Extensive Sampling of Forest Carbon using High Density Power Line Lidar

Estimating carbon sequestration and greenhouse gas emissions from forest management, natural processes, and disturbance is of growing interest for mitigating global warming. Ponderosa pine is common at mid-elevations throughout the western United States and is a dominant tree species in southwestern forests. Existing unmanaged "relict" sites and stand reconstructions of southwestern ponderosa pine forests from before European settlement (late 1800s) provide evidence of forests of larger trees of lower density and less vulnerability to severe fires than today's typical conditions of high densities of small trees that have resulted from a century of fire suppression. Forest treatments to improve forest health in the region include tree cutting focused on small-diameter trees (thinning), low-intensity prescribed burning, and monitoring rather than suppressing wildfires. Stimulated by several uncharacteristically-intense fires in the last decade, a collaborative process found strong stakeholder agreement to accelerate forest treatments to reduce fire risk and restore ecological conditions. Land use planning to ramp up management is underway and could benefit from quick and inexpensive techniques to inventory tree-level carbon because existing inventory data are not adequate to capture the range of forest structural conditions. Our approach overcomes these shortcomings by employing recent breakthroughs in estimating aboveground biomass from high resolution light detection and ranging (lidar) remote sensing. Lidar is an active remote sensing technique, analogous to radar, which measures the time required for a transmitted pulse of laser light to return to the sensor after reflection from a target. Lidar data can capture 3-dimensional forest structure with greater detail and broader spatial coverage than is feasible with conventional field measurements. We developed a novel methodology for extensive sampling and field validation of forest carbon, applicable to managed and unmanaged areas, using high point density lidar collected over transmission line corridors. The lidar metric of quadratic mean height guided our selection of field plots spanning the full range from low to high levels of aboveground biomass across the study region. Before model selection, we minimized two of the major sources of errors in lidar calibration: variance in tree allometry across landscapes and plot edge effects (spatial mismatch between field measurements and lidar points). We tested an assortment of model selection techniques and goodness of fit measures for deriving forest structural metrics of interest. For example, we obtained an R-squared value for aboveground biomass (Mg/ha) of 0.9 using stepwise regression. The forest metrics obtained are being used in the next stage of the project to parameterize biogeochemical models linking terrestrial carbon pools and atmospheric greenhouse gas exchanges.