|
Forest Ecosystem Processes
>
Decision Support Tools
>
Most Similar Neighbor (MSN)
March 2007: MSN is replaced with an R package called yaImpute.
Support for MSN has been discontinued.
August 2003: Version 2.12 Update now available for user testing.
Often it is desirable to impute ground-based measurements to sample units that are not measured on the ground.
For example, consider using a model that requires ground-based forest inventory
for all sample units to simulate the future of a landscape.
Typically, the ground-based inventory required to run the model is not available for all sample units.
The Most Similar Neighbor (MSN) application is a powerful tool used to impute available ground-based
inventory data to non-inventoried units.
The MSN method uses available data from the ground-based sample units and globally
available data measured on all sample units to guide the imputation.
Examples of global information known about all sample units include
topographic data, photo-interpreted data, and satellite imagery.
1. Problem: Ground-based measurements exist for some
sample units but not others. Ground-based data are needed for all
sample units for landscape analysis.
2. Solution: Find the most similar sample unit for which
ground-based measurements exist to represent a sample unit for which
ground-based measurements do not exist. Impute the most similar
ground-based inventory to each sample unit for which no ground-based
inventory exists.
3. Desireable Attributes: Make use of all the available
information for units that have ground-based measurements and
information about those units for which no ground-based measurements
exist.
4. Result: Fully populate the planning area with realistic
ground-based inventory information.
Applications:
- Filling in an inventory. For example, updating a current
vegetation layer.
- Creating mapped data layers. For example, creating a species
composition map.
|
Download MSN Application:
|
 |
Download User's Guide:
|
Publications Available:
Crookston, Nicholas L.; Moeur, Melinda; Renner, David. 2002.
Users guide to the Most Similar Neighbor Imputation Program Version 2.
Gen. Tech. Rep. RMRS-GTR-96.
Ogden, UT: U.S. Department of Agriculture, Forest Service, Rocky Mountain Research Station. 35 p.
Moeur, Melinda. 2000. Extending stand exam data with most similar
neighbor inference. In: Proceedings of the Society of American
Foresters National Convention; 1999 September 11-15; Portland, OR.
SAF Publication 00-1. Bethesda, MD: Society of American Foresters: 99-107.
Moeur, Melinda, and Rachel Rismann Hershey. 1999.
Preserving spatial and attribute correlation in the interpolation
of forest inventory data. Chapter 49 (pp. 419-430) in K. Lowell,
and A. Jaton (eds.), Spatial accuracy assessment: land information
uncertainty in natural resources. Ann Arbor Press.
Crookston, Nicholas L. and Melinda Moeur. 1998. Using most similar neighbor
inference to impute ground-based inventory measurements to sample
units for which no ground-based inventory exists. Poster,
Integrated tools for natural resources inventories in the 21st
century, August 16-19, 1998, Boise, ID (USDA FS, SAF, IUFRO and Boise Cascade, sponsors).
Moeur, Melinda, Nicholas L. Crookston, and Albert R. Stage. 1995.
Most similar neighbor analysis: a tool to support ecosystem management.
Proceedings, Analysis in support of ecosystem management.
Analysis Workshop III, April 10-13, 1995, Fort Collins, CO.
Washington, D.C.: USDA, Forest Service, Ecosystem Management Analysis Center, 360 pp.
Moeur, Melinda and Albert R. Stage. 1995. Most similar neighbor: an
improved sampling inference procedure for natural resource
planning. Forest Science 41(2):337-359.
|
