Purpose and Scope

The purpose of this report is to summarize the methods and results used to estimate annual ET_g for the AWSR study area (fig. 1). The 28,229,000 square meter (m²) AWSR study area is comprised of five GDAs along the AWSR and China Ranch Wash, and in Chicago and California Valleys. This study uses site-scale ET_g measured at two ET-monitoring sites in the AWSR study area—a sparse shrub site and a dense wetland site. Data were collected at these sites from March 2017 to May 2019, and ET_g was calculated for January 2018–19. Total ET was measured using the eddy-covariance method and measurements were adjusted using an energy-balance closure method. Precipitation was subtracted from total adjusted ET to calculate site-scale ET_g. Groundwater discharge areas in the study area were visually delineated using high-resolution aerial imagery, subdivided into open-water, moist-soil, and vegetated ET units, and refined by field verification. An ET unit is an area with a uniform ET rate, such as the open-water and moist-soil ET units, or an area with a variable but predictable ET rate, such as the vegetated ET unit. The open-water and moist-soil ET units were assigned site-scale ET_g based on Laczniak and others (2001) and Jackson and others (2018). The vegetated ET units were assigned site-scale ET_g that varied according to a quadratic equation developed for this study that relates site-scale ET_g to a normalized difference vegetation index (NDVI) derived from aerial imagery acquired July 2, 2020. Annual ET_g for the AWSR study area was calculated by adding the annual ET_g for all the ET units in the study area.

This report presents an estimate of annual ET_g for the AWSR study area and provides baseline groundwater-discharge data to support a mandated comprehensive management plan for the AWSR (Wild and Scenic Rivers Act, October 2, 1968). A supporting geospatial dataset of the ET-unit and GDA boundaries and a supporting dataset of flux and ET data are presented in associated USGS data releases (Damar and Pavelko, 2023, and Pavelko, 2023). The annual ET_g estimate helps constrain past estimates because this study measured ET at sites within the study-area boundaries and used multispectral data that were collected within 2 years of the ET-data collection. Previous ET_g studies measured ET at locations outside of the study area and used multispectral data that were collected before ET was measured (Laczniak and others, 2001) or used ET that was empirically derived and 30-year composite multispectral data (Huntington and others, 2016). The ET-unit and GDA boundaries can be compared with future delineations to help determine any changes to the ET units or GDAs. Data collected during the 1-year period of analysis for the Shrub ET-monitoring site (USGS site 355846116160401) and Wetland ET-monitoring site (USGS site 355918116161801) are stored in the USGS National Water Information System (NWIS) and are publicly accessible at the NWIS web portal (U.S. Geological Survey, 2020) using their NWIS site numbers. Data collected for this study and stored in NWIS include air temperature, relative humidity, wind speed, wind direction, net radiation, atmospheric water-vapor density, soil temperature, volumetric soil-moisture content, soil-heat flux, sensible-heat flux, latent-heat flux, and total ET.

Previous Studies

There have been two reconnaissance-level ET_g studies of the AWSR study area (Laczniak and others, 2001, and Huntington and others, 2016) and two studies that estimated moist-soil ET in Death Valley (DeMeo and others, 2003, and Jackson and others, 2018). Laczniak and others (2001) delineated and subdivided groundwater discharge areas in the study area and used site-scale ET_g measured from 1994 to 1997 in Ash Meadows, Nevada (Laczniak and others, 1999), about 60-km north of the AWSR study area, and values from a 1992 vegetation index to estimate an annual ET_g of 11,015,000 cubic meters (m³). Laczniak and others (2001) subdivided discharge areas based on plant communities and their similarities in multispectral imagery. Huntington and others (2016) used the discharge areas delineated by Laczniak and others (2001) and used empirically derived site-scale ET_g across the discharge areas (Beamer and others, 2013) and a multiple-year composite (1984–2015) of remotely sensed multispectral data to estimate an annual ET_g of 9,923,000 m³.

Study Area

The AWSR study area comprises five GDAs, the Amargosa River GDA in Inyo and San Bernardino Counties, Calif., and the Shoshone, Chicago Valley, Resting Spring, and California Valley GDAs in Inyo County, Calif. (fig. 1). The Amargosa River GDA also includes the tributary China Ranch Wash. The China Ranch Wash and Chicago and California Valleys are included in the study area because they contribute regional groundwater to the AWSR. In the study area, regional groundwater derived from interbasin flow is the primary source of recharge, ET_g is the primary source of discharge, intermittent AWSR flow out of the study area is the secondary source of discharge, and all other water-budget components are considered negligible (Malmberg, 1967; Harrill, 1986; Harrill and Prudic, 1998; Belcher and others, 2019; Halford and Jackson, 2020). Interbasin groundwater flow primarily comes from the east, in Pahrump Valley, where precipitation on the Spring Mountains infiltrates a carbonate mountain block and recharges the regional groundwater flow system. From Pahrump Valley, groundwater flows beneath the Nopah Range and recharges the Chicago Valley, Resting Spring, and California Valley GDAs before flowing toward the Shoshone and Amargosa River GDAs. In the AWSR study area, annual precipitation estimates range from about 500,000 to 1,100,000 m³ (Hevesi and others, 2003; Belcher and others, 2017; Halford and Jackson, 2020) whereas recharge by interbasin groundwater estimates range from about 11,200,000 to 19,900,000 m³, with about 600,000 m³ coming from the Amargosa Desert to the north and the remainder coming from Pahrump Valley to the east (fig. 1; Harrill and others, 1988; Belcher and others, 2019).

Flow in the AWSR is perennial in most of the study area where there are gaining and losing sections. However, flow is ephemeral into the study area and intermittent out of the study area with flow occurring in these areas primarily during and after rainfall (Belcher and others, 2019). Based on geochemical analyses of spring discharges and insufficient precipitation to sustain the spring discharges, Belcher and others (2019) determined most springs in the study area discharge regional groundwater. Study-area springs with discharge greater than about 0.3 liter per second (L/s) and water temperatures greater than about 30-degrees Celsius (°C) include Shoshone Spring in the Shoshone GDA and Borehole, Borax, and Willow Springs, Tecopa Hot Springs, and the unnamed seeps and springs in Amargosa Canyon in the Amargosa River GDA (The Source Group, Inc., 2011). Thom Spring, in the Amargosa River GDA has discharge greater than about 0.3 L/s (The Source Group, Inc., 2011). Spring and seep discharge in the study area does not substantially contribute to regional recharge or discharge because nearly all spring and seep water that discharges to the surface subsequently is discharged as ET_g (Belcher and others, 2019), with the remainder discharging from the study area as intermittent surface-water flow in the Amargosa River. Borehole Spring, Borax Spring, and Thom Spring are geographic features that are not named in the Geographic Names Information System (Yost and Carswell, 2009) but have been identified and named in The Source Group, Inc. (2011) and Belcher and others (2019).

The study area has typical Mojave Desert climate, with long hot summers, where most precipitation is from widespread, long-duration storms during cooler months and from shorter-duration, thunderstorms during warmer months (Hereford and others, 2004). In Shoshone, from 1981 to 2010, the mean annual minimum temperature was about 13 °C and the mean annual maximum temperature was about 28 °C (Western Regional Climate Center, 2020). In Shoshone, the mean annual precipitation from 1981 through 2010 was 123 millimeters (mm), with about 60 percent of precipitation falling in December through March, and the mean-annual air temperature was about 21 °C (Western Regional Climate Center, 2020). In Tecopa, the mean annual precipitation from 2007 through 2020 was 85 mm, and annual precipitation in 2018 was 74 mm (Community Environmental Monitoring Program, 2022), or about 87 percent of the 2007–20 Tecopa average.

Vegetation in the study area predominantly is phreatophytes interspersed with xerophytes, which is typical of areas in the Mojave Desert with shallow regional groundwater (Nichols, 2000). Discharge from phreatophytes often is the largest component of regional discharge in the Basin and Range Province (Harrill and Prudic, 1998; Nichols, 2000). Phreatophytes require a constant water source and have relatively deep root systems, as much as about 20 m below land surface, and can obtain water from deeper regional aquifers, shallow local systems, and soil moisture derived from precipitation (Meinzer, 1923; Robinson, 1958). Xerophytes are plants that generally have relatively shallow root systems that obtain water stored as soil moisture. Phreatophytes and xerophytes grow together in areas where the depth to water is shallow, but in areas where the depth to water is deeper than about 20 m below land surface, only xerophytes grow. Xerophytes that grow among phreatophytes consume shallow regional groundwater and contribute to regional ET_g. Phreatophytes in the study area grow in riparian and wetland areas, woodlands, shrublands, and grasslands. Riparian areas occur along perennial water ways and springs and can be a mixture of wetland vegetation, trees, shrubs, and grasses. Common wetland plants in the study area include American bulrush (Schoenoplectus americanus), common reeds (Phragmites australis), rushes (Juncus sp.), cat tails (Typha sp.), and wiregrass (Juncus tenuis). Common trees in woodland areas in the study area include western honey mesquite (Prosopis glandulosa var. torreyana), desert willow (Chilopsis linearis), tamarisk (Tamarix ramosissima), and date palm (Phoenix dactylifera). Other trees in the study area include screwbean mesquite (Prosopis pubescens), athel tamarisk (Tamarix aphylla), Mexican fan palm (Washingtonia robusta), California fan palm (Washingtonia filifera), and Fremont’s cottonwood (Populus fremontii). A mesquite bosque is a dense, near-monoculture of mesquite trees, typically growing along floodplains or arroyos; in the study area, bosques primarily consist of western honey mesquite. In the study area, common shrubs include greasewood (Sarcobatus vermiculatus), bush seepweed (Suaeda nigra), and quailbush (Atriplex lentiformis), and common grasses include saltgrass (Distichlis spicata) and alkali sacaton (Sporobolus airoides). Some of the most common xerophytes in the study area are creosote bush (Larrea tridentata), white bursage (Ambrosia dumosa), blackbrush (Coleogyne ramosissima), and brittlebush (Encelia farinosa).

For this study, a GDA is defined as an area that has relatively shallow groundwater that naturally discharges from a groundwater system, and GDAs are subdivided into open-water, moist-soil, and vegetated ET units. Open-water ET units consist of ponds and reservoirs. Moist-soil ET units consist primarily of unvegetated areas with soil that is perennially moist from regional groundwater, sometimes salty or with salt encrustations, sparse phreatophytes, or both. Vegetated ET units primarily consist of phreatophytes, which most commonly grow along the banks and floodplains of the AWSR and China Ranch Wash, near regional springs and seeps, and in riparian and wetland areas, woodlands, shrublands, and grasslands.

The Chicago Valley GDA is in central Chicago Valley, which is bounded on the north and west by the Resting Spring Range and on the south and east by the Nopah Range (fig. 1). Interbasin groundwater flow from Pahrump Valley is the primary source of water for the Chicago Valley GDA (Belcher and others, 2019). The Chicago Valley GDA includes one periodically pumped well (Belcher and others, 2019), but the pumping contributes negligibly to regional groundwater discharge. Groundwater that does not discharge in the GDA either flows west, under the Resting Spring Range and toward the AWSR, or flows south and toward the Resting Spring GDA (Belcher and others, 2019; Halford and Jackson, 202041). There is no surface-water outflow from the Chicago Valley GDA and there are no open-water or moist-soil ET units. Vegetated ET units in the Chicago Valley GDA primarily consist of western honey mesquite bosques separated by dry sandy soils, sparse desert shrubs, and other xerophytes. In Chicago Valley, the sizes of bosques generally are less than 3,000 m², but there are longer, linear bosques as much as about 100,000 m² that grow in dry arroyos.

The Resting Spring GDA is in southern Chicago Valley, along the southeastern and southern end of the Resting Spring Range, and it is associated with Resting Spring (fig. 1). Interbasin groundwater flow from Pahrump Valley and California Valley are the primary source of water for the Resting Spring GDA (Belcher and others, 2019). Groundwater that does not discharge in the GDA flows toward the AWSR to the west. All Resting Spring discharge is consumed primarily by phreatophytes and secondarily by Resting Springs Ranch; there is no surface-water outflow, and there are no open-water or moist-soil ET units in the Resting Spring GDA. Water use at the ranch is considered negligible. The GDA primarily consists of dense western honey mesquite bosques growing in dry washes, and the ranch includes turf grass, mixed palm trees, and Fremont’s cottonwood. Bosques generally are less than about 40,000 m² and separated by areas with dry soils and sparse xerophytes, although there is one bosque that is about 500,000 m². Resting Spring is developed and sustains the ranch; discharge from Resting Spring was about 9.5 L/s on January 23, 2011 (The Source Group, Inc., 2011).

The California Valley GDA is in the west-central part of California Valley, which is bounded on the northeast by an indistinct border with foothills of the Nopah and Kingston Ranges, on the southeast by the Kingston Range, on the south by the Dumont Hills, and on the west by the Nopah Range (fig. 1). Interbasin groundwater flow from Pahrump Valley is the primary source of water for the California Valley GDA (Belcher and others, 2019). Regional groundwater that does not discharge from the GDA flows south and discharges into the China Ranch Wash, a tributary of the AWSR (Belcher and others, 2019). There is no surface-water outflow and there are no open-water or moist-soil ET units in the California Valley GDA. Vegetation in the California Valley GDA primarily consists of western honey mesquite bosques separated by sandy soil, sometimes with sparse saltgrass. The California Valley GDA includes Tule Spring, which no longer discharges water (The Source Group, Inc., 2011).

The Shoshone GDA is along the uppermost section of the AWSR, from about 2 km north of Shoshone to about 7 km south of Shoshone, and both ET-monitoring sites used for this study were located there (fig. 1). Interbasin groundwater flow from Pahrump and Chicago Valleys is the primary source of water for the Shoshone GDA (Belcher and others, 2019). Flow in the AWSR is perennial in much of the GDA but flow becomes intermittent near the southern boundary. Groundwater that does not discharge in the GDA flows south through a shallow groundwater system and supplies water to the Amargosa River GDA near Tecopa (Belcher and others, 2019). The Shoshone GDA has open-water, moist-soil, and vegetated ET units. The only open-water ET unit is a pond in Shoshone that is about 2,400 m². Moist-soil ET units generally are found in the southern part of the GDA along the AWSR floodplain and typically are salt-encrusted and have sparse saltgrass; soil-grain sizes are variable, including coarser sediments transported during occasional flooding. Vegetated ET units occur along the AWSR and floodplain and include a variety of phreatophytes, including sparse to dense saltgrass and shrubs, wetland and riparian vegetation, and mesquite bosques. The Shoshone GDA also includes Shoshone Spring, which discharges about 14–16 L/s (The Source Group, Inc., 2011; Belcher and others, 2019) and is the sole municipal water supply for Shoshone. Shoshone Spring discharge is used for drinking water for Shoshone residents, irrigation water for residences, a municipal pool, the pond, a wetlands park near Shoshone Spring, and a grassy park (about 30,000 m²) with large (some taller than about 30 m) phreatophyte trees, including athel tamarisk and Mexican fan palm. All Shoshone Spring discharge, except the amount consumed by humans, is either consumed by ET or flows into the AWSR.

The Amargosa River GDA is along the AWSR, from the Tecopa Hot Springs area to the southern Sperry Hills, and along the China Ranch Wash, from about 2 km east of Tecopa Pass to its confluence with the AWSR below China Ranch (fig. 1). Interbasin groundwater flow from Pahrump and California Valleys is the primary source of water for the Amargosa River GDA (Belcher and others, 2019). Evapotranspiration consumes nearly all water in the GDA, with no substantial subsurface discharge and only intermittent surface-water discharge (Belcher and others, 2019). The Amargosa River GDA includes domestic pumping in the Tecopa area and at China Ranch (Belcher and others, 2019), but the pumping contributes negligibly to regional groundwater discharge. The Amargosa River GDA has open-water, moist-soil, and vegetated ET units. Open-water ET units are found as small ponds and reservoirs in the Tecopa Hot Springs area and near Willow Spring. Moist-soil ET units are found in the playa area west of the Tecopa Hot Springs area and generally consist of clay to fine-grained sediments that sometimes are salty or salt-encrusted, similar to playa sediments described for Death Valley (DeMeo and others, 2003; Jackson and others, 2018). Vegetated ET units include a variety of phreatophytes, including sparse to dense grasses and shrubs, wetland and riparian vegetation, and western honey mesquite, screwbean mesquite, desert willow, and tamarisk woodlands. The Amargosa River GDA includes China Ranch, which grows about 100,000 m² of various date palms, and Willow Spring, which supplies water to a reservoir (about 5,250 m²) for China Ranch. From Tecopa to the confluence with the China Ranch Wash, the AWSR has gaining and losing sections (Belcher and others, 2019). The China Ranch Wash, which is tributary to the AWSR, is intermittent below Willow Spring. Other seeps and springs in the GDA include Borehole Spring, Borax Spring, Thom Spring, Wild Bath Spring, seeps and springs in the Amargosa Canyon and China Ranch Wash, and hot seeps and springs in the Tecopa Hot Springs area (fig. 1). Wild Bath Spring is a geographic feature that is not named in the Geographic Names Information System (Yost and Carswell, 2009) but has been identified and named in The Source Group, Inc. (2011) and Belcher and others (2019).

Four USGS streamgages in the Amargosa River GDA continuously monitor discharge from the Amargosa River, China Ranch Wash, and Borehole Spring (fig. 1), and another streamgage is on the Amargosa River about 6 km downstream from the study area boundary. Data from the streamgages are stored in NWIS and are publicly accessible at the NWIS web portal (U.S. Geological Survey, 2020) by using their NWIS site numbers. The Amargosa River at Tecopa, Calif. (USGS streamgage 10251300), has been measured periodically since 1962, nearly continuously since 1983, and continuously since 2000; the average annual discharge for water years 2000–19 is 70 L/s. The Amargosa River above China Ranch Wash, near Tecopa, Calif. (USGS streamgage 10251330), has been measured continuously since 2007 and the average annual discharge for water years 2007–19 is 84 L/s. Willow Spring, which discharges perennial flow into the China Ranch Wash, is monitored by Willow Creek at China Ranch, Calif. (USGS streamgage 10251335), and has been monitored continuously since 2014, and the average annual discharge for water years 2014–19 is 4 L/s. Borehole Spring channel, near Tecopa Hot Springs, Calif. (USGS streamgage 10251290), which discharges perennial flow to the AWSR near Tecopa Hot Springs, has been monitored continuously since 2014, and the average annual discharge for water years 2014–19 is 7 L/s. The Amargosa River at Dumont Dunes, near Death Valley, Calif. (USGS streamgage 10251375), which is outside of the study area, was periodically measured from 1998 to 2001, continuously measured from 1999 to 2001, and the average annual discharge for water years 2000–01 is 42 L/s.

The Shoshone and Amargosa River GDAs, which are along the perennial Amargosa River and China Ranch Wash, differ from the Chicago Valley, Resting Spring, and California Valley GDAs, which do not have perennial surface water. In the Shoshone and Amargosa River GDAs, there is open water, moist soil, and phreatophytes that generally grow along the AWSR and China Ranch Wash in long and narrow patches that have a high diversity of species and a wide range of vegetation density. In contrast, the Chicago Valley, Resting Spring, and California Valley GDAs, contain no open water or moist soil, and phreatophytes generally consist of pockets of dense woodlands, typically western honey mesquite bosques, separated by dry soils and xerophytes and sometimes with very sparse desert grasses or shrubs. The different vegetation characteristics likely are a result of the shallow groundwater that is available to the phreatophytes along the AWSR as opposed to the deeper groundwater in the other GDAs that is accessible only to phreatophytes with deeper root systems. Groundwater levels measured in Chicago Valley (USGS site 360020116103701; USGS site 360005116114902; and USGS site 360005116114901) and California Valley (USGS site 355129116025701; fig. 1B) generally are more than 6 m to the water table and springs generally have gone dry in these valleys (Belcher and others, 2019). In this setting, mesquite trees, with their deep root systems, can reach the water table, whereas phreatophytic shrubs and grasses cannot.

Energy Budget

The site-scale surface-energy budget for this study includes available energy—net radiation (R_n) and soil-heat flux at land surface (G)—and turbulent fluxes—sensible-heat flux (H) and latent-heat flux (λE). Energy budget components are evaluated to assess and constrain the turbulent-flux measurements that are used to calculate site-scale ET (Bowen, 1926; Penman, 1956; Montieth, 1973; Brutsaert, 1982). Heat flux, as used here, is a term that describes the amount of heat energy transferred to or from the land surface. Available energy is the energy that drives land-surface processes and turbulent fluxes are the energy that is exchanged or consumed by land-surface processes. Based on the law of conservation of energy, available energy equals turbulent fluxes plus any energy that may be in storage (Brutsaert, 1982; Stull, 1988). For this study, energy in storage was considered negligible and therefore,

where

During the daytime, when the sun is unobstructed by clouds, available energy primarily comes from incoming solar radiation (heat energy). Where water is limited, such as in the AWSR study area, most of the heat energy first is transferred into latent energy, which drives ET, and then, as water becomes even more limited due to ET, the incoming heat energy is transferred to sensible heat, which warms the air. Net radiation, which is the heat flux of atmospheric air, is the sum of incoming long- and short-wave solar radiation and reflected long-wave radiation from the Earth’s surface. Net radiation, which is directly measured in this study, is positive during daytime when heating the Earth’s surface (fig. 2). Net radiation is the primary driver of ET, because it heats the Earth’s surface and causes it to be warmer than the overlying air and the underlying soil. Soil-heat flux at land surface is the heat energy transferred by conduction between near-surface soils and near-surface air and is negative when heat energy is transferred into near-surface soils. Sensible-heat flux is the heat energy transferred by convection between the Earth’s surface and near-surface air. When R_n is positive and warming the Earth’s surface, H is negative, as heat energy in the near-surface air is transferred upward and into cooler atmospheric air. Latent-heat flux is the energy required to drive a liquid-to-vapor phase change and when there is available energy and water, λE is positive and ET can occur. Thus, excess daytime available energy heats soils and drives turbulent fluxes and ET. After the sun sets, available energy sharply decreases, the energy-budget becomes unstable, turbulent fluxes fluctuate around zero, and ET essentially stops (fig. 2; Stull, 1988).

Soil-heat flux at land surface is calculated from measurements of soil-heat flux at depth (G_d), soil temperatures (T_p and T_c), and volumetric soil-water content (θ_v):

where Heat capacity of soil is calculated as: where Gravimetric soil-water content, in fraction of total mass, is calculated as: where

Sensible-heat flux is averaged for 30-minute periods with the eddy-covariance method using 10-Hz measurements of vertical wind speed and air temperature as:

where

Latent-heat flux, where λ is the latent heat of vaporization of water, in J/kg, and E is the evaporation rate, in meters per 30-minute period, has units of W/m² and is calculated with the eddy-covariance method using 10-Hz measurements of vertical wind speed and water-vapor density as:

where

Evapotranspiration, in m/s, is then calculated as:

where The latent heat of vaporization of water is temperature dependent and is estimated as: whereT is the air temperature, in °C (Stull, 1988, p. 641).Thus, ET, in mm per 30-minute averaging period, is calculated from measurements of λE and T as: where To determine annual ET, in mm, all 30-minute values are summed for the year.

Based on the law of conservation of energy, available energy is, on average, equal to turbulent fluxes and the surface-energy budget is in balance, but field measurements often result in an imbalance between available energy and turbulent fluxes (Twine and others, 2000; Wilson and others, 2002; Foken, 2008) and, therefore, a potential miscalculation of ET. The energy-balance ratio (EBR) for a given period is

where The EBR is used to determine the degree to which field energy measurements are out of balance. Energy balance occurs when turbulent fluxes ($\Sigma \lambda E+ \Sigma H$) are equal to available energy ($\Sigma R_n- \Sigma G$). For this study, the EBR is considered a measure of the uncertainty of the turbulent-flux measurements, where an EBR close to one is considered ideal. Energy-balance ratios are used to assess uncertainty in eddy-covariance measurements that primarily result from the inability to accurately represent larger scale eddies (Mauder and others, 2020).

Measurements

Air temperature, relative humidity, precipitation, R_n, photosynthetically active radiation (PAR), G, H, λE, and total ET were either directly measured or were calculated from direct measurements of other parameters. Data were collected at the Shrub site from March 8, 2017, to May 1, 2019, and data were collected at the Wetland site from April 20, 2017, to May 7, 2019 (U.S. Geological Survey, 2020). Sites were visited about monthly to inspect sensors and equipment, document site conditions, and download data. All sensors and equipment used at the sites (table 1) were calibrated, maintained, and positioned according to manufacturer recommendations. At both sites, data were collected for more than 1 year and were recorded and stored to a data logger every 30 minutes. Air temperature and relative humidity were directly measured at both sites with a temperature and relative humidity probe (Campbell Scientific, Inc., 2015b).

Precipitation was measured during each site visit with volumetric rain gages at both sites and with a recording tipping-bucket rain gage at the Shrub site. Volumetric rain gages, which do not record data, are more accurate than tipping-bucket rain gages because tipping-bucket rain gages can under-measure during light rains, when precipitation can evaporate before a bucket is tipped, and during intense rains, when precipitation falls too fast to register (Humphrey and others, 1997). Tipping-bucket rain gages accurately record the timing and duration of rainfall, but their volumetric totals must be corrected to the volumetric totals. Because the sites were 1 km apart, it was assumed that the timing of rainfall at the Wetlands site was similar to the measured rainfall at the Shrub site. The volumetric rain gages were standard 20.3-centimeter (cm) diameter gages (National Weather Service, 2020) measured with a 0.254-mm resolution ruler and installed such that the gage intakes were about 0.84 m above ground. A small volume of mineral oil, which floats on water, was added to the precipitation reservoirs of the volumetric gages to help prevent evaporative losses between site visits (Sevruk, 1973). The tipping-bucket rain gage (Campbell Scientific, Inc., 2010d), installed at the Shrub site, was at a height of 1.04 m above ground and recorded with a 0.254-mm resolution. During the study, snowfall was not observed at either site.

Net radiation was directly measured at both sites and PAR was directly measured only at the Shrub site. Net radiation was measured with net radiometers, where the downward-facing field of view was a circular area with a radius of 10 times the sensor height (Campbell Scientific, Inc., 2010c). At each site, the net radiometer was installed at a height and location to measure an area representative of the targeted vegetation with the downward-facing sensor and positioned to have an unobstructed upward-facing view of the sky. At the Shrub site, the height of the net radiometer captured shrubs and bare soil within a ground radius of about 28.8 m, and at the Wetland site, the height of the radiometer captured primarily dense common reeds within a ground radius of about 37.3 m. Photosynthetically active radiation, a measure of incoming photons with wavelengths that plants use during photosynthesis, was measured with a quantum sensor (Campbell Scientific, Inc., 2015a) that was positioned to have an unobstructed upward-facing view of the sky.

Soil-heat flux at land surface was calculated for both sites from below-ground measurements of G, soil temperature, and soil-moisture content (eq. 2). Soil-heat flux at depth was measured 8 centimeters (cm) below land surface with two soil-heat-flux plates (Campbell Scientific, Inc., 2016), soil temperature was measured at 2 and 6 cm below land surface with thermistors (Campbell Scientific, Inc., 2018), and soil-moisture content at 2.5 cm below land surface was measured with a water-content reflectometer (Campbell Scientific, Inc., 2011b). At each site, soil sensors were installed in representative locations of vegetation such that the land surface above the sensors received a corresponding amount of shade and direct sunlight. At the Shrub site, the soil sensors were installed next to shrubs such that the patch of soil above the sensors received a representative amount of shade and direct sunlight throughout the course of a typical day. At the Wetland site, the soil sensors were beneath about a 14-cm thick layer of dying, dead, and decaying common reeds such that the patch of soil above the sensors rarely received direct sunlight.

Sensible-heat flux and λE were calculated using the eddy-covariance method using measurements from two fast-response sensors at each site—a three-dimensional sonic anemometer used to measure vertical wind speeds and air temperatures, and a krypton hygrometer used to measure water-vapor fluctuations (Campbell Scientific, Inc., 2010a, 2010b). Sensible-heat flux was calculated from measurements of vertical wind speed and air temperature (eq. 5), and λE was calculated from measurements of vertical wind speed and water-vapor fluctuations (eq. 6). At each site, the fast-response sensors were installed at heights to capture fluxes from the targeted vegetation and oriented toward an azimuth of 208 degrees from north to best capture prevailing winds. From 2007 to 2016, wind at Tecopa generally was bidirectional, primarily coming from between the south-southeast and south (157.5–180 degrees from north) with a secondary component coming from between the northwest and north-northwest (315–337.5 degrees from north; Community Environmental Monitoring Program, 2022). At the Shrub site, the fast-response sensors were 2.14 m above land surface and measured fluxes primarily affected by sparse shrubs and dry soil. At the Wetland site, the fast-response sensors were 3.64 m above land surface and measured fluxes primarily influenced by common reeds.

Data Processing

For each ET-monitoring site, site-scale ET_g was estimated and annual daytime mean CNF values were calculated. A 1-year period of analysis was selected for each site where each period of analysis was selected to minimize missing and suspect data. Site-scale data were processed with a data-logger program and the software packages LoggerNet (Campbell Scientific, Inc., 2011a) and EdiRe (Clement, 1999), and then the data were examined and post-processed in a spreadsheet to calculate site-scale ET_g and the EBR. A data logger (Campbell Scientific, Inc., 2011c) and program were used at each site to store data and to perform initial data processing, including calculating raw turbulent-flux values using site-specific data. Site-specific data used in each data-logger program were the atmospheric pressure of the site, the orientation (azimuth) of the sonic anemometer, and the calibration constants for the hygrometer, radiometer, quantum sensor, and heat-flux plates. LoggerNet was used to convert logger output files into formats that could be read into EdiRe and a spreadsheet. EdiRe, a software package for micrometeorological applications, was used to process and correct 10-Hz data and aggregate them into 30-minute averages or covariances, and to calculate available energy, turbulent fluxes, and mean CNF values. The processed 30-minute data were examined for missing and suspect values and post-processed in a spreadsheet to calculate available energy and turbulent flux values, the annual EBR, site-scale ET_g, EBR-adjusted turbulent flux values, EBR-adjusted site-scale ET_g, and annual daytime mean CNF values.

The 10-Hz data from the fast-response sensors were corrected and processed into 30-minute H and λE values using EdiRe (Clement, 1999). Data spikes were filtered according to a method described by Højstrup (1993) and corrected to compensate for any signal lag between the anemometer and hygrometer. Raw covariances were mathematically two-dimensionally rotated to force the mean vertical and crosswind velocities to zero, which compensates for errors associated with small misalignments of the sonic anemometer (Kaimal and Finnigan, 1994). Frequency response corrections were applied to the data to compensate for the inability of the sensors to measure the flux contributions from larger (greater than 1 km) and smaller (less than 10 cm) eddies (Moore, 1986). Corrections were applied to vertical wind speeds to compensate for variations in air density that result from fluctuating temperatures and humidity (Webb and others, 1980). Latent-heat flux values were corrected to compensate for the attenuation of the hygrometer signal caused by oxygen in the krypton laser signal path (Tanner and Greene, 1989; Campbell Scientific, Inc., 1998). Sensible-heat flux values were corrected to compensate for air density and sound-path deflection, which affect the sonic-derived temperatures used to calculate H (Schotanus and others, 1983).

Data then were examined and post-processed to perform additional corrections, address missing or suspect data, determine daytime and nighttime, and to calculate site-scale ET_g. Precipitation for the Shrub site, which had a volumetric and tipping-bucket rain gage, was calculated by adjusting the tipping bucket rain gage values so that they equaled the more accurate volumetric rain gage values (World Meteorological Organization, 2008). The volumetric-corrected tipping-bucket data then were wind-speed corrected for periods when wind speeds exceeded 5 m/s because precipitation can be undermeasured if there is wind turbulence at the rain gage (World Meteorological Organization, 2008). For the Wetland site, which only had a volumetric rain gage, precipitation was calculated by first determining the linear relation between the volumetric measurements from both sites measured on the same day and then deriving the corrected amount by applying the linear relation to the measured volumes. The volumetric measurements were corrected based on the assumption that wind speeds were similar at the two sites, which were about 1 km apart. Net radiation values were corrected for periods when wind speeds exceeded 5 m/s to compensate for convective cooling caused by wind blowing across the sensors (Brotzge and Duchon, 2000; Campbell Scientific, Inc., 2010c). For periods when EdiRe-calculated H and λE data were missing or suspect, H and λE values calculated by the data logger program were frequency-corrected and used (Moore, 1986; Massman, 2000).

Data were examined for missing or suspect values, which were censored and replaced with estimated values. Net radiation, PAR, G, H, and λE values were flagged as suspect if they did not conform to expected daily or seasonal trends or if the values exceeded suggested guidelines. Suggested data-range guidelines were that R_n values should not be less than −200 W/m² or greater than 900 W/m², PAR values should not be less than zero or greater than 2,200 micromoles per second per square meter (µmol/s/m²), G values should not be less than −50 W/m² or greater than 50 W/m², and H and λE values should not be less than −150 W/m² or greater than 700 W/m² (Law and others, 2005). If a value exceeded suggested guidelines but conformed to expected trends, then the value was not flagged as suspect. If a value exceeded suggested guidelines or did not conform to expected trends, then it was censored and replaced with an estimated value. Suspect R_n, G, and H data were censored and replaced with linearly interpolated values. Suspect λE values that were collected during the daytime were censored and replaced with linearly interpolated values and suspect λE values that were collected at nighttime were censored and replaced with zero. If a consecutive series of suspect λE values occurred during the transition from nighttime to daytime, or daytime to nighttime, then the nighttime values were edited to zero and the daytime values were linearly interpolated from, or to, zero, respectively. Missing or suspect λE values at nighttime were edited to zero rather than being replaced with linearly interpolated values because there generally is very little or no available energy at nighttime and measured turbulent-flux values tend to fluctuate around zero.

For this study, daytime and nighttime were defined using R_n and PAR data. It was daytime if R_n was greater than −5 W/m² and PAR was greater than 200 µmol/s/m². If the R_n and PAR definitions disagreed (for example, if R_n was more than −5 W/m² but PAR was less than 200 µmol/s/m², or if the R_n and PAR values both did not fit expected daily temporal trends (for example, if R_n was less than −5 W/m² and PAR was less than 200 μmol/s/m² as a result of cloudy skies during the late morning or afternoon), then daytime and nighttime were delineated based on the daily trends preceding and following the suspect data.

For the Shrub site, the 1-year period of analysis, from January 4, 2018, to January 3, 2019, included 17,520 thirty-minute values. There were no missing data but there were 4 R_n values; 5,911 G values; 13 H values; and 141 λE values that were flagged as suspect (Law and others, 2005); no PAR data were suspect. There were 1,042 R_n values that were wind-speed corrected because the corresponding wind speeds exceeded 5 m/s. Suspect R_n and H values did not fit expected daily trends, so they were censored and replaced with linearly interpolated values. All suspect G values fit expected daily and seasonal trends and therefore, no G values were censored. Of the 141 suspect λE values, 90 values were censored and replaced with linearly interpolated values and 51 were replaced with zero; all values that were replaced with zero occurred at nighttime, or within 1 hour of nighttime, as defined by R_n and PAR data. When defining daytime and nighttime, there were 1,005 disagreements between the R_n and PAR data and there were 36 values when R_n and PAR data did not fit expected trends.

For the Wetland site, the 1-year period of analysis, from January 1 to December 31, 2018, included 17,520 thirty-minute values. There were no missing H values but there were 193 missing λE values and 2 missing R_n values; 7 G values, 397 H values, and 877 λE values were considered suspect (Law and others, 2005). There were 146 R_n values that were wind-speed corrected because the corresponding wind speeds exceeded 5 m/s. The suspect R_n and G values were censored and replaced with linearly interpolated values. Of the 397 suspect H values, 330 values fit expected daily and seasonal trends and were not censored, and 67 values were censored and replaced with linearly interpolated values. Of the 1,070 missing (193) and suspect (877) λE values, 756 values fit expected daily trends and were not censored, 179 values were censored and edited to zero, and 135 values were censored and replaced with linearly interpolated values. All 179 λE values that were censored and edited to zero were associated with rain and 165 of them occurred at nighttime or within 1 hour of nighttime. When defining daytime and nighttime, there were 505 disagreements between R_n data at the Wetland site and PAR data at the Shrub site and there were 17 values when R_n and PAR data did not fit expected daily trends.

Site-scale ET_g, which equals total ET minus precipitation, was calculated from the 30-minute available-energy, turbulent-flux, and precipitation data for each site. Total ET was calculated as the average of measured total ET and EBR-adjusted total ET (Moreo and Swancar, 2013). Measured total ET was calculated from measured 30-minute air-temperature and λE (eq. 9). An annual EBR was calculated from measured 30-minute available-energy and turbulent-flux data (eq. 10). Measured turbulent fluxes (H and λE) then were adjusted, while maintaining the Bowen ratio, to achieve energy balance or an annual EBR of one (Twine and others, 2000; Foken and others, 2012b; Garcia and others, 2014). The Bowen ratio is H divided by λE (Bowen, 1926; Sverdrup, 1943). The EBR-adjusted total ET was calculated from air temperature and EBR-adjusted λE (eq. 9). Evapotranspiration from soil moisture was not a consideration when calculating ET_g because the soil-moisture content at the beginnings and endings of the periods of analyses were similar (Allander and others, 2009; Garcia and others, 2014). Uncertainty in site-scale ET_g was calculated as one-half of the percentage of difference between measured and EBR-adjusted ET_g values (Moreo and Swancar, 2013).

The final step in site-scale data processing was calculating annual daytime mean CNF values for selected footprint sizes (Schuepp and others, 1990; Kormann and Meixner, 2001) and determining the footprint sizes that were used in the analyses. For this study, 30-minute CNF values were calculated using EdiRe software (Clement, 1999) for six zones centered on each site corresponding to radii of 25, 50, 75, 100, 200, and 300 m. Annual daytime mean CNF values represent an entire year of analysis but do not include nighttime data, when the surface-energy budget generally is unstable and site-scale ET fluctuates around zero.

Site-Scale Groundwater Discharge by Evapotranspiration

At the Shrub site, site-scale ET_g for the 1-year period of analysis was 0.121±0.005 m, or 3.9 percent (table 2). Site-scale ET_g was calculated by subtracting precipitation (0.068 m) from total ET (0.189 m). Precipitation is based on a wind-speed correction to the tipping-bucket rain gage total (0.048 m), which was corrected for the volumetric rain gage total (0.063 m). Total ET is the average of measured total ET (0.197 m; fig. 6) and EBR-adjusted total ET (0.182 m). The EBR for the Shrub site was 1.085. Based on the annual daytime mean CNF values and the surrounding plant cover, a 75-m radius footprint was selected for the Shrub site, which accounted for about 82 percent of the measured turbulent fluxes, and therefore, 82 percent of the measured ET (table 3). The footprint size was limited to a 75-m radius because the area from 75 to 200 m from the site had uneven plant cover and the area beyond 200 m from the site included a highway and vegetation that predominantly was xerophytes (fig. 4A).

At the Wetland site, site-scale ET_g for the 1-year period of analysis was 1.056±0.076 m, or 7.2 percent (table 2). Site-scale ET_g was calculated by subtracting precipitation (0.073 m) from total ET (1.129 m). Total ET is the average of measured total ET (1.053 m; fig. 6) and EBR-adjusted total ET (1.205 m). The EBR for the Wetland site was 0.873. Precipitation is based on wind-speed corrected volumetric measurements of 0.068 m. Based on the annual daytime mean CNF values and the surrounding plant cover, a 75-m radius footprint was selected for the Wetland site, which accounted for about 85 percent of the measured turbulent fluxes, and therefore, 85 percent of the measured ET (table 3). The footprint size of the Wetland site was limited to a 75-m radius because much of the area from 75 to 100 m from the site included sparse grasses and a large part of the area beyond 100 m from the site included a highway and vegetation that predominantly was xerophytes (fig. 4B).

Processing the Scaled Vegetation Index

A scaled NDVI (Rouse and others, 1974) that encompasses the AWSR study area was calculated from the red and near-infrared bands of 0.6-m resolution 4-band NAIP aerial imagery collected in 2020 (Damar and Pavelko, 2023). The NAIP imagery is collected every other year in most States, including California, and the imagery is well-suited for use in ET studies. In addition to visible spectrum bands of red, green, and blue, NAIP imagery includes a near-infrared spectral band, which is necessary to calculate common vegetation indexes. The NAIP imagery typically is collected during periods when vegetation is considered near the peak of seasonal growth, and the flight paths allow large areas to be collected on a single day (U.S. Department of Agriculture, 2017). A vegetation index derived from imagery collected in a single day, as opposed to over several weeks or months, minimizes variations in sun angle, atmospheric conditions, and ground conditions so that values across the scene are comparable. The scaled NDVI used in this study was derived from imagery collected almost entirely on July 2, 2020, with less than one-half of 1 percent of the study area, at the southern end of the Amargosa River GDA, collected on May 23, 2020. Ideally, the collection date also is at least 30 days after any rainfall in the study area and is contemporaneous with the ET measurements. Antecedent rainfall can cause an unseasonable growth spurt in vegetation that could misrepresent the vigor (health) of the plants for the time of year. Contemporaneous data collection ensures that the imagery is representative of measurement conditions. The 2020 NAIP imagery used in this study was collected with no antecedent rain for at least 30 days, but it was not collected contemporaneously with the ET measurements. Although NAIP imagery also was collected in 2018, the imagery covering the study area was collected over multiple days in late August of that year, when vegetation is not in peak vigor. It was assumed that undisturbed vegetation, such as in the AWSR study area, varied minimally during the 2 years after ET-data collection and, therefore the imagery collected in July 2020 is the better representation of peak growing season conditions in 2018.

The NDVI was selected as the vegetation index for this study because the radiometric correction applied to the NAIP imagery does not include conversion to percentage of reflectance, which is required for more complex vegetation indexes. The NDVI is a normalized ratio of reflectance in the near-infrared band to reflectance in the red band. These two bands can be used as indicators of the presence of vegetation because plant leaves reflect near-infrared energy and chlorophyll absorbs red energy (U.S. Department of Agriculture, 2017). Raw NDVI values range from −1 to 1, with values from 0 to 1 being positively correlated to amount of green biomass (Jensen, 2000). For this study, negative NDVI values were set to zero and all other NDVI values were scaled by 10,000 (Damar and Pavelko, 2023). These scaled values were used in the calculations of this study and are referred to as scaled NDVI values.

A footprint-weighted mean scaled NDVI value for each ET-monitoring site was calculated using site-scale CNF data and scaled NDVI values that corresponded to the selected footprint sizes (table 3). The size of a footprint was selected based on the annual daytime mean CNF values for six areas surrounding the site and the vegetation and other surfaces in those areas, where the largest footprint was selected that contained predominantly the targeted vegetation. Footprint-weighted mean scaled NDVI values were used to develop the quadratic equation because they represent the site footprint, whereas the scaled NDVI value for the site location only corresponds to a 0.6×0.6 m pixel. After the footprint size of a site was determined, it was assumed that all measured fluxes come from within the footprint area. A footprint-weighted mean scaled NDVI value then was determined by calculating the mean scaled NDVI value for each footprint zone around the site (0–25, 25–50, 50–75, 75–100, and 100–200 m; table 3; figs. 7A and 7B), weighting each mean scaled NDVI value for each zone with the corresponding annual daytime mean CNF value, and then adding the weighted means of each zone. The Shrub and the Wetland ET sites have a 75-m radius footprint, each comprised of three footprint zones, 0–25, 25–50, and 50–75 m, therefore the footprint-weighted mean scaled NDVI value is calculated as:

where For the Shrub site, the footprint-weighted mean scaled NDVI is 183, and for the Wetland site, the footprint-weighted mean scaled NDVI is 5,060 (table 3).

Assigning Rates and Uncertainty

Open-water and moist-soil ET units were assigned site-scale ET_g derived from previous studies and vegetated ET units were assigned site-scale ET_g estimated for this study. For open water in the Great Basin, site-scale ET_g ranges from about 1.49 meters per year (m/yr) at Walker Lake to about 2.62 m/yr at Ash Meadows (Laczniak and others, 1999; Berger and others, 2001; Westenburg and others, 2006; Allander and others, 2009; Moreo and Swancar, 2013; Earp and Moreo, 2021). For this study, open-water ET units were assigned a site-scale ET_g of 2.53 m/yr, which is the open-water rate estimated for the Tecopa/California Valley area by Laczniak and others (2001). For moist soil in the Great Basin, estimated site-scale ET_g ranges from about 0 m/yr for salt-encrusted playas of Death Valley (Jackson and others, 2018) to about 0.792 m/yr for moist bare soil in Ash Meadows (Laczniak and others, 1999). Moist-soil ET units were assigned a site-scale ET_g of 0.322 m/yr, which is the average of the site-scale ET_g for the moist bare soil ET unit in the Shoshone and Tecopa/California Valley areas (0.643 m/yr) from Laczniak and others (2001) and the site-scale ET_g for the salt-encrusted playas of Death Valley (0.00000002 m/yr) of Jackson and others (2018).

Vegetated ET units were assigned site-scale ET_g defined by a quadratic relation between site-scale ET_g, as the dependent variable, and scaled NDVI values, as the independent variable, similar to Allander and others (2009) and Beamer and others (2013). The quadratic relation (fig. 8) was developed using four points. The first two points are derived from the site-scale ET_g measurements (table 2) and corresponding footprint-weighted mean scaled NDVI values (table 3). The second two points are theoretical, one point representing a theoretical minimum site-scale ET_g of zero and a scaled NDVI value of zero and a second point representing a maximum site-scale ET_g of 1.222 m/yr and the maximum scaled NDVI value of 10,000. The estimated maximum site-scale ET_g for the AWSR study area was assumed to be 1.222 m/yr because that was the highest site-scale ET_g estimated for the DVRFS and corresponded to dense wetland vegetation in the Shoshone area (Laczniak and others, 2001). The relation was developed following the assumptions that an increase in scaled NDVI values must result in an increase in site-scale ET_g, that scaled NDVI values equal to zero were assigned an ET_g of zero, and that the relation would not extend beyond the maximum scaled NDVI value of 10,000 or the estimated maximum site-scale ET_g of 1.222 m/yr.

The range of annual ET_g for the study area was computed using the lowest and highest estimated site-scale ET_g. The lowest estimated annual ET_g for the study area was computed using an open-water site-scale ET_g of 1.494 m/yr (Allander and others, 2009) and a moist-soil site-scale ET_g of 0.00000002 m/yr (Jackson and others, 2018). The lowest estimated annual ET_g for vegetated areas was computed from a quadratic relation that, for each site, used only the smaller of measured and EBR-adjusted values (table 2) instead of their average. The highest estimated annual ET_g for the study area was computed using an open-water site-scale ET_g of 2.62 m/yr (Laczniak and others, 1999) and a moist-soil site-scale ET_g of 0.792 m/yr (Laczniak and others, 1999). The highest estimated annual ET_g for vegetated areas was computed from a quadratic relation that, for each site, used only the larger of measured and EBR-adjusted values (table 2) instead of their average.