Table 1. Summary of the best models describing troglobitic species composition, species richness, rarity-weighted richness and phylogenetic diversity.
| Response variable | Modela | Predictor variables | Estimate | SE | t∕z | p |
|---|---|---|---|---|---|---|
| Species composition (MDS1) | LMM | Subterranean area (500 m) | 0.13 | 0.04 | 3.12 | 0.002 |
| Declivity 0–10°(500 m) | 0.09 | 0.03 | 2.98 | 0.003 | ||
| Mining area (250 m) | −0.08 | 0.03 | −3.01 | 0.003 | ||
| Species composition (MDS2) | LMM | Declivity 30–40°(500 m) | −0.06 | 0.02 | −2.74 | 0.006 |
| Cave density (50 m) | −0.08 | 0.02 | −3.23 | 0.001 | ||
| Lithology axis 2 (50 m) | −0.10 | 0.04 | −2.51 | 0.012 | ||
| Canga area (50 m) | 0.07 | 0.03 | 2.44 | 0.015 | ||
| Altitude | −0.06 | 0.03 | −1.82 | 0.070 | ||
| Species richness | GLMM | Distance to nearest creek | 0.13 | 0.04 | 3.04 | 0.002 |
| Geomorphology axis 2 (50 m) | 0.20 | 0.05 | 3.85 | <0.001 | ||
| Agriculture area (50 m) | −0.15 | 0.04 | −4.29 | <0.001 | ||
| Cave area | 0.30 | 0.03 | 10.11 | <0.001 | ||
| Presence of guano | 0.35 | 0.07 | 5.33 | <0.001 | ||
| Rarity-weighted richness (log-transformed) | LMM | Distance to nearest creek | 0.28 | 0.1 | 2.87 | 0 |
| Geomorphology axis 2 (50 m) | 0.27 | 0.11 | 2.56 | 0.01 | ||
| Canga area (50 m) | 0.25 | 0.08 | 3.34 | <0.001 | ||
| Cave area | 0.5 | 0.07 | 7.55 | <0.001 | ||
| Presence of guano | 0.54 | 0.14 | 3.89 | <0.001 | ||
| Phylogenetic diversity | LMM | Distance to nearest creek | 176.25 | 64.69 | 2.72 | 0.007 |
| Subterranean area (50 m) | 125.46 | 49.04 | 2.56 | 0.011 | ||
| Geomorphology axis 2 (50 m) | 180.44 | 71.08 | 2.54 | 0.012 | ||
| Agriculture area (50 m) | −203.45 | 51.50 | −3.95 | <0.001 | ||
| Cave area | 551.18 | 53.99 | 10.21 | <0.001 | ||
| Presence of guano | 390.94 | 93.70 | 4.17 | <0.001 |
Notes.
Response variables are shown followed by the type of model employed (Model), the predictor variables included in the best models (selected through likelihood ratio tests with α <0.05), estimates, standard errors (SE), t/z-values and p-values.
a
Linear mixed models (LMM) or Generalized Linear Mixed Models (GLMM). All models accounted for spatial autocorrelation and unaccounted variation between study sites by keeping the highland were the caves were located as a random effect. Eight out of the 473 caves were excluded from these analyses because they contained missing data in at least one field.