Purpose and Scope

The purpose of this report is to present the results of the cooperative study to (1) characterize streamflow conditions for Indian and Kill Creeks in Johnson County, Kans., (2) quantify hydromodification for the urbanized Indian Creek sites relative to the less disturbed Kill Creek site, and (3) identify reliable hydrological indicators that can be used to establish goals for streamflow restoration and assess effectiveness of management practices toward achieving more healthy stream ecosystems. Results presented in this report are intended to provide information to the Johnson County Stormwater Management Program, municipalities in Johnson County, and others to improve understanding of hydromodification in the county. Regionally and nationally, although the same set of streamflow indicators may not apply to other regions, methods and results presented in this report provide guidance, techniques, and perspective that can be useful for future related studies or similar studies elsewhere, particularly those designed to quantify hydromodification of urban streams and monitor the effectiveness of restoration efforts.

Description of Study Area

The study area consisted of the Indian and Kill Creek Basins in Johnson County, in east-central Kansas (fig. 1, table 1). Indian Creek is part of the Kansas City metropolitan area, which is north and east of Johnson County. The urbanization gradient generally decreases moving west, where Kill Creek is, in western Johnson County. Two study sites are on Indian Creek and one study site is on Kill Creek. The Indian Creek Basin (delimited using the USGS streamgage Indian Creek at State Line Road, Leawood, Kans. [06893390; hereafter referred to as “Leawood”]) is about 64 square miles (mi²), and the Indian Creek at Overland Park, Kans. (06893300; hereafter referred to as “Overland Park”), streamgage is centrally located in the basin and has a drainage area of 27 mi². Tomahawk Creek is a major tributary and flows into Indian Creek between the two streamgages. Two Indian Creeks sites were used in the study because one is the most downstream site and is representative of the full basin (Leawood) and the other site (Overland Park) has a longer period of record for streamflow data. Information from the Overland Park site provided additional historical flow context and could help identify streamflow characteristics in the upper basin that differed from the lower basin. The Kill Creek Basin (delimited using the USGS streamgage Kill Creek at 95th Street near DeSoto, Kans. [06892360; hereafter referred to as “Kill Creek”]) is about 53 mi² (fig. 1).

The Indian and Kill Creek Basins are physiographically located in the transition between the Dissected Till Plains and the Osage Plains sections of the Central Lowland Province (Schoewe, 1949). Basin topography typically ranges from nearly level to moderately sloping. Basin soil types include silt loams, clay loams, and silty clay loams that vary from somewhat poorly drained to well drained (Plinsky and others, 1979). The underlying bedrock mostly is limestone and shale of Pennsylvanian age (Kansas Geological Survey, 1964).

The Johnson County climate is characterized by well-defined seasons and variable precipitation. Long-term mean annual precipitation in the county was about 42 inches during 1981–2010 (High Plains Regional Climate Center, 2017). Most of the annual precipitation falls during the growing season (generally, April–September).

Land use differs substantially between the two basins. The Indian Creek Basin is primarily urban with at least 92-percent urban land use for each of the two Indian Creek sites. In contrast, 31 percent of the Kill Creek Basin is urban, and the remainder is predominantly grassland, woodland, and cropland (fig. 1; Homer and others, 2015). In addition, about 15 percent of the Kill Creek Basin includes land that was part of the Sunflower Army Ammunition Plant (fig. 1) where propellants were produced from 1942 to 1992 (Kansas Department of Health and Environment, 2005). Although most of the area is now covered with grassland, some overgrown roads and building foundations remain from the time the plant was in use. The former ammunition plant property is included in the Kill Creek urban land use estimate.

Streamflow in Indian Creek and Kill Creek is affected by wastewater discharges; however, effects are more pronounced at the Indian Creek sites because of the wastewater plant’s closer proximity to streamgages and substantially larger discharge capacity. The Johnson County Douglas L. Smith Middle Basin Plant (hereafter referred to as the “Middle Basin”) wastewater treatment facility (WWTF) is about 3.2 miles (mi) upstream from the Overland Park streamgage and about 8.1 mi upstream from the Leawood streamgage (fig. 1). The Middle Basin WWTF was originally constructed in 1979 with a maximum capacity of 9 million gallons per day (Mgal/d; 13.9 cubic feet per second [ft³/s]) and subsequently has undergone several upgrades that increased capacity. The most recent upgrade during 2007–10 increased average capacity from 12 Mgal/d (18.6 ft³/s) to 14.5 Mgal/d (22.4 ft³/s) (Johnson County, Kansas, 2018). After phase-in of the plant during the early 1980s, wastewater discharge has resulted in increased base flow throughout the year.

The Tomahawk Creek WWTF discharges into Indian Creek downstream from the Overland Park streamgage and 1.4 mi upstream from the Leawood streamgage (fig. 1). Originally constructed in 1955, the plant’s design capacity during the available period of streamflow record at the Overland Park streamgage (2004–18) was 10 Mgal/d (15.5 ft³/s) (Johnson County, Kansas, 2020). Streamflow data are not available to document changes related to wastewater discharges from the Tomahawk Creek WWTF before 2004.

The Kill Creek WWTF is 13.1 mi upstream from the Kill Creek streamgage. The plant was constructed in 2002 with a capacity of 2.5 Mgal/d (3.9 ft³/s). The Kill Creek streamgage was installed in 2003, so available streamflow data do not document streamflow changes related to discharges when the plant began operation.

Kill Creek is treated as a reference site in this study because its basin is relatively undisturbed compared to Indian Creek and other creeks in Johnson County. However, Kill Creek has been affected by urbanization, wastewater discharges, and agricultural land use and is not an undisturbed reference site.

Previous Investigations

Juracek and Eng (2017) assessed streamflow alteration for 129 selected USGS streamgages in Kansas and compared the observed condition (1980–2015) to the predicted expected (least disturbed) condition using 29 streamflow metrics. The predicted least disturbed condition is the condition expected in the absence of hydrologic modifications and was determined using a random forest (Cutler and others, 2007) model. Flow alteration at streamgages for each of the 29 metrics was quantified as the ratio of the observed value to the predicted expected value (O/E) and categorized as either minimally altered, diminished, or inflated. An O/E value close to 1, within the model error range, indicated a minimally altered condition and was considered least disturbed. Any O/E values greater than 1 or less than 1, outside the model error range, indicated inflated or diminished conditions, respectively. The statewide assessment included the Overland Park, Leawood, and Kill Creek streamgages.

Streamflow alteration statewide was likely because of human activity rather than changes in precipitation; urbanized basins had flashier flow regimes, and implemented agricultural land-management practices may have been partly responsible for an inflated magnitude of low flows in the central and eastern parts of Kansas. O/E values indicated a pronounced difference in streamflow conditions between the Kill Creek streamgage and the two Indian Creek streamgages. Most metrics (20 of 29) indicated a minimally altered streamflow condition for Kill Creek. In contrast, virtually all Indian Creek site metrics were indicative of a substantially altered condition (diminished or inflated). Alterations included an inflated condition for mean monthly flows at Indian Creek sites. Indian Creek sites also had flashier flow regimes (shorter lag times and more frequent and higher peak discharges) than what would be expected for a least disturbed condition. The flashier flow regime was indicated by the 10th-, 25th-, 75th-, and 90th-percentile flow pulses where mean annual frequency and magnitude were inflated and mean duration was diminished (Juracek and Eng, 2017).

Stream quality during 2002–10 in Johnson County, including Kill and Indian Creeks, was described by Rasmussen and others (2012) using stream-water and streambed-sediment chemistry, riparian and instream habitat, and periphyton and macroinvertebrate community data. Streamflow metrics at seven sites across the county, including the Indian and Kill Creek sites evaluated in this study, and other environmental variables were used in correlation analysis to assess factors affecting biological stream quality. A total of 20 streamflow metrics characterizing streamflow frequency, duration, magnitude, variability, and rate of change were selected for correlation analysis. Metrics indicative of streamflow magnitude and variability were most strongly correlated with biological indices of stream quality such as multimetric benthic macroinvertebrate scores, and streamflow alteration was correlated with diminished stream health.

Biological conditions reflected a gradient in urban land use. Less disturbed streams were in rural areas of Johnson County and included Kill Creek (Rasmussen and others, 2012). Biological conditions indicated Indian Creek was among the most disturbed streams in Johnson County. In 2010, 2 Kill Creek sites were among only 4 of 20 monitoring sites in Johnson County that were fully supporting of aquatic life according to Kansas Department of Health and Environment aquatic life use criteria (Rasmussen and others, 2012). All the Indian Creek sites were nonsupporting of aquatic life. Environmental variables that consistently were highly negatively correlated with biological conditions were the percentage of impervious surface and percentage of urban land use. The most important habitat variables were sinuosity, length and continuity of natural buffers, riffle substrate embeddedness, and substrate cover diversity, each of which was correlated with all macroinvertebrate metrics. Correlation analysis indicated that if riparian and instream habitat conditions improve, then so might invertebrate communities and stream biological quality. The percentage of impervious surface, as a measure of urban land use, explained 34–67 percent of the variability in biological communities. General urbanization, as indicated by impervious surface area or urban land use, was consistently determined as the fundamental factor causing change in Johnson County stream quality.

The largest loads of sediment, fecal bacteria, and nutrients have originated from urban sources transported to streams during stormwater runoff in Johnson County (Rasmussen and others, 2008; Rasmussen and Gatotho, 2014). Streamflow downstream from wastewater effluent discharges in Johnson County is largely composed of wastewater effluent during base-flow conditions (Wilkison and others, 2002; Lee and others, 2005). Sediment and bacteria concentrations are typically smaller downstream from WWTFs in Johnson County because of the diluting effect of wastewater effluent (Lee and others, 2005; Wilkison and others, 2006, 2009; Rasmussen and others, 2008; Graham and others, 2010). Wastewater effluent discharges to Indian Creek caused changes in stream-water quality that may affect biological community structure and ecosystem processes, including higher concentrations of bioavailable nutrients (nitrate and orthophosphorus) and warmer water temperatures during winter (Graham and others, 2014).

Selection and Computation of Streamflow Indicators

To investigate streamflow conditions for Indian and Kill Creeks, 40 streamflow indicators of key flow regime aspects were selected (table 2). The 40 streamflow indicators consisted of 35 commonly used indicators for characterizing streamflow (Poff and others, 1997, 201041; Olden and Poff, 2003), 2 less common seasonality indicators, and 3 other indicators based on duration curves, runoff hydrographs, and streamflow percentile classes. The 35 commonly used indicators represent 5 major components of the natural flow regime: magnitude, frequency, duration, timing, and rate of change (Poff and others, 1997). These indicators include mean flow (annual, monthly), flow variability, minimum flows, maximum flows, low- and high-flow pulses (frequency, duration, magnitude), flashiness, and rise and fall rate. Indicators also were included to evaluate seasonality of high and low flows (table 2). Additional indicators characterize differences in flow regimes and included duration curves, runoff event hydrographs, and streamflow percentile classes. Collectively, these indicators combined to provide a broad characterization of the various attributes of streamflow that could be ecologically relevant (Poff and others, 1997; Bunn and Arthington, 2002; Konrad and Booth, 2005; Arthington and others, 2010; Kennen and others, 2010; Burns and others, 2012; Carlisle and others, 2017).

Seasonality Indicators

To assess high- and low-flow seasonality, mean seasonal (3-month span) flow event frequency distributions greater than the 90th percentile and less than the 10th percentile (hereafter referred to as “seasonality of high and low flows,” respectively) were calculated using daily mean flow data for winter, spring, summer, and fall (Eng and others, 2019; example provided in fig. 2A–D). For the two seasonality indicators, the available period of streamflow record was used. The period of record was 2004–16 for the Kill Creek and Leawood streamgages and 1985–2016 for the Overland Park streamgage. Although the streamflow record for Overland Park dates back to 1963, data before 1985 were excluded because streamflow characteristics were affected by discharges from a new large WWTF beginning in the early 1980s.

For the seasonality of high flows, winter began in December, and for the seasonality of low flows, winter began in November. The beginning of the low-flow season was shifted 1 month earlier so as not to split apart the period when flows were typically lowest in the United States (August, September, and October; Lins and Slack, 2005; Eng and others, 2016). To summarize season-to-season variability for high- and low-flow seasonality metrics, we calculated the absolute value of the difference of the seasonal frequency and the mean frequency among the four seasons, repeated for all seasons, and summed the four resulting values (fig. 2A–D). Small values of seasonality metrics indicate nonseasonal behavior in which low or high flows can exist with similar frequency among the four seasons (fig. 2B). Large values of the seasonality metrics indicate a strong seasonal pattern in low or high flows (fig. 2A, C, and D). For drainage basins that have substantial human-caused modifications, shifts away from the season that has the highest frequency under natural conditions can exist (fig. 2D). To provide an indication of alteration, a seasonality index was computed by subtracting the estimated least disturbed seasonality metric from the observed seasonality metric. The least disturbed seasonality metric was estimated using the modeling approach described in Eng and others (2019). A negative value for the seasonality index indicates a loss of seasonality, and a positive value indicates a gain of seasonality (that is, a more pronounced seasonal pattern in high and low flows).

Calculation of Basinwide Precipitation

Basinwide mean precipitation was calculated to evaluate annual patterns in streamflow indicators relative to precipitation patterns. Precipitation data for 2004–18 were obtained from the City of Overland Park Stormwatch network (Overland Park Stormwatch, 2019) which provides weather data from a large network of sites that includes the study area. Basinwide annual estimates for Kill, Indian, and Tomahawk Creeks were determined by averaging precipitation measurements from all stations within each basin. By the end of the study period (2018), 8 precipitation stations were in the Kill Creek Basin, 18 stations were in the Indian Creek Basin, and 9 stations were in the Tomahawk Creek Basin.

Assessment of Hydromodification

Using the Kill Creek streamgage as representative of a least disturbed rural baseline condition, hydromodification at both streamgages along urbanized Indian Creek was quantified as the percentage difference from the rural baseline condition for each of the 35 commonly used flow indicators. Kill Creek, although likely somewhat affected by hydromodification because of agriculture activity and wastewater effects, was selected as a reference because it has been documented as less urban and less disturbed (Rasmussen and others, 2012; Juracek and Eng, 2017) and because streamflow data existed for comparison to the urban site. Quantification and comparison of flow indicators between urbanized and rural stream sites make it possible to isolate flow characteristics potentially associated with hydromodification, use the information to set management goals that may restore more natural flow characteristics, and recognize whether management activities are having the desired effect. Specific indicators were determined to be useful for identifying hydromodification in this study if differences were generally consistent (noted each year of the study) and substantial (mean of Indian Creek percentage differences exceeded an arbitrary threshold of 100 percent). Marginally useful indicators were those with differences that were somewhat inconsistent from year to year or had differences that were just less than 100 percent. Indicators that varied inconsistently from year to year and indicated small differences between Indian Creek and Kill Creek were determined not to be useful for identifying hydromodification. Seasonal hydromodification was assessed using two seasonality metrics that compared observed condition to predicted least disturbed condition. Three additional streamflow indicators were used to further characterize hydromodification by making comparisons between historical and more recent (2004–18) flow duration ranges, event hydrographs, and flow percentiles.

Streamflow Conditions in Indian and Kill Creeks

Time-series plots of daily mean streamflow, normalized by drainage area, from 2004 to 2018 (figs. 3A, B and 4) illustrate general flow patterns at the three stream sites and represent flow during a range in annual precipitation conditions. Daily flows were consistently higher at the Indian Creek sites than the Kill Creek site. The two streamflow peaks of record for Indian Creek were in August and July 2017. The Kill Creek site experienced several weeks of zero flow during 2012. During 2004–18, annual precipitation at Kill Creek was lowest in 2012 and highest in 2015, and annual precipitation at Indian and Tomahawk Creeks was lowest in 2012 and highest in 2017(fig. 5).

Annual median and mean streamflows were consistently and substantially higher for Indian Creek than Kill Creek. Compared to Kill Creek, the mean of the annual median flows during 2004–18 at Overland Park and Leawood was about 310 percent and 240 percent larger, respectively (fig. 6A, appendix 2, indicator MED_ANNUAL). Likewise, the mean of the annual mean flows at Overland Park and Leawood during 2004–18 was about 160 percent and 130 percent larger, respectively, than Kill Creek (fig. 6B, appendix 2, indicator MEAN_ANNUAL). Median and mean annual streamflows (normalized for basin size) were larger at the urban sites than the rural site primarily because impervious surfaces, bank armoring, and channelization associated with urbanization lead to rapid runoff rather than infiltration. Contributions from wastewater discharges also may account for differences. Patterns in annual median (fig. 6A) and annual mean (fig. 6B) flows generally reflected annual precipitation patterns (fig. 5). Some differences can be attributed to specific storm events; for example, the largest precipitation event in the Indian Creek basin during the study period was in June 2010, which contributed toward the 2010 spike in annual median and mean flows at the Overland Park site. The annual coefficient of variation of daily mean flows was consistently higher at Kill Creek compared to Overland Park sites possibly because less WWTF discharge resulted in increased variability in low flows (fig. 6C).

Monthly mean streamflows were consistently higher for Indian Creek than Kill Creek, except during September 2015 and 2016 (fig. 7A–L) when more monthly precipitation fell in the Kill Creek drainage basin than normal. The most consistent year-to-year separation between the two streams was evident for the month of August (fig. 7H). For that month, the mean of the daily mean flows during 2004–18 at Overland Park and Leawood was 353 percent and 279 percent larger, respectively, than Kill Creek (appendix 2).

Annual 1-day, 3-day, 7-day, 30-day, and 90-day minimum daily mean streamflows were consistently and substantially higher for Indian Creek than Kill Creek (fig. 8A–E). For example, the mean of the 3-day minimum daily mean flows during 2004–18 at Overland Park and Leawood was 1,900 percent and 1,200 percent larger, respectively, than at Kill Creek (fig. 8B, appendix 2, indicator MIN_3DAY). The primary factor contributing to larger minimum flows in Indian Creek likely was wastewater discharge. Domestic irrigation and leaky water infrastructure also may be contributing factors.

Annual 1-day, 3-day, and 7-day maximum daily mean streamflow differences between Indian Creek and Kill Creek were evident but less pronounced than annual minimum mean flows. Overall, annual maximum mean flows were higher for Indian Creek; however, some annual maximum mean 1-day and 3-day flows at Kill Creek were greater than one or both of those flows at the Indian Creek sites (fig. 9A–C). These exceptions were in 2009, 2014, and 2016 and likely resulted from differences in basin-specific rainfall events.

Annual peak streamflows, in terms of frequency and magnitude, were somewhat higher for Indian Creek than Kill Creek (fig. 10A, B). Peak flow frequency, defined as the annual number of pulses greater than triple the period-of-record median flow (table 2), averaged about 91 percent and only 10 percent higher at Overland Park and Leawood, respectively, compared to Kill Creek (fig. 10A, appendix 2, indicator PEAK_NUM). Peak flow magnitude, defined as the annual mean magnitude of flow for days when the flow was greater than triple the period-of-record median flow (normalized by drainage area; table 2), averaged 218 percent and 161 percent higher at Overland Park and Leawood, respectively, compared to Kill Creek (fig. 10B, appendix 2, indicator PEAK_MAG).

Additional annual low-flow indicators assessed were the number, mean duration, and mean magnitude of flow pulses less than the period-of-record 10th percentile. The ability to compare these indicators among the three streamgages was somewhat constrained because such pulses were not observed for every year for all three sites (fig. 11A–C). In general, the annual number of such pulses was lowest at Kill Creek; however, this was not always the case (fig. 11A). Available data indicated that the annual mean pulse duration typically was longer at Kill Creek with considerable year-to-year variability as to how much longer (fig. 11B). For annual mean pulse magnitude, the available data indicated consistently and substantially higher flows for Indian Creek than Kill Creek (fig. 11C). Specifically, the mean of the annual mean pulse magnitude less than the 10th percentile (normalized by drainage area) was 2,000 percent and 1,400 percent higher at Overland Park and Leawood, respectively, compared to Kill Creek (appendix 2, indicator PUL_MAG_P10).

Additional annual high-flow indicators assessed were the number, mean duration, and mean magnitude of flow pulses greater than the period-of-record 90th percentile. The annual number of such pulses was marginally higher for Indian Creek compared to Kill Creek (fig. 12A). Specifically, the number of such pulses averaged 89 percent and 85 percent higher at Overland Park and Leawood, respectively (appendix 2, indicator PUL_NO_P90). Annual mean pulse duration typically was longer at Kill Creek with considerable year-to-year variability as to how much longer (fig. 12B). Annual mean pulse magnitude was marginally higher for Indian Creek (fig. 12C). Specifically, the mean of the annual mean pulse magnitude (normalized by drainage area) was 113 percent and 93 percent higher at Overland Park and Leawood, respectively, compared to Kill Creek (appendix 2, indicator PUL_MAG_P90).

Neither flashiness indicator (Richards-Baker flashiness index or the fraction of the year the daily mean flow is greater than the annual mean flow [RB_INDEX and TQMEAN, respectively; table 2]) indicated a substantial difference in flashiness between the two streams (fig. 13A–C). This result was unexpected because increased flashiness generally is a classic response to increased urbanization. When Indian Creek streamflow contributed by wastewater effluent was subtracted from total streamflow, the Richards-Baker flashiness index was lower for Kill Creek than the Indian Creek sites, indicating that wastewater may have masked differences in flashiness between basins.

Annual mean rise and fall rates (table 2) were consistently and substantially higher for Indian Creek (fig. 14A, B). The mean of the annual mean rise rate during 2004–18 was 136 percent and 115 percent higher at Overland Park and Leawood, respectively, compared to Kill Creek (fig. 14A, appendix 2, indicator RISE_RATE). Similarly, the mean of the annual mean fall rate was 169 percent and 125 percent higher at Overland Park and Leawood, respectively (fig. 14B, appendix 2, FALL_RATE). At all three streamgages, the flow rise rate was about twice the fall rate.

Changes in flow seasonality were evident for Indian and Kill Creeks (fig. 15A–F). At both streamgages along Indian Creek, a loss of seasonality for high flows (fig. 15A, B) and low flows (fig. 15D, E) was indicated by a “flattening out” of the observed season-to-season flow frequencies and negative values for the associated seasonality index. For Kill Creek, the observed seasonality of high flows was virtually unchanged from the estimated natural condition (fig. 15C); however, the low-flow seasonality at Kill Creek increased (positive seasonality index) and was characterized by fewer low flows in the November–January period and more low flows the rest of the year (fig. 15F).

Streamflow-duration curves for the available period of record for Indian Creek at Overland Park and Kill Creek (fig. 16) illustrate the change in daily mean flow for Indian Creek since the early 1960s. Flows during the more recent 15 years (2004–18) had much lower variability most of the time, and higher flows were more frequent compared to the first 12 years of the record. Lower variability during 2004–18 may be caused by more steady flows from wastewater discharges and other urban contributions such as irrigation runoff and leaking infrastructure. More frequent high flows during 2004–18 may be caused by runoff from more impervious surfaces compared to the earlier 12 years. During 1963–75, the range in normal streamflow (25th–75th percentile) was about 1.5–13 ft³/s, flows exceeded 100 ft³/s about 4 percent of the time, and there was no flow about 6 percent of the time. During 1976–85, the normal flow range had slightly increased to 2.6–19 ft³/s, flows exceeded 100 ft³/s about 5 percent of the time, and all daily mean flows exceeded 0 ft³/s. Flows during 1986–2003 were similar to flows in the more recent period of 2004–18 and much lower variability was in the normal range, about 15–32 ft³/s; flows were less than 10 ft³/s less than 1 percent of the time; and flows exceeded 100 ft³/s about 9 percent of the time. Kill Creek daily mean flows during 2004–18 were most similar to Indian Creek flows during 1976–85 when Indian Creek was less affected by urbanization and wastewater discharges.

Differences in magnitude, duration, and falling limb characteristics of runoff events are evident in runoff hydrographs for Indian Creek and Kill Creek. In one example that is characteristic of such events (fig. 17), during May 29 through June 1, 2013, runoff events with peak flows of 4,900 ft³/s at Kill Creek (just less than the estimated peak flow of 5,410 ft³/s for an event with an exceedance probability of 0.2; table 3) and 18,000 ft³/s at Leawood (just less than the estimated peak flow of 19,000 ft³/s for an event with an exceedance probability of 0.1) indicate characteristic hydrograph differences between urbanized and nonurbanized streams (Rosburg and others, 2017). Although the Leawood magnitude is larger in part because the event is larger (and the basin is larger), the hydrograph also is of shorter duration with a steeper falling limb compared to Kill Creek. Flow volumes for defined runoff events may be monitored and used to set management goals for flow reduction. A similar approach using defined rainfall events also may be useful.

Streamflow plotted in percentile classes for selected periods of record indicated changes in streamflow conditions for Indian Creek since the 1960s and differences compared to Kill Creek. The 7-day mean streamflows indicate that during 1964–73, less than normal flows (less than 25th percentile) on Indian Creek were much more common compared to 2007–16 (fig. 18A, B). Wastewater discharges sustained flow throughout the year during the more recent period. Flows were more variable during 2007–16 compared to 1964–73, as indicated by more frequent peaks. In addition, more of the 2014 hydrograph (representative of about normal rainfall for both streams) plots in the greater than normal range (greater than the 75th percentile; referred to as “above normal” on fig. 18A–C) during 1964–73 compared to 2007–16, which indicates that flow percentiles have shifted upward. Less than normal flows also were much more common at the Kill Creek site during 2004–16 compared to Indian Creek during 2007–16 (referred to as “below normal” on fig. 18A–C). The 7-day mean streamflow was plotted as an example that smooths out the hydrograph for general understanding. Other streamflow measures could be illustrated in a similar way.

Hydromodification of Indian Creek

Consistent with Juracek and Eng (2017), substantial hydromodification of Indian Creek was indicated as compared to Kill Creek, which, for the purposes of this study, was considered representative of a least disturbed rural reference condition. Of the 35 common streamflow indicators evaluated, 19 indicated a generally consistent and substantial difference between Indian and Kill Creeks. Hydromodification of Indian Creek was characterized by larger mean annual and monthly flows (figs. 6B and 7A–L), larger low flows of shorter duration (figs. 8A–E and 11B–C), larger high flows with increased frequency and shorter duration (figs. 10A and 11A–C), and faster rise and fall rates (fig. 14A, B). For the two seasonality indicators, seasonality of high and low flows decreased (fig. 15A–F). Duration curves, runoff event hydrographs, and streamflow percentile classes also indicated differences between the two streams for specific ranges of flow. The primary causes of hydromodification in the Indian Creek Basin include impervious surfaces, altered flow paths, and effluent contributions from sewage treatment plants in the urbanized basin.

Utility of Streamflow Indicators

To be considered useful for hydromodification assessment, indicators would be ecologically relevant and quantifiable, as were the indicators used in this study. In addition, for this study, a candidate indicator ideally varied consistently and substantially between the two Indian Creek streamgages and the Kill Creek streamgage for all years included in the analysis (that is, 2004–18). Based on this criterion, many of the indicators were useful in discriminating between site types for this study area (table 4). Annual median flow (fig. 6A) and annual mean flow (fig. 6B) were similarly useful and could be used interchangeably. Of the monthly mean flows, the most effective months were February, July, August, September, October, November, and December (fig. 7A–L). The minimum mean flows (1-day, 3-day, 7-day, 30-day, and 90-day) were all useful (fig. 8A–E). For flow pulses less than the 10th percentile, the annual number and mean magnitude were useful (fig. 11A–C). The rise and fall rates effectively discriminated between the two streams (fig. 14A, B). Like the indicators, the flow-duration curves (fig. 16), runoff event hydrographs (fig. 17), and streamflow percentile classes (fig. 18A–C) were effective at showing differences in streamflow characteristics among periods and streams. The high-flow and low-flow seasonality indicators, although informative, are not intended for use in annual assessments.

Several indicators were considered marginally useful either because the distinction between the two streams was not pronounced for at least 1 year or the pattern was not consistent for at least 1 year. Included in this category were the monthly mean flows for January, March, April, and June (fig. 7A, C, D, and F). Other marginally useful indicators were peak number (fig. 10A) and pulses greater than the 90th percentile (fig. 12A).

Some indicators were not useful either because the two streams were similar for multiple years or the pattern was not consistent for multiple years. Included in this category were the coefficient of variation of daily mean flows (fig. 6C); the monthly mean flow for May (fig. 7E); annual 1-day, 3-day, and 7-day maximum daily mean flow (fig. 9A–C); the annual mean duration of flow pulses less than the 10th percentile (fig. 11B) and greater than the 90th percentile (fig. 12B); and the flashiness indicators Richards-Baker flashiness index and the fraction of the year the daily mean flow is greater than the annual mean flow (fig. 13A–C). The ineffectiveness of the two flashiness indicators to distinguish urbanized Indian Creek from mostly rural Kill Creek was not anticipated and may be caused by wastewater discharge that masked flashiness at the Indian Creek sites.

Duration curves, runoff event hydrographs, and streamflow percentile classes provide additional opportunities for understanding hydromodification and for monitoring changes in streamflow characteristics that may not be evident using standard indicators. Although those indicators were described in this report more visually, changes could be quantified over time relative to management goals. Duration curves can be constructed using different periods of record (for example, annually, monthly, for defined runoff events, predevelopment, and postdevelopment) to evaluate opportunities for managing specific ranges in flow that might restore more natural flow regimes. Runoff event hydrographs could be used to establish goals for reducing runoff and streamflow volume for defined streamflow events and restoring more natural flow regimes. Additional coordination of feasible management activities and expected restoration results could help set these quantifiable goals.

Streamflow plotted in percentile classes helps facilitate conceptual planning for seasonal streamflows that can support healthy streams and aquatic communities (table 5; DePhilip and Moberg, 2010). Seasonal high flows in Indian and Kill Creeks generally are important for maintaining floodplain connectivity, maintaining channel morphology, and flushing organic matter and fine sediment (table 5). Streamflow variability is needed to support vegetation and habitat and fish spawning and development. Normal flows throughout the year would help to connect habitats, support aquatic life, and maintain water quality. Evaluation of historical, natural, and seasonal streamflow characteristics along with a general understanding of seasonal flows that could help support ecosystems provides a foundation for establishing meaningful goals and identifying management actions that could help restore flow regimes.

Future Assessments and Management Implications

In this study, multiple indicators that can be used to numerically and visually assess hydromodification in urban Johnson County streams, and likely other similar streams in the region, were identified (table 4). Indicators provided a baseline characterization of hydrologic conditions for Indian and Kill Creeks, as well as a means for assessing hydromodification of the former. In addition, indicators may be used to evaluate effectiveness over time of practices implemented to restore urban streams. For multibasin assessments, indicators can be used to rank streams as to the degree of hydromodification and the priority for restoration.

A next step would be to identify management goals for restoring flow regimes that are specific to selected streamflow indicators, ecologically meaningful, and achievable. Goals could be described in terms of percentage of change in specific streamflow indicators. Determining which aspects of the flow regime could be targeted for restoration warrants consideration of feasible management options. Additional work to determine which indicators are more ecologically relevant for these systems may help managers better target restoration efforts. Specific aspects of modified flow that are harmful to ecological populations could be identified for remediation. In addition, changing climate may interfere with hydromodification assessments by making it more difficult to distinguish effects of climate variability on streamflow characteristics from land and water management effects (Palmer and others, 2009).

Only two streams were evaluated in this study. Expanding this assessment to a larger area to include more streams can provide a more complete understanding of streamflow characteristics across a larger range of stream systems and potentially improve the likelihood of effective management and restoration plans. Annual values for the 35 commonly used streamflow indicators at 11 USGS streamgages for the period 1999–2018 are provided in appendix 3. Optionally, a smaller subset of primary indicators could be selected based on usefulness for characterizing hydromodification, potential for implementing management practices that likely would result in improvements, ability to set goals and monitor progress toward achieving those goals, and potential for detecting short-term (less than 10 years) and long-term (more than 10 years) changes. A multimetric scoring and ranking system could be developed to prioritize and monitor changing streamflow characteristics. In the future, periodic reassessments (for example, every 5 years) using selected indicators could be used to determine progress, or lack thereof, toward achieving stream-restoration goals. Results of the reassessments would be intended to support adaptive management in which goals, priorities, and practices may be revised as appropriate.

Another possible path forward for urban stream restoration would involve the implementation of different practices in different basins with attendant monitoring using the various indicators to assess which practices are most effective and at what level of implementation. Once determined, the “best” practices could then be implemented in all basins. Ideally, the hydrologic assessments would be coupled with biological assessments. Such a coupling would serve to determine the extent to which the ecological health of streams changes in response to changes in hydrologic conditions. In particular, if restoration efforts are successful in creating more natural flow conditions, the coupled assessment would provide information to ascertain whether or not the ecological health of the stream also improved.