UGA
Skidaway Institute of Oceanography
search
Research About US Education Faculty Publications R/V Savannah
 Urban Runoff 

Urban Runoff and Oxygen Dynamics on Salt Marsh Platforms

A PDF copy of the Year 1 Final Report to Georgia DNR is available can be found here.
A PDF copy of the Year 2 Final Report to Georgia DNR is available can be found here.

Metropolitan Savannah relies on an extensive network of stormwater drainage canals to channel runoff away from city streets and into the surrounding rivers and estuaries. Drainage of stormwater off impervious surfaces such as streets, parking lots and rooftops, leads to "flashy" inputs of large volumes of freshwater into local receiving waters. Runoff may carry high concentrations of pollutants such as PAHs, metals, and organic and inorganic nutrients (Holland et al. 2005). The focused inputs of stormwater runoff in space and time can exceed the capacity of estuarine systems to respond to these inputs in a healthy fashion, potentially leading to problems such as lowered water column oxygen, nuisance algal blooms, and impairment of animal populations.

Figure 1

Figure 1: Stormwater drainage canals in Savannah. Canals are shown in yellow. The approximate drainage watershed for the Vernon River is outlined in pink. Red circles denote the location of our monitoring sites on the Vernon River (at Halcyon Bluff) and on the Skidaway River (at the SERF site at Skidaway Institute of Oceanography)

In order to document the effects of runoff on local estuaries, we have been monitoring the oxygen concentrations of flood tide water on the marsh platforms on the Vernon River on the Savannah southside and at Skidaway Institute's SERF research facility on Skidaway Island. The Vernon River receives inputs from several major drainage canals, including the Casey Canal, the Kingsland Canal, and The Harmon Canal. The Casey Canal/Hayner's Creek drainage system has been cited by GA EPD for not meeting water quality guidelines for dissolved oxygen, fecal coliform, or for safe fish consumption (http: //www.gaepd.org/). Prior studies have been conducted in the Vernon River to identify sources and concentrations of fecal coliform bacteria in river waters (Richardson 2007, Power et al. 2007). A study of the main channel of the Vernon River has shown that low water column oxygen concentrations can accompany extremely high runoff events (Richardson, 2007). In contrast, the SERF site on Skidaway Island is surrounded by woodlands. There is a small runoff source from the Landings, a large residential development, several kilometers to the south of the study site.

If runoff and canal discharge is negatively affecting water quality (as indicated by low dissolved oxygen concentrations), then we would predict that water column oxygen concentrations will be lower, and oxygen consumption rates will be higher, at the Vernon River than the SERF site. Furthermore, we predict that individual runoff pulses into the Vernon River system will lead to relatively prolonged periods of low oxygen and high O2 consumption as the runoff water is returned to the upper portions of the river with succeeding tides.

Figure 2
Figure 2: Water quality instrumentation. The float and 'garage' assembly at low tide at SERF (left image). An identical setup is installed at Halcyon Bluff. The instrument deployed in the creek at SERF is shown in the right image.


Oxygen concentrations at both sites have been monitored with YSI OMS 600 sondes. The instruments sample conductivity, temperature, water depth, and dissolved oxygen concentrations at 15 minute intervals. Because the optical dissolved O2 sensor must be kept moist, we built 'garages' for the instruments to sit in at low tide. As the tide rises, the instrument floats up to a stop point 15 cm above the sediment surface. As the tide drains from the marsh it returns to its water-filled garage. Only data from the time the sensor is in its sampling position are used in the data analysis. Sondes are swapped out for fresh calibrated units every two weeks. No significant drift in the O2 signals of the instruments have been identified in 7 months of continuous operation.

Figure 3
Figure 3: O2 concentration data for the SERF marsh platform and an adjacent creek.

After an initial evaluation showing a ~constant offset in oxygen concentration between creek and platform waters at SERF (Figure 3), the sonde was moved to an adjacent tidal creek for the remainder of the sampling program. Because the sondes were now at different tidal heights, we restricted our analysis at SERF to those times that corresponded to flooded marsh at Vernon River (Figure 4).

Figure 4
Figure 4: Analytical windows for sondes at Halcyon Bluff (left) and SERF (right). Oxygen concentration was taken as the mean of the individual oxygen readings over the sampling interval. Oxygen consumption was measured as the slope of oxygen content vs time

Except where otherwise noted, data below are taken only from night time high tides. Daytime data are complicated by photosynthetic oxygen production and its strong interaction with both day length and tide. The analysis and interpretation of the daytime data are ongoing.

Salinity data indicate episodes of stormwater discharge into the Vernon River (Figure 5) as well as seasonal inputs of freshwater into the estuary as a result of seasonal changes in freshwater inputs from inland rivers (Ogeechee River basin).

Figure 5
Figure 5: Salinity data for flood tides at Vernon River (Halcyon Bluff: red dots) and Skidaway River (SERF: green dots).


Figure 6Figure 6: Salinity differences between the two sites range from near 0‰ to almost 25‰ (bottom). Increased relative salinity differences are always associated with rainfall events (top) and reflect local stormwater inputs of freshwater to the Vernon River.

Salinity differences between the two sites range from near 0‰ to almost 25‰ (Figure 6: Bottom). Increased relative salinity differences are always associated with rainfall events (Figure 6: Top) and reflect local stormwater inputs of freshwater to the Vernon River.

A plot of nighttime high tide oxygen concentrations at the two sites between late July 2009 and mid-February 2010 shows similar patterns for the two sites (figure 7: upper panel), with slowly declining oxygen concentrations at both sites until mid-October, followed by sharply increasing concentrations through the fall and winter. Applying the creek/platform correction (Figure 3) to the data bring the data to near superposition throughout the sampling period except for the time between ~mid-September and early December, 2009, when oxygen concentration on the Vernon River marsh platform are slightly but consistently lower than at SERF (Figure 7: lower panel). Contrary to our predictions, this is the driest time of the entire sampling period. At neither site do O2 concentrations fall below the critical 2 mg l-1 threshold for extended periods.

Figure 7
Figure 7. Nighttime high tide O2 concentrations at the Vernon River (red) and Skidaway River (green) sites. Open circles in the lower panel show the (downward) shift in [O2] resulting from the creek/platform correction. Points circled in red indicate the portion of the sampling period in which Halcyon O2 concentrations were consistently less than those at SERF. The dotted line shows the 2 mg l-1 concentration threshold.

An interesting aspect of the oxygen concentration data is its apparent strong response to seasonal meteorological forcing. Prior to ~October 15, 2009, average O2 concentrations at Halcyon Bluff were declining over time despite slowly cooling water temperatures (Figure 8a,b). After the 15th, however, water temperatures began to drop more rapidly, and oxygen concentrations began to rise in response. The transition between the two regimes corresponded with the first significant autumn cold front. The data trends for SERF (not shown) were similar to those at Halcyon Bluff.

Figure 8
Figure 8. Response of Vernon River to the seasonal temperature regime. Right image:  Water temperature before (pink) and after (black) 10/15/09 when a strong cold front passed through Savannah. Left image: Response of oxygen concentrations to water temperatures before (pink) and after (black) 10/15/09.

Comparison of oxygen consumption rates between the two sites reveals only small differences (Figure 9). O2 consumption is consistently greater at the Vernon River site (Paired T-test; P < .001), but the magnitude of the difference is only about 10% of the average rate. (mean rates, seasonality). Note that in contrast to oxygen concentration, there is no apparent inflection in O2 consumption rates in October.

Figure 9
Figure 9. Comparison of nighttime high tide O2 consumption rates at the Vernon River (red) and SERF (green)

By comparing differences in salinity between Halcyon and SERF with differences in O2 concentration, we can visualize the effects of freshwater runoff on the oxygen regime on marsh platforms. There appears to be no effect (Figure 10a). There is a significant difference in O2 concentrations between the sites. Halcyon has 0.2 mg l-1 less O2 than SERF on average (paired T-test; P<.001). This small difference is unlikely to be biologically meaningful. However, a similar comparison of relative O2 consumption rates as a function of salinity differences does show an effect. Respiration rates are more rapid on the Vernon River marsh platform when the salinity difference is greater - i.e. , when stormwater runoff into the Vernon River is greatest (Figure 10b). The overall relationship is weak, however, and appears to be driven by a small number of points.

Figure 10
Figure 10: Relative oxygen concentrations and consumption rates as a function of salinity differences (=stormwater inputs) between the two sites.

The instances of high relative respiration rate correspond closely with specific rain events (Figure 11). However, in contrast to our expectations, runoff from single rain events did not produce either prolonged depressions in oxygen concentration or elevation in respiration rate at Halcyon. Once the rain ceased respiration rates returned to mean levels determined by water temperatures.

Figure 11
Figure 11: Relative respiration rate (Halcyon - SERF) contrasted with rainfall. Individual rain events correspond closely with increases in O2 consumption at Halcyon.

The absence of a persistent effect suggests that the stormwater inputs into the Vernon River are moving through the system quite rapidly. To examine this possibility more closely, we have adopted a simple tidal flushing model developed by Larry Sanford at the University of Maryland (Sanford et al, 1992). Flushing time (Tf) for a basin connected to a tidal river isEquation 1 where Vlw is the low-water (LW) volume, P is the tidal prism and r is the return flow factor. Maps were analyzed to delineate the LW and high-water (HW) volumes, the difference of which yields P (Figure 12). As it was impossible to estimate the water depths in the small channels at the upstream end of the tidal, two sets of LW and HW volumes were calculated: a dry-channel set where water depth at LW was set to zero and a wet-channel set where the same area depths were set to 1 m at LW (Table 1).

Equation 1
Equation 1: Flushing time (Tf ) for a basin connected to a tidal river is calculated above where Vlw is the low-water (LW) volume, P is the tidal prism and r is the return flow factor.


Figure 12
Figure 12. Image of Vernon Creek and surrounding intertidal area of the Halcyon site.


Dry upstream channel Wet upstream channel
LW area 1.014 × 106 1.014 × 106
HW area 5.587 × 106 5.587 × 106
LW volume 0.446 × 106 4.706 × 106
HW volume 2.836 × 106 11.674 × 106
Table 1. Areas (m2) and volumes (m3) for the Halcyon site. See text for details.

The return-flow factor (r) in Equation 1 is very important but difficult to estimate. We used several rain events, which injected a high freshwater volume tracer into the system. As an example (Figure 13), we took the LW and HW salinity of each consecutive tide and converted each to a freshwater volume (Q) using a maximum salinity value of 34 (maximum salinity observed at the SkIO fuel dock) to represent a freshwater volume of zero. Return flow for each tidal cycle is calculated as follows: Equation 2 where Qhw and Qlw are the freshwater contents at HW and LW respectively.

Equation 2
Equation 2: Return flow for each tidal cycle is calculated above where Qhw and Qlw are the freshwater contents at HW and LW respectively.

There were two rain events examined for each site (Figure 13 is one of the pairs). Using Equation 2, a range of return flow factors was calculated for each event at each site (Table 2).

Marsh site Minimum Mean Max
Halcyon 1 0.06 0.12 0.23
Halcyon 2 0.17 0.23 0.34
SERF Dock 1 0.11 0.22 0.57
SERF Dock 2 0.23 0.36 0.73
Table 2. Return-flow factors (r) for Halcyon and SERF Dock sites based on freshwater impulse episodes. Event 1 occurred from 1-10 September (Fig. 2) and event 2 occurred from 13-20 October.

Figure 13
Figure 13. Salinity records at the two study sites showing a rain event. Lowest salinity typically occurred at LW while the highest salinity occurred at HW. Water level (m) was measured at SERF Dock and is scaled to plot. The difference between these two within a tidal cycle is used to estimate the return-flow factor (r). See text.

Using Equation 1, the flushing time for the Halcyon site ranges from 0.1 to 0.5 day (Table 3), suggesting that a single tidal cycle completely flushes away a pulse of material with a high biochemical oxygen demand.


r
Tf (dry channel at LW) Tf (wet channel at LW)
0.06 (minimum) 0.10 0.37
0.12 (mean 1) 0.11 0.40
0.23 (mean 2) 0.13 0.45
0.34 (maximum) 0.15 0.53
Table 3. Range of flushing times (days) for Halcyon site based on low-water and high-water volumes estimated from GIS analyses and using r values from Table 1.

A similar analysis can be done for SERF Dock if a map of that site analogous to Figure 12 can be drawn. The data from SERF Dock (Table 2) already suggests that flushing times would be longer at this site. Assuming that the relative scales of the ratio Vlw / P at SERF Dock are similar to those at Halcyon, Equation 1 constrains the flushing times between 0.1 and 1.3 days. The large range is partially due to the fact that estimating the LW volumes at SERF Dock suffer from the same problems encountered at Halcyon, namely that we do not know how much of the creek channels still have water at low tide.

These flushing numbers are very approximate. Our high salinity endmember is very weakly constrained by this model. However, it is significant that the model output support our observations - low salinity, low oxygen, high oxygen demanding water delivered to then Vernon River by storms does not have a long-lasting effect on the system. It appears to pass through the river and mix with the larger volumes of water in the Little Ogeechee River and Green Island Sound over the course of 1 or 2 tidal cycles. The coherence of oxygen concentrations at both the runoff-affected Vernon River site and the relatively unperturbed SERF site suggests that the oxygen regime in these habitats is being set by common properties of the system that act on scales greater than the distance between these two sites.

We are continuing our collect data from the two sites. We have yet to fully analyze the data from daytime high tides. We see both net O2 consumption and O2 production during daytime high tides, depending on tidal stage, season, and light intensity. Our ultimate goal is to develop a comprehensive annual oxygen budget of these two systems based on our dataset.

References:
  Georgia Environmental Protection Division. 2008. 2008 Integrated 305(b)/303(d) List. http://www.gaepd.org/Files_PDF/305b/Y2008_303d/Y2008_Coastal_Streams.pdf
  Holland, F, Sanger, D, et al. 2005. The tidal creeks project: understanding coastal waterways. SC DNR booklet. 24 pp.
  Power C, R. Feldner, B. Scanlon, M. Frischer, and J. Richardson. (2007). Bacterial source tracing in the Vernon River Watershed. Proceedings of the 2007 Georgia Water Resouces conference. University of Georgia.  March 27-29.  Available at: http://www.gwri.gatech.edu/uploads/proceedings/2007/4.6.2.pdf
  Richardson, J. 2007. http: //ceasurf.com/Recentprojects.aspx. Effects of Altering Storm Water Discharge Rates and Volumes on Water Quality in Coastal Rivers and Estuaries.
  Sanford, L.P. , Boicourt, W.C. , and Rives, S.R. (1992) Model for estimating tidal flushing of small embayments. J. Waterway, Port, Coastal and Ocean Engineering 118, 635-655.


Text Only Version

• Skidaway Institute of Oceanography • 10 Ocean Science Circle • Savannah, GA 31411 • USA • (912) 598-2400 •
Copyright © 2014 Skidaway Institute of Oceanography. All rights reserved.