The purpose of this report is to assess long-term changes in surface-water extent within Klamath Marsh. It documents the development of a dataset that quantifies the maximum area of wetland and open water in the marsh for each water year¹ 1

The 12-month period from October 1, for any given year, through September 30, of the following year. The water year is designated by the calendar year in which it ends.

Klamath Marsh is a large wetland in south-central Oregon spanning an area of about 70 square miles (mi²; 45,000 acres). The study area encompasses about 1,300 mi², the part of the Williamson River Basin upstream from the U.S. Geological Survey (USGS) streamgage Williamson River near Klamath Agency, OR (11493500, hereinafter referred to as the Klamath Agency streamgage; fig. 1). The marsh is centrally located within the Williamson River Basin, situated in a topographic low between the Cascade Range to the west and Yamsay Mountain to the east. The altitude of the low-elevation area surrounding Klamath Marsh ranges from 4,500 to 4,600 feet (ft) with Klamath Marsh having a mean altitude of about 4,520 ft. Study area margins range in altitude from about 5,000 to over 8,000 ft at prominent peaks such as Mount Scott, Mount Thielsen, and Yamsay Mountain. The study area is bounded on the north by a broad plateau, separating it from the Little Deschutes River Basin to the north, and on the west by the Cascade Range. Volcanic peaks and fault scarps separate the Williamson River Basin from the Sprague River Basin to the south and the Summer Lake Basin to the east.

The development of the January-through-June annual maximum surface-water extent (MSWE) geospatial dataset (Kennedy, 2023) is based on the Dynamic Surface Water Extent (DSWE) model by Jones (2015; 2019). The DSWE model uses Landsat imagery to detect surface-water inundation within open and vegetated areas using a series of diagnostic tests to determine the probability of the presence of water in each pixel, which is then classified based on the probabilistic confidence that the pixel contains water (U.S. Geological Survey, 2022). Although DSWE data, based on a revised approach, are available as a USGS Landsat Collection 2 Level-3 science product (U.S. Geological Survey, 2022), the MSWE dataset was created using the method previously developed by Jones (2019) for Collection 1 DSWE data production as described in U.S. Geological Survey (2018), and implemented within Google Earth Engine (GEE; Soulard and others, 2020; Walker and others, 2020). GEE was used to leverage the processing power needed to access the entire Landsat period of record in a manner that would fit within the project timeframe and budget constraints.

The DSWE algorithm (Jones, 2019) uses water and vegetation indices to classify each pixel within an image as not water, open water—high confidence, open water—moderate confidence, partial surface water—conservative, partial surface water—aggressive, and cloud/shadow/snow. Pixels classified as open water—high confidence and open water—moderate confidence contain areas of open water. Pixels classified as partial surface water—conservative and partial surface water—aggressive contain surface water and vegetation (Jones, 2019). Pixels classified as not water were assumed to be dry land, and pixels classified as cloud/shadow/snow indicate that the pixel was contaminated by the presence of clouds, cloud shadow, or snow using the Landsat Quality Assessment band.

The initial step for creating the MSWE dataset was to constrain a fixed area of the marsh for which changes in the extent of surface-water inundation would be calculated. Defining a marsh boundary was an iterative process using a 10-meter (m) resolution digital elevation model (DEM; U.S. Geological Survey, 2021) and resulted in the selection of the 4,522-ft elevation contour because it encompassed those parts of the marsh most prone to surface-water inundation. Excepting some dry, upland areas, this elevation contour encompasses the extent of the KMNWR. The contour line was modified to form a closed polygon (hereinafter referred to as the marsh boundary) at the southeast end of the marsh where the marsh converges into the Williamson River channel (fig. 2).

The MSWE dataset integrated two modifications to the GEE code written by Walker and others (2020). First, the script was modified to use a higher-resolution 10-m DEM instead of the 30-m DEM used in the published code. Second, the timeframe during which Landsat imagery was composited was shortened to only include imagery obtained during January 1–June 30 of each year. This time frame was selected for two reasons. First, it coincided with maximum streamflow from snowmelt and the typical period of maximum water extent in the marsh based on visual examination of historical imagery and anecdotal observations. Second, water-management practices diverted water through a system of ditches from wet areas to drier areas within the marsh (U.S. Fish and Wildlife Service, 2014) during late summer. Excluding imagery after June minimized the inclusion of areas that may have been artificially flooded.

The GEE code created annual raster layers of MSWE by compositing every scene within the designated annual timeframe (January through June), selecting only the highest confidence water-containing pixel type from each individual Landsat image. Each annual MSWE raster was then clipped to the extent of the marsh boundary polygon described previously. The resulting output is a January-through-June MSWE raster image for each year (1985–2021) for which useable Landsat imagery exists. Surface-area calculations of each MSWE classification type were then calculated for each year. All MSWE data are made available in a USGS data release² 2

Pixel classification names in this report vary slightly from pixel classification names in the data release. Pixel classifications defined in this report are shown here, with their data release counterpart shown in parentheses: open water—high confidence (Water—high confidence); open water—moderate confidence (Water—moderate confidence); partial surface water—conservative (Wetland—moderate confidence); partial surface water—aggressive (Wetland—low confidence). The classification names not water and cloud/shadow/snow are the same in both this report and in the data release.

MSWE imagery was compared with a high-resolution light detection and ranging (lidar) DEM of Klamath Marsh to ensure that pixels classified as containing water corresponded with areas of lower topography. The imagery was also compared with available high-resolution true-color imagery where standing water existed during a similar time frame. The qualitative evaluation indicated that the algorithm consistently matched MSWE pixels with areas of lower topography in the DEM and areas containing standing water in the true-color imagery. Visual inspection of the 37 annual MSWE images indicated that scenes from 2008 and 2012 contain linear stripes. These stripes are an artifact in Landsat 7 imagery caused by the mechanical failure of the scan-line corrector, resulting in data gaps along the margins of images collected since June 2003 (U.S. Geological Survey, 2015). These data gaps exist in the 2008 MSWE image along the eastern margin of the central part of the marsh. Data gaps in the 2012 image are more noticeable, with linear stripes apparent across the entire extent of the marsh. Although the data gaps appear to have a minimal impact on the 2008 image, the data gaps in the 2012 image appear to underestimate the extent of pixels classified as open water—high confidence. Despite this issue, the imagery and surface-area calculations for 2008 and 2012 provided useful information and were included in the analysis.

Trends in MSWE, stream discharge, groundwater level, total annual precipitation, and annual mean temperature were evaluated using the Mann-Kendall (MK) trend test. A 90-percent significance level (p-value less than [<] 0.10) was used to characterize the likelihood of trends. Correlations among these time-series datasets were evaluated using Spearman’s rank correlation coefficient. Serial correlation was not addressed as part of this study. Locally weighted scatterplot smoothing (LOWESS) lines were used to smooth, visually inspect, and compare time-series data (Gramfort, 2015). Because the various datasets used in the study contained differing numbers of data and periods of record and because a similar degree of smoothness was desired among the datasets for comparison purposes, the value of the smoothing parameter used to generate the LOWESS lines varied from 0.15 to 0.22.

The spatial extent of surface-water inundation was assessed as a percentage of the total marsh boundary area (44,201 acres; 69.1 mi²) for each pixel type (fig. 3). Pixels classified as not water (dry land) and open water—high confidence had the largest year-to-year variability and the largest changes during the 37-year study period. During the first half of the study period, 1985–2003, the mean surface area of pixels classified as not water and open water—high confidence accounted for 34 and 22 percent of total marsh area, respectively. During the latter half of the study period, 2003–21, the mean surface area of pixels classified as not water increased to account for 50 percent of the marsh, whereas the mean surface area of pixels classified as open water—high confidence decreased by half, accounting for 11 percent of total marsh area. By comparison, the other three MSWE classifications had less variability and no substantial long-term changes during the study period. Surface area for 2003 was included in calculations for the first half and latter half of the study period.

The Mann-Kendall trend test was performed for each of the five MSWE classifications (table 1). Statistically significant (p-value <0.1) trends during 1985–2021 were identified for pixel classifications not water (upward), open water—high confidence (downward), open water—moderate confidence (downward), and partial surface water—aggressive (downward). These results corroborate the visual evaluation of the LOWESS line through the data (fig. 3A–E). A change in the area of pixels classified not water is nearly entirely due to a corresponding and opposite change in the area of pixels classified as open water—high confidence and open water—moderate confidence. Subsequent analyses and discussion can therefore be simplified to total open-water extent (TOWE), which is the combined area of pixels classified as open water—high confidence and open water—moderate confidence (fig. 3F).

Continuously gaged streamflow data for the Williamson River have been collected at two sites within the study area (fig. 1; table 2). The OWRD Sheep Creek streamgage is about 9 mi upstream from the KMNWR boundary and has data for WYs 1974–2022 (Oregon Water Resources Department, 2023b). USGS Klamath Agency streamgage serves as the downstream drainage basin boundary of the study area and has data available for WYs 1955–2022 (U.S. Geological Survey, 2023). Total annual discharge by WY for each streamgage was calculated from the mean daily flow values for the period of the MSWE analysis (WYs 1985–2021). Both sites contained periods of missing measurements. The Sheep Creek streamgage had three periods of missing data: December 12, 2001, to February 4, 2002; April 25 to June 19, 2002; and October 1, 2005, to May 1, 2006. The result of these missing data is an absence of total annual discharge calculations for WYs 2002 and WY2006. The mean of the data from the years bracketing the missing WYs was used for the LOWESS line. The Klamath Agency streamgage had one period of missing data from October 25, 1995, to October 1, 1998. Because of the 3-year gap and large year-to-year variability at this location, we deemed it inappropriate to statistically fill the missing values and generated separate LOWESS lines for WYs 1985–95 and 1999–2021.

Streamflow dynamics at the two gaged locations in the study area are considerably different (fig. 4). The difference partly reflects the streamgage locations and the contributing area each integrates. The catchment of the Sheep Creek streamgage comprises 14 percent of the study area, drains the drier eastern part, and is relatively stable because of a large groundwater contribution (Gannett and others, 2007). In contrast, the Klamath Agency streamgage aggregates streamflow for the entire study area; therefore, annual variability in snowpack amount and snowmelt timing in the Cascade Range likely has a major influence on the large variability in streamflow at the site. During and prior to the study period, flow commonly ceased at the Klamath Agency streamgage for 1–3 months each year. In recent years, however, the frequency of zero-flow months has increased; zero flow was measured at the streamgage during April 2020–April 2023, which is the longest dry period on record for the Klamath Agency streamgage. In contrast, annual discharge at the Sheep Creek streamgage has remained within 35 percent of the study period median with no dry periods.

Mann-Kendall trend tests substantiate the trends apparent in the LOWESS lines and indicated statistically significant (p-value <0.1) declines in total annual discharge at both streamgages during the 1985–2021 study period (table 3). The 3-year gap in discharge measurements at the Klamath Agency streamgage introduces an uncertainty in the magnitude of the trend. Additional care in the interpretation of the trend statistic at this location is warranted because of the large historical variability in total annual streamflow, which can vary by an order of magnitude within a 6-year period; for example, from 1986 to 1992 and 1999 to 2005.

Available groundwater monitoring locations are sparse and limited to the northern part of the study area (fig. 1, table 4). All groundwater-level data were obtained from the USGS National Water Information System (U.S. Geological Survey, 2023); any measurement that was interpreted from a geophysical log or was designated as having been affected by pumping was excluded. Groundwater-level measurements were not temporally consistent, and no wells had measurements throughout the full MSWE period of record (1985–2021). KLAM 522 had the longest period of record spanning WYs 1962–2023. Regular, periodic water-level measurements within the basin began in WY 2000 for wells KLAM 637 and KLAM 11774 and in WY 2001 for wells KLAM 527, KLAM 561, and KLAM 562. Well KLAM 527 had consistent quarterly measurements until it was decommissioned in 2012 and monitoring efforts switched to KLAM 522 (about 300 ft away) starting in 2013. Wells KLAM 561 and KLAM 11774 were consistently monitored until WY 2005 when monitoring efforts were stopped. Well KLAM 615 only had two measurements during the study period, one in WY 2000 and one in WY 2021. Quarterly measurements at wells KLAM 561, KLAM 615, and KLAM 11774 resumed in 2021 as part of this project. Wells KLAM 637 and KLAM 562 have been consistently monitored since WYs 2000 and WY2001, respectively. No consistent groundwater-level measurements were collected at any of the wells during the 1990s.

Declining trends in groundwater level were observed in all wells during the study period (fig. 5). An MK trend test was performed on groundwater-level measurements for all wells with more than two measurements; all trends were downward and statistically significant for WYs 2001–21 (p-value <0.1; table 5). Rates of decline were calculated from the slope of a least-squares regression through the groundwater-level record of each well (table 5). Mean annual rate of change was also calculated from the difference of the first and last water-level measurement for each period of analysis and divided by the time spanning the two measurements (table 5). Among wells having a period of record of WYs 2001–21, the mean decline in groundwater level based on the linear regression slope was 18.8 ft, with a mean decline rate of 0.94 ft per year. Two prominent increases in groundwater level were observed in 2006 and 2017. Wells KLAM 561, KLAM 562, and KLAM 637 all had similar rates of groundwater-level decline at 1.2 ft per year during WYs 2001–21. KLAM 615 is closest to areas where most of the groundwater withdrawal for irrigation occurs, and the decline was similar to the decline in other wells throughout the study area.

Groundwater-level measurements at KLAM 522 do not have the same period of record as other wells in the study area, but the well provides a useful historical context for groundwater level near Klamath Marsh. During WYs 1962–87, groundwater level varied from 4,530 to 4,547 ft, likely in response to periods of wetter and drier conditions in the basin. For this period, a small statistically significant increase in groundwater level was observed (+0.31 ft per year; p-value <0.1; table 5). Sometime during WYs 1987–2001, the groundwater system shifted from a state of fluctuating groundwater level (as observed prior to 1987) to one of nearly continuous year-over-year decline. During WYs 2013–21, a statistically significant decline in groundwater level at KLAM 522 was observed at a rate comparable to other wells nearby (table 5).

Groundwater pumpage for agricultural irrigation, fire suppression, manufacturing, and construction within the study area is permitted at 26 well locations, also referred to as a point of diversion (POD) by OWRD (fig. 6). Maximum annual pumpage for permitted wells was estimated from permitted groundwater rights retrieved from OWRD Water Rights Information System (WRIS; Oregon Water Resources Department, 2023c). Irrigation water rights are classified as primary or supplemental. A primary water right designates the POD as the principal water source for the authorized use. A supplemental water right designates that the POD can only be used to make up for a deficiency in available water from a primary water right. Additional pumpage from unpermitted wells is unknown; unpermitted pumpage for irrigation likely is minimal based on an assessment of historical aerial imagery. The pumpage estimates might reflect a maximum value because (1) reporting of actual pumpage or water use is not required of the permittees, (2) many of the water rights are for supplemental irrigation and, therefore, are not used every year, (3) withdrawals might not be at the full permitted rate, and (4) not all well pumps are in working condition. The permitted maximum annual pumpage for each groundwater right within the study area is summarized in table 6. The rates obtained from WRIS (in cubic feet per second) were converted to acre-feet per day and multiplied by the maximum number of days permitted for the POD.

Groundwater permits for irrigation represent 99 percent of the permitted groundwater use in the study area and the maximum permitted groundwater pumpage for all uses is 48,082 acre-feet per year (acre-ft/yr). In terms of both the number of water rights issued and the associated permitted pumpage, most groundwater rights were issued during the 1960s and 1970s (table 6; fig. 7). During those 2 decades, the permitted maximum annual groundwater pumpage for primary irrigation increased to about 25,700 acre-feet (acre-ft) and supplemental irrigation increased to about 12,300 acre-ft. During 1980–2004, the permitted maximum annual groundwater pumpage for primary irrigation and supplementary irrigation increased by about 10,000 acre-ft, and no new permits have been issued since.

Precipitation and temperature data for six sites (table 7) were obtained from the National Oceanic and Atmospheric Administration’s National Centers for Environmental Information (NCEI) Global Summary of the Month dataset (Lawrimore and others, 2016). The sites were selected to represent the spatial variability of climate across the study area. These data were used to assess trends in total annual precipitation and annual mean temperature by WY.

Correlations among the time series of TOWE, stream discharge, groundwater level, precipitation, and temperature were evaluated using the Spearman’s rank correlation test. Correlation coefficients (Spearman’s Rho) were computed for two periods of record, WYs 1991–2021 (fig. 10) and WYs 2001–21 (fig. 11). These periods correspond to the maximum overlap among the datasets. The magnitude and sign of the correlation coefficients are similar among the two periods, although the magnitude of the correlation tends to be somewhat weaker for the shorter period, which was expected because of the smaller sample size. A small number of correlation coefficients that were significant in the longer analysis period were not significant in the shorter analysis period.

A high degree of correlation exists in the temperature records and precipitation records among the six climate sites (figs. 8 and 9), so correlations were first developed among the six climate sites to reduce redundancy. Among the six climate stations, correlation coefficients of annual precipitation records for WYs 1985–2021 ranged from 0.78 to 0.93 and correlation coefficients of annual mean temperature records for WYs 1991–2021 ranged from 0.64 to 0.94. Subsequent analyses and discussions used two climate stations, USS0021F12S and USC00351946. Climate station USS0021F12S was used to represent conditions across most of the study area east of the Cascade Range. This station had the strongest temperature correlations (0.89–0.91) with the other four lower-altitude sites. Climate station USC00351946 at Crater Lake was used to represent conditions at high altitudes in the Cascade Range.

Groundwater level across the Williamson River Basin upstream from the Klamath Agency streamgage has been declining since at least 2000 (fig. 5). Although groundwater-level data for the study area prior to WY 2000 are sparse and the onset of declines is not well documented, sustained decreases in open-water area for the study period began after WY 2000 (fig. 3). Therefore, groundwater level for WY 2000 is useful for illustrating the relation between groundwater and open-water area in the marsh prior to substantial declines in open-water area and provide a contrast to current conditions.

In summer 2000, the groundwater level in two wells on opposite sides of the upper marsh (KLAM 11774, 300-ft deep, open from 4,505 to 4,224 ft above NAVD 88; KLAM 637, 136-ft deep, open from 4,478 to 4,458 ft above NAVD 88) was at or above the altitude of the marsh land surface (fig. 13), about 4,520 ft. The groundwater level has been declining since WY 2000 and was about 10 ft below marsh land surface at KLAM 11774 and about 20 ft below marsh land surface at KLAM 637 by July 2021. Because of the decline in groundwater level over the previous 21 years, a 10- to 20-ft-thick unsaturated zone formed beneath the marsh.

Although information about the subsurface geology under Klamath Marsh is limited, the available data indicate the upper 100 to 200 ft of material are moderately to highly permeable. A surficial layer of peat and diatomaceous silts, varying in thickness from 3 to 5 ft, are present across much of the marsh. The thin surficial layer has a lower permeability than deeper deposits and attenuates the hydraulic interaction between surface water and groundwater (U.S. Fish and Wildlife Service, 2010). Evidence for material with much higher permeability beneath the surficial peat layer comes from multiple sources, including driller descriptions and yield tests from OWRD water well reports; numerous losing streams on the marsh periphery, few established surface-water channels within the marsh and surrounding area, low groundwater-level gradients (Gannett and others, 2007), and relatively high hydraulic conductivities estimated from a calibrated groundwater model (Gannett and others, 2012).

Total open-water extent is well-correlated with annual discharge in the Williamson River at both streamgages throughout the study period (figs. 10, 11, and 14); however, discharge at Sheep Creek has a stronger correlation during the first half of the study period (1985–2003; Spearman’s Rho =0.80, p-value= <0.001), compared to the second half (2003–21; Spearman’s Rho =0.71, p-value =0.001). The shift in the relation between TOWE and discharge at Sheep Creek occurs during the period when the groundwater level in the two near-marsh wells (KLAM 637, KLAM 11774) declined below the land surface of the marsh (figs. 5 and 13). The correlation with TOWE and discharge at the Klamath Agency streamgage is marginally weaker during the first half (Spearman’s Rho =0.69, p-value =0.003) of the study period compared to the second half (Spearman’s Rho =0.72, p-value =0.001). The decline in groundwater level resulted in an increased thickness of the unsaturated zone, providing greater storage capacity for infiltrating stream water beneath and near the marsh and ultimately reducing the extent of open water within Klamath Marsh. Although an increased fraction of the springtime freshet is potentially infiltrating the subsurface, this additional recharge is still less than the amount needed to stop or reverse the groundwater-level declines observed at all wells within the study area.

Despite the groundwater level having been below the marsh land surface continuously since about 2005, the area of open water in Klamath Marsh has expanded greatly during years when the springtime freshet was large, such as in 2006, 2011, and 2017. Presumably, the rate of the inflowing surface water exceeds the infiltration rate through the surficial layer of peat and silt. Although the area of open water in the marsh still expands in response to years of large streamflow, the area of open water in years of less-abundant inflow has declined since about 2005.

Declining groundwater level in the study area likely is responsible for the decrease in flow at the Klamath Agency streamgage over the previous 20 years (fig. 15). Since WY 2000, discharge at this streamgage has been steadily decreasing, with peaks during years of abundant streamflow and precipitation, such as in WYs 2006, 2011, and 2017. Historically, discharge at this streamgage commonly was seasonal with flows declining to zero during the late summer. During WYs 1955–2000, the median number of dry days per year was 73; since WY 2000, the median number of dry days per year was 240. From WY 1955 (the beginning of the record) through WY 2014, this site had not remained dry for an entire year; however, starting in WY 2015 it has remained dry for 3 years, including WYs 2021 and 2022.

Declining streamflow at the Klamath Agency streamgage since 2000 is likely due to the declining groundwater level in the study area, and in particular, the level in the lowlands directly underlying and surrounding Klamath Marsh. Trends in precipitation (fig. 8; table 8) since WY 2001 were not statistically significant and trends in streamflow at the Sheep Creek streamgage (fig. 4; table 3) were of substantially smaller magnitude than trends at the Klamath Agency streamgage. As groundwater level has declined during the previous 2 decades, increasing volumes of surface-water inflow have been infiltrating into the shallow groundwater system leaving less (or no) water available for streamflow through the Williamson River channel downstream from the marsh.

Declining groundwater level is the principal factor influencing changes in the seasonal surface-water expression (TOWE) and surface-water outflow from Klamath Marsh. Rates of decline in groundwater level is similar across the entire study area, including areas far upgradient from Klamath Marsh (figs. 1 and 5). Declines in the surface-water extent of the marsh are likely related to these regional changes in groundwater level rather than to changes in water management and water use in the study area. An assessment of groundwater right permits obtained from OWRD WRIS (Oregon Water Resources Department, 2023c) in the study area supports this conclusion.

As of WY 2021, permitted groundwater withdrawals in the study area totaled 48,082 acre-ft/yr, with 9,190 acre-ft/yr (19 percent) permitted during the study period of WYs 1985–2021 and the latest permit issued in WY 2004 (table 6). Of the pumpage permitted during the study period, 3,739 acre-ft/yr were for primary irrigation and were likely used to their full extent each year. This recent primary irrigation pumpage represents 12 percent of the total permitted pumpage for primary irrigation (30,238 acre-ft/yr) in the study area. Of all permitted groundwater withdrawals during the 1985–2021 study period, 62 percent is associated with one well located along the Williamson River between the marsh and the Sheep Creek streamgage. This permitted pumpage is for supplemental irrigation water and the reported water use for this well indicated that it was only used in 5 years during 2003–22 and was pumped at a maximum rate of 459 acre-ft/yr (Oregon Water Resources Department, 2023c). Considering the actual use of this one large permit, new groundwater withdrawals during the study period were likely less than 4,600 acre-ft/yr. Groundwater-level data from KLAM 522 indicate that groundwater level in the study area fluctuated during 1962–87, but there was no indication of a sustained decline (fig. 5). The relatively small amount of additional pumpage permitted after 1985 is likely not responsible for the sustained declines in groundwater level observed in measurements from multiple wells across the study area since 2000.

One water right transfer was identified within the study area during the study period. The transfer enabled the owner to withdraw water from a second well of similar depth less than 1 mi from the original well and was associated with the installation of several center pivot sprinkler systems west of the marsh sometime during August 2012–July 2014 (based on high-resolution aerial imagery). Although conspicuous on the landscape, the sprinkler systems use groundwater that has been permitted for use since the early 1960s and that water is being withdrawn near the originally permitted well.

The hydrologic response of Klamath Marsh is seemingly at odds with the hydrologic input to the marsh. Changes to the water budget of the marsh were attributed to two potential influences: increasing air temperatures and the reestablishment of forested land within the study area. Since 2000, the extent of TOWE, groundwater level, and streamflow at Klamath Agency streamgage have all declined, whereas no trend in total annual precipitation was observed at any climate station within and adjacent to the study area. Streamflow entering the marsh from the Williamson River has had significant declines during all three analysis periods, WYs 1985–2021, 1991–2021, and 2001–21, with a somewhat pronounced decline during WYs 2019–21.

The MSWE data document declining surface-water inundation within Klamath Marsh since about 2000. Although no trend in the total annual precipitation at climate stations in and near the study area was detected, groundwater level near and surface-water inflows to the marsh declined during the study period. Increasing mean annual temperatures could indicate that evaporative demand is increasing and potentially capturing a part of water that otherwise would recharge the groundwater system. This initial assessment of available climatic and hydrologic data indicated that surface-water extent within Klamath Marsh is a complex issue caused by a combination of factors, and that new and continued monitoring would be useful for tracking and understanding future changes.