The first analytical technique (percent occurrence) is the simplest. It calculates an annual index to a particular species' abundance, as the number of stations on which the species was found, expressed as a percentage of the total number of stations sampled in that year. These annual percentages are regressed on year and the slope is compared to 0 by t-test. This technique does not correct for factors such as the uneven geographic distribution of routes, or the fact that every route is not run in every year.
The other three, more complex analytical techniques were developed with the support of the U.S. Geological Survey, based on their procedures for determining "estimating equations" for federal Breeding Bird Survey data (Link and Sauer 1994). For each route an index to abundance was computed for each species in each year. In the route frequency regression technique the number of stations of occurrence was added (range = 0-10); it is a measure of how many stations a species occurs on, regardless of how abundant it may be on these stations (call index values are not considered). The route index regression technique summed the index values from each station (range = 0- 30). In the route adjusted-index regression technique we arbitrarily assigned to each call index value a number that we felt better estimated the relative abundance of animals represented by that value (call index 1 = 3 calling males, call index 2 = 25 calling males, call index 3 = 50 calling males); these values were then summed for each route as in the preceding technique (range = 0-500). For a given species the route index regression equates 3 stations each recording index values of 1 with a single station recording an index value of 3; the route adjusted-index regression equates about 17 stations recording an index value of 1 with one station recording a value of 3.
Trends for each route were then computed using estimating equations, regressing the appropriate dependent variable (frequency, summed index values, or summed adjusted-index values) on year, and these trends were averaged for all routes in each ecoregion, expressed as mean annual percent change for that ecoregion. Before averaging, however, the trend for each species on each route was weighted according to the relative abundance of the species on that route (routes with a high average frequency of occurrence or a high mean index value contributed more to the estimated ecoregional trend than did routes in which the species was less common), and an estimate of the variance in the trend estimate (routes in which the trend was precise contributed more than routes in which the trend was imprecise). The mean trend for each ecoregion was then weighted according to the area of that ecoregion, and these weighted means were averaged to produce a standard statewide trend (Geissler and Sauer 1990).
Trend results are presented in tabular form, for 3 periods: 1984-88 (ending in the middle of the late 1980s drought years), 1989-98 (beginning in the middle of drought), and 1984-98 (all 14 years of survey). The type of analysis is indicated in parentheses following each species name:
The table columns are as follows:
Interpreting trend results: Keeping in mind the assumptions and limitations of each analytical technique, estimates from all 4 techniques are useful for comparison. For example, when 2 or 3 techniques estimate similar trends for a species it strengthens the evidence for these trends. Note that trend estimates are summaries of population change, and do not provide information on other patterns of population change (such as cycles) over time. Some of these patterns are suggested by the graphs of annual population indices for each species. The statewide trend estimates may also mask trends that are occurring at a smaller (perhaps more meaningful) scale; for example, opposing trends could occur in different regions in the state but "cancel each other out" in the statewide trend estimate. Trends calculated from samples with fewer than about 14 routes should be considered tentative, even when P-values are significant.