Background

Rivers designated as “wild,” “scenic,” or “recreational” (hereinafter referred to as Wild and Scenic Rivers or WSR) may be free of or have very few impoundments or diversions, have varying levels of accessibility and shoreline development depending on the classification, and have one or more ORVs. Although none of the WSR segments are officially designated as “scenic,” every segment in the Owyhee Canyonlands Wilderness is classified with scenic, fish, and wildlife ORVs. Following the passing of the Omnibus Public Lands Management Act, BLM published the Final Management Plan and Environmental Assessment (hereinafter referred to as the wilderness management plan), which addresses future management of ORVs within the Owyhee Canyonlands Wilderness (Bureau of Land Management, 2015) in accordance with the WSR Act.

Streamflow statistics are often used for various water-resource related studies including emergency planning, water-resource management, road infrastructure designs, and for regulatory purposes that describe high and low streamflows. The wilderness management plan describes the importance of high and low streamflows for maintaining ORVs. The BLM commonly uses the bankfull streamflow statistics (hereinafter referred to as annual streamflow statistics) and the 20-, 50-, and 80-percent (hereinafter referred to as Q₂₀, Q₅₀, Q₈₀, respectively) exceedance probability semimonthly daily mean streamflow statistics (hereinafter collectively referred to as semimonthly streamflow statistics) to file water right claims necessary for sustaining ORVs in the Owyhee Canyonlands Wilderness. Semimonthly streamflow statistics refer to streamflows that are equaled or exceeded 20-, 50-, and 80-percent of the time during any given semimonthly period. High streamflow periods, approximately equal to the semimonthly Q_20, are necessary for recreational opportunities such as rafting, canoeing, and kayaking during the spring runoff. Median semimonthly streamflows (Q₅₀) are commonly used to determine whether streamflow is greater than, equal to, or less than the median streamflow conditions. Low streamflow periods, approximately equal to the semimonthly Q_80, are critical for sustaining habitat necessary for aquatic species survivability.

Ries and others (2004) developed the U.S. Geological Survey (USGS) StreamStats program in Idaho to estimate streamflow statistics at ungaged locations throughout the State. Basin characteristics for the Owyhee Canyonlands Wilderness are generalized by StreamStats. As a result, the region is poorly represented by regionalized streamflow regressions in StreamStats. Wood and Fosness (2013) and Wood (2014) developed streamflow regressions and statistics specific to the Owyhee Canyonlands Wilderness after determining StreamStats generated inaccurate statistics using regressions in the Intermountain Semidesert region occupied by the Owyhee Canyonlands Wilderness. In an effort to update the statewide streamflow statistic estimates, Wood and others (2016) used additional streamflow data and new statistical methods to update regional regressions in StreamStats, which did not include the previously developed regressions for the Owyhee Canyonlands Wilderness.

Beginning in water year 2013, the USGS collected streamflow data throughout the Owyhee Canyonlands Wilderness to improve streamflow statistics previously estimated in Wood and Fosness (2013) and Wood (2014), which were later used to file water rights claims for WSR segments. Because of the inherent difficulties with operating and maintaining real-time streamgages in geographically remote regions like the Owyhee Canyonlands Wilderness, only 1 year of streamflow measurements were made as part of Wood and Fosness (2013) and 2 years of streamflow measurements in Wood (2014). Historical streamflow data, where available, were compiled and used with recent streamflow measurements made since water year 2012 to update streamflow statistics from previous studies (Idaho Power Company, 2021; U.S. Geological Survey, 2022).

Purpose and Scope

In 2013, the U.S. Geological Survey (USGS), in cooperation with BLM, continued work in the Owyhee Canyonlands Wilderness in southwestern Idaho to improve annual and semimonthly streamflow statistics used to define the streamflows required to help sustain ORVs for rivers designated “wild,” “scenic,” or “recreational.” The streamflow statistics may support the Bureau of Land Management’s effort to develop and file a federal water rights claim for sustaining outstanding remarkable values for the rivers. The purpose of this study is to provide annual and semimonthly streamflow statistics using updated streamflow regressions. The streamflow regressions developed in this study were updated using streamflow data collected since water year 2013 and other historical data collected before this study began. This report presents the background of the study, methods used to develop streamflow regressions that were used to update annual and semimonthly streamflow statistics, an evaluation of regression performance, and a comparison with previously published statistics for WSR segments in the Owyhee Canyonlands Wilderness.

Description of Study Area

The Owyhee Canyonlands Wilderness comprises 16 WSR segments spanning 325 mi across six Wilderness Areas managed by the BLM. The wilderness areas encompass 517,000 acres with six main river corridors including the North Fork Owyhee River, South Fork Owyhee River, East Fork Owyhee River, Big Jacks Creek, Little Jacks Creek, Bruneau River, Jarbidge River, and Pole Creek (not discussed in this report). Streamflow regressions and statistics were determined at 11 locations along WSR segments in the Owyhee Canyonlands Wilderness (fig. 1). State and Federal owned land encompass the Owyhee Canyonlands Wilderness with neighboring parcels of land owned by private citizens and Native American Tribes (Bureau of Land Management, 2015).

The Owyhee Canyonlands Wilderness is characterized by deeply incised basaltic and rhyolitic canyons and open sagebrush-dominated highlands that make up the scenic ORVs in each WSR segment. Contributing to the recreational ORV associated with some WSR segments, the study area is pristine and void of human development providing excellent opportunities for hiking, camping, and other outdoor activities (Bureau of Land Management, 2015). Characterized by a generally semiarid and cool climate, summertime temperatures exceeding 100 °F are common. The general hydrology can be summarized by high streamflows usually occurring from rain-on-snow events that occur during March–June with timing dependent on the range of elevations in each individual drainage basin. Minimum streamflows generally occur during July–February.

Each WSR segment is characterized by fish and wildlife ORVs. The Owyhee Canyonlands Wilderness provides habitat for many biotic communities including the greater sage grouse (Centrocercus urophasianus), California bighorn sheep (Ovis canadensis californiana), redband trout (Oncorhynchus mykiss), bull trout (Salvelinus confluentus), and Bruneau hot springsnail (Pyrgulopsis bruneauensis). Numerous WSR segments have provided essential resources like access to permanent water, fuel, and other animal and vegetative materials for the Shoshone-Paiute people for countless generations (Bureau of Land Management, 2015). A complete description of the study area including topography, land use, and hydrology is provided in Wood and Fosness (2013).

Streamflow Data Collection

Real-Time Streamgages

Streamflow data from real-time streamgages throughout the region were used to develop synthetic streamflow records at partial-record sites. Currently (2022), the USGS operates six real-time streamgages upstream and downstream of WSR segments and one (USGS 10329500) in an adjacent basin. In addition to USGS streamgages, three Idaho Power Company (IPC) operated real-time streamgages and one U.S. Department of Agriculture (USDA) operated real-time streamgage in the region were evaluated as index sites, or locations where streamflow would be referenced to, in this study (fig. 1, table 2). Unlike partial-record sites, real-time streamgages have continuous streamflow records that vary in length. For example, the East Fork Owyhee River real-time streamgage (USGS 13176400) began operation at the start of this study to develop more accurate synthetic streamflow records at nearby partial-record sites in the East Fork Owyhee River Basin (figs. 1–2, table 2). Conversely, the period of record of the index site at Martin Creek near Paradise Valley, Nevada (USGS 10329500) dates back to 1921 (table 2). Using a developed stage-discharge rating, a continuous record of instantaneous and daily mean streamflow was available for each real-time USGS, IPC, and USDA streamgage (Pierson and others, 2001; Idaho Power Company, 2021; U.S. Geological Survey, 2022). Each USGS real-time streamgage is operated, maintained, and processed according to USGS policies and guidelines (Turnipseed and Sauer, 2010; Mueller and others, 2013).

Table 2.    

Streamgages evaluated as index sites for the development of synthetic streamflow records at partial-record sites, Owyhee Canyonlands Wilderness, southwestern Idaho, 2021.

[All USGS data collected for the study are available in the USGS National Water Information System (U.S. Geological Survey, 2022). Site No.: USGS and Idaho Power streamgages were used in this analysis. In some cases, IPC continued streamgage operation and may have the same site number as the previously operated USGS streamgage. Abbreviations: No., number; NAD 83, North American Datum of 1983; mm-yyyy, month-year; Qd_min, minimum daily mean streamflow measurement; Qd_max, maximum daily mean streamflow measurement; IPC, Idaho Power Company; USDA, U.S. Department of Agriculture; USGS, U.S. Geological Survey; ft³/s, cubic foot per second; mi², square mile]

Table 2.    Streamgages evaluated as index sites for the development of synthetic streamflow records at partial-record sites, Owyhee Canyonlands Wilderness, southwestern Idaho, 2021.

Site No. (fig. 1)	Site name	Latitude (NAD 83, decimal degrees)	Longitude (NAD 83, decimal degrees)	Drainage area (mi2)	Period of record used in analysis (mm-yyyy)	Qdmin (ft3/s)	Qdmax (ft3/s)
Succor Creek Basin
IPC 13172900	Succor Creek near Jordan Valley, Oregon	43.2289	–117.054	80.6	07-2007 to 10-2021	0	504
Reynolds Creek Basin
USDA 036	Reynolds Creek at Critical Zone Observatory outlet	43.2614	–116.753	92	01-1963 to 10-2021	0	3,850
Little Humboldt Creek Basin
USGS 10329500	Martin Creek near Paradise Valley, Nevada	41.5346	–117.418	175	10-1921 to 10-2021	2.00	2,500
Bruneau River and Jarbidge River Basins
USGS 13162225	Jarbidge River below Jarbidge, Nevada	41.8905	–115.429	30.6	04-1998 to 10-2021	1.00	633
1IPC 13162500	East Fork Jarbidge River near Three Creek, Idaho	42.0318	–115.368	83.4	10-1928 to 03-1933, 09-1953 to 09-1971, 03-2006 to 10-2021	2.28	1,120
USGS 13161500	Bruneau River at Rowland, Nevada	41.9332	–115.675	382	06-1913 to 09-1918, 10-1966 to 10-2021	1.70	2,070
1IPC 13167505	East Fork Bruneau River near Hot Spring, Idaho	42.5568	–115.511	502	08-1910 to 04-1915, 12-1948 to 09-1971, 03-2006 to 10-2021	0	791
USGS 13168500	Bruneau River near Hot Spring, Idaho	42.7711	–115.72	2,690	07-1909 to 02-1915, 10-1943 to 10-2021	21.0	6,470
Owyhee River Basin
USGS 13175100	Owyhee River near Mountain City, Nevada	41.8606	–115.988	391	04-1991 to 09-1995, 05-1997 to 10-2021	0.42	1,590
USGS 13176400	East Fork Owyhee River at Crutcher Crossing, Idaho	42.2614	–116.868	2,140	12-2012 to 10-2021	8.00	9,280
USGS 13181000	Owyhee River near Rome, Oregon	42.8664	–117.649	7,690	10-1949 to 10-2021	44.0	46,900

¹

Streamgage previously operated by USGS.

Statistical Analysis

All discrete streamflow measurements at partial-record sites and daily mean streamflow at evaluated index sites were compiled using the R programming language (R Core Team, 2020; De Cicco and others, 2021). Where available, historical data at partial-record sites dating back to 1961 were considered in this study (table 1). Duplicate measurements made for quality assurance and quality control collected on the same day were averaged and zero streamflow measurements were included. Conversely, measurements flagged for not meeting USGS quality standards were not included in this study. In total, 318 streamflow measurements were used to develop streamflow regressions and statistics at partial-record sites.

This study used three methods based on the Maintenance of Variance Extension, Type 1 (MOVE.1) regression technique (Hirsch, 1982) to develop an estimated, or synthetic, streamflow record at each partial-record site in the study area using one or two index sites. In some instances, field crews observed zero streamflow at partial-record sites, which required a unique logistic regression and MOVE.1 method to estimate the presence of streamflow.

Semimonthly Streamflow Statistics

Developing synthetic daily mean streamflow records using methods previously described provided the basis for calculating semimonthly streamflow statistics at each partial-record site in this study. Using the updated streamflow regressions, daily mean streamflow was estimated at each partial-record site for the entire period of record of the index site through the end of water year 2021 (table 2) resulting in a continuous synthetic daily mean streamflow record. The synthetic streamflow records were divided into semimonthly periods to calculate the streamflows corresponding to the Q₂₀, Q₅₀, and Q₈₀ exceedance probabilities where the first semimonthly period was defined as the 1st through the 14th and the second semimonthly period was defined as the 15th through the end of the month. For each semimonthly period, the daily mean streamflow data were ranked in descending order using the Weibull plotting position formula, a standard procedure in Bulletin 17C, the standard reference for determining flood frequencies in the United States (England and others, 2019; Helsel and others, 2020). Daily mean streamflow values corresponding to the Q₂₀, Q₅₀, and Q₈₀ exceedance probability (fig. 3) were extracted from the distribution. If the exceedance probability was not exactly equal to 20-, 50-, and 80-percent, linear interpolation between adjacent streamflow values was used to estimate exceedance probability streamflow. This same method was used to determine streamflows corresponding to the annual 20-, 50-, and 80-percent exceedance probability but instead of compiling semimonthly periods, the entire synthetic streamflow record was used.

Regression Equations for Estimating Synthetic Streamflow

Of the 11 real-time streamgages evaluated as possible index sites, 6 were selected for the development of streamflow regressions at partial-record sites. Each index site was selected based on goodness-of-fit statistics between measured and predicted streamflow values at the partial-record sites. Moreover, the proximity of index sites to the partial-record sites also was considered when selecting an index site for streamflow regressions. Final updated regression coefficients and statistical performance metrics are published in Ducar and Dudunake (2022).

Updated streamflow regression equations were developed using one index site with the Maintenance of Variance Extension, Type 1 (MOVE.1) regression technique to estimate synthetic streamflow at eight partial-record sites (table 3). Because zero streamflow was not observed at these partial-record sites, a logistic regression was not required. Although Martin Creek (USGS 10329500, fig. 1, table 2), a high-elevation basin located in the Santa Rosa Range of northern Nevada, was outside the drainage basin of other sites in this study, it was selected as an index site for Big Jacks Creek (USGS 13161930), Cottonwood Creek (USGS 13169420), and Little Jacks Creek (USGS 13169800) due to similar runoff characteristics. Despite estimated streamflow values at Cottonwood Creek tending to be overbiased and underbiased at Big Jacks Creek compared to the measured streamflow values, Martin Creek was the best available index site in the study area. Similar to Martin Creek, the Bruneau River (USGS 13161500) was selected as an index site for the North Fork Owyhee River (USGS 13177910), which was attributed to the general regression performance and similar basin drainage area between each site. The East Fork Jarbidge River (IPC 13162500), a real-time streamgage located upstream from the partial-record site at the Jarbidge River (USGS 13162670), was selected for estimating streamflow at the downstream partial-record site. The real-time streamgage at East Fork Owyhee River (USGS 13176400), was selected as the best index site (table 3) for partial-record sites located upstream at Battle Creek (USGS 13176195) and Deep Creek (USGS 13176305). Despite surface-water regulation and diversions upstream on the South Fork Owyhee River near Duck Valley Indian Reservation (fig. 1), the Owyhee River (USGS 13181000) index site best represented the hydrologic characteristics at the South Fork Owyhee River (USGS 13177845, table 3).

In the instance of the Bruneau River (USGS 13162050, fig. 1), using one index site overbiased or underbiased streamflows at the partial-record site; therefore, two index sites were selected to develop a synthetic streamflow record with reasonable performance metrics that minimized overall bias (table 4). For example, Bruneau River at Rowland, Nevada (USGS 13161500) estimated low streamflow values better at the Bruneau River above Jarbidge River (USGS 13162050) although the Bruneau River near Hot Spring, Idaho (USGS 13168500) estimated high streamflow values better; therefore, both sites were selected and the MOVE.1 regression equations were combined using a weighted average of the mean square error (Granato, 2009).

A combined logistic regression and MOVE.1 regression approach was used at partial-record sites where zero streamflow was observed to retain all records when using a log transformed MOVE.1 regression (table 5). To account for zero streamflow measurements at Sheep Creek (USGS 13164600), the probability of streamflow was determined using a logistic regression equation. Using the index site at Bruneau River at Rowland, Nevada (USGS 13161500) accurately predicted the presence of streamflow for all streamflow measurements at Sheep Creek (table 5). The Bruneau River near Hot Spring, Idaho (USGS 13168500) was selected as the index site when the logistic regression predicted streamflow at the Sheep Creek partial-record site. Similarly, Martin Creek (USGS 10329500) was used as the index site for the logistic regression, which accurately predicted the presence of streamflow for 63 percent of the measurements at Wickahoney Creek (USGS 13169400) (table 5). Additionally, the Martin Creek index site was used to predict the streamflow at Wickahoney Creek using the MOVE.1 regression technique.

Streamflow Statistics

The updated streamflow regressions at partial-record sites described in the previous section “Regression Equations for Estimating Synthetic Streamflow” are the basis for determining streamflow statistics used by BLM for maintaining ORVs in the Owyhee Canyonlands Wilderness. The final streamflow statistics determined in this study represent the best estimates of the Q₂₀, Q₅₀, and Q₈₀ for semimonthly and annual periods in addition to bankfull streamflow statistics using the available streamflow records at index and partial-record sites. Using additional streamflow measurements and improved regression techniques, these streamflow statistics are an improvement over regional statistics in StreamStats (Ries and others, 2017) and those previously developed in Wood and Fosness (2013) and Wood (2014). A compilation of all calculated semimonthly and annual streamflow statistics for each partial-record site are published in Ducar and Dudunake (2022).

Semimonthly streamflow statistics represent the range of geographic and hydrologic conditions in the Owyhee Canyonlands Wilderness observed during this study. The largest streamflows corresponding to the Q₂₀ in the Big Jacks Creek and Little Jacks Creek Basins generally occurred in the first semimonthly period of May and ranged from 9.52 ft³/s at Wickahoney Creek (USGS 13169400) to 38.5 ft³/s at Big Jacks Creek (USGS 13161930). Conversely, the lowest streamflow characterized by the Q₈₀ typically occurred in August in Big Jacks Creek, Little Jacks Creek, and Cottonwood Creek with streamflows as low as 0.75 ft³/s at Cottonwood Creek and zero streamflow at Wickahoney Creek for all semimonthly periods beginning in mid-June through mid-March.

Streamflow statistics for all semimonthly periods generally were higher at partial-record sites with larger drainage areas (for example, Bruneau River and Jarbidge River) than those with small drainage areas (for example, Cottonwood Creek, Wickahoney Creek, Little Jacks Creek). The largest Q₂₀ streamflows generally occurred from early May to mid-June at Sheep Creek (USGS 13164600), Bruneau River (USGS 13162050), and Jarbidge River (USGS 13162670) partial-record sites. The lowest streamflows characterized by the Q₈₀ typically occurred from early September to early October on the Bruneau and Jarbidge Rivers. Although the drainage area of Sheep Creek is greater than the Bruneau and Jarbidge Rivers partial-record sites, the calculated Q₈₀ for semimonthly periods from mid-July to the end of October was zero streamflow.

Streamflow statistics were updated at four partial-record sites in the Owyhee River Basin. With drainage areas from 98.5 square miles (mi²) at the North Fork Owyhee River (USGS 13177910) to 2,720 mi² at the South Fork Owyhee River (USGS 1377845), partial-record sites in the Owyhee River Basin represent a larger range in drainage areas than other basins in the study. Timing of the largest Q₂₀ streamflows occurred from mid-March to mid-April for the South Fork Owyhee River (USGS 1377845), Battle Creek (USGS 13176195), and Deep Creek (USGS 13176305). The largest Q₂₀ streamflow occurs during the first semimonthly period of May at the North Fork Owyhee River partial-record site. The lowest Q₈₀ streamflow statistic is greater than zero for all partial-record sites in the Owyhee River Basin and occurs in August or September.

The annual bankfull streamflow values at partial-record sites were calculated using updated streamflow regressions and measured peak streamflow data available at each index site (table 6). The bankfull streamflow values represent the best available estimates based on the currently (2022) available streamflow data and exemplify the range of hydrologic conditions within the Owyhee Canyonlands Wilderness. In general, estimated bankfull streamflow at partial-record sites increased with an increase in drainage areas. Exceptions to this included the North Fork Owyhee River (USGS 13177910) Basin, which is situated at a higher elevation than partial-record sites in the Big Jacks Creek and Little Jacks Creek Basins. Similarly, the estimated bankfull streamflows at the Jarbidge River (USGS 13162670) and Deep Creek (USGS 13176305) were higher than other partial-record sites with comparable drainage areas.

Mass Balance of Bruneau River and Jarbidge River Basins

The Bruneau River and Jarbidge River Basins provided a unique opportunity to evaluate the performance of updated streamflow regressions at partial-record sites using a mass-balance approach. The real-time streamgage at Bruneau River near Hot Spring, Idaho (USGS 13168500) provided a continuous record of daily mean streamflow draining four nested basins upstream in the Owyhee Canyonlands Wilderness. Streamflow at three partial-record sites—Bruneau River above Jarbidge River (USGS 13162050), Sheep Creek above mouth (USGS 13164600), Jarbidge River above mouth (USGS 13162670)—and one real-time streamgage at East Fork Bruneau River (IPC 13167505) were accumulated and compared to streamflow at the downstream real-time streamgage. The four nested basins have a combined drainage area of 2,129 mi² accounting for 79 percent of the drainage area of the real-time streamgage at Bruneau River near Hot Spring, Idaho. The unaccounted drainage area generally is low-lying and lacks any major tributaries. Assuming no streamflow was gained or lost downstream of the four basins, the sum of daily mean streamflow in the four nested basins was expected to be approximately equal to the daily mean discharge at the downstream real-time streamgage. This assumption allows for a comparison of streamflow magnitude at each partial-record site and the overall influence they have on the streamflow draining from the Owyhee Canyonlands Wilderness.

For water years 1966–2021 when daily mean streamflow data were available at all index sites, the summation of the nested basins overestimates the daily mean streamflow at the downstream real-time streamgage by an average of 21.4 percent. Overall, the streamflow mass balance performs well at representing streamflow during baseflow conditions and throughout the rising-limb and falling-limb of the hydrograph but usually overestimates streamflow during peak streamflow conditions (fig. 4). The overestimation may be attributed to uncertainty in the streamflow estimates from updated regressions at the upstream partial-record sites. Alternatively, the assumption that gains and losses do not exist in the nested basins may be invalid. Additional peak streamflow measurements at the partial-record sites could improve the overall regression performance, estimated streamflow, and the Bruneau River and Jarbidge River Basins streamflow mass balance. For example, 2,820 ft³/s was measured at Sheep Creek (USGS 13164600), the highest recorded streamflow measurement at that partial-record site, during a short-duration rain-on-snow event on February 9, 2017. The concurrent daily mean streamflow at the Bruneau River index site (USGS 13168500) was 2,990 ft³/s. Assuming the measurement at Sheep Creek represented the daily mean discharge, streamflow at Sheep Creek accounted for 94 percent of the streamflow measured at the index site. Peak streamflow was recorded at the index site (5,140 ft³/s) from the same runoff event but occurred 1 day after the Sheep Creek partial-record site. Based on the updated streamflow regression at Sheep Creek, the estimated streamflow was 7,890 ft³/s, or 154 percent of the streamflow at the index site. Due to the lack of additional streamflow measurements at Sheep Creek when the index site streamflow was at least 3,000 ft³/s or higher, the estimated streamflow at Sheep Creek accounts for more streamflow than what was measured at the downstream index site. Additional streamflow measurements at Sheep Creek during long-duration peak streamflow conditions at the index site may improve the overall streamflow estimates at Sheep Creek. This observation suggests the assumption that measurements at partial-record sites representing daily mean streamflow conditions is invalid and can underestimate or overestimate actual streamflow conditions at a partial-record site. Moreover, an estimate of the lag-time between the partial-record site and index site in the basin also could improve daily streamflow estimates and the mass-balance in the Bruneau River and Jarbidge River Basins. However, the remoteness and difficulty of collecting measurements at partial-record sites in the basin makes estimating lag-time costly. The mass-balance method demonstrates the validity of updated streamflow regressions in the Bruneau River and Jarbidge River Basins and the need to further improve regressions that have high uncertainty to estimate streamflow more accurately at ungaged sites.

PeakFQ and StreamStats Generated Bankfull Estimates

Long peak streamflow records at index sites and updated streamflow regressions provided a way to update annual streamflow statistics at partial-record sites in the Owyhee Canyonlands Wilderness. Bankfull streamflow statistics and confidence intervals at the index sites were computed using annual peak streamflow data in PeakFQ and compared to bankfull streamflow estimates and prediction intervals that were computed at each partial-record site using StreamStats peak-flow regional regression equations (Ries and others, 2017) (table 6).

The bankfull streamflow estimates generated using PeakFQ were better estimates than bankfull estimates from peak-flow regional regression equations, because the most current streamflow data were used from nearby streamgages and not regionalized. PeakFQ bankfull streamflow estimates at each partial-record site were within their computed upper and lower 90-percent confidence intervals but were outside of the 90-percent prediction intervals determined in peak-flow regional regression equations for Big Jacks Creek (USGS 13161930), Cottonwood Creek (USGS 13169420), Wickahoney Creek (USGS 13169400), Little Jacks Creek (USGS 13169800), Deep Creek (USGS 13176305), and South Fork Owyhee River (USGS 13177845). Confidence intervals for the peak-flow regional regression equations bankfull estimate at the South Fork Owyhee River could not be computed because the drainage area was outside the range for which regressions were developed; therefore, a comparison could not be made for the bankfull streamflow estimated using PeakFQ.

Bankfull streamflow estimates using peak-flow regional regression equations at the Big Jacks Creek (USGS 13161930) and Wickahoney Creek (USGS 13169400) partial-record sites (table 6) were higher than all streamflow measurements made at those sites except measurements that occurred during a low-duration high-magnitude rain-on-snow runoff event on February 10, 2017 (table 1). The bankfull streamflow estimate from peak-flow regional regression equations was higher than streamflow measurements collected during runoff events in an above normal water year (for example, 2017) at some partial-record sites, likely due to the poor performance of regionalized regressions in representing site conditions of small drainage areas. This suggested that the PeakFQ bankfull estimate was likely more representative of actual conditions at Big Jacks Creek and Wickahoney Creek. Similarly, the peak-flow regional regression equations bankfull estimate was higher than all streamflow measurements collected at the Cottonwood Creek partial-record site and, in turn, may suggest that PeakFQ bankfull streamflow estimates are more representative of conditions at the partial-record site (table 6). The peak-flow regional regression equations bankfull streamflow estimate also was higher than the PeakFQ estimate and all measurements at the Little Jacks Creek partial-record site. Altogether, each peak-flow regional regression equations bankfull streamflow estimate was higher than the PeakFQ bankfull streamflow estimate in the Big Jacks Creek and Little Jacks Creek Basins (table 6), which further supports the observation that site conditions are unique and may be poorly represented by peak-flow regional regression equations.

Peak-flow regional regression equations bankfull streamflow estimates for each partial-record site in the Bruneau River and Jarbidge River Basins were less than estimates from PeakFQ (table 6). Moreover, the peak-flow regional regression equations bankfull streamflow estimates ranged from 36 to 87 percent less than the maximum streamflow measurement at each partial-record site in the Bruneau River and Jarbidge River Basins. Similar to the Big Jacks Creek and Little Jacks Creek Basins, regional regressions provided in peak-flow regional regression equations do not accurately represent at-site conditions at the Bruneau River (USGS 13162050), Jarbidge River (USGS 13162670), or Sheep Creek (USGS 13164600) partial-record sites. PeakFQ bankfull streamflow estimates also were higher than all peak-flow regional regression equations estimates for partial-record sites in the Owyhee River Basin (table 6) and ranged from 52 to 75 percent less than the maximum streamflow measurement for each partial-record site in the Owyhee River Basin (table 1). This observation may suggest that peak-flow regional regression equations bankfull streamflow estimates may be less than actual bankfull conditions in the Owyhee River Basin and better estimated using PeakFQ.

Big Jacks Creek Basin

Wood and Fosness (2013) calculated semimonthly and bankfull statistics at Big Jacks Creek using a direct computation method at a discontinued streamgage (USGS 13169500) and 65 years of available streamflow data (1939–2004). Peak streamflow at the discontinued streamgage generally occurred in March before drying up for the remainder of the year. Conversely, streamflow measurements collected since 2013 at the partial-record site (USGS 13161930) located approximately 21 mi upstream from the discontinued streamgage indicate streamflow generally is perennial. Based on previous measurements, Big Jacks Creek experienced large streamflow losses between the wilderness boundary and the discontinued streamgage as Big Jacks Creek exits the canyonlands. The loss is explained by the ephemeral and perennial characteristics at the discontinued streamgage and upstream partial-record site, respectively.

The largest updated and previously published Q₂₀ streamflows were similar (35 and 38.5 ft³/s, respectively), but occurred nearly 2 months later using the updated regression (fig. 5A). The lag between Q₂₀ streamflow estimates was likely attributed to index site selection for the updated regressions. Martin Creek (USGS 10329500) is dominated by snowmelt runoff from higher elevations resulting in peak streamflow that occurs later in the water year relative to Big Jacks Creek. Overall, Martin Creek performed better at estimating streamflow at the Big Jacks Creek partial-record site than other index sites (table 3; R²=0.865, percent bias = –19.3 percent). However, establishing a new index site in closer proximity to the Big Jacks Creek partial-record site may improve streamflow estimates and semimonthly and annual statistics. The updated bankfull streamflow estimate using PeakFQ was 42 percent less than the value calculated in Wood and Fosness (2013); however, the values were not comparable because the location represented by bankfull streamflow estimates are 21 mi apart and have different periods of record. To that end, the updated bankfull streamflow estimate using PeakFQ is considered to represent the best available estimate at Big Jacks Creek (table 6).

Streamflow regressions and statistics at Cottonwood Creek (USGS 13169420), the smallest drainage area in the Big Jacks Creek Basin, were originally developed using the Bruneau River index site (USGS 13168500). Although the original streamflow regression performed well, it was developed using only four streamflow measurements at Cottonwood Creek. With 20 additional streamflow measurements at the partial-record site, the Martin Creek (USGS 10329500) index site performed nominally better (table 3) compared to the Bruneau River index site used in Wood and Fosness (2013). The updated streamflow regression is an improvement that is attributed to the selection of an index site with more similar basin and runoff characteristics. Additional streamflow data collection represented a wider range of hydrologic conditions that included a peak streamflow measurement in 2017, an above-average water year, and helped improve the overall regression performance. The updated streamflow regression generally performed poorly (R²=0.609) compared to the other updated regressions in this study, likely attributed to the spring-fed nature of Cottonwood Creek. The largest Q₂₀ streamflow occurred in May using both regressions but was 115 percent greater using the updated regression (9.4 ft³/s, fig. 5B). The estimated bankfull streamflow using PeakFQ at Cottonwood Creek (11.5 ft³/s) was 178-percent greater than the previously estimated bankfull streamflow (4.14 ft³/s) and is likely a result of additional peak streamflow measurements that improved the streamflow regression performance (table 6, fig. 5B). In general, the updated streamflow regressions and statistics are a slight improvement over estimates in Wood and Fosness (2013); however, establishing a new index site that is in closer proximity to the Cottonwood Creek partial-record site may improve streamflow estimates and semimonthly and annual statistics.

An improved regression technique was used to update the streamflow regression and statistics at the Wickahoney Creek partial-record site (USGS 13169400), which included a combined logistic regression and MOVE.1 method (table 4). Unlike the regression developed in Wood and Fosness (2013), which used a Tobit regression model and the Bruneau River index site (USGS 13168500), zero streamflow measurements at Wickahoney Creek were used in a logistic regression instead of censoring the values. Using a logistic regression and streamflow at the Martin Creek index site (USGS 10329500) allowed for the calculation of the probability of zero streamflow occurring at Wickahoney Creek. Additionally, when streamflow was predicted using the logistic regression, a MOVE.1 regression was used with streamflow at the Martin Creek index site to estimate streamflow at Wickahoney Creek. Because zero streamflow was measured for 41 percent of the total streamflow measurements at the partial-record site (table 4), using the combined logistic regression and MOVE.1 method performed better than the Tobit regression model and improved overall streamflow statistics from Wood and Fosness (2013). The largest updated and previously published Q₂₀ streamflows at Wickahoney Creek were similar (8.88 and 9.52 ft³/s, respectively) and both occurred in May. The updated bankfull statistic calculated using PeakFQ was 86 percent greater than the bankfull streamflow estimated in Wood and Fosness (2013) likely attributed to additional streamflow measurements (table 6, fig. 5C). The updated bankfull statistic calculated using PeakFQ was 86 percent greater than the bankfull streamflow estimated in Wood and Fosness (2013) likely attributed to additional streamflow measurements (table 6, fig. 5C). Similar to Big Jacks Creek, Cottonwood Creek, and Wickahoney Creek partial-record sites, an index site in the basin may improve the regression performance and streamflow estimates.

Bruneau River and Jarbidge River Basins

Wood and Fosness (2013) estimated semimonthly and annual streamflow statistics at the Bruneau River partial-record site (USGS 13162050) using a streamflow regression with the downstream real-time index site on the Bruneau River (USGS 13168500). The updated regression used the same index site as Wood and Fosness (2013) but also included an upstream index site on the Bruneau River (USGS 13161500) to better represent the basin characteristics of the partial-record site located at the intermediate location (table 4). The updated regression used 24 additional streamflow measurements, which encompassed the range of updated semimonthly streamflow statistics, unlike the previous estimates that extrapolated beyond the measured values. Additional peak streamflow measurements during above-average water years and measurements during semimonthly periods without measurements may further improve the streamflow statistics.

The largest previously estimated Q₂₀ streamflow from Wood and Fosness (2013) was 745 ft³/s, approximately 165-percent greater than the updated estimate of 464 ft³/s, and occurred one semimonthly period after the new estimate. Moreover, the updated bankfull streamflow estimate (412 ft³/s) was 34 percent less than the previously estimated bankfull streamflow estimate (table 6, fig. 6A). The large difference in estimates was likely attributed to the inclusion of more streamflow measurements in the updated regression, which corresponded well to measured values at the partial-record site (R²=0.981, table 4).

Similar to methods that Wood (2014) used to update the Jarbidge River streamflow regression, Wood and Fosness (2013) created a synthetic streamflow record at Sheep Creek using a drainage-area-ratio method after accumulating streamflows that were calculated for upstream tributaries. As a result, a final streamflow regression was not initially developed for Sheep Creek. As part of this study, additional streamflow measurements were collected at the Sheep Creek partial-record site (USGS 13164600) and upstream tributaries (USGS 13164000, USGS 13163200, and USGS 13163300) to develop a streamflow regression since streamflow statistics were originally estimated in Wood and Fosness (2013). Based on the new streamflow regression using two Bruneau River index sites (USGS 13161500, USGS 13168500), the largest Q₂₀ streamflow occurred in May (1,270 ft³/s) and was similar to the previously estimated value (1,080 ft³/s; fig. 6B). Although the streamflow accumulation method was used to assess the streamflow regression and quantify streamflow at the Sheep Creek partial-record site similar to the previous study, updated streamflow statistics for Sheep Creek represent the best available estimates based on the additional streamflow measurements. Bankfull streamflow was not easily estimated in the prior study because a synthetic record of annual peak streamflow was not calculated. Therefore, the peak-flow regional regression equations bankfull streamflow estimate (735 ft³/s) was considered the best available estimate at the time. However, this study established a streamflow regression used in conjunction with the peak streamflow record at the index sites to estimate a new bankfull streamflow (909 ft³/s) at the Sheep Creek partial-record site (table 6).

Instead of developing a regression to estimate streamflow statistics, Wood and Fosness (2013) directly computed statistics for the Jarbidge River using streamflow data at the real-time streamgage located approximately 40 river miles upstream of the WSR segment. Wood (2014) updated semimonthly and annual streamflow statistics using a drainage-area-ratio method to account for additional tributaries between the upstream real-time streamgage and the mouth of the Jarbidge River (USGS 13162670). To further update streamflow regressions and statistics, 31 additional streamflow measurements were made at the Jarbidge River partial-record site since the initial streamflow statistics were updated in Wood (2014). With additional measurements, a new regression was developed using the real-time index site on the East Fork Jarbidge River (IPC 13162500). Based on the new streamflow regression, the largest Q₂₀ streamflow occurred in June (700 ft³/s), one semimonthly period later than the previously estimated Q₂₀ (1,120 ft³/s, fig. 6C). The lag between updated and previously estimated Q₂₀ streamflows is likely attributed to the index site selection and additional streamflow measurements since Wood (2014) estimated streamflow in the Jarbidge River. Wood (2014) considered using the drainage-area-ratio method to calculate bankfull streamflow but concluded that the bankfull streamflow was best estimated using peak-flow regional regression equations (691 ft³/s) because estimating and adjusting annual peak streamflow statistics using streamgages with varying periods of record provided inaccurate estimates. Conversely, this study updated bankfull streamflow estimates using PeakFQ, which evaluated the peak streamflow record at the nearby index site on the East Fork Jarbidge River. The updated streamflow regression at the partial-record site was used with the peak streamflow record at the index site to calculate a new bankfull streamflow estimate (741 ft³/s) and was considered to represent the best available estimate of bankfull streamflow at the Jarbidge River partial-record site (table 6).

Little Jacks Creek Basin

At the time of this study (2022), streamflow was not monitored using a real-time streamgage in the Little Jacks Creek Basin. Streamflow regressions and statistics at Little Jacks Creek (USGS 13169800), were originally developed using the Martin Creek index site (USGS 10329500). Martin Creek was outside the study area but shared similar runoff characteristics as Little Jacks Creek. After testing all possible index sites (table 2), Martin Creek was selected as the index site and the regression was updated with additional streamflow measurements. Because the updated and previously published regression used the same index site, differences in the streamflow statistics are related to the inclusion of additional streamflow measurements for updating the streamflow regression. The maximum measured streamflow used to update the regression was about nine times larger than the maximum streamflow that Wood and Fosness (2013) used to develop the initial regression, which improved the overall regression performance for estimating high streamflows. The largest Q₂₀ streamflow occurred in May using both regressions but was 162-percent greater using the updated regression (12.2 ft³/s). The updated and previously developed regression estimated the lowest Q₂₀ streamflows at Little Jacks Creek from August to January (fig. 7).

Bankfull streamflow estimated using the developed regression at Little Jacks Creek was 85 percent lower than the previously estimated bankfull streamflow (fig. 7). The previous bankfull streamflow was generated using a discontinued streamgage located downstream (USGS 13170000), which operated from 1938 to 1949. Wood and Fosness (2013) considered the streamflow statistics at the upstream site more representative of conditions in the WSR segment because Little Jacks Creek was a perennial stream while streamflow at the discontinued streamgage was ephemeral. However, due to the lack of high streamflow measurements at the upstream site and 11-year peak streamflow period of record from direct and indirect measurements at the discontinued streamgage, Wood and Fosness (2013) considered the bankfull streamflow from the downstream site a reasonable estimate for the WSR segment. Estimating bankfull streamflow at the discontinued streamgage is reasonable because losses to groundwater were less during peak streamflow conditions relative to baseflow conditions. The bankfull streamflow estimate, 82.2 ft³/s, was considered the best available estimate because it was computed using peak streamflow measurements and not a regression developed with daily mean streamflow.

Owyhee River Basin

The previously developed streamflow regression at Deep Creek (USGS 13176305) used an index site (USGS 13161500) in the Bruneau River Basin that resulted in a poor correlation. Although Deep Creek and Battle Creek are relatively close, the downstream index site on the Owyhee River (USGS 13181000) was previously used for the Battle Creek (USGS 13176195) streamflow regression. Moreover, the measurements at Deep Creek and Battle Creek were originally collected in less than 1 year and included five streamflow measurements used to develop a stage-discharge rating that was further used to develop streamflow regressions. With additional streamflow measurements at Deep Creek and Battle Creek and a reevaluation of index site selection since 2012, streamflow regressions and statistics were updated using the downstream East Fork Owyhee River index site (USGS 13176400, table 3) for both partial-record sites. As previously described, the real-time streamgage on the East Fork Owyhee River is more representative of basin characteristics than the selected index sites in Wood and Fosness (2013) despite a lower R² compared to the previously developed regressions. The decrease in updated regression performance is likely attributed to the few corresponding measurements made at partial-record sites and the short period of record at the index site used to update the regressions (table 3).

The largest updated Q₂₀ streamflow estimate at the Battle Creek partial-record site (USGS 13176195) was 211 ft³/s, consistent with the previous estimate of 242 ft³/s and occurred in the same semimonthly period (fig. 8A). However, the 95-percent prediction interval ranged from 32.6 to 1,360 ft³/s as published in Ducar and Dudunake (2022) and would likely decrease using additional high streamflow measurements to update the streamflow regression. Similarly, the updated bankfull streamflow estimate (393 ft³/s) was consistent with the previous estimate (389 ft³/s; fig. 8A). The largest difference between previous and updated Q₂₀ streamflow estimates occurred at the Deep Creek partial-record site (USGS 13176305) where the largest Q₂₀ streamflow was 3,420 ft³/s and occurred in May while the updated estimate was 640 ft³/s and occurred in March (fig. 8B). The difference in Q₂₀ streamflow estimates is largely associated with index site selection and the length of the period of record at the index and partial-record site. The difference between the previous and updated Q₂₀ suggests that previous estimates were not accurately represented using the Bruneau River index site (USGS 13161500) for the streamflow regression. The updated bankfull streamflow estimate (1,618 ft³/s) was consistent with the previous estimate (1,650 ft³/s) (fig. 8B) despite using different index sites to develop streamflow regressions. The updated bankfull streamflow estimate at Deep Creek was about 15 percent greater than the maximum measured streamflow on March 22, 2019. The estimated bankfull streamflow and semimonthly statistics at Battle Creek and Deep Creek were computed using the most representative index site and, in turn, are considered the best available estimates. However, additional measurements capturing high streamflow events at each partial-record site would likely improve regression performance.

Streamflow at the North Fork and South Fork Owyhee River partial-record sites (USGS 13177910 and USGS 13177845, respectively) was estimated using updated streamflow regressions and the same index sites used to generate previously developed regressions (table 3). Streamflow in the East Fork Owyhee River is monitored using a real-time streamgage (USGS 13176400) as previously described. Together, the three forks of the Owyhee River represent about 64 percent of the total drainage area at the downstream Owyhee River index site (USGS 13181000) with the South Fork Owyhee River being the largest drainage area of all partial-record sites in this study.

In general, the updated streamflow regression for the South Fork Owyhee River partial-record site performed only slightly better than the previously developed streamflow regression despite 20 additional measurements. The largest updated Q₂₀ streamflow estimated at the South Fork Owyhee River partial-record site was 1,030 ft³/s, about 10 percent less than the previous estimate of 1,150 ft³/s and occurred during the same semimonthly period (fig. 8D). Conversely, the largest updated Q₂₀ streamflow estimate at the North Fork Owyhee River partial-record site was 390 ft³/s, approximately 30 percent greater than the previous estimate of 300 ft³/s, which also occurred during the same semimonthly period (fig. 8C). Although the updated and previously developed streamflow regressions both used the Bruneau River index site (USGS 13161500), the updated regression performance decreased slightly with additional streamflow measurements resulting in wide prediction intervals. Moreover, the largest updated Q₂₀ streamflow estimate was slightly larger than the highest measured streamflow value at the partial-record site suggesting the need for additional measurements during high streamflow conditions (fig. 8C). The updated bankfull streamflow estimate at the South Fork Owyhee River decreased about 17 percent to 1,510 ft³/s from the previous estimate and increased at the North Fork Owyhee River by 31 percent to 318 ft³/s. The changes in updated bankfull streamflow estimates were largely associated with the additional measurements at each partial-record site.