Estimation of solar direct beam transmittance of conifer canopies from airborne LiDAR

- Two LiDAR-derived models of solar direct beam canopy transmittance were evaluated.
- A Beer-Lambert bulk model and a raytrace model were compared to in-situ estimates.
- The raytrace model resolved complex seasonal and diurnal variability.
- Beer-Lambert errors highest in late spring and in areas of variable canopy cover.
- LiDAR-derived spatially explicit characterization of shortwave forcing is shown.

The utility of airborne scanning LiDAR data to estimate solar direct beam canopy transmittance in complex, forested terrain was evaluated. Twenty-four hemispherical photos were used to produce ground-based estimates of solar direct beam canopy transmittance. The photo estimates were used to develop and evaluate two spatially distributed canopy transmittance models: 1) a Beer's Law-type transmittance model based on LiDAR-derived canopy metrics, and 2) a solar raytrace model applied to a three-dimensional canopy model. The models were used to estimate solar direct beam canopy transmittance at five-minute resolution for all days between the winter and summer solstices over an 800 m by 700 m domain at 1 m horizontal grid spacing. When compared to estimates from hemispherical photos, the raytrace model resolved the complex seasonal and diurnal variability of solar direct beam canopy transmittance resulting from individual trees and localized canopy structure. The Beer's-type model was unable to resolve these detailed factors. The two models exhibited similar and relatively low normalized daily mean error values from December to early March. Later in the season (01 March-21 June), the model differences were pronounced; the daily mean and standard deviation of the error values for the Beer's-type and raytrace models were 13% ± 10% and 8% ± 6%, respectively. The results confirm previously known limitations of Beer's Law when used to estimate sub-canopy solar beam irradiance under heterogeneous canopy conditions. Averaged over the spatial domain, the Beer's-type model estimated 21% and 48% lower canopy transmittance than the raytrace model on 01 March and 03 May, respectively. The Beer's-type model was unable to represent the seasonal increase in areal average canopy transmission contributed from small canopy gaps. Finally, both distributed models were used to simulate the cumulative solar beam irradiance during the 2010 snowmelt season. The raytrace model was shown to capture a high level of variability necessary to simulate explicit stand-scale solar irradiance that strongly influences spatiotemporal patterns of snowmelt, soil water availability, and the partition and exchange of energy within heterogeneous forest ecosystems.