Landsat Ensemble AGB 1.1.1: Data Ensemble

Change Summary

Notes

• FIA plot data as well as LiDAR-AGB pixel predictions were used as training data in this set of models.
  • Training data from version of FIA models 1.1.1 More Data x LiDAR Zeroes
  • Using training data from Landsat LiDAR models 1.1.0 More Data
• The FIA training data was randomly upsampled to achieve 60% share of the pooled training data set.
• The pooled training data set consisted of 10,000 LiDAR-AGB predictions (40%) and 15,000 FIA AGB samples (60%)
• This test set was not upsampled, and maps directly to the Landsat-FIA model test set

FIA Test Plot Results

	RF (ranger)	GBM (LightGBM)	Ensemble (model weighted)	Ensemble (RMSE weighted)
%RMSE	41.634	42.051	40.969	40.962
RMSE	52.615	53.142	51.776	51.767
MAE	40.199	39.907	39.246	39.336
MBE	5.144	4.686	4.958	4.918
R2	0.516	0.509	0.530	0.530

Training Data Distribution

summary(training$agb_mgha)

   Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
    0.0    73.2   130.9   131.7   187.4   425.0 

Testing Data Distribution

summary(testing$agb_mgha)

   Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
   0.00   78.28  128.09  126.38  174.86  422.62 

Metadata

Ensembles

• RMSE-weighted model weights:

• Linear model weights:

Call:
lm(formula = agb_mgha ~ rf_pred * lgb_pred, data = k_fold_preds)

Residuals:
     Min       1Q   Median       3Q      Max 
-160.797   -2.549    0.273    2.113  170.171 

Coefficients:
                   Estimate Std. Error t value Pr(>|t|)    
(Intercept)      -2.246e+00  4.206e-01  -5.341 9.34e-08 ***
rf_pred           5.933e-01  1.009e-02  58.802  < 2e-16 ***
lgb_pred          4.134e-01  1.005e-02  41.124  < 2e-16 ***
rf_pred:lgb_pred  6.863e-05  2.295e-05   2.991  0.00279 ** 
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Residual standard error: 23.18 on 24996 degrees of freedom
Multiple R-squared:  0.9109,    Adjusted R-squared:  0.9109 
F-statistic: 8.522e+04 on 3 and 24996 DF,  p-value: < 2.2e-16

\(n\) and \(p\)

• 25662 observations
  • 25000 training
  • 662 testing
• 166 predictors
  • aweinsh, delta_aweinsh, aweish, delta_aweish, bcc, delta_bcc, dvi, delta_dvi, embi, delta_embi, exgr, delta_exgr, exr, delta_exr, fcvi, delta_fcvi, gari, delta_gari, gbndvi, delta_gbndvi, gemi, delta_gemi, gli, delta_gli, gndvi, delta_gndvi, gosavi, delta_gosavi, grndvi, delta_grndvi, gvmi, delta_gvmi, ikaw, delta_ikaw, ipvi, delta_ipvi, lswi, delta_lswi, mag, mbi, delta_mbi, mcari2, delta_mcari2, mlswi26, delta_mlswi26, mlswi27, delta_mlswi27, mndwi, delta_mndwi, mrbvi, delta_mrbvi, msavi, delta_msavi, msi, delta_msi, msr, delta_msr, mtvi2, delta_mtvi2, muwir, delta_muwir, nbr, delta_nbr, nbr2, delta_nbr2, nbsims, delta_nbsims, nddi, delta_nddi, ndsi, delta_ndsi, ndvi, delta_ndvi, ndvimndwi, delta_ndvimndwi, ndwi, delta_ndwi, ngrdi, delta_ngrdi, nirv, delta_nirv, nli, delta_nli, nmdi, delta_nmdi, normg, delta_normg, normnir, delta_normnir, normr, delta_normr, nrfig, delta_nrfig, nsds, delta_nsds, nwi, delta_nwi, pisi, delta_pisi, rcc, delta_rcc, rdvi, delta_rdvi, rgbvi, delta_rgbvi, rgri, delta_rgri, s3, delta_s3, si, delta_si, sr, delta_sr, sr2, delta_sr2, swi, delta_swi, tcb, delta_tcb, tcg, delta_tcg, tcw, delta_tcw, tdvi, delta_tdvi, tgi, delta_tgi, trivi, delta_trivi, tvi, delta_tvi, ui, delta_ui, vgnirbi, delta_vgnirbi, vig, delta_vig, vrnirbi, delta_vrnirbi, wi1, delta_wi1, wi2, delta_wi2, yod, precip, tmin, tmax, dem, twi, slope, aspect, chm, lcpri4, lcpri6, lcpri3, lcpri1, lcpri2, lcpri5, lcpri8, lcsec3, lcsec2, lcsec1, lcsec4, lcsec5, lcsec6, lcsec8

Component Models

• Tuning used 5-fold CV
• Final hyperparameters:

Random forest:

$num.trees
[1] 1000

$mtry
[1] 58

$min.node.size
[1] 2

$sample.fraction
[1] 1

$replace
[1] FALSE

$formula
agb_mgha ~ .

LGB:

$params
$params$num_leaves
[1] 44

$params$max_depth
[1] 15

$params$extra_trees
[1] FALSE

$params$min_data_in_leaf
[1] 10

$params$bagging_fraction
[1] 0.9

$params$bagging_freq
[1] 1

$params$feature_fraction
[1] 0.9

$params$min_data_in_bin
[1] 18

$params$lambda_l1
[1] 1

$params$lambda_l2
[1] 9

$params$learning_rate
[1] 0.3

$params$force_col_wise
[1] TRUE

$params$nrounds
[1] 500