Control of experimental material and implementation of treatments is many times easier on paper than it is in practice. I discuss the constraints and trade-offs involved in designing experiments in ecology in the context of two particular cases. The first case is the first U.S.-based experiment on the effectiveness of changing turbine cut-in speed on reducing bat fatality at wind turbines. Our objectives were to 1) determine if delaying turbine rotation until winds had reached pre-determined speeds would result in fewer bat fatalities, and 2) determine the economic costs of the delay. We defined three treatments that were of operational interest and that could be implemented at a turbine over an entire night. I will discuss unique conditions that muddied the definition of the experimental unit, prevented consistent application of the treatments and introduced bias into our observations. And I will discuss how we addressed these issues in our study design. The second case involves genetics studies, in which seedlings collected from certain locations are assumed to be genetically adapted to those particular site conditions. They are then planted at other sites, of varying ‘ecological distance’ from their origins and traits reflecting their productivity are measured and compared. The objectives are to 1) separate site factors from genetic factors that lead to optimal productivity, and 2) predict productivity in the face of climate change. In this study the distinction between random and fixed effects is blurred and what constitutes true replication of a factor is not clear. I discuss these issues and suggest potential approaches to analysis of data of this type.
Kidder Hall (campus map) |
|
350 |
|
Maggie Neel |
|
1 541 737 1981 |
|
neel at science.oregonstate.edu |
|
Statistics (Science) |
