TY - JOUR
T1 - Estimation of forest resources from a country wide laser scanning survey and national forest inventory data
AU - Nord-Larsen, Thomas
AU - Schumacher, Johannes
PY - 2012/4/16
Y1 - 2012/4/16
N2 - The demand for renewable energy has raised a need for efficient mapping of forest fuel resources in Denmark. Airborne laser scanning may provide a means for assessing local forest biomass resources. In this study, national forest inventory (NFI) data was used as reference data for modeling forest basal area, volume, above-ground biomass, and total biomass from laser scanning data obtained in a countrywide scanning survey. Data covered a wide range of forest ecotypes, stand treatments, tree species, and tree species mixtures. The four forest characteristics were modeled using nonlinear regression and generalized method-of-moments estimation to avoid biased and inefficient estimates. The coefficient of determination was 68% for the basal area model and 77-78% for the volume and biomass models. Despite the wide range of forest types model accuracy was comparable to similar studies. Model predictions were unbiased across the range of predicted values and crown cover percentages but positively biased for deciduous forest and negatively biased for coniferous forest. Species type specific (coniferous, deciduous, or mixed forest) models reduced root mean squared error by 3-12% and removed the bias. In application, model predictions will be improved by stratification into deciduous and coniferous forest using e.g. infrared orthophotos or satellite images.
AB - The demand for renewable energy has raised a need for efficient mapping of forest fuel resources in Denmark. Airborne laser scanning may provide a means for assessing local forest biomass resources. In this study, national forest inventory (NFI) data was used as reference data for modeling forest basal area, volume, above-ground biomass, and total biomass from laser scanning data obtained in a countrywide scanning survey. Data covered a wide range of forest ecotypes, stand treatments, tree species, and tree species mixtures. The four forest characteristics were modeled using nonlinear regression and generalized method-of-moments estimation to avoid biased and inefficient estimates. The coefficient of determination was 68% for the basal area model and 77-78% for the volume and biomass models. Despite the wide range of forest types model accuracy was comparable to similar studies. Model predictions were unbiased across the range of predicted values and crown cover percentages but positively biased for deciduous forest and negatively biased for coniferous forest. Species type specific (coniferous, deciduous, or mixed forest) models reduced root mean squared error by 3-12% and removed the bias. In application, model predictions will be improved by stratification into deciduous and coniferous forest using e.g. infrared orthophotos or satellite images.
U2 - 10.1016/j.rse.2011.12.022
DO - 10.1016/j.rse.2011.12.022
M3 - Journal article
SN - 0034-4257
VL - 119
SP - 148
EP - 157
JO - Remote Sensing of Environment
JF - Remote Sensing of Environment
ER -
