Oklahoma Weather Effects on E. Coli in Surface Water ad Produce Safety
Justin McConaghy
Produce Safety Program Coordinator
Oklahoma Department of Agriculture, Food, and Forestry
International Food Protection Training Institute (IFPTI)Fellowship in Food Protection
Abstract
This study examined the correlation between weather data (rainfall, temperature, barometric pressure, wind, solar radiation, dew point, and humidity) and the Escherichia coli (E. coli) counts from surface water sampling in the state of Oklahoma to determine if weather monitoring is a feasible approach to estimate E. coli count in surface water supplies used for fresh produce irrigation. If so, such weather monitoring could allow farmers to determine which days may be safer than other days for irrigation. The study examined weather data from the Oklahoma Mesonet system from 2001 to 2018 and water data from the Oklahoma Water Resources Board (OWRB) and the Oklahoma Water Survey. The study concluded that there were no substantial correlations between the weather data and surface water E. coli overall. However, E. coli in some individual water sources could be correlated with a variety of weather factors, the most common being the amount of rainfall. The study recommended that farmers who regularly test their water also keep weather records to determine if weather is correlated with individual surface water contamination, and if so, adjust their irrigation schedule to possibly limit contaminating fresh produce. The study also recommended that further research should be conducted to determine how weather may affect water sources not included in this study, primarily small ponds and irrigation canals. More research also should be conducted to determine why Oklahoma’s man-made lakes are not impaired by E. coli even though they capture water from rivers and creeks that were contaminated upstream.
Keywords: surface water, irrigation, fresh produce, Produce Safety Rule, E. coli, weather, mesonet, rain, temperature, wind, barometric pressure, solar radiation, dew point, humidity
Background
Fresh produce poses a risk of foodborne illness as fresh produce is usually eaten raw rather than going through a pathogen kill step. Between 1998 and 2008, produce contamination accounted for 38.1% of hospitalizations and 22.9% of deaths attributed to foodborne illnesses (Painter et al, 2013). Despite the risk of eating fresh produce, these foods have important nutritional benefits that suggest increasing their consumption. In an attempt to reduce the prevalence of foodborne illness outbreaks, the Produce Safety Rule of the Food Safety Modernization Act (FSMA) was passed in 2011 to regulate the food safety aspects of the growing, harvesting, packing, and handling of fresh produce on farms.
One controversial portion of the Produce Safety Rule is the testing of irrigation water for generic Escherichia coli (E. coli) in order to limit the fecal coliform contamination of fresh produce due to irrigation (Karst, 2017). This portion of the Rule has been delayed for two additional years to allow for further research on the topic (FDA, 2019). While this rule delay was being proposed, it had received criticism from consumer groups, especially after a recent outbreak associated with romaine lettuce in April 2018 was found to have been caused by contaminated water in the irrigation canal (Fox, 2019; FDA, 2018).
Section 303(d) of the Clean Water Act (United States Environmental Protection Agency [EPA], N.D[CH(1] .) requires states to study and develop a list of public waters in the state that are impaired by various water quality factors. A study of the map of the Section 303(d)-listed surface waters in Oklahoma impaired by E. coli brings up some interesting observations (see Figure 1). Figure 1 illustrates that the western half of the state has considerably more E. coli impaired rivers and streams than the eastern half; additionally, no lakes in the state are impaired by E. coli. A potential explanation of why the lakes are not impaired is that the water could be more diluted and is more stationary thus allowing the E. coli to settle into the sediment on the bottom of the lake.
Figure 1. Map of EPA 303(d) E. coli impaired waters in Oklahoma
Potential explanations of why more western rivers are impaired by E. coli include: lesser rainfall in the western half of the state resulting in the E. coli not being diluted and washed downstream as often, or different vegetation in the western half of the state such as shorter grasses and less trees resulting in increased runoff and less natural filtering of runoff waters.
Another potential explanation of why western rivers are more impaired by E. coli would be a heavier concentration of livestock. However, the United States Department of Agriculture (USDA) (2012) National Agricultural Statistics Service Census maps do not appear to substantially support this hypothesis. Poultry in Oklahoma is concentrated along the eastern counties that border Arkansas and cattle in Oklahoma looks evenly distributed across the entire state. Swine production could explain the E. coli concentration in the panhandle and portions of the northwest quarter of the state, but would not explain the impaired waters in the southwest quarter of Oklahoma.
This project sought to determine if monitoring the local weather could help predict when the E. coli loads in surface waters would increase or decrease, and therefore reduce the potential for contamination of produce fields that are irrigated using surface water. Farmers commonly monitor the weather as a major determining factor in deciding when and what to plant and harvest, which ultimately affects the crop’s profitability. If a farmer could also monitor certain aspects of local weather that is found to affect the E. coli levels in surface waters, then such an approach would provide an inexpensive, easy to follow model that would help reduce the risk of E. coli contamination of fresh produce.
Problem Statement
The relationship between weather events and E. coli load in surface waters in Oklahoma is unknown at this time.
Research Questions
What is the relationship between weather data from the Oklahoma Mesonet system with the Oklahoma Water Resources Board and Oklahoma Water Survey water testing data collected from 2001 to 2018?
Can the relationships found between weather data and water data provide useful guidance for farmers?
Is the method used in this project applicable to other states?
Methodology
Data for this study was collected from the Oklahoma Water Resources Board (OWRB), Oklahoma Water Survey, and Oklahoma Mesonet. The OWRB data used method 9223B (EPA, 2017[CH(4] ) to test for total generic E. coli in 100 ml of water from rivers, streams, and lakes from across the state on random non-consecutive days during the months of May through October, the growing and harvesting months for most produce grown outdoors in Oklahoma, between 2001 and 2015. This water data was one of the data sources used in the determination of EPA 303(d) impaired waters list for the state of Oklahoma. Additional water data was used from the Oklahoma Water Survey’s 2018 water tests on 23 various streams across the state with 10 tests per site collected from May through September.
Daily weather data from the nearest Oklahoma Mesonet station to each water collection site was compared to the E. coli data to determine correlation. Oklahoma Mesonet weather data is collected every five minutes from over 110 weather monitoring sites uniformly scattered across the entire state in locations that minimizes influences from anthropogenic, water, land slope, and wind obstruction factors, and the data goes through a quality assurance system to ensure accuracy (Brock et al., 1995) (McPherson et al., 2007). The weather data that was used in this study as independent variables was the maximum air temperature (MAXTEMP), minimum air temperature (MINTEMP), average air temperature (AVGTEMP), maximum wind speed (WINDMAX), minimum wind speed (WINDMIN), average wind speed (WINDAVG), maximum barometric pressure (PRESSMAX), minimum barometric pressure (PRESSMIN), average barometric pressure (PRESSAVG), total solar radiation (TOTSOL), inches of rainfall (RAIN), maximum rate of rainfall in five minutes (MAXFIVE), soil temperature under sod at 5 cm (SODTEMP), dew point (DEWPNT), and average humidity (HUMAVG).
Additional variables that were calculated from the Mesonet weather data included the change in barometric pressure for the day (PRESSCHG) and the number of days since a daily rainfall of 0.20 inches or more of rain (RAINDEL). Nevers and Whitman (2005) found that there was typically a spike in E. coli in the beaches and a tributary of Lake Michigan following a rainfall of greater than 0.5 cm (.1968 inches) in 24 hours. Whitman and Nevers (2008) also found that wave height was useful in predicting E. coli in water of Chicago beaches, and that increase in wave heights can occur during a change in barometric pressure signifying storm conditions.
To reduce the amount of data to a feasible work load, the water sample data was reduced to only the samples that contained 100 coliform forming units (cfu) or more of E. coli. This decision was made based on the recommendation of Dr. James Bidlack, Professor of Biology at the University of Central Oklahoma (J. Bidlack, personal communication, December 10, 2018). This level is below the geometric mean decision level of 126 cfu that is established in the Produce Safety Rule. This choice of criteria reduced the water data points from over 10,000 water tests to 2,036. In addition, some weather data points were excluded from the analysis due to a loss of data from non-functioning sensors.
Since the 303(d) map (Figure 1) of E. coli impaired water bodies looked to favor the Western half of Oklahoma, a variable was added to the data to account for whether the water test site was on the West or East side of the Indian Meridian, which is a geographic central dividing line of the state.
Analysis of the data was completed by using the Pearson Correlation in the SAS computer program to determine the linear correlation between the E. coli cfu and each independent variable of weather.
Results
A statewide correlation using all of the data had significant p-values in WINDMAX, PRESSMAX, PRESSMIN, PRESSAVG, TOTSOL, RAIN, and MAXFIVE. A p-value of 0.05 or less is considered in most research as the threshold to consider a variable to be statistically significant, and the lower the p-value the less likely of having a false positive. While the p-values of these variables were all 0.05 or less, they all had very low correlations. The highest correlation coefficient was RAIN with 0.1355, while the rest of the variables had correlation coefficients less than 0.1. A positive correlation in rainfall was expected since runoff from surrounding land could pick up fecal material from domestic animals and wildlife and deposit the E. coli bacteria into the tested water supply. This increase in E. coli bacteria from rainfall was short-lived as the number of days since a rainfall 0.2 inches or greater was not a significant variable. Therefore, the data analysis suggests that a rain event only caused an increase in E. coli on the date that the location had rainfall, and the bacteria was likely dispersed downstream, settled out of the water, or had died by the next day.
Examination of the data for western Oklahoma found no weather variable on the more arid western half of the state to be statistically significant (p-value < 0.05). However, in eastern Oklahoma there were several significant weather factors with only WINDMIN, all barometric pressure measurements, and DEWPNT not being statistically significant. The correlations were higher in the East than in the statewide analysis. RAIN in the East once again had the highest correlation coefficient with a value of 0.24366, with TOTSOL and MAXFIVE at -0.14666 and 0.14579, respectively. These coefficients for eastern Oklahoma suggest that the more volume of rain in a day and a higher rate of rainfall would result in slightly higher E. coli counts in the water and that solar radiation had a slight negative effect on the E. coli. All air temperature variables and SODTEMP had negative correlations, meaning a higher temperature resulted in lower E. coli amounts.
Dividing the water sources by stream or lake resulted in streams having significant but very low correlation (under 0.1) in MAXTEMP, AVGTEMP, WINDMAX, PRESSMAX, PRESSMIN, PRESSAVG, TOTSOL, and MAXFIVE. RAIN was once again the highest correlated variable with a coefficient of 0.16479 and a p-value of <0.0001. Lakes were unique in that the only significant variables were PRESSCHG and HUMAVG with coefficients of 0.17128 and -0.19623, respectively. The increase in pressure change resulting in a slight increase of E. coli is similar to what was found from the Lake Michigan study by Whitman and Nevers (2008), which showed changes in pressure change and wave height were interrelated (which may occur during storm conditions) and a greater wave height increased the E. coli in the water along the lake’s beaches (Nevers and Whitman, 2005).
Additionally, correlations were conducted on the 24 individual sample collection sites that had 20 or more samples with an E. coli cfu of 100 or greater. Narrowing down the data by site resulted in many highly correlated weather variables, but each site was definitely affected by different factors. One example of a site highly correlated with weather factors was the Big Cabin Creek (see Table 1) with significant variables (p-value < 0.05) being WINDMAX, PRESSCHG, TOTSOL, RAIN, MAXFIVE, and HUMAVG, and the respective correlation coefficients being 0.63286, 0.57492, -0.48977, 0.92785, 0.89297, and 0.42327. The correlation coefficients being closest to +1 or -1 have the most impact on the E. coli cfu. Therefore, RAIN and MAXFIVE with the respective correlation coefficients at 0.92785 and 0.89297 and very low p-values of < 0.0001 implies that both an increase in the amount and rate of rainfall are very likely to increase the E. coli in the water at the Big Cabin Creek test site.
Table 1
Six sites of the 24 used in this targeted analysis resulted in having no significant correlations with any weather variables. The weather variables with the most significant correlations were RAIN and MAXFIVE with 7 sites that were all positively correlated, which once again suggests that increased rainfall volume and an increase in the maximum rate of rainfall on the day of testing is likely increasing E. coli in water at select sites by picking up fecal material through runoff from the surrounding land. TOTSOL had 6 significant sites that were all negatively correlated, suggesting that solar radiation could be eliminating the E. coli in the water at select locations. While MAXTEMP, WINDMAX, and HUMAVG also had 6 significant sites each, they all had a mix of both positive and negative correlations.
Conclusions
This study has shown that a statewide recommendation on monitoring weather to predict when E. coli concentrations in surface water would not be advisable since the correlation coefficients were quite low. However, individual sites may be highly correlated with weather, and therefore monitoring could be used on certain individual sites. If a farm were to perform a similar in-depth study with each individual water source, there is potential that weather monitoring guidelines could be created for each site to limit the potential of contaminating fresh produce with fecal coliforms.
The most significant and highest correlated weather variable throughout the study was rainfall volume on the day of testing, suggesting that farmers may want to avoid irrigating produce for one day after a rain. The number of days since a 0.2 inch or greater rain was rarely a significant variable, suggesting that waiting more than one day is likely not necessary unless an individual site’s analysis suggests waiting longer. Not irrigating within 24 hours of a rain should already be a common practice as irrigation is not typically needed when there is significant rainfall.
With the Eastern half of Oklahoma having many significant variables and the more arid climate in the Western half of the state having no significant variables, these differences between East and West in this study show that using weather as a guidance in reducing the potential of fecal coliforms (E. coli) in surface water may be more feasible in Eastern states than in the more arid locations found in Western states.
Recommendations
Farmers that are regularly testing surface water should keep weather records to determine if there are weather events that affect each individual surface water source. If there are any correlations found, then that information could be used to adjust their irrigation schedule to limit contaminating fresh produce.
Further research should be conducted to determine how weather may affect small ponds and irrigation canals, not included in this study, which are commonly used as irrigation sources.
In addition, research should also be conducted to determine why Oklahoma’s man-made lakes are not impaired by E. coli even though they capture water from rivers and creeks that were contaminated upstream.
Acknowledgements
First, I greatly appreciate the Oklahoma Water Resources Board, Oklahoma Water Survey, and Oklahoma Mesonet for their work in collecting the data over the years and providing it for use in my research. I also would like to thank Dr. James E. Bidlack of the University of Central Oklahoma for his direction and assistance in programming and analyzing of the data in SAS. I want to recognize the IFPTI staff, mentors, and Paul Dezendorf for their assistance in this project and development of an excellent fellowship program. A special thanks to my mentor, Dan Gump, for his guidance. Finally, I would like to thank my boss, Bryan Buchwald, for recommending and allowing me to participate in the fellowship program, and a special recognition to all of the members of Cohort 7.
References
Bidlack, J. (2018, December 10). Personal communication.
Brock, F. V., Crawford, K. C., Elliott, R. L., Cuperus, G. W., Stadler, S. J., Johnson, H. L., & Eilts, M. D. (1995). The Oklahoma mesonet: A technical overview. Journal of Atmospheric and Oceanic Technology, 12. 5-19.
Fox, M. (2019, January 17). Food safety falls short in the U.S., consumer group says. NBC News. Retrieved from https://www.nbcnews.com/health/health-news/food-safety-falls-short-u-s-consumer-group-says-n960011
Karst, T. (2017, November 14). Produce rule’s water testing delay will cause more illness, groups say. The Packer. Retrieved from https://www.thepacker.com/article/produce-rules-water-testing-delay-will-cause-more-illness-groups-say
McPherson, R. A., Fiebrich, C., Crawford, K. C., Elliott, R. L., Kilby, J. R., Grimsley, D. L., … Demko, D. B. (2007). Statewide monitoring of the mesoscale environment: A technical update on the Oklahoma mesonet. Journal of Atmospheric and Oceanic Technology, 24, 301–321.
Nevers, M. B., & Whitman, R. L. (2005). Nowcast modeling of Escherichia coli concentrations at multiple urban beaches of southern Lake Michigan. Water Research, 39(20), 5250-5260.
Painter, J., Hoekstra, R., Ayers, T., Tauxe, R., Braden, C., Angulo, F., & Griffin, P. (2013). Attribution of foodborne illnesses, hospitalizations, and deaths to food commodities by using outbreak data, United States, 1998-2008. Emerging Infectious Diseases, 19(3), 407-415.
U.S. Food and Drug Administration (FDA). (2018). Environmental assessment of factors potentially contributing to the contamination of romaine lettuce implicated in a multi-state outbreak of E. coli O157:H7. Retrieved from https://www.fda.gov/Food/RecallsOutbreaksEmergencies/Outbreaks/ucm624546.htm
U.S. Food and Drug Administration (FDA). (2019). FDA finalizes new compliance dates for agricultural water requirements. Retrieved from https://www.fda.gov/Food/NewsEvents/ConstituentUpdates/ucm633483.htm
United States Department of Agriculture (USDA). (2012). Census of agriculture: 2012 census ag atlas maps – livestock and animals. Retrieved from https://www.nass.usda.gov/Publications/AgCensus/2012/Online_Resources/Ag_Atlas_Maps/Livestock_and_Animals/
United States Environmental Protection Agency (EPA). (2017). Analytical methods approved for compliance monitoring under the revised Total Coliform Rule. Retrieved from https://www.epa.gov/sites/production/files/2017-02/documents/rtcr_approved_methods.pdf
United States Environmental Protection Agency (EPA). (N.D.). Overview of identifying and restoring impaired waters under Section 303(d) of the CWA. Retrieved from https://www.epa.gov/tmdl/overview-identifying-and-restoring-impaired-waters-under-section-303d-cwa
Whitman, R. L., & Nevers, M. B. (2008). Summer E. coli patterns and responses along 23 Chicago beaches. Environmental Science Technology, 42, 9217-9224.