Rapid assessments of recreational water quality River Rally May 4, 2008 Donna Francy, Amie Brady, and Rebecca Bushon USGS Ohio Water Science Center
Today’s Agenda Introduction to public health microbiology and bacterial indicators Recreational water quality monitoring and assessment Rapid analytical methods Predictive models
Waterborne disease is a public health concern Pathogens ingested from drinking water recreational water contaminated fish or shellfish Potential sources of pathogens treated and untreated sewage septic tanks combined sewer overflows landfills animal waste
Incidence of recreational waterborne disease outbreaks in the US Associated Illness Incidence Gastroenteritis 30 (48.4%) Dermatitis 13 (21.0%) Acute Respiratory Illness 7 (11.3%) Others: Amebic Meningoencephalitis, Meningitis, Leptospirosis, Otitis Externa, Mixed Illnesses 12 (19.3%) Etiologic Agents Identified: Bacteria 32.3% Protozoa 24.2% Virus 9.7% Chemical or Toxin 4.8% Centers for Disease Control and Prevention (CDCP)
Factors effecting incidence of waterborne Illness in the US Increasingly greater threat to public health Increase in population Aging water-treatment systems Aging population Inadequately managed animal wastes Lack of integrated regulatory approach
Types of waterborne pathogens Protozoa (larger, complex cells) Viruses (tiny, non-living) Bacteria (medium size, simple cells)
Pathogens and swimming-associated illnesses Disease Cryptosporidium parvum Cryptosporidiosis Giardia lamblia Diarrhea Naegleria fowleri Headache, fever, and vomiting Campylobacter jejuni Gastroenteritis E. coli pathogenic strains Salmonella, Shigella, and Vibrio species Various enteric fevers, gastroenteritis, cholera, dysentery, septicemia Adenovirus Respiratory and gastrointestinal infections Hepatitis Infectious hepatitis (liver malfunction); also may affect kidneys and spleen Norovirus
Why don’t we test directly for waterborne pathogens? Safety May require direct manipulation of pathogenic organisms Time and cost Each pathogen must be detected using a different test Requires processing of large volumes of sample Pathogens usually are present in low concentrations
Indicators of fecal contamination Indicator organisms usually NOT pathogenic "indicate" the possible presence of pathogenic organisms Used to directly detect the presence of fecal contamination from warm-blooded animals E. coli
An Ideal Indicator Organism Has an easy testing procedure Is of human or animal fecal origin Survives as long as, or longer, than pathogens Present at densities related to the degree of fecal contamination Is a "surrogate" for many different pathogens Useful in fresh and saline waters Enterococci
Indicator Organisms—Problems Present when there is no fecal contamination Total coliforms and C. perfringens are found in soil so they are not exclusively indicators of fecal contamination Absent when pathogens are present E. coli may die off faster than viral pathogens Density may not always relate well to the density of pathogens E. coli can reproduce in warm, tropical waters
Methods for detecting indicators Membrane filtration method
Methods for detecting indicators Enzyme substrate tests
Tests for Escherichia coli Membrane filtration method Modified mTEC agar E. coli colonies are magenta colored after incubation.
Tests for Escherichia coli Enzyme substrate test Colilert E. coli positive wells fluoresce under UV light. (Total coliforms positive wells are yellow under ambient light.)
Membrane filtration method Enterococci colonies have a blue halo. Tests for Enterococci Membrane filtration method mEI agar Enterococci colonies have a blue halo.
Enterococci positive wells fluoresce blue under UV light. Tests for Enterococci Enzyme substrate test Enterolert Enterococci positive wells fluoresce blue under UV light.
Recreational water quality monitoring and assessment
Definitions Criteria Are not rules and do not have regulatory impact. Scientific data and guidance on the potential human health risk involved in the water’s use (or in acceptable limits for aquatic life).
Definitions Standards Have regulatory impact Rules set forth by the state or USEPA to protect users of waters (based on water-quality criteria) For indicator bacteria, based on a quantifiable relation between density of the indicator in water the potential human health risk by the water’s use
Criteria for recreational waters Relation between E. coli and swimming-associated gastrointestinal illness
Criteria for recreational waters Relation between fecal coliforms and swimming-associated gastrointestinal illness
USEPA criteria for recreational waters—1986
USEPA BEACH PROGRAM Beaches Environmental Assessment and Coastal Health Act of 2000 (BEACH) Required the states to adopt criteria into standards Required USEPA to develop new criteria by late 2005 Required USEPA to address research topics such as modeling/monitoring, exposure and health effects Provided grants to the states and local governments to develop new monitoring programs Provided national beach guidance http://www.epa.gov/beaches
Need for rapid assessments At Ohio beaches, advisories are issued if E. coli is > 235 Culture results take 18-24 hours Water quality may change during that time
Solutions for rapid assessments Rapid analytical methods Description of qPCR and IMS/ATP Rapid method studies Predictive models State of the science Nowcasting at Ohio beaches and a recreational river
SURFACE RECOGNITION (IMS/ATP) SOLUTIONS? RAPID ANALYTICAL METHODS MOLECULAR (qPCR) SURFACE RECOGNITION (IMS/ATP)
qPCR RAPID METHOD Target and amplify specific DNA sequences Standard curve is created from analyzing known quantities of target organism Unknown sample values are interpolated from the standard curve Unknowns
Field Testing of qPCR Method USGS and Northeast Ohio Regional Sewer District Samples were collected from July – September 2006 and 2007 at two Lake Erie beaches - Edgewater and Villa Angela Objective: Compare results obtained by qPCR to those of the conventional membrane-filtration method Project funded by the Ohio Department of Health
E. coli Standard Curve y = -3.5551x + 44.835 R2 = 0.9497
Correlations between qPCR and membrane filtration for E. coli
qPCR results – Villa Angela 2007 Predictive tool Sample size Correct Sensitivity Specificity qPCR 38 87% (33/38) 91% (21/23) 80% (12/15) Previous day’s E. coli 33 73% (24/33) 70% (14/20) 77% (10/13)
Next Steps for qPCR DNA extraction Test different extraction kits Analyze samples daily or weekly – not in batch format qPCR Find alternate source for assay reagents Determine the best data interpretation procedure Transfer technology to local agencies
IMS/ATP RAPID METHOD Immunomagnetic separation (IMS) Uses antibody-coated magnetic beads which bind to antigens present on the surface of cells Adenosine triphosphate (ATP) Energy molecule in all cells Reported in Relative Light Units (RLU) M A G N E T
Field Testing and Technology Transfer of IMS/ATP Method Water samples from Ohio Lake Erie beaches In cooperation with local and state cooperators (2005-2007) Water samples from a recreational river (CVNP) In cooperation with federal cooperators (2004-2006) Sewage samples from Ohio, NC, and CA plants and water samples from Avalon Beach, CA In cooperation with Southern California Coastal Water Research Project (SCCRWP)
Correlations between IMS/ATP and membrane filtration for E. coli Cuyahoga River at Jaite, 2006 r=0.55 r=0.67 r=0.89
Villa Angela 2007: Relations to E. coli Early Summer Late Summer IMS/ATP method r = 0.49 p = 0.0034 r = -0.0094 p = 0.9628 Turbidity r = 0.40 p = 0.0228 r = 0.25 p = 0.2013 Number of birds r = 0.41 p = 0.0166 r = 0.15 p = 0.4739
Next steps for IMS/ATP method Continue refinement of IMS/ATP Identify additional antibodies that include most strains Optimize the beads and reagents Test at other locations Test whether it is a stand alone method or can be used in existing models, integrate in predictive models Epidemiological study - SCCWRP Transfer technology to local agencies
SOLUTIONS? PREDICTIVE MODELS RAINFALL BASED ALERTS Stamford, Ct (20 years) Door County and Milwaukee, Wi Southern Ca (10 years) Delaware (12 sites) Myrtle Beach, SC Boston Harbor Ozaukee County, Wis (may use in 2008)
MULTI-VARIABLE STATISTICAL MODELS SOLUTIONS? MULTI-VARIABLE STATISTICAL MODELS Linear relations between variables and E. coli Use statistical techniques such as multiple linear regression Beach specific models Does not require identification of the source r=0.56 P<0.0001
SOLUTIONS? STATISTICAL MODELS OPERATIONAL MODELS Project SAFE, IN (4 beaches, 3 yrs) SwimCast, Lake County, IL (3 beaches, 1−3 yrs) Chicago, (2 beaches, begin in 2008?) Nowcast, Ohio, (1 beach, 2 years)
NOWCASTING AT OHIO BEACHES Rainfall Turbidity Wave height Lake level Water temp Wind direction Day of the year
NOWCASTING AT OHIO BEACHES Threshold probability ranges from 27 to 32% Huntington and Edgewater Wave height Turbidity 48 hr weighted rainfall (Airport) Antecedent dry days Radar rainfall Day of the year Lake level Huntington, Bay Village Output from the model is the probability that E. coli will be >235 CFU/100 mL Threshold probability ranges from 27 to 32%
Edgewater wave height buoy
Ohionowcast.info
NOWCAST results in 2007 Model 78 82.1% 50% (11/22) 94.6% 66 66.7% Edgewater, Cleveland, Ohio Predictive tool Sample size Correct Sensitivity Specificity Model 78 82.1% 50% (11/22) 94.6% Previous day’s E. coli 66 66.7% 33.3% (6/18) 79.2%
Next steps for NOWCAST Villa Angela and Lakeshore, Ohio Standard was exceeded on the majority of days tested Correct responses—Model 61-63%, Current method 73- 75% Consider using QPCR or IMS/ATP Huntington and Edgewater Improve performance of the model Enable real-time measurements
Cuyahoga Valley National Park Cuyahoga River
Develop predictive models IMS/ATP rapid method results Streamflow Turbidity Rainfall
Environmental variables in relation to E. coli concentrations Pearson’s r correlation coefficients (number of samples) Year IMS/ATP rapid method Rain* Estimated Discharge Log10 turbidity 2004-5 0.55 (78) 0.56 (140) 0.43 (125) 0.78 (141) 2004-6 0.61 (118) 0.59 (179) 0.56 (164) 0.82 (180) *Rainfall data gathered from the Automated Flood Warning System: www.afws.net
CVNP – Turbidity model 2007 Data: Model results vs. Actual concentrations
CVNP – Turbidity model Jaite – 2007 data Predictive tool Sample size Correct Sensitivity Specificity Model 31 81% 94% 64% Previous day’s E. coli 26 68% 62% 83% Cuyahoga Valley N.P. Resource Management Office and Lab
Next Steps - CVNP Funded projects: Continue data collection in 2008 Implement the models to the public Continue data collection in 2009 -10 Post model results on NOWCAST website Outreach – news releases, factsheet Transfer technology to NPS Other: Use of real-time turbidity sensor at site
Acknowledgements Christopher Kephart, Erin Bertke, Robert Darner, USGS Ohio Water Science Center Jill Lis, Cuyahoga County Board of Health Suzanne Britt, Ann Gliha, and Patricia Boone, Cuyahoga County Sanitary Engineers Eva Hatvani, Mark Citriglia, Lester Stumpe, NEORSD Mark Pfister, Lake County Health Department, IL Richard Whitman, USGS, Porter, IN Calum McPhail, Scottish Environmental Protection Agency Meg Plona, Cuyahoga Valley National Park Ohio Lake Erie Commission Ohio Water Development Authority