Next Article in Journal
Phenological Metrics Derived over the European Continent from NDVI3g Data and MODIS Time Series
Previous Article in Journal
A New Tree Cover Percentage Map in Eurasia at 500 m Resolution Using MODIS Data
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Actual Evapotranspiration (Water Use) Assessment of the Colorado River Basin at the Landsat Resolution Using the Operational Simplified Surface Energy Balance Model

1
ASRC Federal InuTeq, Contractor to the US Geological Survey Earth Resources Observation and Science Center, 47914 252nd Street, Sioux Falls, SD 57198, USA
2
US Geological Survey Earth Resources Observation and Science Center, 47914 252nd Street, Sioux Falls, SD 57198, USA
3
Stinger Ghaffarian Technologies Inc., Contractor to the US Geological Survey Earth Resources Observation and Science Center, Sioux Falls, SD 57198, USA
4
Southwest Watershed Research Center, US Department of Agriculture, Tucson, AZ 85719, USA
*
Author to whom correspondence should be addressed.
Remote Sens. 2014, 6(1), 233-256; https://0-doi-org.brum.beds.ac.uk/10.3390/rs6010233
Submission received: 10 November 2013 / Revised: 28 November 2013 / Accepted: 16 December 2013 / Published: 20 December 2013

Abstract

:
Accurately estimating consumptive water use in the Colorado River Basin (CRB) is important for assessing and managing limited water resources in the basin. Increasing water demand from various sectors may threaten long-term sustainability of the water supply in the arid southwestern United States. We have developed a first-ever basin-wide actual evapotranspiration (ETa) map of the CRB at the Landsat scale for water use assessment at the field level. We used the operational Simplified Surface Energy Balance (SSEBop) model for estimating ETa using 328 cloud-free Landsat images acquired during 2010. Our results show that cropland had the highest ETa among all land cover classes except for water. Validation using eddy covariance measured ETa showed that the SSEBop model nicely captured the variability in annual ETa with an overall R2 of 0.78 and a mean bias error of about 10%. Comparison with water balance-based ETa showed good agreement (R2 = 0.85) at the sub-basin level. Though there was good correlation (R2 = 0.79) between Moderate Resolution Imaging Spectroradiometer (MODIS)-based ETa (1 km spatial resolution) and Landsat-based ETa (30 m spatial resolution), the spatial distribution of MODIS-based ETa was not suitable for water use assessment at the field level. In contrast, Landsat-based ETa has good potential to be used at the field level for water management. With further validation using multiple years and sites, our methodology can be applied for regular production of ETa maps of larger areas such as the conterminous United States.

1. Introduction

Water management is becoming more challenging, particularly in the arid western United States because of changes in climate, land use, and demography. One climate change study has indicated that annual runoff in the Colorado River Basin (CRB) will be reduced by 14%, 18%, and 17% for the periods 2010–2039, 2040–2069, and 2070–2098, respectively [1]. The reduced availability of water in the basin could also prevent water allocation requirements of the Colorado River Compact from being met [2]. Increased demand for water from various sectors has motivated water managers to demand more accurate water budgets at higher resolution so that available water resources may be better allocated.
Remote sensing has great potential but remains underutilized by practicing water resource managers [3,4]. Accurately estimating consumptive water use using remotely sensed data helps water managers in planning, distribution, and management of water resources. Agro-meteorological models based on remote sensing are the most suited for estimating crop water use at the field and regional scales [5]. Some of the remote sensing based models for estimating actual evapotranspiration (ETa) include the surface energy balance index (SEBI) [6], two source model (TSM) [7], surface energy balance algorithm for land (SEBAL) [8], simplified surface energy balance index (S-SEBI) [9], surface energy balance system (SEBS) [10], ET mapping algorithm (ETMA) [11], atmosphere-land exchange inverse (ALEXI) [12], mapping evapotranspiration at high resolution with internalized calibration (METRIC) [13], simplified surface energy balance (SSEB) [14], wet METRIC (wMETRIC) [15], and operational simplified surface energy balance (SSEBop) [16]. Various publications have reviewed some of these models and other methods for estimating ETa using remotely sensed data [5,1719].
The US Bureau of Reclamation of the Department of the Interior (DOI) is the nodal agency for managing water in the Colorado River Basin. As per “the 1964 Decree” of the U.S. Supreme Court in Arizona vs. California et al., the Secretary of the DOI must provide complete, detailed, and accurate records of consumptive use of water from the Colorado River. Furthermore, one of the six science strategic directions of the US Geological Survey (USGS) in the decade 2007–2017 is a Water Census (water availability and use assessment) of the United States for quantifying, forecasting, and securing freshwater for America’s future [20]. The DOI has launched WaterSMART (Sustain and Manage America’s Resources for Tomorrow) to achieve a sustainable water strategy to meet the Nation’s water needs [21]. The CRB is one of the focus areas under WaterSMART initiatives for developing and testing suitable methods to meet the Water Census objectives [22].
Our objective of this study is to produce seamless ETa estimates of the Colorado River Basin at the Landsat scale using the SSEBop model and evaluate the model performance using field measurements, water balance study, and Moderate Resolution Imaging Spectroradiometer (MODIS)-based ETa estimates.

2. Methods and Materials

The operational Simplified Surface Energy Balance (SSEBop) model was used for estimating ETa within the CRB. This model uses pre-defined hot and cold boundary conditions unique to each pixel. Independent weather datasets from the Global Data Assimilation System (GDAS) are used to compute the reference ET (ETo). ETo is constrained by ET fraction (ETf) based on Landsat surface temperature and Parameter-Elevation Regressions on Independent Slopes Model (PRISM) air temperature data. A brief overview of the SSEBop model is presented here. Readers may refer to Senay et al. (2013) [16] for more details on model parameterization, hypothesis, and justification.

2.1. The Operational Simplified Surface Energy Balance (SSEBop) Model

The SSEBop Model uses a pre-defined hot and cold boundary condition for computing daily evapotranspiration ETa (mm·d−1) as
ET a = ET f × k × ET o
where ETo is short grass reference ET (mm·d−1), ETf is ET fraction (−), and k is a scaling coefficient (−) based on calibration. In this study we used k as 1. Senay et al. (2013) [16] suggest a calibration process to determine the k value since the calculation of the pre-defined parameters may already incorporate a compensating bias.
ET fraction, ETf (−), is computed as
ET f = T h T s T h T c
where Th is pre-defined idealized reference hot pixel temperature (K), Ts is land surface temperature (K) obtained from Landsat images, and Tc is pre-defined idealized reference cold pixel temperature (K).
In the SSEBop model, each pixel has a predefined hot and cold boundary values based on maximum air temperature and differential temperature (dT). Pre-defined cold pixel temperature is computed as a fraction of maximum air temperature. Maximum air temperature is not selected over the Landsat scene or the modeling domain but each pixel has its own predefined boundary values. The cold pixel temperature is approximated as being close to the corresponding air temperature based on the assumption that for a given clear-sky day, the land surface will experience an ETa rate equal to the potential rate for healthy and well watered vegetation when its Ts is close to the near-surface air temperature (i.e., little or no sensible heat flux). Calibration of the SSEBop Model using MODIS-based images has shown this coefficient as 0.993 [16]. However, our calibration using Landsat images has shown a different value; hence, we computed Tc as
T c = 0.985 × T max
where Tmax is the maximum air temperature (K).
Pre-defined hot pixel temperature is computed as
T h = T c + dT
Differential temperature (dT, K) is computed based on the assumption that latent heat flux and heat storage at the daily time scale for a dry, bare soil will be negligible. Thus based on energy balance, sensible heat flux (H) will be equal to net radiation (Rn) at the dry, bare soil. So we can replace H with Rn in the conventional H formulation and compute dT as
dT = R n × r ah ρ a × c p
where rah is the aerodynamic resistance to heat transfer from a hypothetical dry, bare surface (110 s·m−1), ρa is the density of air (kg·m−3), and cp is the specific heat of air at constant pressure (1,004 J·kg−1·K−1). The calculations of net radiation (Rn) under clear-sky condition and air density (ρa) are adaptions of Allen et al. (1998) [23] as described in Senay et al. [16]. It is important to note that dT is unique for each period and location, but the value does not vary from year to year since it is calculated under a clear-sky condition.

2.2. Study Area

The Colorado River originates in the Rocky Mountains of the western United States and flows about 2,300 km through seven states (Wyoming, Colorado, Utah, New Mexico, Arizona, Nevada, and California) before draining into the Gulf of California (Figure 1). The CRB has an area of about 630,000 km2, much of it arid. The Colorado River is a major source of water supply to the southwestern United States. The Colorado River supplies water to more than 25 million people and irrigates more than 12,000 km2 of cropland across the seven basin states [22]. The annual flow of the river has ranged from 6.5 billion cubic meters (BCM) to 29.6 BCM during 1906–2000 [1]. The combined reservoir storage capacity (74.0 BCM) within the basin is about four times the long-term average annual flow (16.7 BCM).
Based on the National Land Cover Database (NLCD, 2006) [24], shrubland (61%) is the most dominant land cover in the study area followed by evergreen forest (19%) (Table 1). Most of the pastureland is in the upper basin, most of the cropland is in the lower basin, and shrubland is distributed throughout the basin (Figure 2). Land use/land cover of the CRB is also used as an input in the Lower Colorado River Accounting System (LCRAS), a water accounting system developed by the USGS and the Bureau of Reclamation [25]. Congalton et al. (1998) [26] developed a procedure to accurately map agricultural crops and other land cover in the lower CRB. For additional detail on geo-topographic and climatic conditions of the study area, see Kumar and Duffy (2009) [27].

2.3. Processing of Landsat Images

The CRB is covered by 44 Landsat scenes spread over paths 33–44 and rows 30–38 (Figure 3). First, we downloaded all the Landsat images (Thematic mapper and Enhanced thematic mapper plus) of the study area for 2010 with less than 10% cloud cover from the Earth Explorer site ( http://earthexplorer.usgs.gov/). We then selected images that were nearly cloud free or had cloud only along the edges (total 328 images). In general, one Landsat image per month is ideal, but two images are desired for cropland, particularly during the crop growing season. However, based on our Landsat processing experience, reasonable results can be obtained based on 10–12 images per year.
Processing of Landsat images for obtaining land surface temperature (Ts) was carried out as follows. First, the digital numbers (DN) of downloaded Landsat images were converted to at-sensor radiance (L) as
L = ( L max L min ) × ( D N Q cal min ) ( Q cal max Q cal min ) + L min
where L is spectral radiance (W·m−2·sr−1·μm−1), Lmax is maximum rescaling factor (W·m−2·sr−1·μm−1), Lmin is minimum rescaling factor (W·m−2·sr−1·μm−1), DN is quantized calibrated pixel value (−), Qcalmax is quantized calibrated pixel value corresponding to the Lmax (−), and Qcalmin is quantized calibrated pixel value corresponding to the Lmin (−). The values of Lmax, Lmin, Qcalmax, and Qcalmin for each band are provided in the metafile of unzipped downloaded images.
The at-sensor radiance for the shortwave bands was converted to top of atmosphere reflectance (ρ) as
ρ = π × L ESUN × Cos θ × d r
where ρ is planetary top of atmosphere reflectance (−), dr is earth-sun distance parameter (−), ESUN is mean exoatmospheric solar irradiance (W·m−2·μm−1), and θ is solar zenith angle (degree). ESUN values are given in Chander et al. (2009) [28].
Top of atmosphere albedo and at surface albedo were calculated as
α toa = ( ρ × c )
α = α toa α path τ sw 2
where αtoa is albedo at the top of the atmosphere (−), α is at surface albedo (−), c is weighting coefficient (based on Tasumi et al., 2008) [29], αpath is albedo path radiance ranging from 0.025 to 0.04 (−), and τsw is transmittance as computed below.
τ sw = 0.75 + 2 × 10 5 × Z
where Z is elevation above the mean sea level (m).
Surface emissivity and land surface temperature were computed as presented in Allen et al. (2007) [13] without using any atmospheric radiative transfer simulation model such as MODTRAN. In some desert areas where albedo is high (>0.3), the radiometric land surface temperature tends to decrease, potentially due to reduced net radiation [16]. We corrected those Ts values to avoid erroneous ET fraction as
T s _ co = T s + 50 × ( α 0.3 )
where Ts_co is corrected land surface temperature (K).
The SSEBop model algorithm was implemented with Landsat images using Model Maker in Erdas Imagine 2011 (version 11.0.4) (Intergraph Corporation, Huntsville, AL, USA). Each individual scene was processed separately for computing ETa on the day of the satellite overpass. All the available daily ETa images were used for upscaling to compute annual ETa as discussed in the Computation of Annual ET section.

2.4. Other Supporting Data

Monthly minimum and maximum temperature and precipitation (P) gridded 4 km data for 2010 were downloaded from the Parameter-Elevation Regressions on Independent Slopes Model (PRISM) website. We used these data in their original format without any spatial interpolation to preserve the data integrity. PRISM is a knowledge-based system that uses ground-based meteorological station records, a digital elevation model (DEM), and many other geographic datasets to generate gridded estimates of monthly climatic parameters [30]. Monthly runoff (Q) based on Hydrologic Unit Code (HUC) for 2010 was downloaded from the USGS Water Watch website ( http://waterwatch.usgs.gov/index.php). The estimates of HUC runoff are generated by the USGS by combining historical flow data from stream gauges, the drainage basins of the stream gauges, and the HUC boundaries [31]. Annual ETa based on spatial water balance at the HUC8 level was carried out by subtracting annual HUC-based runoff from the annual PRISM precipitation. Only PRISM-based air temperature (not precipitation) was used in the SSEBop model for computing Landsat-based ETa. Thus ETa based on water balance (residual of precipitation and runoff, i.e., P–Q) should serve as an independent data source for validation of Landsat-based annual ETa. We assume that there is no change in annual storage of water; therefore, PQ is equivalent to ETa.
Topographic elevation data of the study area were derived from the February 2000 National Aeronautics and Space Administration (NASA) Shuttle Radar Topography Mission (SRTM), a seamless and complete void-filled spatial dataset [32]. Data were downloaded from the SRTM data portal. The daily reference evapotranspiration (ETo) (Senay et al., 2008) [33] was computed from climate parameter data extracted from global data assimilation system (GDAS) based 6-hourly weather datasets [34]. The daily ETo data are available at the Famine Early Warning Systems Network website ( http://earlywarning.usgs.gov/adds/global/index.php). The global ETo data available at 1° ground resolution were downscaled to 10 km resolution [14]. Gridded flux data are a global, spatially and temporally explicit estimates of latent flux upscaled from flux measurements based on remote sensing indices, climate and meteorological data, and information on land use [35]. These data were obtained from the Max Planck Institute for Biogeochemistry server. MODIS-based ETa maps using the SSEBop model [16] were also used for result comparison.

2.5. Computation of Annual Evapotranspiration

ETa on the day of the satellite overpass was computed using the SSEBop model. Computation of seasonal/annual ETa becomes challenging when daily ETa is not available because of Landsat’s 16-day repeat cycle and/or a lack of cloud-free images. We have used upscaling of daily ETa to annual ETa based on the ratio of actual ET to reference ET (ET coefficient). A study by Singh (2009) [36] with multiyear datasets showed that there was no single interpolation method that worked better than other methods under all conditions. Suitability of the method depended on the number of images per season, number of days between two consecutive images, extreme values on any particular day, and duration of the season [37]. The fixed method worked well for longer duration (annual). So we used the fixed method for computing annual ET based on ET coefficient:
ET annual = i = 1 n ( ET oi × ET fi )
where ETannual is the total annual ET (mm), EToi is the reference ET (mm) for period i (days), and ETfi is the representative ET coefficient (−) for period i.
After computing the annual ETa for each path/row, all annual ETa images were mosaicked. Mosaicking of annual ETa for different path/row is done to obtain a nearly seamless ETa map. For this we used average values for the overlapped area. Finally, the study area was extracted from the mosaicked images using a boundary shape file of the river basin.

2.6. Field Validation

Modeled ETa was compared with the eddy covariance measurements carried out at the different eddy covariance measurement sites (Table 2). These sires are Flagstaff managed forest [38], Flagstaff unmanaged forest [38], Flagstaff wildfire [38], Santa Rita Creosote [39], Santa Rita Mesquite [40], Kendall grassland [41], and Charleston Mesquite [42]. For comparison, we used the ETa values of pixels that have a flux tower without any knowledge of the spatial extent and relative importance of upwind source areas (footprint). Because all of these sites except for the Charleston Mesquite site are part of FLUXNET [43], we have included additional details and discussion on the Charleston Mesquite site. The data used from the FLUXNET sites are Level 2 data, so they have data gaps ranging from about 7% (Flagstaff wildfire site) to 22% (Santa Rita Creosote site). The data collection and methodology for FLUXNET sites were according to Ameriflux protocol described in the references given in Table 2. The Charleston Mesquite site is located at an elevation of 1200 m, about 16 km northeast of Sierra Vista, Arizona, on the east side of the San Pedro River. This site is dense riparian woodland dominated by velvet mesquite (Prosopis velutina), while the understory is composed of sacaton grass (Sporobolus wrightii), greythorn shrubs (Zizyphus obtusifolia), and other summer active annual herbaceous species. The average canopy cover is about 70% with leaf area index ranging from about 1.2 prior to leaf-out to about 2 during most of the growing season. A three-dimensional sonic anemometer (Model CSAT3, Campbell Scientific Inc., Logan, UT, USA) and an open path infrared gas analyzer (IRGA; Model LI-7500, LI-COR Inc., Lincoln, NE, USA) were used at the site. The variables were sampled at 10 Hz by a datalogger (CR5000, Campbell Scientific Inc., Logan, UT, USA) and averaged over 30-min. The average depth to groundwater was about 10 m. Other details of the Charleston Mesquite site and measurement techniques are described elsewhere [42,44].

3. Results and Discussion

3.1. Annual ETa of Different Land Use/Land Cover

Spatial distribution of the first ever annual ETa of the CRB at the Landsat scale showed different water use patterns within the basin (Figure 4). As expected, open water had the highest mean annual water loss (994 mm) followed by cropland (538 mm) as a result of evapotranspiration (Table 3). There was a wide spatial variation (high standard deviation) in annual ETa of different land cover classes (Table 3). The standard deviation of mean annual ETa ranged from 144 mm (grassland) to 699 mm (water). A wide range of mean annual ETa may be attributed to many variables including thematic accuracy of the NLCD 2006 map, changes in land use from 2006 to 2010, topography, cropping system, cropping intensity, irrigation, local climatic conditions, and Landsat scene availability for ETa mapping. In general, thematic accuracy of the Level II NLCD map is less than 80% [45]. This means many land cover classes may be misclassified, resulting in a wide range of annual ETa. However, the classification errors aggregated over a larger area do not result in significant error [46]. Cultivation of different crops will also result in a wide variation of actual ET as different crops have different water requirements based on plant physiology and crop duration. Similarly, growing more than one crop per year will result in higher ETa than growing only one crop per year. There will also be a difference in ETa of the same crop under irrigated and non-irrigated conditions because irrigation leads to increased ETa. A study on the effect of irrigation on the water and energy balance of the CRB has shown decreased stream flow and surface temperature [47]. The climatic gradient within the basin due to its large size and relief will also cause differences in the water use pattern.
As an example, the histogram of annual ETa for some of the land cover classes for path 37/row 37 (Figure 3) is presented in Figure 5. A majority (76%) of the barren land pixels had an annual ETa within 100–200 mm, and very few pixels (4%) had an annual ETa greater than 500 mm. Similarly, about half of the grassland (47%) had an annual ETa between 200 and 300 mm and less than 5% exceeded 500 mm. In contrast, annual ETa was greater than 800 mm in the majority of cropland. Some of the cropland was fallow based on visual interpretation during 2010 resulting in an annual ETa similar to grassland and barren land. In addition, the effect of misclassification in NLCD 2006 and/or change of land cover may be attributed to low annual ETa of cropland and high annual ETa of barren land and grassland.
Westenburg et al. (2006) [48] collected ETa data from three sites in Havasu National Wildlife Refuge during 2002–2004 using Bowen ratio stations. These sites included a large area of medium-to-high density, homogeneous saltcedar (Tamarix spp.); an area of medium density mixed vegetation; and a homogeneous area of low-to-medium density arrowweed. They reported annual ETa for 2003 for saltcedar, mixed vegetation, and arrowweed as 1,076 mm, 728 mm, and 716 mm, respectively. Our results for these sites showed annual ETa for 2010 as 1,291 mm, 1,089 mm, and 1,333 mm, respectively. It appears that ETa increased during these 7 years particularly for mixed vegetation and arrowweed. Doody et al. (2011) [49] listed ETa of some of the cover types based on reviewed literatures. They reported annual ETa of arrowweed as 1,370–1,590 mm, very close to our result for arrowweed. They found a wide range of annual ETa for saltcedar, ranging from 220 to 1,460 mm with a mean value of 765 ± 413 mm. They reported annual ETa for bare soil as 307 mm and for open water as 1,156 mm. Thus, overall our results for these locations and cover types are similar and comparable.
Nagler et al. (2005) [50] calculated spatial ET for the Lower Colorado River stretch at Havasu National Wildlife Refuge based on MODIS-enhanced vegetation index and air temperature. They found that the annual ETa of the area during 2000–2003 ranged from 832 mm (2000) to 881 mm (2002) with a mean ETa of 851 mm. Our zonal analysis of the same region showed mean annual ETa of 865 mm. Though we did not use exactly the same areal extent because our delineation slightly differed from theirs, the comparable mean ETa shows that our results are reliable. Even though there is a wide range in the standard deviation of different land cover classes, the mean annual ETa values are comparable to some of the measured and reported ETa in the CRB.

3.2. Validation with Eddy Covariance Measurement

Figure 6 shows the comparison of annual ETa measured by eddy covariance against Landsat annual ETa for all seven validation sites within the study area. Overall, the SSEBop-estimated ETa was in good agreement (R2 = 0.78) with the measured ETa for a wide range of elevations (991–2,270 m) and tower heights (4–23 m) (Table 2). The mean difference between measured and estimated annual ETa was about 10%, which is comparable with results from other energy balance based models [5,17]. The root mean square error (RMSE) was 106 mm, mainly due to the high discrepancy at the Flagstaff unmanaged forest and Flagstaff wildfire sites (Figure 6). If these two sites are removed, then R2 improves to 0.95 (Y = 0.997 X − 53.767) and RMSE decreases to 68 mm. This is remarkable considering we had only 6–9 cloud-free Landsat scenes for 2010 (Table 2) and we had not carried out any calibration of our model using eddy covariance measurements.
It should be noted that we have compared the measured ETa only with the ET of the pixel that has a flux tower. However, the point distribution in Figure 6 shows that the SSEBop model captured the ET variability well even without footprint analysis. In general, the footprint of a point of observation is affected by thermal stability, surface roughness, observation levels, and wind speed and direction [51]. Since the wind direction is variable throughout the year, an exact footprint analysis is simply impossible to carryout [52]. The advantage of having a remote sensing based ET map is that we get spatial and temporal distribution of ET, thus not requiring any footprint analysis. For example, Figure 7 shows the ET values for 5 × 5 Landsat pixels centered over the eddy covariance tower. We can see there is variability in ET values even within 150 m × 150 m (5 × 5 pixels), an area smaller than a generally used 100-to-1 fetch-to height ratio for sites (considering the tower height given in Table 2). Based on measured ET values, we can get a good sense of footprint direction and our results are in reasonable agreement. For example, wind direction at the Flagstaff unmanaged forest site was mostly from the southwest, where pixels had lower ETa values (Figure 7).
Though we have not carried out uncertainty analysis in the present study, like any other models, our model result may have some uncertainty. There are also systematic and random errors associated with the eddy covariance measurements [5355]. The total annual precipitation and ETa measured at the Charleston Mesquite site during 2010 were 266 mm and 673 mm, respectively. Our SSEBop model-estimated ETa at the site was 730 mm, which is only about 8% higher than the measured value. At this site, there is a lack of energy balance closure by about 11% (on a daily basis) in the eddy covariance measurements, and this is generally the case at most eddy covariance sites. Wilson et al. (2002) [54] evaluated eddy covariance measurements across 22 FLUXNET sites (50 site-years) and reported that the mean imbalance between turbulent energy fluxes (sensible and latent heat fluxes) and available energy (net radiation minus soil heat flux) is on the order of 20%. For forced energy balance closure we scaled the latent and sensible heat fluxes conserving the measured Bowen ratio. The scaled up ETa for energy balance closure at the site was 761 mm. Thus, there was only 4% difference between the scaled up measured ETa and the modeled ETa, giving us good confidence in the annual value at the pixel level. Our result is within the range of general ETa estimation accuracy reported for seasonal ETa using remote sensing models.
Jung et al. (2009) [35] used eddy covariance data from FLUXNET and developed a global, spatially and temporally explicit ETa map (gridded flux data) using model tree ensemble, a machine learning technique. We used this gridded flux data (written communication 2013) to compare our result. The mean annual ETa of the CRB for 2010 using gridded flux data was 292 mm. Our result showed that the mean annual ETa of the CRB is 312 mm. These values are comparable (within 7%) in spite of very coarse (50 km) spatial resolution of gridded flux data against this high resolution (30 m) Landsat data.

3.3. Comparison with Annual ETa Based on Water Balance Analysis

The Landsat-based mean annual ETa of HUC8 ranged from 144 mm (Upper Green-Slate in Upper Colorado basin) to 613 mm (Lower Salt in Lower Colorado basin), while mean annual ETa based on water balance ranged from 145 mm (Yuma desert in Lower Colorado basin) to 629 mm (Tonto sub-basin in Lower Colorado basin). Comparison of Landsat-based mean annual ETa with that of water balance-based mean annual ETa at the HUC8 level has shown that some HUC8 had very good correspondence while other HUC8 were scattered (Figure 8). This finding is expected as ETa based on water balance assumes mass balance closure at the annual scale. In reality, basin runoff exhibits strong memory from year to year and water balance may not close even over a year due to time lag. Overall, there was poor correlation (R2 = 0.16) between annual ETa based on Landsat and annual ETa based on water balance. Most of the mismatch between Landsat and water balance-based ETa is in sub-basins having large irrigated cropland or water bodies or where vegetation has access to groundwater thus invalidating the premise of water balance closure. We used three sub-criteria to remove some HUC8 sub-basins where water balance may not close because (1) mean runoff is more than mean precipitation, i.e., sub-basins where base flow is dominant; (2) mean ETa is more than mean precipitation, i.e., sub-basins where mostly irrigated or open water bodies are dominant; and (3) the runoff ratio (Q/P) is greater than 0.55, indicating large seepage loss/base flow/percolation loss. Out of 144 sub-basins within the CRB, 28 sub-basins were removed from the analysis based on these three criteria. This resulted in significant improvement of R2 from 0.16 to 0.85, and RMSE decreased from 120 mm to 99 mm. The regression line shows that annual ETa based on water balance was about 12% higher than Landsat-based ETa. This is reasonable considering that water balance-based ETa is computed neglecting the storage term in the water balance equation thus overestimating annual ETa. In general, the accuracy of ETa products at a higher HUC level (HUC8) is lower than the accuracy at a lower HUC level (HUC2) [56]. Our results show that even at the HUC8 level, Landsat-based ETa explained about 85% variability in ETa based on water balance. Compared to mean annual ETa, there was wide range in areal extent of sub-basins within the CRB, ranging from 381 km2 (Cloverdale in Lower Colorado basin) to 11,845 km2 (Chaco in Upper Colorado basin). A sub-basin having smaller mean ETa but larger area can have larger volume of water lost due to ETa. Thus, spatial distribution of ETa based on remotely sensed data at the sub-basin level can help water managers identify the water surplus and water deficit sub-basins and quantify the water availability.

3.4. Comparison with MODIS-Based Annual ETa

We also compared our Landsat-estimated ET with that obtained from MODIS-based annual ET. Senay et al. (2013) [16] used the SSEBop model with the MODIS images for 2005, and their monthly ET validation with 45 Ameriflux data sites for the conterminous United States resulted in an R2 of 0.64 with RMSE of 27 mm. With more years (2000–2007), MODIS-based annual ETa using the SSEBop model has resulted in a high skill score across all the climate zones in the western United States [56]. Spatial distribution of annual ETa using 2010 MODIS images has shown consumptive water use distribution within the CRB (Figure 9). In general, we can see that the ETa distribution pattern using MODIS and Landsat images is similar. Our comparison of Landsat-based annual ETa and MODIS-based annual ETa for all 144 HUCs (level 8) showed that there was good agreement (Figure 10). The coefficient of determination (R2 = 0.79) indicated that we can obtain a comparable ETa estimate even with the coarse resolution (1 km) of MODIS at the sub-basin (HUC8) level. However, the advantages of Landsat-based annual ETa become obvious when ETa is required at fine resolution (field scale). Figure 11 shows the annual ETa distribution in a small part of Colorado using Landsat and MODIS images. The ETa variation within individual center-pivot irrigation fields is clearly visible in the Landsat-based ETa estimate, but there is no distinction between fields in the MODIS-based ETa estimates. This is of high importance particularly for implementing water management decisions at the local level. For example, water conservation measures can be adopted in the fields that have higher ETa than surrounding fields. Besides improving efficiency, this can save energy and water that can be used elsewhere. As tackling water challenges is one of the highest priority initiatives under the DOI WaterSMART Program, ETa estimates at the field scale will help in water conservation, availability, and water conflict resolution.
As modelers, we recognize that model development is a continuous process. As with any other models, our model has continued evolving since its first publication [14]. With the availability of better data sources, our model output will also improve as a result of using better input data. For example, the use of Daymet (Thornton et al., 1997) [57], daily climatological data is expected to give better results than monthly PRISM air temperature data. Similarly, North American Land Data Assimilation System (NLDAS) (Mitchell et al., 2004) [58], which is a quality-controlled and spatially and temporally consistent land surface model dataset (1/8th degree spatial resolution), may improve ETo and thus ETa estimates.

4. Conclusions

Accurate information on water availability and water use is necessary for planning sustainable use of water, particularly in an arid region like the southwestern United States. Though detailed information on water availability in the Colorado River Basin is available, there is a need for reliable estimation of spatially distributed water use. We have developed a basin-wide ETa map of the Colorado River Basin using images from Landsat, the longest existing polar orbiting satellite. The first ever ETa map of the Colorado River Basin at Landsat scale has shown the distribution pattern of ETa, which is based on the scientifically robust and operationally viable SSEBop model. We have demonstrated that, in spite of the complexities of the CRB, seamless ETa based on remote sensing can be estimated at the Landsat scale. Though cropland covers less than 1% of the CRB, it has the highest ETa after open water bodies. There was a wide variation in ETa within most land use/land cover classes in the basin due to hydro-climatic variation within the basin and uncertainty in the land cover classification. Validation of Landsat-based estimated ETa with ETa measured using eddy covariance has shown that there was good correspondence (R2 = 0.78) and the model captured the variability of ETa over a wide range of elevation and tower height. The R2 improved to 0.95 with a slope of 0.997 when two sites (FUF and FWF) were removed from validation. The SSEBop model-estimated ETa was within 4% of the measured ETa at two sites (SRC and CMS). Comparison with ETa based on water balance at the HUC8 level showed good agreement (R2 = 0.85) for sub-basins most likely having water balance closure. ETa maps using Landsat and MODIS images showed a similar water use pattern at the regional scale. However, the advantages of the Landsat images were clearly observed when the ETa map was zoomed in for field-level ETa distribution. In contrast with MODIS-based ETa maps, individual fields are clearly distinguishable in Landsat-based ETa maps. Thus, Landsat-based ETa mapping is better suited for planning and management of water resources at the field scale.
ETa based on remote sensing may have biases, but these biases can be easily identified and removed based on validation using reliable field data. Once the bias is removed, consistent ETa maps can be generated for regular ETa mapping and water resources planning and management. A regular production of Landsat-based ETa maps using a uniform method and data source at a seasonal and annual scale has the potential for resolving inconsistencies arising from using different methods and variable data sources. With further validation using multiple years and sites, our methodology can be applied for larger areas such as the conterminous United States.

Acknowledgments

This work was performed under U.S. Geological Survey (USGS) contract G23PC00028 and G10PC00044 in support of the WaterSMART program. We acknowledge with thanks the internal review of the manuscript by Roger Auch, USGS Earth Resources Observation and Science Center. We gratefully acknowledge the use of Ameriflux data for our model validation. We are thankful to the four anonymous reviewers for their thoughtful comments and helpful suggestions on an earlier version of the manuscript. Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the US Government.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Christensen, N.S.; Wood, A.W.; Voisin, N.; Lettenmaier, D.P.; Palmer, R.N. The effects of climate change on the hydrology and water resources of the Colorado River Basin. Clim. Chang 2004, 62, 337–363. [Google Scholar]
  2. McCabe, G.J.; Wolock, D.M. Warming may create substantial water supply shortages in the Colorado River Basin. Geophys. Res. Lett 2007, 34, L22708. [Google Scholar]
  3. Bastiaanssen, W.G.M.; Molden, D.J.; Makin, I.W. Remote sensing for irrigated agriculture: Examples from research and possible applications. Agric. Water Manag 2000, 46, 137–155. [Google Scholar]
  4. Nagler, P.L.; Glenn, E.P.; Didan, K.; Osterberg, J.; Jordan, F.; Cunningham, J. Wide-area estimates of stand structure and water use of Tamarix spp. on the Lower Colorado River: Implications for restoration and water management projects. Restor. Ecol 2008, 16, 136–145. [Google Scholar]
  5. Gowda, P.H.; Chavez, J.L.; Colaizzi, P.D.; Evett, S.R.; Howell, T.A.; Tolk, J.A. ET Mapping for agricultural water management: Present status and challenges. Irrig. Sci 2008, 26, 223–237. [Google Scholar]
  6. Menenti, M.; Chaudhury, B.J. Parameterization of Land Surface Evapotranspiration Using a Location Dependent Potential Evapotranspiration and Surface Temperature Range. Proceedings of the Exchange Processes at the Land Surface for a Range of Space and Time Scales; Bolle, H.J., Feddes, R.A., Kalma, J.D., Eds.; International Association of Hydrological Sciences: Rennes, France, 1993; 212, pp. 561–568. [Google Scholar]
  7. Norman, J.M.; Kustas, W.P.; Humes, K.S. A two-source approach for estimating soil and vegetation energy fluxes from observations of directional radiometric surface temperature. Agric. For. Meteorol 1995, 77, 263–293. [Google Scholar]
  8. Bastiaanssen, W.G.M.; Menenti, M.; Feddes, R.A.; Holtslag, A.A.M. The surface energy balance algorithm for land (SEBAL): Part 1 Formulation. J. Hydrol 1998, 212–213, 198–212. [Google Scholar]
  9. Roerink, G.J.; Su, Z.; Menenti, M. S-SEBI: A simple remote sensing algorithm to estimate the surface energy balance. Phys. Chem. Earth Part B 2000, 25, 147–157. [Google Scholar]
  10. Su, Z. The Surface Energy Balance System (SEBS) for estimation of turbulent heat fluxes. Hydrol. Earth Syst. Scie 2002, 6, 85–99. [Google Scholar]
  11. Loheide, S.P., II; Gorelick, S.M. A local-scale, high resolution evapotranspiration mapping algorithm (ETMA) with hydroecological applications at riparian meadow restoration sites. Remote Sens. Environ 2005, 98, 182–200. [Google Scholar]
  12. Anderson, M.C.; Norman, J.M.; Mecikalski, J.R.; Otkin, J.A.; Kustas, W.P. A Climatological study of evapotranspiration and moisture stress across the continental United States based on thermal remote sensing: 1 Model formulation. J. Geophys. Res 2007. [Google Scholar] [CrossRef]
  13. Allen, R.G.; Tasumi, M.; Trezza, R. Satellite-based energy balance for mapping evapotranspiration with internalized calibration (METRIC)—Model. ASCE J. Irrig. Drain. Eng 2007, 133, 380–394. [Google Scholar]
  14. Senay, G.B.; Budde, M.E.; Verdin, J.P.; Melesse, A.M. A coupled remote sensing and Simplified Surface Energy Balance (SSEB) approach to estimate actual evapotranspiration from irrigated fields. Sensors 2007, 7, 979–1000. [Google Scholar]
  15. Singh, R.K.; Irmak, A. Treatment of anchor pixels in the METRIC model for improved estimation of sensible and latent heat fluxes. Hydrol. Sci. J 2011, 56, 895–906. [Google Scholar]
  16. Senay, G.B.; Bohms, S.; Singh, R.K.; Gowda, P.H.; Velpuri, N.M.; Alemu, H.; Verdin, J.P. Operational evapotranspiration mapping using remote sensing and weather datasets: A new parameterization for the SSEB approach. J. Am. Water Resourc. Assoc 2013, 49, 577–591. [Google Scholar]
  17. Gowda, P.H.; Chavez, J.L.; Colaizzi, P.D.; Evett, S.R.; Howell, T.A.; Tolk, J.A. Remote sensing based energy balance algorithms for mapping ET: Current status and future challenges. Trans. Am. Soc. Agric. Biol. Eng 2007, 50, 1639–1644. [Google Scholar]
  18. Kalma, J.D.; McVicar, T.R.; McCabe, M.F. Estimating land surface evaporation: A review of methods using remotely sensed surface temperature data. Surv. Geophys 2008, 29, 421–469. [Google Scholar]
  19. Glenn, E.P.; Neale, C.M.U.; Hunsaker, D.J.; Nagler, P.L. Vegetation index-based crop coefficients to estimate evapotranspiration by remote sensing in agricultural and natural ecosystems. Hydrol. Process 2011, 25, 4050–4062. [Google Scholar]
  20. US Geological Survey (USGS). Facing Tomorrow’S Challenges—U.S. Geological Survey Science in the Decade 2007–2017. US Geol. Surv. Circ 2007, 1309, 1–69. [Google Scholar]
  21. US Department of the Interior (DOI). Fiscal Year 2011 The Interior Budget in Brief. In WaterSMART: Departmental Highlights; US Department of the Interior: Washington, DC, USA, 2010; pp. 19–25. [Google Scholar]
  22. Bruce, B.W. WaterSMART—The Colorado River Basin Focus Area Study; US Geological Survey: Washington, DC, USA.
  23. Allen, R.G.; Pereira, L.S.; Raes, D.; Smith, M. Crop Evapotranspiration: Guidelines for Computing Crop Water Requirements; FAO: Rome, Italy, 1998. [Google Scholar]
  24. Fry, J.; Xian, G.; Jin, S.; Dewitz, J.; Homer, C.; Yang, L.; Barnes, C.; Herold, N.; Wickham, J. Completion of the 2006 national land cover database for the conterminous United States. Photogramm. Eng. Remote Sens 2011, 77, 858–864. [Google Scholar]
  25. Bureau of Reclamation (BOR). Lower Colorado River Accounting System, Demonstration of Technology; Lower Colorado Regional Office: Boulder City, NV, 1997. [Google Scholar]
  26. Congalton, R.G.; Balogh, M.; Bell, C.; Green, K.; Milliken, J.A.; Ottman, R. Mapping and monitoring agricultural crops and other land cover in the Lower Colorado River Basin. Photogramm. Eng. Remote Sens 1998, 64, 1107–1114. [Google Scholar]
  27. Kumar, M.; Duffy, C.J. Detecting hydroclimatic change using spatio-temporal analysis of time series in Colorado River Basin. J. Hydrol 2009, 374, 1–15. [Google Scholar]
  28. Chander, G.; Markham, B.L.; Helder, D.L. Summary of current radiometric calibration coefficients for Landsat MSS, TM, ETM+, and EO-1 ALI sensors. Remote Sens. Environ 2009, 113, 893–903. [Google Scholar]
  29. Tasumi, M.; Allen, R.; Trezza, R. At-surface reflectance and albedo from satellite for operational calculation of land surface energy balance. J. Hydrol. Eng 2008, 13, 51–63. [Google Scholar]
  30. Daly, C.; Gibson, W.P.; Taylor, G.H.; Johnson, G.L.; Pasteris, P. A knowledge-based approach to the statistical mapping of climate. Clim. Res 2002, 22, 99–113. [Google Scholar]
  31. US Geological Survey (USGS). Waterwatch–Maps, Graphs, and Tables of Current, Recent, and Past Streamflow Conditions. Available online: http://pubs.usgs.gov/fs/2008/3031/ (accessed on 31 December 2012).
  32. Reuter, H.I.; Nelson, A.; Jarvis, A. An evaluation of void filling interpolation methods for SRTM data. Int. J. Geogr. Inf. Sci 2007, 21, 983–1008. [Google Scholar]
  33. Senay, G.B; Verdin, J.P.; Lietzow, R.; Melesse, A.M. Global reference evapotranspiration modeling and evaluation. J. Am. Water Resourc. Assoc 2008, 44, 969–979. [Google Scholar]
  34. Kanamitsu, M. Description of the NMC global data assimilation and forecast system. Weather Forecast 1989, 4, 335–342. [Google Scholar]
  35. Jung, M.; Reichstein, M.; Bondeau, A. Towards global empirical upscaling of FLUXNET eddy covariance observations: Validation of a model tree ensemble approach using a biosphere model. Biogeosciences 2009, 6, 2001–2013. [Google Scholar]
  36. Singh, R.K. Geospatial Approach for Estimating Land Surface Evapotranspiration. Ph.D. Dissertation,. University of Nebraska-Lincoln: Lincoln, NE, USA, 2009. [Google Scholar]
  37. Singh, R.K.; Liu, S.; Tieszen, L.L.; Suyker, A.E.; Verma, S.B. Estimating seasonal evapotranspiration from temporal satellite images. Irrig. Sci 2012, 30, 303–313. [Google Scholar]
  38. Dore, S.; Montes-Helu, M.; Hart, S.C.; Hungate, B.A.; Koch, G.W.; Moon, J.B.; Finkral, A.J.; Kolb, T.E. Recovery of ponderosa pine ecosystem carbon and water fluxes from thinning and stand-replacing fire. Glob. Chang. Biol 2012, 18, 3171–3185. [Google Scholar]
  39. Kurc, S.A.; Benton, L.M. Digital image-derived greenness links deep soil moisture to carbon uptake in a creosotebush-dominated shrubland. J. Arid Environ 2010, 74, 585–594. [Google Scholar]
  40. Scott, R.L.; Jenerette, G.D.; Potts, D.L.; Huxman, T.E. Effects of seasonal drought on net carbon dioxide exchange from a woody-plant-encroached semiarid grassland. J. Geophys. Res.: Biogeosci 2009, 114, G04004. [Google Scholar]
  41. Scott, R.L.; Hamerlynck, E.P.; Jenerette, G.D.; Moran, M.S.; Barron-Gafford, G. Carbon dioxide exchange in a semidesert grassland through drought-induced vegetation change. J. Geophys. Res.: Biogeosci 2010. [Google Scholar] [CrossRef]
  42. Scott, R.L.; Edwards, E.A.; Shuttleworth, W.J.; Huxman, T.E.; Watts, C.; Goodrich, D.C. Interannual and seasonal variation in fluxes of water and carbon dioxide from a riparian woodland ecosystem. Agric. For. Meteorol 2004, 122, 65–84. [Google Scholar]
  43. Baldocchi, D.; Falge, E.; Gu, L.; Olson, R.; Hollinger, D.; Running, S.; Anthoni, P.; Bernhofer, C.; Davis, K.; Evans, R.; et al. FLUXNET: A new tool to study the temporal and spatial variability of ecosystem-scale carbon dioxide, water vapor, and energy flux densities. Bull. Am. Meteorol. Soc 2001, 82, 2415–2431. [Google Scholar]
  44. Barron-Gafford, G.A.; Scott, R.L.; Jenerette, G.D.; Hamerlynck, E.P.; Huxman, T.E. Temperature and precipitation controls over leaf and ecosystem-level CO2 flux along a woody plant encroachment gradient. Glob. Chang. Biol 2012, 18, 1389–1400. [Google Scholar]
  45. Wickham, J.D.; Stehman, S.V.; Gass, L.; Dewitz, J.; Fry, J.A.; Wade, T.G. Accuracy assessment of NLCD 2006 land cover and impervious surface. Remote Sens. Environ 2013, 130, 294–304. [Google Scholar]
  46. Stehman, S.V.; Milliken, J.A. Estimating the effect of crop classification error on evapotranspiration derived from remote sensing in the Lower Colorado River basin, USA. Remote Sens. Environ 2007, 106, 217–227. [Google Scholar]
  47. Haddeland, I.; Lettenmaier, D.P.; Skaugen, T. Effects of irrigation on the water and energy balances of the Colorado and Mekong River basins. J. Hydrol 2006, 324, 210–223. [Google Scholar]
  48. Westenburg, C.L.; Harper, D.P.; DeMeo, G.A. Evapotranspiration by Phreatophytes along the Lower Colorado River at Havasu National Wildlife Refuge, Arizona; US Geological Survey Scientific Investigations Report 2006–5043; US Department of the Interior: Washington, DC, USA, 2006; p. 44. [Google Scholar]
  49. Doody, T.M.; Nagler, P.M.; Glenn, E.P.; Moore, G.W.; Morino, K.; Hultine, K.R.; Benyon, R.G. Potential for water salvage by removal of non-native woody vegetation from dryland river systems. Hydrol. Process 2011, 25, 4117–4131. [Google Scholar]
  50. Nagler, P.L.; Scott, R.L.; Westenburg, C.; Cleverly, J.R.; Glenn, E.P.; Huete, A.R. Evapotranspiration on western US rivers estimated using the enhanced vegetation index from MODIS and data from eddy covariance and Bowen ratio flux towers. Remote Sens. Environ 2005, 97, 337–351. [Google Scholar]
  51. Leclerc, M.Y.; Thurtell, G.W. Footprint prediction of scalar fluxes using a Markovian analysis. Bound.-Lay. Meteorol 1990, 52, 247–258. [Google Scholar]
  52. Bastiaanssen, W.G. SEBAL-based sensible and latent heat fluxes in the irrigated Gediz Basin, Turkey. J. Hydrol 2000, 229, 87–100. [Google Scholar]
  53. Twine, T.E; Kustas, W.P.; Norman, J.M.; Cook, D.R.; Houser, P.R.; Meyers, T.P.; Prueger, J.H.; Starks, P.J.; Wesely, M.L. Correcting eddy covariance flux underestimates over a grassland. Agric. For. Meteorol 2000, 103, 279–300. [Google Scholar]
  54. Wilson, K.; Goldstein, A.; Falge, E.; Aubinet, M.; Baldocchi, D.; Berbigier, P.; Bernhofer, C.; Ceulemans, R.; Dolman, H.; Field, C.; et al. Energy balance closure at FLUXNET sites. Agric. For. Meteoro 2002, 113, 223–243. [Google Scholar]
  55. Hollinger, D.Y.; Richardson, A.D. Uncertainty in eddy covariance measurements and its application to physiological models. Tree Physiol 2005, 25, 873–885. [Google Scholar]
  56. Velpuri, N.M.; Senay, G.B.; Singh, R.K.; Bohms, S.; Verdin, J.P. A comprehensive evaluation of two MODIS evapotranspiration products over the conterminous United States: Using point and gridded FLUXNET and water balance ET. Remote Sens. Environ 2013, 139, 35–49. [Google Scholar]
  57. Thornton, P.E.; Running, S.W.; White, M.A. Generating surfaces of daily meteorological variables over large regions of complex terrain. J. Hydrol 1997, 190, 214–251. [Google Scholar]
  58. Mitchell, K.E.; Lohmann, D.; Houser, P.R.; Wood, E.F.; Schaake, J.C.; Robock, A.; Cosgrove, B.A.; Sheffield, J.; Duan, Q.; Luo, L.; et al. The multi-institution North American land data assimilation system (NLDAS): Utilizing multiple GCIP products and partners in a continental distributed hydrological modeling system. J. Geophys. Res 2004. [Google Scholar] [CrossRef]
Figure 1. Location of the study area showing the Colorado River Basin.
Figure 1. Location of the study area showing the Colorado River Basin.
Remotesensing 06 00233f1
Figure 2. Land use/land cover distribution in the Colorado River Basin based on National Land Cover Database 2006 [24].
Figure 2. Land use/land cover distribution in the Colorado River Basin based on National Land Cover Database 2006 [24].
Remotesensing 06 00233f2
Figure 3. Coverage of Landsat TM/ETM+ images (path/row) in the Colorado River Basin based on the worldwide reference system (WRS-2).
Figure 3. Coverage of Landsat TM/ETM+ images (path/row) in the Colorado River Basin based on the worldwide reference system (WRS-2).
Remotesensing 06 00233f3
Figure 4. Spatial distribution of evapotranspiration in the Colorado River Basin for 2010 using Landsat images.
Figure 4. Spatial distribution of evapotranspiration in the Colorado River Basin for 2010 using Landsat images.
Remotesensing 06 00233f4
Figure 5. Histogram of the annual evapotranspiration for selected land cover classes (a) Barren land; (b) Grassland; and (c) Cropland within the Landsat path 37/row37.
Figure 5. Histogram of the annual evapotranspiration for selected land cover classes (a) Barren land; (b) Grassland; and (c) Cropland within the Landsat path 37/row37.
Remotesensing 06 00233f5
Figure 6. Validation of Landsat-based annual evapotranspiration with eddy covariance measured evapotranspiration at Flagstaff managed forest (FMF), Flagstaff unmanaged forest (FUF), Flagstaff wildfire (FWF), Santa Rita Creosote (SRC), Santa Rita Mesquite (SRM), Kendall Grassland (WKG), and Charleston Mesquite (CMS) sites. Removal of FUF and FWF sites resulted in y = 0.997x − 53.767 (R2 = 0.95) and RMSE of 68 mm.
Figure 6. Validation of Landsat-based annual evapotranspiration with eddy covariance measured evapotranspiration at Flagstaff managed forest (FMF), Flagstaff unmanaged forest (FUF), Flagstaff wildfire (FWF), Santa Rita Creosote (SRC), Santa Rita Mesquite (SRM), Kendall Grassland (WKG), and Charleston Mesquite (CMS) sites. Removal of FUF and FWF sites resulted in y = 0.997x − 53.767 (R2 = 0.95) and RMSE of 68 mm.
Remotesensing 06 00233f6
Figure 7. Annual evapotranspiration values for 5 × 5 Landsat pixels centered over the eddy covariance tower at the Flagstaff managed forest (FMF), Flagstaff unmanaged forest (FUF), Flagstaff wildfire (FWF), Santa Rita Creosote (SRC), Santa Rita Mesquite (SRM), Kendall Grassland (WKG), and Charleston Mesquite (CMS) sites.
Figure 7. Annual evapotranspiration values for 5 × 5 Landsat pixels centered over the eddy covariance tower at the Flagstaff managed forest (FMF), Flagstaff unmanaged forest (FUF), Flagstaff wildfire (FWF), Santa Rita Creosote (SRC), Santa Rita Mesquite (SRM), Kendall Grassland (WKG), and Charleston Mesquite (CMS) sites.
Remotesensing 06 00233f7
Figure 8. Comparison of mean annual evapotranspiration based on water balance (Precipitation–runoff) and Landsat-based mean annual evapotranspiration for (a) all HUC8 sub-basins; and (b) HUC8 sub-basins meeting three criteria: precipitation > runoff, Evapotranspiration < precipitation, and runoff coefficient (Q/P) < 0.55.
Figure 8. Comparison of mean annual evapotranspiration based on water balance (Precipitation–runoff) and Landsat-based mean annual evapotranspiration for (a) all HUC8 sub-basins; and (b) HUC8 sub-basins meeting three criteria: precipitation > runoff, Evapotranspiration < precipitation, and runoff coefficient (Q/P) < 0.55.
Remotesensing 06 00233f8
Figure 9. Spatial distribution of evapotranspiration in the Colorado River Basin for 2010 using MODIS images.
Figure 9. Spatial distribution of evapotranspiration in the Colorado River Basin for 2010 using MODIS images.
Remotesensing 06 00233f9
Figure 10. Comparison of mean annual evapotranspiration of sub-basins (HUC8) using Landsat and MODIS images.
Figure 10. Comparison of mean annual evapotranspiration of sub-basins (HUC8) using Landsat and MODIS images.
Remotesensing 06 00233f10
Figure 11. Example showing the spatial distribution of annual evapotranspiration map of the same area within the Colorado River Basin with (a) zoomed in view of the Landsat-based evapotranspiration map for the area in the black square; and (b) zoomed in view of the MODIS-based evapotranspiration map in the black square area. Legend and scale shown at the bottom are for images Figure 11a,b.
Figure 11. Example showing the spatial distribution of annual evapotranspiration map of the same area within the Colorado River Basin with (a) zoomed in view of the Landsat-based evapotranspiration map for the area in the black square; and (b) zoomed in view of the MODIS-based evapotranspiration map in the black square area. Legend and scale shown at the bottom are for images Figure 11a,b.
Remotesensing 06 00233f11
Table 1. Land use/land cover distribution within the study area based on National Land Cover Database [24].
Table 1. Land use/land cover distribution within the study area based on National Land Cover Database [24].
Land Use/Land CoverNLCD Code%
Water110.40
Ice/snow120.05
Open space development210.79
Low intensity development220.51
Med. intensity development230.24
High intensity development240.05
Barren land313.34
Deciduous forest413.51
Evergreen forest4219.35
Mixed forest430.38
Shrubland5260.56
Grassland717.73
Pasture/Hay811.44
Cropland820.83
Woody Wetland900.58
Herbaceous Wetland950.22
Table 2. Details of eddy covariance tower validation sites used in this study.
Table 2. Details of eddy covariance tower validation sites used in this study.
Sl. No.Site NameCodeLatitude (Degree)Longitude (degree)Elevation (m)Tower Height (m)Land CoverLandsat Path/RowNo. of Cloud-Free ImagesReference
1Flagstaff managed forestFMF35.1426−111.7273216023Ponderosa pine forest37/366Dore et al. (2012) [38]
2Flagstaff unmanaged forestFUF35.089−111.762218023Ponderosa pine forest37/366Dore et al. (2012) [38]
3Flagstaff wildfireFWF35.4454−111.771822704Ponderosa pine forest37/357Dore et al. (2012) [38]
4Santa Rita CreosoteSRC31.9083−110.83959914.25Open shrub land36/389Kurc and Benton (2010) [39]
5Santa Rita MesquiteSRM31.8214−110.866111206.4Woody Savannas36/389Scott et al. (2009) [40]
6Kendall GrasslandWKG31.7365−109.941915316.4Grassland35/387Scott et al. (2010) [41]
7Charleston MesquiteCMS31.6637−110.1776120014Riparian woodland35/387Scott et al. (2004) [42]
Table 3. Annual evapotranspiration (mm) during 2010 for different National Land Cover Database [24] land use/land cover within the study area.
Table 3. Annual evapotranspiration (mm) during 2010 for different National Land Cover Database [24] land use/land cover within the study area.
Land Use/Land coverMean Annual ETa (mm)Standard Deviation (mm)
Water994699
Ice/snow405168
Open space development323227
Low intensity development374245
Med. intensity development397213
High intensity development386222
Barren land283198
Deciduous forest434189
Evergreen forest447223
Mixed forest515186
Shrubland262182
Grassland224144
Pasture/Hay365271
Cropland538393
Woody Wetland421301
Herb. Wetland337283

Share and Cite

MDPI and ACS Style

Singh, R.K.; Senay, G.B.; Velpuri, N.M.; Bohms, S.; Scott, R.L.; Verdin, J.P. Actual Evapotranspiration (Water Use) Assessment of the Colorado River Basin at the Landsat Resolution Using the Operational Simplified Surface Energy Balance Model. Remote Sens. 2014, 6, 233-256. https://0-doi-org.brum.beds.ac.uk/10.3390/rs6010233

AMA Style

Singh RK, Senay GB, Velpuri NM, Bohms S, Scott RL, Verdin JP. Actual Evapotranspiration (Water Use) Assessment of the Colorado River Basin at the Landsat Resolution Using the Operational Simplified Surface Energy Balance Model. Remote Sensing. 2014; 6(1):233-256. https://0-doi-org.brum.beds.ac.uk/10.3390/rs6010233

Chicago/Turabian Style

Singh, Ramesh K., Gabriel B. Senay, Naga M. Velpuri, Stefanie Bohms, Russell L. Scott, and James P. Verdin. 2014. "Actual Evapotranspiration (Water Use) Assessment of the Colorado River Basin at the Landsat Resolution Using the Operational Simplified Surface Energy Balance Model" Remote Sensing 6, no. 1: 233-256. https://0-doi-org.brum.beds.ac.uk/10.3390/rs6010233

Article Metrics

Back to TopTop