Table 1. Model selection among variants of the Fokker-Planck diffusion model describing beetle dispersal in the field experiment.

			Model parameters
Model	NLL	ΔAIC	μ 1	μ 2	ξ	ω 0	ω 1	π 2
			(m2 d−1)	(m2 d−1)	(d−1)	(m−2)	(m−2)	(-)
1	944.0	0.0	215		0.066	0.17	0.09	1.5
2	944.0	2.0	218	212	0.066	0.17	0.09	1.5
3	948.9	7.9	216		0.066	0.15	0.09
4	949.0	10.0	228	220	0.065	0.15	0.09
5	959.5	29.1	180		0.071	0.13		1.2
6	960.7	29.6	183		0.071	0.12
7	959.5	31.1	175	187	0.071	0.13		1.2
8	960.7	31.5	175	194	0.071	0.12
9	997.3	104.6	217			1.46	0.35	1.8
10	997.3	106.6	219	216		1.47	0.35	1.8
11	1000.9	111.9	165	288		1.27	0.35
12	1002.1	112.3	214			1.25	0.35
13	1023.2	154.6	200			0.95		1.5
14	1023.4	154.8	140	316		0.89
15	1022.7	155.4	159	256		0.94		1.3
16	1025.5	157.1	202			0.86

A single parameter value is shown when habitats or trap types were not distinguished in a model variant. The value of π ₁ is 1 throughout. If no value for π ₂ is shown, it was set to 1 and not included in the calibration. The negative log-likelihood (NLL) is a measure of the goodness of fit of the model to the data. ΔAIC is the difference in Akaike's information criterion between a model variant and the model variant with most support of the data (model 1). Models variants of which the ΔAIC is smaller than two are considered equivalent and have equal support from the data.

μ ₁: motility in oilseed radish; μ ₂: motility in rye; ξ: relative loss rate due to removal other than recapture (e.g. mark wear and mortality); ω ₀: trap-coefficient for trapping stations without screens; ω ₁: trap-coefficient for trapping stations with screens; π ₁: multiplication factor of the flux of beetles from oilseed radish to rye; π ₂: multiplication factor of the flux of beetles from rye to oilseed radish.