HR: 1340h
AN: H33D-0506    [Abstracts]
TI: Evaluating the Management Implications of Scientific Research
AU: * Krogstad, F
EM: fkrogsta@u.washington.edu
AF: University of Washington, College of Forest resources, seattle, wa 98195-2100
AB: When doing scientific research, we collect some data, apply statistical analysis, and then discuss its management implications. These management implications generally consist of interpretations and/or calculations to make the inferred statistical statements meaningful to decision makers. As with any other simplification, a management implication is an imperfect representation, which can recommend management actions quite unlike those that would have been recommended by the data or statistical results. There is no rigorous framework however for evaluating management implications that is comparable to the use of experimental design in the collection of data or statistical inference in the evaluation of hypotheses. A framework for evaluating management implications can be constructed by noting that the goal of management implications is to guide management decisions. A good management implication can be defined as one that recommends management actions similar to those that would have been recommended by the data or statistical results. The true unsimplified management implication of the data can in turn be defined as, "the probability of alternate outcomes of a proposed action given the observed outcomes of past actions." This is know in statistics as the posterior predictive distribution and can be directly calculated. Applying the posterior predictive distribution to alternate management actions identifies the actions that are recommended by the data. These recommendations in turn can be compared to the recommendations of alternate simplified management implications to identify the best simplified management implication. A less rigorous framework can be created by simply trying to use the results of classical statistics to predict the consequences of proposed actions. The potential consequences of the non-rigorous approach (and the utility of this proposed framework) can be seen by revisiting the previous analyses of the impacts of timber harvest on peak streamflow in the H.J. Andrews experimental forest. Applying this framework to the same data and models shows that contrary to the previous management implications, there actually is strong evidence that timber harvest has large impact on peak streamflows and that this impact increases with larger floods. Similar results could also have been produced if the previous results had been used to predict the consequences of alternate actions.
DE: 6309 Decision making under uncertainty
DE: 1860 Runoff and streamflow
DE: 1869 Stochastic processes
SC: Hydrology [H]
MN: 2004 AGU Fall Meeting