Table 9.
The relative importance of explanatory variables in explaining spatial variation in the abundance of SI (stream intersect), VC (variably connected) and NCO (no connection observed) wetland classes
| SI wetlands | Spearman rank correlation | Sum of “Akaike weights” (all models) (ratio)* | AWsum (%) | Hierarchical partitioning (I + J)* | Hierarchical partitioning (I %) | Gini importance (randomForest) | Gini importance (%) | Conditional Permutation (cforest) (CP) | CP (%) |
|---|---|---|---|---|---|---|---|---|---|
| Autocovariate | 0.80* | 1 | 11.2 | −1.76E−01 | 19.8 | 5.17E−02 | 31.3 | 2.72E−04 | 54.5 |
| Stream density | 0.72* | 1 | 11.2 | −1.24E−01 | 17.5 | 3.05E−02 | 18.5 | 6.33E−05 | 12.7 |
| Wetland to stream distance | −0.64* | 1 | 11.2 | −8.76E−02 | 13.6 | 1.89E−02 | 11.5 | 5.96E−05 | 12.0 |
| Lake abundance | −0.31* | 1 | 11.2 | −1.65E−02 | 5.5 | 2.64E−03 | 1.6 | 2.05E−07 | 0.0 |
| Lake areal abundance | −0.25 | 0.39 | 4.4 | −8.56E−03 | 4.5 | 2.92E−03 | 1.8 | 1.06E−06 | 0.2 |
| Max surface-water extent | −0.51* | 0.99 | 11.1 | −5.44E−02 | 8.3 | 6.85E−03 | 4.1 | 5.49E−06 | 1.1 |
| Change in surface-water extent | −0.45* | Excluded | Excluded | Excluded | Excluded | 4.64E−03 | 2.8 | 1.76E−06 | 0.4 |
| Wetland to wetland distance | 0.21 | 1 | 11.2 | −2.80E−03 | 7.3 | 6.49E−03 | 3.9 | 2.33E−05 | 4.7 |
| Total wetland density | −0.10 | 0.63 | 7.1 | −3.35E−03 | 5.4 | 8.27E−03 | 5.0 | 7.28E−06 | 1.5 |
| Wetland areal abundance | −0.32* | Excluded | Excluded | Excluded | Excluded | 3.37E−03 | 2.0 | 1.73E−06 | 0.3 |
| Percent drained | −0.26 | 1 | 11.2 | −2.29E−02 | 6.0 | 6.60E−03 | 4.0 | 2.69E−05 | 5.4 |
| Elevation coeffic. of var. | 0.47* | 0.64 | 7.2 | −2.96E−02 | 5.9 | 8.61E−03 | 5.2 | 1.31E−05 | 2.6 |
| Melton ruggedness number | 0.50* | 0.28 | 3.1 | −3.52E−02 | 6.2 | 1.29E−02 | 7.8 | 2.27E−05 | 4.6 |
| VC wetlands | Spearman rank correlation | Sum of “Akaike weights” (all models) (AWsum)* | AWsum (%) | Hierarchical partitioning (I + J)* | Hierarchical partitioning (I %) | Gini importance (randomForest) | Gini importance (%) | Conditional Permutation (cforest) (CP) | CP (%) |
|---|---|---|---|---|---|---|---|---|---|
| Autocovariate | 0.73* | 1 | 14.4 | −2.51E−01 | 22.3 | 1.72E−01 | 32.2 | 1.4E−03 | 77.8 |
| Stream density | −0.04 | 0.98 | 14.1 | −1.31E−03 | 7.7 | 2.39E−02 | 4.5 | 3.9E−05 | 2.2 |
| Wetland to stream distance | 0.16 | 0.72 | 10.4 | −1.94E−02 | 7.0 | 2.33E−02 | 4.3 | 2.2E−05 | 1.3 |
| Lake abundance | 0.25 | 0.26 | 3.7 | −1.91E−02 | 7.1 | 1.31E−02 | 2.4 | −6.3E−07 | 0.0 |
| Lake areal abundance | 0.08 | 0.32 | 4.6 | −4.02E−05 | 6.4 | 1.30E−02 | 2.4 | 1.4E−05 | 0.8 |
| Max surface-water extent | 0.22 | 1 | 14.4 | −3.48E−02 | 8.5 | 1.78E−02 | 3.3 | 1.9E−05 | 1.1 |
| Change in surface-water extent | 0.29* | Excluded | Excluded | Excluded | Excluded | 2.73E−02 | 5.1 | 3.7E−05 | 2.1 |
| Wetland to wetland distance | −0.40* | 0.99 | 14.3 | −1.14E−01 | 12.8 | 6.21E−02 | 11.6 | 1.3E−04 | 7.1 |
| Total wetland density | 0.42* | 0.58 | 8.4 | −8.09E−02 | 11.0 | 8.52E−02 | 15.9 | 9.2E−05 | 5.2 |
| Wetland areal abundance | 0.30* | Excluded | Excluded | Excluded | Excluded | 1.64E−02 | 3.1 | 6.7E−06 | 0.4 |
| Percent drained | −0.18 | 0.31 | 4.5 | −7.68E−02 | 4.1 | 5.43E−02 | 10.1 | 4.8E−05 | 2.7 |
| Elevation coeffic. of var. | 0.11 | 0.32 | 4.6 | −3.94E−04 | 6.4 | 1.49E−02 | 2.8 | −2.5E−06 | −0.1 |
| Melton ruggedness number | −0.01 | 0.46 | 6.6 | −5.00E−04 | 6.4 | 1.22E−02 | 2.3 | −3.0E−06 | −0.2 |
| NCO Wetlands | Spearman rank correlation | Sum of “Akaike weights” (all models) (AWsum)* | AWsum (%) | Hierarchical partitioning (I + J)* | Hierarchical partitioning (I %) | Gini importance (randomForest) | Gini importance (%) | Conditional Permutation (cforest) (CP) | CP (%) |
|---|---|---|---|---|---|---|---|---|---|
| Autocovariate | 0.85* | 1 | 16 | −7.77E−01 | 45.4 | 8.04E−01 | 34 | 6.27E−03 | 81.2 |
| Stream density | −0.57* | 0.99 | 16 | −2.31E−01 | 10.8 | 1.50E−01 | 6 | 2.29E−04 | 3.0 |
| Wetland to stream distance | 0.69* | 0.26 | 4 | −2.94E−01 | 9.9 | 3.73E−01 | 16 | 1.85E−04 | 2.4 |
| Lake abundance | 0.47* | 0.46 | 7 | −1.36E−01 | 4.2 | 1.16E−01 | 5 | 1.38E−04 | 1.8 |
| Lake areal abundance | 0.00 | 0.95 | 15 | −7.79E−03 | 1.4 | 6.93E−02 | 3 | 7.05E−05 | 0.9 |
| Max surface-water extent | 0.35* | 0.26 | 4 | −9.71E−02 | 2.7 | 6.12E−02 | 3 | 2.64E−05 | 0.3 |
| Change in surface-water extent | 0.55* | Excluded | Excluded | Excluded | Excluded | 1.20E−01 | 5 | 2.11E−05 | 0.3 |
| Wetland to wetland distance | −0.86* | Excluded | Excluded | Excluded | Excluded | Excluded | Excluded | Excluded | Excluded |
| Total wetland density | 0.90* | Excluded | Excluded | Excluded | Excluded | Excluded | Excluded | Excluded | Excluded |
| Wetland areal abundance | 0.51* | Excluded | Excluded | Excluded | Excluded | 2.24E−01 | 10 | 1.30E−04 | 1.7 |
| Percent drained | −0.55* | 1 | 16 | −3.75E−01 | 21.6 | 3.14E−01 | 13 | 5.87E−04 | 7.6 |
| Elevation coeffic. of var. | 0.04 | 0.98 | 15 | −6.32E−04 | 2.3 | 5.88E−02 | 3 | 2.43E−05 | 0.3 |
| Melton ruggedness number | 0.04 | 0.45 | 7 | −2.34E−03 | 1.6 | 6.05E−02 | 3 | 4.52E−05 | 0.6 |
Individual Spearman rank correlations are also shown with the Bonferroni correction applied. Significant correlations are starred. “Important” variables within each approach are identified using natural breaks in the normalized results (italic font). Variables identified as “important” in most approaches are in bold font. Variables identified as “important” in some of the approaches are in italic font. This notation is meant only to highlight results and does not represent an objective analysis of the results
*n = 1024, 1024 and 256 models for SI, VC and NCO, respectively, for Akaike weights and hierarchical partitioning approaches. The summed independent (I) and conjoined (J) contribution of each variable is shown for the hierarchical partitioning approach