Models trained using BIOWIDE stratification

For these models the training data was split according to the BIOWIDE stratification.

Variable importance

Variable importance for this boosted regression tree model.

##                                                             var  rel.inf
## dtm_10m                                                 dtm_10m 9.436428
## Sand_utm32_10m                                   Sand_utm32_10m 9.410692
## treetype_bjer_dec                             treetype_bjer_dec 7.586392
## Clay_utm32_10m                                   Clay_utm32_10m 6.717277
## Soc_utm32_10m                                     Soc_utm32_10m 6.288340
## ns_groundwater_summer_utm32_10m ns_groundwater_summer_utm32_10m 5.873385
## dtm_10m_sd_110m                                 dtm_10m_sd_110m 5.429778
## amplitude_sd                                       amplitude_sd 5.258751
## canopy_height                                     canopy_height 5.022070
## vegetation_density                           vegetation_density 4.780821
## canopy_height_sd_110m                     canopy_height_sd_110m 4.767433
## ns_groundwater_summer_sd_110m     ns_groundwater_summer_sd_110m 4.674164
## solar_radiation                                 solar_radiation 4.625719
## vegetation_density_sd_110m           vegetation_density_sd_110m 4.329235
## normalized_z_sd                                 normalized_z_sd 4.155538
## foliage_height_diversity               foliage_height_diversity 3.908227
## amplitude_mean                                   amplitude_mean 3.760337
## slope                                                     slope 2.515989
## openness_difference                         openness_difference 1.459423

Performance in BIOWIDE regions:

Performance map based on the independent validation data:

Performance table based on the independent validation data:

Measure Overall Bornholm Fune_Lolland Nordjlland Oestjylland Sjaelland Vestjylland
Accuracy 0.81 0.81 0.78 0.90 0.80 0.74 0.89
Error 0.19 0.19 0.22 0.10 0.20 0.26 0.11
Sensitivity (True Positive Rate) 0.83 0.88 0.85 0.94 0.84 0.78 0.71
Specificity (True Negative Rate) 0.78 0.70 0.63 0.86 0.76 0.70 0.96
Fall-out (False Positive Rate) 0.22 0.30 0.37 0.14 0.24 0.30 0.04
Positive predictive value (User Accuracy) 0.80 0.84 0.83 0.88 0.80 0.73 0.87

Performance table based on the dependent training data:

Measure Overall Bornholm Fune_Lolland Nordjlland Oestjylland Sjaelland Vestjylland
Accuracy 0.99 1.00 0.99 1.00 1.00 0.99 0.99
Error 0.01 0.00 0.01 0.00 0.00 0.01 0.01
Sensitivity (True Positive Rate) 0.99 1.00 1.00 0.99 0.99 0.99 0.97
Specificity (True Negative Rate) 0.99 0.99 0.99 1.00 1.00 0.99 1.00
Fall-out (False Positive Rate) 0.01 0.01 0.01 0.00 0.00 0.01 0.00
Positive predictive value (User Accuracy) 0.99 1.00 1.00 1.00 1.00 0.99 1.00

Perfromance in Derek’s regions:

Performance map based on the independent validation data:

Performance table based on the independent validation data:

Measure Overall Region 1 Region 2 Region 3
Accuracy 0.81 0.78 0.79 0.88
Error 0.19 0.22 0.21 0.12
Sensitivity (True Positive Rate) 0.83 0.84 0.85 0.76
Specificity (True Negative Rate) 0.78 0.69 0.73 0.92
Fall-out (False Positive Rate) 0.22 0.31 0.27 0.08
Positive predictive value (User Accuracy) 0.80 0.79 0.81 0.80

Performance table based on the dependent training data:

Measure Overall Region 1 Region 2 Region 3
Accuracy 0.99 0.99 0.99 0.99
Error 0.01 0.01 0.01 0.01
Sensitivity (True Positive Rate) 0.99 0.99 0.99 0.98
Specificity (True Negative Rate) 0.99 0.99 0.99 1.00
Fall-out (False Positive Rate) 0.01 0.01 0.01 0.00
Positive predictive value (User Accuracy) 0.99 0.99 1.00 1.00

Performance by forest type (boradleaf vs. coniferous)

Performance table based on the independent validation data:

Measure Overall Broadleaf Coniferous
Accuracy 0.81 0.78 0.88
Error 0.19 0.22 0.12
Sensitivity (True Positive Rate) 0.83 0.86 0.63
Specificity (True Negative Rate) 0.78 0.68 0.95
Fall-out (False Positive Rate) 0.22 0.32 0.05
Positive predictive value (User Accuracy) 0.80 0.80 0.80

Performance table based on the dependent training data:

Measure Overall Broadleaf Coniferous
Accuracy 0.99 0.99 0.99
Error 0.01 0.01 0.01
Sensitivity (True Positive Rate) 0.99 1.00 0.97
Specificity (True Negative Rate) 0.99 0.99 1.00
Fall-out (False Positive Rate) 0.01 0.01 0.00
Positive predictive value (User Accuracy) 0.99 0.99 0.99

Models trained using Derek’s stratification

Variable importance

Variable importance for this boosted regression tree model.

##                                                             var  rel.inf
## Sand_utm32_10m                                   Sand_utm32_10m 9.292837
## dtm_10m                                                 dtm_10m 9.063782
## treetype_bjer_dec                             treetype_bjer_dec 7.359120
## Clay_utm32_10m                                   Clay_utm32_10m 6.703277
## ns_groundwater_summer_utm32_10m ns_groundwater_summer_utm32_10m 6.048256
## Soc_utm32_10m                                     Soc_utm32_10m 6.036391
## amplitude_sd                                       amplitude_sd 5.627812
## dtm_10m_sd_110m                                 dtm_10m_sd_110m 5.429703
## canopy_height                                     canopy_height 5.251818
## canopy_height_sd_110m                     canopy_height_sd_110m 4.772935
## ns_groundwater_summer_sd_110m     ns_groundwater_summer_sd_110m 4.552738
## vegetation_density_sd_110m           vegetation_density_sd_110m 4.520161
## solar_radiation                                 solar_radiation 4.495919
## vegetation_density                           vegetation_density 4.397915
## normalized_z_sd                                 normalized_z_sd 4.318571
## foliage_height_diversity               foliage_height_diversity 4.116388
## amplitude_mean                                   amplitude_mean 3.919426
## slope                                                     slope 2.403913
## openness_difference                         openness_difference 1.689039

Performance in BIOWIDE regions:

Performance map based on the independent validation data:

Performance table based on the independent validation data:

Measure Overall Bornholm Fune_Lolland Nordjlland Oestjylland Sjaelland Vestjylland
Accuracy 0.81 0.82 0.81 0.89 0.80 0.76 0.89
Error 0.19 0.18 0.19 0.11 0.20 0.24 0.11
Sensitivity (True Positive Rate) 0.84 0.86 0.90 0.93 0.85 0.79 0.69
Specificity (True Negative Rate) 0.79 0.75 0.63 0.86 0.75 0.72 0.95
Fall-out (False Positive Rate) 0.21 0.25 0.37 0.14 0.25 0.28 0.05
Positive predictive value (User Accuracy) 0.80 0.86 0.83 0.88 0.78 0.75 0.84

Performance table based on the dependent training data:

Measure Overall Bornholm Fune_Lolland Nordjlland Oestjylland Sjaelland Vestjylland
Accuracy 1 1 1 1 1 0.99 0.99
Error 0 0 0 0 0 0.01 0.01
Sensitivity (True Positive Rate) 1 1 1 1 1 0.99 0.98
Specificity (True Negative Rate) 1 1 1 1 1 0.99 1.00
Fall-out (False Positive Rate) 0 0 0 0 0 0.01 0.00
Positive predictive value (User Accuracy) 1 1 1 1 1 0.99 1.00

Perfromance in Derek’s regions:

Performance map based on the independent validation data:

Performance table based on the independent validation data:

Measure Overall Region 1 Region 2 Region 3
Accuracy 0.81 0.79 0.80 0.87
Error 0.19 0.21 0.20 0.13
Sensitivity (True Positive Rate) 0.84 0.84 0.86 0.77
Specificity (True Negative Rate) 0.79 0.70 0.72 0.92
Fall-out (False Positive Rate) 0.21 0.30 0.28 0.08
Positive predictive value (User Accuracy) 0.80 0.81 0.80 0.78

Performance table based on the dependent training data:

Measure Overall Region 1 Region 2 Region 3
Accuracy 1 1.00 1 1.00
Error 0 0.00 0 0.00
Sensitivity (True Positive Rate) 1 1.00 1 0.99
Specificity (True Negative Rate) 1 0.99 1 1.00
Fall-out (False Positive Rate) 0 0.01 0 0.00
Positive predictive value (User Accuracy) 1 1.00 1 1.00

Performance by forest type (boradleaf vs. coniferous)

Performance table based on the independent validation data:

Measure Overall Broadleaf Coniferous
Accuracy 0.81 0.95 0.97
Error 0.19 0.05 0.03
Sensitivity (True Positive Rate) 0.84 0.97 0.90
Specificity (True Negative Rate) 0.79 0.93 0.99
Fall-out (False Positive Rate) 0.21 0.07 0.01
Positive predictive value (User Accuracy) 0.80 0.96 0.96

Performance table based on the dependent training data:

Measure Overall Broadleaf Coniferous
Accuracy 1 0.95 0.97
Error 0 0.05 0.03
Sensitivity (True Positive Rate) 1 0.97 0.90
Specificity (True Negative Rate) 1 0.93 0.99
Fall-out (False Positive Rate) 0 0.07 0.01
Positive predictive value (User Accuracy) 1 0.95 0.96