tree frction nd thus snow ledo. The ledo of vegettion tht ws not covered y snow ws prescried ccording to the present-dy sitution. The nnul cycle of ledo (Fig. 8) chnges sustntilly etween the control runs ACTRL nd ACTRL + nd the insoltion experiments AINS nd AINS +. The difference etween ky BP in the AINS experiment nd its control run (ACTRL ) is cused solely y chnges in the time (ky B.P.) - ACTRL INS AINS AINS - Fig. 7 Annul men tree frction (), surfce ledo () nd surfce ir temperture (c) for the lnd surfce N (excluding ice sheets). Shown re 0.8 ky running mens from the unccelerted experiments (CTRL + nd INS + ), the control runs (ACTRL nd ACTRL + ) nd the ccelerted insoltion experiments (AINS nd AINS + ). Tree frctions from ACTRL nd AINS re simulted for dignosticl purposes only nd re not influencing the surfce properties s seen y the tmosphere c surfce ledo (-) sored solr rdition (W m -2 ) Jn Fe Mr Apr My Jun Jul Aug Sep Oct Nov Dec 0 Jn Fe Mr Apr My Jun Jul Aug Sep Oct Nov Dec time ACTRL - ACTRL + AINS - AINS + AINS - F + Fig. 8 Sesonl cycle of lnd surfce ledo () nd sored shortwve rdition () y the lnd surfce for the northern hemisphere high ltitudes (60 90 N) for ky BP. Shown re 10-dy running mens snow cover: in spring snow melt occurs erlier nd fster, resulting in lrger snow-free re during summer. This difference is enhnced with the interctive lnd surfce: lnd surfce ledo is oth in summer nd in winter elow the control run vlues for ky BP in the AINS + experiment. The nnul cycle (Fig. 8) for ledo in the AINS F + experiment is similr to the experiment with interctive lnd surfce (AINS + ) during winter nd spring, nd similr to the experiment with fixed lnd surfce (AINS ) during summer. Chnges in winter re lrger thn chnges in summer. However, ecuse the incoming shortwve rdition is much higher during summer, the summer ledo chnges re in generl of more importnce. Figure 8 shows tht the sesonl cycle of sored solr rdition for the INS F + experiment differs mostly from the AINS experiment in spring nd erly summer. The yerly totl sored solr rdition for the lnd surfce etween 60 nd 90 N is J for the AINS, J for the AINS + nd J for the AINS F + experiment. For the control runs this is much lower: J for the ACTRL experiment, nd J for the ACTRL + experiment. This sets the increse in sored shortwve rdition due to tree frction induced snow ledo chnges to two-thirds of the totl yerly increse due to lnd surfce chnges, leving one-third for vegettion ledo chnges nd possile synergetic effects.

12 368 G. Schurgers et l.: The effect of lnd surfce chnges on Eemin climte vegettion cover (-) ledo (-) Fig. 9 Time evolution of the zonl men vegettion cover for Northwest Afric (lnd surfce 10 E 20 W) for the Eemin insoltion experiment (INS + ). Shown re 10-yer running mens Shr desert During the first hlf of the coupled experiment (INS + ), Northwest Afric, including the Shr region, is covered with vegettion (Fig. 9). Mjor chnges in the vegettion pttern occur here during the course of the experiment: The cover of evergreen trees, oth tropicl nd temperte, present in the re etween 5 nd 15 N t the eginning, reduce, s well s the cover of herceous plnts. Between 15 nd 30 N, the cover of herceous plnts reduces clerly during the experiment, nd evergreen trees reduce slightly s well. Around 122 ky BP n lmost rupt trnsition of the Shr vegettion etween 15 nd 30 N tkes plce towrds drier conditions with less vegettion. This retret of the vegettion cuses the Shr region to ecome desert (Fig. 9). Enhnced vegettion cover ccompnied y n increse in precipittion (reltive to the respective control runs), is simulted for the period 120 ky BP in ll insoltion experiments. The dignostic simultion of vegettion for the uncoupled trnsient run (AINS ) shows cler increse of oth vegettion nd precipittion compred to ll control runs (CTRL +, ACTRL nd ACTRL +, Fig. 10). Chnges in the insoltion due to chnges in the erth s orit cuse heting of the lnd surfce in the sutropicl nd temperte regions in summer compred to present, nd therey n incresed temperture grdient etween the lrge lnd msses nd the ocen, therey enhncing the tmospheric moisture trnsport of the monsoons. The positive feedcks etween presence of vegettion, lnd surfce ledo nd precipittion cuse n ovious mplifiction of this effect y pproximtely fctor of 3 for the erly Eemin in the coupled trnsient runs (INS + nd AINS + ). A similr mplifiction ws simulted for Indi, lthough the mgnitude is smller due to smller differences etween vegettion nd soil ledo. precipittion (mm y -1 ) The ptterns of winds t 850 hp nd verticlly integrted tmospheric moisture content of the tmosphere for the summer months for ky BP re shown in Fig. 11 for the monsoon re over North Afric nd Indi. There re cler differences etween the experiment with prescried (AINS ) nd the experiment with interctive (AINS + ) lnd surfce: Over West Afric the southwesterly winds coming from the Atlntic Ocen re stronger in cse of n interctive lnd surfce, therey dvecting more moisture into the Western Shr region. The continentl winds coming from the north re clerly weker. This results in n increse in tmospheric moisture content, especilly over lrge prts of Western Afric. The positive feedck etween vegettion nd precipittion is supposed to e minly driven y chnges in ledo. Chrney (1975) presented mechnism which links decrese in ledo to n increse in rising ir nd in convection nd therey n increse in precipittion, relted to n increse of energy ner the surfce of the tmosphere from enhnced sensile nd ltent het loss of the surfce. According to Elthir (1996), Elthir nd Gong (1996) nd Brconnot et l. (1999) this energy increse drives the in time (ky B.P.) CTRL + ACTRL - AINS + ACTRL + INS + AINS - Fig. 10 Annul men vegettion cover (), surfce ledo () nd precipittion(c) for the Shr region (10 E 20 W, N). Shown re 0.8 ky running mens from the unccelerted experiments (CTRL + nd INS + ), the control runs (ACTRL nd ACTRL + ) nd the ccelerted insoltion experiments (AINS nd AINS + ). Vegettion covers from ACTRL nd AINS re simulted for dignosticl purposes only nd re not influencing the surfce properties s seen y the tmosphere c

13 G. Schurgers et l.: The effect of lnd surfce chnges on Eemin climte 369 c Fig. 11 Averge summer (JJA) winds t pressure level 850 hp (rrows, indicted rrow is 10 m s 1 ) nd verge summer integrted tmospheric moisture content (kg m 2, colours) for the uncoupled control run (ACTRL )() nd for ky BP () with fixed lnd surfce (AINS ) nd with interctive lnd surfce (AINS + )(c). Shown re the Africn nd Indin monsoon regions cresed moisture trnsport from the ocen into Northern Afric, which drives the precipittion increse. It is hrd to distinguish etween them, ecuse oth effects re in principle resulting in n enhncement of the monsoon circultion. Both effects re cused y the chnge in ledo, s shown in Fig. 10, nd result in the modified circultion pttern in Fig. 11. Besides this ledo chnge, the hydrologicl cycle is further mplified y recycling of wter: the presence of vegettion enhnces evportion nd therey precipittion gin. In Northwest Afric, more thn hlf of the precipittion etween 15 nd 25 N tht occurs during the summer monsoon (Fig. 12e) is evported gin. To verify the importnce of the lnd surfce ledo, the dditionl experiment AINS + A ws performed. In this experiment, ckground ledo ws prescried, ut ll other lnd surfce prmeters were llowed to vry with the simulted vegettion chnges. In Fig. 12, the AINS + A experiment is compred with the AINS + nd AINS experiments. The intensifiction of the monsoon for the period round ky BP compred to present is shown clerly, oth the uncoupled nd the coupled experiments show northwrd expnsion of the precipittion pttern in summer, nd in generl higher precipittion rtes. The simulted ptterns in the AINS + A experiment re siclly the sme s in the experiment without interctive lnd surfce (AINS, compre Fig. 12c, d, e), wheres the precipittion in the experiment with fully interctive lnd surfce (AINS + ) expnds much further northwrd. These results confirm tht ckground ledo, the only prmeter tht differs etween these experiments, is the min feedck fctor for the enhncement of Shr rinfll due to dynmic vegettion chnges during the erly Eemin. 3.4 Effects of lnd surfce chnges on the glol energy lnce Figure 13 shows n overview of the glol energy lnce in the control run, nd the nomlies for nd 115 ky BP. As ws discussed ove, the glolly nd nnully integrted incoming rdition chnges only slightly etween control run, nd 115 ky BP, with the nomlies for nd 115 ky BP eing quite similr. However, the chnges in the sptil nd temporl distriution cuse lrger nomlies for other components of the energy lnce. For most udgets the nomlies in the insoltion experiment differ sttisticlly significntly from the control run (Fig. 13). These differences re up to one order of mgnitude lrger thn the chnges in nnully integrted glol incoming rdition, which indictes tht these sptil nd temporl chnges re of gret importnce. For ky BP, totl cloud cover is higher, especilly in the monsoon res of the northern hemisphere (Fig. 4), cusing the solr rdition reflected nd sored y the tmosphere to increse, nd the mount of solr rdition reching the erth surfce to decrese. This effect is stronger over lnd thn over the ocens. Due to lower ledoes of the ocen (cused y decrese of se ice cover) nd the lnd surfce (cused y the greening of the Shr region nd the extension of orel forests in high ltitudes), the mount of solr rdition reflected y the surfce decreses even more thn the decrese of incoming shortwve rdition, especilly for the lnd surfce ( 1.1 W m 2 ). The negtive ledo nomlies for ky BP cuse slight increse of sored shortwve rdition t the surfce for lnd nd ocen, despite the smller

14 370 G. Schurgers et l.: The effect of lnd surfce chnges on Eemin climte Fig. 12 Zonlly verged nnul cycle of precipittion on lnd (mm month 1 ) for the Western Shr region (10 E 20 W), for the control experiments (left) nd for the ccelerted insoltion experiments (right) for ky BP. The top pnels show the experiments with fixed lnd surfce, the ottom pnels show the experiments with fully interctive lnd surfce, nd the middle pnel shows the experiment with interctive lnd surfce ut fixed ckground ledo c d e incoming flux. For ky BP, the outgoing shortwve rdition t the top of the tmosphere shows cler positive nomly in the monsoon re over Southest Asi, Northern Afric nd the Arin Se compred to present, where cloud cover increses (see Fig. 4), nd slight negtive nomly compred to present in the high ltitudes of the northern hemisphere, where the effective ledo decreses due to lnd surfce chnges. The high cloud cover for ky BP cuses s well ig chnges in the longwve rdition t the surfce: downwrd longwve rdition increses y 2.5 W m 2 over lnd nd y 2.0 W m 2 over the ocen, which is prtilly counterlnced y n increse of outgoing longwve rdition due to incresed tempertures of the surfce (+1.4 W m 2 over lnd, +1.7 W m 2 over the ocen). Becuse cloud cover chnges re much smller for 115 ky BP, the nomlies in the energy fluxes mentioned ove re ll much smller. Due to slight cooling for 115 ky BP (Fig. 4), the incoming nd outgoing therml rdition t the surfce decrese slightly for 115 ky BP, with in generl stronger effects for the lnd thn for the ocen, cused y lrger temperture nomlies. For ky BP, the fluxes of ltent nd sensile het show considerle chnges compred to the control run, minly for the lnd surfce. Although the solute chnges re comprle to those of other fluxes, the reltive chnge of especilly ltent het over the lnd surfce is very lrge. Due to higher tempertures, nd n increse of vegettion cover nd soil wetness minly in the Shr region, evportion increses drsticlly, cusing n increse in ltent het loss for the lnd surfce (+1.6 W m 2 or +13.8%) compred to the control run. This is only to smll prt counterlnced y decrese in sensile het loss ( 0.2 W m 2 ). Ltent het loss is the min fctor compensting the imlnce in longwve rdition t the surfce. The min chnges in ltent s well s sensile het loss tke plce in the tropics (Fig. 13), with dominnt role for North Afric. The effects on the incoming rdition t the surfce re in generl much lower for the experiment without lnd surfce coupling (AINS ) thn for the experiment with lnd surfce coupling (AINS + ). Lnd surfce chnges re significnt mplifier of the energy lnce effects due to chnges in insoltion for this time period, even on glol scle. 4 Conclusions nd discussion A prmetriztion for lnd surfce properties ws presented, which descries the chnges in lnd surfce properties in dependence of the simulted vegettion. Trnsient

15 G. Schurgers et l.: The effect of lnd surfce chnges on Eemin climte 371 Fig. 13 Overview of the glol energy lnce for the control run CTRL + nd nomlies for ky BP (upper vlue) nd ky BP (lower vlue) from the INS + experiment to the control run. All fluxes re in W m 2, the reltive chnges re given for nd ky BP s well. Anomlies tht re not significntly different from the control run ( = 0.01) re printed in itlics. Positive numers indicte n enhncement in the direction of the rrow experiments covering the entire lst interglcil were performed with complex erth system model. Comprle experiments were so fr only performed with intermedite complexity models. The newly clculted vegettion nd tree frction differ sustntilly from the originl input s used for the ECHAM3 tmosphere generl circultion model. This is is relted to the difference etween the oserved vegettion s ws used for ECHAM3 nd the present-potentil vegettion s modelled in this erth system model. However, the control run with interctive lnd surfce did not differ widely in ledo vlues nd climte from the control run with prescried ECHAM3 prmeters. Climte chnge during the Eemin ws forced y chnges in the sptil nd temporl ptterns of incoming solr rdition. It cused cler chnges in the lnd surfce cover, especilly in the high ltitudes of the northern hemisphere, nd in the monsoon regions in the sutropics of the northern hemisphere. For the period from to 120 ky BP, the simulted tree frction in the high ltitudes is lrger thn in the control run, with mximum round ky BP nd grdul decrese fterwrds. Due to the positive feedck etween temperture, forest growth nd ledo, the decrese of forest re during the insoltion experiments cuses n increse of the surfce ledo nd enhnces the temperture decrese in the high ltitudes. A comprison etween experiments with fixed lnd surfce nd experiments with interctive lnd surfce shows tht roughly two-thirds of this enhncement re due to the effect of forest presence on snow ledo, nd one-third is due to the difference in ckground ledo etween trees nd grsses nd due to synergetic effects etween these two processes. The enhnced monsoon circultion etween nd 122 ky BP compred to the present sitution over North Afric nd Southest Asi cuses n increse in trnsport of wter vpour to these regions. The enhnced vegettion growth nd increse in soil moisture mplify this effect. The chnges in the surfce ledo re y fr the most importnt prmeter for this positive feedck etween climte nd lnd surfce. Overll, these chnges in the lnd surfce t high northern ltitudes nd in the northern hemisphere monsoon elt (in oth cses relted to chnges in the surfce ledo), hve glol implictions, nd tend to intensify climte chnge s induced y chnges in the erth s orit. This intensifiction ws simulted for key climte fctors, such s temperture nd precipittion. Our simultions show no res with cler negtive feedck on lnd surfce ledo. The mplifiction due to ledo chnges ws simulted for the Eemin efore y Kutzki et l. (2000) nd Crucifix nd Loutre (2002). Together with ledo chnges from ice sheet nd se ice chnges, lnd surfce ledo