Remote sensing data from the NOAA-11 AVHRR satellite were collected over the USDA-Agricultural Research Service Walnut Gulch Experimental Watershed in southeastern Arizona during the MONSOON 90 field campaigns. An energy balance model which relies primarily on remotely sensed inputs was used to extrapolate evapotranspiration (ET) estimates from one location containing near-surface meteorological data to other areas in the basin. Satisfactory results were obtained under a wide range of environmental conditions. However, the ET values are essentially instantaneous and therefore do not necessarily provide reliable estimates of daytime or daily ET fluxes required for many hydrological and resource management applications. An operational technique was developed to extrapolate one time of day ET estimates to daytime averages using the evaporative fraction concept and empirical methods for converting midday available energy to daytime average values. Model derived daytime average ET fluxes were in reasonable agreement with local ground-based measurements. The technique also was used to estimate daily ET at the basin scale.

INTRODUCTION

Reliable methods for estimating actual evapotranspiration (ET) are critical for accurately assessing the water balance at basin and regional scales. Operational techniques using standard meteorological input variables have been developed (Bouchet, 1963; Morton, 1971; Brutsaert and Stricker, 1979; Kovaks, 1987) and tested in different climatic regions (Bouchet, 1963; Brutsaert and Stricker, 1979; Morton, 1983; Ali and Mawdsley, 1987; Byrne et al., 1988; Granger and Gray, 1990; Parlange and Katul, 1992). While these approaches have shown some promise, a major drawback for evaluating regional ETs is that they utilize near-surface meteorological observations which represent local environmental conditions. This limits their application to basins containing uniform vegetation, soil moisture, and topography.

Remote sensing has the potential for providing synoptic surface information that is relevant to energy balance modeling over regions containing significant spatial variation in vegetation cover and soil moisture (Price, 1990). Yet, remote sensing cannot provide important atmospheric variables such as wind speed, air temperature, and vapor pressure. This difficulty has been alleviated somewhat by employing atmospheric boundary layer models which can simulate these quantities given initial conditions from conventional large-scale weather observations (Carlson et al., 1981; Wetzel et al., 1984) or by utilizing information provided by mesoscale atmospheric models (Noilhan et al., 1991). Therefore, it appears feasible to combine more detailed soil-plant-atmosphere models that make use of remote sensing data (Soer, 1980; Taconet et al., 1986; Camillo, 1991) with models that simulate atmospheric forcing variables (e.g., wind speed, air temperature, and humidity). This type of an approach, however, will be difficult to implement over large heterogeneous areas without a considerable amount of surface information for all the required input parameters. Attempts to minimize the number of required model inputs while maintaining the basic physics for quantifying regional ET have shown some success (e.g., Noilhan and Planton, 1989; Bougeault et al., 1991).

Other approaches that have operational capabilities combine remote sensing observations with ancillary meteorological data. One such technique utilizes surface air temperature differences near midday to compute daily ET (Jackson et al., 1977; Seguin and Itier, 1983). This method and refinements to it has been successful in mapping ET (especially at time scales of days to weeks) over large areas containing minimal meteorological and other surface information (Rambal et al., 1985; Seguin et al., 1989; Lagouarde, 1991). In addition, remote sensing models requiring minimal inputs for mapping ET have been used in hydrologic models to improve water balance calculations (e.g., Sucksdorff and Ottl?, 1990; Ottl? et al., 1989).

Results from these studies suggest that ET models that incorporate remote sensing information have the capability of quantifying basin and regional scale ET. However, relatively few studies have verified model estimates of spatially distributed ET fluxes with actual measurements, nor have there been many investigations into the potential errors in flux estimates caused by interpolating near-surface meteorological observations over large heterogeneous areas.

In this article, a method for mapping ET over a semiarid rangeland watershed using satellite observations from NOAA- 11 AVHRR with remotely sensed data from low level aircraft measurements and local ancillary meteorological is discussed. The data were collected during the MONSOON 90 field campaign (Kustas et al., 1991a). The approach involved computing a reference ET with local meteorological data and extrapolating this estimate to other parts of the basin using remotely sensed inputs (Gash, 1987; Kustas et al., 1990). The computed energy fluxes are essentially instantaneous and hence are not very useful for hydrologic modeling or for water resource management decisions. This shortcoming was addressed by taking the "instantaneous" ET fluxes evaluated near midday and converting them to "instantaneous" values of evaporative fraction (EF). The EF represents the fraction of available energy at the surface that is actually used for ET. Estimates of EF from this experiment and other field studies (Shuttleworth at al., 1989) have shown midday EF to be highly correlated to daytime average EF. This result was utilized to compute and map daytime average and daily ET for the basin, which is a more useful quantity for hydrologic applications.

MATERIALS AND METHODS

Site Description

The MONSOON 90 field experiment was conducted in the Walnut Gulch Experimental Watershed (31.5EN 110EW) near Tucson, Arizona. The watershed is approximately 150 km² in area and is part of the San Pedro Basin. The watershed topography is characterized by rolling hills 10-20 m in height with separation between ridge tops on the order of 500 m. The main ephemeral gullies run in a west-southwest direction with numerous smaller ridge tops on the order of 500 m. The main ephemeral gullies ridge tops on the order of 500 m. The main ephemeral gullies run in a west-southwest direction with numerous smaller channels, running north-south, significantly dissecting the watershed. The mean elevation of the watershed is 1500 m above mean sea level (MSL). The vegetation cover is generally sparse and highly variable (i.e., from 10% to 70% cover) with a greater amount of vegetation growing in the ephemeral channels. The plant communities are generally described as either shrub- or grass-dominated with the shrub communities primarily occupying the western half of the watershed.

Experimental Observations

Field campaigns in 1990 were held during the dry season (June) and during the wet or "monsoon" season (July-September). Details concerning the measurements made during these field campaigns are summarized by Kustas et al. (1991a).

Ground Data

The ground-based data used in this study were obtained from eight meteorological-energy flux (METFLUX) stations situated along two parallel transects and mainly on the ridge tops. The METFLUX network covered the shrub- and grass-dominated plant communities and the transition between them (Fig. 1). The sites were separated by 2-4 km and covered about 1/3 the area of the watershed. Measurements included air temperature (T.), wind speed (u), and wind direction at 4-4.5 m above ground level (agl), net radiation (R_n), incoming solar radiation (R_s) soil heat flux (G), near-surface soil temperature and soil moisture (-2.5 cm to -5 cm), and surface temperature. A detailed description of the instrumentation can be found in Kustas et al. (1994). The standard deviation in air temperature at 4 m was recorded and used to compute the sensible heat flux H by the variance method (see, e.g., Tillman, 1972; Wesely, 1988; Weaver, 1990). The latent heat flux was solved as a residual in the energy balance equation, namely:

where fluxes away from the surface are negative. Comparison of daytime fluxes between the variance method and eddy correlation techniques showed satisfactory agreement (Kustas et al., 1991a; 1994). Data from the METFLUX network were collected continuously during the main field campaign, in July-August, with 20-min averages being recorded. During the June and September field studies, only the two METFLUX stations at the eastern- and westernmost sites, namely, Lucky Hills (site 1) and Kendall (site 5), were in operation (see also Figs. 1 and 2). This provided estimates of the fluxes for the shrub- and grass-dominated ecosystems.

Figure One

Aircraft Data

Remote sensing measurements from a low flying aircraft (100 mg agl) were used in this analysis to supplement and help interpret the satellite observations. This included performing empirical adjustments to vegetation index estimates from satellite-based reflectances for atmospheric effects. Data were collected with nadir viewing sensors in the visible, near-infrared, and thermalinfrared wavelengths. The instruments had a 150 field of view, which flying at a nominal altitude of 100 m agl produced about a 25 m pixel on the surface. The aircraft flight speed averaged about 50 m s^-1 with instruments sampling at 1 Hz. A black-and-white video camera on board recorded aircraft location and was used later to geolocate the measurements. Due to the relatively low flying altitude, there were difficulties in maintaining constant aircraft speed, altitude, and direction. In order to obtain remotely sensed quantities that were representative of conditions near the METFLUX sites, averaging of several pixels along the transects was required. This aircraft and instrument package has been used in a number of remote sensing experiments (e.g., Jackson et al., 1987; Kustas et al., 1989; 1990; Moran et al., 1989). The aircraft flew the three transects illustrated in Figure 2, which covered the METFLUX network and main study area. Lines 1 and 2 were about 10 km in length while line 3 was about 5 km. A mission normally consisted of flying line 3 several times followed by lines 1 and 2, flying them twice in opposite directions. A mission took approximately 30 min to complete, but actual flight time over any of the flight lines took less than 5 min.

Satellite Observations

The NOAA-11 AVHRR local area coverage (LAC) data were collected during the field campaigns. The fivechannel LAC data were obtained in digital 1B format. The visible, near-infrared, and two thermal channels (i.e., Channels 1, 2, 4, and 5) were processed. All images were coregistered to within I pixel accuracy using the 1:2,000,000 digital line graph data provided by the U.S. Geological Survey. Calibration of the visible and thermal data were performed according to NOAA guidelines (NOAA, 1988a,b). The thermal channel data were corrected for the nonlinear response of the sensors using published corrections as a function of scene brightness temperature and internal blackbody temperature (NOAA, 1988b). For the days considered, the large scan angles (see Table 1) resulted in pixel resolutions that were four to seven times the 1.1 km resolution of AVHRR LAC. data at nadir. After georegistration of the images, the data were resampled to create 1. 1 km pixels. At this resolution, 163 pixels comprised the watershed.

Figure Two

Table One

DATA ANALYSIS

Remote Sensing Observations

Analysis of the NOAA-11 images collected during the field campaigns gave three daytime images containing minimal cloud cover under different environmental conditions. One image was collected on DOY 156 (5 June) during the dry season while the other two were during the "monsoon" season, viz., DOY 209 (28 July) and DOY 216 (4 August). Data also were collected with the low altitude aircraft. A summary of the remote sensing observations is given in Table 1.

Good agreement between the low altitude aircraft remote sensing observations and ground-based measurements (Moran et al., 1991a) suggested that atmospheric corrections were not necessary. Atmospheric water vapor correction to the thermal channels of the NOAA-11 AVHRR was performed by Doraiswamy and Perry (1991). They compared the split window model of Price (1984) with LOWTRAN 7 (Kneizys et al., 1988) and WNDO (Price, 1983) radiative transfer models for DOYs 156 and 209. The results showed satisfactory agreement between the split window and radiative transfer models. Thus the split window technique was considered suitable. Potential errors in the split window approach were evaluated by Kustas et al. (1991b) and estimated to be on the order of ? 2.5 K. This is in agreement with comparisons made between ground and satellite observations by Cooper and Asrar (1989).

For this study, atmospheric attenuation of the visible and near-infrared channels on the NOAA-11 was neglected. Hence the satellite-derived values of normalized difference vegetation index (NDVI) are top of the atmosphere estimates (NDVI.,o,,,). Values of NDVI.,.,, will be significantly less than the surface NDVI (e.g., Arino et al., 1992). Figure 3 is a comparison of NDVI evaluated at the eight METFLUX sites using the NOAA-11 satellite and the low altitude aircraft observations. Although the disagreement between the two measurements are largely due to atmospheric effects, differences in sensor look angle and overpass times also contribute (Gutman, 1991).

Clouds were in the region during the aircraft and satellite overpasses. A cloud screening algorithm for satellite observations was used by Kustas et al. (1991b). For the aircraft measurements, the effects of clouds were detected by significant deviations in the magnitude of the reflectance factors along the aircraft transect. No attempt was made to correct cloud contaminated pixels; instead they were eliminated from the analysis.

Figure Three

Figure Four

Surface Conditions

The three days represent different soil moisture and vegetation conditions. For 5 June, DOY 156, the vegetation was dormant, and the soil moisture profile was dry. By 28 July, DOY 209, several rainfall events earlier in the month restored ample moisture to the root zone and revived the vegetation. However, lack of any significant precipitation the prior week resulted in low values of near-surface (i.e., 0-5 cm) soil moisture existing on this day. In early August, several rainfall events between 1

The above paragraph continues on a scanned page (Warning: 406 KB). The last paragraph on the scanned page continues below.

the energy balance components to new locations. However, the above expression will not yield reliable results if there are significant changes in roughness or local atmospheric conditions. On the other hand, simulations with a surface parameterization for use in mesoscale weather prediction models by Wetzel and Chang (1988) showed that variation in other surface conditions, including soil moisture and percent cover, can strongly contribute to the spatial variation in LE fluxes. This is where a remote sensing technique for quantifying these changes over the watershed and computing area averaged values may be very useful. In fact, methods for estimating regional ET are difficult to verify with ground observations if there is significant spatial variation in the measured fluxes (Brutsaert and Sugita, 1992). Thus the remote sensing method described above may provide estimates of LE more representative of the regional values.

Temporal Extrapolation Scheme

Remote sensing data are essentially instantaneous, and hence the fluxes are "instantaneous" as well. For many applications in hydrology, however, information on daytime or daily ET is much more useful. As previously mentioned, use of remotely sensed surface temperature in the afternoon to infer daily ET has shown some potential. The approach taken here was to compute from the "instantaneous" fluxes an "instantaneous" evaporative fraction (EF), viz.:

These "instantaneous" values of EF, R_n , and G then were related to daily quantities which were used in Eq. (8) to compute daily ET. The value of EF varies from 0 to 1 under daytime conditions with minimal advection. Its magnitude essentially indicates the fraction of available energy at the surface that is converted to latent energy. For selected days of FIFE, Shuttleworth et al. (1989) found that EF was nearly constant for most of the daytime period. They concluded that an estimate of EF around midday would be representative of the daytime average.

This hypothesis was evaluated with observations from the METFLUX sites taken over the June and July-August experimental period. The midday EF was computed from the 20-min averages of the energy balance components, LE, R_n , and G taken between 1030 and 1430 MST (i.e., + 2 h of solar noon) in order to reduce variability inherent in the flux data (Kustas et al., 1994). The daytime estim00ates from the METFLUX sites were obtained by summing the 20-min values of R_n, G, and LE between 0700 and 1800 MST. This was usually the period when R_n > 100 Wm ^-2. The relation between midday and daytime EF is illustrated in Figure 5. A slight bias results in midday EF, EF_{midday ,} underestimating daytime EF, EF_daily, by approximately 8% on average. The data represent a wide range of cloud cover and soil moisture conditions which do not appear to alter the relationship or cause significant scatter. The result in Figure 5 suggests that midday EF is representative of the daytime average value.

The relationship between midday and daytime average R_nand G was assessed by computing the ratios G_daytime / G_midday and R_ndaytime / R_nmidday using average values of R_nand G from the eight METFLUX sites. The daily ratios are shown in Figure 6 for DOY 206 to DOY222. The ratio R_ndaytime / R_nmidday stays relatively constant throughout the period, while larger deviations in G_daytime / G_midday are observed especially for DOYs 213 and 218. An afternoon convective storm on DOY 213 caused large temporal variations in the magnitude of G while DOY 218 was overcast resulting in little change in G-values during the day. Average values for the ratios did not change significantly whether or not these days were excluded from the calculation. For R_ndaytime / R_nmidday the average is 0. 71 ( ? 0.05) and for G_daytime / G_midday the mean is 0.61 (? 0.20). These values were used to convert "instantaneous" values of G and R_n estimated by the remote sensing model to daytime average values. Values of EF computed by the model were then used in (8) to calculate daytime average values of LE and daytime ET in mm day^-1.

Figure Five

Figure Six

RESULTS

Extrapolating Instantaneous Fluxes over the Study Area

The METFLUX site at Lucky Hills (site 1) was chosen as the reference site. Measurements of R_s., T_a, and vapor pressure at the reference site were used in the computation of R_n .. The aerodynamic roughness parameters in (5), z_0m and d₀, were estimated from Kustas et al. (1994). The aircraft spectral data were used to estimate surface albedo for all eight sites and values of NDVI_TOA were multiplied by a constant for each day so that they agreed, on average, with the aircraft estimates (Figure 3).
    Field observations indicate that G / R_n is not constant throughout the day (e.g., Kustas and Daughtry, 1990). To investigate whether estimates of G / R_n using NDVIhad to be adjusted for the temporal variability in this ratio, remote sensing estimates of G / R_n. were compared to 20-min values of G / R_n from the METFLUX sites as a function of time. To reduce the inherent noise in G / R_n-values from individual sites, average values of G and R_n from the METFLUX network were used together with an average NDVI computed for the study area. The ratio of the quantity G / R_n from the METFLUX network, (G / R_n)_meas, to that computed with the average NDVI, (G / R_n)_NDVI ,was determined as a function of time for several days which were typical of environmental conditions during the field campaigns. The values of (G / R_n)_meas,/(G / R_n)_NDVI as a function of time are plotted in Figure 7, with each point being an average value computed from the sampled days. For the given satellite overpass times (Table 1), values of (G / R_n)_NDVI for DOY 156 were adjusted by 0.8 while, for DOYs 209 and 216, the ratio was close enough to 1.0 (especially given the scatter in this relationship for individual days) that no adjustment in the values of (G / R_n )_NDVI were made.

The comparison between measured and estimated R_n and G for the eight sites and on the three days is illustrated in Figure 8. Also listed are the root mean square error (RMSE) and mean percent difference (MPD). Values of RMSE were computed with the standard formula (e.g., Willmott, 1982), and MPD was computed by taking the average of absolute differences between modeled and measured values divided by the measured. There is generally good agreement between modeled and measured fluxes, although model sensitivity to the spatial variation in R_n. and G among the sites is somewhat limited.

The lack of model sensitivity is due at least partly to the significant difference in the scale of the remote sensing measurements used as input to the model and the ground-based measurements of R_n and G. The radius of influence of the upwelling and reflected radiation from the surface on the net radiometer was calculated by Stannard et al. (1994) to be approximately 20 m. Individual soil heat flux measurements were influenced by an area having a radius on the order of 10^-2 m. In contrast, the remote sensing observations for estimating albedo, upwelling radiation and soil heat flux had a radius of influence between 10² m for the aircraft and 10³ mfor the satellite. Therefore, it appears that the source area for the ground-truth data was several orders of magnitude smaller than the remote sensing observations. This probably contributed to a reduced variability in the remote sensing data and hence model sensitivity.

For estimating H at the reference site, it was discovered that different values of the coefficient in (6), s_kB, were required for each day in order to obtain good agreement with measured fluxes. Values of the coefficient were 0.07 for DOY 156, 0.13 for DOY 209, and 0.25 for DOY 216. The variation in s_kBindicates other factors affecting the radiometric temperature observations, such as the sensor view angle and solar zenith angle (see Table 1) and canopy architecture, are not explicitly treated in (6). These factors can affect the magnitude of kB^-1significantly (Brunet et al., 1991) and thus the value of s_kBrequired in (6). However, a more physically based model for estimating kB^-1will need input parameters that may not be readily available.

With reliable estimates of H_r , H for the other METFLUX sites estimated by (4)and the resulting LE-valuesgiven by (7)are compared to the measurements in Figure 9. The figure shows relatively close agreement between modeled and measured H, albeit the variation in H is relatively small. For LE, model estimates are not as variable as the measurements. This is caused by the fact that model estimates of R_nand G were not as variable as the measured values (see Fig. 8). Still, the values of MPD are within the average difference ( 20%) found among several flux measurement systems (Kustas et al., 1994; Stannard et al., 1994).

Figure Seven

Figure Eight

Extrapolating Instantaneous Fluxes over Time

With estimates of the instantaneous fluxes for each of the METFLUX sites, the daytime average LE-valueswere computed using (8) and the relation for R_ndaytime / R_nmiddayand G _daytime/ G_midday. The result is shown in Figure 10. A MPD < 15% suggests that average daytime fluxes can be estimated satisfactorily with this approach. The average daytime LE-valueswereconverted to daytime total ET (mm) by multiplying the averages by the number of seconds representing the daytime period when most of the evaporation occurred (i.e., 0700-1800MST) and dividing by the latent heat of vaporization.

In order to obtain daily ET, an estimate of the nighttime contribution to the daytime total is needed. Observations from Owe and van de Griend (1990)suggest nighttime ET can be anywhere from 10% to 30%of the daily total with some dependency on soil moisture conditions. This means that daytime ET can represent from 70%to 90%of the daily total. However, the data used by Owe and van de Griend (1990)were probably collected under advective conditions not usually experienced in natural semiarid rangeland environments. Measurements of nighttime ET at the METFLUX sites averaged about 13% of the daily total while ET measurements made by eddy correlation systems at several of the sites (Stannard et al., 1994) gave around a 5%nighttime contribution, on average. Therefore, a conservative value of 10% was assumed to be the contribution of nighttime ET to the daily total. The agreement between daily ET from the model and that determined by the METFLUX sites is illustrated in Figure 11. A satisfactory result is obtained with the relatively small mean bias error (MBE .0.17 mm). This supports the use of a 10% adjustment to the daytime values to account for nighttime ET.

Daily ET at Basin Scale

The above remote sensing model for extrapolating ET in space and time was applied to the whole Walnut Gulch basin using the NOAA-11 AVHRR observations of T_sand NDVI_TOA,for DOYs 209 and 216. For DOY 209, the surface soil moisture was uniformly dry over the watershed. Therefore, variability in ET would mainly be related to the amount and type of vegetation cover. But, for DOY 216, the east-west gradient in rainfall may have caused additional spatial variability in ET. The surface albedo was estimated by correlating NDVI_TOA with ground-based measurements of incoming and reflected solar radiation made at Lucky Hills (shrub-dominated) and Kendall (grass-dominated) sites. A threshold in the value of NDVI_TOA,which essentially indicated whether the pixel was primarily composed of grasses or shrubs, was used to assign a value of a_s of 0.2 for shrubs and 0.15 for grasses. In future analyses a technique similar to Saunders (1990) for converting AVHRR visible and near-infrared radiances to a broad band surface albedo will be used.

Maps of daily ET over the watershed for DOYs 209 and 216 are represented by false color images (Fig. 12) generated by the model at the 1.1 km scale (i.e., using the resampled AVHRR pixel values). For the image on DOY 216, some clouds over the eastern end of the watershed prohibited model estimates for that area. The model output for DOY 209 (Fig. 12a) indicates that, except for a few areas of smaller ET rates near the eastern and western end of the watershed, ET was fairly uniform over the basin. For DOY 216, the general pattern in ET correlates with the distribution in the recent rainfall totals (Fig. 4). In other words, there was higher ET for most of the watershed except for the western end, which had received little precipitation.

Figure Nine

Figure Ten

Figure Eleven

Figure Twelve

The range in daily ET computed by the model for both days is represented in Figure 13 as a distribution using a 0.5 mm /day interval. Average daily ET given by the METFLUX network for DOY 209 was 3.4mm / day and for 216 it was about 4.4mm /day. For both days these values fall in the ET intervals having the highest percentage of pixels computed by the model. Basin averages were computed by simply adding up all the pixels. This yielded a value close to 3.4mm /day and 4.1mm /day for DOYs 209 and 216, respectively. Thus for DOY 209, the average ET for the study area was essentially equal to the basin average. For DOY 216, the basin average ET is slightly lower than the study area average.

This result agrees qualitatively with the observation that significantly higher rainfall totals from the recent precipitation events existed for the study site compared to other areas in the basin (Figure 4). However, the model did not compute any ET-values comparable to ones obtained in the study area just before the rains (i.e., on the order of 2 mm / day). This was expected for pixels at the west end of the watershed since this area received little precipitation from the rains in early August. Some assumptions in the model may have caused this result. For instance, using only the deviation in T_s to quantify the change in H from the reference site [cf. Eq. (4)] is an oversimplification (Novak, 1990). Moreover, the accuracy and resolution of remote sensing data from satellites also will have significant effects on model output. This is because spatial variation in ET computed by the model is predominately a function of the spatial variation in surface properties quantified by the remotely sensed data.

CONCLUSION

An approach for extrapolating ET to the basin scale using AVHRR data and a reference site containing local meteorological data was compared to fluxes estimated at seven other sites during the MONSOON 90 study under three widely different environmental conditions. Empirical relations between midday and daytime EF, R_n, and G allowed the conversion of "instantaneous" ET fluxes to daily values. There was satisfactory agreement between measured and modeled fluxes.

However, the model in its present form cannot account for changes in surface roughness which can significantly affect the estimate of H. In addition, variation in sensor view angle and its effects on the remotely sensed data were not considered. The impact of sensor view angle on the magnitude of T_s may have forced adjustments to the coefficient s_kB in (6) in order to obtain good agreement with measured H. Obviously, this extrapolation scheme will only work for cases where advection can be neglected. Thus it will give reliable estimates only where deviations in near-surface meteorological variables (especially u and T_a) across the watershed are small (Kustas et al., 1990).

Finally, it is shown that the spatial resolution of the remote sensing observations has an impact on the sensitivity of the model output of the fluxes and, hence, in the verification with ground-based measurements. At the scale of NOAA AVHRR data, it may only be possible for the model to compute variations in ET that reflect the influence of large scale (i.e., on the order of 5-10 km) landscape features.

Figure Thirteen

ACKNOWLEDGMENTS

The cooperation and assistance of the USDA-ARS Southwest Watershed Research Center in Tucson, Arizona and on site personnel who maintain the Walnut Gulch Experimental Watershed are gratefully acknowledged. Collection and processing of the aircraft data were performed by Dr. Paul J. Pinter, Jr. and Tom R. Clarke from the USDA-ARS U.S. Water Conservation Lab. The rainfall data were provided by Dr. David C. Goodrich from the USDA-ARS Southwest Watershed Research Center. The false color ET maps (Fig. 12) were generated by Mr. Rob Parry of the USDA-ARS Hydrology Lab. Funding from NASA Interdisciplinary Research Program in Earth Sciences (NASA Reference No. IDP-88-086) and funds from USDA-ARS Beltsville Area Office provided the necessary financial support to conduct this study.

REFERENCES

Ali, M. F., and Mawdsley, J. A. (1987), Comparison of two recent models for estimating actual evapotranspiration using only recorded data, J. Hydrol. 93:257-276.

Arino O., Dedieu, G., and Deschamps, P. Y. (1992), Determination of land surface spectral reflectances using Meteosat and NOAA / AVHRR shortwave channel data, Int. J. Remote Sens.
13:2263-2287.

Bouchet, R. J. (1963), Evapotranspiration reelle et potentielle signification climatique. General Assembly Berkley, Int. Assoc. Sci. Hydrol. 62:134-142.

Bougeault, P., Noilhan, J., Lacarrere, P., and Mascart, P. (1991), An experiment with an advanced surface parameterization in a mesobeta-scale model. Part 1: Implementation, Mon. Weather Rev. 119:2358-2373.

Brest, C. L., and Goward, S. N. (1987), Deriving surface albedo measurements from narrow band satellite data, Int. J. Remote Sens. 8:351-367.

Brunet, Y., Paw U., K. T., and Prevot, L. (1991), Using radiative surface temperature in energy budget studies over plant canopies, in Fifth international Colloquium, Physical Measurements and    Signatures in Remote Sensing, 14-18 January, Courchevel, France.

Brutsaert, W. (1975), On a derivable formula for longwave radiation from clear skies, Water Resour. Res. 11:742-744.

Brutsaert, W. (1982), Evaporation into the Atmosphere: Theory, History and Applications, D. Reidel, Hingham, MA, 299 pp.

Brutsaert, W., and Stricker, H. (1979), An advection-aridity approach to estimate actual regional evapotranspiration, Water Resour. Res. 15:543-550.

Brutsaert, W., and Sugita, M. (1992), Regional surface fluxes under nonuniform soil moisture conditions during drying, Water Resour. Res. 28:1669-1674.

Byrne, G. F., Dunin, F. X., and Diggle, P. J. (1988), Forest evaporation and meteorological data: a test of a complementary theory advection-aridity approach, Water Resour. Res. 24:30-34.

Camillo, P. J. (1991), Using one- or two-layer models for evaporation estimation with remotely sensed data, in Land Surface Evaporation: Measurement and Parameterization (T. J. Schmugge and J.-C. Andre, Eds.), Springer-Verlag, New York, pp. 183-197.

Carlson, T. N., Dodd, J. K., Benjamin, S. G., and Cooper, J. N. (1981), Remote estimation of surface energy balance, moisture availability and thermal inertia, J. Appl. Meteorol. 20:67-87.

Chamberlain, A. C. (1968), Transport of gases to and from surfaces with bluff and wave-like roughness elements, Quart. J. Roy. Meteorol. Soc. 94:318-332.

Cooper, D. I., and Asrar, G. (1989), Evaluating atmospheric correction models for retrieving surface temperatures from the AVHRR over a tallgrass prairie, Remote. Sens. Environ.     27:93-102.

Doraiswamy, P. C., and Perry, E. M. (1991), The relationship between satellite-based surface temperature and vegetation indices over a semi-arid rangeland watershed, AMS Preprint ofthe Tenth Conference on Biometeorology and Aerobiology and Special Session on Hydrometeorology, 10-13 September, Salt Lake City, UT, American Meteorological Society, pp.226-229.

Garratt, J. R. (1992), The Atmospheric Boundary Layer, Cambridge University Press, New York, 316 pp.

Gash, J. H. C. (1987), An analytical framework for extrapolating evaporation measurements by remote sensing surface temperature. Int. J. Remote Sens. 8:1245-1249.

Granger, R. J., and Gray, D. M. (1990), Examination of Morton's CRAE model for estimating daily evaporation from field-sized areas, J. Hydrol. 120:309-325.

Gutman, G. G. (1991), Vegetation indices from AVHRR: an update and future prospects, Remote Sens. Environ. 35: 121-136.

Jackson, R. D., Reginato, R. J., and Idso, S. B. (1977), Wheat canopy temperature: a practical tool for evaluating water requirements, Water Resour. Res. 13:651-656.

Jackson, R. D., Moran, M. S., Gay, L. W., and Raymond, L. H. (1987), Evaluating evaporation from field crops using airborne radiometry and ground-based meteorological data, Irrig. Sci.     8:81-90.

Kneizys, F. X., Shettle, E. P., Abreu, L. W., et a]. (1988), Users Guide to LOWTRAN 7, Environmental Research Papers, Report AFGL-TR-88-0177, Air Force Geophysical Lab.,   Hanscom AFB, Bedford, MA, 137 pp.

Kovaks, G. (1987), Estimation of average area] evapotranspiration-proposal to modify Morton's model based on the complementary character of actual and potential evapotranspiration, J. Hydrol.     95:227-240.

Kustas, W. P. (1990), Estimates of evapotranspiration with a one-and two-layer model of heat transfer over partial canopy cover, J. Appl. Meteorol. 29:704-715.

Kustas, W. P., and Daughtry, C. S. T. (1990), Estimation of the soil heat flux / net radiation ratio from spectral data, Agric. For. Meteorol. 49:205-223.

Kustas, W. P., Choudhury, B. J., Moran, M. S., et al. (1989), Determination of sensible heat flux over sparse canopy using thermal infrared data, Agric. For. Meteorol. 44:197216.

Kustas, W. P., Moran, M. S., Jackson, R. D., et al. (1990), Instantaneous and daily values of the surface energy balance using remote sensing and a reference field in an and environment, Remote Sens. Environ. 32:125-141.

Kustas, W. P., Goodrich, D. C., Moran, M. S., et al. (1991a), An interdisciplinary field study of the energy and water fluxes in the atmosphere-biosphere system over semiarid rangelands: description and some preliminary results, Bull. Am. Meteorol. Soc. 72:1683-1705.

Kustas, W. P., Perry, E. M., and Doraiswamy, P. C., (1991b), MONSOON 90- a remote sensing feasibility study of the energy and water balance of a semiarid basin, preprint from the 42nd    Congress of the International Astronautical Federation, 5-11 October, Montreal, Canada, 6 pp.

Kustas, W. P., Blanford, J. H., Stannard, D. I., Daughtry, C. S. T., Nichols, W. D., and Weltz, M. A. (1994), Local energy flux estimates for unstable conditions using variance data in semi-arid rangelands, Water Resourc. Res. 30:13511361.

Lagouarde, J.-P. (1991), Use of NOAA AVHRR data combined with an agrometeorological model for evaporation mapping, Int. J. Remote Sens. 12:1853-1864.

Massman, W. J. (1992), A surface energy balance method for partitioning evapotranspiration data into plant and soil components for a surface with partial canopy cover, Water Resour. Res. 28:1723-1732.

Moran, M. S., Jackson, R. D., Raymond, L. H., Gay, L. W., and Slater, P. N. (1989), Mapping surface energy balance components by combining LANDSAT Thermatic Mapper and ground-based meteorological data, Remote Sens. Environ. 30:77-87.

Moran, M. S., Kustas, W. P., Bach, L. B., Weltz, M. A., Huete, A.R., and Amer, S. A. (1991a), Use of remotely sensed spectral data for evaluation of hydrologic parameters in semiarid rangeland, AMS Preprint of the Tenth Conference on Biometeorology and Aerobiology and Special Session on Hydrometeorology, 10-13 September, Salt Lake City, UT, American Meteorological Society, pp. 238-241.

Moran, M. S., Kustas, W. P., Vidal, A., Stannard, D. I., and Blanford, J. (1991b), Use of ground-based remotely sensed data for surface energy balance calculations during MONSOON 90, in Proc. IEEE Geosci. and Remote Sensing Symp., Helsinki, Finland, 3-6 June, IEEE, New York, pp. 33-37.

Morton, F. 1. (1971), Catchment evaporation and potential    evaporation: further development of a climatological relationship, J. Hydrol. 12:81-99.

Morton, F. 1. (1983), Operational estimates of area] evapotranspiration and their significance to the science and     practice of hydrology, J. Hydrol. 66:1-76.

Nichols, W. D. (1992), Energy budgets and resistances to energy transport in sparsely vegetated rangeland, Agric. For. Meteorol. 60:221-247.

NOAA (1988a) NOAA polar orbiter data users guide (K. B. Kidwell, Ed.), NOAA-11 update December.

NOAA (1988b) Data extraction and calibration of TIROS-N NOAA radiometer (W. G. Planet, Ed.), NOAA Tech. Memo. NESS 107-Rev.1.

Noilhan, J., and Planton, S. (1989), A simple parameterization of land surface processes for meteorological models, Mon.Weather Rev. 117:538-549.

Noilhan, J., Andre, J. C., Bougeault, P., Goutorbe, J. P., and Lacarrere, P. (1991), Some aspects of the HAPEXMOBILHY Programme: the data base and modelling strategy, Surv.     Geophys. 12:31-61.

Novak, M. D. (1990), Micrometeorological changes associated with vegetation removal and influencing desert formation, Theor. Appl. Climatol. 42:19-25.

Ottl? C., Vidal-Madjar, D., and Girard, G. (1989), Remote sensing applications to hydrologic modeling, J. Hydrol. 105: 369-384.

Owe, M., and van de Griend, A. A. (1990), Daily surface moisture model for large area semi-arid land application with limited climate data, J. Hydrol. 121:119-132.

Panofsky, H. A., and Dutton, J. A. (1984), Atmospheric Turbulence- Models and Methods for Engineering Applications, Wiley, New York, 397 pp.

Parlange, M. B., and Katul, G. G. (1992), An advection-aridity evaporation model, Water Resour. Res. 28:127-132.

Price, J. C. (1983), Estimating surface temperatures from satellite thermal infrared data-a simple formulation for the atmospheric effect, Remote Sens. Environ. 13:353-361.

Price, J. C. (1984), Land surface temperature measurements from the split window channels of the NOAA 7 advanced very high resolution radiometer, J. Geophys. Res. 89:723 17237.

Price, J. C. (1990), Using spatial context in satellite data to infer regional scale evapotranspiration, IEEE Trans. Geosci. Remote Sens. 28:940-948.

Rambal, S., Laeaze, B., Mazurek, H., and Debussche, G. (1985), Comparison of hydrologically simulated and remotely sensed actual evapotranspiration from some Mediterranean vegetation    formations, Int. J. Remote Sens. 6: 1475-1481.

Saunders, R. W. (1990), The determination of broad band surface albedo from AVHRR visible and near-infrared radiances, Int. J. Remote  Sens. 11:49-67.

Seguin, B., and Itier, B. (1983), Using midday surface temperature to estimate daily evaporation from satellite data, Int. J. Remote Sens. 4:371-383.

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, Int. J. Remote Sens. 10:1101-1117.

Shuttleworth, W. J., and Gurney, R. J. (1990), The theoretical relationship between foliage temperature and canopy resistance in sparse crops, Quart. J. Roy. Meteorol. Soc. 116: 497-519.

Shuttleworth, W. J., Gurney, R. J., Hsu, A. Y., and Ormsby, J. P. (1989), FIFE, the variation on energy partition at surface flux sites, in Proc. IAHS Third Int. Assembly, IAHS Publ. No. 186,      pp. 67-74.

Smith, R. C. G., and Choudhury, B. J. (1991), Analysis of normalized difference and surface temperature observations over southeastern Australia, Int J. Remote Sens. 12: 2021-2044.

Soer, G. J. R. (1980), Estimation of regional evapotranspiration and soil moisture conditions using remotely sensed crop surface temperatures, Remote Sens. Environ. 9:27-45.

Stannard, D. I., Blanford, J. H., Kustas, W. P., et al. (1994), Interpretation of surface-flux measurements in heterogeneous terrain during Monsoon '90 experiment, Water Resour. Res.     30:1227-1239.

Stewart, J. B., Shuttleworth, W. J., Blyth, K., and Lloyd, C. R. (1989), FIFE: a comparison between aerodynamic surface temperature and radiometric surface temperature over sparse     prairie grass, AMS Preprint of the 19th Conference Agricultural and Forest Meteorology and 9th Conference Biometeorology and Aerobiology, Charleston, SC, American Meteorological Society, pp. 144-146.

Sucksdorff, Y., and Ottl? C. (1990), Application of satellite remote sensing to estimate area] evapotranspiration over a watershed, J. Hydrol. 121:321-333.

Taconet, 0., Carlson, T., Bernard, R., and Vidal-Madjar, D. (1986), Evaluation of surface /vegetation parameterization using satellite measurements of surface temperature, J. Clim. Appl.    Meteorol. 25:1752-1767.

Tillman, J. E. (1972), The indirect determination of stability, heat and momentum fluxes in the atmospheric boundary layer from simple scalar variables during dry unstable conditions, J. Appl.     Meteorol. 11:783-792.

Weaver, H. L. (1990), Temperature and humidity flux-variance relations determined by one-dimensional eddy correlation, Boundary-Layer Meteorol. 53:77-91.

Wesely, M. L. (1988), Use of variance techniques to measure dry air-surface exchange rates, Boundary-Layer Meteorol. 44:13-21.

Wetzel, P. J., and Chang, J.-T. (1988), Evapotranspiration from nonuniform surfaces: a first approach for short-term numerical weather prediction, Mon. Weather Rev. 116: 600-621.

Wetzel, P. J., Atlas, D., and Woodard, R. (1984), Determining soil moisture from geosynchronous satellite infrared data: a feasibility study, J. Clim. Appl. Meteorol. 23:375-391.

Willmott, C. J. (1982), Some comments on the evaluation of model performance, Bull. Am. Meteorol. Soc. 11:13091313.