Summary and Conclusions


Water allocation in the Klamath Basin is important to many different groups of people, and to the health and well‑being of many ecosystems within the basin. Competition for the limited water supply in this semiarid basin is intense, and knowledge of water losses to evapotranspiration (ET) is key information for deciding optimal water-use strategies. One of the central hubs for water distribution in the basin is Upper Klamath Lake, near Klamath Falls, Oregon. The U.S. Geological Survey (USGS) and the Bureau of Reclamation conducted a study to quantify ET from extensive wetlands in the Upper Klamath National Wildlife Refuge (NWR) at the northwest periphery of the lake, and evaporation from the open-water portion of the lake. Data collection spanned the period from May 1, 2008, to September 29, 2010.


Two wetland sites were selected for study to typify vegetation communities and hydrologic conditions most frequently occurring in the NWR. Vegetation at one wetland site consisted of a virtual monoculture of bulrush (formerly Scirpusacutus, now Schoenoplectus acutus), while the other site was a mixture dominated by bulrush, with minor amounts of cattail (Typha latifolia), wocus (Nuphar polysepalum), open water, and trace amounts of other vegetation. The bulrush site is at an altitude of 1,262.03 m (4,140.02 ft) asl and the mixed site is about 0.15 m (0.50 ft) higher. As a result of controlled lake levels, the wetlands are periodically inundated with lake water on an annual basis. Standing water typically occurs from January or February through July or August, for an average hydroperiod of about 6 months at the bulrush site and 5 months at the mixed site. Water level typically fluctuates about 1.3 m, from about 0.8 to 0.9 m above land surface at the bulrush site in spring and early summer to about 0.4 to 0.5 m below land surface in October, with correspondingly lower levels at the mixed site. Minimum lake level in fall 2009 was unusually low, resulting in a late water-level rise and reduced maximum water level in 2010. The root zone of the bulrush site remained saturated during the entire study period (volumetric water content, θ, was 0.8–0.9 m³/m³), whereas the mixed site root zone partially dewatered to a θ of around 0.5 m³/m³ during late summer and re-saturated in late winter or spring. Canopy height at both sites typically reaches a maximum of 2.2 to 2.3 m during the summer, although maximum height at the mixed site in 2010 was only 1.9 m. When the canopy senesces in October, the dead stalks remain vertical for some time, and eventually lodge over in response to snow and wind loading. The bent-over stalks form a loosely distributed mat, generally about 0.6 to 1.0 m high.


The eddy-covariance (EC) method was used to measure ET from two wetland sites. A source-area model estimates that at the bulrush site, 98 to 99 percent of the measured ET originated from within the wetland, and at the mixed site, 95 to 96 percent originated from within the wetland, indicating that contamination of the EC measurement by other surface types was insignificant. A chronic inability to close the surface energy balance using the EC method was remedied by adjusting the turbulent flux upward to equal the long‑term available energy, while maintaining the ratio between the sensible- and latent-heat fluxes (the Bowen ratio, or β). This common adjustment (Foken, 2008) made the estimates of wetland and open-water evaporation consistent with one another.


Partitioning of available energy (AE) into sensible-heat flux (H) and latent-heat flux (LE) varies dramatically during the seasonal cycle (LE is the energy equivalent of the ET rate). At the beginning of the calendar year, the remnant dead vegetation mat from previous years is the dominant exchange surface with the atmosphere and solar radiation, although rising water levels soon overtop the land surface, forming a secondary, mostly shaded water surface beneath the dead canopy. When the canopy is dry, β tends to be greater than one. Precipitation briefly shifts the balance toward LE for one to a few days, dropping β below 1, but evaporation and infiltration of intercepted water to lower layers soon return the partitioning to β > 1. As a result of these repeated rapid reversals, on average, β ≈ 1 during most of the winter. As AE increases with growing sun angles in the spring, both H and LE grow in response, maintaining a β of around one. The new green transpiring vegetation typically emerges from the water and dead stalk mat in May or June, at which time partitioning begins to shift rapidly toward LE. By July LE has reached peak values, and H has actually decreased from its spring values, even though AE has increased to peak values. Through much of the growing season, LE is much greater than H, as indicated by β values around 0.26 at the bulrush site and 0.13 at the mixed site. Possibly the greater partitioning toward LE at the mixed site is in response to the greater proportion of open water at that site, in spite of lower standing water levels. As the vegetation canopy senesces during September, transpiration decreases, and partitioning ceases to favor LE. During the study, the Octobers were unusually dry, receiving 52 percent of mean 1999 through 2011 precipitation at Klamath Falls. This lack of precipitation and transpiration, coupled with water levels below land surface, typically resulted in β of around 2 or 3. Greater precipitation in November and December (the two greatest precipitation months historically) again restores the average β to around one.


Measured ET at the wetland sites is compared to reference ET (ET_r) computed from data collected at the nearby Bureau of Reclamation Agency Lake weather station (AGKO), to compute crop coefficients (K_c) at daily, biweekly, and annual time steps. Approximate formulas are given to estimate daily values of growing season K_c, thereby allowing computation of daily ET using ET_r from the AGKO weather station. Biweekly values of growing season K_c are computed from ensemble average values of ET and ET_r during the 3-year study period growing seasons, and a single, mean value of K_c is computed for the non-growing season. Together, these provide relatively accurate estimates of biweekly ET during the study (RMSE = 0.396 and 0.347 mm d^-1, r ² = 0.962 and 0.971 at the bulrush and mixed sites, respectively). A fourth‑order polynomial fit of the growing season values to day of year provides a more automated form of ET computation.


Annual values of ET from both study sites are computed for the calendar years of 2008, 2009, and 2010. Annual values for 2008 and 2010 are synthesized by substituting non-growing season ET data from 2009 for periods before the study began (January 1–April 30, 2008) and after the study ended (September 30–December 31, 2010). Little uncertainty is introduced by this substitution because ET is small during these times. Annual ET at the bulrush site ranged from 0.850 to 1.013 m/yr, with a 3-year mean of 0.953 m/yr (table 11). Annual ET at the mixed site ranged from 0.814 to 0.966 m/yr, with a 3-year mean of 0.903 m/yr. Based on a satellite-image estimate that the bulrush site typifies 70 percent of the Upper Klamath Lake National Wildlife Refuge (NWR) and the mixed site typifies 30 percent, the resulting estimate of mean NWR ET during the 3 years is 0.938 m/yr. At both sites, minimum annual ET occurred during 2010, a year of unusually low water level during spring and early summer, and a year that had 16 percent less precipitation than in 2008 through 2009. Although the lower 2010 water level almost surely did not stress the vegetation, the water surface remained more shaded and decoupled from the atmosphere, leading to less surface water evaporation. Interannual variability (about ± 8.5 percent) was substantially greater than intersite variability (about ± 2.5 percent). A paired t-test conducted on site differences indicated that daily ET values from the two sites were statistically, but only marginally different from one another at the α = 0.05 level.


Study-period averages of solar radiation (R_s), air temperature (T_a), and precipitation (P) measured at nearby AgriMet stations all were lower than their corresponding 11-year averages from 2001–2011. Normally, lower R_s is associated with greater P, but in this case, the study period was less sunny and drier than the long-term average. These climatic conditions all suggest that ET measured during the study period probably is less than the expected long‑term average. However, because R_s and T_a appear in the expression for reference ET (ET_r), and the effect of P on ET in this wetland setting is minor, the crop coefficient approach developed here should remain relatively robust in other climatic settings.


A comparison of measured growing-season wetland ET with alfalfa and pasture ET was made by tabulating alfalfa and pasture ET totals posted for the Klamath Falls station (KFLO) on the Bureau of Reclamation AgriMet Web site (http://www.usbr.gov/pn/agrimet/ETtotals.html) and obtaining growing season durations (http://www.usbr.gov/pn/agrimet/chart/kflo08et.txt). During the 190-day average alfalfa growing seasons from 2008 through 2010, wetland ET at our study sites was 0.779 m, about 7 percent less than alfalfa mean ET of 0.838 m. During the 195-day average pasture growing seasons from 2008 through 2010, wetland ET at our study sites was 0.789 m, about 18 percent greater than pasture mean ET of 0.671 m. A comparison of annual ET can be made by assuming alfalfa and pasture ET are equal to wetland ET during the non‑growing season. This assumption leads to annual estimates for 2008 through 2010 of 0.997, 0.938, and 0.820 m of ET from alfalfa, wetland, and pasture land covers, respectively.


An earlier ET study of this wetland used short‑term (1- to 2-day) EC measurements of ET made four times during the growing season of 1997 to calibrate a Penman‑Monteith ET model, driven by continuous onsite weather data, to compute growing-season ET (Bidlake, 2000). The study site was located about 200 m northwest of our bulrush site, with virtually identical vegetation and altitude. Total growing‑season ET from that study was comparable to our 2008 and 2009 ET, with notable short‑term differences that appear to be related to differences in cloudiness. The two studies are consistent, and together they provide a well-documented estimate of Upper Klamath NWR evapotranspiration.


Bidlake (2002) also studied ET from fallowed cropland in the Tule Lake NWR during the growing season of 2000, using the Bowen-ratio energy balance method supplemented with modeling to fill in data gaps when sensors malfunctioned, and to extend the period of record to the full growing season (defined as May 1–October 31 in that study). Although the seasonal timing of ET varied considerably among the three types of fallow surfaces studied, total ET was remarkably constant, at 0.435 ± 0.009 m during the growing season. During the same months, mean ET measured at the Upper Klamath Lake NWR wetland in the current study was 0.718 m, or about 65 percent greater than the fallowed‑cropland ET. If the non-growing-season ET measured at the wetland (0.220 m) also is assumed to occur from the fallowed cropland, the resulting annual ET from the wetland during 2008 through 2010 (0.938 m) is about 43 percent greater than the fallowed cropland ET (0.655 m) during 2000.


Open-water evaporation was measured at two midlake locations during 23 warm season biweekly periods during the study, using the Bowen-ratio energy balance method. Seasonal patterns of open-water evaporation were similar to those of wetland ET, but exhibited less of a seasonal cycle. Open-water and wetland magnitudes were nearly equal during late June to early August, when wetland vegetation was green and abundant. During late summer, open-water evaporation consistently exceeded wetland ET as the vegetation canopy began to senesce, while the lake released much of the energy stored during the summer, as latent‑heat flux. Spring comparison data are few, but suggest that open‑water evaporation exceeds wetland ET when emergence of the new vegetation canopy out of the old, dead stalk mat is not complete. Overall, measured open-water evaporation was 20 percent greater than wetland ET during the same periods.


First posted March 4, 2013