Appendix S1: Description of study landscapes and sampling
We sampled in four 10,000-ha landscapes located in the Atlantic plateau of São Paulo, encompassing parts of the municipalities of Ribeirão Grande/ Capão Bonito, Tapiraí/ Pietade, and Cotia/ Ibiúna, in the State of São Paulo, Brazil. The altitude of the study area varies between 800-1,000 m above sea level (Ross & Moroz 1997). Annual rainfall is between 1,220 and 1,810 mm and mean minimum and maximum temperatures are 17.3ºC and 28.4ºC for the warm-wet season (October to March) and 12.1ºC and 24.9ºC for the cool-dry season (April to September). The whole region was once covered with Atlantic forest, classified as Lower Montane Atlantic Rain Forest (Oliveira-Filho & Fontes 2000). Remaining forest patches in the landscapes are secondary regrowth in intermediate to advanced stages of regeneration, while the surrounding matrix is dominated by cattle pastures and agriculture. Although similar in terms of general forest characteristics (e.g. topography, relief, climate, type of forest, and type of human-use) and deforestation history, the fragmented landscapes differ in the proportion of remaining native forest cover, varying from ca. 86% (Cotia), to 49% (Tapiraí/ Piedade), 31% (Ibiúna) and 11% (Ribeirão Grande/ Capão Bonito). However, our study species, Marmosops incanus, was not captured in the 11% landscape. See Pardini et al. (2010) for further detail on the fragmented landscapes.
A total of 37 forest sites were sampled between 2002 and 2007: 12 sites in the 86% cover landscape, 14 fragments in the 49% cover landscape, and 11 forest fragments in the 31% cover landscape. Sample sizes per site did not differ significantly among landscapes (ANOVA, F1,2 = 0.7384, p = 0.485). There were no differences in climatic or study site conditions which might have biased our results. The mean size of sampled patches was 24.6 ha (standard deviation 44.3 ha), and 59.9 ha (standard deviation 70.1 ha) in the 49% and 31% cover landscape, respectively. Mean nearest neighbor distance of sampled fragments was 26 m (standard deviation. 27 m) and 19 m (standard deviation 20 m) in the 49% and 31% cover landscape, respectively. These values did not differ between the two landscapes (mean patch size: ANOVA, F1, 23 = 1.57, p = 0.223; mean distance to the nearest neighbor: ANOVA, F1, 23 = 0.4833, p = 0.494). Note that patch size and nearest neighbor distances cannot be calculated in the continuously-forested landscape (86%).
Additional references (not listed in main text)
Oliveira-Filho AT, Fontes MA (2000) Patterns of floristic differentiation among Atlantic Forests in Southeastern Brazil and the influence of climate. Biotropica 32: 793-810.
Ross JLS, Moroz IC (1997) Mapa Geomorfológico do Estado de São Paulo: escala 1:500.000. São Paulo: FFLCH-USP, IPT and FAPESP.
Table S1: Size range, fluorescent dye used and number of alleles detected per microsatellite locus within each landscape and across landscapes. For some loci, the genotyping was multiplexed and identical mix numbers indicate which loci very genotyped simultaneously.
Locus / Size range / Dye / Multiplexing / 86% / 49% / 31%
Mdo228 / 271-297 / HEX / Mix1 / 12 / 14 / 11 / 15
Mdo294 / 197-215 / 6-FAM / Mix5 / 5 / 4 / 4 / 5
Dm09 / 224-234 / 6-FAM / Mix3 / 6 / 6 / 3 / 6
Mdem09 / 244-268 / HEX / Mix5 / 13 / 12 / 9 / 13
Minc1 / 180-220 / HEX / Mix3 / 8 / 11 / 7 / 11
Minc2 / 121-143 / HEX / Mix1 / 7 / 8 / 6 / 9
Minc3 / 208-262 / HEX / Mix2 / 16 / 16 / 11 / 20
Minc4 / 150-166 / 6-FAM / Mix1 / 3 / 4 / 5 / 5
Minc7 / 178-208 / 6-FAM / Mix6 / 11 / 10 / 10 / 14
Minc8 / 368-400 / 6-FAM / Mix6 / 14 / 15 / 13 / 17
Minc9 / 375-393 / 6-FAM / Mix7 / 4 / 4 / 5 / 5
Minc10 / 201-207 / 6-FAM / Mix3 / 3 / 3 / 3 / 3
Total / 102 / 107 / 87 / 123
Table S2: Observed (Ho) and expected (He) heterozygosities in each sampling site within the three study landscapes.
.
Study Landscape / Name of sampling site / He / Ho86% landscape (12 sampling sites) / C / 0.561 / 0.565
Caucaia / 0.623 / 0.639
Ecletor / 0.583 / 0.590
Gigante / 0.526 / 0.524
Grilos / 0.626 / 0.583
Jararaca / 0.528 / 0.542
M1 / 0.575 / 0.619
M2 / 0.616 / 0.575
M3 / 0.616 / 0.614
Pelos / 0.609 / 0.582
Psicotica / 0.621 / 0.611
Quilombo / 0.647 / 0.627
Mean / 0.594 / 0.589
49% landscape (14 fragments) / Alce / 0.536 / 0.544
Bicudinho / 0.576 / 0.613
Cristo / 0.620 / 0.574
D_Yvonne / 0.553 / 0.587
Estrada / 0.583 / 0.611
Fuzue / 0.588 / 0.640
Haras / 0.510 / 0.542
Horacio / 0.550 / 0.617
Janzao / 0.520 / 0.558
Janzinho / 0.511 / 0.532
Medico / 0.584 / 0.573
Osasco / 0.596 / 0.591
Romulo / 0.585 / 0.613
Torres / 0.566 / 0.614
Mean / 0.563 / 0.586
31% landscape (11 fragments) / Dito / 0.621 / 0.644
DitoAndre / 0.595 / 0.544
Dragao / 0.578 / 0.597
Godoy / 0.585 / 0.617
Japones / 0.401 / 0.421
Lacerda / 0.510 / 0.480
Lila / 0.553 / 0.543
Nelson / 0.569 / 0.556
Pedroso / 0.538 / 0.533
Takimoto / 0.555 / 0.598
Tucano / 0.502 / 0.540
Mean / 0.545 / 0.551
Table S3: ANOVA results for the reduced data set (i.e., using only fragments with at least 10 sampled individuals).
ANOVA / F / p / F / p / F / p
50.39 / <0.0001 / 4.5928 / 0.0236 / 6.3662 / 0.008
Tukey's HSD / Difference / Adjusted p / Difference / Adjusted p / Difference / Adjusted p
86 vs. 49% / 0.110 / 0.051 / -0.056 / 0.003 / -0.026 / 0.007
86 vs. 31% / 0.466 / <0.0001 / -0.049 / 0.056 / -0.045 / 0.012
49 vs. 31% / 0.356 / <0.0001 / 0.007 / 0.689 / -0.019 / 0.121
Table S4: Overview of best models explaining genetic diversity and genetic differentiation in the 49% and 31% landscape when varying the assumed mean dispersal distance from 50 m to 1,600 m. The metrics leading to the best model(s) are given, and the highest adjusted R² values are shown. Results discussed in the main text are based on a mean dispersal distance of 300 m (shown in bold). This distance was chosen because it gave highest adjusted R² when modeling genetic variation in the 31% landscape. Note that in the 49% landscape, the constant model was still the best model (i.e., ∆AIC = 0). In the 31% landscape, the cumulative area of surrounding forest (‘metric 2’) and patch proximity (‘metric 6’) still led to the best models.
49% landscape / 31% landscapeAssumed mean
dispersal distance (m) / Resulting
alpha / Allelic
Richness / Genetic
Differentiation / Allelic
Richness / Genetic
Differentiation
50 / 0,02 / NA (constant model) / NA (constant model) / Metric 2, metric 6 / 0,36 / Metric 6 & others* / 0,29
100 / 0,01 / NA (constant model) / NA (constant model) / Metric 2, metric 6 / 0,36 / Metric 6 & others / 0,33
200 / 0.005 / NA (constant model) / NA (constant model) / Metric 2, metric 6 / 0,36 / Metric 6 / 0,41
300 / 0,0033 / NA (constant model) / NA (constant model) / Metric 2, metric 6 / 0,36 / Metric 6 / 0,42
400 / 0,0025 / NA (constant model) / NA (constant model) / Metric 2, metric 6 / 0,36 / Metric 6 / 0,39
500 / 0.002 / NA (constant model) / NA (constant model) / Metric 2, metric 6 / 0,36 / Metric 6 / 0,38
600 / 0,0017 / NA (constant model) / NA (constant model) / Metric 2, metric 6 / 0,36 / Metric 6 & others / 0,38
700 / 0,0014 / NA (constant model) / NA (constant model) / Metric 2, metric 6 / 0,36 / Metric 6 & others / 0,37
800 / 0,0013 / NA (constant model) / NA (constant model) / Metric 2, metric 6 / 0,36 / Metric 6 & others / 0,36
900 / 0,0011 / NA (constant model) / NA (constant model) / Metric 2, metric 6 / 0,36 / Metric 6 & others / 0,35
1000 / 0.001 / NA (constant model) / NA (constant model) / Metric 2, metric 6 / 0,36 / Metric 6 & others / 0,34
1300 / 0,0008 / NA (constant model) / NA (constant model) / Metric 2, metric 6 / 0,36 / Metric 6 & others / 0,34
1600 / 0,0006 / NA (constant model) / NA (constant model) / Metric 2, metric 6 / 0,36 / Metric 6 & others / 0,27
*Prox & others means that while the model involving Prox was best (delta AIC = 0), several other metrics also led to a delta AIC <= 2
Table S5: Comparison of patch-level models in the 49% landscape. AR = allelic richness, GD = genetic differentiation, AdjR² = adjusted R², AICc = AIC adjusted for small sample sizes.
.
Genetic Variable / Patch Metric / AdjR² / AICc / deltaAICcAR / NA (constant model) / NA / -21.517 / 0
AR / 5 / 0.047 / -21.306 / 0.211
AR / 2 / 0.044 / -21.267 / 0.25
AR / 6 / 0.022 / -20.956 / 0.562
AR / 3 / -0.049 / -19.969 / 1.548
AR / 1 / -0.05 / -19.955 / 1.562
AR / 4 / -0.053 / -19.91 / 1.607
GD / NA (constant model) / NA / -59.613 / 0
GD / 3 / -0.05 / -58.052 / 1.56
GD / 5 / -0.073 / -57.754 / 1.859
GD / 2 / -0.079 / -57.663 / 1.95
GD / 6 / -0.08 / -57.657 / 1.956
GD / 1 / -0.081 / -57.637 / 1.976
GD / 4 / -0.083 / -57.613 / 2
Figure S1:
Figure S1: Comparison of allelic richness estimates using the full and reduced data set (i.e., only fragments with at least 10 sampled individuals). Full and reduced data are displayed by black and grey triangles, respectively.
Figure S2:
Figure S2: Comparison of four alternative estimates of genetic differentiation using the full data set (black triangles) and the reduced data set (i.e., only fragments with at least 10 sampled individuals; grey triangles).