Table 2.

Performance of microclimate- vs. habitat area-based metapopulation models in predicting observed occupancy of habitat patches by Hesperia comma in two time periods. 500 simulations were run for the periods 1982–2000 and 2000–2009, starting with observed occupancy. Table shows the Akaike Information Criterion (AIC) based on the log-likelihood that observed occupancy was correctly predicted a) for individual habitat patches (612 in 2000; 906 in 2009), or b) for all habitat patches grouped per 5 km square in which they occur (n = 115 × 5 km squares in 2000, 149 in 2009). b) represents a more conservative measure which takes account of likely spatial autocorrelation of nearby patches. ΔAIC is positive where the microclimate model out-performed the area model

Test data	Start year	1982	2000
	End year	2000	2009
a) Independent patches	AIC Habitat area model	728.2	841.4
	AIC Microclimate model	688.6	799.1
	ΔAIC	39.5	42.3
b) All patches per 5 km square	AIC Habitat area model	440.8	550.9
	AIC Microclimate model	434.9	535.5
	ΔAIC	5.8	15.4