Sample-based estimation of biennial forest disturbance rates for the conterminous United States
Robert
E.
Kennedy, Dept. of Forest Science / Oregon State University, robert.kennedy@oregonstate.edu
(Presenting)
Warren
B.
Cohen, PNW Research Station / USDA Forest Service, wcohen@fs.fed.us
Gretchen
G.
Moisen, RM Research Station / USDA Forest Service, gmoisen@fs.fed.us
Samuel
N.
Goward, Dept. of Geography / UMD, sgoward@umd.edu
Michael
Wulder, Pacific Forestry Center / Canadian Forest Service, mwulder@pfc.cfs.nrcan.gc.ca
Scott
L.
Powell, PNW Research Station / USDA Forest Service, spowell@fs.fed.us
Jeffrey
Masek, Biospheric Sciences Branch / GSFC, jeffrey.g.masek@nasa.gov
Chengquan
Huang, Dept. of Geography / UMD, cqhuang@umd.edu
Sean
Healey, RM Research Station / USDA Forest Service, seanhealey@fs.fed.us
Characterization of forest disturbance and regrowth is critical for understanding the role of forests in national-scale carbon dynamics. Remote sensing data have the potential to capture these two processes, but the spatial grain size and temporal measurement density needed to map them accurately makes wall-to-wall mapping impractical. Here, we describe the sampling and estimation strategy created for the North American Forest Disturbance (NAFD) project to estimate biennial forest disturbance rates for the conterminous United States. The sampling strategy is designed to estimate national biennial forest disturbance and recovery rates using detailed observations in Landsat time series stacks (LTSS). Sample units are defined by the Thiessen polygon describing the non-overlapping area of each TM path/row address in the conterminous U.S. (referred to as Thiessen scene areas, or TSAs), stratified into eastern and western forest areas. An unequal probability sampling approach was used: A series of constraints were used to cull from 100,000 random ordered TSA lists (ROTLs) a small set of potential sample sets, the final sample of which was chosen randomly from the culled set with known inclusion probabilities. The final sample balances four competing objectives: 1. Scene dispersion, 2. Capture of diverse forest types, 3. Inclusion of TSAs with high forest area and 4. Inclusion of TSAs where existing data and analysis have been conducted. Estimation is based on standard Horvitz-Thompson estimators, with variance estimates provided by two different approaches (collapsed stratum and modeled).
NASA Carbon Cycle & Ecosystems Active Awards Represented by this Poster: