Table 2. The top-ranked set of ordinal regression models relating leaf damage of A. hippocastanum with the number of years that C. ohridella was predicted to have been present.

Model type*	Number of C. ohridella	Break-point	ΔAIC	AIC
	generations to model distribution†	(years)		weight
Log-linear segmented	2-generation model	3	0	0.395
Linear segmented	2-generation model	3	0.4	0.319
Linear segmented	3-generation model	4	2.7	0.105
Log-linear segmented	2-generation model	4	4.0	0.053
Log-linear segmented	3-generation model	4	4.7	0.038
Log-linear segmented	2-generation model	7	6.1	0.019
Log-linear segmented	2-generation model	n/a	6.2	0.018
Log-linear segmented	2-generation model	6	6.9	0.013
Log-linear segmented	2-generation model	2	8.5	0.006
Linear segmented	2-generation model	2	8.5	0.006
Log-linear segmented	3-generation model	7	8.6	0.005
Linear	3-generation model	n/a	9.0	0.004
Log-linear	3-generation model	n/a	9.0	0.004
Log-linear segmented	3-generation model	6	9.1	0.004
Log-linear segmented	2-generation model	5	9.5	0.003
Linear segmented	2-generation model	4	9.6	0.003

Models are ordered by relative model fit (ΔAIC) and only those with ΔAIC<10 are shown here.

The form of the relationship of years present with leaf damage. The segmented regression models show a linear or log-linear relationship up to the break-point, after which the relationship is constant.

All the top-ranked regression models were based on the modeled distribution of C. ohridella, assuming either 2 or 3 generations per year. Regression models including the directly observed year of arrival were included in the candidate model set, but all had ΔAIC>10.7, so are not shown here.