A general predictive model for estimating monthly ecosystem evapotranspiration

A general predictive model for estimating monthly ecosystem evapotranspiration
of 11
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
  ECOHYDROLOGY  Ecohydrol.  4 , 245–255 (2011)Published online 29 December 2010 in Wiley Online Library( DOI: 10.1002/eco.194 A general predictive model for estimating monthly ecosystemevapotranspiration Ge Sun, 1 * Karrin Alstad, 2 Jiquan Chen, 2 Shiping Chen, 3 Chelcy R. Ford, 4 Guanghui Lin, 3 Chenfeng Liu, 5 Nan Lu, 2 Steven G. McNulty, 1 Haixia Miao, 3 Asko Noormets, 6 James M. Vose, 4 Burkhard Wilske, 2 Melanie Zeppel, 7 Yan Zhang 5 and Zhiqiang Zhang 5 1  Eastern Forest Environmental Threat Assessment Center, USDA Forest Service, Raleigh, NC, USA 2  Department of Environmental Sciences, University of Toledo, Toledo, OH, USA 3  Institute of Botany, Chinese Academy of Sciences, Beijing, China 4 Coweeta Hydrologic Laboratory, USDA Forest Service, Otto, NC, USA 5 College of Soil and Water Conservation, Beijing Forestry University, Beijing, China 6  Department of Forestry and Environmental Resources, North Carolina State University, Raleigh, NC, USA 7  Department of Biology, Macquarie University, Sydney, Australia ABSTRACT Accurately quantifying evapotranspiration (ET) is essential for modelling regional-scale ecosystem water balances. This studyassembled an ET data set estimated from eddy flux and sapflow measurements for 13 ecosystems across a large climatic andmanagement gradient from the United States, China, and Australia. Our objectives were to determine the relationships amongmonthly measured actual ET (ET), calculated FAO-56 grass reference ET (ET o ), measured precipitation ( P ), and leaf areaindex (LAI)—one associated key parameter of ecosystem structure. Results showed that the growing season ET from wetforests was generally higher than ET o  while those from grasslands or woodlands in the arid and semi-arid regions were lowerthan ET o . Second, growing season ET was found to be converged to within š 10% of   P  for most of the ecosystems examined.Therefore, our study suggested that soil water storage in the nongrowing season was important in influencing ET and wateryield during the growing season. Lastly, monthly LAI,  P , and ET o  together explained about 85% of the variability of monthlyET. We concluded that the three variables LAI,  P , and ET o , which were increasingly available from remote sensing productsand weather station networks, could be used for estimating monthly regional ET dynamics with a reasonable accuracy. Suchan empirical model has the potential to project the effects of climate and land management on water resources and carbonsequestration when integrated with ecosystem models. Copyright  ©  2010 John Wiley & Sons, Ltd. KEY WORDS  climate change; ET; eddy flux; modelling; sapflow; water balance  Received 27 January 2010; Accepted 4 December 2010 INTRODUCTIONEvapotranspiration (ET) accounts for over half of thetotal water loss from most terrestrial vegetated ecosys-tems (Zhang  et al ., 2001; Lu  et al ., 2003). For example,in water-limited semi-arid and arid regions, ET can com-prise an even greater percentage of the total water loss(Wang  et al ., 2010) and can equal precipitation. Changesin land use/land cover and climate can also directlyimpact water supply and demand and the regional hydro-logical cycle (DeWalle  et al ., 2000; Jackson  et al ., 2001;Foley  et al ., 2005; Liu  et al ., 2008; Sun  et al ., 2008a) byaltering the ET processes. Although ET is a key vari-able that links hydrological and biological processes inmost ecosystem models (Hanson  et al ., 2004), ET is oneof the most difficult water budget components to quan-tify (Allen, 2008; Shuttleworth, 2008). Worldwide hightemporal scale ET measurements based on soil water *Correspondence to: Ge Sun, Eastern Forest Environmental ThreatAssessment Center, USDA Forest Service, Raleigh, NC, USA.E-mail: ge balance, sapflow, and eddy covariance methods offerednew insights in ecohydrological sciences and helped toadvance our understanding of the ET processes. Sev-eral techniques for quantifying ET exist; for example,the watershed water balance method of precipitation ( P )inputs minus streamflow outputs ( Q ), or ET D  P  Q , istypically limited to long-term average, when the changein water storage component is negligible (Wilson  et al .,2001; Ford  et al ., 2007). At the other temporal extreme,sapflow- and eddy covariance-based ET estimates agreewell with other techniques for uniform stands with largefootprints (i.e. continuous coverage), but are less reliablein complex stands and small or nonuniform footprints(i.e. canopy gaps) (Wullschleger  et al ., 1998; Wilson et al ., 2001; Ewers  et al ., 2002; Law  et al ., 2002; Arain et al ., 2003; Paw U, 2006; Ford  et al ., 2007; Sun  et al .,2008a, 2009, 2010; Barker  et al ., 2009). Eddy covari-ance and sapflow methods have gained popularity forsimultaneously measuring both water and carbon fluxesbecause of their ability to resolve fluxes on a short time-step, offering high temporal resolution. This is largely Copyright  ©  2010 John Wiley & Sons, Ltd.  246  G. SUN  et al . due to performance improvements and reduced costs of fast-response monitoring equipment in recent years. Ageneral predictive model of ET at a monthly scale couldhelp land managers to maximize the ecosystem servicesbecause ET is highly coupled with carbon gain (Law et al ., 2002; Jackson  et al ., 2005; Noormets  et al ., 2006)and other ecosystem services such as biodiversity (Currie,1991).Biophysical modelling has been the most popularapproach for estimating the regional ET using mass-and energy balance theories and empirical relationshipsamong potential ET, precipitation or soil moisture sta-tus, and/or land cover type (Zhang  et al ., 2001; Lu  et al .,2003; Amatya and Trettin, 2007; Zhou  et al ., 2008).Energy and water balances of terrestrial ecosystems aretightly coupled through the ET processes at multiplescales. The long-term ET for a large area is mainlycontrolled by water and energy availability and by landsurface characteristics to a minor extent (Milly, 1994;Zhang  et al ., 2001, 2004). Although a comparison of Budyko-type models that describe such energy–waterrelationships is found in Zhang  et al . (2004), quantifyingET of vegetated surfaces at a fine spatial and temporalscale (e.g. watershed, daily, monthly) remains challeng-ing. For example, the process-based Penman-MonteithET model requires several climatic variables that areoften not available, nor can the parameters be derived forlarge areas. Even the widely used FAO-56 grass referenceET (ET o ) method (Allen  et al ., 1994), a simplified ver-sion of the Penman-Monteith equation, needs substantialcorrections to provide ET estimates for certain landscapes(e.g. forests) at a daily or monthly scale (Sun  et al .,2010). Generally, because  in situ  ET measurements arerarely available at the watershed scale, most hydrologi-cal models are validated with run-off rates measured atthe watershed outlets only; thus, those models have largeuncertainties in describing the full hydrological cycle(Sun  et al ., 2008b). However, tree sapflow and eddy fluxmeasurements from many types of ecosystems around theglobe offer an opportunity to derive ET and water bal-ance models at a higher temporal resolution than werepreviously possible.Our overall goal in this study was to develop a sim-ple monthly ET model that can be used for regionalapplications in modelling ecosystem services (i.e. predict-ing water yield, carbon sequestration, and biodiversity).Our hypothesis was that monthly ET could be estimatedfrom three environmental controls that include availableenergy (i.e. ET o ), water (i.e. precipitation,  P ), and sea-sonal vegetation dynamics (i.e. leaf area index, LAI).We assembled data from ten United States-China CarbonConsortium (Sun  et al ., 2009) sites and three forestedsites with intensive sapflow measurements in the UnitedStates and Australia. Our specific objectives of this syn-thesis study were to: (1) contrast monthly ET and envi-ronmental controls ( P , LAI, and ET o ) among the 13 sites,and (2) develop an empirical monthly ET model that canbe readily used to estimate ET at the site or over a largeregion.METHODS  Monthly ET, P, and LAI  We assembled a database from 13 research sites that rep-resent a range of biomes. Sites span a large climatic gradi-ent, ranging from subtropical rain forests (CWWP) in thehumid Appalachians in the southeastern United States,to the hot dry woodlands in eastern Australia (AUWS,AUPA), and from forested wetlands (NCLP, NCCC) onthe Atlantic coastal plain in the southeastern United Statesto the grasslands (DLSP, XLDS, XLFC) and shrub lands(KBSB) and cultivated croplands (DLCP) in the semi-arid Inner Mongolia region in northern China (Figure 1;Table I). Management practices also vary widely. Thedata set includes two loblolly pine plantations (NCLP,NCCC) on a drained wetland landscape and two poplarplantation sites (BJPL, KBPL) that were subject to brief irrigation during the growing seasons. For the same grass-land ecosystem type, the data set consists of an ecosystemthat was under annual grazing (XLDS) and one underprotection (XLFC) from human disturbances (i.e. fenced,no grazing). The geographic range of the sites varies inlatitude from 43 Ð 5 ° N to 33 Ð 7 ° S and in longitude from83 Ð 8 ° W to 150 Ð 8 ° E. The annual mean air temperatureranges from 0 Ð 6 to 17 Ð 6 ° C and mean annual precipitationfrom 300 to over 1800 mm year  1 . Details of the physi-cal characteristics, site codes, research methods, and keyreferences that have published the ET data for each siteare listed in Table I.Monthly total ET from each site was scaled fromhalf-hour measurements using either the standard eddycovariance methods or the sapflow and interceptionmethods (Table I). Although most of the ET data hadbeen published, ancillary data, such as monthly averagedLAI,  P , and climatic variables, were assembled fromvarious sources.To be consistent, we defined the growing seasonin the northern hemisphere to be May–September andOctober–April in the southern hemisphere. We acknowl-edge that there was no distinct growing season for thetwo Australian forests used here and the tree growth wasgenerally limited to water availability. As some sites didnot have year-round measurements, therefore, this studyfocused on growing season ET when cross-site compar-isons were made. Calculated grass reference evapotranspiration (ET  o ) Potential evapotranspiration (PET) is a nebulous termand can evoke confusion because PET does not clearlyspecify what land surface it refers to. For example,the ‘potential’ amount of water that a forest couldevaporate and transpire would be much higher thana grassland ecosystem could under the same ‘waterunlimited’ conditions due to the larger leaf area of theforest compared to the grassland. Thus, forest PET valuesshould be much higher than grassland PET under thesame climate. When the differences of PET methodsare ignored and a general PET method for grassland orcrops is used for a forest-dominated landscape, serious Copyright  ©  2010 John Wiley & Sons, Ltd.  Ecohydrol.  4 , 245–255 (2011)DOI: 10.1002/eco  A MODEL FOR ESTIMATING ECOSYSTEM EVAPOTRANSPIRATION  247 OakOpening, Toledo,Ohio(OHOO)Beijing Poplar Plantation(BJPL)DuolunSteppe (DLSP) , IMXilinhotGrazed Grassland (XLDS),IMAustralia, Castlereagh,Western Sydney(AUWS)Australia, Paringa, Open woodland (AUPA)White pine,Coweeta,NC(CWWPLoblolly pine(NCLP)Clearcut(NCCC)XilinhotFencedGrassland (XLSP),IMKubuqiShrubland(KBSB),Inner Mongolia (IM)KubuqiPoplar Plantation (KBPL) ,Inner Mongolia (IM)DuolunCroplands (DLCP) , IM Figure 1. Geographic distribution and characteristics of 13 ecosystems across a climatic and management gradient. underestimation of actual forest ET is expected (Sun et al ., 2010). To allay this confusion and normalizethe vegetated land surface to which PET refers to,the term grass reference ET (ET o ) has gradually beenreplacing the PET term as a standard way to represent theenergy conditions for a particular region and makes PETestimates comparable worldwide (Allen  et al ., 1994).Using the process-based Penman-Monteith ET equation,actual daily ET of a hypothetical well-watered grass (i.e.ET o ) that has a 0 Ð 12-m canopy height, a leaf area of 4 Ð 8,a bulk surface resistance of 70 s m  1 , and an albedo of 0 Ð 23 is estimated as follows:ET o  D 0 Ð 408 ⊲R n  G⊳ C ⊲C/⊲T C 273 ⊳⊳u 2 ⊲ e s  e a ⊳ C ⊲ 1 C 0 Ð 34 u 2 ⊳,⊲ 1 ⊳ where ET o  D  grass reference ET (mm)   D  slope of the saturation water vapour pressure atair temperature  T  (kPa ° C  1 )  D 2503e 17 Ð 27 T/⊲T C 237 Ð 3 ⊳ ⊲T C 237 Ð 3 ⊳ 2  R n  D  net radiation (MJ m  2 ); G  D  soil heat flux (MJ m  2 );    D  the psychrometric constant (kPa ° C  1 );e s  D  saturation vapour pressure (kPa);e a  D  actual vapour pressure (kPa); u 2  D  mean wind speed (m s  1 ) at 2 m height; C  D  unit conversion factor with a value of 900.Details of the computation procedures are found in Allen et al . (1994). Monthly ET o  rates were calculated as thesum of daily values in this study.  Empirical ET model development  We pooled all published data of monthly ET,  P , andLAI that were measured onsite using various methods(Table I), and the monthly ET o  estimated by Equation (1)as described above. The observation time length variedfrom one full growing season to 3 years (Table I). Thisdatabase contains 270 records (i.e. 270 site-months).All data analyses were performed using the SAS 9 Ð 2software (SAS Institute Inc., 2008). Regression modelsthat relate ET, ET o ,  P , and LAI for the entire data setwere developed using the SAS’s regression procedure.Different combinations of the independent variables ( P ,LAI, and ET o ) were tested to derive the best fit of observed data. Influences of ET o ,  P , and LAI on ETfor each site were determined by the Pearson correlationcoefficients with significant level at  ˛ D 0 Ð 05.RESULTS  ET  o  , P, and ET in the growing season The 13 sites covered a large range of climatic regimesas indicated by average air temperature and annual totalprecipitation (Table I), resulting in a large difference inecosystem structures (i.e. LAI) and water balance pat-terns. For example, the Coweeta site (CWWP) in the Copyright  ©  2010 John Wiley & Sons, Ltd.  Ecohydrol.  4 , 245–255 (2011)DOI: 10.1002/eco  A MODEL FOR ESTIMATING ECOSYSTEM EVAPOTRANSPIRATION  249     T   a    b    l   e    I .    (     C   o   n    t    i   n   u   e     d     ) .     S    i    t   e    S    i    t   e   c   o    d   e    P    l   a   n    t   c   o   m   m   u   n    i    t   y   a   n    d    d   o   m    i   n   a   n    t   s   p   e   c    i   e   s    L   o   c   a    t    i   o   n   a   n    d   e    l   e   v   a    t    i   o   n    (   m    )    M   e   a   n   a   n   n   u   a    l   a    i   r    t   e   m   p   e   r   a    t   u   r   e    (             °     C    )    A   n   n   u   a    l       P     (   m   m    )    A   n   n   u   a    l    E    T    (   m   m    )    A   n   n   u   a    l    E    T    o     (   m   m    )    S   o    i    l   s    M   a   n   a   g   e   m   e   n    t    O    b   s   e   r   v   a    t    i   o   n   p   e   r    i   o    d    M   e   a   s   u   r   e   m   e   n    t   m   e    t    h   o    d   s    (    S    F ,   s   a   p    fl   o   w   ;    E    F ,   e    d    d   y    fl   u   x    )   a   n    d   r   e    f   e   r   e   n   c   e   s    X    i    l    i   n    h   o    t ,    I   n   n   e   r    M   o   n   g   o    l    i   a ,    C    h    i   n   a    X    L    D    S    D   e   g   r   a    d   e    d    S    t   e   p   p   e   g   r   a   s   s    l   a   n    d   s    (     L   e   y   m   u   s   c     h    i   n   e   n   s    i   s ,    S    t    i   p   a     k   r   y     l   o   v    i    i ,    A   r    t   e   m    i   s    i   a    f   r    i   g    i     d     )    4    3             °     3    3     0     N ,    1    1    6             °     4    0     0     E   ;    1    2     5    0   m    0    Ð     6    1     5    4  –    1    8    4    a  ,    c     3     5    0     b     ¾     1    9    2    c     6    0    8  –    6    9    9    c     S   a   n    d   y   s   o    i    l   s    (    b   u    l    k    d   e   n   s    i    t   y    D     1    Ð     3    3   g   c   m        3     )    U   n    d   e   r    l   o   n   g  -    t   e   r   m   g   r   a   z    i   n   g    2    0    0    6 ,    2    0    0    7   g   r   o   w    i   n   g   s   e   a   s   o   n    (    M   a   y  –    S   e   p    t   e   m    b   e   r    )    E    F    (    M    i   a   o    e    t   a     l  . ,    2    0    0    9    )    X    i    l    i   n    h   o    t ,    I   n   n   e   r    M   o   n   g   o    l    i   a ,    C    h    i   n   a    X    L    F    C    S    t   e   p   p   e   g   r   a   s   s    l   a   n    d   s    (     L   e   y   m   u   s   c     h    i   n   e   n   s    i   s ,    S    t    i   p   a   g   r   a   n     d    i   s ,    A   r    t   e   m    i   s    i   a    f   r    i   g    i     d     )    4    3             °     3    3     0     N ,    1    1    6             °     4    0     0     E   ;    1    2     5    0   m    0    Ð     6    1     5    4  –    1    8    4    a  ,    c     3     5    0     b     ¾     1    9    0    c     6    6    2  –    7    0    8    c     S   a   n    d   y   s   o    i    l   s    (    b   u    l    k    d   e   n   s    i    t   y    D     1    Ð     2    2   g   c   m        3     )    F   e   n   c   e    d    i   n    2    0    0     5   a    f    t   e   r    l   o   n   g  -    t   e   r   m   g   r   a   z    i   n   g    2    0    0    6 ,    2    0    0    7   g   r   o   w    i   n   g   s   e   a   s   o   n    (    M   a   y  –    S   e   p    t   e   m    b   e   r    )    E    F    (    M    i   a   o    e    t   a     l  . ,    2    0    0    9    )    a     P   r   e   c    i   p    i    t   a    t    i   o   n    l    i   s    t   e    d    d   e   n   o    t   e   s    t    h   e   r   a   n   g   e   o   r    t   o    t   a    l    d   u   r    i   n   g    t    h   e   s    t   u    d   y   p   e   r    i   o    d .     b     P   r   e   c    i   p    i    t   a    t    i   o   n    l    i   s    t   e    d    d   e   n   o    t   e   s    t    h   e   r   a   n   g   e   o   r    t   o    t   a    l    d   u   r    i   n   g    t    h   e    l   o   n   g  -    t   e   r   m   m   e   a   n .    c     G   r   o   w    i   n   g   s   e   a   s   o   n    (    M   a   y  –    S   e   p    t   e   m    b   e   r    )   m   e   a   s   u   r   e   m   e   n    t   s   o   n    l   y . southeastern United States had the highest annual pre-cipitation ( > 2000 mm) with a temperate climate, thussupported a plantation conifer forest with the highest LAI(peak LAI  D  7 Ð 1) among all sites examined. In contrast,The Kubuqi shrub (KUSB) and poplar plantation (KUPL)sites in a desert environment of western China’s InnerMongolia had an annual precipitation of   < 300 mm andlow air temperature of 6 Ð 3 ° C (Table I). Thus, those twosites supported plant communities with a low LAI (LAI <  0 Ð 4). The Paringa site on the Liverpool Plain in east-ern Australia had the highest annual ET o  ( ¾ 1070 mm)and moderate annual precipitation ( P D 680 mm) with arather high seasonal and annual variability. A combina-tion of high ET o  and uneven distribution of rainfall mightexplain the periodic water stress that resulted in low LAI(maximum LAI  <  1 Ð 3) for this water-limited ecosystem(Zeppel  et al ., 2006).In addition to the contrasting differences in annualaveraged climate, the 13 sites had contrasting patternsof   P  and ET o  during the growing seasons (Figure 2).The CWWP had the highest precipitation (1153 mm)but the lowest ET o  (482 mm), while the KBPL receivedthe lowest precipitation (228 mm) and the AUPA hadthe highest ET o  (1070 mm) (Figure 2a). Across the 13sites, it appears that the 400 mm precipitation line sep-arated the grassland ecosystems from temperate forestsand water-stressed open woodlands in eastern Australia(Figure 2a). Energy received by the grassland regions onthe Mongolian Plateau and other drier forest sites (Bei- jing and Toledo) were comparable to the forest sites inthe southeastern United States, suggesting that the aridand semi-humid ecosystems were not energy limited forET during the growing season, but rather limited by  P .The total growing season ET was linearly correlatedwith  P  (  R 2 D 0 Ð 96,  p <  0 Ð 001) with a slope of 0 Ð 99, withCWWP being an exception (Figure 2b) to the overallrelationship as a group. Among the 13 sites, except for thewettest (CWWP) and driest (KUPL) sites, ET was within10% of   P  (Figure 2c). Precipitation barely matched theET demand at BJPL, AUWS, AUPA, DLSP, and KUSBand was less than ET at the NCLP, NCCC, OHOO,KUPL, XLDS, XLFC sites during the growing season.In contrast, the CWWP had the lowest ET o , but thehighest ET (Figures 2a–b). The CWWP received 50%more  P  (1153 mm) than needed for ET consumptionduring the growing season. Thus, severe droughts werenot likely for this site, and a perennial stream existedat this relatively wet site (ET o  /  P <  0 Ð 5) (Ford  et al .,2007). Therefore, unlike the other 12 sites that weresomewhat water-limited as indicated by the aridity index(ET o  /  P ), the CWWP was an energy-limited system. Thegroundwater table, an indicator of soil water storage,declined dramatically during the growing season at theNCLP, NCCC (Sun  et al ., 2010), and OHOO sites(Figure 3).  Monthly ET  o  , P, and ET  Monthly ET values varied from less than 10 mm month  1 to as high as 170 mm month  1 , reflecting the biophysical Copyright  ©  2010 John Wiley & Sons, Ltd.  Ecohydrol.  4 , 245–255 (2011)DOI: 10.1002/eco
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks