|
Activity Number:
|
261
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Tuesday, August 10, 2004 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistics and the Environment
|
| Abstract - #302238 |
|
Title:
|
Nationwide Mapping of Forest Type and Biomass: Comparing Nonparametric Modeling Tools
|
|
Author(s):
|
Gretchen Moisen*+
|
|
Companies:
|
U.S. Forest Service
|
|
Address:
|
507 25th Street, Ogden, UT, 84401,
|
|
Keywords:
|
forest inventory ; MARS ; GAM ; CART ; Random Forest ; satellite imagery
|
|
Abstract:
|
The USDA Forest Service Forest Inventory and Analysis program (FIA) collects forest inventory data on a systematic grid across all lands in the United States. Moderate resolution maps of forest biomass and forest type are currently being developed nationwide by modeling response data collected on FIA sample plots as nonparametric functions of numerous ancillary predictor layers. These layers include: 16-day Moderate Resolution Imaging Spectrometer (MODIS) composites and associated vegetation indices and MODIS percent tree cover; vegetative diversity and type synthesized from the National Land Cover Dataset; topographic variables derived from Digital Elevation Models; monthly and annual climate parameters; and other ancillary variables. The nonparametric modeling tools of choice for these production mapping efforts include See5 for the categorical response (forest type) and Cubist for the continuous response (biomass). We compare the performance of these two tools against other nonparametric alternatives using data collected in the interior Western United States.
|
