Description of Study Area

Fountain Creek watershed has a drainage area of 927 square miles (mi²) and flows into the Arkansas River at Pueblo, Colo. (USGS, 2019b; fig. 1). The Fountain Creek watershed has a maximum elevation of 14,100 feet (ft) at the summit of Pikes Peak, and a minimum elevation of 4,630 ft at the confluence of Fountain Creek and the Arkansas River. The Fountain Creek watershed comprises two physiographic regions: the Front Range of the Southern Rocky Mountains and the Colorado Piedmont (Hansen and Crosby, 1982; Miller and Stogner, 2017; fig.1). The Front Range region makes up the western one-third of the watershed and is primarily underlain by granite and other crystalline rocks (von Guerard, 1989). The eastern two-thirds of the Fountain Creek watershed is made up of the Colorado Piedmont region, a subregion of the Front Range (Hansen and Crosby, 1982; Kohn and others, 2014). The geologic formations of the Colorado Piedmont are sedimentary in nature; soils are readily erodible and are thus a key source of sediment to Fountain Creek (Larsen, 1981). A more detailed description of surface geology and soils in the Fountain Creek watershed is presented by Larsen (1981) and von Guerard (1989).

The main stem channel of Fountain Creek can be separated into three distinct segments with differing characteristics. The upper reach, located upstream from Manitou Springs (fig. 1), has pool-and-riffle channel morphology, and bed material comprises sand and gravel to cobble- and boulder-size particles (Stogner and others 2013; Miller and Stogner, 2017). The middle segment, which begins at the confluence of Monument Creek and ends at USGS site 07105800 (Fountain Creek at Security, Colo.; fig. 1), is channelized and dominated by runs with intermittent pools. The bed material consists of sand, gravel, and cobble and is locally scoured to bedrock in some areas. Additionally, some locations in this section have been lined with concrete to lessen streambank erosion during high flow (Stogner and others, 2013; Miller and Stogner, 2017). The lower reach extends downstream from USGS site 07105800 (Fountain Creek at Security, Colo.) to the confluence with the Arkansas River. The channel is braided through this section with meanders more prevalent than in the upper reaches. The channel is wider throughout the lower reach, resulting in shallower stream depths. The streambed in the lower reach is almost exclusively composed of sand and small gravel, which can be easily transported at the full range of flows (Stogner and others 2013; Miller and Stogner, 2017).

Tributaries of Fountain Creek important to this study include Camp Creek, South Cheyenne Creek, Monument Creek, Bear Creek, Sand Creek, Jimmy Camp Creek, and Little Fountain Creek (USGS, 2019b; fig. 1). Monument Creek is the main tributary to Fountain Creek and originates in the Rampart Range. From its origin, Monument Creek flows eastward toward Palmer Lake and then south toward Colorado Springs and its confluence with Fountain Creek. The upper reach of Monument Creek is meandering and contains pools, riffles, and runs. Bed material in the upper reaches of Monument Creek consists primarily of sand, gravel, and cobble. In the lower reach of Monument Creek (beginning at the confluence of Cottonwood Creek), the channel is often braided and characterized by finer bed material, including sand and small gravel. Additionally, Monument Creek is affected by urbanization in the lower reaches, with intermittent concrete-lined stream banks and extensive channelization (Kohn and others 2014; Miller and Stogner, 2017).

Land Use and Climate

According to the 2021 National Land Cover Database, 21.0 percent of the Fountain Creek watershed can be described as developed (urban) land, 24.9 percent is forested, 37.6 percent is covered by herbaceous growth including grassland, and 12.8 percent can be described as shrubland (Dewitz, 2023). Although the watershed has a variety of land uses, increased population and urban land use in the upper watershed has the potential to intensify sediment transport in the lower reaches of Fountain Creek. Urban land use commonly creates impermeable surfaces, which can accelerate storm runoff, increase the magnitude of storm runoff peaks, and thereby increase the amount of stream energy available for bed and bank erosion and sediment transport (Kohn and others, 2014).

The Fountain Creek watershed has experienced urban development and population growth in the last century, with a steady increase beginning in the 1980s (Colorado Department of Local Affairs, 2024). El Paso County, in which a majority of the Fountain Creek watershed lies, is the most populous county in the State as of 2022 and experienced the fifth largest population increase in Colorado from 2021 to 2022. The population of El Paso County has grown from about 515,000 in 2000 to about 740,000 in 2022 (Miller and Stogner, 2017; Colorado Department of Local Affairs, 2024).

Mean annual precipitation in Fountain Creek watershed is 18.85 inches (in) and generally decreases with decreasing elevation (Miller and Stogner, 2017; USGS, 2019b). Mean monthly precipitation ranges from a minimum of 0.32 in in January to a maximum of 3.34 in in August (Weather Service, 2019). Mean annual precipitation in the study area varies considerably from year to year; however, most precipitation occurs from May through August (Miller and Stogner, 2017; Weather Service, 2019). Flow in Fountain Creek has more influence from snowmelt from April through June and is characterized by diurnal patterns affected by heating during the day and cooling during the night, and from increased daily flows from wastewater-treatment-plant output (Stogner and others, 2013). From mid-June through October, flow regimes are defined by rapid increases in streamflow in response to convectional thunderstorms that build throughout the day. These rapid rises result in flash floods, have the potential to transport high loads of sediment, and can cause drastic changes in stream channel geometry (Stogner and others, 2013).

Selection of Data for Trend Analysis

Monthly mean SSF (expressed as tons per day) and monthly mean streamflow (expressed as ft³/s) at 10 USGS streamgage sites (table 1) within the Fountain Creek watershed were analyzed for trends during their respective operational periods, ranging from 1998 to 2022. Monthly mean SSF and monthly mean streamflow, used for trend analysis, were calculated from daily mean SSL and daily mean streamflow, compiled from USGS NWIS database (USGS, 2024), and published by Downhour and Hennessy (2024). Daily mean data were used to calculate monthly mean values to avoid introducing autocorrelation in robust linear regression analysis (Theil-Sen estimator). Monthly mean was found to be the most appropriate statistic because data-point frequency remained unchanged during the period of analysis and ensured each summary statistic was computed using a near equal number of observations (Helsel and others, 2020). Monthly mean SSF and monthly mean streamflow were transformed using natural log prior to use in trend analysis. When a log transformation was used, quantile plots displayed a more normal distribution, and residual plots were more heteroscedastic. Additionally, linearity was improved when using log transformation. Data used for analysis are summarized in table 1.

To further assess stream behavior, the Richards-Baker flashiness index (RBI), a measure of flashiness, was analyzed (Baker and others, 2004). The RBI was calculated monthly for the same months as streamflow and suspended sediment (April through September) using all available streamflow data for a given site (USGS, 2024). The resulting monthly values were then assessed for trends using the Mann-Kendall trend test as described in the “Temporal Trends Statistical Analysis” section of this report. Daily mean SSC and daily mean streamflow data were used to calculate daily mean SSL independently for each seasonal period (using GCLAS; Porterfield, 1972; USGS, 2019a) in the analysis, allowing any year-to-year changes in the relation between streamflow and SSL to be observable when plotted in a double-mass curve. Monthly total (April through September) SSL (expressed in tons; calculated from sum of daily mean load for each month) and monthly total streamflow (expressed in acre-feet [acre-ft]; calculated from sum of daily mean streamflow for each month), were used to create double-mass curves for all study sites regardless of the presence of a significant trend in either SSF or streamflow data.

Temporal Trends Statistical Analysis

A Mann-Kendall trend test, in conjunction with the Thiel-Sen slope (robust linear regression) to quantify rate of change, was used for evaluating trends in monthly mean SSF and streamflow to detect monotonic trends for 9 out of 10 sites in this analysis (app. 1, fig. 1.1–1.18). Monotonic trends are those that have generally consistent increases or decreases through time (Helsel and others, 2020). No assumption of normality is needed for the Mann-Kendall trend test that determines whether the median changes through time (Helsel and others, 2020). Because of the presence of occasional outliers in suspended sediment and streamflow data during periods of high flow or floods, ordinary least squares regression slope calculations may have been disproportionately affected, making the Thiel-Sen slope better suited for the analysis. Although log transformations improved the data distribution, some nonnormality was still observed in exploratory plots, violating the assumption of a linear relation necessary to use ordinary least squares for obtaining a best linear unbiased estimator (Helsel and others, 2020). Ordinary least squares regression may have been appropriate for some sites with normal distribution; however, a consistent method that works for all sites is preferred. One site, 07103700 (Fountain Creek near Colorado Springs, Colo.), required a Wilcoxon signed-rank test as there were two distinct time periods available for analysis, with a relatively long gap between them. There is no specific rule on how long a gap in data needs to be to require step trend procedures; however, the general rule of gap length greater than a third of the entire period of data collection was followed (Helsel and others, 2020). Site 07104905 (Monument Creek at Bijou Street at Colorado Springs, Colo.) had a gap in the data from 2011 to 2016; however, it was not long enough to require step trend procedures.

Double-Mass Curves

Double-mass curves were used to check the consistency of the relation between streamflow and SSL. Both monthly sediment load and monthly streamflow from the same site were used as measured variables in each of the double-mass curves, which is discussed further in the “Other Considerations” section of this report. The method used in this study can show changes in the relations between the variables but can lack power to determine which variable is responsible for the break in certain circumstances. Additionally, this method cannot determine if changes in watershed characteristics or hydrology are the cause of any observed shifts in the relation.

Of the 10 sites included in this study, 9 displayed observable break points in the double-mass plots, determined by application of a nonparametric LOESS smooth line (fig. 2.1–2.10). Permutation tests run on these nine sites determined that the period (used as the interaction variable) was statistically significant at the 95 percent confidence level (p<0.05), indicating an interaction effect was present. All sites with an observable break point in the plot and a significant interaction effect were then analyzed using regression analysis and hypothesis tests. The slope values for each period (determined by observed break points in time) were calculated and compared pairwise (table 3). The results of the slope comparison determined that the null hypothesis (no difference) could be rejected for all sites and all periods, indicating that the slope change was statistically significant at the 95-percent confidence level (p<0.05). One exception to these results was at USGS site 07105600 (Sand Creek above Mouth at Colorado Springs, Colo.). The null hypothesis for slope comparison of periods A (2003–05) and C (2008–11) at site 07105600 could not be rejected, indicating there is not a significant difference in the relation of SSL and streamflow between these two groups. Known watershed changes that coincide with these break points in the double-mass curves are discussed in the “Characterization of Suspended Sediment Flux and Streamflow Trends” section of this report.

Flashiness Index

One method to classify streams is to position them on a spectrum ranging from extremely stable groundwater-based streams on one end to very flashy streams on the other. The RBI is a quantitative measure that can be used to establish this continuum (Baker and others, 2004). The highest mean RBI values, which represent more flashy streamflow behavior, were shown by USGS sites 07105600 (Sand Creek above Mouth at Colorado Springs, Colo.; fig. 1) and 07103990 (Cottonwood Creek at Mouth at Pikeview, Colo.; fig. 1), which are two small tributaries to Fountain Creek that drain entirely from the Colorado Piedmont region (fig 1; table 4). In contrast, the least mean flashy behavior was shown by USGS sites 07103700 (Fountain Creek near Colorado Springs, Colo.; fig. 1) and 07103970 (Monument Creek above Woodmen Road at Colorado Springs, Colo.; fig.1), which were the most upstream sites on the main stems of Fountain Creek and Monument Creek and include the Front Range physiographic region in their headwaters. USGS site 07105530 (Fountain Creek below Janitell Road below Colorado Springs, Colo.; fig. 1) had a similarly low mean RBI value, but streamflow at that site was moderated by substantial discharge from an upstream wastewater treatment plant (Bern and others, 2024). Farther downstream along the main stem of Fountain Creek, mean RBI values were intermediate to those calculated for upstream sites (table 4).

Monthly mean RBI decreased insignificantly through time at all but one site, but percent change per year was 0 percent for all sites (table 4). At USGS site 07103970 (Monument Creek above Woodmen Road at Colorado Springs, Colo.), there was a significant (p=0.06) increase in RBI through time, but as with the other sites, the magnitude of change was negligible. Because changes in flashiness through time are negligible, changes in flashy stream behavior can be excluded as a potential driver of observed changes in SSL in the watershed.

Other Considerations

This report focused on trends of SSF, which are calculated by multiplying SSC, from samples collected at varying stages along the hydrograph, by streamflow. Increases or decreases in monthly mean streamflow can result in a related increase or decrease in SSF if the SSC remains unchanged. For this reason, it was necessary to also evaluate trends in streamflow in this report to provide a possibly more accurate perspective of any significant trends identified in SSF. Two sites (Fountain Creek below Janitell Road below Colorado Springs, Colo. and Fountain Creek near Piñon, Colo.) where a statistically significant negative trend in monthly mean SSF was identified also had a statistically significant negative trend in monthly mean streamflow (table 2). For these two sites, caution may be beneficial when evaluating the negative trend in SSF. Further analysis would be required to quantify how much of the observed trend in SSF is due to the associated decrease in streamflow. Three sites (Cottonwood Creek at Mouth at Pikeview, Colo., Monument Creek at Bijou Street at Colorado Springs, Colo., and Fountain Creek at Security, Colo.) that had a statistically significant negative trend in SSF did not have an associated significant negative trend in streamflow (table 2). For these three sites, it can be assumed that the decrease in sediment load is most likely due to a decrease in SSC at the sites as there is less sediment per volume of water at those locations (USGS, 2024).

Caution needs to be used when drawing conclusions from the double-mass curve plots in this report. When using two measured variables (rather than using the preferred one measured variable and the cumulation of a pattern of many records in an area), the double-mass analysis may give indefinite results because it is not always possible to determine which variable is responsible for the break in slope (Searcy and others, 1960). The presence of known large-scale watershed changes that have occurred during the operational periods could help to overcome this limitation. Additionally, the SSF data used in this report are not directly measured; rather, they are calculated by the interpolation method described by Porterfield (1972) using the GCLAS software (Koltun and others, 2006). The inherent variability of hydrologic data needs to be considered as well when considering spurious breaks in the double-mass curve plots. Numerous sites in this study show a visible break in the double-mass curves between 2013 and 2015 (app. 2, fig. 2.1–2.10) and may be due to effects from the Waldo Canyon Fire in 2012 and the abnormally high precipitation observed during the 2013 and 2015 seasonal periods (Cole and others, 2014; USGS, 2024). Although these break points are observable on the graph, any break in a curve that persists for fewer than 5 years is not considered reliable. Breaks that persist for 5 years or longer could be due to chance or a real change, and unless the break coincides with a known change in the watershed, statistical methods would be beneficial to evaluate the differing slopes (Searcy and others, 1960).