Advertisement

  • Loading metrics

Open Access

Peer-reviewed

Research Article

Urbanization in Denver produces more streamflow because of contributions from excess irrigation, leaking pipes, and stormwater

Abstract

Urban streams in the Denver, Colorado, USA region flow more often than undeveloped grassland streams. We sought to identify the sources of this increased flow using water stable isotope data and an analysis of streamflow responses to rain events. We collected and assessed 402 urban stream, 522 tap, and 38 precipitation samples across 2019, 2021, and 2022. Two endmember mixing analysis was utilized to obtain the percentage of precipitation-derived groundwater and tap water contributing to urban baseflow. Our endmember mixing results revealed that a major portion of stream water came from tap water, through excess lawn irrigation returning to the stream and leaking water pipes. The average portions of streamflow that come from tap water and lawn irrigation return flow were 76% and 47% respectively across 2019, 2021, and 2022. Uncertainty related to estimation of tap contribution and lawn irrigation return flow ranged from 3 – 29%. We also observed an increasing correlation between lawn irrigation return flow in urban streams and imperviousness of the watersheds in the Denver area. In semi-arid and arid cities in the USA, including in Denver, urban irrigation consumes a large portion of city water. Through an analysis of spatiotemporal variations in streamflow, we observed that tap water is a larger contributor to urban streamflow than increased stormflow during most months. The joint contributions of tap water and directly-connected impervious area driving increased stormwater lead to profound alterations in the urban streamflow regime compared to grassland streamflow. This study provides insights into how urban irrigation and stormwater together increase streamflow, aiding water managers in implementing effective water management strategies in water-scarce cities.

Citation: Al Fatta A, Bhaskar AS (2025) Urbanization in Denver produces more streamflow because of contributions from excess irrigation, leaking pipes, and stormwater. PLOS Water 4(7): e0000299. https://doi.org/10.1371/journal.pwat.0000299

Editor: Pierre Horwitz, Edith Cowan University - Joondalup Campus: Edith Cowan University, AUSTRALIA

Received: September 3, 2024; Accepted: May 26, 2025; Published: July 9, 2025

Copyright: © 2025 Al Fatta, Bhaskar. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: Al Fatta, A., A. S. Bhaskar (2025). Datasets: Urbanization in Denver produces more streamflow because of contributions from excess irrigation, leaking pipes, and stormwater, HydroShare, https://doi.org/10.4211/hs.b213390987b74771bd9317c1588277a9

Funding: Funding for this research was provided by NSF grants 2045340, 2318903, and 2115169. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Competing interests: The authors have declared that no competing interests exist.

1. Introduction

Arid and semi-arid cities are increasingly facing water scarcity due to urbanization and climate change [1–3]. These changes intensify the gap between water supply and demand, making it crucial to minimize this imbalance for a sustainable water supply [4,5]. Urban landscape irrigation, predominantly for lawns, can account for 40–75% of the total domestic water supply in arid and semi-arid regions of the United States [6–12]. Reducing water used for urban irrigation can effectively contribute to water security in these water-scarce regions. A management concern about limiting urban irrigation is that may significantly impact urban stream baseflow, and then riparian zones, and surrounding ecosystems [13,14]. Urbanization generally affects urban stream baseflow, in ways that are complex and vary by setting [15,16], influencing stream ecology and the surrounding environment [17,18]. For example, Passarello et al. [19] studied urban-induced recharge in Austin, TX, finding that leaky pipelines, wastewater leakage, and urban irrigation return flow significantly contributed to recharge. Managing the effect of changes in specifically urban irrigation to baseflow requires a quantitative understanding of the temporal and spatial patterns of irrigation contributions to urban streams.

In semi-arid and arid areas, urban irrigation constitutes a large portion of urban water use, as well as contributes to streamflow [9,17]. Lawn and landscaping irrigation often contribute to peak demand, making them a target for conservation approaches [12]. Excess lawn irrigation can be transported to urban streams directly, through storm sewer systems, or by recharging the subsurface and discharging into streams [17]. For example, Manago & Hogue [20] evaluated the impacts of imported water and water conservation policies on streamflow dynamics in the semi-arid Los Angeles area, finding that before conservation efforts, heavy irrigation led to increased streamflow in the urbanized watershed [20].

Hydrometric measurements and climatic data are used to characterize the hydrologic cycle and contributions of various sources to streamflow, but it is often challenging to get insight into highly urbanized areas due to complex water systems [21,22]. Water stable isotope (δ2H and δ18O) techniques are widely used to better understand the hydrologic cycle, especially in natural environments [23–25]. Water stable isotopes have been also used for understanding the hydrologic phenomena in heavily urbanized water supply systems [26,27], including the distribution of tap water in urban areas [26,28–31]. Water-stable isotopes have also been used to identify the sources of streamflow and stormflow. Isotopic analysis showed that groundwater and wastewater effluent accounted for dominant streamflow sources in Berlin, Germany [32], but in San Diego, California, USA, stormwater was mostly baseflow with little direct contribution from rainfall or imported water [31]. Geochemical analysis and isotopic studies had also been used to quantify sources of streamflow; for example, municipal water contributes the majority of flow to a stream in Austin, Texas, USA [33]. These studies demonstrate the usefulness of water stable isotope techniques for evaluating the fraction and spatiotemporal distribution of urban streamflow sources.

Previously, in the Denver, Colorado, USA region, Fillo et al. [17] analyzed contributions of tap water to urban baseflow using water stable isotopes, but only focusing on August – September 2019. In this study, we combine the previous Fillo et al. [17] analysis with two additional years (2021–2022) of sampling using the water stable isotope technique along with water provider reports to quantify the contribution of streamflow sources to the urban streams for three years. We also compared the contributions of tap water to baseflow to the changes in stormflow with urban development from an lengthening of the analysis of Wilson et al. [34] and the overall changes in streamflow across the urban-rural gradient in the Denver metropolitan area. The research questions of this study are:

  1. What are the contributions of precipitation-derived groundwater, water losses, and lawn irrigation return flows to urban baseflow in the Denver area?
  2. How does the alteration to urban baseflow compare to the alteration to urban stormflow in the Denver metropolitan area?

2. Methods

To address research question 1, this study assessed the contributions of tap water and precipitation to urban streamflow in the Denver metropolitan area through isotopic analyses. We collected baseflow water samples from 13 urban watersheds, as well as from precipitation collectors and tap water sources. Baseflow conditions were identified using real-time streamflow data and a digital filtering method to focus on non-storm contributions. The isotopic composition (δ²H and δ¹⁸O) of water samples was analyzed to quantify the relative contributions of tap water and precipitation using a two-endmember mixing model. In our study area, all municipal tap water providers sourced their supply from high-elevation snowmelt originating in the Rocky Mountains west of Denver—except for the Centennial Water and Sanitation District, which supplemented surface water with local groundwater. For this reason, tap water as an endmember was considered a proxy for snowmelt-fed water. The precipitation endmember represented local precipitation that could contribute to groundwater recharge and eventually stream baseflow discharge. Tap water fractions were further separated into contributions from lawn irrigation return flow (LIRF) and infrastructure water loss. We conducted an uncertainty analysis to quantify errors in the calculation of streamflow separation and tap water partitioning. Further, we investigated the relationship between tap water contributions and watershed characteristics. Finally, to address research question 2, we compared monthly streamflow, stormflow, and tap water contributions to streamflow to understand how tap water contributes to the overall hydrologic system.

2.1. Study area and site selection criteria

In the Denver, Colorado, USA metropolitan area, average annual precipitation is 396 mm (15.58 in), annual average high temperature is 17.8ºC (64ºF), and average annual low temperature is 2.2ºC (36ºF) (U.S. Climate Data, 2022). There were 13 urban watersheds selected for isotope sampling (Fig 1). The urban watersheds had USGS stream gages and isotope samples were taken near the gages with permission from USGS. We used 2 precipitation samplers in the Denver metropolitan area, 17.2 km apart from each other (Fig 1) with permission from the City of Westminster and an anonymous urban farm. We collected tap samples from a total of 75 locations such as gas stations, restaurants, and superstores (Fig 1). There are six water providers in the study area, with Denver Water being the largest (Fig 1).

thumbnail
Download:
Fig 1. (a) Study area showing the watersheds sampled for isotope analysis, water providers, and sampling locations, and (b) Red star representing the location of Denver, Colorado on US Map.

https://doi.org/10.1371/journal.pwat.0000299.g001

The urban watersheds were delineated previously by Fillo et al. [17] using the 10 m digital elevation model (DEM) and the National Hydrography Dataset [35,36]. The mean elevation of the watersheds ranged from 1660 m to 1892 m above sea level [36]. The drainage area of the watersheds ranged from 5.6 km2 to 63.3 km2 for urban watersheds, as extracted from ArcGIS Pro. Slopes ranged from 2% to 9.3% for urban watersheds calculated in ArcGIS Pro. Our study watersheds did not have any wastewater discharges and there was no significant contribution from deep groundwater upwelling in the streams being studied [17].

2.2. Stream, tap, and precipitation sampling for isotope analysis

We collected and analyzed 402 urban stream, 522 tap, and 38 precipitation samples over the three-year period (2019, 2021, 2022). We sampled stream baseflow and tap water every week alternating weeks between northern and southern sites (Table 1). Since we wanted to focus on stream samples of baseflow, the weather was observed to see if there were any precipitation events before the day of sampling or during the sampling day. The USGS real-time streamflow data [37] was also used to check streamflow conditions before taking the samples. We used narrow mouth HDPE 60 mL bottles for collecting stream and tap samples. In the stream, samples were collected by submerging the bottles and tightening the mouth under water. Most of the tap sample collection sites were manually operated faucets. Most of the time, cold water was collected, but in a few instances, tap water was hot. There was no convenient way to submerge the bottle underwater while sampling for tap samples. If there were any bubbles in the bottle, the bottle was shaken, and efforts were made to take the samples without bubbles.

thumbnail
Download:
Table 1. Streamflow sample and precipitation sample collection sites. Water provider acronyms: CA: City of Arvada; CG: City of Golden; CMWC: Consolidated Mutual Water Company; CW: City of Westminster; DW: Denver Water; CWSD: Centennial Water and Sanitation District.

https://doi.org/10.1371/journal.pwat.0000299.t001

Stream, tap, and precipitation samples were quickly transferred to a cooler to minimize evaporation during transport, and next samples were transferred to the refrigerator and stored at 4°C at Colorado State University. The 60 mL HDPE bottles were washed and dried carefully after transferring water samples for lab analysis to use them for further sampling.

2.3. Precipitation sample collection

Precipitation samples were collected once a month from two volume-weighted precipitation samplers (Rain Sampler RS1, Palmex Ltd.). The samplers were designed to prevent evaporation and collect rain samples monthly in a 3-liter plastic bottle through the funnel and the intake tube of the collection system [38]. There were no tall buildings or large trees near our precipitation sampler so that the samples were taken to represent the local precipitation. We conducted precipitation sampling following the precipitation sampling guidelines from Global Network of Isotopes in Precipitation (GNIP) [17]. Precipitation samples collected on 16 June 2022 were excluded from our analysis due to a collection error. Also, we could only collect the precipitation samples from P01 (Anonymous urban farm) for September 2022 and October 2022 as we found no water in P02 (Tree Limb Recycling Center) precipitation sampler.

2.4. Sample preparation and lab analysis

We used the University of Utah SIRFER lab facilities in 2019 and 2021 for isotope () analysis using the Picarro L2130i Laser Water Isotope Analyzer. We transferred the water samples from 60 mL bottles to 1.8 mL glass vials (Thermo Fisher Scientific, Waltham, MA). We used a syringe and 0.2 μm sterile syringe filters (Thermo Fisher Scientific, Waltham, MA) for filtering the water samples. The vials were sealed by the septa caps using a crimper tool. Then, water samples were transported to SIRFER. The samples were kept in the cooler and shipped overnight so that no evaporative events occurred during the shipments, following the same process as in [17].

In 2022, Colorado State University’s EcoCore Analytical Facility was used to analyze the samples for δ2H and δ18O values with the Picarro L2130i Laser Water Isotope Analyzer (the same instrumentation used at SIRFER). We used sterile syringes (Thermo Fisher Scientific, Waltham, MA) and 0.22 μm syringe filters (Tisch Scientific, Cleves, Ohio) for filtering the samples before introduction into the Picarro. We used 9 mm glass screw thread vials (Thermo Fisher Scientific, Waltham, MA) and 9 mm autosampler vial screw thread caps with septa (Thermo Fisher Scientific, Waltham, MA) for transferring the samples from 60 mL bottles. Finally, we set up the samples in the isotope analyzer where samples were injected seven times. We ignored the first three injections to minimize the previous sample memory effects.

For EcoCore, the typical δ18O uncertainty (2 sigma) was less than 0.2‰, and the δ2H uncertainty (2 sigma) was less than 0.6‰, measured as the reproducibility of a given injection. The δ18O uncertainty (1 sigma) for EcoCore was similar to SIRFER (< 0.1‰), but the δ2H uncertainty (1 sigma) was larger (< 1‰) than SIRFER because we used a much larger range of δ2H to constrain the standards line.

We obtained the results in the form of an isotope ratio δ relative to the Vienna Standard Mean Ocean Water (VSMOW) for the isotope analysis [39].

(1)

where and are the or ratios for and in normal and VSMOW standard condition.

To assess consistency across laboratories, we reanalyzed the same 20 samples (tap, precipitation, and stream water) in 2022 at the EcoCore Lab at Colorado State University, which had previously been analyzed at the SIRFER Lab at the University of Utah. The and values from EcoCore were more negative, with the percentage differences of up to 6% (-1.42 to -11.68 ; median - 4.98 ) for and up to 16% (-0.57 to -3.40 ; median -1.49 ) for . Only data was used for the calculations and analysis in this study.

2.5. Baseflow analysis

We verified whether the streams were in baseflow condition before conducting the two end member analysis, as our goal was to identify contributors to stream baseflow. Many approaches have been developed to define baseflow such as digital filter methods, graphical methods, mass balance methods, and numerical simulation methods [40–42]. Digital filter methods have been used extensively for reproducibility; however, digital filter parameters can vary with size and type of watersheds [42]. In our study, we used the BaseflowSeparation function in R package ‘EcoHydRology’ [43]. We used a digital filter parameter of 0.99 along with 3 passes to separate baseflow and quickflow [34,44]. Then, we used threshold values considering four streamflow metrics: streamflow rate, quickflow, instantaneous quickflow minus minimum quickflow of previous 6 hours, and instantaneous quickflow minus minimum quickflow of subsequent 12 hours for each of the watersheds [34,44]. We considered the same threshold values from Wilson et al. [34] in these same watersheds. If any streamflow obtained from the USGS gaging stations exceeded threshold values of one or more metrics, we considered streamflow to not be in baseflow condition.

2.6. Data processing and comparison with GMWL and LMWL

The obtained isotopic values from the lab were plotted against the Global Meteoric Water Line (GMWL) and the Local Meteoric Water Line (LMWL). Craig [45] first documented GMWL that represents the global linear relationship between hydrogen (δ2H) and oxygen (δ18O) stable isotopic ratios in meteoric waters and the equation for GMWL is [45]. Whereas, LMWL corresponds to the regional long-term correlation between the stable isotopic ratios of hydrogen (δ2H) and oxygen (δ18O) [46]. Harvey [47] conducted a study on of precipitation in northeastern Colorado and found an equation of local meteoric water line (LMWL) as .

2.7. Precipitation and tap water fraction by two end member mixing analysis

We used two endmember mixing analysis for partitioning precipitation and tap water contributing to streamflow. Here precipitation and tap water were the two end members. We selected stream samples for analysis following the criteria from Fillo et al. [17], where (1) there were at least 6 or more tap samples collected in the antecedent period (14 days prior to streamflow sampling) from the water provider serving the watershed of that stream sample, (2) average tap values (‰) over the antecedent period were more negative than streamflow , (3) average precipitation values (‰) were less negative than streamflow , (4) area-normalized streamflow depth (mm/d) was greater than estimated area-normalized infrastructure depth (mm/d), and (5) uncertainty associated with precipitation fraction and tap fraction did not exceed 30%. The following equations were used to determine the fractions of tap and precipitation water [48].

(2)(3)

Here, and were precipitation and tap water fractions contributing to streamflow and , , and were tap, precipitation, and stream isotope ratios.

2.8. Infrastructure water loss determination within the watersheds

Infrastructure water loss was calculated in this section which was needed in the next section to estimate the LIRF to the streams. We collected the infrastructure water loss reports for each watershed from the Colorado Water Conservation Board’s Water Efficiency Data Portal [49]. Infrastructure water loss data was available from 2013 to 2020 for most of the water providers. The metered potable treated water was subtracted from distributed potable treated water to obtain the calculated loss of potable treated water. The City of Golden’s 2014 and 2015 loss data was excluded from the calculation because potable treated distributed water was found to be less than the metered water. For Centennial Water and Sanitation District, there was no data available for 2016, 2019, and 2020; and water loss data for 2020 was missing for City of Arvada as well. We assumed that loss occurred uniformly over the service area. We divided the calculated loss value by the service area and took the average of all available years to obtain the area normalized mean daily loss (mm/d) and weighted standard deviation (mm/d) for each water provider. We excluded the 2020 water loss data when calculating the infrastructure water loss for 2019. Then the percentage of watershed area contributed by water providers was multiplied by the area normalized mean daily loss contributions (mm/d) of each water provider contributing to watersheds to obtain the area normalized mean daily loss contributions (mm/d) of each watershed. The fraction of infrastructure water loss, was then obtained dividing the daily infrastructure loss depth (loss) by area-normalized mean daily streamflow (mm/d). All these parameters were shown in Table A in S1 Text and Table 1.

2.9. Tap water partitioning into lawn irrigation return flow and infrastructure water loss

The two end member mixing model provided us with precipitation and tap water fractions contributing to urban streamflow. We hypothesized that tap water can be sourced either from water infrastructure loss (e.g., leaking drinking distribution pipes) or from excess lawn irrigation contributing to the stream through subsurface or overland pathways. In this section, we calculated the LIRF to the streams by subtracting the fractions of infrastructure water loss in the streams, from the tap water fractions in the stream, .

(4)

2.10. Calculating fraction uncertainty

Uncertainty may arise from analytical errors and measurement errors in endmember mixing analysis, which contribute to uncertainty in the estimates of major streamflow sources. We calculated the associated uncertainty and the contributing fractions for each of the end members to determine the overall uncertainty in the streamflow separation estimates [25]. We followed the uncertainty calculations from [17] and [48],

(5)

where is the 70% confidence limit for uncertainty associated with and , while , , and denoted the estimated uncertainty related to the precipitation, tap, and stream samples, respectively [17,48]. The uncertainty in identifying tap water as either lawn irrigation or infrastructure loss was determined by the temporal variability in infrastructure loss values. The 70% confidence limit for uncertainty in infrastructure loss was calculated by multiplying the standard deviation of infrastructure loss from 2013 to 2021 by the corresponding 70% t-value [17,35,48]. Since we did not have data on the spatial distribution of infrastructure water loss, it was assumed to be consistent across the service area and throughout the year, equated entirely to infrastructure leakage contributing to streamflow, despite the potential for unauthorized use, inaccurate metering, and leakage leading to evapotranspiration. Infrastructure loss was used as a proxy for infrastructure leakage due to a lack of detailed estimates, and a criterion was applied to ensure that infrastructure loss contributions did not exceed total streamflow. The analysis further assumed that contributions from other potential sources of tap water, such as wastewater pipe leakage and soil-wetting, were insignificant and could be disregarded [17].

2.11. Relationship between tap water and watershed characteristics

We compared tap water contributions to streamflow with various watershed characteristics using simple linear regression, including imperviousness, latitude, longitude, drainage area extracted from ArcGIS Pro, mean elevation and slope calculated in ArcGIS Pro [36], percentage of open space, development and grassland percentage from NLCD [50]. R-squared values were used as a measure of the explanatory power of each watershed characteristic. Additionally, we analyzed consumptive water use data obtained from Denver Water for its service area by ZIP code, covering the years 2019–2022 and organized by billing cycle. Area-weighted averages for annual and summer consumptive water use, measured in millimeters, were calculated to assess water consumption across different watersheds.

2.12. Comparison of urban streamflow, grassland streamflow, stormflow to tap water contribution

To put the tap contribution to streamflow in context, we compared it with the area-normalized mean monthly streamflow and stormflow of urban and grassland streams. For urban streamflow, we used the same watersheds used for isotope sampling shown in Fig 1, except we did not use U02 as that stream gage does not have as complete of a streamflow record. The urban streams included have a range of drainage area of 5.6 km2 to 63.3 km2 and range from 22%-44% impervious surface cover (Table 1). For grassland streamflow, we used two sources of data. Because of the extensive urbanization in the Denver region, there is only one watershed gaged by USGS below 10% impervious surface cover within a similar elevation range as the urban watersheds – First Creek at Bel Buckley Rd, at Rocky Mountain Arsenal, CO. This watershed NE of Denver is 76 km2 and has 8.9% impervious surface cover [34]. We also used the small grassland watersheds NW of the Denver in Rocky Flats, gaged by the US Department of Energy (DOE) (B5INFLOW, GS10, GS12, GS13, GS33, SW027, SW093 – see map in Wilson et al. [34]) which range in size from 0.73 km2 to 1.41 km2 and from 0.8 to 7.3% impervious surface cover [34].

Urban and grassland stormflow data in the same study watersheds were previously analyzed from 2013 to 2020 using a semi-automated method to pair rainfall and 15-minute streamflow by Wilson et al. [34]. Using the same approach as Wilson et al. [34], we analyzed 2021–2022 stormflow events. We used updated DOE streamflow data from Rocky Flats, USGS streamflow data for urban streams, rainfall from the Mile High Flood District rain gage network, as well as updated snow dates using NDSI (Snow Index) from MODIS Terra 1 km Daily from ClimateEngine.org for the entire Denver metro area. We used monthly analyses to bring datasets of different time resolutions to a single time scale for comparison. Stormflow depth is over a storm, and there are varying numbers of storms per month, which were summed up for comparison to streamflow and tap values.

To convert tap water contributions that were estimated on a daily time scale for one or more days each month (Section 2.7) to monthly values, we took the mean of all tap contribution values (mm/day) in a month and multiplied by the number of days in that month. We used streamflow data from 2019 to 2022 for grassland and urban streams, and 2019, 2021, and 2022 for tap water contributions in our analysis.

3. Results

3.1. Isotopic composition of water samples

We compared our isotope measurements with the GMWL and LMWL to gain insights into local hydrological systems and the factors affecting isotopic composition. Tap, stream, and precipitation δ2H and δ18O values were observed below the GMWL and LMWL (Fig 2), which could result from air temperature differences, evaporative processes, and the seasonality of precipitation [39,47,51]. Precipitation isotopic values closely followed the GMWL and LMWL in the more negative δ2H and δ2O value range but deviated below these lines in the less negative values (Fig 2c). We found a slope of 7.1 and intercept of 7.9 when constructing a regression line using our precipitation data where Harvey [47] found a slope of 7.86 and intercept of 7.66 for northeastern Colorado. Stream samples (Fig 2a) and tap samples (Fig 2b) were shifted to the right of the GMWL and LMWL lines in the less negative δ2H and δ2O value range, indicating more evaporative losses. In contrast, δ2H and δ2O values of stream and tap samples exhibited close alignment with the lines in the more negative side.

thumbnail
Download:
Fig 2. δ2H and δ2O values of (a) Stream (b) Tap, and (c) Precipitation against the Global Meteoric Water Line (Blue line) and the Local Meteoric Water Line (Red line) to gain insights into the local hydrological systems and the processes influencing isotopic composition.

https://doi.org/10.1371/journal.pwat.0000299.g002

3.2. Precipitation isotopes variation

In this section, we analyzed how precipitation isotopes (δ2H and δ2O) varied over time and compared our values with Global Network of Isotopes in Precipitation (GNIP) database [52] and Waterisotopes database [53]. Our measured average monthly δ2H values from the collected precipitation samples and from the GNIP database using the WISER data portal showed the same general patterns [52] (Fig 3). The magnitude and trends of measured average monthly δ2H values also matched the calculated precipitation δ2H values derived from raster data obtained from Waterisotopes Database [53] using ArcGIS Pro 3.0.0 software (Fig 3). We did not collect precipitation samples for the months of November and December in 2019, 2021, and 2022. We observed an increasing trend towards less negative δ2H values from February to July where an increasing trend was observed up to August for the calculated values obtained from GNIP and Waterisotopes Database. Then average monthly δ2H values started to descend towards more negative δ2H values, with an outlier in September. Higher temperature causes larger evaporation and greater isotope fractionation [54], which fits with the observed pattern.

thumbnail
Download:
Fig 3. Average δ2H values of local precipitation measured from our collected samples in 2019, 2021, and 2022, compared with those collected from Global Network of Isotopes in Precipitation (GNIP) database (2009-2015) [52], and calculated from Waterisotopes Database (2002 – 2018) [53].

https://doi.org/10.1371/journal.pwat.0000299.g003

3.3. Tap water variation

This section examined the variation in tap water sources and isotopic composition among water providers, highlighting their reliance on high-elevation surface water from the Rocky Mountains and the spatial differences in δ²H and δ¹⁸O values. Most tap water providers in the study area sourced their supply from the Rocky Mountains west of Denver. Only the Centennial Water and Sanitation District also relied on groundwater in addition to mountain water. Centennial Water and Sanitation District had 51 deep aquifer wells along with 3 raw water surface reservoirs and 8 potable water reservoirs as sources of water [55], and tap water with less negative mean isotopic values (δ2H and δ2O) than other water providers (Fig 4). All other water providers used surface water sources. Denver Water collected approximately 50% of its tap water from tributaries of the Colorado River situated on the west of Continental Divide and rest of the water from the South Platte River and its tributaries located on the east side of Continental Divide where it stored water in 12 major reservoirs [56]. Consolidated Mutual Water Company has 90% of source water from Clear Creek watershed accompanied by high elevation mountainous reservoirs that accumulate runoff from snowmelt, and the remaining 10% water came from Denver Water [57]. Consolidated Mutual Water Company and Denver Water showed larger spreads in δ2H and δ2O values (Fig 4). For Denver Water, variation of mean elevation (2,622 m – 3,555 m) of water sources and multiple sources as well as the large distribution areas could be explain larger spread in δ2H values [17]. The City of Golden, Consolidated Mutual Water Company, City of Westminster, and Denver Water showed similar mean δ2H values. The City of Arvada collected its water from two surface water sources: Denver Water’s North System and Clear Creek [58]. The City of Golden diverted water from Clear Creek and its tributaries [59]. The City of Westminster also accumulated its water by snowmelt from Clear Creek and its tributaries to Standley Lake [60]. Although City of Golden and City of Westminster collected their water from the same watershed, they showed different mean δ2H values over the three-year periods. The δ2H (‰) values of tap water ranged from -133.7‰ to -72.6‰ where mean was -112.7‰. Spatial variation in tap water was shown in S13 Fig.

thumbnail
Download:
Fig 4. Boxplots showing δ2H values of each water provider over the three-year period (2019, 2021, 2022) indicating differences due to water sources.

https://doi.org/10.1371/journal.pwat.0000299.g004

3.4. Stream isotopes variation

In this section, we explored seasonal and annual variation in stream isotopic composition (δ²H) and its relationship with tap water and precipitation to understand streamflow sources and mixing processes. We observed distinctive ranges of isotopic (δ2H) values of stream, tap, and precipitation, especially in late summer (Fig 5). Stream water δ2H (‰) was bounded by more negative tap δ2H (‰) and less negative precipitation δ2H (‰). In 2019, July and August exhibited less negative stream water δ2H (‰) values compared to other months. But August and September showed less negative δ2H (‰) values for 2021 and 2022. We found that 47 out of 61 of our streamflow dates satisfying all criteria for endmember mixing analysis (Section 2.7) in August and September of 2019 and 2020. In 2022, we identified 69 streamflow dates, meeting all the criteria for endmember mixing analysis spanning from May to September due to our increased sample collection that year.

thumbnail
Download:
Fig 5. Variation of δ2 H isotopes over 2019, 2021, and 2022 for urban stream, tap, and precipitation samples.

https://doi.org/10.1371/journal.pwat.0000299.g005

3.5. Tap and Lawn irrigation return flow contribution analysis

We observed the highest mean LIRF percentage in August 2019 and the lowest mean LIRF percentage in September 2021. We did not find any streamflow dates during April in any year for endmember mixing analysis fulfilling our criteria (Section 2.7). Table 2 presents data on the percent contribution of tap water and LIRF across 2019, 2021, and 2022. Streamflow was made up of a mean of 72% to 80% tap water and within that, a mean of 42% to 57% LIRF (Table 2). The percent contributions for both sources varied widely, with tap water ranging from 38%-98% and LIRF from 5%-85% over the years (S1 Fig (a), S2 Fig, S4 Fig). The uncertainty associated with these contributions ranged from ±2% to ±29% and was higher in 2021 and 2022 (S1 Fig (a), S2 Fig, S4 Fig). Tap water contributions ranged from a minimum of mm in 2021 to a high of mm in 2019, with uncertainties between mm and mm (Table C in S1 Text, S1 Fig (b), S3 Fig, S5 Fig). LIRF contributions ranged from a minimum of mm in 2021 to a high of mm in 2019, with uncertainties ranged from mm to mm (Table C in S1 Text, S1 Fig (b), S3 Fig, S5 Fig). The mean contribution of tap water and LIRF to the urban streamflow over a three-year period is shown in Fig 6.

thumbnail
Download:
Table 2. Contribution to streamflow from tap, LIRF, precipitation, and infrastructure water loss for each year. The percent contribution is the entire range (minimum to maximum) over all sampling locations and dates for that year where end-member mixing analysis was performed. The mean percent contribution is the mean over all the samples for which end-member mixing analysis was performed. Median percent contribution is over the same samples with the inter-quartile range given in parenthesis.

https://doi.org/10.1371/journal.pwat.0000299.t002

thumbnail
Download:
Fig 6. (a) Median and 1st and 3rd quartiles contribution of tap water and lawn irrigation return flow to the urban streamflow for all samples and dates and (b) Mean contribution of tap water and lawn irrigation return flow to the urban streamflow for 2019, 2021, 2022.

https://doi.org/10.1371/journal.pwat.0000299.g006

We observed an increasing trend in lawn irrigation return flow percentage with increasing imperviousness (R2 = 0.14, p-value = 10-5) (Fig 7). We applied simple linear regression on the tap water or lawn irrigation return flow against the watershed characteristics consumptive water use by ZIP code, watershed area, grass percentage, mean slope, minimum and maximum elevation, percent development, latitude, and longitude. We did not find any significant relationship between any of these individually and the tap water or lawn irrigation return flow percentage contributions (S6 to S12 Figs).

thumbnail
Download:
Fig 7. The relationship between lawn irrigation return flow percentage and watershed imperviousness.

https://doi.org/10.1371/journal.pwat.0000299.g007

3.6. Comparison of urban streamflow, grassland streamflow, stormflow to tap water contribution

Grassland streamflow magnitude was almost always lower than urban streamflow on a monthly basis (Fig 8). The exception to this pattern was May 2021, when the grasslands watersheds not only had higher streamflow than the urban ones, but also had the highest monthly streamflow during the 2019 – 2022 time period. On an annual average basis, urban streamflow was 2.5 times larger (94 mm) than grassland streamflow (37 mm). In addition to almost always having less streamflow, the grassland streamflow also exhibited distinct seasonal patterns compared to the urban streams. The grassland streamflow peaked earlier in the spring (April or May) and then rapidly decreases with minimal streamflow by August and the rest of the fall. The urban streams in contrast did not have as large of a seasonal cycle, with a later peak in streamflow (mid-summer) and only a gradual decline in streamflow in the early fall. The difference in streamflow between the grassland and urban streams was most pronounced in late summer and early fall when the urban streams were maintained at a higher level and did not go dry whereas the grassland streams had little to no streamflow. In terms of interannual variability, the urban streams maintained relatively consistent streamflow across years, but the year-to-year variability in streamflow was high in the grassland streams. For example, in 2021 the grassland streams had the highest streamflow for the entire time period, and then in 2022 there was only a short period of flow in the spring with most of the summer and fall seeing dry grassland streams.

thumbnail
Download:
Fig 8. Comparison of monthly streamflow (area-normalized, in mm) from January 2019 – December 2022.

For each month, a boxplot is shown for both grassland and urban streamflow. The boxplot for grassland streamflow shows the median, quartiles, and outliers across all grassland watersheds for that month (and similarly for the urban boxplot).

https://doi.org/10.1371/journal.pwat.0000299.g008

Comparing stormflow and streamflow during summer (Fig 9), grassland stormflow was small and almost zero in 2020 and 2022. There were some months where there was no grassland stormflow as there were no storms recorded in the grassland streams, even though there was sometimes baseflow contributing to grassland streamflow. This was not the case in the urban streams. Urban stormflow was always smaller than urban monthly streamflow, but in some cases was larger than total grassland streamflow and a significant fraction of urban streamflow. Examining the available area-normalized monthly tap contributions, we observed that these remained between the grassland and urban streamflow and exhibited closer values to urban streamflow. The earlier analysis (Section 3.5) examined the percent contribution of tap water to streamflow on sampled days, where all sampled days were baseflow days (i.e., not elevated flow in response to a storm event). But now, scaling those daily tap contributions to monthly values, tap water made up a large part of total urban streamflow, not just baseflow. In many months, the tap contribution was larger than urban stormflow (e.g., August 2019, September 2019, September 2021, June 2022) or the two were similar (e.g., July 2021, August 2021, August 2022, September 2022), whereas in a few months urban stormflow was larger than tap water (e.g., in May 2022, July 2022).

thumbnail
Download:
Fig 9. Summer monthly streamflow for grassland and urban streams (also shown for the full calendar year in Fig 8), along with grassland stormflow, urban stormflow, and tap contributions to streamflow (only in months where estimated), all in values of mm over each month.

Summer is examined as tap contributions were only quantified in the summer. The boxplots show the median, quartiles, and outliers across: grassland watersheds, stormflow in grassland watersheds, tap sampling sites within that month, urban watersheds, and stormflow in urban watersheds, respectively.

https://doi.org/10.1371/journal.pwat.0000299.g009

4. Discussion

The results of this study provided insights on the dynamics and sources of urban and grassland streamflow in the Denver metropolitan area, which may inform hypotheses of behavior in other semi-arid and arid urban areas. Using isotopic analysis techniques, we gained an understanding of the contributions of different water sources to urban baseflow which is a growing concern due to the impacts of urbanization and climate change [15,16,18]. The study extended prior work by Fillo et al. [17], which analyzed one year of data, by incorporating two additional years (2019, 2021, and 2022), thereby offering a more comprehensive understanding of streamflow sources. The results of the three-year analysis were largely similar and supported the findings of previous work [17], indicating that the variability between years in tap and lawn irrigation return flow contributions to streamflow in the region are smaller than the variability between sampling dates and locations in a single year (Fig 6). Tap water consistently dominated streamflow contributions across all years (2019, 2021, and 2022), with mean contributions ranging from 72% to 80%, and LIRF formed a significant proportion within tap water contributions (42% to 57%). Quantifying LIRF contributions can provide a first-order estimate of how urban stream baseflow may decline with more efficient irrigation practices. Urban streams in the Denver area will have less flow with less excessive urban irrigation, and potentially be dry for parts of the year, returning to a streamflow regime closer to their pre-development grassland conditions (Fig 8). This has implications for water quality, as LIRF may carry substantial nutrient loads [61–63].

Similar patterns of anthropogenic influence on urban hydrology have been observed globally. Various processes, such as infrastructure leakage and excessive irrigation—can lead to elevated baseflow in urban streams [64]. In Southern California, for instance, lawn irrigation and leaking infrastructure were found to increase dry-season flows in streams [65]. In Santiago, Chile and Perth, Australia, inefficient urban water use contributed significantly to groundwater recharge and stream baseflow [15,66]. Likewise, treated wastewater and shallow groundwater dominated streamflow in a heavily urbanized catchment in Berlin, Germany [32]. Together, these findings underscore how human water use can fundamentally alter stream hydrology across varied climates.

Our isotopic analysis also revealed key environmental drivers [30,67,68]. Deviations from the Global Meteoric Water Line (GMWL) and Local Meteoric Water Line (LMWL) observed in Fig 2 indicated the impact of evaporative processes, temperature fluctuations, and seasonal precipitation on isotopic signatures [47,51,68]. For instance, Centennial Water and Sanitation District’s reliance on groundwater resulted in less negative δ²H and δ¹⁸O values in tap water isotopic composition, whereas other providers in the Denver area primarily used mountainous surface water sources. These spatial isotopic variations highlight how water source types and infrastructure contribute to regional heterogeneity.

We observed a positive relationship between LIRF and watershed imperviousness (Fig 7), although imperviousness explained only part of the variation. This may be because imperviousness serves as a proxy urbanization, where more urbanized areas also have a greater fraction of pervious landscapes highly irrigated. Another possible reason is as imperviousness increases, a greater portion of applied irrigation water runs off into streams.

Previous work in this same region had the surprising finding that the volume of stormflow was similar between urban and grassland streams when compared on an event-scale [34], which differed from the increases in stormflow compared to pre-development found in other urban areas [69,70]. When examining those storm events that were observed in streamflow, grassland streams in the Denver metro area had relatively large responses [34], and grassland streams showed a comparatively large amount of annual variability in streamflow (Fig 8). From the present work, we can also see however that examined on a monthly scale, grassland stormflow was consistently lower on an area-normalized basis compared to urban stormflow. Grassland streams may have big responses to storms occasionally, but they did not respond to storms nearly as often [34]. Grassland and urban streams had similar stormflow when just considering those events that led to a stormflow response. But looking at storms that did not produce a response, we see the pattern that is observed elsewhere with urbanization, but in a more extreme way: urban streams in the Denver area had more stormflow than grassland streams considering that many storms did not produce any response in the grassland watersheds. The biological impact of increasing baseflow with urbanization is thought to be larger in streams that are naturally ephemeral [71]. In the watersheds studied, the increase in baseflow combined with the effect of conventional stormwater drainage which led to increased stormflow.

The total streamflow in urban streams in the Denver area was larger than that of grassland streams in the area. We can see from this work that this is not just because of the commonly attributed change in streamflow from impervious surfaces leading to more stormwater runoff, although that plays a role (Fig 9). The larger contributor in most summer months was contributions to streamflow from leaking tap water and lawn irrigation return flow. All of these were leading to profound changes in streamflow regime between grassland and urban streams (Fig 8).

This study provides actionable insights for water-scarce urban regions. Reducing potable water losses through targeted infrastructure maintenance and leak detection can help conserve limited water supplies and prevent unintended contributions to streamflow. Improving irrigation efficiency—through strategies such as xeriscaping, optimized scheduling, and smart irrigation systems—can further reduce lawn irrigation return flows [72]. In urban areas such as Denver, a shift toward more efficient water use may lead to reductions in baseflow and a return to more ephemeral stream conditions, with important implications for water quality and aquatic ecosystems.

Beyond these management considerations, our study demonstrates the broader utility of stable isotope techniques in complex urban environments. The ability to trace contributions from infrastructure water loss, irrigation return flows, and precipitation provides a powerful diagnostic tool for understanding altered hydrologic regimes and informing targeted interventions. As urbanization continues to reshape the water cycle, the application of isotope hydrology can play a critical role in supporting integrated, evidence-based approaches to urban water resource management in diverse climatic and infrastructural settings.

4.1. Assumptions and limitations

To conduct this analysis, we made some assumptions. One was that we took the average of infrastructure water loss over the whole area of water providers, assuming the same infrastructure water loss across the area. There should be spatial variation in infrastructure water loss, but we did not have any data or measurement technique to capture the spatial variability of infrastructure water loss. Second, we took 14 days as the antecedent period for tap water and precipitation to reach the streams in our endmember mixing analysis, without further information on the transit time for tap water or precipitation-derived groundwater to reach the streams. Third, there was no way to separate the tap water into lawn irrigation water loss or infrastructure water loss as they had similar isotopic values. For this reason, they could not be used as separate endmembers. Lastly, in estimating area-normalized consumptive water use across the watersheds, we assumed that the data provided by Denver Water was uniformly distributed across each ZIP code area. This assumption was necessary because the available data only provided total consumptive water use for each ZIP code, without any spatial distribution details within those areas. As a result, when clipping the data for specific watersheds, this assumption may introduce potential errors in the calculations.

5. Conclusions

We sought to understand the sources of increased flow observed in the urban streams as compared to grassland streams in the Denver area. To achieve this, we combined isotopic analysis of 402 urban stream, 522 tap, and 38 precipitation samples with assessment of streamflow and stormflow variability over the same period. The contributions of different sources to urban baseflow were largely similar between 2019, 2021, and 2022 (Fig 6). First, the largest contributor to baseflow was tap water (median: 77% of streamflow, 67–86% for first and third quartile values), with the remainder coming from precipitation-derived groundwater during this time period. Within tap water, lawn irrigation return flow was largest (median: 49% of streamflow, 36–62% for first and third quartile values). The remainder of tap water was infrastructure water loss. Second, tap water showed variations in δ2H values across different water providers. Lawn irrigation return flow percentage exhibited a positive trend with increasing imperviousness (Fig 7). Third, grassland stormflow was consistently small (and sometimes non-existent when there was no grassland response to storms) compared to urban stormflow in the Denver metro area (Fig 9). For urban streams, the tap contribution to streamflow is often larger than the monthly stormflow – indicating that the large increase in streamflow in urban streams results both from urban baseflow and urban stormflow increases. These findings demonstrate that human water use—particularly through irrigation return flow and infrastructure loss—profoundly alters streamflow regimes in urban settings. As cities pursue more efficient outdoor water use and adopt nature-based stormwater strategies, understanding the dynamics of urban streamflow will be critical to anticipating hydrologic change and manage urban water challenges. This study provides a foundation for future analyses investigating how urban streamflow can respond to reductions in outdoor water use and ongoing urban development.

Supporting information

S1 Text.

Table A. Infrastructure Water Loss Calculation for 2019. Table B. Infrastructure Water Loss Calculation for 2021 and 2022. Table C. Estimated contributions to streamflow depth from tap water (Tap), lawn irrigation return flow (LIRF), and precipitation during summer months for selected urban stream sites in 2019, 2021, and 2022. Values are reported in millimeters and represent the range of area-normalized depth contributions observed across all sampling events and sites within each year. The uncertainty ranges reflect uncertainty in the associated endmembers, as derived from Equation 5.

https://doi.org/10.1371/journal.pwat.0000299.s001

(DOCX)

S1 Fig. (a) Streamflow percentage and (b) depth contribution of lawn irrigation return flow (LIRF), infrastructure loss (Loss), and precipitation to urban streamflow for 2019.

https://doi.org/10.1371/journal.pwat.0000299.s002

(TIFF)

S2 Fig. Percent contribution of lawn irrigation return flow (LIRF), infrastructure loss (Loss), and precipitation to urban streamflow for 2021.

https://doi.org/10.1371/journal.pwat.0000299.s003

(TIFF)

S3 Fig. Streamflow depth contribution of lawn irrigation return flow (LIRF), infrastructure loss (Loss), and precipitation to urban streamflow for 2021.

https://doi.org/10.1371/journal.pwat.0000299.s004

(TIFF)

S4 Fig. Percent contribution of lawn irrigation return flow (LIRF), infrastructure loss (Loss), and precipitation to urban streamflow for 2022.

https://doi.org/10.1371/journal.pwat.0000299.s005

(TIFF)

S5 Fig. Streamflow depth contribution of lawn irrigation return flow (LIRF), infrastructure loss (Loss), and precipitation to urban streamflow for 2022.

https://doi.org/10.1371/journal.pwat.0000299.s006

(TIFF)

S6 Fig. Tap water percentage against (a) imperviousness and (b) grass percentage.

https://doi.org/10.1371/journal.pwat.0000299.s007

(TIFF)

S7 Fig. Tap water isotope against (a) latitude and (b) longitude.

https://doi.org/10.1371/journal.pwat.0000299.s008

(TIFF)

S8 Fig. Tap water percentage against (a) mean slope and (b) percent development.

https://doi.org/10.1371/journal.pwat.0000299.s009

(TIFF)

S9 Fig. Stream water isotope against (a) drainage area and (b) imperviousness.

https://doi.org/10.1371/journal.pwat.0000299.s010

(TIFF)

S10 Fig. Lawn irrigation return flow percentage versus grass percentage.

https://doi.org/10.1371/journal.pwat.0000299.s011

(TIFF)

S11 Fig. Lawn irrigation return flow percentage versus (a) latitude and (b) longitude.

https://doi.org/10.1371/journal.pwat.0000299.s012

(TIFF)

S12 Fig. Lawn irrigation return flow percentage versus (a) mean slope and (b) percent development.

https://doi.org/10.1371/journal.pwat.0000299.s013

(TIFF)

S13 Fig. Plot showing variation in tap δ2H values in early and late summer of 2019, 2021, and 2022.

https://doi.org/10.1371/journal.pwat.0000299.s014

(TIFF)

Acknowledgments

We gratefully acknowledge field work assistance by Connor Williams, Claire McWilliams, Samuel Carles Pedroza, Amber Boyle, Liam Milton, and Jillian Lukez. We thank Dr. Jeremy Rugenstein and Dan Reuss for their training and the opportunity to utilize the Picarro isotope analyzer facilities. Special thanks to Dr. Ryan Smith, current advisor to the lead author, for his continued support and encouragement during the completion of this manuscript. We appreciate the financial support received from the Whitney Borland Scholarship, Morton W Bittinger Scholarship, Walter Scott, Jr. Graduate Fellowship from Colorado State University. We also thank George Squibb for sharing streamflow data for Rocky Flats and Stacy Wilson for assisting with streamflow analysis.

References

  1. 1. Brown TC, Mahat V, Ramirez JA. Adaptation to Future Water Shortages in the United States Caused by Population Growth and Climate Change. Earth’s Future. 2019;7(3):219–34.
  2. 2. He C, Liu Z, Wu J, Pan X, Fang Z, Li J, et al. Future global urban water scarcity and potential solutions. Nat Commun. 2021;12(1):4667. pmid:34344898
  3. 3. Kahil MT, Dinar A, Albiac J. Modeling water scarcity and droughts for policy adaptation to climate change in arid and semiarid regions. J Hydrol. 2015;522:95–109.
  4. 4. Dang X, Li L, Fan L. Spatiotemporal variation of household water consumption and coping strategies in water-stressed city of arid and semiarid Northwest China. Front Environ Sci. 2022;10.
  5. 5. Heidari H, Arabi M, Warziniack T, Sharvelle S. Effects of Urban Development Patterns on Municipal Water Shortage. Front Water. 2021;3.
  6. 6. Gober P, Quay R, Larson KL. Outdoor water use as an adaptation problem: insights from North American cities. Water Resour Manag. 2016;30(3):899–912.
  7. 7. Quesnel KJ, Ajami NK. Large Landscape Urban Irrigation: A Data‐Driven Approach to Evaluate Conservation Behavior. Water Resources Res. 2019;55(1):771–86.
  8. 8. Shurtz KM, Dicataldo E, Sowby RB, Williams GP. Insights into Efficient Irrigation of Urban Landscapes: Analysis Using Remote Sensing, Parcel Data, Water Use, and Tiered Rates. Sustainability. 2022;14(3):1427.
  9. 9. Bijoor NS, Pataki DE, Haver D, Famiglietti JS. A comparative study of the water budgets of lawns under three management scenarios. Urban Ecosyst. 2014;17(4):1095–117.
  10. 10. Milesi C, Running SW, Elvidge CD, Dietz JB, Tuttle BT, Nemani RR. Mapping and modeling the biogeochemical cycling of turf grasses in the United States. Environ Manage. 2005;36(3):426–38.
  11. 11. Maupin MA, Kenny JF, Hutson SS, Lovelace JK, Barber NL, Linsey KS. Estimated use of water in the United States in 2010 [Internet]. Circular. U.S. Geological Survey; 2014 [cited 2022 Sep 20]. Report No.: 1405. Available from: https://pubs.er.usgs.gov/publication/cir1405
  12. 12. Yue C, Cui M, Kong X, Watkins E, Barnes M. Landscape irrigation and water conservation in urban areas: An analysis of information-based strategies. HortTechnology. 2022;32(2):213–25.
  13. 13. Imberger SJ, Thompson RM, Grace MR. Urban catchment hydrology overwhelms reach scale effects of riparian vegetation on organic matter dynamics. Freshw Biol. 2011;56(7):1370–89.
  14. 14. Majumdar A, Avishek K. Riparian zone assessment and management: an integrated review using geospatial technology. Water Air Soil Pollut. 2023;234(5):319.
  15. 15. Bhaskar AS, Beesley L, Burns MJ, Fletcher TD, Hamel P, Oldham CE, et al. Will it rise or will it fall? Managing the complex effects of urbanization on base flow. Freshwater Science. 2016;35(1):293–310.
  16. 16. O’Driscoll M, Clinton S, Jefferson A, Manda A, McMillan S. Urbanization Effects on Watershed Hydrology and In-Stream Processes in the Southern United States. Water. 2010;2(3):605–48.
  17. 17. Fillo NK, Bhaskar AS, Jefferson AJ. Lawn Irrigation Contributions to Semi‐Arid Urban Baseflow Based on Water‐Stable Isotopes. Water Resour Res. 2021;57(8):e2020WR028777.
  18. 18. Rolls RJ, Leigh C, Sheldon F. Mechanistic effects of low-flow hydrology on riverine ecosystems: ecological principles and consequences of alteration. Freshw Sci. 2012;31(4):1163–86.
  19. 19. Passarello MC, Sharp JM, Pierce SA. Estimating urban-induced artificial recharge: A case study for Austin, TX. Environ Eng Geosci. 2012;18(1):25–36.
  20. 20. Manago KF, Hogue TS. Urban streamflow response to imported water and water conservation policies in Los Angeles, California. JAWRA J Am Water Resour Assoc. 2017;53(3):626–40.
  21. 21. Kuhlemann LM, Tetzlaff D, Soulsby C. Spatio-temporal variations in stable isotopes in peri-urban catchments: A preliminary assessment of potential and challenges in assessing streamflow sources. J Hydrol. 2021;600:126685.
  22. 22. Pataki DE, Boone CG, Hogue TS, Jenerette GD, McFadden JP, Pincetl S. Socio‐ecohydrology and the urban water challenge. Ecohydrology. 2011;4(2):341–7.
  23. 23. Zhou J, Liu G, Meng Y, Xia C, Chen K, Chen Y. Using stable isotopes as tracer to investigate hydrological condition and estimate water residence time in a plain region, Chengdu, China. Sci Rep. 2021;11(1):2812.
  24. 24. Sprenger M, Carroll RWH, Marchetti D, Bern C, Beria H, Brown W, et al. Stream water sourcing from high-elevation snowpack inferred from stable isotopes of water: a novel application of d-excess values. Hydrol Earth Syst Sci. 2024;28(7):1711–23.
  25. 25. Congjian S, Yaning C, Weihong L, Xingong L, Yuhui Y. Isotopic time series partitioning of streamflow components under regional climate change in the Urumqi River, northwest China. Hydrol Sci J. 2016;61(8).
  26. 26. Du M, Zhang M, Wang S, Chen F, Zhao P, Zhou S, et al. Stable Isotope Ratios in Tap Water of a Riverside City in a Semi-Arid Climate: An Application to Water Source Determination. Water. 2019;11(7):1441.
  27. 27. Ehleringer JR, Barnette JE, Jameel Y, Tipple BJ, Bowen GJ. Urban water – a new frontier in isotope hydrology. Isotopes Environ Health Stud. 2016;52(4–5):477–86.
  28. 28. de Wet RF, West AG, Harris C. Seasonal variation in tap water δ2H and δ18O isotopes reveals two tap water worlds. Sci Rep. 2020;10(1):13544. pmid:32782259
  29. 29. Jameel Y, Brewer S, Good SP, Tipple BJ, Ehleringer JR, Bowen GJ. Tap water isotope ratios reflect urban water system structure and dynamics across a semiarid metropolitan area. Water Resources Research. 2016;52(8):5891–910.
  30. 30. Landwehr JM, Coplen TB, Stewart DW. Spatial, seasonal, and source variability in the stable oxygen and hydrogen isotopic composition of tap waters throughout the USA. Hydrol Process. 2014;28(21):5382–422.
  31. 31. Wallace S, Biggs T, Lai CT, McMillan H. Tracing sources of stormflow and groundwater recharge in an urban, semi-arid watershed using stable isotopes. J Hydrol Reg Stud. 2021;34:100806.
  32. 32. Marx C, Tetzlaff D, Hinkelmann R, Soulsby C. Isotope hydrology and water sources in a heavily urbanized stream. Hydrol Process. 2021;35(10):e14377.
  33. 33. Beal L, Senison J, Banner J, Musgrove M, Yazbek L, Bendik N, et al. Stream and Spring Water Evolution in a Rapidly Urbanizing Watershed, Austin, TX. Water Resources Research. 2020;56(4).
  34. 34. Wilson S, Bhaskar AS, Choat B, Kampf S, Green TR, Hopkins KG. Urbanization of grasslands in the Denver area affects streamflow responses to rainfall events. Hydrol Process. 2022;36(10):e14720.
  35. 35. Fillo NK. Quantifying Lawn Irrigation Contributions to Semi-Arid, Urban Stream Baseflow with Water-Stable Isotopes - ProQuest [Internet]. 2020 [cited 2022 Nov 11]. Available from: https://www.proquest.com/openview/8886da85ea727fa1fd7fc408c1dc9ff5/1.pdf?pq-origsite=gscholar&cbl=51922&diss=y
  36. 36. USDA. USDA:NRCS:Geospatial Data Gateway:Home [Internet]. 2022 [cited 2022 Dec 5]. Available from: https://datagateway.nrcs.usda.gov/
  37. 37. USGS. USGS WaterWatch -- Streamflow conditions [Internet]. 2022 [cited 2022 Dec 19]. Available from: https://waterwatch.usgs.gov/?m=real&r=co
  38. 38. Palmex Ltd. Palmex Rain Sampler RS1 [Internet]. 2022 [cited 2022 Dec 16]. Available from: http://www.rainsampler.com/portfolio-page/rain-sampler-rs1/
  39. 39. Kendall C, McDonnell JJ. Isotope Tracers in Catchment Hydrology - 1st Edition [Internet]. 1999 [cited 2022 Oct 11]. Available from: https://www.elsevier.com/books/isotope-tracers-in-catchment-hydrology/kendall/978-0-444-81546-0
  40. 40. Arnold JG, Allen PM, Muttiah R, Bernhardt G. Automated base flow separation and recession analysis techniques. Groundwater. 1995;33(6):1010–8.
  41. 41. Kang T, Lee S, Lee N, Jin Y. Baseflow Separation Using the Digital Filter Method: Review and Sensitivity Analysis. Water. 2022;14(3):485.
  42. 42. Liu Z, Liu S, Ye J, Sheng F, You K, Xiong X, et al. Application of a Digital Filter Method to Separate Baseflow in the Small Watershed of Pengchongjian in Southern China. Forests. 2019;10(12):1065.
  43. 43. Fuka D, MT W, JA A, TS S, ZM E. Package ‘EcoHydRology’. 2015.
    • 44. Hopkins KG, Bhaskar AS, Woznicki SA, Fanelli RM. Changes in event-based streamflow magnitude and timing after suburban development with infiltration-based stormwater management. Hydrol Process. 2020;34(2):387–403. pmid:32063664
    • 45. Craig H. Isotopic Variations in Meteoric Waters. Science. 1961;133(3465):1702–3. pmid:17814749
    • 46. Putman AL, Fiorella RP, Bowen GJ, Cai Z. A Global Perspective on Local Meteoric Water Lines: Meta‐analytic Insight Into Fundamental Controls and Practical Constraints. Water Res Res. 2019;55(8):6896–910.
    • 47. Harvey FE. Stable hydrogen and oxygen isotope composition of precipitation in northeastern colorado. J Am Water Resources Assoc. 2005;41(2):447–60.
    • 48. Genereux D. Quantifying uncertainty in tracer‐based hydrograph separations. Water Resources Research. 1998;34(4):915–9.
    • 49. Colorado Water Conservation Board. Water Efficiency Data Portal [Internet]. 2022 [cited 2022 Dec 16]. Available from: http://www.cowaterefficiency.com/home_page
    • 50. Homer CG, Fry JA, Barnes CA. The National Land Cover Database [Internet]. Fact Sheet. U.S. Geological Survey; 2012 [cited 2024 Sep 3]. Report No.: 2012–3020. Available from: https://pubs.usgs.gov/publication/fs20123020
    • 51. Clark ID, Fritz P. Environmental isotopes in hydrogeology. Boca Raton: CRC Press; 1997. 342 p.
      • 52. IAEA. IAEA WISER [Internet]. 2022 [cited 2022 Dec 19]. Available from: https://nucleus.iaea.org/wiser/index.aspx
      • 53. Waterisotopes Database. WaterIsotopes.org [Internet]. 2022 [cited 2022 Dec 19]. Available from: https://wateriso.utah.edu/waterisotopes/pages/information/oipc_info.html
      • 54. Zhang S, Bu J, Li C, Xu X, Wang X, Liu Q. Analysis of water source contributions and their impacts on hydrological structural connectivity in plain urban river network areas based on stable isotopes (δ2H, δ18O). Hydrol Process. 2022;36(12):e14782.
      • 55. CWSD. Centennial Water’s Infrastructure | Centennial Water Sanitation District, CO [Internet]. 2022 [cited 2022 Nov 27]. Available from: https://www.centennialwater.org/about-centennial/about-us/infrastructure
      • 56. Adams J. Where does your water come from? [Internet]. Denver Water. 2017 [cited 2022 Nov 27]. Available from: https://www.denverwater.org/tap/where-does-your-water-come
      • 57. CMWC. Consolidated Mutual Water Company | Water Sources and Storage Levels [Internet]. 2022 [cited 2022 Nov 27]. Available from: https://www.cmwc.net/your-water/sources-storage-levels/
      • 58. City of Arvada. 2022 Water Quality Report - City of Arvada [Internet]. 2022 [cited 2022 Nov 27]. Available from: https://arvada.org/residents/services-and-sustainability/water-sewer-general-information/water-report
      • 59. City of Golden. Water Supply | City of Golden, Colorado [Internet]. 2022 [cited 2022 Nov 27]. Available from: https://www.cityofgolden.net/government/departments-divisions/water/water-supply/
      • 60. City of Westminster. Where does your drinking water come from? [Internet]. City of Westminster; 2022 [cited 2022 Nov 27]. Available from: https://www.cityofwestminster.us/News/where-does-your-drinking-water-come-from
        • 61. McPherson TN, Burian SJ, Stenstrom MK, Turin HJ, Brown MJ, Suffet IH. DRY AND WET WEATHER FLOW NUTRIENT LOADS FROM A LOS ANGELES WATERSHED. J Am Water Resources Assoc. 2005;41(4):959–69.
        • 62. Stein ED, Ackerman D. Dry weather water quality loadings in arid, urban watersheds of the Los Angeles basin, California, USA. JAWRA J Am Water Resour Assoc. 2007;43(2):398–413.
        • 63. Toor GS, Occhipinti ML, Yang Y-Y, Majcherek T, Haver D, Oki L. Managing urban runoff in residential neighborhoods: Nitrogen and phosphorus in lawn irrigation driven runoff. PLoS One. 2017;12(6):e0179151. pmid:28604811
        • 64. Christian LN, Banner JL, Mack LE. Sr isotopes as tracers of anthropogenic influences on stream water in the Austin, Texas, area. Chem Geol. 2011;282(3–4):84–97.
        • 65. Townsend-Small A, Pataki DE, Liu H, Li Z, Wu Q, Thomas B. Increasing summer river discharge in southern California, USA, linked to urbanization. Geophys Res Lett. 2013;40(17):4643–7.
        • 66. Sanzana P, Gironás J, Braud I, Muñoz J, Vicuña S, Reyes‐Paecke S, et al. Impact of Urban Growth and High Residential Irrigation on Streamflow and Groundwater Levels in a Peri‐Urban Semiarid Catchment. J American Water Resour Assoc. 2019;55(3):720–39.
        • 67. Bhuiyan SA, Jameel Y, Chartrand MMG, St-Jean G, Gibson J, Bataille CP. Spatial variations in tap water isotopes across Canada: Tracing water from precipitation to distribution and assess regional water resources. PLOS Water. 2023;2(1):e0000068.
        • 68. Meng Y, Liu G, Xiang Q, Liu Y. Spatio-temporal variation characteristics of stable isotopes of tap water and its potential as a proxy for surface water in Sichuan, China. Sci Total Environ. 2024;912:168755.
        • 69. Leopold LB. Hydrology for urban land planning--a guidebook on the hydrologic effects of urban land use. Washington D.C.: U.S. Geological Survey; 1968. 18 p.
          • 70. Bhaskar AS, Hopkins KG, Smith BK, Stephens TA, Miller AJ. Hydrologic Signals and Surprises in U.S. Streamflow Records During Urbanization. Water Resources Research. 2020;56(9).
          • 71. Walsh CJ, Fletcher TD, Vietz GJ. Variability in stream ecosystem response to urbanization: Unraveling the influences of physiography and urban land and water management. Prog Phy Geogr Earth Environ. 2016;40(5):714–31.
          • 72. Mini C, Hogue TS, Pincetl S. Estimation of residential outdoor water use in Los Angeles, California. Landsc Urban Plan. 2014;127:124–35.