Bulk download data by holding shift and dragging a rectangle across the screen above the sites you wish to obtain data for. A .zip file containing CSVs of the data for all available selected locations will then be preparred for you to download. Click icons for the latest data at a particular location, as well as an option to plot or download data historical and up-to-date data for that site.
Choose to plot Raw Data, Weekly Aggregated Data, or Normalized Weekly Aggregated Data by Location. Choose from over 140 parameters to plot either Raw Data or Weekly Aggregated Data. The Raw Data will always take a bit longer to load because some files exceed 100MB in size. Columns in tables containing greater than 20% NANs were removed from the aggregate and normalized datasets entirely. Rows of the remaining columns which contained NANs were removed from the aggregate and normalized datasets before aggregation. Normalized datasets were normalized using scikit MinMaxScaler. Relationships between variables are most easily explored using the aggregate and normalized datasets.
Please note the following features for data exploration when using the interactive plotting tools:
Download selected data in CSV format. Data for the following sites are updated weekly:
Collaborative efforts between multiple agencies, inlcuding Wayne State University and Healthy Urban Waters are beginning a real-time monitoring system at Lake St. Clair Metropark. You can visit this portal to view real-time data of your local metroparks!
We have labled all discussed measurement units of Escherichia coli into only CFU/100ml on the publc facing water quality portal.
The standard measurement unit for Escherichia coli is cfu/100 mL, where cfu is an abbreviation for colony forming units.
The term cfu was initially based on actual counting of colonies on filters place on agar plates (or equivalent). An E. coli specific substrate is usually used, along with moderately selective media to enable counting only E. coli colonies.
Methods like IDEXX Quantitrays use liquid culture methods (hence does not actually count colonies), does multiple cultures of known volume and dilution, and infers the number of cfu from the number of cultures that have E coli growth versus those that do not have growth (again using E. coli specific substrates and selective media). The inferred number is determined by a statistical method, with the number given as being the MPN, or most probable number, and takes into account the possibility that a positive culture could have one or more than one E. coli to start with; whereas, a negative culture is known to have none. The statistical method can also calculate the 95% confidence interval of the MPN.
Now, you might be wondering why use a method that is only "most probable" rather than an actual count. The IDEXX method is the simpler method since it doesn't involve filtering onto filters and statistically gives the same average numbers. The decision to open or close a beach is a statistically based one anyway, involving calculating the geometric mean of several replicate samples. In effect, since you are using averages anyway, you are still dealing with a most probable number (in a general statistical sense, not in the technical MPN method sense). Also, even the counting on plates may have large variability between replicates, depending on dilutions of the sample that might be made before plating, the actual volume filtered through the filter, or, in related methods, the actual volume plated.
A comparison between a plate counting method and the IDEXX Quantitray method can be found in the this paper that was published by Vail et alia a number of years ago. In this paper, Petrifilm plates were used to do a colony count method that employed only 1 mL samples--much less volume than the standard 100 mL filtered volume. It was compared to two plate count methods and the IDEXX Quantitray (see Figure 2 in the paper--you will note that we used "cfu/100 mL" as the units in all cases). We asked: how many replicate plates should we do with 1 mL volume in order that the 95% confidence interval of the average count is at least as good as the IDEXX Quantitray. At the "beach closing" criterion level of 300 cfu/100 mL (which would be an average of just 3 colonies on each 1 mL plate), we found that using 3 replicate plates gave about the same 95% confidence interval. Obviously, if we had a plate that did an actual 100 mL, there would be no need to use statistics to get "cfu/100 mL". But there is always a need for replicates anyway, so the Petrifilm is a reasonable method--it's very easy, suitable for citizen science, and equivalent with adequate replicates to methods that use a full 100 mL. In the discussion, we also noted the following: "a membrane filtration measurement of water having 300 cfu/100 mL would typically use a 10-fold dilution to yield 30 cfu on the filter (to be in the count range of the method). As noted in American Public Health Association (1998; Method 9222) 'membrane counts really are not absolute' and are assumed to follow a Poisson distribution."