Sample representativeness: A must for reliable regional lake condition estimates

Reliable environmental resource estimates are essential to informed regional scale decisions regarding protection, restoration, and enhancement of natural resources. Reliable estimates depend on objective and representative sampling. Probability-based sampling meets these requirements and provides accuracy estimates (confidence limits). Non-probability-based (judgment or convenience) sampling often is biased, thus less reliable (no accuracy estimates), and potentially misleading. We compare results from a probability- and a non-probability-based Secchi transparency sampling of lakes in the northeastern geographic region of the United States and its three primary ecoregions. Results from these samplings are compared on the basis of sample representativeness relative to the regional lake population and subsequent reliability of lake condition estimates. Statistically derived sampling indicates the northeast lake population median lake size to be about 9.5 (+/- 2.3) ha and the Secchi disk transparency (SDT) to be about 2.4 (+/- 0.4) m. On the basis of judgment sampling estimates, the median SDT for lakes in the same area would be 4.2 m. However, only about 15% of the regional lake population based on statistically designed sampling estimates has a SDT greater than or equal to 4.2 m. Estimate unreliability of this magnitude can have profound effects on lake management decisions. Thus, regional extrapolation of non-probability-based sampling results should be avoided.