Confidence Limits for Estimates of Population Density in Line Transects

Publisher Website
Google Scholar
Abstract
The SD of density estimates, obtained for breeding land birds in line transects of the Finnish type, were studied in 2 regions in southern Finland. A multiple regression analysis revealed that SD can be estimated on the basis of 3 variables: the estimated average density of the species, the number of transects (median length 4.5-5 km) censused and a species-specific coefficient used in estimating densities. The regression models explained 94-95% of the variation in log SD. The parameters in the models were extremely similar, and therefore a new model was derived by pooling the data. Mean belt data yielded greater values of SD than survey belt data, especially if the density of the species was less than 5 pairs/km2. The models (one for the main belt and the other for the survey belt) were tested by using data from 11 independent regions in Finland and northern Norway. The observed SD values correlated highly significantly with the predicted ones (almost all correlations above 0.95). The correlation was equally high when all regions were analyzed simultaneously, showing that the pattern was similar in all regions. The estimated density did not correlate with the error made in the prediction. Because of the good fit of the predictions the method can be used to derive estimates of SD (and, as a result, confidence limits) for density estimates obtained in line transects, probably in wide geographical regions in Europe (or elsewhere in areas where general habitat structure is similar).

This publication has 0 references indexed in Scilit: