When Big Data are Too Much: Effects of LiDAR Returns and Point Density on Estimation of Forest Biomass

Abstract

Analysis of light detection and ranging (LiDAR) data is becoming a mainstream approach to mapping forest biomass and carbon stocks across heterogeneous landscapes. However, large volumes of multireturn high point-density LiDAR data continue to pose challenges for large-area assessments. We are beginning to learn when and where point density can be reduced (or aggregated), but little is known regarding the degree to which multireturn data—at varying levels of point density—improve estimates of forest biomass. In this study, we examined the combined effects of LiDAR returns and data reduction on field-measured estimates of aboveground forest biomass in deciduous and mixed evergreen forests in an urbanized region of North Carolina, USA. We extracted structural metrics using first returns only, all returns, and rarely used laser pulse first returns from reduced point densities of LiDAR data. We statistically analyzed relationships between the field-measured biomass and LiDAR-derived variables for each return type and point-density combination. Overall, models using first return data performed only slightly better than models that utilized multiple returns. First-return models and multiple-return models at one percent point density resulted in 14% and 11% decrease in the amount of explained variation, respectively, compared to models with 100% point density. In addition, variance of modeled biomass across all point densities and return models was statistically similar to the field-measured biomass. Taken together, these results suggest that LiDAR first returns at reduced point density provide sufficient data for mapping urban forest biomass and may be an effective alternative to multireturn data.