This document provides summary stats (area) for the forest conservation value projections. We show the statistics for all four models tested in our analysis.

Content:

  1. Forest area in Denmark according to Bjerreskov et al. 2021.
  2. Training data area: High conservation value and low conservation value forests.
  3. Disturbance detected in forests overall.
  4. Gradient Boosting model summary stats (BIOWIDE).
  5. Random Forest model summary stats (BIOWIDE).
  6. Gradient Boosting model summary stats (Derek’s stratification).
  7. Random Forest model summarz stats (Derek’s stratification).

Forests area in Denmark according to Bjerreskov et al. 2021

Below you can find the total area of forest in the forest mask from Bjerreskov et al. 2021. This is the reference for the total area of forest used in our project. The mask is based on the tree type layer from the same publication (see predictor description). We generated the forest mask by refining the treetype layer into a forest mask by applying a minimum mapping filter removing all continuous forest patches smaller than 500 m2 (see also Bjerreskov et al. 2021).

Layer Area [km²]
forest mask 6345.3

Training data area: High conservation value and low conservation value forests

Here you can see the area covered by our training data. Including both the high conservation value forests with designations (p15, p25 and private old growth), as well as the low conservation value training polygons. The proportions are given relative to the total area of forest according to the forest mask generated from Bjerreskov et al. 2021 (see above).

category Area [km²] Proportion of all forest [%]
p25 111.71 2e-06
private_old_growth 31.98 1e-06
p15 17.79 0e+00
total_high_quality 161.13 3e-06
ikke_p25 46.70 1e-06
NST_plantations 59.33 1e-06
total_low_quality 106.03 2e-06

Disturbance overall

We used a disturbance layer generated by Cornelius (Senf and Seidl 2021) to estimate the disturbance in Denmark’s forests since the lidar data for EcoDes-DK15 was collected.

Please note that this disturbance mask was projected and down-sampled from a 30 m Landsat grid to the 10 m EcoDes-DK15 grid (nearest neighbour algorithm), potentially adding small uncertainties to the area estimates. Currently, we also only account for disturbances from 2016 till 2020.

Name Area [km²] Proportion [%]
disturbed forest 84.49 1.30
total forest 6345.30 100.00

Gradient Boosting projections summary stats (BIOWIDE)

This gradient boosting model was trained based on the “BIOWIDE” stratification.

Type Area [km²] Proportion [%]
high conservation value forest 1979.92 31.20
low conservation value forest 4307.60 67.90
total forest 6345.30 100.00

Disturbance statistics:

Type Area [km²] Proportion [%]
disturbed high conservation value forest 18.20 0.90
total high conservation value forest 1979.92 100.00
Type Area [km²] Proportion [%]
disturbed low conservation value forest 66.29 1.50
total low conservation value forest 4307.60 100.00
Type Area [km²] Proportion [%]
disturbed high conservation value forest 18.20 21.50
disturbed low conservation value forest 66.29 78.50
total disturbed forest 84.49 100.00

Random Forest projections summary stats (BIOWIDE)

This random forest model was trained based on the “BIOWIDE” stratification.

Type Area [km²] Proportion [%]
high conservation value forest 1999.61 31.50
low conservation value forest 4287.91 67.60
total forest 6345.30 100.00

Disturbance statistics:

Type Area [km²] Proportion [%]
disturbed high conservation value forest 17.81 0.90
total high conservation value forest 1999.61 100.00
Type Area [km²] Proportion [%]
disturbed low conservation value forest 66.69 1.60
total low conservation value forest 4287.91 100.00
Type Area [km²] Proportion [%]
disturbed high conservation value forest 17.81 21.10
disturbed low conservation value forest 66.69 78.90
total disturbed forest 84.49 100.00

Gradient Boosting projections summary stats (Derek’s stratification)

This gradient boosting model was trained based on the “Derek’s” stratification.

Type Area [km²] Proportion [%]
high conservation value forest 1986.18 31.30
low conservation value forest 4301.33 67.80
total forest 6345.30 100.00

Disturbance statistics:

Type Area [km²] Proportion [%]
disturbed high conservation value forest 18.19 0.90
total high conservation value forest 1986.18 100.00
Type Area [km²] Proportion [%]
disturbed low conservation value forest 66.30 1.50
total low conservation value forest 4301.33 100.00
Type Area [km²] Proportion [%]
disturbed high conservation value forest 18.19 21.50
disturbed low conservation value forest 66.30 78.50
total disturbed forest 84.49 100.00

Random Forest projections summary stats (Derek’s stratification)

This random forest model was trained based on the “Derek’s” stratification.

Type Area [km²] Proportion [%]
high conservation value forest 2007.38 31.60
low conservation value forest 4280.13 67.50
total forest 6345.30 100.00

Disturbance statistics:

Type Area [km²] Proportion [%]
disturbed high conservation value forest 17.97 0.90
total high conservation value forest 2007.38 100.00
Type Area [km²] Proportion [%]
disturbed low conservation value forest 66.52 1.60
total low conservation value forest 4280.13 100.00
Type Area [km²] Proportion [%]
disturbed high conservation value forest 17.97 21.30
disturbed low conservation value forest 66.52 78.70
total disturbed forest 84.49 100.00