You are on page 1of 18

This article was downloaded by: [Ingenta Content Distribution Psy Press Titles] On: 7 November 2009 Access

details: Access Details: [subscription number 911796918] Publisher Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

International Journal of Remote Sensing


Publication details, including instructions for authors and subscription information: http://www.informaworld.com/smpp/title~content=t713722504

Remote sensing based estimation of evapotranspiration rates


E. Boegh a; H. Soegaard a a Institute of Geography, University of Copenhagen, 1350 Copenhagen K, Denmark Online Publication Date: 01 July 2004

To cite this Article Boegh, E. and Soegaard, H.(2004)'Remote sensing based estimation of evapotranspiration rates',International

Journal of Remote Sensing,25:13,2535 2551


To link to this Article: DOI: 10.1080/01431160310001647975 URL: http://dx.doi.org/10.1080/01431160310001647975

PLEASE SCROLL DOWN FOR ARTICLE


Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf This article may be used for research, teaching and private study purposes. Any substantial or systematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material.

INT. J. REMOTE SENSING, VOL.

25,

NO.

10 JULY, 2004, 13, 25352551

Remote sensing based estimation of evapotranspiration rates


E. BOEGH* and H. SOEGAARD
Institute of Geography, University of Copenhagen, Oester Voldgade 10, 1350 Copenhagen K, Denmark (Received 19 June 2002; in nal form 18 August 2003 )
Downloaded By: [Ingenta Content Distribution Psy Press Titles] At: 10:42 7 November 2009

Abstract. A remote sensing based method is presented for calculating evapotranspiration rates (lE) using standard meteorological eld data and radiometric surface temperature recorded for bare soil, maize and wheat canopies in Denmark. The estimation of lE is achieved using three equations to solve three unknowns; the atmospheric resistance (rae), the surface resistance (rs) and the vapour pressure at the surface (es) where the latter is assessed using an empirical expression. The method is applicable, without modication, to dense vegetation and moist soil, but for a dry bare soil, where the effective source of water vapour is below the surface, the temperature of the evaporating front (Ts *) can not be represented by the measured surface temperature (Ts). In this case (Ts-Ts is assessed as a linear function of the difference between surface *) temperature and air temperature. The calculated lE is comparable to latent heat uxes recorded by the eddy covariance technique.

1.

Introduction Spatially distributed estimates of evapotranspiration (lE) rates are signicant both for agricultural management planning and climate studies. In agricultural management planning, expressions based on the atmospheric demand are normally used to assess lE, and the physiological control of the vegetation is considered through empirical crop coefcients (i.e. Allen et al. 1998). More precise solutions are usually conned to research experiments where surface uxes and/or leaf stomatal resistance data are recorded. The PenmanMontieth (PM) equation is a well established physically based method for estimating evapotranspiration rate (Monteith and Unsworth 1990). Apart from standard meteorological input data, the PM equation requires information about surface roughness and stability conditions to calculate the atmospheric resistance to heat transfer (ra); and the surface resistance (rs) needs to be estimated. Measuring representative values of rs for vegetation canopies is very time-demanding, while modelling the vegetation surface (stomatal) resistance is difcult because plant physiological processes are controlled both by the physical environment and by the selective behaviour of plant species with respect to optimal functioning (Norman 1993). Empirical models are available for predicting the
*Corresponding author; e-mail: evb@geogr.ku.dk
International Journal of Remote Sensing ISSN 0143-1161 print/ISSN 1366-5901 online # 2004 Taylor & Francis Ltd http://www.tandf.co.uk/journals DOI: 10.1080/01431160310001647975

2536

E. Boegh and H. Soegaard

Downloaded By: [Ingenta Content Distribution Psy Press Titles] At: 10:42 7 November 2009

stomatal resistance based on climatic and soil moisture input data (i.e. Jarvis 1976, Lindroth and Halldin 1986), but they are not generally applicable without prior local determination of the empirical constants. Instead, semi-empirical models based on functional relationships between photosynthesis and stomatal resistance at the leaf scale (Ball et al. 1987, Leuning 1995) may be preferred. In this case, the evapotranspiration/rs-model is coupled to a mechanistic photosynthesis-model (i.e. Collatz et al. 1991, Sellers et al. 1992, Leuning 1995, Boegh et al. 1999, Thorgeirsson and Soegaard 1999, Boegh et al. 2002b, among many others). Even though the process of computing the evapotranspiration rate becomes more complex, the process of tting empirical stomata models to a given location and situation is substituted by functional parameters related to vegetation type. When the surface is only partly covered by vegetation, the estimation of evapotranspiration is further complicated because net radiation needs to be partitioned between soil and vegetation, and a more complex two-layer model structure (i.e. Shuttleworth and Wallace 1985) is required to allow for the exchange of energy between soil and vegetation (i.e. within-row advection). From an operational standpoint, two-layer models are particularly difcult to apply because of the additional need to estimate the surface resistance of the soil. In practice, empirical models relating the soil resistance to soil surface moisture do not give good estimates of hourly and daily evaporation (Daamen and Simmonds 1996), and alternative methods may be needed for estimating soil evaporation rates (Ham and Heilman 1991, Yamanaka et al. 1997, van der Keur et al. 2001). Theoretically, the application of the PM equation (to one or two layers) is not well suited for calculating evaporation rates from dry bare soils because its derivation is based on the assumption that the sources of water vapour and heat are represented at the same temperature, Ts (i.e. Monteith and Unsworth 1990). In reality, the temperature of a dry bare soil surface may be much higher than the temperature of the evaporating front (Ts*) which is located deeper in the soil. Nevertheless, if remote sensing data of Ts are available, estimating the bulk stomatal resistance, the soil surface resistance (for bare soil/sparse vegetation) and the surface humidity (of dry soils) can be avoided. In such cases, Ts is taken to represent the aerodynamic temperature at canopy source height (T0) to enable the calculation of the sensible heat ux (H):  W m{2 H~rcp T0 {Ta ra 1 where rcp is the heat capacity (J m23 C21). The evapotranspiration rate can then be assessed as the residual in the surface energy balance equation (equation (2)) after estimating or measuring net radiation and soil heat ux (i.e. Seguin et al. 1989, Humes et al. 1997). However, two signicant drawbacks of this method are that Ts may exceed T0, even for dense canopies (Choudhury et al. 1986, Huband and Monteith 1986, Kalma and Jupp 1990) and that the atmospheric resistance between the surface and the air (rae) may be larger than the atmospheric resistance between canopy source height and the air above the canopy (ra) (i.e. Stewart et al. 1994). Numerous empirical and theoretical studies have been conducted to determine an excess resistance to heat transfer between the surface and source height (i.e. Kustas et al. 1989, Sugita and Brutsaert 1990, Stewart et al. 1994, Troueau et al. 1997, Verhoef et al. 1997, Massman 1999). The classical approach is to use the quantity kB21 to estimate the excess resistance, rex~kB21/(ku.), where k is the von Karman constant, u. is the friction velocity (m s21), and kB21 is the natural logarithm of the ratio of the roughness length for momentum (z0) and roughness

Estimation of evapotranspiration rates

2537

Downloaded By: [Ingenta Content Distribution Psy Press Titles] At: 10:42 7 November 2009

length for heat (z0h). In remote sensing studies, the kB21 is normally determined empirically. For dense canopies, it is approximately constant (around 2), but for sparse canopies, it is usually higher (Kustas et al. 1989, Stewart et al. 1994, Troueau et al. 1997, Verhoef et al. 1997), and its dependence on both vegetation structure, climate and surface conditions has been demonstrated (Troueau et al. 1997). Very complex models are needed to predict this kB21 dependency (Massman 1999), and currently alternative methods are being developed to estimate lE from Ts where the estimation of the excess resistance to heat transfer is not needed. Instead, simultaneous measurements of Ts from two different view angles can be used to calculate lE (Norman et al. 1995, Kustas and Norman 1997) or, assuming the atmospheric resistance of dry and wet soils to be similar, the surface temperature of a dry soil column can be included for calculating soil evaporation rates (Qiu et al. 1998). Methods to assess the atmospheric resistance were suggested by Bastiaanssen et al. (1998) who used the slope of the Ts-albedo relationship of dry surfaces to assess the atmospheric resistance from Earth observations, and Boegh et al. (2002a) who introduced an empirical equation for the vapour pressure at the surface (es) to facilitate mathematical solutions for rae and rs. Because the method of Boegh et al. (2002a) used the measurement of Ts to assess the surface saturation vapour pressure, this method needed adjustment to assess the surface humidity (and lE) for dry bare soils where the effective source area of water vapour is below the surface. In the present paper, the method of Boegh et al. (2002a) is further elaborated and expanded to include an adjustment function accounting for the difference in temperature between the surface and the evaporating front (Ts-Ts*) of dry soil surfaces. The method requires inputs of surface temperature, net radiation, soil heat ux, air temperature and vapour pressure. Because the latter two variables are standard meteorological data and the remaining inputs can be obtained from remote sensing data, the method could be of particular relevance for monitoring spatially distributed evapotranspiration using satellite data such as Landsat, NOAA-AVHRR or EOS-MODIS. In the present paper, the utility of the methodology is demonstrated using thermal infrared eld data which were recorded for bare, sparsely and densely vegetated soils in Denmark. In Boegh et al. (2004), the method is applied to AVHRR observations and meteorological data which were predicted by a high resolution weather forecast model.

2.

Methodology The evapotranspiration rate (lE) relates to the surface energy balance as lE~Rn {H{G W m{2 2

where Rn is net radiation (W m22) and G is the soil heat ux (W m22). In equation (2), H and lE may be expressed as  W m{2 H~rcp Ts {Ta rae 3  W m{2 lE~ rcp c es {ea =rae 4 where Ts is the surface temperature (C), c is the psychrometer constant (Pa C21), ea is the vapour pressure above the canopy (Pa), es is the vapour pressure at the surface (Pa), and rae is the atmospheric resistance to heat transfer between the surface and the air above the canopy (s m21), which is given by rae~razrex.

2538

E. Boegh and H. Soegaard

Assuming the atmospheric resistance to be the same for temperature and water vapour, an expression for rae is obtained by combining equations (2), (3) and (4) {1 Ts {Ta zes {ea =c rae ~rcp 5 sm Rn {G An equation for the surface resistance s) is obtained on the basis of the following (r    equations; lE~ rcp c e {es rs ~ rcp c es {ea =rae , where the saturated s vapour pressure at the evaporating front (es*) is evaluated at the temperature (Ts*). By rearrangement, rs is given by  {1 rs ~rae e {es es {ea sm 6 s
Downloaded By: [Ingenta Content Distribution Psy Press Titles] At: 10:42 7 November 2009

* For dense canopies and moist soils, Ts equals Ts, but for a dry bare soil, the evaporating front is located beneath a dry surface layer exhibiting different thermal properties, and Ts may be much higher than Ts* in this situation. Because an increase in Ts due to greater solar radiation increases both the soil and the atmospheric sensible heat ux (Garratt 1992, Cellier et al. 1996), the difference * between Ts and Ts may be inferred from the observations of (Ts-Ta). The close coupling of G and H through the surface temperature allows G to be expressed as a fraction of H, i.e. G~aH (Berkowicz and Prahm 1982, Novak 1986). Since  G~ks Ts {Ts zd (ks is thermal conductivity of dry soil and zd is the thickness of the upper dry soil layer), insertion in equation (3) leads to the following expression *) for (Ts-Ts Ts {Ts ~BTs {Ta 0 C 7 where B~arcpzd/(raeks). Equation (7) implies the use of the depth of the evaporating surface instead of the more traditional use of surface-moisture availability for calculating soil-limited evaporation rates. A simple energy-balance model applying zd instead of surface-moisture for calculating lE was also proposed by Yamanaka et al. (1997). In this study, the coefcient B is evaluated using eld data and literature values of zd and ks of the dry soil layer (4.3). To solve rae (equation (5)) and rs (equation (6)), an equation for the third unknown, es, is required. Because es is regulated at the surface by the amount of feedback which is mediated both by the atmospheric boundary layer and the surface response functions, its evaluation should take place within a larger scale environment, such as that represented by the PM equation (Jones 1992; p. 117). By rewriting the PM equation, the vapour pressure decit at the surface is given by (Jarvis and McNaughton 1986): 8 es {es ~V e {es eq z1{V e {ea s a where V is the decoupling coefcient, V~(D/cz1)/(D/cz1zrs/rae), which quanties the degree of atmospheric coupling between the surface and the air; and D is the slope of the saturation vapour pressure versus temperature relationship which is evaluated at (TszTa)/2. With an efcient vertical mixing of the air, rae decreases and Vp0 in which case the atmospheresurface coupling is good. If the surface is completely coupled to the atmosphere (V~0), the vapour pressure decit of the atmosphere (ea*-ea) is imposed at the surface (equation (8)). The surface vapour pressure decit during decoupled conditions, (es*-es)eq, is  solved by e {es eq ~lEeq rs c rcp , where lEeq~D(Rn-G)/(Dzc) is the equilibrium s evapotranspiration rate which is approached during decoupled conditions (Vp1) (Jarvis and McNaughton 1986).

Estimation of evapotranspiration rates

2539

Downloaded By: [Ingenta Content Distribution Psy Press Titles] At: 10:42 7 November 2009

When (Ts-Ta) increases above 510C, the linear approximation of D in the PM equation is invalid (Paw U and Gao 1988). This temperature difference is typically reached for bare dry soils in Denmark. When nonlinear solutions are used for the saturated vapour pressure function instead of the linear approximation, it can be shown (Paw U and Gao 1988) that the limit of lE during decoupled conditions approaches the available energy (Rn-G) which may be very different from the lEeq used to calculate (es*-es)eq in equation (8). The application of equation (8) for evaluation of es is therefore in error when rs/raep0. In addition, it was already discussed that differences between Ts and Ts* may be substantial for bare dry soils which further invalidates the applicability of PM-theory in such situations. In the present study, measurements rather than predictions are used to assess the limit of es at any time. If the surface is decoupled from the atmosphere, the atmospheric vertical transfer of vapour is weak. In this case, water vapour tends to *. accumulate at the surface until, eventually, it approaches es The limit of es during decoupled conditions may therefore be calculated from the Ts measurement, or the Ts* estimate (for a non-saturated surface). On the contrary, if the surface is well coupled to the atmosphere (efcient vertical mixing of air), es is expected to approach ea. In this paper, the feasibility of V as a weighting factor to place es between its (measurement-based) limit values, es and ea, is investigated, i.e. * es ~A V e z1{V ea s Pa 9 Substituting equation (9) in equation (6) it can be seen that the empirical factor A appears to be analytically related to ea/es by *  crae zcrs zDrae zDrea e s A~ 10 crae zcrs zDrae zDrs For ea/es*~1, or rs~0, then A~1. For lower values of ea/es*, Av1. In this paper, a constant value of A (A~0.9) in equation (9) facilitates the calculation of evapotranspiration rates during a wide range of conditions. For this purpose, the hypothetical expression for es (equation (9) using A~0.9) is solved together with equations (5) and (6) using NewtonRaphson iteration (appendix A), thereby allowing the derived rae and rs to be used for computing lE:   lE~ rcp c e {ea rae zrs W m{2 11 s Apart from evaluating the usefulness of the simple empirical expression for es by comparing the ensuing calculations of rae and lE with measurement-based values of rae (rae~rcp(Ts-Ta)/H) and eddy covariance of lE data, the dependency of es on V is also evaluated relative to the results given by a more complex canopy scale vegetationatmosphere transfer (VAT) model. The VAT model applied in this paper is a slightly simplied version of the model described in Boegh et al. (2002b). It provides an iterative solution for surface uxes and surface conditions within a larger scale environment, and it is validated using the eld measurements of lE and Ts. A summary of the VAT model is given in appendix B.

3.

Location and equipment For the present study, data collected from two elds cropped with maize and winter wheat, respectively, are used. The elds are located in an agricultural area in Denmark (9.423 E, 56.486 N) characterized by sandy loam soils, and the measurement campaign took place during the spring and summer of 1998.

2540

E. Boegh and H. Soegaard

Downloaded By: [Ingenta Content Distribution Psy Press Titles] At: 10:42 7 November 2009

Generally, the growth season in 1998 was very humid (gure 1(a)). The winter wheat was sown in Autumn 1997, and the maize was sown in May 1998. During the measurement campaign, the leaf area index ranged from 0 to 4.2 for winter wheat, and from 0 to 4.4 for maize (gure 1(b)). Leaf area index was measured each week using destructive sampling (Boegh et al. 2002b). A continuous record of the eld surface temperature was made in 1998, except for the period 24 July to 6 August for wheat, and in the period 4 August to 9 August for maize. On each eld, thermal infrared radiometers (KT17, Heimann, D; Wiesbaden, Germany) were mounted on masts at 2 m height with azimuth angles of 60 and pointing 45 towards the surface. An emissivity of 0.99 was applied to convert the thermal radiance to surface temperature. Net radiation (Q*6, REBS Inc., USA), soil temperature, air temperature and air humidity (MP300, Rotronic; Bassersdorf, Switzerland) were measured continuously throughout the study period. Sensible and latent heat uxes were measured using the eddy covariance technique (Baldocchi et al. 1988). The winter wheat eld size is 600 m6800 m, and the minimum fetch is 300 m. The maize eld, which is long and narrow, is located in an eastwest direction. For this eld, data were abandoned in the period between day 151 and 196 due to an insufcient fetch (Soegaard et al. 2002). The equipment at each eld consisted of a 3-dimensional sonic anemometer (Gill Solent; Lymington, UK), and an IRGA (Infrared Gas Analyzer; Li-Cor, USA) was used for measuring water vapour concentrations. The sampling tube (Bev-a-line, LiCor; Lincoln, NE, USA) had an inner diameter of 3.2 mm and it was 10 m in length. The sensor head was mounted 2.5 m above the top of the canopy. The air was sucked through the sampling tube by a membrane pump at a ow rate of 9 l min21. The Edisol software package was used for data collection and processing (Moncrieff et al. 1997). The ux data availability is indicated in gure 1(b). 4. Results 4.1. The dependency of es on V for dense vegetation Because the stomatal cavities of leaves are usually close to saturation, it can be * assumed that Ts~Ts for dense vegetation canopies. In order to study the dependency of es on V in such situations, the VAT model (appendix B) is applied to

Figure 1. (a) Mean midday temperature (11.0014.00 h) and precipitation in the experimental period. (b) Leaf area index (LAI) of wheat (empty circle) and maize (lled circle) in the experimental period. The horizontal lines represent the ux data availability for wheat (upper line) and maize (lower line).

Estimation of evapotranspiration rates

2541

Downloaded By: [Ingenta Content Distribution Psy Press Titles] At: 10:42 7 November 2009

evaluate the es-V relationship of a wheat canopy in periods with a dense vegetation cover (L~4). For inclusion of variability in the environment, the VAT model is applied both to a dry period (1821 May 1998) and a wet period (2730 May 1998). The chosen dry period is towards the end of the longest dry period in the study-area that year (the last rain fell on 12 May 1998), whereas the wet period follows the occurrence of heavy rainfall (17 mm) on 26 May 1998 (gure 1(a)). The simulated uxes and surface conditions by the VAT model are in quite good agreement with the eld measurements of lE and Ts for wheat (gure 2). Assuming that the model is also capable of predicting es and rs correctly (rae is calculated from measurements; appendix B), the VAT model can be used to study the dependency of es on V. As would be expected, the measurements of ea at reference height are independent of variations in V (gure 3(a)). The surfaceatmosphere system is well coupled (Vp0) during early mornings and late afternoons because the leaf stomatal (surface) resistance is higher during these periods. In such situations, the simulated es (gure 3(b)) remains in the same range as ea. Towards the less well coupled conditions prevailing during midday (Vp1), es increases (gure 3(b)), and the same pattern is recognized in the relationship between es* and V (gure 3(c)). While es is, in fact, linearly related to es*, it is even better related to the results derived from the empirical expression for es suggested by equation (9) (gure 4). Using the VAT model for calculating the factor A (equation (10)), its mean value is found to be 0.87 (SD~0.06) during the 6 days depicted in gures 2 and 3. 4.2. lE calculation based on empirical expression for es The calculations to estimate lE (equations (5), (6), (9) and (11)) were run on a half-hourly basis using the inputs of Rn, Ts, Ta, ea and G during the periods when ux data are available for validation (gure 1(b)). For this purpose, G is determined as the residual in the energy balance equation (equation (2)) using the input data of lE, H and Rn. At rst, the calculations of lE are conducted assuming Ts~Ts* which is valid for dense canopies. In this case, it is seen (gure 5) that the lE predictions compare well with the measurements when LAI>3 (dense vegetation).

Figure 2. (a) Comparison between the evapotranspiration rates (lE) simulated by a vegetationatmosphere transfer (VAT) model (appendix B) and eld measurements of lE. (b) Comparison of surface temperature (Ts) simulated by a VAT model and eld measurements of Ts. The gures were produced on the basis of data representing dense vegetation, including both a dry period (18 May to 21 May) and a wet period (27 May to 30 May, 1998).

2542

E. Boegh and H. Soegaard

Downloaded By: [Ingenta Content Distribution Psy Press Titles] At: 10:42 7 November 2009

Figure 3. (a) Measurements of atmospheric vapour pressure (ea) versus the decoupling coefcient (V) simulated by a vegetation-atmosphere transfer (VAT) model, (b) VAT simulations of the surface vapour pressure (es) versus V, (c) VAT simulations of the saturated vapour pressure at the surface (es versus V. The gures were produced on *) the basis of data representing dense vegetation, including both a dry period (18 May to 21 May, 1998) and a wet period (27 May to 30 May, 1998).

Figure 4. Comparison between the vapour pressure at the surface (es) estimated using the hypothetical equation (8), es ~AVe z1{Vea , and es modelled by a more complex s vegetationatmosphere transfer model. (a) using A~1 in equation (9), (b) using A~0.9 in equation (9). The gures were produced on the basis of data representing dense vegetation, including both a dry period (18 May to 21 May, 1998) and a wet period (27 May to 30 May, 1998).

Estimation of evapotranspiration rates

2543

Downloaded By: [Ingenta Content Distribution Psy Press Titles] At: 10:42 7 November 2009

Figure 5. Comparison between evapotranspiration rates (lE) calculated assuming Ts~Ts * (using equations (5), (6), (8) and (9)) and eddy covariance data of lE. Results are shown separately for bare/sparse canopies where LAIv3 (empty circle) and for dense canopies where LAI>3 (lled circle). (a) Wheat, (b) maize.

On many occasions, the results also perform well for LAIv3, in particular for wheat. The better performance for the wheat eld than for maize relates to differences in (Ts-Ta). Generally, the larger values of (Ts-Ta) predominate at low LAI, but these are much lower for wheat (gure 6(a)) than for maize (gure 6(b)) because the periods with sparse vegetation at the wheat eld were characterized by frequent rain events (gure 1(a)). In contrast, the exposed soil patches at the maize eld cause (Ts-Ta) to increase dramatically during the dry spells in MayJune as can be seen in gure 1(b)). Overall, the lE predictions (assuming Ts~Ts*) are in good accordance with eld measurements when (Ts-Ta)v5C (gure 7). However, when (Ts-Ta)w5C, the calculated lE is overestimated. In this situation, the effective source of water vapour from the dry soil patches at the maize eld is located below an upper dry layer exhibiting distinct thermal properties. This causes the measurement of Ts to exceed the hypothetical temperature of the evaporating front which is located *) deeper in the soil (Ts which, in turn, causes es* and lE to be overestimated.

Figure 6. Field data of surface temperature (Ts) and air temperature (Ta) are used to illustrate the variations in (Ts-Ta) for (a) wheat and (b) maize during the experimental period.

2544

E. Boegh and H. Soegaard

Downloaded By: [Ingenta Content Distribution Psy Press Titles] At: 10:42 7 November 2009

Figure 7. Comparison between evapotranspiration rates (lE) calculated assuming Ts~Ts * (using equations (5), (6), (8) and (9)) and eddy covariance data of lE. Results are shown separately for (Ts-Ta)w5C (empty circle) and for (Ts-Ta)5C (lled circle), where (Ts-Ta) is the difference between surface temperature and air temperature. (a) Wheat, (b) maize.

4.3. Assessment of Ts* of dry bare soil The estimation of Ts* during dry conditions can be made using equation (7) if information on zd is available. Even though the moisture condition in the upper thin surface layer of soils is generally considered to be very important for determining rs and soil-water limited rates of lE, only very few registrations of zd have been reported in the literature. For a sandy loam soil in Japan with a particle composition (81% sand, 9% silt, 10% clay) very similar to that of the agricultural topsoils in the present study (80% sand, 10% silt, 10% clay), the maximum depth of the dry layer was found to be 0.02 m while for a sand dune site (96% sand, 2% silt, 2% clay), the maximum observed zd was 0.08 cm (Yamanaka and Yonetani 1999). Using zd~0.02 m to represent the soil condition during the very dry period, 1821 May 1998, in this study, measurement-based estimates of G/H (equation (2)) and rae (equation (3)) can be used to calculate B. In this period, the average G/H was found to be 0.66 (SD~0.29) and the average rae was 53 s m21 (SD~13 s m21). Using ks~0.25 W m21 C21 to represent the thermal conductivity of the upper dry soil layer, B is given by B~(0.66*1227*0.02)/(53*0.25)~1.2. Theoretically, the factor B should then vary from 0 (in which case Ts~Ts*) for a wet soil to 1.2 for a dry soil. Because the transition from energy-limited to soil-limited evaporation tends to be abrupt and is accompanied by an increase in surface temperature (Amano and Salvucci 1999), it is simply assumed that Ts*<Ts when (Ts-Ta)v5C (4.2) while equation (7) (with B~1.2) is adopted for calculating lE when (TsTa)>5C. The sensitivity of the calculated lE to B is evaluated in 4.4. 4.4. lE calculation using empirical estimates of es and Ts* In the following, the calculation of lE for surfaces characterized by (Ts* Ta)v5C is based on the assumption that Ts~Ts while the calculation of lE for warm surfaces with (Ts-Ta)>5C uses the relationship in equation (7) to assess Ts *. This causes predictions to be comparable with measurements (gure 8). Figure 9 facilitates a more detailed evaluation of the results at the wheat eld and maize eld, respectively. For each eld, diurnal predictions of lE and rae are compared with measurement-based values of lE and rae on three different days varying in vegetation density and surface temperature (table 1).

Estimation of evapotranspiration rates

2545

Downloaded By: [Ingenta Content Distribution Psy Press Titles] At: 10:42 7 November 2009

Figure 8. Comparison between predicted evapotranspiration rates (lE) and eld measurements of lE for (a) wheat and (b) maize. The 1:1 lines are also shown. The comparison includes results from all days with available ux data (gure 1(b)).

Towards the mornings and late evenings, the predictions of rae obtain large (positive or negative) values when the denominator in equation (5) (Rn-G) approaches zero. The case is the same for the measurement-based rae when H approaches zero. During midday, the predictions of rae are in quite good agreement with the measurement-based rae. For the dense maize eld (gure 9( f )), Ts typically comes very close to Ta during afternoons, and it may even happen that Ts becomes lower than Ta. Simultaneous observations of negative (Ts-Ta) and positive H indicate the presence of counter-gradient uxes in the tall maize canopy which causes the measurement-based rae to become negative (gure 9( f )). This situation illegitimates both the measurement-based rae and the applied methodology for calculating lE. On day 222 (gure 9( f )), both (Ts-Ta) and H turn negative during late afternoon which indicates that sensible heat is being used to increase the transpiration rates. The sensitivity of the calculated lE in gure 9 to the various inputs and parameter settings is illustrated in table 2. Overall, there is a quite broad range of sensitivities to the individual parameters. For sparse/dense canopies with (TsTa)v5C, the sensitivity is highest to the observations of Ts, Rn and ea while the sensitivity to the choice of parameter values for A and B is less. For the dry bare soil and sparsely vegetated surfaces where (Ts-Ta)>5C, the sensitivity to the empirical surface humidity parameter A is highest, and the sensitivity to the parameter B is also important. Nevertheless, it seems feasible to use a xed parameter value, A~0.9, regardless of vegetation density and dryness condition (gure 8). Optimally, the setting of parameter B should use information on the dry surface layer development (4.3) to improve the predicted transition from energy limited to soil water limited evapotranspiration rates. However, in this study reasonable results were obtained using a setting of B~1.2 to represent a sandy loam soil. 5. Conclusion A method is presented to calculate lE using standard meteorological data and remote sensing based estimates of Ts, Rn and G. The method uses physical descriptions of the surface energy exchange together with an empirical assessment of the surface humidity (equation (9)) and, for dry surfaces, the temperature of the

2546

E. Boegh and H. Soegaard

Downloaded By: [Ingenta Content Distribution Psy Press Titles] At: 10:42 7 November 2009

Figure 9. Comparison between predictions of the atmospheric resistance between the surface and the air (rae, empty circle/dotted line), measurement-based values of rae (thick line), predictions of evapotranspiration rates (lE, lled circle) and eld measurements of lE (thin line). Results are shown for wheat and maize on selected days characterized by variations in leaf area development and surfaceair temperature differences (see table 1).

evaporating front needs to be evaluated (equation (7)). It is interesting that, when we use a more basic and traditional, although more complicated, model, we nd that the hypothetical equation (9) is approximately true. For humid surface conditions (Ts*<Ts), the introduction of this hypothetical equation into a set of simpler equations also describes our eld data well. Because Ts constitutes the lower

Estimation of evapotranspiration rates

2547

Table 1. Leaf area index, LAI, and the difference between surface temperature and air temperature at 13.00 h (Ts-Ta)13h, on selected days for (a) wheat and (b) maize. See gure 9 for results of evapotranspiration rates and atmospheric resistance on these days. (a) Wheat LAI (Ts-Ta)13h (b) Maize LAI (Ts-Ta)13h Day 138 0 15.1 Day 164 0.1 6.4C Day 222 4.2 2.1C Day 112 0.3 4.4C Day 138 3.8 3.6C Day 189 0 4.5C

Downloaded By: [Ingenta Content Distribution Psy Press Titles] At: 10:42 7 November 2009

Table 2. Fractional change of the calculated evapotranspiration rate to independent increases of 10% in input and parameter settings. Results are shown for wheat and maize on selected days at midday (13.00 h). See also table 1 and gure 9. (Ts, surface temperature; Rn, net radiation; Ta, air temperature; ea, vapour pressure; A, parameter in equation (9); B, parameter in equation (7). Wheat Day 112 Ts Rn Ta ea A B 20.57 0.09 0.05 20.05 0.08 0 Day 138 20.45 0.08 0.07 20.06 0.07 0 Day 189 0.26 20.63 0.07 20.03 0.03 0 Day 138 20.26 0.09 0.26 0.01 0.39 20.14 Maize Day 164 20.17 0.11 0.01 20.31 0.43 20.18 Day 222 0.14 0.09 0.13 0.33 0.02 0

boundary condition for the atmospheric uxes and the upper boundary condition for the soil uxes (Ts-Ts*) and (Ts-Ta) are closely associated. Theoretically, this relationship is dependent on the depth of the evaporating surface, but in practice, a constant coefcient representing the ratio of (Ts-Ts*) and (Ts-Ta) for a dry soil was found useful for estimating Ts* and calculation of lE for dry surfaces with (Ts-Ta)>5C. Should the expressions for es and Ts* prove to be reasonable approximations in other studies, they may provide an effective and simple way to calculate evapotranspiration from standard, easily obtainable meteorological data and thermal infrared remote sensing data. In Boegh et al. (2004), the application of the method for calculating lE in Denmark is demonstrated based on inputs of Earth observations (NOAA AVHRR) and climate predictions which were computed by a high-resolution weather forecast model. Appendix A The NewtonRaphson Method states that if x~c is an approximation to the solution of the equation f(x)~0, then a better approximation is given by cnew ~c{ f c f 0 c A1

2548

E. Boegh and H. Soegaard

When estimating lE, the function f is given by f x~rnew {rae ae where rnew ae is calculated using equation (5). The derivative of f follows as f 0 x~ where drnew drnew des ae ~ ae dre des dre with
Downloaded By: [Ingenta Content Distribution Psy Press Titles] At: 10:42 7 November 2009

A2

drnew ae {1 dre

A3

A4

rcp drnew ae ~ des cRn {G and   des D=cz1 D=cz1 ~e A r r2 { r r2 e s 2 s ae 2 s ae a drae D=cz1zrs =rae D=cz1zrs =rae

A5

A6

Appendix B. Description of VAT model The VAT-model consists of a coupled photosynthesis/stomatal resistance model which is connected to the above atmosphere by an atmospheric boundary layer. The canopy photosynthetic rate is simulated to allow the prediction of stomatal (surface) resistance which, in turn, is used to evaluate the transpiration rate and the surface energy balance. The model is applied to a dense canopy only. The model applied in this paper is a slightly simplied version of the model described in Boegh et al. (2002b). The same model parameters representing wheat (taken from the literature) are applied, but in this paper, more input data are based on measurements (rather than predictions) in order to maximize the accuracy of the simulations. As such, inputs of measured net radiation are used in this version (rather than global radiation), and the atmospheric resistance between the surface and the atmosphere above the canopy (rae) is assessed using the measurements of H, Ts and Ta:  {1 sm rae ~rcp Ts {Ta H B1 In Boegh et al. (2002b), all leaf level calculations were made separately for shaded and sunlit leaves, but in this version, only the photosynthetic rate is computed seperately for sunlit and shaded leaves. The radiation absorbed by sunlit and shaded leaves is evaluated, and an exponential prole of leaf Rubisco capacity in the canopy is distributed between sunlit and shaded leaves. The canopy photosynthesis (Ac) is the sum of shaded and sunlit fractions. The (big-leaf) stomatal conductance (gs~1/rs) is then calculated as (Leuning 1995): a wsf Ac P mol m{2 s{1 gs ~gs0 Lz B2 ps {C1zDs =D0 where P is atmospheric pressure (Pa), ps is the CO2 partial pressure at the surface (Pa), C is the CO2 compensation point (Pa), Ds is the vapour pressure decit at the surface, and gs0, a and D0 are empirical coefcients for wheat. The wsf is a waterstress factor (set to unity). Because Ds is not known, Ds and gs are solved by iteration. In Boegh et al. (2002b), the net-isothermal form of the PM equation was used to compute the transpiration rates, but in this version, it is preferable to keep

Estimation of evapotranspiration rates

2549

the VAT model completely independent of the linearly approximated denition of D which is used in the empirical expression for es (to be compared with the VATmodelled es). The big-leaf transpiration rate is therefore computed by   W m{2 lE~ rcp c e {ea rae zrs B3 s and the big-leaf surface temperature (Tl) is calculated as  0 C Tl ~Hrae rcp zTa B4

Downloaded By: [Ingenta Content Distribution Psy Press Titles] At: 10:42 7 November 2009

where H is the sensible heat ux (H~Rn-lE). The evaluation of the big-leaf temperature, rather than a separate evaluation of energy balances for shaded and sunlit leaves (as in Boegh et al. 2002b), allows direct validation of the modelled Tl relative to the Ts measurements. Finally, es is calculated by  0 C B5 es ~e {lE rs c rcp s The gs, Tl, lE and Ac are solved simultaneously with the computation of es, ps and pi (leaf internal CO2 pressure) through utilizing a NewtonRaphson iterative solution process for pi, as introduced by Collatz et al. (1991). In order to simulate a dense winter wheat canopy, the leaf area index is set to 4, and the leaf photosynthetic capacity at the top of the canopy, Vmax0, is set to 150 mmol m22s21 to represent wheat (Wang and Leuning 1998). Acknowledgments The paper was prepared within the projects EO-ux-budget and RS-model which is nanced by the Danish research council, Copenhagen. Information on leaf area index and precipitation was kindly provided by Anton Thomsen and Tom Jensen from The Danish Institute of Agricultural Sciences in Foulum, Denmark. References
ALLEN, R., PEREIRA, L. S., RAES, D., and SMITH, M., 1998, Crop evapotranspiration. Irrigation and Drainage Paper 56 (Rome, Italy: FAO). BALDOCCHI, D. D., HICKS, B. B., and MEYERS, T. P., 1988, Measuring biosphere atmosphere exchanges of biologically related gases with micrometeorological methods. Ecology, 69, 13311340. BALL, J. T., WOODROW, I. E., and BERRY, J. A., 1987, A model predicting stomatal conductance and its contributors to the control of photosynthesis under different environmental conditions. In Progress in Photosynthesis Research, Vol. IV, edited by J. Biggens (Dordrecht: Martinus Nijhoff), pp. 221225. BASTIAANSSEN, W. G. M., MENENTI, M., FEDDES, R. A., and HOLTSLAG, A. A. M., 1998, A remote sensing surface energy balance algorithm for land (SEBAL). 1. Formulation. Journal of Hydrology, 212213, 198212. BERKOWICZ, R., and PRAHM, L. P., 1982, Sensible heat ux estimated from routine meteorological data. Journal of Applied Meteorology, 21, 18451864. BOEGH, E., SOEGAARD, H., FRIBORG, T., and LEVY, P. E., 1999, Models of CO2 and water vapour uxes from a sparse millet crop in the Sahel. Agricultural and Forest Meteorology, 93, 726. BOEGH, E., SOEGAARD, H., and THOMSEN, A., 2002a, Evaluating evapotranspiration rates and surface conditions using Landsat TM to estimate atmospheric resistance and surface resistance. Remote Sensing of Environment, 79, 329343. BOEGH, E., SOEGAARD, H., BROGE, H., HASAGER, C. B., JENSEN, N. O., SCHELDE, K., and THOMSEN, A., 2002b, Airborne multispectral data for quantifying leaf area index, nitrogen concentration, and photosynthetic efciency in agriculture. Remote Sensing of Environment, 81, 179193. BOEGH, E., SOEGAARD, H., CHRISTENSEN, J. H., HASAGER, C. B., JENSEN, N. O., NIELSEN, N. W., and RASMUSSEN, M. S., 2004, Combining weather prediction and remote

2550

E. Boegh and H. Soegaard

sensing data for the calculation of evapotranspiration rates: application to Denmark. International Journal of Remote Sensing, in press. CELLIER, P., RICHARD, G., and ROBIN, P., 1996, Partition of sensible heat uxes into bare soil and the atmosphere. Agricultural and Forest Meteorology, 82, 245265. CHOUDHURY, B. J., REGINATO, R. J., and IDSO, S. B., 1986, An analysis of infrared temperature observations over wheat and calculation of latent heat ux. Agricultural and Forest Meteorology, 37, 7588. COLLATZ, G. J., BALL, J. T., GRIVET, C., and BERRY, J. A., 1991, Physiological and environmental regulation of stomatal conductance, photosynthesis and transpiration: a model that includes a laminar boundary layer. Agricultural and Forest Meteorology, 54, 107136. DAAMEN, C. C., and SIMMONDS, L. P., 1996, Measurement of evaporation from bare soil and its estimation using surface resistance. Water Resources Research, 32, 13931402. GARRATT, J. R., 1992, Extreme maximum land surface temperatures. Journal of Applied Meteorology, 31, 10961105. HAM, J. M., and HEILMAN, J. L., 1991, Aerodynamic and surface resistance affecting energy transport in sparse crops. Agricultural and Forest Meteorology, 53, 267284. HUBAND, N. D. S., and MONTEITH, J. L., 1986, Radiative surface temperature and energy balance of a wheat canopy. 1. Comparison of radiative and aerodynamic temperature. Boundary-Layer Meteorology, 36, 117. HUMES, K. S., KUSTAS, W. P., and GOODRICH, D. C., 1997, Spatially distributed sensible heat ux over a semi-arid watershed. Part 1: Use of radiometric surface temperatures and a spatially uniform resistance. Journal of Applied Meteorology, 36, 281293. JARVIS, P. G., 1976, The interpretations of the variation in leaf water potential and stomatal conductance found in canopies in the eld. Philosophical Transactions of the Royal Society of London, Series B, 273, 593610. JARVIS, P. G., and MCNAUGHTON, K. G., 1986, Stomatal control of transpiration. Scaling up from leaf to region. Advances in Ecological Research, 15, 149. JONES, H. G., 1992, Plants and Microclimate, 2nd edn (Cambridge: Cambridge University Press). KALMA, J. D., and JUPP, D. L. B., 1990, Estimating evaporation from pasture using infrared thermometry: evaluation of a one-layer resistance model. Agricultural and Forest Meteorology, 51, 223246. KUSTAS, W. P., CHOUDHURY, B. J., MORAN, M. S., REGINATO, R. J., JACKSON, R. D., GAY, L. W., and WEAVER, H. L., 1989, Determination of sensible heat ux over sparse canopy using thermal infrared data. Agricultural and Forest Meteorology, 44, 197216. KUSTAS, W. P., and NORMAN, J. M., 1997, A two-source approach for estimating turbulent uxes using multiple angle thermal infrared observations. Water Resources Research, 33, 14951508. LEUNING, R., 1995, A critical appraisal of a combined stomatal-photosynthesis model for C, plants. Plant, Cell Environment, 18, 339355. LINDROTH, A., and HALLDIN, S., 1986, Numerical analysis of pine forest evaporation and surface resistance. Agricultural and Forest Meteorology, 38, 5979. MASSMAN, W. J., 1999, A model study of kB{1 for vegetated surfaces using localized nearH eld Lagrangian theory. Journal of Hydrology, 223, 2743. MONTEITH, J. L., and UNSWORTH, M. H., 1990, Principles of Environmental Physics, 2nd edn (New York: Edward Arnold). NORMAN, J. M., 1993, Scaling processes between leaf and canopy levels. In Scaling Physiological Processes. Leaf to globe, edited by J. R. Ehleringer and C. B. Field (New York: Academic Press, Inc.), pp. 4177. NORMAN, J. M., KUSTAS, W. P., and HUMES, K. S., 1995, Source approach for estimating soil and vegetation energy uxes in observations of directional radiometric surface temperature. Agricultural and Forest Meteorology, 77, 263293. NOVAK, M. D., 1986, Theoretical values of daily atmospheric and soil thermal admittances. Boundary-Layer Meteorology, 24, 1734. PAW, U. K. T., and GAO, W., 1988, Applications of solutions to non-linear energy budget equations. Agricultural and Forest Meteorology, 43, 121145. QIU, G. Y., YANO, T., and MOMII, K., 1998, An improved methodology to measure

Downloaded By: [Ingenta Content Distribution Psy Press Titles] At: 10:42 7 November 2009

Estimation of evapotranspiration rates

2551

Downloaded By: [Ingenta Content Distribution Psy Press Titles] At: 10:42 7 November 2009

evaporation from bare soil based on comparison of surface temperature with a dry soil surface. Journal of Hydrology, 210, 93105. SEGUIN, B., ASSAD, E., FRETEAUD, J. P., IMBERNON, J., KERR, Y., and LAGOUARDE, J. P., 1989, Use of meteorological satellites for water balance monitoring in Sahelian regions. International Journal of Remote Sensing, 10, 11011117. SELLERS, P. J., BERRY, J. A., COLLATZ, G. J., FIELD, C. B., and HALL, F. G., 1992, Canopy reectance, photosynthesis and transpiration. III. A reanalysis using improved leaf models and a new canopy integration scheme. Remote Sensing of Environment, 42, 187216. SHUTTLEWORTH, W. J., and WALLACE, J. S., 1985, Evaporation from sparse cropsan energy combination theory. Quarterly Journal of the Royal Meteorological Society III, 469, 839857. SOEGAARD, H., JENSEN, N. O., BOEGH, E., HASAGER, C. B., SCHELDE, K., and THOMSEN, A., 2002, Carbon dioxide exchange over agricultural landscape using eddy correlation and footprint modeling. Agricultural and Forest Meteorology, 114, 153173. STEWART, J. B., KUSTAS, W. B., HUMES, K. S., NICHOLS, W. D., MORAN, M. S., and DE BRUIN, H. A. R., 1994, Sensible heat uxradiometric surface temperature relationship for 8 semi-arid sites. Journal of Applied Meteorology, 33, 11101117. SUGITA, M., and BRUTSAERT, W., 1990, Regional surface uxes from remotely sensed skin temperature and lower boundary layer measurements. Water Resources Research, 26, 29372944. THORGEIRSSON, H., and SOEGAARD, H., 1999, Simulated carbon dioxide of leaves of barley scaled to the canopy and compared to measured uxes. Agricultural and Forest Meteorology, 98, 479489. TROUFLEAU, D., LHOMME, J. P., MONTENY, B., and VIDAL, A., 1997, Sensible heat ux and radiometric temperature over sparse Sahelian vegetation. I. An experimental analysis of the kB21 parameter. Journal of Hydrology, 188189, 815838. VAN DER KEUR, P., HANSEN, S., SCHELDE, K., and THOMSEN, A., 2001, Modication of DAISY SVAT model for use of remotely sensed data. Agricultural and Forest Meteorology, 106, 215231. VERHOEF, A., DE BRUIN, H. A. R., and vAN DER HURK, B. J. J. M., 1997, Some practical notes on the parameter kB21 for sparse vegetation. Journal of Applied Meteorology, 36, 560573. WANG, Y.-P., and LEUNING, R., 1998, A two-leaf model for canopy conductance, photosynthesis and partitioning of available energy I: Model description and comparison with a multi-layered model. Agricultural and Forest Meteorology, 91, 89111. YAMANAKA, T., and YONETANI, T., 1999, Dynamics of the evaporation zone in dry sandy soils. Journal of Hydrology, 217, 135148. YAMANAKA, T., TAKEDA, A., and SUGITA, F., 1997, A modied surface-resistance approach for representing bare-soil evaporation: wind tunnel experiments under various atmospheric conditions, Water Resources Research, 33, 21172128.

You might also like