Recovery Plan for the Northern Spotted Owl - DRAFT

impractical at present. We describe components of a monitoring plan that webelieve would provide the data needed to estimate trends either for considerationof delisting or to gain assurance that serious declines would be detected.One step in such an analysis is estimating the magnitude of fluctuations thatmight be expected in spotted owl populations that were stable and 'healthy."No data for such calculations are available for spotted owls, but the estimatesare important because we would not expect trends to be exactly zero, even if apopulation were fully recovered. In any given period, a population probablywould be increasing or decreasing slightly and would have roughly a 50percent chance of declining slightly. Thus, some effort must be made tounderstand the magnitude of trend that might be considered normal and toincorporate this information into sample size guidelines for the monitoringprogram.We examined the Breeding Bird Survey data sets described earlier to helpdetermine natural levels of variation in populations that are stable or close tostable. Four of the fifteen populations showed positive and negative trendsduring the 25-year period. We estimated both trends in these cases, obtaininga total of 19 trends. Five were negative and 14 were positive. About half (42percent) of the absolute trends exceeded 3 percent per year and two-thirdsexceeded 2 percent per year. The preponderance of positive values may havebeen caused by a slight overall increase in these populations at the regional ornational level (Droege, pers. comm.) or perhaps by a general increase in surveyorskill (Peterjohn, pers. comm.). If we subtract 2 percent from all values,then 9 of the 19 trends (i.e., about hall) are negative and half are positive. Inthis case, 37 percent of the absolute trends exceeds 3 percent per year and 42percent exceeds 2 percent per year. These results suggest that average annualchanges, over periods of up to 25 years in statewide populations of raptors, arecommonly as large as 2.5 percent or 3 percent. Smaller populations probablyexhibit somewhat stronger fluctuations, so annual changes in a single provinceof 3.5 to 4 percent may be common.A second factor to consider in developing sample size guidelines is the cost ofobtaining the estimate. Estimated sample sizes for detecting trends of 2percent, 2.5 percent, and 3.5 percent, with either 8 or 12 years of data, arepresented in Table A. 10. They were computed initially from the general regressionequation described earlier and were modified by inspecting tables of actualpower as calculated by the computer simulations. The regression equationand the results in Table A. 10 suggest that detecting a trend in 8 years takes500 to 700 more stations per year than detecting the same trend in 12 years(compare adjacent columns in Table A. 10). This guideline holds only if thetrend during the 8-year period is linear. If trends are nonlinear then 8 yearsmay be insufficient for estimating trends regardless of how many stations arevisited per year. If 12 years of data are available for the estimate, then detectingtrends of 2.5 percent would require approximately 800 stations (dependingon number of owls recorded per 100 stations), and detecting trends of 3.5percent would require approximately 300 stations per year. Similar calculationshave not been carried out yet for the demographic analyses, but theconfidence intervals for the estimated trend, based on 5 to 7 years of data,varied from 0.04 to 0.09 suggesting that 10 or more years would be required todetect trends as small as 2.5 percent, and longer might be needed if substantialyear effects were present.A final factor worth considering in developing sample size guideline is how theestimate of trend will be combined with other information in determiningwhether populations should be delisted. We believe that conclusions about thelong-term stability of the population should not depend solely, or even primarily,on empirical estimates of trend. On the contrary, these data probably276