The purpose of this report is to aid the BLM and other agencies in (1) understanding the current distribution of streamflow within the study area, in particular, streamflow diverted into irrigation ditches and (2) determining how surface water extents will change under the proposed restoration. In order to help answer these questions, the study includes measurement of streamflow in natural channels and irrigation ditches as well as evaluation of surface water extent and hydraulic characteristics using a 2-dimensional hydraulic model of the proposed restoration area.

The spring complex is located at the north end of the Black Rock Desert–High Rock Canyon Emigrant Trails National Conservation Area (NCA). The Soldier Meadows ACEC is only a small part of the much larger NCA but covers almost the entire study area. The complex consists of dozens of springs and seeps, some isolated, others clustered in groups. In some places, spring flow merges to form stream channels that flow downslope to the valley bottom and eventually to Mud Meadow Reservoir. The study area is a small section of the spring complex (fig. 2) that contains the largest contiguous desert dace habitat and is a subsection of area 4 as defined by Vinyard (1996).

Stream channels, irrigation ditches, and emergent wetlands occur within the study area. Many locations with water have very dense emergent wetland vegetation dominated by hardstem bulrush (Schoenoplectus acutus) and other wetland plants. Heavy thatch from the bulrush means that the true spatial extent of surface water is difficult to discern. Stream channels tend to be more open, especially in areas with gravel or cobble substrate, however, in many locations, vegetation forms a closed canopy over the stream channel water surface (fig. 3). Hubbs and Miller (1948) as well as Nyquist (1963) both provide early descriptions of vegetation and channel shape in the spring complex that are very similar to current conditions.

Streamflow in the study area is complex, and gaining a better understanding of streamflow distribution is part of this study. In August 2019, streamflow discharged from wetland 1 in irrigation ditch 1 and flowed east (figs. 2 and 4 in blue). The natural stream channel (fig. 4 in red) that goes southeast from wetland 1 was completely dewatered by diversion into irrigation ditch 1. After about 575 ft, near the intersection with wetland 2, almost all of the streamflow discharged from irrigation ditch 1 and flowed southeast into an incised channel that returns flow to the natural channel. Irrigation ditch 1 continued east from the incised channel but was dry within several hundred feet. The streamflow in the incised channel flowed back to the natural channel just downstream of the first road crossing. The natural channel then flowed to the southeast for about 2,000 ft, where irrigation ditch 2 then diverted a small amount of streamflow to the east. The remaining streamflow in the natural channel flowed downstream for about 2,800 ft to another road and a fish barrier/culvert, where irrigation ditches 3 and 4 diverted a small amount of streamflow to the north and south. Finally, the remaining streamflow in the natural channel entered a wetland that is beyond the extent of the current restoration and study area.

During the field survey within the dewatered natural channel, several features were found illustrating the consequences of the dewatering. Parts of the channel had dry, friable root masses (fig. 5). The remaining wetland plants were stunted, and the channel was being colonized by rabbitbrush (Chrysothamnus sp.), cheatgrass (Bromus tectorum), and sagebrush (Artemisia sp.), indicating a transition from wetland to phreatophytic-upland vegetation that has been observed in other Great Basin springs as a result of drying (Patten and others, 2008). The last occurrence of streamflow in the dewatered channel is unknown; however, satellite images from 1981 and 1994 both show much denser vegetation than what is currently present, perhaps indicating some streamflow may have occurred in the dewatered channel during those periods.

The BLM is proposing to restore streamflow to the natural channel through a sequence of actions, from upstream to downstream areas. At the upstream end of the study area, irrigation ditch 1 would be filled for about 200 ft with the goal of returning streamflow to the currently dewatered natural channel. The incised channel would be completely filled for its entire length from irrigation ditch 1 to the natural channel. Sediment that has filled in under two bridges would be removed to allow streamflow restored to the previously dewatered natural channel to flow through. From the natural channel downstream for about 65 ft, irrigation ditch 2 would be filled in. Finally, at the downstream end of the study area, the first 200 ft of irrigation ditches 3 and 4 downstream from the natural channel would be filled in.

Desert dace have been studied by several researchers, primarily focusing on fish biology and ecology, with less emphasis on the hydrology and geomorphology of the Soldier Meadows spring complex (U.S. Fish and Wildlife Service, 1997). Measurements in 1987 (Vinyard, 1988) found most of the streamflow continued southeast in the natural channel past irrigation ditch 1 (whether through the incised channel or the currently dewatered natural channel isn’t clear), with an estimated mean streamflow of 2.09 cubic feet per second (ft³/s). Farther downstream, only about 0.53 ft³/s continued past irrigation ditch 2 in the natural channel. In 1989, irrigation ditch 1 was re-dredged and nearly all the streamflow was diverted from the natural channel into the ditch, with no streamflow reaching the downstream wetland 3 by the natural channel. Vinyard (1996) found fish catches dropped about 70 percent from May 1988 (pre-dredging) to October 1989 (post-dredging) in the study area. Even 6 years after the dredging, 81 percent of the total streamflow was still diverted into irrigation ditch 1, and fish catches were still about 50 percent less than before dredging. Fish trapping occurred again in 2002–03 and from 2014 to present using different sampling methods (Rissler and Scoppettone, 2004; Bauman, 2019). Trapping in the study area found a 333-percent increase in fish catches from 2002–2003 to 2014, but then a 52-percent decrease from 2014 to 2019 (Bauman, 2019).

The U.S. Geological Survey (USGS) made discrete streamflow measurements in 1980 at a number of sites and in 2013 at one site in the spring complex. Those measurements represent discharge at springheads and not the stream channels and irrigation ditches that are the main focus of this study. Continuous streamflow data for South Antelope Spring (USGS gage 412537119094001), just north of the spring complex, show no variation in streamflow over the period of operation, from November 2012 to June 2014 (U.S. Geological Survey, 2019a). This gage, however, was located directly at a springhead where the effect of evapotranspiration on stream discharge would be limited as a result of limited vegetation. Farther downstream, evapotranspiration from vegetation could result in more variable streamflow. The collection of additional USGS discrete streamflow measurements began in August 2019 at several sites in the spring complex; however, only limited data were available for use in this report.

In order to provide streamflow data for the hydraulic model and quantify the amount of streamflow in irrigation ditches, 11 streamflow measurements were made at 10 locations in the study area (table 1). Measurements were made with either an acoustic Doppler velocimeter or portable Parshall flumes (3-inch or 1-inch) following the procedures detailed in Turnipseed and Sauer (2010) and are published in the USGS National Water Information System (U.S. Geological Survey, 2019b).

Streamflow is diverted from the main channel to irrigation ditch 1 below the incised channel and also from the main channel to irrigation ditches 2–4. Because of the small quantity of streamflow diverted to the irrigation ditches and the natural stream channel shape (mostly deep, narrow, and several tenths of feet below the surrounding land surface), the increase in habitat extent likely would be minimal if the irrigation ditches were to be filled in. To quantify changes to channel depth, streamflow measurements were adjusted to account for the increase in discharge that would result if the irrigation ditches were filled. The measured streamflow (Q_m) at natural channel locations can be described as: Q m = A m V m = D m W m V m (1)where

The estimated streamflow (Q_e) at the same natural channel location with the irrigation ditches filled in is the measured streamflow of the irrigation ditches (Q_ID) added to the measured streamflow of the channel location (Q_m):

Because the stream channels are confined within nearly vertical walls, the additional streamflow would only increase the stream channel depth, not the width. Additionally, because the amount of additional streamflow is so small, the change in mean channel velocity is thought to be essentially zero. Therefore, the only variable changing is the mean depth (D_e): Q e = D e W m V m (3)which can be rearranged to solve for mean depth as

In August 13–15, 2019, Global Navigation Satellite System (GNSS) surveys were completed to determine locations of bridges and a culvert, as well as topographic points such as dry and flowing stream channel locations, streamflow measurements sites, and possible high-water marks. The GNSS surveys consisted of single-baseline Online Positioning User Service-Static (OPUS-S) surveys at one location for survey control, and single-base, real-time kinematic (RTK) GNSS surveys using the methods outlined in Rydlund and Densmore (2012). Owing to the distance to published benchmarks, OPUS-S surveys were completed on a temporary benchmark near the downstream end of the model area, to level 1 quality (0.18 ft elevation uncertainty). During the RTK surveys, points were collected at 1-second intervals for 3 minutes (180 epochs) at survey control, blunder checks, bridges, and culvert locations. Topographic points were collected at 1-second intervals for 5 seconds (5 epochs). The topographic points include 14 locations within the flowing natural stream channel at the downstream end of the model reach, where the channel bottom was surveyed, and the water depth was measured to determine a water surface elevation for use in model calibration (fig. 6). Both OPUS-S and RTK GNSS surveys were completed using Leica GS14 GNSS receivers. All RTK points were of level 4 quality (>0.32 ft elevation uncertainty).

In April 2018, aerial light detection and ranging (lidar) data were acquired in the spring complex for the BLM by Quantum Spatial, Inc. The lidar data had a vertical accuracy of 0.12 ft at the 95-percent confidence level for non-vegetated areas (root mean square error [RMSE] of 0.06 ft) and of 0.57 ft at a 95-percent confidence level for the vegetated areas (RMSE of 0.29 ft). These values are within the accuracy of quality level one lidar data (U.S. Geological Survey, 2019c). To verify the suitability of combining the lidar data with the GNSS surveyed data, 265 topographic points were surveyed during the GNSS survey in both lightly and densely vegetated areas. The vertical accuracy at the 95-percent confidence level between the GNSS and nearest lidar-derived topographic points was acceptable at 0.19 ft (RMSE of 0.16 ft). The filtered ground points from the lidar and GNSS topographic points were combined to create a 1-ft triangulated irregular network (TIN) of the model area. Topographic data were released as part of a USGS data release (Morris, 2021).

The primary data input in the model was the 1-ft TIN, and to simulate the proposed restoration scenario, several adjustments were made to the TIN. Near the upstream end of the model area, the elevations of the 200 ft of irrigation ditch 1 were replaced with surrounding land surface elevations, to represent filling in of the ditch. The elevations of the 60 ft of incised channel near the downstream end were similarly replaced with surrounding land surface elevations. The proposed restoration action calls for the entire length of the incised channel to be filled, but only the downstream end of the incised channel affects the model; therefore, only the downstream end of the incised channel was replaced with surrounding land surface elevations.

Currently (2019), the streamflow in the incised channel goes under the road at a cattle guard that has been placed on concrete footers, forming a small bridge over the channel. Two more cattle guard bridges are present on the road at the dewatered natural channel; however, both were filled with sediment. Future restoration efforts would remove sediment under the bridges to allow streamflow to continue downstream. Consequently, an estimated 0.1–2 ft of elevation was subtracted from the land surface elevations to simulate this sediment removal and was extended about 10 ft upstream and downstream of the bridges for grading to smooth the transition. Finally, at present, there is a large 1–2 ft elevation drop from the dewatered natural channel into the watered natural channel at the junction with the incised channel. This drop would act as a knickpoint, possibly leading to habitat fragmentation and new channel incision. A 20-ft section of elevation was graded to smooth the transition from dewatered to watered natural channel. Because the streamflow was modeled to never reach the bottom of the bridge decks, the bridges were conceptualized as concrete box culverts with natural channel bottoms in the model. All the locations of topographic adjustments are shown in figure 6.

The downstream boundary condition was set at the surveyed water surface of 4,444.08 ft because the downstream end of the model area was not expected to change with restoration. The upstream boundary condition was set at the measured streamflow of 1.9 ft³/s and was held steady over the simulated period of 5 hours to ensure the streamflow fully equilibrated in the model.

In addition to the topography of the land surface, the water depth and velocity are also controlled by friction with vegetation and channel material. This roughness, defined by Manning’s n-value, was computed at five of the streamflow measurement locations. The methods described in Jarrett and Petsch (1985) were used to compute the roughness coefficient by rearranging Manning’s equation: n= 1.486 Q A R 2/3 S 1/2 (5)where

Estimates of roughness coefficient are presented in table 1. The computed roughness coefficients ranged from 0.066 to 0.139.

Measurement sites were in locations of well-defined channels and where vegetation was at a minimum; therefore, the computed roughness coefficients are less applicable to areas of shallow water depths and dense vegetation that occurs in the center and upper ends of the modeled area. Roughness coefficients at these locations were set to 0.25 based on field observations and estimated using similar sites with dense bulrush and cattails (Hall and Freeman, 1994; Soong and others, 2012). In the middle and lower sections of the modeled area, channels were narrower and more similar to measurement sites, and roughness coefficients were set to between 0.07 and 0.1. Finally, although upland locations were not expected to be inundated, roughness coefficients of 0.2 were assigned to areas of very shallow depths in open sagebrush (Weltz and others, 1992). The specific response of vegetation production and resulting density to water being returned to the dewatered natural channel is unknown. The roughness coefficients for the currently dewatered natural channels are only estimates within a range of possible values.

Model calibration was accomplished by (1) developing a model geometry that represents the proposed restoration design and (2) iteratively adjusting computed roughness coefficients to minimize differences between simulated and measured water surface elevations and streamflow velocities. The model geometry included the following primary changes to the 1-ft TIN that represents the current (2019) land surface:

`Roughness coefficients were iteratively adjusted during the calibration process within a range of computed values to provide an additional constraint on minimizing differences in simulated and measured water-surface elevations and streamflow velocity.

The dataset used for model calibration included fourteen water-surface elevation measurements and a mean streamflow velocity at downstream locations of the natural channel and model area (fig. 6). All water surface and velocity measurements are for August 2019, and the mean streamflow velocity represents two measurements at the SMDD natural channel location (table 2). Calibration datasets were limited to surface-water data for unaltered areas of the hydraulic model geometry; consequently, water-surface elevation and streamflow velocity measurements for irrigation ditch 1 and the incised channel (fig. 2; table 1) were not used for model calibration.

The calibrated model simulated surveyed surface water elevations by an average of 0.10 ft too high (fig. 7), indicating roughness coefficients might be too large, whereas simulated mean velocity was 0.17 feet per second (ft/s) too large, indicating roughness coefficients might be too small. Examining the simulated versus surveyed water surface elevations, the largest differences were located near the center of the natural channel reach used for calibration, near a channel bend, where five simulated water surface elevations were on average 0.18 ft too high (fig. 7). Moreover, there are numerous locations of undercut banks in the study area where the TIN may be incorrect. The undercut banks in this section of the natural channel may be the cause of the larger model calibration error in water surface elevations. Both upstream and downstream of this section of the natural channel, the difference between simulated and measured water surface elevations was only 0.01 ft and 0.05 ft, respectively, indicating good calibration elsewhere in the natural channel.

In the upstream reaches of the dewatered natural channel (fig. 5), sections of dry, friable root masses and stunted wetland plants were found that may indicate high-water marks and a total of 37 of these points were surveyed. Results of the model calibration indicate that the final simulated water surface had a mean difference from the surveyed possible high-water marks of –0.06 ft. Surveyed elevations of possible high-water marks were not used as a dataset for model calibration because of uncertainty on the magnitude of stream discharge for past flows and on whether the roots were submerged or not; however, the surveyed points may provide a general idea of past water surface elevations.

The results of the study have several limitations and uncertainties. Calibration data were limited to surface water elevations and streamflow velocity for downstream natural channel reaches that would remain unaltered for the proposed restoration design. The specific response of vegetation production and resulting density to water being returned to the natural channel also is unknown for the hydraulic model. Elsewhere in the spring complex, vegetation grows densely in locations with water, but the density of the vegetation is variable. If vegetation were to regrow more densely than estimated in the model, the roughness coefficients would increase and the conveyance would decrease, resulting in larger surface water depths, a larger spatial extent of streamflow, and lower streamflow velocities. Conversely, if vegetation were to regrow less densely than estimated in the model, water depths would be smaller, the spatial extent of streamflow smaller, and the velocities higher. Additionally, vegetation regrowth would occur over time, so water depths and spatial extent of water would increase over the regrowth period.

Several sections of the model used estimated channel modifications (removal and addition of channel material). The actual change in the channel will likely be different than specified in the model and could result in differences in simulated surface water elevation, velocity, and spatial extent.

Model simulations assume that streamflow is steady and does not vary seasonally. This assumption is supported by a 1.5-year continuous discharge record from nearby South Antelope Spring in the Soldier Meadows spring complex (U.S. Geological Survey, 2019a). However, some seasonal variation may occur to the steady discharge assumption used in the model. For example, streamflow measurements at SMDD Natural Channel made on January 2020 revealed a 22-percent increase in streamflow from the August 2019 measurements, whereas in April 2020, streamflow was only increased about 5 percent above the August 2019 measurements. January and April are typically periods of low evapotranspiration, therefore the additional streamflow measured in January compared to April is thought to be related to recent snowmelt or rainfall. The January streamflow measurements suggest that snowmelt and or rainfall during these months can increase streamflow and represent a seasonal maximum streamflow. Streamflow measurements used in this study to determine an average steady discharge of 1.9 ft³/s were made in August, the driest month of the year and at near maximum evapotranspiration conditions (National Oceanic and Atmospheric Administration, 2020), and are considered a seasonal minimum streamflow. Therefore, the simulation represents a seasonal minimum of surface water extent, streamflow channel depth, and streamflow velocity.

Additional streamflow measurements over time would improve understanding of variations in streamflow and thus the variability in restored channel conditions and desert dace habitats. Finally, since streamflow is relatively small and most channels are narrow, measurements are generally considered poor. Small variations of measured streamflow during this study probably reflect uncertainty in measurements rather than actual changes in streamflow.