The major application of digital remotely sensed data has traditionally been viewed as wall-to-wall mapping.  More recently, however, RCR personnel have been involved in developing new, and sometimes re-discovering forgotten applications that use remotely sensed imagery.  The imagery is used either in a sampling mode, as elements of a multi-stage sample that includes several imagery types and field data, or as ancillary data that support the FIA pre-field classification of plots. 

In the western United States, fluvial riparian ecosystems often comprise less than five percent of the total landscape, making them difficult to sample adequately with standard field sampling procedures. The objective of this project was to develop a method of inventorying these sparsely distributed resources for a relatively large area (Wyoming). The method was a multistage, nested area frame sample that used imagery at different resolutions to iteratively refine riparian vegetation cover estimates. Moderate resolution imagery (MODIS time series) was used to create five strata.  Primary sampling units (PSUs) were chosen in each stratum with the number of primary sampling units dependent upon the variability in the remote sensing imagery. Using a DEM to define valley bottoms and mid-resolution imagery (Landsat) to classify vegetation, the vegetation in the PSUs was classified as riparian or non-riparian. Riparian vegetation estimates were calculated for the individual stratum and for the entire state of Wyoming using this riparian / non-riparian classification.

In the National Forest System, the increasing need for statistically defensible estimates of inventory variables leaves National Forests with a difficult choice, use mid-level vegetation maps “as is” with an inadequate sample size of associated FIA data, or revert to the use of biased and outdated stand-exam data that cannot provide statistically defensible estimates and have no explicit relationship to their mid-level map products used for forest plan revision.  Alternatives to this untenable choice include several expensive approaches, including intensifying the base grid to provide an adequate sample size, or implementing a traditional two-stage sample of the map features depicting vegetation pattern.  The overarching objective of this project is to evaluate statistical efficiency and cost / benefit relationship of an intensification of the base grid using photo data.  The plan is to test these relationships in at least two different locations, which will include both a western softwood and eastern mixed hardwood environment.

RCR personnel developed a multi-stage sampling design to estimate Pinyon Pine mortality on the Williams Ranger District of the Coconino National Forest.  The multi-stage sample used three strata: 1) Dominance Type (pinyon-juniper) derived from Landsat TM imagery; 2) Canopy Density derived from Landsat TM merged with Digital Orthophoto Quads, and; 3) Mortality derived from natural color digital photography.  The results showed 7% of the land area had mortality while 20% of the trees were dead.  The methodology provided a cost-effective means of obtaining mortality estimates with statistically valid error estimates.