Mapping tropical forest carbon on the Osa Peninsula, Costa Rica

Background/Question/Methods

Old growth tropical rainforests play a large role in the global carbon (C) cycle by storing about 350 Pg C in aboveground tree biomass, a reservoir that is shrinking due to widespread forest clearing and degradation.  A shifting climate may also cause substantial changes in intact forest C stocks.  And yet, our understanding of landscape level variations in and controls over aboveground carbon density (ACD) remains far from optimal. Airborne LiDAR (Light Detection and Ranging) has emerged as a powerful tool to measure ACD, and offers a platform to rapidly assess ACD across critical environmental gradients that can give insight into the controls over terrestrial C balance. In January 2012, the Carnegie Airborne Observatory collected LiDAR data over the Osa Peninsula region of SW Costa Rica. In combination with data from twelve 0.5 hectare plots that represent a successional chronosequence, we evaluated the accuracy of LiDAR top-of-canopy height (TCH) measurements for predicting field measurements of aboveground carbon stocks. We compared a traditional model (i.e. TCH-to-ACD alone) to both an ecoregional and universal model that incorporate plot-aggregate allometries, including basal area and wood density, for the forests of Osa Peninsula and the global tropics, respectively.

Results/Conclusions

Aboveground C stocks in the Osa region display notably high ACD values. The traditional model explained 81% of the regional variation in aboveground carbon stocks derived from detailed inventory measurements. The ecoregional model, which was parameterized using a stocking coefficient (i.e. the ratio of basal area to TCH) derived from local plot inventory measurements, performed the best, explaining 86% of the variation in LiDAR-ACD with a root mean square error (RMSE) of 21.1 Mg C/ha. The universal model, which integrates plot allometries from 13 ecoregions across the tropics, also performed well, yielding a regression coefficient of 0.82, RMSE of 23.88 Mg C/ha and relative error of 17%. Much of this uncertainty reflects the plot-size used for calibration, as plots below 1 hectare are sensitive to edge effects. Nevertheless, these results strongly suggest that the universal equation for mapping forest carbon can be effectively applied to high biomass rainforests, such as those found on the Osa Peninsula. We will conclude with an examination of controls over biomass in the region, then upscale LiDAR-based ACD regionally using a hybrid stratification – regression approach, with uncertainties estimated in each step of the process.