Posted by Stephanie A. Smith, Ph.D., on Fri, Jan 3, 2014 @ 11:00 AM
If you’re standing at the bank of your favorite swimming hole with your rubber ducky ring around your waist and your flippers on, and the water has a scum of blue-green algae on it, are you going to jump in? Simon sampled on Tuesday and got 500 ppb microcystin; Simona sampled a few hours later but hers came back reading only 5 ppb microcystin, well below the World Health Organization’s recreational guidelines. It’s Wednesday. Are you going to swim, or aren’t you? And while we’re at it, why aren’t you at work on a Wednesday anyway?
Let’s do a little recap of Simple Samples Parts I & II for the purpose of addressing this scenario. I attempted in Part I to introduce the concept of sampling, and how important careful sampling is to scientists in order to draw conclusions. How do you know, for instance, whether Simon’s or Simona’s samples (or neither!) were representative of the situation you’re faced with today? I hoped that Part II would demonstrate that most lake sampling that can be done will be biased, due to time, space, and money constraints. Maybe Simon is a geek and was just curious how high microcystin could get so he went for the thick stuff. Maybe Simona charges folks $5 for swimming and so sampled at a time when the water looked more like it does when people would typically swim.
Note that neither Simon nor Simona was more “correct” with their sampling—both were biased. The challenge in our scenario regards extrapolation. Is it a good idea to extrapolate either Simon or Simona’s data to this very place and point in time?
And therein lies the rub. You might believe it is safe to swim, based on Simona’s microcystin numbers, when in fact it may not be, and the scum before you contains a very high concentration of microcystin toxin to which you will be exposed. However, if you believe it’s dangerous to swim, when it’s actually perfectly safe, you may as well take off those ridiculous flippers and go back to work. That’s not an awesome outcome, either, and Simona will be upset because it hurts her business.
It is my opinion that extrapolation can be as dangerous as or worse than biased sampling. Sampling bias can be disclosed, and sometimes is better tolerated if supporting information was collected: date and time, exact site, depth, and size of the sample. If you’re lucky accompanying measurements like pigments, turbidity, or temperature may have been taken. These are things that I advise people to ask about before accepting someone else’s extrapolation, or statements like “It’s perfectly safe!” or “I wouldn’t touch that water with a 10-foot pole!”
Find out more about how your swimming hole is sampled for microcystin, and whether other data are collected. Better yet participate in the sampling process through programs like CLAM, and suggest other pieces of information that might be helpful. Beagle will be happy to make recommendations in that regard. For now, ditch the ducky and get back to work.