Using large footprint lidar to predict forest canopy height and aboveground biomass in high biomass tropical forests : A challenging task

Abstract

In order to assess the impact of deforestation on climate change, reliable estimates of aboveground biomass are needed. Estimates based on field measurements can be extended over broader spatial scales using remote sensing techniques. Although LiDAR (Light Detection And Ranging) shows no saturation at the biomass levels that represent the limits for optical and radar systems, it is not clear how it behaves at extremely high biomass densities (500 Mg ha-1 and above). Our study site in Corcovado National Park (Costa Rica) presents challenges for LiDAR use because of very high biomass conditions and the small size of the plots (0.07 ha). Because of the low co-registration (spatial overlap) between field plots and LiDAR footprints, LiDAR metrics could not significantly predict canopy heights. Biomass on the other hand was significantly predicted but with low accuracy (RMSE above 50%). We suggest that a plot size of at least 0.2 ha is needed to limit the biomass variability between plots, which may otherwise cause considerable model errors. Additionally, field maximum tree height (Hmax) proved a good predictor of plot level biomass in plots of small size, while dominant tree height (Hdom) and mean tree height (Hmean) seemed to outperform Hmax as plot size increased. We used a model based on Hmax to predict biomass at footprint level and obtained mean biomass densities at swath level of 281.5 Mg ha-1 for Corcovado and 194.8 Mg ha-1 for our other field site, the La Selva Biological Station in Costa Rica. These values are comparable to other results found in the Neotropics.