1. Introduction
A major objective of the Global Energy and Water Cycle Experiment Continental-scale International Project is to improve the predictive capability of coupled hydrologic–meteorological models. To achieve this, improvements in modeling the large-scale soil moisture dynamics and their interaction with the atmosphere along with observations for verification is essential. Remote sensing and hydrometeorological data collected during the 1997 Southern Great Plains (SGP97) experiment provide observations necessary to explore this interaction. More specific, the aircraft soil moisture mapping using passive microwave radiometry over an approximately 10 000 km2 area at a spatial resolution compatible with known data interpretation (∼1 km) provides a domain large enough to investigate interactions between near-surface soil moisture and regional-scale atmospheric dynamics.
Passive microwave radiometry at wavelengths on the order of 20 cm (L band) have been used for monitoring near-surface soil moisture for a wide range of vegetation cover and climatic conditions (Jackson and Schmugge 1989; Schmugge and Jackson 1994; Jackson et al. 1995). Unfortunately a satellite-based passive microwave system at these long wavelengths will have an inherent spatial resolution problem that is due to technological limitations on antenna size. However, innovative approaches such as synthetic aperture radiometry (LeVine et al. 1994) and other technological alternatives/advances are likely to provide passive L-band microwave soil moisture observations having a resolution on the order of 10–30 km, which is still very useful for regional atmospheric models that typically use a 10-km grid mesh. The SGP97 experimental domain affords the opportunity to evaluate the effects of scaling up near-surface soil moisture to the projected satellite pixel size.
Land surface–atmosphere schemes (LATS) are emerging that can assimilate near-surface soil moisture observations (e.g., Entekhabi et al. 1994; Calvet et al. 1998; Houser et al. 1998; Li and Islam 1999), and other approaches have been developed to use the soil moisture product directly with ancillary data for computing the large-scale heat fluxes (e.g., Kustas et al. 1998, 1999). With either modeling strategy, the effect of such coarse resolution in near-surface soil moisture observations from a satellite-based sensor on model simulations will need to be evaluated. The impact of model grid size and/or resolution of the input data on energy balance predictions has been studied for selected cases (e.g., Famiglietti and Wood 1995; Sellers et al. 1995; Kustas and Jackson 1999). However, to have a satisfactory level of confidence in these results, one must first assess the reliability of model-simulated heat fluxes.
Validation of LATS heat flux predictions has mainly been performed using tower-based flux observations. These tower observations represent an insignificant fraction of the study area; hence, model output of the fluxes for most of the region cannot be validated. Furthermore, there is the issue of mismatch in spatial or pixel resolution of the model output and the effective area contributing to the tower flux measurements, which typically represent a patch with spatial dimensions on the order of 100 m (Schuepp et al. 1990). This mismatch has led others to perform model validation indirectly via other hydrometeorological observations such as soil moisture and rainfall (e.g., Mohr et al. 2000).
The primary objective of this investigation is to evaluate the reliability of a surface energy balance model for computing heat fluxes over the SGP97 experimental domain not only by comparing model output with tower-based flux observations but also by investigating “regional-scale” estimates through use of aircraft-based flux measurements. At the higher (800 m) resolution, the model outputs are compared with flux tower observations made during the experiment. For selected areas, the spatially distributed fluxes computed by the model upwind of the flux aircraft transects are averaged to compare with these regional-scale flux observations.
Given the limitations in the validation of model output of the heat fluxes with observations, it would seem imperative that the uncertainty in LATS surface energy balance computations on a regional-scale basis needs to be investigated by evaluating differences in spatially distributed output among models, similar to the efforts made in the LATS (or more commonly referred to as soil–vegetation–atmosphere transfer, or SVAT) modeling community (Henderson-Sellers et al. 1995). A first step toward a comparison of spatially distributed output is performed by evaluating differences in regionally averaged fluxes between a LATS using high-resolution surface temperature data (∼30 m) and the current model using the 800-m microwave-derived soil moisture data. The utility of the spatially distributed output of other model variables for LATS research is also discussed.
2. Data description
Details of the SGP97 region and experimental design are described by Jackson et al. (1999). At the time of writing, a description of the experiment and all measurement activities additionally was available on the Internet (http://hydrolab.arsusda.gov/sgp97/). This region is exceptionally well instrumented for hydrometeorological research. In Fig. 1, a map of the SGP97 region is illustrated along with the three main facilities (Central Facility, El Reno, and the Little Washita Watershed Facility), the Oklahoma Mesonet, the United States Department of Agriculture (USDA) Agricultural Research Service Micronet, the Department of Energy Atmospheric Radiation Measurement (ARM) Program Cloud and Radiation Testbed study region, and the ARM atmospheric boundary layer facilities.
The L-band passive microwave data were collected using the Electronically Scanned Thinned Array Radiometer (ESTAR) flown by the P-3B aircraft operated by the National Aeronautics and Space Administration (NASA) Wallops Flight Center. ESTAR observations were made over a 30-day period from 18 June [day of year (DOY) 169] to 17 July (DOY 198). The instrument was installed to provide horizontally polarized data. Recent experiments, such as Washita '92, at the watershed scale have demonstrated the reliability of this instrument (Jackson et al. 1995). Details of the processing of the ESTAR data are given in Jackson et al. (1999). The algorithm for converting the microwave brightness temperatures to volumetric soil moisture requires other remotely sensed information also used by the model. A land cover classification is needed for estimating roughness parameters and was taken from an analysis performed by Doraiswamy et al. (1998) using a Landsat Thematic Mapper (TM) scene and on-site surveys as part of this supervised approach. The normalized difference vegetation index (NDVI) was also computed for the Landsat TM scene collected on 25 July 1997 (DOY 206) and integrated to the 800-m pixel size. A soil texture database was derived from the State Soil Geographic Database (STATSGO) developed by the USDA Natural Resources Conservation Service. One of the products available is a soil texture classification of the surface soil on a 1-km grid, which was resampled to the 800-m grid. Thus land cover, NDVI, and soil texture were integrated into Geographic Information System layers for the SGP97 ESTAR domain.
Meteorological data, which include air temperature, wind speed, solar radiation and relative humidity, were defined for each 800-m pixel using the Mesonet (Fig. 1). For the main region covered by the ESTAR observations, data from 15 stations were used in an interpolation scheme for defining these forcing variables for each pixel. The wind speed observations for all stations were available at 10 m above ground level (AGL) and air temperature was available at 1.5 m AGL. Details of the measurements and quality control of the data are described by Shafer et al. (2000).
Tower flux observations were made at the three main facilities by using Bowen ratio and eddy covariance techniques. Half-hourly averaged fluxes were recorded. Twine et al. (2000) provide detailed description of the observations and an analysis of the measurement uncertainties. A key finding was that the sum of eddy covariance measurements of sensible heat H and latent heat LE fluxes was consistently less than the available energy, namely, net radiation Rn less soil heat flux G. This lack of “closure” in the surface energy balance ranged from 10% to 30% as quantified by taking the ratio of (H + LE)/(Rn − G), suggesting that, when available energy measurement errors are known and modest, the eddy covariance measurements of sensible and latent heat fluxes should be adjusted to obtain closure.
Aircraft flux and atmospheric boundary layer profiling observations were made by the National Research Council Canada Twin Otter atmospheric research aircraft at multiple locations and altitudes within the SGP97 study area. This aircraft has been involved in numerous field campaigns, including large-scale field experiments over grasslands with moderate topography [First International Satellite Land Surface Climatology Project Field Experiment (FIFE); Sellers et al. 1992a] and forests (Boreal Ecosystem–Atmosphere Study; Sellers et al. 1997). A summary of the aircraft observations, instrument description, data processing, and preliminary flux estimates for the run-averaged data for each flux run is given in the data report (MacPherson 1998). At the time of writing, a description of the data was also available on the Internet (http://daac.gsfc.nasa.gov/CAMPAIGN_DOCS/SGP97/air_boundary.html). Besides the flux instrumentation on the Twin Otter aircraft, a nadir-viewing Barnes Precision Radiometer Thermometer-5 (PRT-5) infrared radiometer was present, measuring surface temperature along the flight line. These data will be compared with model-simulated values over the transects described below.
Two of the flight lines with multiple flux runs over the course of the experimental period are used in the current validation study. One flight track, ∼15 km in length, was over the El Reno facility, with the aircraft flying essentially east–west (called the El Reno or ER track). Another flight track, ∼27 km in length, was north-northeast of the El Reno facility and was a north–south transect (called the Kingfisher or KF track). For both ER and KF tracks, the low-level flux runs were flown at an altitude of approximately 30 m AGL. Preliminary comparisons of tower-based and aircraft-based sensible and latent heat flux estimates have been made (MacPherson et al. 1999a,b). The comparisons indicate that, when using segments of the flight legs having similar land cover in the upwind fetch to that of the towers, there is good agreement in heat flux predictions.
The approximate locations of the Mesonet stations used in the interpolation of the meteorological data, flux towers, and aircraft transects are displayed in Fig. 2 along with the land use/land cover map. In Table 1 is a list of the land cover type and facility location associated with the flux towers used in the validation of model output.
In Table 2, a list is provided of days having ESTAR observations commensurate with either tower and/or aircraft flux observations and general near-surface soil moisture conditions that existed at the three main facilities and the region surrounding the KF transect. Because tower-based flux observations ran more or less continuously, they were available for essentially all ESTAR flights. The ESTAR flights were typically conducted over the period of 0800–1300 central standard time (CST). Hence for this preliminary validation study, meteorological input data and comparisons with the flux tower observations for the half-hourly period 1000–1030 CST were used. The 1030 CST output time was reasonably close to the mean time of ESTAR coverage (i.e., 1000–1015 CST) and provides adequate time from the morning transitional period for a well-developed turbulent atmospheric boundary layer to exist, providing suitable conditions for flux measurements and model application. Additional model runs for other output times were performed on days for which aircraft flux observations were available for other time periods (see Table 2).
3. Model description and parameter estimation
a. Model description
A LATS originally developed to use remotely sensed surface temperature for estimating component heat fluxes from both the soil and vegetation canopy [i.e., a two-source model; Norman et al. (1995)] has been recently modified to employ microwave-derived near-surface soil moisture as the key surface boundary condition for estimating component fluxes.
The Two-Source Energy Balance model using near-surface soil moisture as the key boundary condition (TSEBSM) is described in Kustas et al. (1998, 1999). A schematic (Fig. 3) illustrates the resistance network of the model, where the subscripts c and s indicate canopy and soil surface, respectively. The symbols Ta and ea are the temperature and vapor pressure of the air above the plant canopy; eac, ec, and es are the vapor pressure of the air within the canopy, the air on the leaf surface, and the air on the soil surface; and Tac is the temperature in the canopy air space. Symbols Tc and Ts are the temperatures of the leaves and the soil surface. The resistance symbols ra, rac, and ras represent the aerodynamic resistance of the air above the canopy, the leaf boundary layer, and the soil surface boundary layer, respectively, to water vapor and sensible heat transfer. The resistance symbol rss represents the surface soil resistance to water vapor transfer within the soil. The parameter e∗(Ts) is the saturation vapor pressure at soil surface temperature Ts, and hr is the relative humidity of the soil layer (see below). Last, W is the volumetric soil water content of the near-surface soil layer representative of the 0–5-cm layer.
b. Model parameter estimation
The TSEBSM model requires several vegetation parameters that help to determine surface momentum roughness z0M and displacement height d0 (Brutsaert 1982; Raupach 1994), wind speed, and resistances inside the canopy and above the soil surface (Norman et al. 1995). These parameters include vegetation height hc, leaf width lw, and fractional vegetation cover fc or LAI. A detailed ground survey of vegetation type and cover was conducted over the SGP97 region (http://daac.gsfc.nasa.gov/CAMPAIGN_DOCS/SGP97/sgpvegdatrptv3b.html); hence, these data were used to define values of hc and lw.
The soil texture–dependent parameters Ψs and b required in Eq. (7) and values of Ws were estimated from Clapp and Hornberger (1978) using the STATSGO near-surface soil texture information, which was resampled to 800 m for estimating soil moisture with the ESTAR data.
c. Overview of simplifications in LATS formulations
The TSEBSM model was designed to be a physically based approach, yet to be relatively simple for large-scale applications for which detailed soils and vegetation data are not routinely available. As a result, several of the simplifications in model formulations are at times likely to cause significant discrepancies between computed and observed fluxes. For example, using a constant Priestley–Taylor coefficient αPT for all vegetation types may not be appropriate given that plant physiological differences exist and hence affect the magnitude of αPT (McNaughton and Jarvis 1991); this was actually observed by Kustas et al. (1996) during Washita '92. On the other hand, the model will not permit nonphysical solutions such as LEs < 0 or condensation during the daytime. In this case, the Priestley–Taylor approximation is dropped, and several approximations are used, including assigning LEs = 0, which allows for direct computation of Hs(=Rns − G); then by adopting a parallel resistance network [see Fig. 1 in Norman et al. (1995)], this allows direct computation of the aerodynamic resistances of the soil, canopy, and surface layer. With Hs, Ta, and soil aerodynamic resistance, Ts is estimated and Tc is derived from Eq. (3). Then with Ta, Tc, and canopy-surface-layer aerodynamic resistance, Hc is computed, and LEc = Rnc − Hc.
Capehart and Carlson (1997) using a soil profile model show a significant “decoupling” between surface soil moisture (≈0.5 cm) and the moisture at 5 cm as the soil dries, suggesting that the surface energy balance is more strongly coupled to surface moisture conditions than at deeper layers. This is consistent with the discrepancies observed by Calvet et al. (1998) when comparing their LATS-derived surface temperatures with radiometric temperature measurements. However, both the soil hydraulic properties and the antecedent moisture conditions greatly influence the degree of decoupling between these layers (Capehart and Carlson 1997). Furthermore, comparisons between moisture availability simulated by a LATS model using remotely sensed surface temperature and area-averaged 0–5-cm soil moisture observations show significant correlation (Gillies et al. 1997). This is in contrast to similar comparisons made with individual ground-based samples (Carlson et al. 1995) in which spatial variability in soil moisture can add considerable noise to such relationships (Famiglietti et al. 1999).
In the current formulation, the surface soil resistance rss to water vapor transport, estimated from single near-surface soil moisture observation, is assumed to be constant for the daytime period. This is an oversimplification of the temporal and functional behavior observed in other studies (e.g., Daamen and Simmonds 1996; van de Griend and Owe 1994). However, in practice, errors from computing daily heat fluxes using a single midmorning near-surface soil moisture observation in the model were found to be minor in comparison with other assumptions, such as using a uniform radiation field over the basin (Kustas et al. 1998, 1999).
Although more sophisticated soil evaporation models have been proposed (e.g., Yamanaka et al. 1997), they typically require estimates of model parameters not routinely available over large areas. For example, the model of Yamanaka et al. (1997) requires an estimate of the depth of the wetting front, which cannot be measured with remote sensing.
More physically based approaches linking microwave emission and transfer through the soil to a model of simultaneous heat and water flow in the soil profile (e.g., Microwave Emission—Soil, Water, Energy, and Transpiration (MICRO-SWEAT); Burke et al. 1997; Simmonds and Burke 1998) have been proposed. These models, however, are applicable only at the patch scale and have not been modified to work at the landscape or regional scales. Moreover, a sensitivity analysis of MICRO-SWEAT by Burke et al. (1997) indicates that model predictions are seriously affected by errors in Rn: an uncertainty of ±10 W m−2 in Rn translates to an ∼5-K error in the model-predicted microwave brightness temperature. Discrepancies between observed and remotely sensed estimates of Rn are typically 5%–10% or ∼50 W m−2 near midday (Kustas and Norman 1996), which would translate into an uncertainty in microwave brightness temperature of approximately 25 K. For certain conditions, such an uncertainty in microwave brightness temperature will be unacceptable.
4. Results and discussion
a. Flux footprint issues
To compare either the tower- or aircraft-based fluxes with the model output is not straightforward, because of variations in the flux footprint or source area affecting the measurements, which depend on surface roughness, wind speed, atmospheric stability, and other turbulence characteristics (Schmid 1994). The flux footprints are determined from diffusion models, with some of the analytical approaches being fairly easy to implement (e.g., Schuepp et al. 1990; Horst and Weil 1992). However, because of the relatively coarse resolution of the model output (800 m), such approaches are not feasible with the tower observations. At the typical 2-m AGL measurement height under unstable daytime conditions, these fluxes are affected by an upwind fetch on the order of 100–200 m, with the maximum contribution coming from an upwind area ∼50 m from the tower (Schuepp et al. 1990). This means that the tower-based measurements are sampling at subpixel resolution. With the aircraft flux observations, the flux footprint upwind of the flight leg is significantly larger; however, the influence of the boundary layer–scale eddies and other larger-scale turbulent processes significantly complicates the interpretation of the aircraft flux data and any simple methods of determining source areas (Mahrt 1998).
Another difficulty, which is related to the flux footprint issue, comes from selecting the appropriate pixels of model output to compare with the flux observations, particularly for the tower-based fluxes. In this case, the closest pixel to the tower may actually be in a field having a totally different land use (e.g., the tower is in a bare soil field and the closest pixel covers a heavily vegetated rangeland site). Therefore, the procedure was to select all pixels within 1 km of the tower and average all the pixels having similar land cover. For the aircraft transects, the pixels selected were south and slightly west of the ER flight line, because winds were predominantly from the south and southwest. The rectangular “box” around the El Reno facility (see Fig. 2) represents the flux footprint for the aircraft and is approximately 12.5 km east–west by 2 km north–south. Because the aircraft flew farther west than the area covered by ESTAR, 22 pixels of a potential 32 were used in averaging to compare with the run-average fluxes from the aircraft. For the KF transect, the comparison with the aggregated pixels is less straightforward because the flights were north–south, in the direction of the prevailing winds (see Fig. 2). In this case, a rectangular box approximately 1.5 km by 26 km was defined, resulting in a total of 31 pixels used in the aggregation to compare with the aircraft.
From scalar conservation requirements, Raupach (1995) shows formally that the area-averaged scalar fluxes, H and LE, require elemental flux densities (i.e., fluxes from each surface type) to be weighted by the fraction of area occupied and summed across the landscape. This is done implicitly using a model relying on remotely sensed data to define land use/land cover and fractional vegetation cover. If it is assumed that horizontal fluxes between surface types are small in comparison with vertical fluxes, then, by energy conservation, Rn and G also average linearly. Model simulations suggest that this assumption is reasonable for a fairly wide range of conditions (Raupach and Finnigan 1995).
b. Comparison with tower-based flux observations
In Fig. 4, comparisons of the four energy balance components are made with the flux tower observations from the three main study areas, namely, the Little Washita Watershed, El Reno, and the Central Facility. The agreement between modeled and observed energy fluxes is reasonable, with the points generally scattered about the one-to-one line. The summary statistics are illustrated in Fig. 5, which shows mean observed values along with the root-mean-square difference (rmsd) between observed and model-estimated energy fluxes (Willmott 1982). Relative to the mean, the rmsd values are not insignificant but are acceptable (see below), except for soil heat flux. Also computed was the mean absolute percent-difference statistic (MAPD); this is computed by taking the average of the absolute differences between model and observed, divided by observed and multiplied by 100 to yield a percentage. The scatter in Fig. 4 yields MAPD values of 6% for Rn, 60% for G, 40% for H, and 18% for LE, suggesting significant discrepancies in G and H but values within measurement uncertainty for Rn and LE (Twine et al. 2000).
The relatively large discrepancies in G are not unexpected given that the comparison involves G observations representative of a few-square-meter patch close to the flux tower being compared with the 800-m grid-resolution output of the model. For H, MAPD of about 40% is larger than the typical measurement uncertainty (i.e., ∼25%), yet this level of disagreement is not inconsistent with other comparisons made between regional-scale models and individual-tower flux observations over this study area (e.g., Mecikalski et al. 1999; Mohr et al. 2000) and points to the difficulty in resolving the scale mismatch between model and observations (e.g., Norman et al. 2000).
Investigation of the discrepancies between modeled and observed heat fluxes for individual sites indicates that two of the three winter wheat sites (ER13 and CF02) had the largest uncertainty, with rmsd values on the order of 80 W m−2 for H and 100 W m−2 for LE. These larger discrepancies in fluxes at ER13 are due in part to the fact that nearly one-half the field was pasture grass and the other one-half was bare soil, thus making it difficult to have a single 800-m bare soil pixel to compare with the observations. The CF02 site and surrounding region contained wheat stubble that was rapidly becoming revegetated with weeds caused by frequent rainfall during the study period. Therefore, using the land use designation of “winter wheat” to obtain 800-m pixels near the CF02 flux tower yielded a relatively wide range in vegetation cover conditions and hence model-computed heat fluxes, due to different stages of revegetation in the various winter wheat fields.
c. Comparison with aircraft-based flux and surface temperature observations
For the comparisons with the run-average aircraft-based Rn, H, and LE estimates, multiple runs made by the aircraft for the time period were averaged (a single run took between 5 and 10 min). This usually resulted in computing a mean for two to three run-averaged flux values to compare with a half-hourly average output from the model. Comparison of the regional heat fluxes suggests reasonable agreement between model and observations, albeit a notable bias for the KF transect (Fig. 6), for which the average difference between modeled and observed H was about +60 W m−2 and was about −35 W m−2 for LE. This bias cannot be attributed to lack of closure by the aircraft observations given that both ER and KF had closure values averaging ∼0.9 when using G estimated by the model with Rn, H, and LE from the aircraft. The average fractional vegetation cover estimated for these transects was approximately 0.4 for ER and 0.25 for KF, which translates to an effective regional LAI of about 1 for ER and about 0.5 for KF. A significant bias in the estimated fractional cover could result in higher H and lower LE estimated by the model or it may suggest the bare-soil evaporation algorithm is underpredicting in some cases.
The rmsd values combining both the ER and KF transects were reasonable when compared with the average heat fluxes observed (see Fig. 5). Rmsd values were approximately 30 W m−2 for Rn, and about 50 W m−2 for H and LE. The MAPD values were ∼5% for Rn, 30% for H, and 15% for LE. The better agreement between modeled and observed heat fluxes with the aircraft-based versus tower-based measurements may be due to several factors already discussed that relate to the spatial resolution of the model output and the area sampled by the tower-based flux sensors. The significant mismatch in spatial resolution between the relatively coarse model output (800-m pixel resolution) and the “subpixel resolution” of the tower fluxes (i.e., the tower-based measurements represent an approximately 100-m upwind patch) is not an issue with the aircraft flux data. The modeled fluxes actually need to be aggregated or averaged over multiple pixels (i.e., ∼20–30 pixels) along the aircraft transect. This averaging not only reduces the variability in the modeled fluxes but also avoids the issue of trying to provide model output that matches both the land cover type and fractional cover condition being sampled at subpixel resolution by the tower-based measurements.
For the same set of aircraft transects, surface temperature observations from the PRT-5 radiometer averaged over the flight lines were compared with the simulated Tsurf values from TSEBSM averaged over the same flight leg. The aircraft data were “adjusted” for atmospheric attenuation and emissivity effects based on comparisons with thermal infrared imagery corrected for such effects over the ER transect (French et al. 2000). This indicated that a correction on the order of +1.5°C was needed, which is comparable to what has been applied to low-altitude aircraft radiometric temperature observations in other studies (e.g., Kustas et al. 1994; Goetz et al. 1995). The comparison is illustrated in Fig. 7 and shows good agreement between the TSEBSM-simulated and observed Tsurf values over the ER and KF transects, with an rmsd of ∼1.2°C. This result provides further evidence that the model parameterizations for estimating soil evaporation/temperature and vegetation transpiration/temperature components are reliable for the environmental conditions existing during SGP97.
d. Model sensitivity to uncertainty in remotely sensed inputs
Kustas et al. (1998) evaluated model sensitivity to the uncertainty expected in key input variables provided by remote sensing, namely, LAI and W. They found that, with the likely uncertainty in remotely sensed LAI (±50%) and the ±30% uncertainty in remotely sensed W, both inputs could cause variations in flux predictions of ≈5% for Rn, ≈15% for G, and ≈25% for H and LE, on average. With modeled H for both the tower and aircraft comparisons averaging about 150 W m−2 and LE averaging about 300 W m−2, a 25% uncertainty would translate into approximately ±35 and ±75 W m−2 variation in modeled H and LE. Thus the discrepancies between modeled and observed H and LE for both the tower and the aircraft fall within the likely uncertainty in these key model parameters.
e. Comparison with a LATS using radiometric temperature observations
A final comparison is made with the related LATS model using remotely sensed surface temperature, namely the TSEB with radiometric temperature (TSEBTR; Norman et al. 1995). The TSEBTR model was run over the El Reno area using thermal infrared and visible near-infrared data acquired on DOY 183 at ∼1015 CST (French et al. 2000). The thermal infrared data for estimating radiometric surface temperature came from the airborne Thermal Infrared Multispectral Scanner (TIMS), and the Thematic Mapper Simulator (TMS) visible near-infrared imagery was used to create an NDVI map for the region. The Landsat TM imagery from DOY 206 was used to derive a land cover/land use product. Because the land use information was at the 30-m pixel resolution, the higher-pixel-resolution (∼10 m) TIMS and TMS data were aggregated to 30 m. French et al. (2000) extracted several pixels surrounding the four flux towers (ER sites) and found that TSEBTR modeled heat flux predictions agreed to within 50 W m−2 of the tower-based observations, on average. In comparing the two model predictions, the TSEBSM model output for the area covered by the TIMS and TMS instruments (∼9 km east–west by 4.5 km north–south) was extracted, yielding 64 TSEBSM model pixels. The energy balance components of the 64 pixels were arithmetically averaged and were compared with the arithmetic average of the TSEBTR model output, which at 30-m resolution contained nearly 44 000 pixels. As discussed above, this simple linear averaging of the energy balance components is supported by the theoretical work of Raupach (1995).
In Fig. 8, the comparison of the regional average energy balance components for the day analyzed by French et al. (2000) is remarkably good, especially considering that the TSEBTR model uses radiometric temperature as the key boundary condition instead of ESTAR-derived near-surface soil moisture, and with a pixel resolution of about 1/25 the size of the ESTAR data. The average fractional vegetation cover estimated for the area is very similar, with the 64 TSEBSM model pixels yielding an average fc ≈ 0.48, whereas the 44 000 TSEBTR model pixels yielded an average fc ≈ 0.43. The average Tsurf simulated by TSEBSM was 36.8°C, which is 1°C higher than the average observed for the image (Tsurf ≈ 35.8°C). Given that differences in Tsurf between TIMS and tower-based observations were about 1°C, the discrepancy between the TSEBSM-simulated and TIMS-observed Tsurf is within the level of uncertainty in the Tsurf measurement (French et al. 2000).
f. Utility of TSEBSM output
The spatially distributed meteorological input and surface energy balance output products from this model provide unique spatial information for LATS research. In particular, attempts at comparing surface fluxes from regional-scale models with individual tower observations typically result in significant discrepancies that are difficult to resolve objectively (Jiang and Islam 1999; Mohr et al. 2000). Furthermore, these tower observations represent an insignificant fraction of the study area; hence, model output of the fluxes for most of the region cannot be validated. The use of aircraft-based flux measurements can provide regional-scale validation, potentially reducing scale mismatch issues and registration problems between model grid/pixel output and tower observations, but the footprint issue at the larger scale still exists, and there is no way to validate individual pixel values. Therefore, it would seem imperative that the uncertainty in LATS surface energy balance computations on a regional-scale basis needs to be investigated by evaluating differences in spatially distributed output among models. There has been little, if any, attempt to perform model intercomparisons of their spatially distributed output.
An example of spatially distributed model input of the meteorological and soil moisture data for the SGP97 region (Fig. 9) and resulting output of spatially distributed energy fluxes (Fig. 10) is illustrated for DOY 183 at the 1030 CST output time. From Fig. 9, one observes significant spatial variability in wind speed, soil moisture, and screen-level air temperature under relatively uniform solar radiation conditions that existed over the SGP97 region for the 1000–1030 CST period. A large area surrounding and to the north of the Central Facility had the highest wind speeds and near-surface soil moisture from recent rains and the lowest air temperatures, while a significant area between El Reno and the Central Facility had the highest air temperature, lowest wind speed, and some of the lowest near-surface soil moisture values. The region around the Little Washita Facility also had low near-surface soil moisture, while the area around El Reno was still relatively wet (see Table 2). Both the El Reno and Little Washita facilities had air temperatures and wind speeds between the two extremes that existed to the north. This variation in moisture and meteorological conditions is reflected in the energy balance components (Fig. 10), particularly in the turbulent heat fluxes, H and LE, for which the highest H and lowest LE values are computed for this same hot, dry region between El Reno and the Central Facility as well as dry areas around the Little Washita. In contrast, the wetter areas have significantly lower H and higher LE values.
The spatially distributed Tsurf simulated by TSEBSM is illustrated along with the soil moisture map for this same day (Fig. 11). Clearly there is a qualitative correlation that indicates areas that are generally wetter have lower Tsurf; however, other factors such as the fractional vegetation cover and meteorological conditions have a significant effect on model results. For example, some of the highest Tsurf values came from the region having the lowest wind speeds and highest air temperatures (see Fig. 8), while an area of low W around the Little Washita facility does not have correspondingly high Tsurf values but does have relatively high H (Fig. 10). This same area around the Little Washita facility is primarily range land, pasture, and forage (see Fig. 2), resulting in fractional vegetation covers fc of greater than 0.65, which will strongly modulate the range in Tsurf via Eqs. (3) and (4).
A plot of simulated Tsurf versus W shows a triangular shape (Fig. 12) with the apex having the lowest Tsurf values and highest values of W. The range in Tsurf values increases with decreasing soil moisture, because the amount of vegetation cover also significantly affects Tsurf via the canopy temperature, Tc; Tc does not vary as significantly as the soil temperature Ts because of Eq. (4) and significant differences in the magnitudes of aerodynamic resistances of the canopy and soil (Norman et al., 1995). When Tsurf values simulated by TSEBSM are plotted versus NDVI, the resulting figure (Fig. 13) also has a triangular or trapezoid shape very similar to what has been observed with remote sensing data analyzed over this same region during SGP97 (Jiang and Islam 1999).
In fact, by having a model that simulates surface temperature, comparisons with remotely sensed surface temperature such as observed by Jiang and Islam (1999) in NDVI–Tsurf space can be used to indicate when model parameters need adjustment. With a relatively simple model such as TSEBSM, there are only a handful of parameters that can be modified, the primary one being the Priestley–Taylor coefficient. However, a data assimilation scheme using TSEBSM-derived and remotely sensed surface temperatures will need to consider how the effects of decoupling between surface and near-surface soil moisture will affect simulated surface temperatures from such a model (Capehart and Carlson 1997). The preliminary comparisons made here between simulated Tsurf and the PRT-5 observations from the Twin Otter (cf. Fig. 7) and the airborne TIMS indicate that TSEBSM simulates surface temperatures that are within the uncertainty in remotely sensed surface temperature observations.
Another potential application of the NDVI–Tsurf output would be to use it as input to LATS models that utilize the NDVI–Tsurf relationship to define key model boundary conditions and/or parameters (e.g., Moran et al. 1994; Gillies and Carlson 1995; Jiang and Islam 1999). In particular, the so-called triangle method uses this relationship between observed Tsurf and NDVI to invert the LATS and predict near-surface soil moisture and heat fluxes (Gillies et al. 1997). The output in Fig. 13 affords a very efficient way to compare flux predictions from various LATS that rely on these data as input, because the same set of boundary conditions can be used, and the output from the different models can be compared on a pixel-by-pixel basis.
5. Concluding remarks
This preliminary analysis of comparing the TSEBSM model heat flux predictions with tower- and aircraft-based heat fluxes suggests the model is performing satisfactorily in general. There do appear to be larger discrepancies in the partitioning of available energy into sensible and latent heat fluxes for the tower-based sites and aircraft transects having lower vegetation cover. These discrepancies may be caused by errors in soil moisture and fractional vegetation cover estimates as well as limitations in the soil evaporation parameterization. A further complication is the mismatch in the scale of the observations and model resolution, which is a bigger problem with the tower observations given that they are sampling at the subpixel resolution. Further studies using field-scale soil moisture data derived from ground sampling networks (Famiglietti et al. 1999), which are at a more appropriate spatial resolution for comparing model flux computations with the tower-based observations, are planned to investigate this issue.
Comparison with an LATS having similar modeling structure but using radiometric surface temperature collected at significantly higher pixel resolution (French et al. 2000) yielded area-average heat fluxes in close agreement with TSEBSM. In addition, the area-average observed radiometric surface temperature differed by 1°C with TSEBSM-simulated area-average value. Additional evidence of good agreement between TSEBSM-simulated Tsurf and observed surface temperature, yielding an rmsd of 1.2°C, came from comparisons with the Twin Otter PRT-5 measurements averaged over the ER and KF transects (Fig. 7).
Although TSEBSM is a relatively simple model, it contains the essential physics governing energy exchanges from the two primary sources, soil and vegetation, and hence may help to interpret some empirical results obtained recently using multiregression fits between the heat fluxes and remotely sensed parameters for the SGP97 region (MacPherson et al. 1999b). Their preliminary analysis indicates that the remotely sensed soil moisture data were not a major factor in accounting for the variance in the aircraft-based H and LE observations. The multiregression models, based on remotely sensed Rn, Tsurf − Ta, and a “greenness index,” estimated H and LE with standard errors on the order of 25 and 45 W m−2, respectively. However, the W–Tsurf and NDVI–Tsurf relationships illustrated in Figs. 12 and 13 suggest that the impact of near-surface soil moisture on the heat fluxes is strongly modulated by the amount of vegetation cover. Thus the correlation between W and the heat fluxes will depend on the range in fractional canopy cover and moisture conditions existing in the specific area selected.
For example with DOY 183, 1030 CST, output, which had a wide range in soil moisture conditions, the correlation coefficients R between H and LE computed by the model versus W for the whole SGP97 region were about −0.70 and 0.52, respectively. For the ER transect, there were stronger correlations between H and LE computed by the model versus W, with R ≈ −0.83 and 0.81, respectively; yet, the correlations are significantly lower for the KF transect where R ≈ −0.49 and 0.33, respectively. This example clearly demonstrates that the area selected as well as moisture conditions for a particular day can strongly influence the correlations between the heat fluxes and near-surface soil moisture (Mohr et al. 2000).
The heat flux and surface temperature “maps” simulated by the TSEBSM model can be made available for comparison with output from other LATS models using remote sensing data such as from the Advanced Very High Resolution Radiometer (Gao et al. 1998; Jiang and Islam 1999), the Geostationary Operational Environmental Satellite (e.g., Anderson et al. 1997; Mecikalski et al. 1999), and the Special Sensor Microwave Imager (e.g., Lakshmi et al. 1997). This output can also be compared with regional-scale simulations using models from the SVAT community (e.g., Mohr et al. 2000). The current approach provides independent estimates of the heat fluxes to compare with the models that utilize the NDVI–Tsurf relationship on a pixel-by-pixel basis (e.g., Moran et al. 1996; Gillies et al., 1997). This is impossible to do routinely with tower-based heat flux observations.
Acknowledgments
This work was supported by the NASA EOS and Land Surface Hydrology programs. In particular, funding under NASA NRA 98-OES-11 supported this research investigation. The Southern Great Plains 1997 Hydrology Experiment involved the efforts of over 100 individuals. Collecting the data necessary for this particular part of SGP97 was only possible because of the exceptional efforts of the soil moisture and vegetation sampling teams, tower and aircraft flux teams, and the NASA Wallops Flight Center P-3B aircraft team that made the ESTAR mission possible. We thank Dr. Ming Ying Wei of NASA Headquarters, whose guidance and support made this project of greater value than its individual components. All data collected as part of the SGP97 have been compiled into a public archive as part of the NASA Distributed Active Archive Center at Goddard Space Flight Center at the following Internet site: http://daac.gsfc.nasa.gov/CAMPAIGN_DOCS/SGP97/sgp97.html.
REFERENCES
Anderson, M. C., Norman J. M. , Diak G. R. , Kustas W. P. , and Mecikalski J. R. , 1997: A two-source time-integrated model for estimating surface fluxes from thermal infrared satellite observations. Remote Sens. Environ., 60 , 195–216.
Brutsaert, W., 1979: On a derivable formula for long-wave radiation from clear skies. Water Resour. Res., 11 , 742–744.
——,. 1982: Evaporation into the Atmosphere. D. Reidel, 299 pp.
Burke, E. J., Gurney R. J. , Simmonds L. P. , and Jackson T. J. , 1997: Calibrating a soil water and energy budget model with remotely sensed data to obtain quantitative information about the soil. Water Resour. Res., 33 , 1689–1697.
Calvet, J-C., Noilhan J. , and Bessemoulin P. , 1998: Retrieving the root-zone soil moisture from surface soil moisture or temperature estimates: A feasibility study based on field measurements. J. Appl. Meteor., 37 , 371–386.
Camillo, P. J., and Gurney R. J. , 1986: A resistance parameter for bare soil evaporation models. Soil Sci., 141 , 95–105.
Capehart, W. J., and Carlson T. N. , 1997: Decoupling of surface and near-surface soil water content: A remote sensing perspective. Water Resour. Res., 33 , 1383–1395.
Carlson, T. N., and Ripley D. A. , 1997: On the relation between NDVI, fractional vegetation cover, and leaf area index. Remote Sens. Environ., 62 , 241–252.
——, Gillies, R. R., and Schmugge T. J. , 1995: An interpretation of methodologies for indirect measurement of soil water content. Agric. For. Meteor., 77 , 191–205.
Choudhury, B. J., 1987: Relationships between vegetation indices, radiation absorption, and net photosynthesis evaluated by a sensitivity analysis. Remote Sens. Environ., 22 , 209–233.
——, Ahmed, N. U., Idso S. B. , Reginato R. J. , and Daughtry C. S. T. , 1994: Relations between evaporation coefficients and vegetation indices studied by model simulations. Remote Sens. Environ., 50 , 1–17.
Clapp, R. B., and Hornberger G. M. , 1978: Empirical equations for some soil hydraulic properties. Water Resour. Res., 14 , 601–604.
Daamen, C. C., and Simmonds L. P. , 1996: Measurement of evaporation from bare soil and its estimation using surface resistance. Water Resour. Res., 32 , 1393–1402.
Doriaswamy, P., Stern A. J. , and Cook P. W. , 1998: A classification technique for mapping biophysical parameters in the U.S. southern Great Plains. Proc. Int. Geoscience and Remote Sensing Symp., Seattle, WA, IEEE 007803-4403, 862–865.
Entekhabi, D., Nakamura H. , and Njoku E. G. , 1994: Solving the inverse problem for soil moisture and temperature profiles by sequential assimilation of multifrequency remotely sensed observations. IEEE Trans. Geosci. Remote Sens., 32 , 438–448.
Famiglietti, J. S., and Wood E. F. , 1995: Effects of spatial variability and scale on areally averaged evapotranspiration. Water Resour. Res., 31 , 699–712.
——, and. Coauthors. . 1999: Ground-based investigation of soil moisture variability within remote sensing footprints during the Southern Great Plains 1997 (SGP97) Hydrology Experiment. Water Resour. Res., 35 , 1839–1851.
French, A. N., Schmugge T. J. , and Kustas W. P. , 2000: Surface fluxes over the SGP site with remotely sensed data. Phys. Chem. Earth, 25B , 167–172.
Gao, W., Coulter R. L. , Lesht B. M. , Qiu J. , and Wesely M. L. , 1998: Estimating clear-sky regional surface fluxes in the Southern Great Plains Atmospheric Radiation Measurement site with ground measurements and satellite observations. J. Appl. Meteor., 37 , 5–22.
Gillies, R. R., and Carlson T. N. , 1995: Thermal remote sensing of surface soil water content with partial vegetation cover for incorporation into climate models. J. Appl. Meteor., 34 , 745–756.
——, Cui, J., Carlson T. N. , Kustas W. P. , and Humes K. S. , 1997: Verification of a method for obtaining surface soil water content and energy fluxes from remote measurements of NDVI and surface radiant temperature. Int. J. Remote Sen., 18 , 3145–3166.
Goetz, S. J., Halthore R. N. , Hall F. G. , and Markham B. L. , 1995: Surface temperature retrieval in a temperate grassland with multiresolution sensors. J. Geophys. Res., 100 , 25 397–25 410.
Gutman, G., and Ignatov A. , 1998: The derivation of the green vegetation fraction from NOAA/AVHRR data for use in numerical weather prediction models. Int. J. Remote Sens., 19 , 1533–1543.
Henderson-Sellers, A., Pitman A. J. , Love P. K. , Irannejad P. , and Chen T. H. , 1995: The Project for Intercomparison of Land Surface Parameterization Schemes (PILPS): Phases 2 and 3. Bull. Amer. Meteor. Soc., 76 , 489–503.
Horst, T. W., and Weil J. C. , 1992: Footprint estimation for scalar flux measurements in the atmospheric surface layer. Bound.-Layer Meteor., 59 , 279–296.
Houser, P. R., Shuttleworth W. J. , Famiglietti J. S. , Gupta H. V. , Syed K. , and Goodrich D. C. , 1998: Integration of soil moisture remote sensing and hydrologic modeling using data assimilation. Water Resour. Res., 34 , 3405–3420.
Jackson, T. J., and Schmugge T. J. , 1989: Passive microwave remote sensing system for soil moisture: Some supporting research. IEEE Trans. Geosci. Remote Sens., 27 , 225–235.
——, Le Vine, D. M., Swift C. T. , Schmugge T. J. , and Schiebe F. R. , 1995: Large area mapping of soil moisture using the ESTAR passive microwave radiometer in Washita '92. Remote Sens. Environ., 53 , 27–37.
——, ——, Hsu, A. Y., Oldak A. , Starks P. J. , Swift C. T. , Isham J. D. , and Haken M. , 1999: Soil moisture mapping at regional scales using microwave radiometry: The Southern Great Plains Hydrology Experiment. IEEE Trans. Geosci. Remote Sens., 37 , 2136–2151.
Jiang, L., and Islam S. , 1999: A methodology for estimation of surface evapotranspiration over large areas using remote sensing observations. Geophys. Res. Lett., 26 , 2773–2776.
Kustas, W. P., and Norman J. M. , 1996: Use of remote sensing for evapotranspiration monitoring over land surfaces. Hydrol. Sci. J., 41 , 495–516.
——, and Jackson, T. J., 1999: The impact on area-averaged heat fluxes using remotely sensed data at different resolutions: A case study using Washita '92 data. Water Resour. Res., 35 , 1539–1550.
——, Moran, M. S., Humes K. S. , Stannard D. I. , Pinter Jr. P. J. , Hipps L. E. , Swiatek E. , and Goodrich D. C. , 1994: Surface energy balance estimates at local and regional scales using optical remote sensing from an aircraft platform and atmospheric data collected over semiarid rangelands. Water Resour. Res., 30 , 1241–1259.
——, Stannard, D. I., and Allwine K. J. , 1996: Variability in surface energy flux partioning during Washita 92: Resulting effects on Penman–Monteith and Priestley–Taylor parameters. Agric. For. Meteor., 82 , 171–193.
——, Zhan, X., and Schmugge T. J. , 1998: Combining optical and microwave remote sensing for mapping energy fluxes in a semiarid watershed. Remote Sens. Environ., 64 , 116–131.
——, ——, and Jackson, T. J., 1999: Mapping surface energy flux partitioning at large scales with optical and microwave remote sensing data from Washita '92. Water Resour. Res., 35 , 265–277.
Lakshmi, V., Wood E. F. , and Choudhury B. J. , 1997: A soil–canopy–atmosphere model for use in satellite remote sensing. J. Geophys. Res., 102 , 6911–6927.
Le Vine, D. M., Griffis A. J. , Swift C. T. , and Jackson T. J. , 1994: ESTAR: A synthetic aperture microwave radiometer for remote sensing applications, Proc. IEEE, 82 , 1787–1801.
Li, J., and Islam S. , 1999: On the estimation of soil moisture profile and surface fluxes partitioning from sequential assimilation of surface layer soil moisture. J. Hydrol., 220 , 86–103.
MacPherson, J. I., 1998: NRC Twin Otter operations in the 1997 Southern Great Plains Experiment. Institute for Aerospace Research Rep. LTR-FR-146, National Research Council Canada, 122 pp.
——, Dobosy, R., Verma S. , Kustas W. P. , Prueger J. H. , and Williams A. , 1999a: Intercomparisons between flux aircraft and towers in SGP97. Preprints, 14th Conf. on Hydrology, Dallas, TX, Amer. Meteor. Soc., 125–128.
——, Schuepp, P. H., Desjardins R. L. , Bastin M. , Ogunjemiyo S. , and Riznek R. , 1999b: Relating boundary layer flux measurements to remotely-sensed radiometric data. Proc. Fourth Int. Airborne Remote Sensing Conf./21st Canadian Symp. on Remote Sensing, Vol. I, Ottawa, ON, Canada, ERIM International, Inc., 605–614.
Mahfouf, J. F., and Noilhan J. , 1991: Comparative study of various formulations of evaporation from bare soil using in situ data. J. Appl. Meteor., 30 , 1354–1365.
Mahrt, L., 1998: Flux sampling errors for aircraft and towers. J. Atmos. Oceanic Technol., 15 , 416–429.
McNaughton, K. G., and Jarvis P. G. , 1991: Effects of spatial scale on stomatal control of transpiration. Agric. For. Meteor., 54 , 279–301.
Mecikalski, J. R., Diak G. R. , Anderson M. C. , and Norman J. M. , 1999: Estimating fluxes on continental scales using remotely sensed data in an atmospheric–land exchange model. J. Appl. Meteor., 38 , 1352–1369.
Mohr, K. I., Famiglietti J. S. , Boone A. , and Starks P. J. , 2000: Modeling soil moisture and surface flux variability with an untuned land surface scheme: A case study from the Southern Great Plains 1997 Hydrology Experiment. J. Hydrometeor., 1 , 154–169.
Moran, M. S., Rahman M. A. , Washburne J. C. , Goodrich D. C. , Weltz M. A. , and Kustas W. P. , 1996: Combining Penman–Monteith equation with measurements of surface temperature and reflectance to estimate evaporation rates of semiarid grasslands. Agric. For. Meteor., 80 , 87–109.
Norman, J. M., Kustas W. P. , and Humes K. S. , 1995: A two-source approach for estimating soil and vegetation energy fluxes from observations of directional radiometric surface temperature. Agric. For. Meteor., 77 , 263–293.
——, ——, Prueger, J. H., and Diak G. R. , 2000: Surface flux estimation using radiometric temperature: A dual temperature difference method to minimize measurement error. Water Resour. Res., 36 , 2263–2274.
Pielke, R. A., and Avissar R. , 1990: Influence of landscape structure on local and regional climate. Landscape Ecol., 4 , 133–155.
Priestley, C. H. B., and Taylor R. J. , 1972: On the assessment of surface heat flux and evaporation using large-scale parameters. Mon. Wea. Rev., 100 , 81–92.
Raupach, M. R., 1994: Simplified expressions for vegetation roughness length and zero-plane displacement as functions of canopy height and area index. Bound.-Layer Meteor., 71 , 211–216. Corrigendum,. 76 , 303–304.
——,. 1995: Vegetation–atmosphere interaction and conductance at leaf, canopy and regional scales. Agric. For. Meteor., 73 , 151–179.
——, and Finnigan, J. J., 1995: Scale issues in boundary-layer meteorology: Surface energy balances in heterogeneous terrain. Hydrol. Processes, 9 , 589–612.
Ross, J., 1975: Radiative transfer in plant communities. Vegetation and the Atmosphere, J. L. Monteith, Ed., Vol. 1, Academic Press, 13–15.
Schmugge, T. J., and Jackson T. J. , 1994: Mapping surface soil moisture with microwave radiometers. Meteor. Atmos. Phys., 54 , 213–223.
Schuepp, P. H., Leclerc M. Y. , MacPherson J. I. , and Desjardins R. L. , 1990: Footprint prediction of scalar fluxes from analytical solutions of the diffusion equation. Bound.-Layer Meteor., 50 , 355–373.
Sellers, P. J., Hall F. G. , Asrar G. , Strebel D. E. , and Murphy R. E. , 1992a: An overview of the First International Satellite Land Surface Climatology Project (ISLSCP) Field Experiment (FIFE). J. Geophys. Res., 97 , 18 345–18 371.
——, Heiser, M. D., and Hall F. G. , 1992b: Relations between surface conductance and spectral vegetation indices at intermediate (100 m2 to 15 km2) length scales. J. Geophys. Res., 97 , 19 033–19 059.
——, and. Coauthors. . 1995: Effects of spatial variability in topography, vegetation cover and soil moisture on area-averaged surface fluxes: A case study using FIFE 1989 data. J. Geophys. Res., 100 , 25 607–25 629.
——, and. Coauthors. . 1997: BOREAS in 1997: Experiment overview, scientific results, and future directions. J. Geophys. Res., 102 , 28 731–28 769.
Schmid, H. P., 1994: Source areas for scalars and scalar fluxes. Bound.-Layer Meteor., 67 , 293–318.
Shafer, M. A., Fiebrich C. A. , Arndt D. S. , Fredrickson S. E. , and Hughs T. W. , 2000: Quality assurance procedures in the Oklahoma mesonetwork. J. Atmos. Oceanic Technol., 17 , 474–494.
Simmonds, L. P., and Burke E. J. , 1998: Estimating near-surface soil water content from passive microwave remote sensing—an application of MICRO-SWEAT. Hydrol. Sci. J., 43 , 521–534.
Twine, T. E., and Coauthors. 2000: Correcting eddy-covariance flux underestimates over a grassland. Agric. For. Meteor., 103 , 279–300.
van de Griend, A. A., and Owe M. , 1994: Bare soil surface resistance to evaporation by vapor diffusion under semiarid conditions. Water Resour. Res., 30 , 181–188.
Willmott, C. J., 1982: Some comments on the evaluation of model performance. Bull. Amer. Meteor. Soc., 63 , 1309–1313.
Yamanaka, T., Takeda A. , and Sugita F. , 1997: A modified surface-resistance approach for representing bare-soil evaporation: Wind tunnel experiments under various atmospheric conditions. Water Resour. Res., 33 , 2117–2128.
SGP97 experimental region, showing the main area mapped by the ESTAR instrument, the approximate locations of the three main facilities, the Mesonet and Micronet sites, and ARM boundary layer facilities
Citation: Journal of Hydrometeorology 2, 3; 10.1175/1525-7541(2001)002<0254:VOPARS>2.0.CO;2
Land cover map for the ESTAR instrument mapping area along with the locations of the flux towers, aircraft flux transects, and Mesonet stations used in creating spatially distributed meteorological inputs for the model
Citation: Journal of Hydrometeorology 2, 3; 10.1175/1525-7541(2001)002<0254:VOPARS>2.0.CO;2
Schematic of resistance network for the TSEBSM model. Variables are defined in the text
Citation: Journal of Hydrometeorology 2, 3; 10.1175/1525-7541(2001)002<0254:VOPARS>2.0.CO;2
Comparison of TSEBSM model output of the energy balance components with the tower flux observations from the various locations across the SGP97 study area. Points are referenced to flux tower sites listed in Table 1. Line represents perfect agreement between modeled and observed energy fluxes
Citation: Journal of Hydrometeorology 2, 3; 10.1175/1525-7541(2001)002<0254:VOPARS>2.0.CO;2
Average energy balance components and resulting rmsd values with model predictions for all days from (a) flux tower observations and (b) flux aircraft observations
Citation: Journal of Hydrometeorology 2, 3; 10.1175/1525-7541(2001)002<0254:VOPARS>2.0.CO;2
Comparison of TSEBSM model output of the energy balance components with the run-average aircraft flux observations from the El Reno (ER) and Kingfisher (KF) transects. Line represents perfect agreement between modeled and observed energy fluxes
Citation: Journal of Hydrometeorology 2, 3; 10.1175/1525-7541(2001)002<0254:VOPARS>2.0.CO;2
Comparison of TSEBSM model-simulated surface temperature Tsurf vs the PRT-5 thermal infrared radiometer observations averaged over the ER and KF transects. Line represents perfect agreement between modeled and observed Tsurf
Citation: Journal of Hydrometeorology 2, 3; 10.1175/1525-7541(2001)002<0254:VOPARS>2.0.CO;2
Comparison of area-averaged energy balance components estimated by the TSEBSM and TSEBTR models using radiometric surface temperature imagery collected over the El Reno facility on DOY 183 at ∼1015 CST (see text)
Citation: Journal of Hydrometeorology 2, 3; 10.1175/1525-7541(2001)002<0254:VOPARS>2.0.CO;2
Spatially distributed meteorological and near-surface soil moisture inputs to the TSEBSM model for DOY 183 at 1030 CST output time over the main ESTAR mapping area
Citation: Journal of Hydrometeorology 2, 3; 10.1175/1525-7541(2001)002<0254:VOPARS>2.0.CO;2
Resulting output of spatially distributed energy balance components computed by the TSEBSM model for DOY 183 at 1030 CST output time over the main ESTAR mapping area
Citation: Journal of Hydrometeorology 2, 3; 10.1175/1525-7541(2001)002<0254:VOPARS>2.0.CO;2
A map of simulated Tsurf from the TSEBSM model and the near-surface soil moisture W map for DOY 183 at 1030 CST output time over the main ESTAR mapping area
Citation: Journal of Hydrometeorology 2, 3; 10.1175/1525-7541(2001)002<0254:VOPARS>2.0.CO;2
A scatterplot of Tsurf simulated by the TSEBSM model for DOY 183 at 1030 CST output time vs near-surface soil moisture derived from the ESTAR data. The total number of points plotted is approximately 15 200
Citation: Journal of Hydrometeorology 2, 3; 10.1175/1525-7541(2001)002<0254:VOPARS>2.0.CO;2
A scatterplot of Tsurf simulated by the TSEBSM model for DOY 183 at 1030 CST output time vs NDVI from the Landsat TM image for the main ESTAR mapping area. The total number of points plotted is approximately 15 200
Citation: Journal of Hydrometeorology 2, 3; 10.1175/1525-7541(2001)002<0254:VOPARS>2.0.CO;2
Description of flux tower sites used in validating TSEBSM energy balance predictions. Site codes are defined in Table 2.
Summary of days with ESTAR observations commensurate with tower and/or aircraft-based flux measurements. Near-surface soil moisture conditions for the three main facilities, Central Facility (CF), El Reno (ER), and Little Washita (LW) and for the area surrounding the Kingfisher (KF) flight track (see Fig. 2) are listed
