Digital Terrain Model for British Columbia at 25-m spatial resolution based on GEDI, ICESat-2, AW3D, CDED, GLAD layers, MERIT DEM and GLO-30

DOI10.5281/zenodo.4665926Zenodo4665926MaRDI QIDQ6689003FDOQ6689003

Dataset published at Zenodo repository.

Publication date: 6 April 2021

Copyright license: Creative Commons Attribution 4.0 International

We used GEDI and ICESat-2 space-based LiDAR survey data (cca 7M points) as ground control points to train a model to predict bare surface terrain height using as covariates: AW3D, GLO-30, MERIT DEM, CDED-25, GLAD canopy height, surface water occurrence probability, GLAD (Global Forest Watch) bare earth probability and tree cover fractions for 2000, 2010 and 2014. The training points were extracted using the lowest observed altitudes from the GEDI (Level 2A; elev_lowestmode) and ICESat-2 (ATL08; h_te_mean) surveys. Prior to regression modeling the WGS84 ellipsoid altitudes were converted toEGM2008 using the global difference grids. The Ensemble Models were then fitted using spatial 5-fold cross-validation with refitting, as implemented in the mlr package for Machine Learning, and random forest (ranger), cubist and glmnet as the base learners. The results achieved a reported RMSE of 8.6-m with an R-squared of 0.9998, with random forest and cubist identified as equally important learners. Data description: dtm_elev.lowestmode_gedi.cded.eml_mf_25m_0..0cm_2000..2018_bc.epsg3005_v0.1.2.tif: Ensemble estimate of the terrain height in decimeter (dm). dtm_elev.lowestmode_gedi.cded.eml_md_25m_0..0cm_2000..2018_bc.epsg3005_v0.1.2.tif: Error of the ensemble estimate of terrain heights in meter (m). Model summary: Variable: elev_lowestmode.f R-square: 1 Fitted values sd: 610 RMSE: 8.61 Ensemble model: Call: stats::lm(formula = f, data = d) Residuals: Min 1Q Median 3Q Max -59.725 -3.772 0.291 3.769 267.910 Coefficients: (1 not defined because of singularities) Estimate Std. Error t value Pr(|t|) (Intercept) -0.2358210 0.0189646 -12.44 2e-16 * regr.ranger 0.5504095 0.0013799 398.89 2e-16 * regr.glmnet 0.0074061 0.0005931 12.49 2e-16 * regr.cubist 0.4422925 0.0014321 308.83 2e-16 * --- Signif. codes: 0 '*' 0.001 0.01 '*' 0.05 '.' 0.1 ' ' 1 Residual standard error: 8.605 on 5385151 degrees of freedom Multiple R-squared: 0.9998, Adjusted R-squared: 0.9998 F-statistic: 9.02e+09 on 3 and 5385151 DF, p-value: 2.2e-16 The variable importance table shows: variable importance 3 bc_CDED_DEM_30m_ 653702254033 4 bc_MERITDEM_30m_ 488290057186 1 bc_AW3Dv2012_30m_ 434604342980 2 bc_GLO30_30m_ 363673956969 7 bc_bare2010_30m_ 43879671456 6 bc_treecover2010_30m_ 13043530952 8 bc_treecover2000_30m_ 5833643430 9 bc_canopy_height_30m_ 854769631

This page was built for dataset: Digital Terrain Model for British Columbia at 25-m spatial resolution based on GEDI, ICESat-2, AW3D, CDED, GLAD layers, MERIT DEM and GLO-30