Supplementary figures
Fig. S1

Fig. S2

Fig. S3

Fig. S4

Fig. S5

Fig. S6

Fig. S7

Supplementary figure captions
Fig. S1
Observed cline in mean allelic richness (mean allelic richness over 22 loci per population as a function of distance to the south-westernmost population). Note that distances are in deme units (50 km). R
2
= 0.779.
Fig. S2
Observed isolation by distance (pairwise F
ST
as a function of pairwise distances). Note that pairwise distances are in deme units (50 km). R 2 = 0.310.
Fig. S3
Simulated map of the sampling locations and sampling sizes. Sampling sizes and sampling locations as for the observed dataset are indicated. Similar sampling locations and sampling sizes are generated for the simulated dataset. Colonisable demes are indicated in grey, sea demes (not colonisable) are indicated in white. The Iberian glacial refugium demes are indicated in dark grey. We use a Europe Albers Equal Area Conic projection to adequately represent surfaces.
Fig. S4
Observed isolation by distance for the microsatellites (pairwise F
ST
as a function of pairwise distances, same data as in Fig. S1) and for the colour data (pairwise P
ST
as a function of pairwise distances). Note that distances are in kilometres.
Fig. S5

Posterior distributions of the estimated parameters for the one-refugium carrying-capacitycline model based on 1000 retained simulations over ~10
6
simulations. Grey lines show prior density and black lines the posterior distributions. Note that carrying capacity and mutation rate are in logarithmic scale. The same smoothing parameters were used for all parameter estimates (i.e. Dirac peak width = 0.02). The estimated modes for the carrying capacities of the south-west and north-east corner of the map are very similar (205 vs. 166 respectively with massive overlap of the posterior distributions). The estimated model is thus very similar to the one-refugium one-carrying-capacity model and the difference in carrying capacity is in the expected direction.
Fig. S6
Posterior distributions of the estimated parameters for the one-refugium extinction-ratecline model based on 1000 retained simulations over ~10
6
simulations. Grey lines show prior density and black lines the posterior distributions. Note that carrying capacity and mutation rate are in logarithmic scale. The same smoothing parameters were used for all parameter estimates (i.e. Dirac peak width = 0.02). Both estimates for extinction rate fall in the lower part of the prior distribution, especially the one in the south-west, which would have a strong effect on the fate of the simulations.
Fig. S7
Posterior distributions of the estimated parameters for the one-refugium two-migration-rate model based on 1000 retained simulations over ~10
6
simulations. Grey lines show prior density and black lines the realised distributions. Note that carrying capacity and mutation rate are in logarithmic scale. The same smoothing parameters were used for all parameter

estimates (i.e. Dirac peak width = 0.02). The posterior estimates for the two migration rates are very similar. It is interesting to note that mode of the current migration rate is somehow higher than the one during colonisation, which might be constrained by the low genetic structure observed in the barn owl today.

Table S1. Multiplex composition and primer concentration for microsatellite genotyping.
Multiplex
Multiplex 1
Multiplex 2
Multiplex 3
Multiplex 4
Locus
Ta-206
Ta-210
Ta-216
Ta-306
Ta-218
Ta-220
Ta-204
Ta-214
Ta-305
Ta-310
Ta-413
Ta-202
Ta-212
Ta-215
Ta-402
Ta-408
FEPO42
54f2
Tgu06
Dye
FAM
HEX
FAM
NED
HEX
FAM
HEX
FAM
FAM
NED
NED
FAM
DYO630
FAM
NED
HEX
FAM
NED
HEX
Final Conc. [μM]
0.45
0.105
0.135
0.165
0.178
0.11
0.25
0.5
0.5
0.25
0.25
0.25
1
1
0.25
0.5
0.24
0.24
0.48
Multiplex 5
Calex-05
RBG18
DYO630
FAM
0.48
0.72
Oe053 HEX 0.96
Primer concentration is indicated for both forward and reverse primer together.