Statistically rigorous estimation of forest carbon stocks in northwestern US combining stakeholder forest inventory and remote sensing data, USFS FIA plots, Landsat and GEDI lidar

Abstract:

It is critically important to be monitoring carbon storage in forests and rangelands in this time of rapid global change. It is likely to fall on Earth systems scientists to objectively monitor the effects of land use decisions on carbon sequestration. The work we present here represents a phase 2 continuation of a previous CMS project focused on pulling together field and airborne lidar remote sensing datasets collected by various stakeholders to generate annual maps of forest aboveground carbon from 2000 to 2016 for the northwestern US. It is incredibly valuable to bring together as much field and high-quality remote sensing as possible to inform regional-scale forest carbon maps. However, combining datasets that were collected using different field protocols and different sensor systems to inform one unified model can lead to prediction biases that are not easy to correct. As a part of our phase 2 continuing development of a useful forest carbon MRV system for the northwestern US, we aim to address this issue. Here, we demonstrate the use of a design-unbiased model-assisted estimation approach that combines probabilistically sampled USFS Forest Inventory and Analysis (FIA) plots, with stakeholder data-driven forest carbon maps, Landsat variables and lidar waveform data collected by GEDI to estimate forest aboveground carbon at the state and county levels for the years 2000 to 2016. Results show that leveraging stakeholder forest carbon maps and space-borne remote sensing leads to a marked improvement over forest carbon estimates based on FIA data alone. By reducing uncertainty in forest carbon estimates decision makers can make better informed decisions when it comes to managing forests for climate mitigation.