Remote sensing in support of high-resolution terrestrial carbon monitoring and modeling

Friday, 19 December 2014: 2:25 PM
George C Hurtt1, Maosheng Zhao1, Ralph Dubayah1, Chengquan Huang1, Anu Swatantran1, Jarlath ONeil-Dunne2, Kristofer D Johnson3, Richard Birdsey3, Justin Fisk1, Steve Flanagan1, Ritvik Sahajpal1, Wenli Huang1, Hao Tang1 and Amanda Hildt Armstrong4, (1)University of Maryland, College Park, MD, United States, (2)University of Vermont, Burlington, VT, United States, (3)USDA Forest Service Northern Research Statiuon, Newtown Square, PA, United States, (4)NASA Goddard Space Flight Center, Greenbelt, MD, United States
As part of its Phase 1 Carbon Monitoring System (CMS) activities, NASA initiated a Local-Scale Biomass Pilot study. The goals of the pilot study were to develop protocols for fusing high-resolution remotely sensed observations with field data, provide accurate validation test areas for the continental-scale biomass product, and demonstrate efficacy for prognostic terrestrial ecosystem modeling. In Phase 2, this effort was expanded to the state scale. Here, we present results of this activity focusing on the use of remote sensing in high-resolution ecosystem modeling. The Ecosystem Demography (ED) model was implemented at 90 m spatial resolution for the entire state of Maryland. We rasterized soil depth and soil texture data from SSURGO. For hourly meteorological data, we spatially interpolated 32-km 3-hourly NARR into 1-km hourly and further corrected them at monthly level using PRISM data. NLCD data were used to mask sand, seashore, and wetland. High-resolution 1 m forest/non-forest mapping was used to define forest fraction of 90 m cells. Three alternative strategies were evaluated for initialization of forest structure using high-resolution lidar, and the model was used to calculate statewide estimates of forest biomass, carbon sequestration potential, time to reach sequestration potential, and sensitivity to future forest growth and disturbance rates, all at 90 m resolution. To our knowledge, no dynamic ecosystem model has been run at such high spatial resolution over such large areas utilizing remote sensing and validated as extensively. There are over 3 million 90 m land cells in Maryland, greater than 43 times the ~73,000 half-degree cells in a state-of-the-art global land model.