Estimating groundwater evapotranspiration rates using diurnal water-table fluctuations in a semi-arid riparian zone Laura K. Lautz

Abstract In semi-arid climates, phreatophytes draw on shallow aquifers, and groundwater evapotranspiration (ETG) is a principal component of groundwater budgets. Diurnal water table fluctuations, which often are a product of ETG, were monitored in the riparian zone of Red Canyon Creek, Wyoming, USA. These fluctuations were higher in a riparian wetland (2–36 mm) than a grasscovered meadow (1–6 mm). The onset and cessation of water-table fluctuations correspond to daily temperatures relative to freezing. Spatial differences were due to vegetation type and specific yield, while temporal changes were due to vegetation dormancy. Ratios of ETG to potential evapotranspiration (PET), Kc,GW, were similar to ratios of actual evapotranspiration (ET) to PET, Kc, in semi-arid rangelands. Before vegetation senescence, Kc, GW increased between precipitation events, suggesting phreatophytes pull more water from the saturated zone as soil moisture decreases. In contrast, Kc decreases with soil moisture following precipitation events as ET becomes increasingly water-limited. Error in ETG is primarily from estimates of specific yield (Sy), which is difficult to quantify in heterogeneous sediments. ETG values may be more reliable because the range of acceptable Sy is smaller than Kc and Sy does not change with vegetation type or soil moisture. Résumé Sous les climats semi-arides, les plantes qui puisent dans les aquifères phréatiques et l’évapotranspiration (ETG) sont les composantes principales des bilans hydrogéologiques. Les fluctuations diurnes du niveau de la nappe, qui sont souvent le résultat de l’ ETG, ont été suivies dans la zone riparienne du Canyon de Red Creek, au Wyoming, USA. Ces fluctuations étaient plus imporReceived: 26 March 2007 / Accepted: 22 October 2007 Published online: 13 November 2007 © Springer-Verlag 2007 L. K. Lautz ()) Department of Forest and Natural Resources Management, SUNY College of Environmental Science and Forestry, 207 Marshall Hall, 1 Forestry Drive, Syracuse, NY 13210, USA e-mail: [email protected] Tel.: +1-315-4704765 Fax: +1-315-4706535 Hydrogeology Journal (2008) 16: 483–497

tantes dans une zone riparienne marécageuse (2 à 36 mm) que dans une prairie herbeuse (1 à 6 mm). Le début et la cessation des fluctuations de la nappe correspondent aux températures journalières par rapport au gel. Les différences spatiales étaient dues au type de végétation et à la porosité efficace, tandis que les changements dans le temps étaient dus à la dormance des plantes. Les taux de l’ETG par rapport à l’évapotranspiration potentielle (PET), Kc,GW, sont similaires aux taux de l’évapotranspiration réelle (ET) par rapport à la PET, Kc, dans les prairies semi-arides. Avant la sénescence végétale, Kc,GW augmente entre les évènements pluvieux, ce qui laisse supposer que les plantes puisent plus d’eau dans la zone saturée lorsque l’humidité du sol diminue. Au contraire, Kc décroît avec l’humidité du sol après les précipitations alors que l’ET devient de plus en plus limitée. L’erreur sur l’ ETG `est liée aux estimations de la porosité efficace ( Sy), qui est difficile à quantifier dans les sédiments hétérogènes. Les valeurs d’ETG pourraient être plus fiables car l’échelle des valeurs de Sy acceptables est plus restreinte que pour Kc et le Sy ne varie pas avec le type de végétation ou l’humidité du sol. Resumen En climas semiaridos, los freatofitos consumen agua de los acuiferos someros, y la evapotranspiración (ETG) es un componente principal en los balances hídricos. Las fluctuaciones diurnas del nivel de agua subterránea, que frecuentemente son un producto de ETG, fueron monitoreadas en la zona rivereña de Red Canyon Creek, Wyoming, EU. Estas fluctuaciones fueron mayores en un humedal rivereño (2–36 mm) que en una pradera cubierta por pasto (1–6 mm). La aparición y finalización de las fluctuaciones de agua subterránea corresponden con temperaturas diarias relacionadas con congelamiento. Las diferencias espaciales se debieron al tipo de vegetación y rendimiento específico, mientras que los cambios temporales se debieron a vegetación latente. Los radios de ETG con evapotranspiración potencial (PET), Kc,GW, fueron similares a los radios de la evapotranspiración real (ET) con PET, Kc, en tierras de pastoreo semi-aridas. Antes que la vegetación envejeciera, la razón Kc,GW, incrementó entre eventos de precipitación, sugiriendo que los freatofitos tomán más agua de la zona saturada en la medida que la humedad del suelo decrece. En contraste, Kc decrece con la humedad del suelo que sigue a los eventos de precipitación, al mismo tiempo que ET se vuelve crecientemente limitado DOI 10.1007/s10040-007-0239-0

484

al agua. El error en ETG proviene principalmente de las estimaciones de rendimiento específico ( Sy), el cual es difícil de cuantificar en sedimentos heterogéneos. Los valores de ETG podrían ser más confiables debido a que el rango aceptable de Sy es más pequeño que Kc y Sy no cambia con el tipo de vegetación o humedad del suelo. Keywords Diurnal water table . Evapotranspiration . Arid regions . Phreatophytes . USA

Introduction In arid and semi-arid areas of the Western United States, evapotranspiration by phreatophytes is a principal groundwater sink and a significant component of regional groundwater budgets (Nichols 1994). Although it is essential to quantify transpiration of water from the subsurface, and particularly from groundwater, by phreatophytes in waterpoor areas, few direct methods exist. Consumptive water use by plants from the saturated zone is different from evapotranspiration from near the land surface. Evapotranspiration from the land surface (ET) includes evaporation of water from surface soils and transpiration of water by plants and groundwater evapotranspiration (ETG) is a component of ET. ET is controlled by water availability, the amount of incident solar radiation, temperature, wind speed and humidity (Shuttleworth 1993) and is typically quantified using a combination of field measurements (i.e. evaporation pans or lysimeters), empirical or process-based models, the eddy covariance method, and/ or remote sensing. Empirical models can be temperaturebased or solar-radiation based (Priestley and Taylor 1972; Thornthwaite 1948). Process-based models estimate the radiant energy available for evaporation and the diffusion mechanism that removes water vapor from the site (Penman 1948). In contrast to evapotranspiration near the land surface, direct or indirect consumptive water use by plants from the saturated zone and capillary fringe is a flux across the water table. Because ETG occurs beneath the surface and is only a portion of ET, the factors controlling the flux of water from an aquifer may be somewhat distinct from those controlling the total flux from the land surface and rates of ET and ETG may differ. Others have observed that ETG can be between 9 and 126% of ET from an evaporation pan in semi-arid settings (Meyboom 1967). Although the factors controlling ETG may be somewhat distinct from those controlling ET, there are few tools available to estimate ETG directly. Evapotranspiration losses from the water table are often estimated by first quantifying near-surface ET using the methods described above (and reviewed in detail in Shuttleworth 1993); then ETG is estimated based on a simple linear decrease of ET as a function of the depth of the water table from the land surface (Banta 2000). ETG is assumed to become zero when the water table is below a fixed depth, termed an extinction depth. Recent work indicates an exponential decay function better describes the change in ETG with depth and that the extinction depth varies with vegetation Hydrogeology Journal (2008) 16: 483–497

cover and soil properties (Shah et al. 2007). For groundwater modeling applications, where ETG can be an important flux boundary at the water table, a direct measurement of groundwater evapotranspiration may be more appropriate than simplified models based on nearsurface ET. There are several practical advantages to using diurnal water-table fluctuations instead of simplified models of changes in ET as a function of depth of the water table. ETG measurements based on diurnal water-table fluctuations are advantageous because they are a direct measurement, are accurate over daily time steps and relatively simple to calculate. While temperature-based models of ET (i.e. Thornthwaite 1948) may also be easy to use, they are inaccurate over short time periods and are best used for monthly or seasonal ET values (Thompson 1999). On the other end of the spectrum, complex process-based models of ET (i.e. Penman 1948) are more accurate for shorter time steps, but require a number of input variables, assumptions and complex calculations, which can preclude their use in management applications. In contrast to both empirical and process-based ET models, estimating ETG using diurnal water-table fluctuations requires relatively inexpensive instrumentation, as compared to monitoring a myriad of meteorological parameters such as humidity, wind speed, temperature, and incident radiation. Computations of ETG are also straightforward and, therefore, may be more readily applied to a variety of research and management applications. Monitoring diurnal water-table fluctuations requires less intensive fieldwork than monitoring evaporation pans or lysimeters—the operator does not need to refill the pan or irrigate because groundwater replenishes the water supply naturally and data can be logged frequently for long periods of time using water level data loggers. ETG estimates from diurnal water-table fluctuations may be more applicable for remote locations with limited field access or for limited budgets that do not allow for meteorological stations with significant amounts of instrumentation. Also, water-table monitoring wells serve multiple purposes and are typically already installed onsite in hydrology studies, unlike privately owned or US National Weather Service (NWS)-maintained meteorological stations, which may be unavailable or located several kilometers from the site of interest. Water-table measurements can be a direct observation of ETG and reveal fine scale spatial and temporal patterns. Changes in ETG daily during the growing season reflect changes in transpiration during different stages of plant development and changes in moisture availability during dry periods. While meteorological observations are recorded daily, they do not directly capture changes in plant development or available water, so empirical or processbased estimates of ET do not directly reflect changes in plant transpiration due to life stage of the vegetation or antecedent moisture conditions. Also, spatial differences in ETG observations reflect variability between different plant communities. The direct observation of ETG may be more effective than using a model of near-surface ET, which DOI 10.1007/s10040-007-0239-0

485

requires an estimated crop coefficient to reflect both the plant community and the time during the growing season (Shuttleworth 1993). Crop coefficients are difficult to estimate because they vary temporally during the growing season and between different plant types. Crop coefficients are also used with the assumption that water is not a limiting variable controlling ET. Direct measurements of water-table fluctuations integrate a number of factors, including the time during the growth cycle of the plant, the plant type and moisture availability. Other methods of estimating ET such as remote sensing, capture spatial variability of vegetation and water stress, but fail to produce a continuous time series of ET. Eddy flux towers capture changes in vegetation, type, life stage and water stress, but do not capture small-scale spatial variability, although this can also be a limitation of using well networks to observe the water table. In addition, both remote sensing and eddy flux measurements still need to be adjusted to account for differences between ET and ETG. Although diurnal water-table fluctuations may be a better measure of ETG than meteorology-based values, there are few studies presented in the literature where diurnal water-table fluctuations are used and even fewer where results are validated by comparison with independent measurements of near-surface ET. In addition to the paper originally highlighting how to interpret diurnal water-table fluctuations (White 1932), others have focused on improving the technique by refining how specific yield is estimated (Gerla 1992), describing how other parameters can impact water-table fluctuations (Meyboom 1967) and modeling how flow system geometry, sediment texture, and water consumption in the unsaturated zone can interfere with the interpretation of diurnal water-table fluctuations (Loheide et al. 2005). There are few attempts to validate ETG measurements by making comparisons of ETG with ET near the land surface. Only recently have water-table fluctuations generated by ETG been linked to sap flow measurements, vegetation type and vitality (Butler et al. 2007; Engel et al. 2005). Linking meteorologically-based ET estimates with ETG is a useful exercise because many empirical and process based models of ET have been previously validated (Fisher et al. 2005; Winter et al. 1995) and spatial and temporal patterns of ETG should be comparable to ET, even if smaller in magnitude. In the one study where measurements of ETG from water-table fluctuations were compared with ET measurements at the land surface, Dolan et al. (1984) found that annually, empirical temperature-based ET values (Thornthwaite 1948) produced similar results to annual ETG values, but on a monthly basis ETG was between 0.61 and 2.5 times ET. Evaporation pan measurements were closer to ETG on a monthly interval, but ETG was higher than pan ET in the summer and lower than pan ET in the winter. There is still a need to validate ETG measurements using more robust models of surface ET. In this study, diurnal fluctuations of the water table were used to quantify groundwater evapotranspiration rates in a semi-arid riparian zone. Then groundwater evapotranspiration rates across the water table were Hydrogeology Journal (2008) 16: 483–497

compared to radiation-based estimates of ET for the land surface. The goals of this project were to: (1) use diurnal water-table fluctuations to estimate ETG for a semi-arid riparian area, (2) observe patterns of daily water-table fluctuations that are driven by groundwater recharge and evapotranspiration, both temporally and spatially, and (3) compare direct observations of ETG with meteorologybased models of potential evapotranspiration (PET).

Study site The Red Canyon Creek watershed is located in westcentral Wyoming, in the Wind River Range of the Rocky Mountains (Fig. 1). The watershed is semi-arid, receiving less than 35 cm of precipitation annually, with potential evapotranspiration rates up to 10 times the precipitation rate (Patten 1998). Precipitation is seasonal with over 40% received in the spring, between March and May. Winter and late summer are the driest seasons—during these six months (December through February and July through September) the site receives less than a third of the total annual precipitation, on average. The US Forest Service classifies the rangelands of Wyoming and Montana states as part of the “intermountain semi-desert ecoprovince” based on their semi-arid climate, cold winters, and hot/dry summers (Bailey 1995). Sagebrush (Artemisia spp.) and short grasses occur on the valley slopes, whereas willows (Salix spp.), rushes (Juncus spp.) and sedges (Carex spp.) occupy riparian zones near the creek, which is incised through fluvial deposits in the valley bottom. Baseflow to Red Canyon Creek is supplied by the slow release of groundwater mostly recharged during spring snowmelt to terraces and the flood plain (Patten 1998). The study site is located near the base of the watershed, in a meadow that has been described in detail in Lautz et al. (2006) and Lautz and Siegel (2006). At this site, the creek is incised about 1.5 m into the floodplain sediments, which consist of 6–7 m of unconsolidated alluvium. As a result, the water table is between approximately 1 and 2 m below the land surface in riparian meadows. Several soil borings in the area indicate the upper 2 m of alluvium consists of semi-moist fluvial silt and clay with thin interbedded sand lenses (Fig. 2). The silt caps a 1 m thick section of sand and gravel with a silt and sand matrix. This high permeability layer is underlain by fluvial silt and red shales from the Triassic Chugwater Formation. The somewhat heterogeneous floodplain sediments do not have uniform physical properties, making it difficult to estimate the material properties of the sediments at each well. More detailed discussion is found below.

Methods Evapotranspiration (ET) based on solar radiation Meteorological observations were made at a National Oceanic and Atmospheric Administration US Climate Reference Network (USCRN) station. The USCRN is a DOI 10.1007/s10040-007-0239-0

486 Fig. 1 Site maps showing a the location of Lander, Wyoming, USA, b the proximity of the Red Canyon Creek Watershed to Lander, c the location of the NOAA CRN meteorological station relative to the study site in the Red Canyon Creek Watershed, and d the locations of monitoring wells, water-table elevation contours, the direction of groundwater flow through the site and wetland areas

network of over 85 climate monitoring stations that are designed to provide highly accurate and reliable long-term meteorological observations that will be used to detect present and future climate change (NOAA-NCDC 2006). The USCRN station used in this study is located within the Red Canyon Creek watershed, approximately 700 m from the research site, on a glacial terrace about 50 m higher in elevation (Fig. 1). Data recorded at the station include average, maximum and minimum daily temperatures, daily precipitation, and daily total and maximum solar radiation. Potential evapotranspiration (PET) for a reference crop was estimated using the Priestley-Taylor equation, which models PET based on the solar radiation received at a given site (Priestley and Taylor 1972).   D Rn PET ¼
ð1Þ D þ + 1v where Δ is the gradient of the saturated vapor pressure (kPa °C−1), + is the psychrometric constant (kPa °C−1), 1v Hydrogeology Journal (2008) 16: 483–497

is the latent heat of vaporization (MJ/kg), Rn is the observed net radiation (MJ m−2) and α is a constant. Both + and 1v are a function of the daily average air temperature recorded in the field. For arid locations, α has been estimated as 1.74 (Shuttleworth 1993). Combination equations such as the Penman equation, are often preferred to purely radiation-based equations such as Priestley-Taylor, because they include two terms—a term to estimate the radiant energy available for evaporation and a term to describe the diffusion mechanism that removes water vapor from the site. Because the available energy term often exceeds the diffusion term by a factor of 4, the Priestley-Taylor equation can be used when wind speed and humidity data are not available, as in this study (Shuttleworth 1993). In previous work, the PriestleyTaylor method has been shown to compare closely to energy budget values and to results from the Penman equation (Winter et al. 1995). It is also a method commonly used by practitioners and is therefore appropriate for comparison in this study. DOI 10.1007/s10040-007-0239-0

487

compared to traditional Kc estimates from results at other similar sites. Kc;GW ¼

ETG PET

ð3Þ

Groundwater evapotranspiration (ETG) based on water-table fluctuations

Fig. 2 Sediment boring logs and well construction log for W1 and W2. The elevation of the casing (m above sea level) is given at the top of the sediment boring for each well. The height of the water table indicated on the log is the height at the beginning of the observations described here (13 July 2005). Well construction was identical at both sites and is indicated in the well construction log in the center of the figure. At W2, the sediment core was not recovered between 1.2 and 1.8 m below the land surface and was inferred based on the sediment texture bounding that interval and at nearby sites

The actual evapotranspiration rate (ET) is related to the crop-specific potential evapotranspiration (PET) through the crop coefficient (Kc). Kc reflects the percentage of PET realized, given the vegetation type and time during the growing season. Kc ¼

ET PET

ð2Þ

Values for Kc are derived by taking the ratio of actual ET, which can be measured at the land surface using a lysimeter, and PET, which is calculated using mathematical models such as Priestley-Taylor. Many crop coefficients are given in the literature for irrigated field crops, where water availability does not limit either transpiration or soil water evaporation (Shuttleworth 1993). Because the field site is not irrigated, general Kc values may be too high for this site where water availability often limits ET. In rangelands dominated by tall grasses and/or sedges and shrub willows such as this, soil water evaporation is often water-limited and, therefore, Kc for each crop is difficult to estimate. Due to the inherent difficulty of estimating an accurate, time-variable Kc for this site, ETG (see the following) was compared directly to PET. A groundwater specific coefficient (Kc,GW) was Hydrogeology Journal (2008) 16: 483–497

To observe water-table fluctuations at the study site, two monitoring wells were installed and instrumented with automated water level recorders between July and October of 2005 and between May and June of 2006. The wells are located within a few meters of the stream channel, one located adjacent to a small man-made log dam and associated wetland (W1) and one located adjacent to a riffle sequence downstream of the dam (W2) (Fig. 1). The borings for the wells were installed using a Geoprobe (Kejr, Inc., Salina, Kansas), which uses both static force and percussion to advance the sampling tools into the subsurface. Sediment cores were collected and characterized during well installation to a depth of just over 3 m (Fig. 2). The wells were constructed based on ASTM (American Standard for Testing and Materials) standards for 2-inch (5.1 cm) wells. They were constructed by installing a 5 cm diameter PVC casing attached to a well screen in the boring, pouring a sand pack to 30 cm above the screen, followed by a 30-cm bentonite plug, backfilling the hole with drill cuttings and finishing with a concrete cap. The wells were screened at an interval from 1.5 to 3.0 m below the land surface, which spans the range of observed watertable levels at these locations. Water levels in the wells were recorded every 20 min using TruTrack WT-HR Water Height Data Loggers, which have a resolution of ±1 mm. The groundwater evapotranspiration rate (ETG) was estimated based on the diurnal water-table fluctuations observed at the monitoring wells (White 1932). Watertable elevations at the study site fluctuated daily due to the consumption of water by phreatophytic vegetation, as has been observed elsewhere (Gerla 1992; Loheide et al. 2005; Rosenberry and Winter 1997; White 1932). Others have observed that water-table elevations are generally at a minimum between 18:00 and 19:00, following a period of drawdown during the daytime due to consumptive water use by vegetation. Water-table elevations are generally at a maximum between 07:00 and 09:00, following a period of recharge by a net influx of groundwater during night, when transpiration is minimal. The magnitude of the diurnal fluctuation can be used to estimate the daily consumption of groundwater due to evapotranspiration. White (1932) developed a method to quantify daily groundwater evapotranspiration (ETG, m) using the following equation:
ETG ¼ Sy 24rgw
s

ð4Þ

where Sy is the specific yield of the aquifer sediments (unitless), rgw is the rate of water-table rise between 00:00 DOI 10.1007/s10040-007-0239-0

488

and 04:00 (m hr−1) and s is the net rise or fall of the water table during the 24-h period (m). rgw was quantified at each site daily as the slope of a best-fit line, using linear regression of water height and time, for all the water heights recorded between 00:00 and 04:00, with the assumption that there are minimal fluxes, other than groundwater flow (i.e. evapotranspiration), in or out of the given site during that time. Assuming that the groundwater influx rate is constant throughout the day, the water-table response to evapotranspiration was quantified by taking the difference between the actual water-table elevation at the end of the day and what the water-table elevation was expected to be if the only flux of water in or out of the site were due to groundwater flow (Fig. 3a). The four key assumptions of the White method are: (1) evapotranspiration by plants causes the observed diurnal water-table fluctuations, (2) evapotranspiration is minimal, relative to the influx of groundwater, between 00:00 and 04:00, (3) the rate of groundwater influx to the site is constant throughout the day, and (4) specific yield accurately reflects the volume of water extracted from the saturated zone, per drop in the water table, per unit area of the site (Loheide et al. 2005). This method and the associated assumptions are described in detail in White (1932) and Loheide et al. (2005). There are other causes of water-table fluctuations in addition to ETG, including changes in barometric pressure and temperature. Barometric pressure changes are rarely significant, relative to the effect of groundwater use by phreatophytes, particularly in shallow systems (Meyboom 1967). The water temperature in the wells was recorded by the TruTrack data loggers, described above. The temperatures did not fluctuate on a daily time scale and changed

<3°C over the entire period of observation (9.5–12.1°C at W1 and 9.1–11.8°C at W2). Given the very consistent timing and magnitude of the observed water-table fluctuations, they are consistent with water use by phreatophytic plants and not with changes in barometric pressure or temperature. The assumption that groundwater influx is constant throughout the day can be violated if there are other sources of water to the site such as recharge due to precipitation, pumping of nearby wells or large changes in stream stage. At the Red Canyon Creek site, there are no nearby pumping wells and changes in stream stage are small and not associated with changes in groundwater levels. This can be seen by comparing changes in stream and groundwater heights over the course of several days during periods of low ET (October 2005) and high ET (June 2006; Fig. 4). Stream stage is monitored hourly at a gauging station located approximately 200 m downstream of the wells. In October of 2005, stream stage in Red Canyon Creek varied by as much as 3 cm daily in response to water withdrawals in the upper portion of the watershed. There were no associated changes in the watertable height at W1, which is located closest to the stream channel (Fig. 1). In June of 2006, the water table fluctuated by as much as 2.5 cm daily at W1, and fluctuations in stream stage were similar, but lower in amplitude with a lag in the timing of the maximum and minimum water heights. The lower amplitude and lag time show the stream is responding to changes in the water table rather than vice-versa, as has been observed elsewhere (Chen 2007). There is no evidence that changes in the water-table height are driven by changes in stream stage on a daily time interval.

Fig. 3 Examples of diurnal water-table fluctuations observed at well W1, with explanations of the variables used in the White (1932) method. a Water-table fluctuation over 24 h, with no recharge from precipitation. b Water-table fluctuation over 24 h, with recharge from precipitation. In b, during the first 12 h, only groundwater evapotranspiration (ETG) and groundwater recharge (rgw) are influencing the water-table position. After 12 h, recharge by precipitation causes an additional increase in the height of the water table. Sy is the readily available specific yield and s is the net change in the water-table elevation during the 24 h Hydrogeology Journal (2008) 16: 483–497

DOI 10.1007/s10040-007-0239-0

489

water drains instantly from the saturated zone as the water table drops. However, one expects water will slowly drain due to the force of gravity—the release of water is not instantaneous. In fact, the release of water described by traditional estimates of specific yield can occur over years and if shorter times between water level rises are considered, smaller values of specific yield should be used (Healy and Cook 2002). To account for this, a “readily available specific yield,” which reflects the amount of water that is released from the saturated zone, per unit drop in the water table per unit land surface area, in the time frame of the diurnal fluctuations, should be used (Loheide et al. 2005; Meyboom 1967). In this study, values for specific yield were selected based on the guidelines presented in Loheide et al. (2005), which provide details on selecting a value for readily available specific yield, given the sediment texture and depth to the water table. The guidelines presented by Loheide et al. (2005) provide a general framework for estimating specific yield and that framework is based on numerical simulations of a homogeneous unsaturated zone, rather than field data. Given that the water table is more than 1 m below the land surface and the substrate at that depth consists of semi-moist, fluvial clayey silt with interbedded sand lenses, a readily available specific yield of 7% was selected (Fig. 10 in Loheide et al. 2005).

Fig. 4 Examples of simultaneous stream-stage observations in Red Canyon Creek and groundwater elevation measurements at W1 from a 20–27 October 2005, and b from 5–12 June 2006

The water table at the study site did show a measurable response to recharge by precipitation during events when more than 1 mm was received in a day (Fig. 5). On these days, the recharge due to precipitation can be so large that the water-table elevation at the end of the day at times exceeds the predicted water-table elevation given only groundwater influx and no evapotranspiration (Fig. 3b). The result can be a negative evapotranspiration rate using the White method. For this reason, the White method was not applied on days where the recorded precipitation was greater than 1 mm. Of the 88 days between 13 July and 8 October, only 15 days (17%) were excluded. Of the 34 days between 18 May and 20 June, only 10 days (29%) were excluded. A key source of error when using the White method results from the value used for specific yield. Specific yield, the storage term for unconfined aquifers, is the volume of water that will drain from a site per unit drop in water-table elevation per land surface area of the site (Freeze and Cherry 1979). This term is necessary in the White equation, because it links the total change in watertable elevation due to evapotranspiration to the total volume of water consumed by evapotranspiration. Specific yield is a storage term that does not have a time component. If values for traditional specific yield were used in the White equation, one would assume that all the Hydrogeology Journal (2008) 16: 483–497

Results and discussion Water-table response to environmental variables

Distinct diurnal water-table fluctuations were observed at both well sites (Fig. 5). The timing of the daily maximum and minimum water-table levels was consistent with other studies that have attributed such fluctuations to water use by plants (Butler et al. 2007; Loheide et al. 2005; Meyboom 1967; White 1932). Butler et al. (2007) found a tight coupling between daily variations in plant water use, from sap flow data, and the water-table response, with only 1–2-h lag times between the start of water uptake and corresponding changes in the water-table elevation. Watertable elevations in W1 and W2 at Red Canyon Creek were at a minimum around 18:00 or 19:00 each day, due to plant transpiration during the day. Water-table elevations were at a maximum around 08:00 or 09:00, following recharge by groundwater at night, when transpiration was minimal. At W1, from July until early September of 2005, the magnitude of the fluctuations averaged between 20 and 30 mm, with a maximum fluctuation of 32 mm on 21– 22 July. By the second week in September, the magnitude was generally less than 10 mm and diminished to between 1 and 3 mm by early October. At the onset of diurnal fluctuations in late May of 2006, the fluctuations at W1 increased rapidly from about 4 mm a day initially to about 30 mm a day by mid-June. At W2, the fluctuations were consistently smaller. In mid-July 2005, the magnitude of the fluctuations averaged about 5 mm (Fig. 5a). This magnitude continued to decrease over time to a magnitude of about 1 mm in late August 2005. DOI 10.1007/s10040-007-0239-0

490

Fig. 5 Observed water-table fluctuations at W1 and W2 and precipitation recorded at the USCRN meteorological station in 2005–2006. a–c Shows detailed observations at different time periods during the study. d Shows data for the full periods of monitoring, along with precipitation. The land surface is at an elevation of 1695.6 m asl at both sites

Daily water-table fluctuations were detected until early October 2005 at W1 and early September 2005 at W2 (Fig. 5b and c). Although water-table fluctuations were, at times, small (1–2 mm) at both sites, they were consistently observed and the timing of the maximum and minimum water-table elevations was consistent with earlier observations, indicating they were still generated by evapotranspiration. Seasonal trends in diurnal water-table fluctuations have been observed elsewhere, with higher fluctuations in spring and summer and lower to no fluctuations in winter (Butler et al. 2007; Dolan et al. 1984). Although the transition to dormancy and the associated decline in watertable fluctuations can be gradual, sudden drops in air temperature can also kill off many annual plants and cause an abrupt termination of water-table fluctuations (Butler et al. 2007). The seasonal decline in the water-table fluctuations at the studied sites in the late summer and fall is likely due to changes in the transpiration rate as vegetation either dies off, in the case of annual grasses, or becomes dormant, in the case of perennial wetland vegetation. The average growing season for sagebrush grassland sites in the region typically extends from mid-May to early October (Wight and Hanson 1990). The sudden decline in water-table fluctuations at W1 on 13 September 2005 can be linked to a sudden drop in air temperature that may have triggered riparian wetland plants into dormancy (Fig. 6). A similar pattern was observed in May 2006, as the onset of Hydrogeology Journal (2008) 16: 483–497

transpiration around 18 May initiated diurnal fluctuations at W1. The timing of this onset coincided with the time at which daily minimum temperatures no longer fell below zero. Temporal changes in the relative magnitude of the diurnal fluctuations at W1 and W2 may reflect differences in the growing season for the contrasting vegetation between the sites. The smaller water-table fluctuations at W2, relative to W1, may result from differences between the hydrogeologic settings at the two wells, differences in vegetative cover or differences in the material properties at the depth of the water table. W1 is located in a riparian wetland, adjacent to a log dam that pools stream water, causing groundwater recharge from the stream (Fig. 1). The elevated water level in the stream generates a locally elevated water table in the riparian wetland, which makes water more readily available for phreatophytes. In contrast, at W2 groundwater inflow comes from the meadow and the water table is deeper. The water table at W1 is about 1.8 m below the land surface and the water table at W2 is just over 2.0 m below the land surface–a difference of about 20 cm. Others have also observed that diurnal water-table fluctuations or evapotranspiration rates from groundwater decrease as the depth to the water table increases, but these observations were generally for larger differences in water-table height (Cooper et al. 2006; Rosenberry and Winter 1997; Zhang and Shilling 2006). Cooper et al. (2006) found that a 1.6 m drop in the water DOI 10.1007/s10040-007-0239-0

491

Fig. 6 Water-table elevations at W1 and minimum daily air temperatures recorded at the USCRN meteorological station from a 18 August to 8 October 2005, and b from 1 May to 22 June 2006

table resulted in a 62% reduction in ETG in the San Luis Valley in southern Colorado, USA, but this was due to large-scale groundwater pumping that lowered the water table below natural conditions. Between two wetlands in the Cottonwood Lake area, North Dakota, USA, Rosenberry and Winter (1997) found that as the depth to the water table increased by 0.6–0.8 m, diurnal head fluctuations decreased, indicating a decline in groundwater evapotranspiration over time. Although Rosenberry and Winter (1997) observed that diurnal fluctuations at individual wells declined as the water-table depth increased, the overall magnitude of water-table fluctuations at different wells did not correspond to water-table depth differences between those sites. This suggests that other factors such as spatial variability in transpiration rates or specific yield cause spatial differences in water-table fluctuations at the Red Canyon Creek site. The Nature Conservancy of Wyoming installs log dams in Red Canyon Creek to re-saturate riparian areas and support more wetland vegetation. Phreatophytes are most common in areas where the water table is closer to the land surface and, as a result, W1 is surrounded by sedges and rushes, in contrast to W2, which is surrounded by prairie grasses. Although the water table is higher at W1, the depth of the water table below the land surface is still Hydrogeology Journal (2008) 16: 483–497

over 1.5 m below the land surface and it is not likely that direct evaporation occurs from saturated zone. Sedges and rushes are expected to have higher transpiration rates due to their higher leaf area and would therefore generate larger diurnal water-table fluctuations. Others have observed spatial variability in ETG associated with spatial changes in vegetative cover (Rosenberry and Winter 1997; Zhang and Shilling 2006). Butler et al. (2007) observed very different diurnal fluctuations in five wells located along a transect adjacent to the Arkansas River in Kansas, USA. Diurnal water-table fluctuations were much larger in magnitude in the center of the riparian zone, relative to the edge of the riparian zone, and imperceptible in a pasture outside of the riparian zone. Although W1 and W2 are much closer in proximity than the wells in the Butler et al. (2007) study (25 m apart, as compared to between 75 and 300 m apart, respectively), the wells’ location relative to the riparian wetland and the variability in the water-table fluctuations are similar. Also, the Arkansas River is much larger in size and has a more expansive riparian zone than at Red Canyon Creek, so the relative locations of the wells are similar between the sites. Although both W1 and W2 are installed in similar flood plain materials, differences in the material properties at the depth of the water table may contribute to differences in the observed water-table fluctuations. The water table at W1 is located in fluvial clayey silt; whereas at W2, the water table is around the depth of a thin sand lens interbedded in the fluvial silt. Sand has a higher specific yield and therefore water-table fluctuations due to ETG are muted in sandy sediments. In coarser sediments, small water-table fluctuations release a greater volume of water than in finer sediments, and therefore, equal evapotranspiration rates from the water table will generate smaller water-table fluctuations in coarser sediments. The potential difference in specific yield between the two sites is explored in a later section.

Temporal patterns of groundwater evapotranspiration The daily groundwater evapotranspiration rates (ETG) at W1 and W2, calculated based on the diurnal water-table fluctuations, show that the ETG generally decreases, as expected, between July and October (Fig. 7). The decline in ETG at both sites during this time reflects the decrease in the transpiration rate as vegetation matures and then either dies or transitions to dormancy between July and October. At W1, ETG is about 3 mm/day in mid-June and July, when vegetation is in the middle of the growing season, decreasing to less than 0.5 mm/day in early October, after vegetation has browned and foliage has largely died back. The peak ETG rate at W1 was 4.6 mm/day, observed on June 15th. ETG rates observed at W2 are much smaller, corresponding to the smaller observed water-table fluctuations. At W2, ETG is about 0.5 mm/day in mid-July and decreases to essentially 0 mm/day in early October. The peak ETG rate at W2 was 0.8 mm/day on 18 July. The timing and magnitude of peak ETG are consistent with DOI 10.1007/s10040-007-0239-0

492 Fig. 7 Daily potential evapotranspiration (PET) and groundwater evapotranspiration (ETG) at W1 and W2 over the time period of this study (July– October 2005 and May–June 2006)

peak ET rates reported for mountain sagebrush sites in southwestern Idaho, USA, which were approximately 5 mm/day and occurred around 30 June 1990 (Flerchinger et al. 1996). The daily ETG rates reported here are also consistent with ET rates measured using lysimeters in mid to late June for shortgrass prairie (2.5–5.5 mm/day; Gillette, Wyoming) and sagebrush grass (1.1–5.8 mm/day; Reynolds, Idaho; Wight and Hanson 1990). ETG at the Red Canyon Creek site is also consistent with results from other studies that also used the White (1932) method (Table 1). Radiation-based potential evapotranspiration near the surface (PET) generally increases slightly between March and June and then decreases between July and October, as expected based on the changing photoperiod as the season progresses (Fig. 7). Daily fluctuations in PET are the result of daily variability in solar radiation and temperature, primarily due to changes in cloud cover. PET is about 15 mm/day in June and mid-July, when the photoperiod is the longest, and decreases to less than 10 mm/day in late September and early October, as the number of daylight hours decreases. The peak PET rate was 17.3 mm/day on 13 June 2005.

ETG should be smaller than near-surface PET on both seasonal and daily time scales, because PET reflects the potential for evapotranspiration, regardless of water availability, and can therefore exceed the available water amount (i.e. precipitation) by a factor of ten in semi-arid regions (Patten 1998). As expected, the magnitude of PET was higher than the magnitude of ETG at W1 and W2 on both daily and seasonal time scales and PET exceeded the precipitation received (Fig. 7; Table 1). The total ETG at W1 over the entire period of observation was only about 15% of the total PET and comparable to the total precipitation recorded. The total ETG at W2 over the entire period of observation was only about 1% of the total PET (Table 2). More details regarding comparison of PET and ETG can be found in the next section. Daily patterns of ETG are largely controlled by the meteorological variables that control PET such as net incoming solar radiation and temperature (Butler et al. 2007). As a result, higher daily ETG rates at both sites correspond to higher daily PET and there is a positive correlation between ETG and PET for both sites, particularly during June, July and August, when ETG rates are the highest (Fig. 8). Although there is a linear relationship

Table 1 ETG rates based on diurnal water-table fluctuations as reported in other studies, including the site description, time of year and depth of water table associated with the reported rates Site

Period

Depth of water table (m)

ETG (mm/day)

Reed Canary Grass riparian zone of Walnut Creek (Iowa, USA; Schilling 2007; Schilling and Kiniry 2007)

12–21 July 2004 2 August–22 July 2004 5–27 August 2004 31 August–30 September 2004 1–21 October 2004 May 1977 and 1978 June 1977 July 1977 August 1977 September 1977 October 1977 1–25 July 1990 1–10 August 1990 26–28 August 2002

0.0–0.7 0.2–0.8 0.0–1.0 0.0–1.2 1.2–1.4 ~0.2

3.1 2.1 1.9 1.8 0.4 3.2–8.4 3.0–7.6 3.9–9.5 1.9–7.8 2.1–10.0 1.4–5.6 1–8 1–3 4.2

Marsh/Freshwater Wetland of Palatlakaha River (Florida, USA; Dolan et al. 1984)

Between Prairie Pothole Wetlands (North Dakota, USA; Rosenberry and Winter 1997) Cottonwood, Mulberry, Willow riparian zone of Arkansas River (Kansas, USA; Butler et al. 2007) Hydrogeology Journal (2008) 16: 483–497

<1.2 1.2–2.0 ~1.9

DOI 10.1007/s10040-007-0239-0

493 Table 2 Biweekly averages (Avg.) and standard deviations (St. Dev.) for potential evapotranspiration (PET) and groundwater evapotranspiration (ETG) at W1 and W2

n

Avg.

St. Dev.

ETG,W1 (mm/day) Sy=7% Avg. St. ETG,W1/PET (Kc,GW) Dev.

12 12 12 12 12

13.8 14.0 15.0 12.1 10.1

2.0 2.5 1.6 2.1 3.5

1.5 2.9 3.1 2.3 1.4

0.6 1.0 0.5 0.7 0.9

0.11 0.21 0.20 0.19 0.13

0.5 0.2 0.1

0.1 0.1 0.2

0.04 0.02 0.01

11.3 7.8 6.4 5.4 40.2

14

11.9

1.1

2.1

0.3

0.18

0.1

0.1

0.01

2.0

12

9.4

1.2

0.4

0.4

0.05

0.0

0.1

0.00

5.2

11

7.7

1.7

0.1

0.2

0.01

0.0

0.1

0.01

72.4

97

11.9 1,144.1

3.0

1.6 168.4

1.2

0.14 0.15

0.2 11.4

0.2

0.01 0.01

150.7

Biweekly dates

18 May–6 June 2006 7 June–20 June 2006 13 July–26 July 2005 27 July–9 August 2005 10 August– 23 August 2005 24 August– 6 September 2005 7 September– 20 September 2005 21 September– 8 October 2005 Average Total (mm)

PET (mm/day)

ETG,W2 (mm/day) Sy=7% Avg. St. ETG,W1/PET (Kc,GW) Dev.

Precipitation (mm)

Biweekly average values for Kc,GW are calculated as the ratio of biweekly average ETG to biweekly average PET for W1 and W2. Total PET, ETG at W1 and W2, and precipitation for the entire study are also given

between PET and ETG at both W1 and W2, there is substantial scatter around the trendline. This scatter is likely the result of changes in other environmental factors besides solar radiation and temperature such as available soil moisture, which can impact evapotranspiration and are not included in radiation-based PET. In semi-arid ecosystems, the maximum ET rate during the growing season occurs immediately following rain events and then rapidly decreases within days if no additional rainfall occurs (Kurc and Small 2004). Others have found a strong, exponential relationship between actual evapotranspiration rates and declines in soil moisture over time (Teuling et al. 2006). The radiation-based PET inherently

assumes that soil moisture availability does not vary–no soil moisture variable is considered in the Priestley-Taylor method. This assumption is not valid for the Red Canyon Creek site, which is not irrigated. Because the diurnal water-table evapotranspiration rate (ETG) reflects an actual measurement of the drawdown resulting from phreatophytic water use, it inherently accounts for changes in ET that are caused by changes in water sources. In semi-arid settings, including grasslands, shrublands and sagebrush areas, decreases in measured surface ET correspond to decreases in surface soil moisture (Kurc and Small 2004; Prater and DeLucia 2006). As soil moisture declines during dry periods between precipitation events

Fig. 8 Daily radiation-based potential evapotranspiration (PET) versus groundwater evapotranspiration (ETG) at W1 and W2. Linear regression lines are fit to all of the W1 or W2 data and forced through the origin. Coefficients in the regression equations are significantly different than zero with a p-value <0.001

Hydrogeology Journal (2008) 16: 483–497

DOI 10.1007/s10040-007-0239-0

494

Fig. 9 The ratio of ETG to PET (Kc,GW) over time between precipitation events (where the precipitation exceeded 1 mm/day) during July and early August 2005. Daily precipitation amounts are also shown

and ET becomes water-limited, plants may draw more water from the water table, causing a relative increase in ETG. In contrast, following precipitation events, when soil moisture content is high, plants may reduce their dependency on phreatic water because soil water is so readily available. In contrast to surface ET, fluctuations of the water table and estimates of ETG based on the White method primarily reflect plant water consumption from the saturated zone, so there may be an inverse relationship between soil moisture and ETG between precipitation events. Rosenberry and Winter (1997) hypothesized that upland plants became “opportunistic,” drawing water from the water table when it is within reach of their roots. Butler et al. (2007) presented a similar hypothesis, supported by their observations of larger water-table fluctuations when the soil moisture content was at the wilting point, in contrast to smaller water-table fluctuations when the soil moisture content was near the field capacity. The ratio of ETG to PET can be used to identify changes in the percentage of PET that is derived from the water table between precipitation events. At Red Canyon Creek, the ratio of ETG to PET increases following precipitation events during July and early August at W1, when PET is highest (Fig. 9). This pattern supports the hypothesis that ETG increases as soil water availability in the unsaturated zone decreases and phreatophytes draw more from the groundwater during these dry periods. In July and early August, when demand for water is highest, an increasingly larger percentage of total near-surface ET is derived from the water table. This pattern does not persist in the later half of the study (late August to September), as overall transpiration rates decrease due to vegetation senescence.

differences in estimated ETG at W1 and W2. At W1, the sediment texture in proximity to the water table is uniform clayey silt and the readily available specific yield of 7% is most reliable at this location. At W2, the sediment texture is more heterogeneous in proximity to the water table; notably, there is a thin sand lens that likely has a higher readily available specific yield (15–30%, Loheide et al. 2005). Although one can estimate the potential differences in specific yield based on the guidelines presented by Loheide et al. (2005), the guidelines presented there are based on numerical simulations of homogeneous sediments and are less reliable in heterogeneous sediments such as those found at W2. If a higher specific yield is used for W2, to correspond to the thin sand lens located at the approximate depth of the water table (i.e. 30%, Loheide et al. 2005), the ETG estimates are larger. Given the relationship between ETG and specific yield shown in Eq. 4, increases in specific yield result in linear increases in ETG. ETG rates for W2 were recalculated using the higher specific yield of 30% and totaled 48.9 mm between 13 July and 8 October, still only about 45% of the ETG at W1 for the same period (Fig. 10). Due to the uncertainty of both the W1 and W2 specific yield values, it is difficult to assess whether the differences between the two wells are solely due to differences in material properties. For example, if the specific yield at W1 were only 3.5%, thereby cutting ETG at W1 in half, the ETG rates at W1 and W2 would be comparable. Although uncertainty of specific yield limits the conclusions one can draw from these changes, if material properties accounted solely for the differences in W1 and W2, one would expect ETG rates from the two sites to follow a linear trend. From Fig. 10, significant scatter can be seen around the regression line. Therefore, it is unlikely that specific yield is the only source of the difference between the two wells. Other variables such as the vegetation type and water sources during periods of drought, likely account for some of the differences in ETG between W1 and W2.

Spatial patterns of groundwater evapotranspiration

The overall smaller diurnal water-table fluctuations and corresponding lower ETG at W2 are likely due to the differences in vegetation from W1, as described earlier, or differences in the material properties of the sediment from W1. Different material properties may contribute to the Hydrogeology Journal (2008) 16: 483–497

Fig. 10 Evapotranspiration from the water table (ETG) at W1 versus W2, given a higher readily available specific yield (Sy) of 30% at W2. W1 has Sy=7% and W2 has a Sy=30%. The linear regression is forced through the origin and the coefficient is significantly different than zero, with a p-value <0.001 DOI 10.1007/s10040-007-0239-0

495

Linkages between potential evapotranspiration and groundwater evapotranspiration In many applications, PET is converted to the actual evapotranspiration rate (ET) using the crop coefficient (Kc). Kc is the ratio ET/PET and indicates the percent of PET actually occurring, given limitations caused by water availability, vegetation type and time during the growing season. To compare ETG and PET, a groundwater specific coefficient, Kc,GW, was calculated as the ratio of ETG to PET. Kc,GW indicates the percent of PET that is from the water table and is potentially comparable to field estimates of Kc from similar sites, which have been computed from actual ET (measured in lysimeters or via process-based equations) and radiation-based PET calculated based on meteorological variables. Comparisons between Kc and Kc,GW should be made with caution, considering the uncertainty of specific yield (see the preceding). The Kc,GW values computed in this study range from 0 to 39% at W1 and 0 to 5% at W2. Wight and Hanson (1990) calculated Kc values between about <10 to 60% in western prairie and sagebrush grass rangelands between 13 July and 1 September (when the vegetation had reportedly senesced), with the largest values associated with precipitation events when water availability was highest. Daily Kc values (i.e. ET/PET) in semi-arid grasslands and shrublands in New Mexico, USA were also highest during precipitation events and generally below 50% on dry days between June and September (Kurc and Small 2004). Values of Kc,GW established as part of this study are consistent with these previous observations of Kc over time (Table 2). Although direct comparisons of Kc and Kc,GW are difficult to interpret due to uncertainty of specific yield, the variability of Kc and Kc,GW over time reveal differences between temporal patterns of ET and ETG that are not discernable without normalizing to PET. Due to the linear scaling of ETG with specific yield (see the preceding), temporal patterns shown here would be present regardless of the absolute magnitude of Kc,GW. Others have observed large daily variability in Kc at similar sites, which they attributed to variations in environmental conditions on a daily bases such as changes in soil moisture availability following precipitation events (Kurc and Small 2004; Wight and Hanson 1990). In particular, they noted that Kc was highest following precipitation events, when soil water was not limiting ET, and decreased daily between precipitation events. The opposite relationship was observed at W1 during July and early August, when ETG rates were highest. Immediately following precipitation events (>1 mm/day), Kc,GW was between 5 and 15% and increased to between 27 and 30% over the several days following the precipitation event (Fig. 9). Kc,GW was lowest immediately after precipitation, when soil moisture was more readily available to plants, and increased during dry periods as soil moisture decreased and vegetation relied more heavily on phreatic water. On a biweekly basis at W1, the average Kc,GW fell to 13% between 10–23 August, which corresponds to the biweekly period receiving the second highest precipitation Hydrogeology Journal (2008) 16: 483–497

rate during this study (40.2 mm). The 2 weeks before and after show higher average Kc,GW (19 and 18%, respectively), and much lower precipitation received during those periods (5.4 and 2.0 mm, respectively). These patterns are not observed at W2, where the ETG rates are much lower and it is more difficult to discern subtle changes in ETG and Kc,GW over time. For measurements of ETG based on diurnal water-table fluctuations, the greatest error comes from the readily available specific yield value (Sy), as discussed earlier. For traditional measurements of ET based on solar radiation (i.e. Kc × PET), error comes from the poorly-defined crop coefficient value (Kc), which is a complex variable that is dependent on meteorology, vegetation type and its stage during development (Shuttleworth 1993). A sensitivity analysis of Sy and Kc for the two methods was used to qualitatively compare the uncertainty in ETG and traditional ET estimates for July through October, when observations were made at both W1 and W2. Based on guidelines presented by Loheide et al. (2005), the readily available specific yield for a sandy or clayey silt can range from about 4% to about 10%, depending on the grain size distribution, so ETG was calculated for this range of Sy at W1 and W2 (Fig. 11). Kc is highly variable, depending on the time during the growing season, vegetation type and

Fig. 11 Sensitivity analysis of evapotranspiration from the land surface (ET) and from the water table (ETG) at W1 and W2. The Sy values indicated for W1 are used for both W1 and W2. Dashed lines show the original values used in this study DOI 10.1007/s10040-007-0239-0

496

available soil moisture. Based on the results presented in Wight and Hanson (1990) and Kurc and Small (2004), Kc in prairie and sagebrush grass rangelands during dry periods in July and August can range from about 0.1 to about 0.3, so ET was calculated for this range of Kc (Fig. 11). For the sensitivity analysis, Kc was assumed to be constant throughout the study period. At W1, the sensitivity of ETG, given the range of reasonable Sy values, is comparable to the sensitivity of ET at the surface, given the range of reasonable Kc values. This suggests that the error derived from selecting Kc versus Sy is comparable. Although the apparent errors of ET and ETG are comparable, Sy is easier to estimate than Kc because it depends only on the material properties of the subsurface. Also, the range of acceptable Sy values is generally smaller. For all sediment types, Loheide et al. (2005) indicate a range of Sy from 0.01 to 0.30. In contrast, Kc depends on the type of vegetation, the time during the growing season and the available soil moisture. As a result, Kc is not constant over the course of the year. In fact, Wight and Hanson (1990) report Kc values for sagebrush rangelands ranging from about 0.05 to over 1.2 during the entire growing season (May–August). If Kc values for other vegetation types were included, the range would be even larger and it would be even more difficult to pick an appropriate value for both sites.

Conclusions Diurnal water-table fluctuations were observed during the growing season at two sites in the riparian zone of Red Canyon Creek, Wyoming, USA. Seasonal declines in water-table fluctuations are associated with the transition to dormancy and abrupt onset and cessation of water-table fluctuations were associated with daily temperatures rising above or falling below freezing. Differences in vegetation and sediment texture between the two sites accounted for differences in the magnitude of water-table fluctuations observed. The larger magnitude fluctuations at W1 were associated with the presence of wetland plants that have higher transpiration rates and the more uniform sediment texture of the subsurface, which likely had a lower specific yield. ETG was estimated using diurnal water-table fluctuations and daily rates of ETG in the riparian wetland were generally between 0.5 and 3 mm/day. Outside of the wetland, in the stream floodplain, ETG was generally less than 0.5 mm/day. Heterogeneity of the sediment outside the wetland, at W2, may have resulted in an underestimation of ETG. A fine sand lens interbedded within the clayey silt in proximity of the water table should have a relatively higher specific yield. ETG scales linearly with specific yield, so an increase in specific yield from 7% (clayey silt) to 30% (sand) would cause the ETG estimate to increase by 4.3 times. Although the scaling of ETG at W2 may be affected by specific yield, the temporal differences in ETG at W1 and W2 are also caused by differences in vegetation between the two sites. The Hydrogeology Journal (2008) 16: 483–497

wetland supports phreatophytic vegetation, which draws heavily from the groundwater for transpiration. Although meadow grasses outside the wetland do not draw as much water from the aquifer, they still generate readily observable water-table fluctuations caused by transpiration. Ratios of ETG to PET in the riparian wetland show that ETG becomes an increasingly higher percentage of PET over time during dry periods between precipitation events. This suggests that as soil moisture near the land surface declines, plants may become opportunistic, drawing a larger amount of water from the water table. This observation is in contrast to observations of ET near the land surface. Actual ET at the land surface (relative to PET) declines as soil moisture is depleted and ET becomes water limited. ETG rates based on diurnal water-table fluctuations may be more reliable than simplified estimates of ETG that are based on a linear decrease of PET (or ET) with depth and therefore may be more appropriate for groundwater modeling applications. Although there is comparable error in both methods, derived mainly from the selected values of Sy or Kc, it is easier to reliably estimate readily available specific yield. Sy is based solely on the material properties of the subsurface, which do not change with vegetation type or available soil moisture. In contrast, Kc is highly variable over time due to changes in available soil moisture and plant growth, seasonally. A limitation of the White (1932) method is the need for a high density of wells to record the spatial variability of ETG. Therefore, this technique will be most cost-effective at sites that require dense well networks for other purposes such as collecting water-table depth information for groundwater model calibration. Diurnal water-table fluctuations are readily observed in semi-arid catchments, where there are prolonged dry periods with little groundwater recharge from precipitation. During periods with enough precipitation to recharge the aquifer, diurnal water-table fluctuations are masked by large changes in the elevation of the water table that are not due to evapotranspiration. For this reason, this technique has more limited application in humid catchments or during wet seasons in semi-arid settings. Acknowledgements This work was supported by the National Science Foundation under grant number 0450317. I would like to thank the University of Missouri for supporting the fieldwork for this project and the Nature Conservancy of Wyoming for access to the research site. I also thank two anonymous reviewers and the associate editor of the Hydrogeology Journal for their constructive comments that improved the quality of the manuscript.

References Bailey RG (1995) Descriptions of the ecoregions of the United States. Miscellaneous Publication No. 1391, US Forest Service, Washington, DC Banta ER (2000) MODFLOW 2000, the US Geological Survey modular ground-water model: documentation of packages for simulating evapotranspiration with a segmented function DOI 10.1007/s10040-007-0239-0

497 (ETS1) and drains with return flow (DRT1), US Geol Surv Open-File Rep 00-466, 127 pp Butler JJ, Kluitenberg GJ, Whittemore DO, Loheide SP, Jin W, Billinger MA, Zhan XY (2007) A field investigation of phreatophyte-induced fluctuations in the water table. Water Resour Res 43, W02404. DOI 10.1029/2005WR004627 Chen XH (2007) Hydrologic connections of a stream-aquifervegetation zone in south-central Platte River valley, Nebraska. J Hydrol 333:554–568 Cooper DJ, Sanderson JS, Stannard DI, Groeneveld DP (2006) Effects of long-term water table drawdown on evapotranspiration and vegetation in an arid region phreatophyte community. J Hydrol 325:21–34 Dolan TJ, Hermann AJ, Bayley SE, Zoltek J (1984) Evapotranspiration of a Florida, USA, freshwater wetland. J Hydrol 74:355– 371 Engel V, Jobbagy EG, Stieglitz M, Williams M, Jackson RB (2005) Hydrological consequences of eucalyptus afforestation in the argentine pampas. Water Resour Res 41, W10409. DOI 1029/ 2004WR003761 Fisher JB, DeBiase TA, Qi Y, Xu M, Goldstein AH (2005) Evapotranspiration models compared on a Sierra Nevada forest ecosystem. Environ Model Softw 20:783–796 Flerchinger GN, Hanson CL, Wight JR (1996) Modeling evapotranspiration and surface energy budgets across a watershed. Water Resour Res 32:2539–2548 Freeze RA, Cherry JA (1979) Groundwater. Prentice-Hall, Englewood Cliffs, NJ Gerla PJ (1992) The relationship of water-table changes to the capillary fringe, evapotranspiration, and precipitation in intermittent wetlands. Wetlands 12:91–98 Healy RW, Cook PG (2002) Using groundwater levels to estimate recharge. Hydrogeol J 10:91–109 Kurc SA, Small EE (2004) Dynamics of evapotranspiration in semiarid grassland and shrubland ecosystems during the summer monsoon season, central New Mexico. Water Resour Res 40(9), W0930515. DOI 10.1029/2004WR003068 Lautz LK, Siegel DI (2006) Modeling surface and ground water mixing in the hyporheic zone using MODFLOW and MT3D. Adv Water Resour 29:1618–1633 Lautz LK, Siegel DI, Bauer RL (2006) Impact of debris dams on hyporheic interaction along a semi-arid stream. Hydrol Process 20:183–196 Loheide SP, James J, Butler J, Gorelick SM (2005) Estimation of groundwater consumption by phreatophytes using diurnal water table fluctuations: a saturated-unsaturated flow assessment. Water Resour Res 41, W07030. DOI 07010.01029/ 02005WR003942 Meyboom P (1967) Groundwater studies in the Assiniboine River drainage basin, Part II: hydrologic characteristics of phreato-

Hydrogeology Journal (2008) 16: 483–497

phytic vegetation in south-central Saskatchewan. Geol Surv Can Bull 139:1–64 Nichols WD (1994) Groundwater discharge by phreatophyte shrubs in the Great Basin as related to depth to groundwater. Water Resour Res 30:3265–3274 NOAA-NCDC (2006) Program overview. http://www.ncdc.noaa. gov/oa/climate/uscrn/. Cited March 2007 Patten DT (1998) Riparian ecosystems of semi-arid North America: diversity and human impacts. Wetlands 18:498–512 Penman HL (1948) Natural evaporation from open water, bare soil, and grass. Proc R Soc Lond A193:120–146 Prater MR, DeLucia EH (2006) Non-native grasses alter evapotranspiration and energy balance in Great Basin sagebrush communities. Agric For Meteorol 139:154–163 Priestley CHB, Taylor RJ (1972) On the assessment of surface heat flux and evaporation using large scale parameters. Mon Weather Rev 100:31–92 Rosenberry DO, Winter TC (1997) Dynamics of water-table fluctuations in an upland between two prairie-pothole wetlands in North Dakota. J Hydrol 191:266–289 Schilling KE (2007) Water table fluctuations under three riparian land covers, Iowa (USA). Hydrol Process 21:2415–2424, DOI 10.1002/hyp.6393 Schilling KE, Kiniry JR (2007) Estimation of evapotranspiration by reed canary grass using field observations and model simulations. J Hydrol 337:356–363 Shah N, Nachabe M, Ross M (2007) Extinction depth and evapotranspiration from ground water under selected land covers. Ground Water 45:329–338 Shuttleworth WJ (1993) Evaporation. In: Maidment DR (ed) Handbook of hydrology. McGraw-Hill, New York, pp 4.1–4.53 Teuling AJ, Seneviratne SI, Williams C, Troch PA (2006) Observed timescales of evapotranspiration response to soil moisture. Geophys Res Lett 33, L23408. DOI 10.1029/2006GL028178 Thompson SA (1999) Hydrology for water management. Balkema, Brookfield, VT Thornthwaite CW (1948) An approach towards a rational classification of climate. Geogr Rev 38:55–94 White WN (1932) A method of estimating ground-water supplies based on discharge by plants and evaporation from soil: results of investigations in Escalante Valley, Utah. US Geol Surv Water Suppl Pap 659-A Wight JR, Hanson CL (1990) Crop coefficients for rangeland. J Range Manage 43:482–485 Winter TC, Rosenberry DO, Sturrock AM (1995) Evaluation of 11 equations for determining evaporation for a small lake in the north central United States. Water Resour Res 31:983–993 Zhang YK, Shilling KE (2006) Effects of land cover on water table, soil moisture, evapotranspiration, and groundwater recharge: a field observation and analysis. J Hydrol 319:328–338

DOI 10.1007/s10040-007-0239-0