Table 1 Linear regression result for the most parsimonious models carried out under a Bayesian framework explaining the functional (FDis), taxonomic (Simpson) and phylogenetic (MPD) rates of diversity changes as a function of climatic and soil drivers.

From: Long-term droughts may drive drier tropical forests towards increased functional, taxonomic and phylogenetic homogeneity

Metric

Parameter

Median

HDI (l)

HDI (h)

HDI (l)

HDI (h)

HDI (l)

HDI (h)

ROPE

Rhat

50%

89%

95%

ΔFDisr

Intercept

6.34E−05

3.71E−05

9.35E−05

−1.69E−06

1.38E−04

−2.11E−05

1.54E−04

0.13

1.00

ΔMCWDAbs

1.40E−04

1.15E−04

1.73E−04

6.59E−05

2.08E−04

5.44E−05

2.35E−04

0.00

1.01

Plot area

−6.28E−05

−9.37E−05

−3.37E−05

−1.42E−04

8.89E−06

−1.63E−04

3.00E−05

0.15

1.00

ΔSimpsonr

Intercept

1.19E−04

7.19E−05

1.65E−04

2.92E−06

2.28E−04

2.35E−05

2.62E−04

0.08

1.00

ΔMCWDAbs

1.61E−04

1.02E−04

1.99E−04

3.58E−05

2.80E−04

8.16E−06

3.12E−04

0.00

1.00

Plot area

−8.15E−06

−5.23E−05

4.24E−05

−1.22E−04

1.04E−04

1.46E−06

1.30E−04

0.41

1.00

ΔMPDr

Intercept

−0.17

−2.03E−01

−1.50E−01

−0.24

−9.77E−02

−2.56E−01

−7.02E−02

0.00

1.00

PC1

−0.06

−6.71E−02

−4.32E−02

−0.09

−2.27E−02

−9.81E−02

−1.02E−02

0.00

1.00

PC2

−0.04

−4.78E−02

−2.59E−02

−0.07

−5.29E−03

−7.98E−02

4.51E−03

0.10

1.00

PC3

0.01

−7.65E−03

2.94E−02

−0.03

6.34E−02

−4.66E−02

8.42E−02

0.51

1.00

ΔVPDAbs

0.20

1.76E−01

2.28E−01

0.13

2.67E−01

1.01E−01

2.84E−01

0.00

1.00

ΔMCWDFull

−0.12

−1.63E−01

−9.79E−02

−0.21

−3.66E−02

−2.26E−01

−4.11E−03

0.00

1.00

ΔMCWDAbs

−0.03

−5.09E−02

−1.11E−03

−0.09

3.91E−02

−1.11E−01

5.91E−02

0.31

1.00

Plot area

0.06

4.35E−02

6.76E−02

0.02

8.74E−02

1.90E−02

1.07E−01

0.00

1.00

PC1: ΔVPDAbs

0.06

3.93E−02

8.23E−02

0.00

1.11E−01

−1.59E−02

1.32E−01

0.08

1.00

PC2: ΔVPDAbs

−0.02

−2.63E−02

−4.08E−03

−0.04

1.36E−02

−5.61E−02

2.05E−02

0.57

1.00

PC3: ΔVPDAbs

0.15

1.32E−01

1.84E−01

0.07

2.12E−01

4.82E−02

2.34E−01

0.00

1.00

PC1: ΔMCWDFull

−0.06

−9.72E−02

−3.46E−02

−0.15

2.23E−02

−1.58E−01

5.86E−02

0.12

1.00

PC2: ΔMCWDFull

−0.05

−6.26E−02

−3.45E−02

−0.08

−1.02E−02

−9.25E−02

2.29E−03

0.05

1.00

PC3: ΔMCWDFull

−0.17

−1.97E−01

−1.46E−01

−0.24

−1.05E−01

−2.50E−01

−7.43E−02

0.00

1.00

PC1: ΔMCWDAbs

0.03

1.71E−02

4.76E−02

−0.01

7.57E−02

−2.69E−02

8.32E−02

0.23

1.00

PC2: ΔMCWDAbs

0.08

6.28E−02

9.40E−02

0.04

1.17E−01

2.16E−02

1.29E−01

0.00

1.00

PC3: ΔMCWDAbs

0.11

8.78E−02

1.28E−01

0.05

1.56E−01

3.54E−02

1.78E−01

0.00

1.00

  1. Several different models were fitted (see Supplementary Tables S4 and  S6) to investigate the drivers of changes of each diversity facet. The most parsimonious model, shown above, was selected based on the leave one out cross-validation information criterion (LOOIC) and expected log predicted density (ELPD). Only the most statistically important interactions (lowest ROPE values, i.e., <0.10) are shown in Fig. 3.
  2. HDI highest density interval, l low, h high, ROPE region of practical equivalence to test the importance of parameters with values of 0 or close to 0 reporting a more significant effect, Rhat potential scale reduction statistic.