Ronald E. McRoberts 
,
Daniel G. Wendt,
Greg C. Liknes
Stratified estimation of forest inventory variables using spatially summarized stratifications
McRoberts R. E., Wendt D. G., Liknes G. C. (2005). Stratified estimation of forest inventory variables using spatially summarized stratifications. Silva Fennica vol. 39 no. 4 article id 478. https://doi.org/10.14214/sf.478
Abstract
Large area natural resource inventory programs typically report estimates for selected geographic areas such as states or provinces, counties, and municipalities. To increase the precision of estimates, inventory programs may use stratified estimation, with classified satellite imagery having been found to be an efficient and effective basis for stratification. For the benefit of users who desire additional analyses, the inventory programs often make data and estimation procedures available via the Internet. For their own analyses, users frequently request access to stratifications used by the inventory programs. When data analysis is via the Internet and stratifications are based on classifications of even medium resolution satellite imagery, the memory requirements for storing the stratifications and the online time for processing them may be excessive. One solution is to summarize the stratifications at coarser spatial scales, thus reducing both storage requirements and processing time. If the bias and loss of precision resulting from using summaries of stratifications is acceptably small, then this approach is viable. Methods were investigated for using summaries of stratifications that do not require storing and processing the entire pixel-level stratifications. Methods that summarized satellite image-based 30 m x 30 m pixel stratifications at spatial scales up to 2400 ha produced stratified estimates of the mean that were generally within 5-percent of estimates for the same areas obtained using the pixel stratifications. In addition, stratified estimates of variances using summarized stratifications realized nearly all the gain in precision that was obtained with the underlying pixel stratifications.
Keywords
bias;
precision;
classified satellite imagery;
Internet;
variance
-
McRoberts,
North Central Research Station, USDA Forest Service, 1992 Folwell Avenue, Saint Paul, Minnesota, USA 5510
E-mail
rmcroberts@fs.fed.us
- Wendt, Region 9, USDA Forest Service, 626 East Wisconsin Avenue, Milwaukee, Wisconsin 53202, USA E-mail dgw@nn.us
- Liknes, North Central Research Station, USDA Forest Service, 1992 Folwell Avenue, Saint Paul, Minnesota, USA 5510 E-mail gcl@nn.us
Received 28 December 2004 Accepted 5 September 2005 Published 31 December 2005
Views 3122
Available at https://doi.org/10.14214/sf.478 | Download PDF
