Statistics for environmental monitoring

Abstract

Environmental studies can be broadly categorised as either: i) Observational experiments, or ii) Manipulative experiments. An observational experiment is one where the data is collected by observing some existing process (Manly 1992). For example, the population of Hector's dolphins around Banks Peninsula was surveyed each year from 1990 to 1995 to estimate the trend in the population size once the Banks Peninsula Marine Mammal Sanctuary was put in place. Generally, observational studies are conducted where the environmental process is of interest but the mechanism driving the process may not be well understood. In contrast, a manipulative experiment is conducted when specific variables are altered and the outcome, or effect, is estimated (Manly 1992). If the monitoring of Hector's dolphins had begun well before the establishment of the Sanctuary the survey could have been considered an "experiment" in that the population trend prior to and after the Sanctuary establishment could have been compared. In this example, the protection status of the dolphins would have been the variable that was changed. While there is a lot more control over confounding effects in designed experiments often they can not be conducted. For example, it is not possible to design an experiment to measure the effect of an oil spill in a harbour because ii would be difficult to get permission to create an artificial spill. · As another example of an observational study compared with an experiment consider a study on the effects of changes in weather on the likelihood that a possum will be caught in a trap. The study could be conducted as an observational experiment. By repeatedly trapping a population of possums over time differences in the number of traps catching a possum could be related to differences in weather, such as rain or fine night, air temperature, wind direction. This study would need to be carried out over a long time period to be able to capture as much variation in weather as possible. There is a potential problem with such a design though. Any observed differences in possum-catch are likely to be confounded by changes in possum behaviour - a possum caught once maybe weary of being caught again. For example, perhaps the weather progressively deteriorated throughout the study so that at the start of the study there were fine-weather nights and bad-weather nights near the end. An observed reduction in the proportion of traps catching a possum may appear to be due to differences in weather but in fact due to possums having a learnt aversion to traps. There may be confounding factors that were not even considered. Possum-catch may reduce over time simply because the possum diet had shifted between the start and end of the study. A manipulative experiment would involve using captured possums in pens and artificially creating different temperatures, rainfall, and wind within each pen. The study could be then carried out over one night. With such a design there would be strong evidence that the observed differences in trap-catch was an effect of different temperature, or rainfall, or wind etc. rather than some behavioural aversion.