Impact of upstream plant level pollution on downstream water quality: evidence from the clean water act

Abstract

This is the first study to find empirical evidence that pollutant inputs from major point sources worsen downstream water quality, net of upstream pollution levels, and controlling for location-specific factors. We utilize panel data on monthly biochemical oxygen demand (BOD) concentration for a sample of 87 municipal and industrial plants located in the states of Maryland, Pennsylvania, and Virginia, for the period 1990-2003. Monthly dissolved oxygen (DO) from 67 locations is the measure of water quality. We find that an increase in aggregate BOD (for multiple plants) results in downstream net of upstream DO to decline by 0.001 mg/l. Despite the small magnitude (due to natural attenuation), the results are robust to distance traveled by pollutant and seasonal considerations of high temperature or low stream flow. We infer that point sources have a significant negative impact on ambient water quality net of non-point sources of pollution at upstream locations.

Keywords:

• US Clean Water Act
• ambient water quality model
• dissolved oxygen
• total maximum daily loads
• over-compliance

Acknowledgements

I thank the journal editor Hilda Blanco, anonymous reviewers; I am grateful for the generous guidance provided by Ted McConnell, Jay Shimshack, Erik Lichtenberg, John Horowitz and Maureen Cropper at the inception of this research.

Notes

1 The only significant policy change since the 2000s is that regulators focused on controlling non-point source pollution through flexible market-based instruments such as water quality based trading in some states (Olmstaed 2010).

2 The natural diurnal (daily) cycle of DO concentration is well documented. Dissolved oxygen concentrations are generally lowest in the morning, climbing throughout the day due to photosynthesis and peaking near dusk, then steadily declining during the hours of darkness.

3 It is defined as the minimum of 7 consecutive day average stream discharge having a recurrence interval of 10 years (7Q10). It is so called because the ability of a water body to assimilate pollutants without exhibiting adverse impacts is at a minimum

4 Quantity-based limit (lbs./day) = Design Flow (mgd) × Concentration based limit (mg/l) × 8.34 (lbs./day)/(mgd)(mg/l). Flow or volume of effluents is measured in million gallons per day (mgd).

5 Except, ${DO}_r$ (DO measured immediately upstream from effluent outfall) and ${DO}_w$ (DO measured in wastewater) are used rather than $L_r$ and $L_w.$

6 TMDLs utilize a steady-state model that is a modified Streeter-Phelps DO sag equation. The in-stream DO target for a TMDL is a daily average of not less than 5.0 mg/l for surface water.

7 Using the basic concept of Streeter-Phelps, many increasingly complex mathematical models have cropped up to accurately simulate DO dynamics in streams. “Most were developed to simulate parameters associated with [the NPDES] permits” (Vellidis et al. 2006, 1007), while some specifically simulated DO, others were broader in-stream water quality models, and yet others were watershed-scale transport models incorporating the contribution of non-point sources to water quality degradation. QUAL2E (Enhanced Stream Water Quality Model) is one of the two most popular (one-dimensional, steady-state) models for developing DO TMDLs (USGS 2005), while HSPF (Hydrological Simulation Program- Fortran) is a dynamic model. Soil and Water Assessment Tool (SWAT) is another example of a river basin model that quantifies the impact of land management practices in large watersheds, at the same time as simulating in-stream processes such as DO.

8 Most other water quality parameters such as nutrients, BOD, and biological indicators such as biomass and bacterial counts have been found to be non-normally (in particular, log normally) distributed (USEPA 1991; Gilliom and Helsel 1986).

9 Results are less precisely estimated upon using BOD concentration without the log transformation; however, negative impact of pollution is statistically significant for distance to downstream monitors less than 25 miles.

10 For example, there are three pairs of upstream and downstream monitoring locations with one plant on a tributary, whereas the others are on the main river. For these plants, we assign water quality data from the upstream and downstream monitoring stations on the main river, i.e. before and after the tributary joins it. There is no monitoring data available from the tributaries. Another plant could not be included since we could not identify its relevant upstream station. The plant discharged its effluents near the point of confluence of two tributaries.

11 Location specific “physical” effects such as velocity and depth (determining natural attenuation rates), pressure (and topography) and salinity (determining saturated oxygen levels) might be captured reasonably well by upstream water quality if the two monitors are close to each other.

12 For example, differences in land use across downstream locations since this is unlikely to change rapidly in a short period of time.

13 Seasonal average sum of BOD5 concentration was 12 mg/l during winter, 10.2 in spring, 8.5 in summer, and 10.1 in fall. Plants can reduce discharges during higher temperatures that support better efficiency in the biological processes involving wastewater treatment technology.