Albert de Jong
sub-adult cachinnans: December
Identification of Caspian Gulls: phenotypic variability and the field characteristics of hybrids
Summary of key findings
• First-winter Caspian Gulls varied enormously, with few stable traits. The traits most useful for separation from Herring Gull were bill shape (length/depth ratios), greater-covert pattern, scapular moult and primary projection. Nonetheless, species’ scores for each of these traits overlapped to some degree, indicating that they should not be used in isolation.
Appendix 2. Principal Component Analysis (PCA).
Output from a hypothetical Principal Component Analysis is shown in the diagram below. It visualises the distribution of a number of sample birds (dots) and the traits (arrows A-C) used to characterise each one. In this example, trait A can be interpreted as being the most important one in driving the overall differences between the birds, because it has the longest arrow. Conversely, trait C, because it has the shortest arrow, varies least between the birds and so is not responsible for the major differences between them. Trait B is intermediate, but clearly less important than A and rather similar to C in the proportion of the overall variance that it explains. Trait A lies nearly parallel with the horizontal axis (PC1) and so the separation of birds from left to right in the diagram relates to their values of this trait. The arrowheads point in the positive direction of the trait values; thus, if trait A was the overall size of the bird, then larger birds would be to the right. For example, bird 1 in the figure is a large bird as it sits close to the end of the trait arrow (where it falls on the arrow is shown by the vertical red line). Traits radiate from the origin (0,0), and by convention are shown only in their increasing (positive) direction. The equivalent part of the trait arrow going in the opposite direction, on the other side of the origin, is not normally shown. This decreasing part of the arrow has to be visualised as the 180° projection of the increasing one, back through the origin. Just for illustrative purposes here, this decreasing line tor trait A is shown as a dashed line in the figure. Birds 2 and 3 sit at the extreme end of this negative line and so have the lowest values in the sample for this trait (i.e. they are the smallest birds). Arrows for traits B and C lie very close together. This indicates that these two traits are correlated, i.e. those birds which have high values for trait B will also have high values for trait C. If they were traits related to bill and leg brightness, for example, the analysis would show that the birds with the brightest legs also have the brightest bills, and vice versa. These traits point tangentially (more or less 90°) to trait A. This means that they are not correlated, i.e. there is no relationship between bare-part brightness and the size of the bird.
Appendix 3. Relating trait scores to multivariate space.
3a. Relationship between PCA axis 1 and trait scores for sample adult gulls. The principal component value can be approximated by the equation given on the scatter; while the correlation is not ideal (as some portion of original variation was captured by further principal components), it is very high and sufficient to show that scores characterize phenotype well enough to detect between-species differences. By reading from the regression line, it is possible to convert a sum trait score to a PC1 position, and thus locate a bird within multivariate space. Known hybrids are given by red and blue symbols: they all have one or two known parents (most birds of the latter group), are offspring of different pairs and have been scored as adults. Red circles are known F1 hybrids (offspring of Caspian x Herring Gull
3b. Relationship between PCA axis 1 and trait scores for sample first-winter birds. Triangles show Caspian Gulls, open circles Herring Gulls and red circles hybrids.
|Larus cachinnans 4CY UKK L-005760 December 11-12 2012, Oss, the Netherlands. Picture: Theo Muusse & Lützen Portengen.|
|Larus cachinnans hybrid 4CY & 5CY 07P2 December 17 2009 and January 14 2010, Pohlsche Heide, Germany. Picture: Armin Deutsch.|
|Larus cachinnans hybrid 4CY-7CY 81P1 December 2009 - November 2012, Roermond, the Netherlands.|
|Larus cachinnans 4CY & 8CY 7P16 February 06 2007, December 11 2007 & December 07 2011, Mur de Lixhe, Belgium. Picture: Charly Farinelle.|
|Larus cachinnans 2CY-4CY XNCD 2013-2015, France & the Netherlands. Picture: Pierre Tillier, Ran Schols & Willy Raitière.|
| Larus cachinnans 3CY-5CY PLG DN-03919 January 2006 - November 2008, Deponie Pohlsche Heide - Minden, Germany (52°23'05N, 08°46'45E).
Picture: Armin Deutsch.
|Larus cachinnans sub-adult, 01 December 2011, Boulogne-sur-Mer, NW France, Picture: Jean-Michel Sauvage.|
|Larus cachinnans sub-adult, December 29 2009, Westkapelle, the Netherlands. Picture: Ies Meulmeester.|
|Larus cachinnans 4CY, December 21 2012, Ashdod, Israel. Picture: Amir Ben Dov.|
|Larus cachinnans sub-adult, December 21 2012, Ashdod, Israel. Picture: Amir Ben Dov.|
|Larus cachinnans sub-adult, December 02 2008, Dambovita canal, Bucharest. Picture: Cristian Mihai.|
|Larus cachinnans sub-adult, December 23 2010, Ashdod, Israel. Picture: Amir Ben Dov.|