ASSESSING TEMPORAL CHANGES IN HIGHLY VARIABLE FECAL COLIFORM AND ENTEROCOCCUS DATA IN BOSTON HARBOR AND ITS TRIBUTARIES BY RANDOMIZED BLOCK FACTORIAL ANOVA

Hypothesis

Sewage indicator bacteria in waters receiving CSO have decreased in the period 1989-1996


INTRODUCTION

The challenge: find a way to measure changes in water pollution effects over time, despite many sources of variation in the environment.

Like many older urban areas, Greater Boston has an antiquated combined sewer system. These combined sewers carry wastewater to a treatment plant and also function as storm sewers; runoff from streets is carried in the same pipes as sanitary sewage. During heavy rainstorms, rainwater runoff mixed with sewage is discharged to rivers and Boston Harbor through combined sewer overflows (CSO's) as shown on Figure 1.

CSO's are a major source of bacterial pollution to Greater Boston's waterways, including beaches and shellfish beds.

Remediation of CSO's, at a cost of $500 million, is being implemented by the Massachusetts Water Resources Authority; the 20-year project is at its midpoint. Most CSO's are being eliminated or treated.

MWRA has been monitoring water quality in areas affected by CSOs--Boston Harbor and its tributary rivers--since 1989. Bacterial contamination is the worst problem caused by CSOs; monitoring has focused on fecal coliform and Enterococcus.

The challenge is to measure statistically significant changes in bacterial water quality over time as CSO's have been remediated. This is difficult, because the environmental impact of CSO's on bacteria counts in the water is strongly affected by many variables. These variables include duration and intensity of rainfall, groundwater levels, tidal cycles, water temperature, and salinity. Other sources of bacteria include stormwater runoff, animals, and wastewater treatment plants.

Initial analyses to detect significant changes in bacteria counts over time included linear regression, multiple regression, and analysis of covariance. These methods failed to detect significant changes in bacteria counts, because the huge variation in counts swamped the ability to detect change over time. The work presented here shows a new application of a classical statistical method, a Factorial Analysis of Variance Using Randomized Blocks, which by grouping data with similar sources of environmental variation, allows us to explore the effects of the factors we are interested in: rainfall and time.

The existing database

Fecal coliform and Enterococcus were enumerated using Standard Methods membrane filtration techniques (m-FC and m-Enterococcus media).

METHODS

Select experimental factors

2 Time periods

3 Rainfall variables

Identify key variables for blocking

8 Geographic regions

3 Tide conditions

3 Seasons

Partition data using randomized blocks

Perform the factorial ANOVA

Experimental factors:

Time

To simplify the analysis, the time period was divided into two variables:

  1. Before sewer system and CSO improvements (1989-1991), and
  2. After improvements began (1992-1996)

Rain

Because CSO's occur during wet weather, we wanted to explore whether significant changes in bacteria counts occurred during dry weather, heavy rains, or light rains. CSO's in this area generally only discharge if during storms >0.25 inches of rain. If significant changes in bacteria counts were found in dry weather or light rain only, then the changes probably would not be related to CSO improvements. Similarly, if changes were found in heavy rains, this would be consistent with changes in CSO effects. We created 3 rainfall variables based on the root-mean-square of three days of rainfall:

  1. RMS = 0 inches
  2. 0 inches < RMS <0.25 inches
  3. RMS > 0.25 inches

 

Key variables for blocking

Geographic area, tide, and season are expected to be significant sources of variation in bacteria counts beyond rainfall and sewer system condition.

Geographic areas

The rivers and embayments all have different potential CSO loadings and other characteristics. Based on our understanding of the area, eight geographic locations were chosen, and are shown on the map (Figure 1).

Tide

Combined sewer outlets are generally intertidal, and tidegates often prevent discharge except on the low tide. Three tidal conditions were defined: High tide was above mean sea level, low tide was below mean sea level, and freshwater was not affected by tides.

Season

Bacteria have different survival rates at different temperatures, and runoff characteristics are likely to vary seasonally. Three seasons were defined: Fall/winter, Spring and Summer.

Analysis

RESULTS

Fecal coliform

 

TREATMENTS

OVERALL

RMS=0 in

0 in<RMS<.25 in

RMS>.25 in

89-91

92-96

89-91

92-96

89-91

92-96

89-91

92-96

Mean (colonies/100 ml)

83

62

121

67

264

170

138

89

% Reduction

25%

45%

36%

36%

 

Enterococcus

 

TREATMENTS

OVERALL

RMS=0 in

0 in<RMS<.25 in

RMS>.25 in

89-91

92-96

89-91

92-96

89-91

92-96

89-91

92-96

Mean (colonies/100 ml)

23

25

30

31

94

68

40

37

% Reduction

-9%

-3%

28%

8%

 

 

Factorial ANOVA Results

Fecal coliform

Source of Variation

Degrees of Freedom

F

Blocks

32

34.69

Overall Treatments

5

25.63

Sampling Year Treatments

1

22.89

Rainfall Treatments

2

51.76

Rain/no rain

1

45.44

High/low rain

1

58.07

Error

153

 

Total

190

    

Factorial ANOVA Results

Enterococcus

Source of Variation

Degrees of Freedom

F

Blocks

32

27.39

Overall Treatments

5

22.25

Sampling Year Treatments

1

0.62

Rainfall Treatments

2

53.73

Rain/no rain

1

43.87

High/low rain

1

63.59

Error

153

 

Total

190

RESULTS

DISCUSSION

The statistical method was successful.

The F-values for the blocks easily exceed the critical values for the 0.5% level of significance for both fecal coliform and Enterococcus, showing that the blocking scheme captured variability due to geographic location, tidal condition, and seasonality. The randomized blocking scheme successfully allowed testing the hypothesis of decreasing bacteria counts during differing rain conditions by accounting for variations produced by environmental variables.

Enterococcus and fecal coliform showed different results over time with respect to rainfall.

While the mean overall fecal coliform counts dropped 36% after CSO controls were implemented, overall Enterococcus counts did not change significantly. Fecal coliform counts decreased between the year groups in all three rainfall categories: dry, moderate rain and heavy rain. Enterococcus counts were not significantly different over time during either dry weather conditions or moderate rain. However, Enterococcus counts measured during the wettest weather did drop by 28% after CSO remediation.

These different results for Enterococcus and fecal coliform may be due to several factors:

Future work: refine the analyses

More analyses are being done to focus within individual geographic areas to see where pollution remediation efforts have been most successful.