Carbon Conversion - Take 4: Forest Only

Modeling forest carbon pools as a function of aboveground biomass and static climate and topographic predictors.

Changes

This round differs from the last iteration in the following ways:

• Plots with any non-forest conditions were removed.

• Elastic-Net regression introduced.

• PC Regression was conducted with LOO-CV (as in the latst iteration) and with a traditional (single) test-set.

AGC Model

• \(n\) train: 1851
• \(n\) test: 473
• Predictors: agb, mag, nbr, delta_nbr, ndvi, nsds, delta_nsds, tcb, delta_tcb, tcg, delta_tcg, tcw, delta_tcw, yod, dem, slope, aspect, twi, tmin, tmax, precip, chm, lcpri4, lcpri6, lcpri2, lcpri3, lcsec3, lcsec2, lcsec1, lcsec4

Random Forest Model - Test Set 1:1

Params

$formula
agc ~ .

$num.trees
[1] 1750

$mtry
[1] 13

$min.node.size
[1] 10

$replace
[1] TRUE

$sample.fraction
[1] 0.85

PC Regression - LOO-CV 1:1

Params

• Number of principle components: 23

PC Regression - Test Set 1:1

Params

• Number of principle components: 23

Elastic Net Regression - Test set 1:1

Params

• alpha: 1
• lambda: 0.0770014

Test Set Accuracy

	RF	PC - LOO-CV	PC	Elastic Net
RMSE	25.955	26.515	27.023	26.601
% RMSE	38.571	39.902	40.158	39.531
MAE	20.695	20.994	21.532	21.141
% MAE	30.754	31.593	31.998	31.417
MBE	0.153	-0.020	-0.982	-1.246
R2	0.370	0.346	0.317	0.338
% Improvement	20.705	19.129	17.443	18.732

---

BGC Model

• \(n\) train: 1837
• \(n\) test: 470
• Predictors: agb, mag, nbr, delta_nbr, ndvi, nsds, delta_nsds, tcb, delta_tcb, tcg, delta_tcg, tcw, delta_tcw, yod, dem, slope, aspect, twi, tmin, tmax, precip, chm, lcpri4, lcpri6, lcpri2, lcpri3, lcsec3, lcsec2, lcsec1, lcsec4

Random Forest Model - Test Set 1:1

Params

$formula
bgc ~ .

$num.trees
[1] 2000

$mtry
[1] 14

$min.node.size
[1] 1

$replace
[1] TRUE

$sample.fraction
[1] 0.8

PC Regression - LOO-CV 1:1

Params

• Number of principle components: 23

PC Regression - Test Set 1:1

Params

• Number of principle components: 23

Elastic Net Regression - Test set 1:1

Params

• alpha: 0.15
• lambda: 0.0622222

Test Set Accuracy

	RF	PC - LOO-CV	PC	Elastic Net
RMSE	5.093	5.209	5.299	5.237
% RMSE	37.109	38.341	38.607	38.157
MAE	4.076	4.134	4.242	4.182
% MAE	29.701	30.429	30.908	30.470
MBE	0.045	-0.003	-0.179	-0.207
R2	0.347	0.326	0.293	0.310
% Improvement	19.292	17.949	16.032	17.012

---

Deadwood C Model

• 5 filtered
  • Outlier threshold = [(5 * IQR) + 0.75th percentile] = 54.941
• \(n\) train: 374
• \(n\) test: 97
• Predictors: agb, mag, nbr, delta_nbr, ndvi, nsds, delta_nsds, tcb, delta_tcb, tcg, delta_tcg, tcw, delta_tcw, yod, dem, slope, aspect, twi, tmin, tmax, precip, chm, lcpri4, lcpri6, lcsec2, lcsec3, lcsec1, lcsec4

Random Forest Model - Test Set 1:1

Params

$formula
deadwood ~ .

$num.trees
[1] 250

$mtry
[1] 3

$min.node.size
[1] 7

$replace
[1] FALSE

$sample.fraction
[1] 1

PC Regression - LOO-CV 1:1

Params

• Number of principle components: 22

PC Regression - Test Set 1:1

Params

• Number of principle components: 22

Elastic Net Regression - Test set 1:1

Params

• alpha: 1
• lambda: 0.1796461

Test Set Accuracy

	RF	PC - LOO-CV	PC	Elastic Net
RMSE	5.889	6.570	6.442	6.278
% RMSE	77.954	80.943	85.273	83.105
MAE	4.494	4.795	4.888	4.687
% MAE	59.491	59.082	64.708	62.038
MBE	0.905	0.008	0.858	0.019
R2	0.176	0.114	0.014	0.063
% Improvement	9.694	5.946	1.215	3.727

---

Litter C models

• \(n\) train: 92
• \(n\) test: 28
• Predictors: agb, mag, nbr, delta_nbr, ndvi, nsds, delta_nsds, tcb, delta_tcb, tcg, delta_tcg, tcw, delta_tcw, yod, dem, slope, aspect, twi, tmin, tmax, precip, chm, lcpri4, lcsec3, lcsec1

PC Regression - LOO-CV 1:1

Params

• Number of principle components: 19

PC Regression - Test Set 1:1

Params

• Number of principle components: 19

Elastic Net Regression - Test Set 1:1

Params

• alpha: 0.1
• lambda: 4.9824199

Test-set Accuracy

	PC - LOO-CV	PC	Elastic Net
RMSE	11.302	15.700	8.341
% RMSE	80.829	92.551	49.174
MAE	7.915	10.213	6.608
% MAE	56.603	60.206	38.953
MBE	-0.570	-6.652	0.000
R2	-0.125	-0.639	0.537
% Improvement	-5.620	-25.731	33.197

---

Soil C model

• Predictors: agb, mag, nbr, delta_nbr, ndvi, nsds, delta_nsds, tcb, delta_tcb, tcg, delta_tcg, tcw, delta_tcw, yod, dem, slope, aspect, twi, tmin, tmax, precip, chm, lcpri4, lcsec3
• \(n\) train: 55
• \(n\) test: 17

PC Regression - LOO-CV 1:1

Params

• Number of principle components: 18

PC Regression - Test Set 1:1

Params

• Number of principle components: 18

Elastic Net Regression - Test Set 1:1

Params

• alpha: 1
• lambda: 6.810426

Test-set Accuracy

	PC - LOO-CV	PC	Elastic Net
RMSE	68.492	47.615	41.264
% RMSE	64.175	42.454	36.792
MAE	40.899	34.778	29.609
% MAE	38.321	31.009	26.400
MBE	0.092	0.032	0.000
R2	-0.053	0.360	0.519
% Improvement	-1.874	22.363	32.718