Filtering ground noise from LiDAR returns produces inferior models of forest aboveground biomass in heterogenous landscapes

Abstract

Airborne LiDAR has become an essential data source for large-scale, high-resolution modeling of forest aboveground biomass and carbon stocks, enabling predictions with much higher resolution and accuracy than can be achieved using optical imagery alone. Ground noise filtering – that is, excluding returns from LiDAR point clouds based on simple height thresholds – is a common practice meant to improve the `signal’ content of LiDAR returns by preventing ground returns from masking useful information about tree size and condition contained within canopy returns. However, ground returns may be helpful for making accurate aboveground biomass predictions in heterogeneous landscapes that include a patchy mosaic of vegetation heights and land cover types. In this paper, we applied several ground noise filtering thresholds while mapping forest AGB across New York State (USA), a heterogenous landscape composed of both contiguously forested and highly fragmented areas with mixed land cover types. We fit random forest models to predictor sets derived from each filtering intensity threshold and compared model accuracies, paying attention to how changes in accuracy correlated with landscape structure. We observed that removing ground noise via any height threshold systematically biases many of the LiDAR-derived variables used in AGB modeling, with mean correlation (Spearman’s $$\rho $$) between variables increasing from 0.183 to 0.266. We found that that ground noise filtering yields models of forest AGB with lower accuracy than models trained using predictors derived from unfiltered point clouds, with RMSE increasing by up to 2.2 Mg ha-1 statewide. Although we only modeled AGB for forest cover types, models fit to predictors derived from filtered point clouds performed worse as landscape heterogeneity (as measured by patch density and edge density) increased, suggesting ground returns are particularly useful when modeling edge forests. Our results suggest that ground filtering should be a carefully considered decision when mapping forest AGB, particularly when mapping heterogeneous and highly fragmented landscapes, as ground returns are more likely to represent useful `signal’ than extraneous `noise’ in these cases.

Keywords

• aboveground biomass
• ground noise
• lidar
• machine learning
• random forest
• introduction