Testing the validity of discriminant function analyses based on bird morphology: the case of migratory and sedentary blackcaps Sylvia atricapilla wintering in southern Iberia
Published: Volume 54(1), June 2007. Pages 81-91.
Aims: This study introduces a method for validating the suitability of discriminant function analyses (DFA) based on morphological differences among groups of birds, using migratory and sedentary blackcaps Sylvia atricapilla wintering in southern Iberia as a case study.
Location: During five summers between 1997 and 2005, migratory blackcaps were studied in Álava and Madrid, and sedentary blackcaps were studied in the Campo de Gibraltar area. Wintering birds were also studied in the Campo de Gibraltar area (nine winters: 1997 - 2006), where migratory and sedentary blackcaps occur in sympatry.
Methods: Morphological traits (wing length, wing pointedness and tail length) of birds captured during summer were used to carry out a DFA for distinguishing between migratory and sedentary individuals during winter. The performance of DFA was tested by (1) observation of the percentage of correct classifications of sedentary blackcaps (ringed birds captured in both seasons), and (2) analysing variation between summer and winter in posterior classification probabilities (PC) of birds classified as either migratory or sedentary.
Results: The DFA correctly classified 88 % of summer blackcaps, and it still worked well in winter as shown by 20 out of 21 wintering birds known to be sedentary (by ringing recoveries) being correctly classified. PC increased during winter in birds classified as migratory, due to the arrival to southern Iberia of northern birds with higher migratory-like trait values. As expected from the fact that the sedentary population stays in the area through the year, sedentary blackcaps were classified with similar PC in both seasons.
Conclusions: DFA based on the morphological differentiation between migratory and sedentary blackcaps correctly classifies individuals during sympatric wintering periods, which is when the DFA is meant to be used. This study shows that a suitable way to validate morphological DFA is comparing the PC of birds used to generate the functions with the PC of birds that need to be classified. Analysis of PC is suggested as a routine test of the performance of DFA in ornithological research.