Journal of Geophysical Research: Biogeosciences

Abiotic and biotic controls of organic matter cycling in a managed stream



[1] Urbanization often alters the physical, chemical, and biological structure of aquatic ecosystems embedded within them, creating managed ecosystems with different structure and functioning as compared to their unmanaged counterparts. Our work focused on patterns in dissolved organic carbon (DOC) along a managed stream in Phoenix, Arizona. We documented longitudinal changes in DOC concentrations and quality (defined as chemical complexity and measured as specific ultraviolet absorbance at 254 nm, SUVA) along a 66 km stream dominated by treated wastewater effluent. DOC concentrations along the stream declined by an average of 64%, and chemical complexity increased substantially. We posed four hypotheses to explain changes in downstream water chemistry; including hydrologic dilution, microbial mineralization, abiotic sorption to suspended sediments, and photodegradation by ultraviolet (UV) radiation. Only the second and fourth hypotheses represent permanent removal mechanisms. Our data most strongly supported predictions from the dilution hypothesis and microbial mineralization as an explanation for the changes in DOC chemistry. Surface-subsurface water linkages were important but altered from unmanaged streams, as deep groundwater was used to augment surface flows. Variation in the use of groundwater was linked to human decision making and engineering related to water management. Reduction in geomorphic complexity increased the importance of dilution in explaining patterns but also increased the importance of UV oxidation as a mechanism influencing DOC chemistry. Our findings suggest urban stream management has shifted dependence on microbially mediated C removal mechanisms to hydrologic dilution to reduce output concentrations. This shift lowers contaminant removal potential and increases dependence on limited groundwater resources.

1. Introduction

[2] Urban ecosystems are of interest to ecologists not only because they are home to half of the world's human population, but also because they provide a new arena for testing ecological theory [Grimm et al., 2000]. Aquatic ecosystems are features of urban landscapes that provide recreational opportunities for humans and habitat for animals while often receiving higher loadings of inorganic and organic compounds than in nonurban areas [Mulholland et al., 1997; Müller et al., 1998; Senn and Hemond, 2002]. In particular, managed streams (defined here as streams whose physical form or chemical composition have been and currently are being significantly altered by human activity to meet the needs of a populated area) are of interest because they drain the urban land area. Therefore, managed streams are transport vectors for materials moving from a city to downstream recipient ecosystems.

[3] Paul and Meyer [2001] reviewed the current state of knowledge regarding changes in physical, chemical, and biological attributes a stream may experience due to urbanization in the catchment. Those relevant to this study include changes in hydrograph behavior, channel geomorphology, nutrient and salt loading, microbial community structure and function, and nutrient and organic matter retention. Alterations suggested by Paul and Meyer [2001] are a function of the landscape in which the city is embedded and the pattern of development the city followed. For example, cities in colder climates apply salt to roads to inhibit snow accumulation; therefore, aquatic ecosystems in those cities will differ in terms of chloride dynamics than those in cities where salting is unnecessary. Grimm et al. [2004] suggested five key characteristics of arid land streams that are likely to be strongly modified in urban areas: flood disturbance, nutrient limitation, surface-groundwater interactions, land-water interactions, and landscape hot spots of nutrient retention [Grimm et al., 2004]. The biogeochemical implications of these types of alterations in managed ecosystems can be inferred from published work in nonurban areas, but remain unmeasured in urban ones.

[4] The influence of cities on stream nutrient concentrations is well documented, as many human activities result in export of N and P to local waterways; including agriculture, lawn fertilization, sewage treatment, and atmospheric deposition of automobile emissions [Baker et al., 2001; Faerge et al., 2001; Perakis and Hedin, 2002]. However, the influence of human activity on dissolved organic carbon (DOC) in streams is relatively unexplored [David et al., 1999; Drewes and Fox, 2000; Hope et al., 1994; Hyer et al., 2001; Westerhoff and Anning, 2000]. DOC is one of the primary sources of carbon fueling microbial decomposition and respiration in aquatic ecosystems; therefore, nutrient cycling and greenhouse gas production are closely tied to DOC consumption [Findlay, 2003; Kaplan and Bott, 1982; Sobczak and Findlay, 2002]. In addition to DOC quantity, its chemical complexity regulates microbial activity in many disparate ecosystems [Baker et al., 1999; Dahm et al., 2003; Hopkinson et al., 1998; Pomeroy et al., 2000; Vitousek and Hobbie, 2000]. DOC complexity also regulates contaminant transport in aquatic ecosystems by influencing the solubility of hazardous organic compounds and heavy metals [Baeyens et al., 1997; Gao et al., 1998; Haitzer et al., 1998; Shi et al., 2007; Waeles et al., 2005].

[5] By evaluating which processes are lost or altered through human management of waterways, we gain insight into stream functioning and can predict the dominant mechanisms of DOC retention in managed ecosystems. Managers who want to reduce export of excess organic carbon that might stimulate pathogenic or nuisance microbial populations, or who hope to use the water for crop irrigation and human consumption, must understand mechanisms of downstream DOC retention [Pinney et al., 2000]. Decline in stream DOC concentrations can result with or without C removal from the ecosystem. Therefore, we evaluated mechanisms of each type, including dilution or sorption of DOC to suspended sediments (no removal), and microbial mineralization or UV oxidation (removal) (Table 1). These hypothesized mechanisms were tested in a managed stream near the confluence of the Salt and Gila Rivers, two large tributaries of the Colorado River within the Phoenix metropolitan area of central Arizona, USA. The study stream originates as outflow from a tertiary wastewater treatment plant (WWTP) and terminates 66 km downstream. The most downstream site represents the hydrologic output for the lower Gila River and the city of Phoenix. Sampling along this waterway by the Central Arizona–Phoenix Long-term Ecological Research Project (CAP LTER) revealed that concentration of dissolved organic carbon (DOC) decreased by an average of 50% as water flowed downstream, while the average chemical complexity of the DOC pool increased (unpublished data). These preliminary data presented an opportunity to explore how modifications to stream geomorphology influence carbon cycling and retention in an important component of the urban landscape.

Table 1. Summary of Hypotheses and Predictions
HI: Hydrologic additions of low-DOC, high-complexity waters mixed with existing surface flow (dilution hypothesis)P1a: Groundwater inputs, manifested as increases in surface water salt concentrations, will be positively correlated with changes in DOC concentrations and SUVA at different points along the stream
 P1b: Sampling dates when groundwater wells are not pumping into the system will not exhibit the same DOC decline and increase in SUVA
 P1c: Mass balance calculations of groundwater additions to the surface stream will accurately predict changes in DOC concentrations and SUVA
H2: Microbial community respiration in channel sediments preferentially utilizes low-complexity DOC from surface waters, converting it to CO2 and leaving higher complexity DOC to be exported by surface water flow (microbial hypothesis)P2a: The extent of change in DOC quantity and complexity will be positively correlated with water temperatures due to the well-established temperature dependence of biological respiration.
 P2b: Measured sediment respiration rates will be sufficiently large to explain the decrease in DOC concentration in the surface waters
H3: DOC of low-complexity sorbs to suspended particles (adsorption hypothesis)P3a: Laboratory experiments will demonstrate DOC association with sediment particles and an accompanied change in SUVA values
H4: UV oxidation of DOC changes the quantity and complexity (UV hypothesis)P4a: UV penetration through the water column can generate sufficiently high oxidation rates at all depths and sites to explain the DOC decline and the changes in chemical complexity
 P4b: Published rates of UV oxidation are sufficiently high as to explain conversion of DOC to CO2 or to more complex organic matter during the time water spends in transport along the stream

2. Methods

2.1. Study Site

[6] The managed stream we studied begins approximately 30 miles west of Phoenix, Arizona, in an agricultural community that is experiencing rapid urbanization. The population in the area's three incorporated cities increased 30, 102, and 202% from 1990 to 2000, and at the center of this area, the population of the city of Buckeye increased nearly 200% from 2000 to 2006, making it the second fastest growing suburb in the nation (U.S. Census Bureau for 2000 and 2006). Outflow from the 91st Avenue wastewater treatment plant (WWTP) initiates surface streamflow in a riverbed that is dry because of upstream diversions into canals (latitude 33°23′21″N, longitude 112°15′12″W). The treatment plant receives approximately 56% of the municipal wastewater generated in the Phoenix metropolitan area (population > 4 million; U.S. Census Bureau 2005) [Lauver and Baker, 2000]. Sixty-four percent of the effluent leaving the treatment plant is discharged directly into the Salt River (Figure 1, site 1), which then joins the Gila River 5 km downstream of the treatment plant (Figure 1, site 2). All of the water at the Salt–Gila confluence is then diverted into the Buckeye Canal 7 km downstream (Figure 1, site 3). The water travels 35 km in the canal before discharging into the dry riverbed of the Hassayampa River (Figure 1, site 6). Once in the Hassayampa, the water travels 4.5 km back to the Gila River. The final reach (14.5 km) of the stream is in the Gila River, which is diverted at site 8 into irrigation canals at a privately owned dam built in 1913 (Figure 1). For ease of discussion, the 66 km stream has been divided into three reaches. Reach 1 is the first 13 km (sites 1–3), before the canal begins. Reach 2 is the 35 km of the canal that ends where it empties into the Hassayampa River (sites 3–6), and reach 3 is the last 19 km of flow in the Hassayampa and Gila rivers (sites 6–8).

Figure 1.

State map of Arizona indicating relative position of study site within Maricopa County. Larger map shows details of the study area. On the zoom map, numbers mark locations of sampling points, the dark line indicates the stream (water flows from east to west), gray areas are agricultural fields. Hydrologic inputs (HI) of water are found at three key locations indicated by the arrows. Reach from sampling sites 1 to 3 will be referred to as reach 1, from sampling sites 3 to 6 as reach 2, and sampling sites 6 to 8 as reach 3. Letter A indicates the sampling location used to characterize water moving through the Gila River before it joins the Hassayampa River.

[7] Groundwater wells are located throughout the area to supplement irrigation waters and for water-monitoring purposes. Thirty-four wells are located along the Buckeye Canal to supplement water supplies during high-demand periods (usually April to October). Ten drainage wells, located 1–2 km north of the Gila River bed pump water year round into the dry Gila River bed to keep groundwater levels below the rooting zone of the crops in the area. This water creates some surface flow in the Gila River, which rejoins the stream at the confluence of the Hassayampa and Gila rivers. Because the water table is so close to the surface in this area of the Gila River watershed, both shallow (<2 m) and deep (>10 m) inputs of groundwater play a role in creating the hydrologic template for the stream ecosystem. USGS stream gages were used to monitor surface water discharge at sites 3, 6, and 8 (Figure 1). Discharge at all other sites was estimated using measurements of cross-sectional area and velocity taken during sampling.

[8] Surface stream width along the 66 km stream varied from 5 to 20 m, and water depth ranged between 0.2 and 1.7 m. Attached algal mats were not observed along the stream, probably owing to the high turbidity and velocity. Correspondingly, PAR irradiance reached 40–65% of surface value at a depth of only 20 cm at the three most downstream sites (Table 2). Macrophytic growth was limited to only one ∼500 m reach at the confluence of the Salt and Gila River, where discharge slowed. We did not consider primary producers to be a significant source of C in this high-volume stream and therefore do not further discuss primary producer-mediated DOC dynamics as a mechanism contributing to changes in DOC chemical complexity. Our reasoning was (1) the apparent absence of substantial standing crop of algae at any location and (2) the assumption that in-stream algal biomass would actually increase DOC concentrations and lower SUVA values, an opposite pattern to the one we observed. While standing crop of primary producers was low, nutrient concentrations in the ecosystem were very high. Average nitrate concentration in the surface waters was consistently >2 mg (L NO3-N)−1 and reached as high as 8 mg L−1 at the end of the stream, while phosphate concentration ranged from 0.5 to 2.5 mg (L PO43−-P)−1.

Table 2. Site Characteristics as Average Values From All Sampling Datesa
Distance Downstream (km)Site DescriptionVHG (cm cm−1)Subsurface DO (mg L−1) (Temp (°C))PAR at 20 cm Depth (% of Surface)TDS (g L−1)TSS (mg L−1)OM in TSS (%)TDN (mg L−1)TDP (mg L−1)N:P of DOM
  • a

    The values for VHG and subsurface DO were taken at a sediment depth interval of 10–15 cm and are individual measurements representing variation due to geomorphology and temperature. NA indicates parameter is not applicable at a site, and ND indicates the data were not collected at the site. Values in parentheses are standard errors followed by the number of samples (n) except for subsurface DO.

0.01Salt River bedNDNDND0.97 (0.03, 5)14 (7, 5)33 (26, 3)4.8 (0.6, 4)1.83 (0.39, 4)6 (3, 3)
5Gila River below confluence with Salt+0.274.65 (30.0)ND1.29 (0.04, 9)26 (4, 9)24 (3, 8)8.0 (1.0, 8)2.33 (0.28, 8)8 (4, 6)
  −0.04, +0.052.95 (30.6)       
  −1.444.11 (30.2)       
13Irrigation canalNANAND1.34 (0.05, 3)93 (52, 3)15 (3, 3)5.2 (0.9, 3)1.76 (0.44, 4)8 (4, 2)
20Irrigation canalNANAND1.74 (0.12, 11)92 (17, 11)10 (1, 10)8.1 (0.4, 11)1.61 (0.12, 11)21 (10, 9)
30Irrigation canalNANAND1.88 (0.16, 11)103 (21, 11)10 (1, 10)9.5 (0.8, 11)1.49 (0.19, 11)31 (16, 8)
48Hassayampa River bed+0.2712.1 (22.7)391.99 (0.17, 11)72 (8, 11)11 (1, 10)11.1 (1.2, 11)1.32 (0.21, 10)28 (12, 9)
   12.5 (23.7)       
   8.4 (20.6)       
59Gila River bed+0.351.6 (21.2)542.89 (0.15. 9)95 (15, 9)11 (1, 8)10.4 (0.9, 9)0.99 (0.24, 8)129 (62, 6)
   5.1 (21.7)       
   9.1 (21.0)       
67Gila River bed−0.18, +0.448.0 (20.3)652.90 (0.15, 12)70 (11, 11)14 (1, 10)9.7 (0.5, 11)0.87 (0.22, 9)41 (27, 7)
   11.3 (16.0)       
   11.1 (16)       

2.2. Water Sample Collection and Analysis

[9] Surface water was sampled at roughly 10 km intervals along the stream on three occasions (March, June, October) in 2000, approximately every 6 weeks during 2001, and immediately following two winter floods in 2001. The sampling protocol followed one developed originally by the USGS and modified by the CAP LTER. Conductivity and water temperature were measured at each site using an Orion field meter. A depth-integrated, channel-wide water sample was collected at each site into a 3 L Nalgene® bottle, transported to the laboratory on ice, and then split into 10 subsamples. Three subsamples were filtered through a precombusted Whatman GF/F filter, and the filtrate was used for DOC analysis using high-temperature combustion on a Shimadzu TOC-5000 carbon analyzer. Three other subsamples were filtered through a membrane filter of porosity 0.45 μm, followed by dissolved nutrient analyses using colorimetric methods on a LACHAT “Quick Chem” 8000 flow injection autoanalyzer, and sulfate analysis using a Dionex 4000i ion chromatograph. Three remaining subsamples were used to calculate total dissolved solids (TDS).

[10] In addition to quantifying the concentration of DOC in the water, two measures of DOC chemical complexity were used. One was specific ultraviolet absorbance (SUVA), which is the absorbancy at 254 nm of a filtered, acidified water sample measured using a Hitachi U-2010 UV-Vis spectrophotometer, divided by the concentration of DOC in the sample [Weishaar et al., 2003]. SUVA is an indicator of the quantity of double-bonded C atoms. An aromatic carbon ring has half of its carbon as C = C, and since most natural organic matter is aromatic, not aliphatic, SUVA is a good measure of the aromaticity of a DOC sample [Edzwald, 1993; Nguyen et al., 2002]. The second measure used was the fluorometric index. Water samples are excited on a Shimadzu RF 551 fluorometer at 370 nm and scanned from 400 to 600 nm. The intensity of the scan at 450 nm divided by the intensity at 500 nm is the fluorometric index (FI). McKnight et al. [2001] found that samples with a FI < 1.5 had high-complexity organic material, while a FI value closer to 2.0 came from aquatic systems with low-complexity, algae-dominated, organic matter pools.

2.3. Methods for Hypothesis Testing

[11] We tested 4 alternative hypotheses explaining the decline in quantity and quality of DOC along the managed stream, considering the role of hydrology and in-stream processing. These hypotheses and their resulting predictions are outlined in Table 1. Hypothesis testing was done with a combination of in-stream measurements, mass balance calculations, laboratory measurements, and the use of literature values to determine the most likely explanatory mechanism. To test the dilution hypothesis (hypothesis 1), the chemistry and quantity of major hydrologic inputs (HI) along the stream was characterized. However, we were not able to sample HI 1 (Figure 1) due to its inaccessibility on private land. For HI 2, information regarding inputs of groundwater from the deep wells pumped directly into the canal, as well as the quantity of drainage well water that is routed directly to the Gila River bed was supplied by Buckeye Irrigation District Managers. Well water chemistry was collected in conjunction with sampling by the Arizona Department of Environmental Quality (summer 2002) and chemically analyzed using the water chemistry methods described above. Discharge and water quality for HI 3 was sampled at location A (Figure 1). Discharge was measured at every sampling location without a USGS gauging station on every date using a flowmeter and cross-sectional area measurements. Instantaneous flux (discharge in L s−1 multiplied by DOC concentration as mg C L−1) was calculated at the upstream and downstream end of each reach to determine the change in DOC load. Applying an average SUVA value and DOC concentration to HI 1 and 3, change in stream water chemistry could be predicted using mass balance calculations for locations where shallow and deep groundwater inputs were known to be entering and had been volumetrically quantified.

[12] To test the microbial hypothesis (hypothesis 2), sediment respiration was measured in chambers constructed from clear plastic pipe (32 cm long, 4.4 cm inside diameter). Channel hyporheic sediments were collected to fill chambers (to 50% of chamber volume) by first removing any large (≥5 cm diameter), exposed rocks at the sediment surface and then coring to a depth of 5 cm with a beveled plastic corer (12 cm diameter). Five cores were taken along a transect across the channel, combined, and redistributed to each chamber. Only rocks too large to fit in the chamber were excluded, otherwise mixed diameter sediment materials were used and respiration rates thus encompass variation due to particle size. The remainder of the chamber volume was then filled with water from the surface stream, gently inverted three times to allow any air trapped within sediments to escape, and sealed on both ends with rubber stoppers without introducing air. Sealed chambers were incubated in the dark for 4 h at varying temperatures (Table 3). Dissolved oxygen consumption was measured using a YSI-85 oxygen probe. Sediment dry mass, ash-free dry mass, and particle size distribution were determined for every chamber. Along the canal, benthic sediment accumulation was very small; therefore, respiration associated with the water column in the canal was measured in place of sediment respiration by placing unfiltered water from the water column into the same cores described above and monitoring depletion of oxygen over 4 h. Daily aerobic consumption of carbon by sediment-associated bacteria was calculated for reaches 1 and 3 assuming a respiratory quotient of 1 and using the following two equations:

equation image

where MRR is microbial respiration rate, g C m−2, R is DO consumption d−1 g dry sediment−1, BD is sediment bulk density, g cm−3, and D is sediment depth of 10 cm m−2:

equation image

where MRRtop is microbial respiration rate measured at the top of the reach, MRRbottom is microbial respiration rate measured at the bottom of the reach, and A is area of streambed within the reach, m2; reach 1, ∼260,000 m2, and reach 3, ∼105,000 m2.

Table 3. Benthic Sediment Respiration Measurements for Each Site and Datea
Distance Downstream (km)ReachMonths SampledStream Water Temp (°C)Average Chamber Incubation Water Temp (°C)Average Respiration Rate (g DO Consumed d−1 m−2)Average Percentage Sediment <2 mmAverage Sediment Organic Matter (%)
  • a

    Temperature and sediment characteristics are included for comparison.

  • b

    Water column respiration, units of g DO d−1 g TSS−1.

01Mar 200023.519.84.2180.3
  Oct 200030.023.82.0250.3
51Dec 200019.118.66.5500.4
  Feb 200118.417.91.6440.3
  Apr 200124.620.82.2170.9
  Jun 200127.721.81.0400.4
  Sep 200129.528.31.3300.6
  Oct 200126.220.01.0200.4
132Mar 200018.821.80.21b1002.8
202Mar 200018.721.20.19b1004.1
302Mar 200017.321.70.33b1005.4
483Mar 200018.613.21.4300.6
  Oct 200020.920.40.8270.5
  Feb 200112.817.90.7350.2
  Apr 200123.420.90.3360.6
  Jun 200125.422.40.2380.6
  Oct 200120.520.00.2400.7
593Mar 200015.119.40.9972.0
  Feb 200121.016.70.8990.5
  Apr 200120.525.92.0912.1
  Jun 200122.227.72.21000.5
  Sep 200129.725.01.8980.5
  Oct 200119.818.05.1956.3
673Mar 200016.217.13.9801.5
  Jun 200021.431.13.5970.4
  Jul 200033.030.41.9540.5
  Oct 200024.422.32.5990.8
  Feb 200115.615.91.5951.2
  Apr 200121.823.01.4670.7
  Oct 200119.918.07.6787.0

[13] The percentage of the water column that would need to exchange with the sediments in order to consume the entire quantity of DOC retained within the reach on a particular sampling day was calculated using

equation image

where DOCR is DOC retention s−1, which is the difference between the instantaneous load at the top and bottom of a reach on a particular day.

[14] To estimate the direction of exchange of stream water between benthic and subsurface sediments, vertical hydraulic gradient was measured using minipiezometers [Grimm et al., 2007]. Piezometers (tubes with lateral perforations near the tip) were inserted to a depth of 10–15 cm below the streambed and hydraulic head (cm) was measured on a simple manometer as the difference in water column heights between water columns drawn simultaneously from the piezometer and surface stream. VHG was calculated as the hydraulic head divided by piezometer depth (cm cm−1). Positive VHG suggests that the direction of water movement is from subsurface sediments up into the surface waters, whereas negative VHG suggests that surface waters are downwelling into subsurface sediments. However, sediment porosity also influences the exchange of water between the surface stream and subsurface, so VHG is not a definitive measure of actual exchange.

[15] To test the sorption hypothesis (hypothesis 3), an isotherm was constructed to estimate the sorptive capacity of the suspended sediment for DOC. Suspended material was collected in the field, concentrated by centrifugation, and washed using Nanopure® water and 0.01 N NaOH. The sediments were then sterilized by high-temperature drying or autoclaving. Water from the head of the stream (outflow from the WWTP) was filtered through a Whatman GF/F filter and 250 ml was added to 0.25 L plastic incubation containers. Triplicate incubations plus varying concentrations of sediment (0, 25, 75, 125, and 175 mg L−1) were shaken on a Eberbach 6010 shaker at 1140 rotations per minute for 24 h in the dark. Water was then filtered and analyzed for DOC and SUVA using the methods described above. Adsorbed DOC was calculated as the difference between preincubation and postincubation concentrations. An additional test was completed to determine the temporal pattern of DOC sorption, using only 0 and 10 mg L−1 total suspended solids (TSS), and bottles were sampled at 0, 4, 8, and 14 h after incubation.

[16] Testing the UV hypothesis (hypothesis 4) involved using literature values to calculate the potential rates of UV transformation of DOC. We estimated the moles of photons (spectral downward plane irradiance, Ed(λ), with λ = 290–490 nm) reaching below the surface of the water using the System for Transfer of Atmospheric Radiation (STAR) implemented as in the work by Fichot and Miller [2010] and assuming that all photons entering the water were absorbed by the chromophoric DOC (CDOM). Using three separately derived quantum yield models from the literature [Belanger et al., 2006; Johannessen and Miller, 2001; Vahatalo et al., 2000], an upper limit estimate of conversion of DOC to DIC could be calculated using

equation image

3. Results

3.1. Changes in Water Chemistry

[17] DOC concentration decreased downstream along the 66 km stream on all sampling dates, but to varying degrees (mean DOC decrease is 64%, range 47–80%). The decline in DOC was gradual and approximately linear through reaches 1 and 2, while the pattern between the last three sites was not consistent through time (Figure 2). SUVA was less spatially variable than the original LTER data set suggested, with no increase in SUVA through reach 1 and part of reach 2, a slight decline along the last half of reach 2, and a considerable increase in SUVA through reach 3 (Figure 2). An increase in SUVA indicates an increase in the average chemical complexity of the DOC pool, since the measurement is standardized to account for differences in DOC quantity. The fluorometric index (FI) for samples along the stream also indicated an increase in chemical complexity, with higher values upstream and lower (indicating high complexity DOC) values at the end of the stream (Figure 2). For all four hypotheses presented, we assessed the influence of each on DOC concentration and complexity. DOC concentration can change dramatically without influencing complexity, and vice versa. Therefore, the importance of each hypothesis in explaining downstream patterns may vary considerably according to the dependent variable of interest (i.e., DOC concentration or DOC complexity).

Figure 2.

(a) DOC as a percentage of the DOC concentration at the head of the stream, (b) SUVA, and (c) fluorometric index (FI) versus distance downstream. Error bars represent ±SE. Arrows below the x axis indicate location of the three reaches defined in Figure 1.

3.2. Role of Hydrologic Inputs

[18] Dilution of water with high DOC concentration and low chemical complexity with a source having lower concentrations and/or higher complexity DOC (hypothesis 1) could explain the DOC decline documented along this managed stream. Sulfate was used as an indicator of groundwater additions as sulfate concentrations were very high in the regional aquifer (∼800 mg L−1). While sulfate can be generated from alternative sources such as the mineralization of autochthonous organic S compounds or through oxidation of sulfide in sediment pore waters, neither of these two mechanisms was likely to quadruple sulfate concentrations in the water column, given the low organic matter content found in the sediments (Table 3). Concentrations of SO42−-S increased at several points along the stream indicating groundwater additions to the surface water as predicted by hypothesis 1 (P1a, Table 1 and Figure 3a). Correlations between surface water sulfate concentration at the end of reach 2 (canal) and the end of the entire stream were highly correlated (p < 0.05) with number of groundwater wells pumping (r2 = 0.99 and r2 = 0.63, respectively). When only one to three of the 30 deep groundwater wells along reach 2 were pumping, changes in DOC concentration were insignificant (P1b, Table 1 and Figure 3b). In contrast to reach 2, on low-pumping days, DOC decline was evident in reach 3, where groundwater from the Gila River joined the Hassayampa River (HI 3). In reach 3, an increase in surface discharge resulted in a proportional drop in DOC concentration (p = 0.05, r2 = 0.4, slope of −1.0). This decline in DOC concentration in reach 3 was not evident on days with large inputs from deep groundwater wells along the canal (reach 2).

Figure 3.

(a) Average sulfate concentrations in the surface stream along the stream as compared to an average groundwater concentration for this area (dotted line). (b) DOC along the stream as a percentage of the DOC concentration at the head of the stream. Four dates with varying discharge rates from groundwater wells along reach 2 are shown, with discharge as cubic meters per second (cm s−1) included in legend.

[19] To evaluate the third prediction from hypothesis 1 (P1c, Table 1), decrease in DOC concentration and increase in SUVA along each reach were calculated based on known groundwater inputs and associated chemistry (Figure 4a). Given the lack of water samples from HI 1, we removed reach 1 from these calculations. Most of the change in DOC concentration along reach 2 (from wells adjacent to the canal) was explained by mass balance calculations, leaving 7–25% of the variation unexplained on 8 of the 12 sampling dates (Figure 4b). In reach 3, the expected decrease in DOC due to inputs of shallow groundwater underestimated the actual DOC decline by 1–22% on all but two sampling dates (data not shown). Change in SUVA values along the reach were predicted based on an average groundwater SUVA from samples collected during the study. Since this average value was higher than any SUVA value recorded along reach 2, groundwater additions were unable to explain SUVA patterns along the reach. Dilution-corrected SUVA values declined along reach 2 on every date (Figure 4b), but in reach 3 increases in SUVA calculated from measured water chemistry collected directly upstream of the confluence of the Hassayampa and Gila Rivers underestimated increases in SUVA by 4 to 43% for all but one of the 12 sampling dates (Figure 4b). In all, mass balance calculations do support the third prediction from the dilution hypothesis, but additional, unexplained decline in DOC concentrations, decline in SUVA values along reach 2, and increases in SUVA values along reach 3, remain after dilution is accounted for.

Figure 4.

(a) Study site with hydrologic inputs and extractions during a high irrigation period indicated with arrows that are proportional to size of the water input/export. (b) Changes along reaches 2 and 3 in DOC concentration and SUVA value on each sampling date after correcting for dilution due to subsurface inputs. Negative values indicate a lowering of DOC concentration or SUVA; positive values indicate an increase in either.

3.3. Role of Heterotrophic Microbial Community

[20] Although much of the decline in DOC concentration can be explained by dilution, consumption of surface water DOC through microbial respiration was occurring (Table 3). Our challenge was to determine if these rates could contribute substantially to surface water patterns. The first prediction from the microbial hypothesis (P2a, Table 1) was not supported, as water temperature and the decline in DOC concentration in reaches 2 and 3 were not correlated (data not shown). Calculations of daily aerobic consumption of carbon in the benthic sediments ranged from 362 to 1576 kg C in reach 1, based on the limits of respiration rates measured during the study (equation (2)). By comparing these C consumption values to the DOC load leaving reach 1, we find an average respiration rate could more than account for the decline in DOC concentration measured along reach 1, assuming the water exchanged with the entire benthos to a depth of 10 cm (data not shown). The expected decline in DOC concentration from sediment-associated respiration in reach 2 was small, only 0.08 mg L−1 assuming the maximum respiration rate and TSS concentration recorded in this reach (Figure 4b). In reach 3, where we measured sediment-associated respiration on each date, benthic respiration could explain the entire dilution-corrected decline in DOC concentration on 5 dates, and 22–83% on the other 4 dates (Figure 4b). Our calculation assumes complete exchange of the water column with the benthos down to 10 cm, and that all benthic respiration is fueled by surface water DOC (discussed further below). The influence of respiration on SUVA is harder to predict, but the expectation is that mineralization would cause an increase in SUVA. SUVA values did increase in reach 3 but not in reach 2, consistent with measurements suggesting respiration is unable to account for the DOC concentration decline in reach 2 but is an important retention mechanism in reach 3.

3.4. Role of Sorption Kinetics

[21] Laboratory experiments supported the prediction from hypothesis 3, attributing a portion of DOC retention along the flow path to association with particles (Table 1). Based on the DOC sorption curve created in the laboratory, 0.7 mg L−1 (±0.03) of the DOC at the outflow of the wastewater treatment plant may have associated with suspended sediments in reach 1 (Figure 5a). Experimental suspended-sediment concentration ranged from 25 to 175 mg L−1, which encompassed all TSS values recorded along the stream with the exception of one sampling date in June. SUVA did not change as DOC partitioned with sediments in the laboratory experiment (Figure 5b). In the second sorption experiment, the proportion of the DOC pool that sorbed to the sediments was essentially the same as in the first experiment, and the majority of the sorption occurred during the first 4 h of incubation (data not shown).

Figure 5.

Adsorption test to determine (a) DOC abiotic uptake capacity of washed suspended sediments and (b) change in chemical complexity (SUVA). Error bars represent ±1 SE. Line drawn for illustrative purposes only. Values significantly different from zero (as determined by ANOVA) are indicated with an asterisk.

3.5. Role of UV Transformation

[22] Dilution-corrected DOC concentration decline along reaches 2 and 3 combined ranged from 0.3 to 2.3 mg L−1 when considering all sampling dates. Modeled estimates of UV photons penetrating the water column were used in conjunction with literature values for quantum yield to calculate the theoretical moles of DIC produced per mol of radiation received on each sampling date. Using these literature estimates, an upper limit of 0.7 mg L−1 of DOC could have been oxidized to CO2 through UV photobleaching (hypothesis 4) within the residence time of this managed stream [Amado et al., 2006; Bertilsson and Tranvik, 2000]. However, the assumption that all photons were absorbed by CDOM along the entire length of the stream probably results in an overestimate of DOC oxidation by UV, as suspended sediments were high in the canal and would reduce penetration of radiation into the water column [Belmont et al., 2009].

3.6. Influence of Flooding

[23] Urban centers in arid lands are engineered to maximize rainfall retention, either through reservoirs, neighborhood retention basins, or groundwater recharge sites. Phoenix is no exception, and therefore much of the overland flow that occurred in the catchment of the Gila River never made it to this urban stream. During 2001 two floods did occur in the lower Gila River, both during the winter rainy season. These two events presented an opportunity to determine whether flooding changed chemistry along the stream, and to determine the time required for reestablishment to previous longitudinal patterns. DOC concentration in floodwaters, while slightly higher than during the March flood, followed the same pattern of downstream decline seen at base flow (Figure 6a). SUVA was also higher than base flow values during one of the two floods, but followed the same spatial pattern downstream that has been previously recorded (Figure 6b). Reestablishment of DOC concentration to base flow values took <19 days for both floods (Figure 6).

Figure 6.

(top) DOC concentrations along the stream during and following two floods in Spring 2001. (bottom) SUVA along the stream during and following two floods in Spring 2001. Error bars represent ±1 SE.

4. Discussion

4.1. Hypothesis Testing

[24] While the four hypotheses tested are not mutually exclusive, this research was intended to identify the relative importance of each. Dilution explained 26–100% of the decline in DOC concentration in reach 2 (median of 68%), and 0–100% in reach 3 (median of 29%). Significant alteration of channel geomorphology along the canal (reach 2) created an environment where primary production was low, surface-subsurface exchange was eliminated, and heterotrophic community activity was inconsequential in explaining surface water organic chemistry patterns. Water pumped directly into the canal was from wells at depths > 30 m, which supplied very low DOC concentration water (Figure 7a). The importance of geomorphic changes to mechanisms influencing biogeochemical cycling has been discussed in the literature, but our study is novel in its documentation of the effect in an urban ecosystem [Doyle et al., 2003; Graf, 1975]. Although dilution was the dominant explanatory mechanism for the decrease in DOC concentration in reach 2, DOC patterns were not fully explained by hydrologic additions. Additionally, the decline in SUVA values along reach 2, and the increase in SUVA through reach 3, was consistently underpredicted by groundwater additions. Our alternative hypotheses, therefore, provide partial explanation for the observed pattern of DOC decline.

Figure 7.

(a) DOC concentration in water samples collected from groundwater wells (summer 2002) throughout the study area, graphed as a function of the depth of the well. (b) DOC concentrations at the outfall from the 91st Avenue wastewater treatment at the head of the managed stream (data from CAP LTER database). (c) Timing of pumping of groundwater wells along the canal (data from Buckeye Irrigation District managers).

[25] Quantifying the influence of microbial mineralization of DOC to CO2 (hypothesis 2) required an assumption about the amount/duration of exchange between the surface water and the subsurface sediments so that benthic respiration could be related to chemical changes in the water column. In our calculations, we assumed the entire DOC surface water load exchanged with the subsurface and was bioavailable to hyporheic microorganisms in the time frame of downstream transport. Given the high velocity of surface flow along the stream (0.1–0.8 m s−1 in reach 1, 0.3–1.0 m s−1 in reach 3) and the very small change in elevation, limited surface–subsurface exchange might be expected. If the exchange between shallow subsurface water and surface water was small, these two subsystems were, in effect, acting as separate systems, reducing the influence of benthic microbial activity on retention of nutrients and carbon as water flowed downstream. Using the range of sediment-associated rates of microbial carbon consumption and assuming all respiration was fueled by carbon in particulate organic matter, the residence time of particulate organic carbon (POC) ranged from 0.1 to 31 years in reach 1 and 0.4 to 18 years in reach 3. This calculation suggests it is possible that benthic respiration was fueled primarily by benthic particulate organic matter. But, organic carbon content of sediments only explained 39% of the variation in sediment aerobic respiration across all dates (p < 0.0001), a finding that is inconsistent with sediment OM being the primary fuel for microbial metabolism. Calculations of the contribution microbial respiration could have made to DOC decline indicated that it could be a very important mechanism for C retention in reach 1 and was also important in explaining the decline in the DOC concentration in reach 3.

[26] The microbial hypothesis could also explain the dilution-corrected increases in SUVA values through reach 3. SUVA values at the end of reach 3 were higher than expected after water entered the system at HI 3. Laboratory incubations in other ecosystems have shown SUVA values to increase as microbial utilization of DOC occurs, presumably as mineralization of less chemically complex DOC leaves the more complex material behind [Maurice et al., 2002; Pinney et al., 2000; Strauss and Lamberti, 2002]. Reach 3 had the largest concentration of fine sediments in the benthos, and this abundance of fine material likely supports high microbial biomass, resulting in preferential removal of low-complexity DOC during microbial decomposition.

[27] While our first two hypotheses appeared to be important in controlling the decline in DOC concentration along this urban flow path, our third hypothesis was more difficult to test with regards to its contribution to changes in DOC. Net DOC sorption to sediments (hypothesis 3) was small but not inconsequential in our laboratory experiment (<10% of the concentration at the head of the stream) and did not discriminate between DOC of different SUVA values. This may be a limitation of the sensitivity of the SUVA method for detecting changes in DOC chemical complexity. Nondiscriminatory binding of DOC to sediments has been documented previously in a headwater stream, however specific portions of the DOC pool (e.g., humic acids) are known to preferentially interact with aluminum and iron oxides contained in particles [Kaplan and Newbold, 2000; McDowell, 1985]. Given the short time period required for DOC to form a sediment association in our laboratory work (less than 4 h), sorption would not generate a longitudinal DOC decline along this flow path without the addition of new particle-associated sorption sites. The extent to which water column DOC might encounter available sorption sites, either in suspended or benthic sediment, is very difficult to model. Therefore, adsorption of DOC to organic-matter-rich suspended particles appears to be responsible for a portion of the pattern in DOC decline we documented here, however our data do not support the hypothesis that sorption was influencing SUVA patterns.

[28] Our fourth hypothesis required determining the influence of UV radiation on downstream patterns in DOC chemistry, which involves a complex set of photochemical reactions. UV oxidation can break apart and at times mineralize large DOC molecules, which would lower DOC concentrations but also lower SUVA values [Bertilsson and Tranvik, 1998; Miller et al., 2002]. This pattern of lower DOC, lower SUVA was observed along reach 2 but not reach 3. Researchers have also shown that DOC pools with initially low complexity may be transformed into larger, more complex compounds upon exposure to UV radiation, a mechanism that could explain the increase in SUVA in reach 3 [Moran and Covert, 2003; Tranvik and Kokalj, 1998]. Our estimate of UV reactivity might have been too high, as penetration of UV light through the water column is inhibited by high suspended sediment loads, high conductivity and high pH, all of which were found along reaches 2 and 3 [Bertilsson and Tranvik, 2000]. However, it appears this mechanism could be particularly important in the well-mixed, high-light environment found in the canal portion of the urban stream.

4.2. Reinventing Streams in Urban Environments

[29] We return to our original goal of considering which mechanisms of DOC transformation were altered as the Gila River was transformed to its current configuration. As predicted, we found significant changes in flood disturbance, surface-groundwater interactions, and the location of hot spots of DOC retention [Grimm et al., 2004]. Temporal variability of stream chemistry in response to disturbance by flooding, an important characteristic of unmodified streams in the region [Fisher et al., 1982; Holmes et al., 1998; Jones et al., 1996], was essentially lost from the managed stream. Hydrologic alteration through manipulation of geomorphology, reduction in floodwater delivery, and temporal consistency of the dominant inputs (groundwater), created a physicochemical environment that quickly reestablishes following disturbance events such as floods. There was little to no impact of flooding on DOC dynamics. Because of this stability in physicochemical characteristics, the communities in the ecosystem experienced only weak disturbances. In stark contrast, chemistry of unmanaged desert streams responds immediately to flooding, with higher concentrations of all major constituents in floodwaters and lowered spatial heterogeneity along the stream immediately following a flood [Dent and Henry, 1999; Fisher et al., 1982]. Flooding in unmanaged ecosystems also decimates algal and macroinvertebrate stream abundance, resetting these communities and initiating stream succession [Fisher et al., 1982].

[30] Given these dramatic changes in ecosystem properties in unmanaged desert ecosystems in response to flooding, communities are both resilient and resistant [Fisher et al., 1982; Grimm and Fisher, 1989; Webster, 1983]. Adaptation to a dynamic, disturbance-driven hydrologic regime is unlikely in the microbial communities living within the managed stream ecosystem we studied, given the reduced effect of flooding on both the water chemistry and geomorphology. As a consequence, this urban stream may not maintain its current status with regard to C cycling if a significant change in geomorphic or chemical structure were to occur (i.e., altered water sources, dredging, canalization, or wastewater treatment practices). Following these human-induced disturbances, biological communities may quickly move out of their current domain of attraction into an alternative stable state [see Holling, 2001], as biological communities in managed ecosystems are likely highly specialized for current conditions and would require restructuring to adjust to a new physicochemical template.

[31] The dominance of dilution as a mechanism creating a downstream decline in DOC concentration in this urban stream should have dampened seasonal variability in patterns of reactive elements, as abiotic mechanisms are not heavily influenced by ambient water temperatures or living biomass to the extent that biotic mechanisms are. In our work, however, DOC supply to the stream was temporally variable due to human engineering and decision making. Chemistry at the head of this managed stream was a function of wastewater treatment efficiency. The large wastewater treatment plant releases effluent with variable concentrations of DOC, which set the baseline water chemistry of the stream (Figure 7b). Additional temporal variability was introduced by irrigation water in the area. Demand for groundwater was highest during the warmest months of the year (Figure 7c); therefore, DOC surface water concentrations were closely related to air temperatures. So, while dominant DOC sources (effluent and groundwater) were consistently present, the relative abundance of each was spatially and temporally variable. Information on decisions being made by managers, therefore, is required to understand biogeochemical cycling in a managed stream, and traditional ecological theory must be modified to include human-mediated mechanisms.

5. Conclusions

[32] Much of the spatial variability in DOC concentrations in this study was created by pumping large quantities of DOC-poor, deep groundwater into the managed stream ecosystem, establishing a longitudinal pattern. Work in nonurban desert streams found hot spots of biological carbon utilization where surface water moves down into the subsurface sediments or laterally into the riparian zone [Heffernan and Sponseller, 2004; Holmes et al., 1996; Jones et al., 1995]. In our system and similar to unmanaged ecosystems, surface-subsurface exchange continued to be important, but at depths much greater than those considered in arid stream ecosystems [Dent et al., 2001; Grimm and Fisher, 1984; Stanley and Boulton, 1995; Valett et al., 1994]. Canalization and the reduction or elimination of complex geomorphic structure increased water velocity, and as a consequence, surface water was isolated from subsurface processing [Haggard et al., 1999]. Input of chemically distinct deep groundwater by pumping changed surface water chemistry, to the extent that the number of groundwater wells being utilized on a particular day controlled 26–100% of the downstream decline in DOC concentration. High nutrient concentrations accompanied the low DOC in the deep groundwater additions, increasing nutrient loading to a stream where there was very little primary production for uptake and retention of these nutrients [Roach et al., 2008]. In place of biological utilization in reach 2 (canal) we found abiotic transformation through UV oxidation to be a plausible hypothesis to partially explain downstream declines in DOC concentration, and abiotic adsorption to suspended sediments to be important only at the initiation of the stream in reach 1. However, biological utilization of surface water DOC to support sediment respiration remained an important mechanism in the first and third reaches, explaining 22–100% of the dilution-corrected downstream decline in DOC concentration, and likely explaining 15–73% of the increase in SUVA in the last third of the flow path that was not accounted for by dilution. Because dilution is not a retention mechanism, improving surface water quality in this stream is less a function of biological processing and more a function of groundwater input. Loss of ecosystem services (nutrient uptake and carbon mineralization) is a symptom of the “urban stream syndrome” and is used to justify restoration efforts [Grimm et al., 2008; Paul and Meyer, 2001; Walsh et al., 2005]. As the population of Phoenix metro continues to expand, new pressures will be placed on local streams, and their inability to ameliorate increased OM loading could force significant changes in human behavior and management.


[33] Thanks to Tom Collela, Lisa Dent, Aisha Goettl, Jim Heffernan, Julia Henry, Shero Holland, Darrel Jenerette, Cathy Kochert, Matt Luck, John Roach, John Schade, Ryan Sponseller, and Jill Welter for help with field sampling, lab analysis, and conceptual development. Modeling of potential UV oxidation was completed through collaboration with Cedric Fichot, whose help was invaluable. Comments from anonymous reviewers improved the paper tremendously and were appreciated. This work was based on work supported by funding from Proctor and Gamble, Inc. (J.W.E.), and the National Science Foundation under grant DEB-0423704, Central Arizona–Phoenix Long-Term Ecological Research (CAP LTER).