Environmental Monitoring

Environmental Monitoring
   



The guts of any experimental program, and what will define its limits, are the choices about where?, when?, what?, and how?.

While sampling sites should be realistically accessible, to ensure OH&S, the primary concern is that they make sense in terms of the questions you are asking (and not just be easy to get to!).

Typically in sampling you may need sites that represent:

  • control and impact (positively or negatively);
  • before and after impact;
  • an unbiased sampling of all relevant environments;
  • relevant events (e.g. flood vs normal flow);
  • the natural variability in the system, including natural cycles (e.g. diurnal or seasonal patterns) and ensure that similar times are compared;
  • a meaningful measure of what is important (rather than what is easy to measure)!


These choices will be part of determining the kind of design and, consequently, the kind of analysis you can do. Not thinking enough about these issues can result in an invalid (i.e. useless) program whereas even in the worst situations good thinking can ensure you do the best you can with the least waste of resources.

Inappropriate use of statistical tests can give you invalid interpretations. Most statistical tests are based on assumptions about the characteristics of the data. If a test assumes normal data (e.g. the mean) test the distribution of your data.


Many tests incorporate randomness to eliminate bias but sensible design will ensure you get the best possible results. "Stratification" can make sure you sample appropriately for the problem: make sure you get representative samples of relevant communities/habitats; sample events, times, etc that are relevant to the question not just at "random". See Hurlbert for a more extensive discussion of this issue.

It has been said that "statistics don't prove anything, they just show how surprised you are"! If you are very surprised by your results you don't necessarily reject the hypothesis automatically; you need to reconsider your assumptions (which are part of your hypothesis) and to think rigorously about the design and interpretation to ensure that there is a physical basis for the (unexpected) results. This where the learning can really happen.


Can you afford the program you think you need? If not, what can you afford that is worth doing?


   
 

Hosted by Keysoft Pty Ltd