> Research > Intergration of Remote Sensing


This research was conducted on a contiguous four county region in the northeast part of Mississippi (Figure 1).  This area was chosen for its biophysical variability and contains most inventory scenarios present in the state.  In addition to geo-correcting all of the Landsat data used in the study, analysis of the data was performed by Veridian ERIM-International to identify forested areas and assess the age and composition of these areas (Roller 2000) (Figure 2).  Forest area mapping was based on spectral categorization of the leaf-on 1998 Thematic Mapper (TM) data. Forest age was assessed using a first and last date forest stratified, hybrid change detection procedure. This procedure yielded six forest age classes based on the available Landsat Multispectral Scanner (MSS) and TM data.  Forest composition assessment was based on 1999 leaf-off  TM data. The assessment was performed using a forest and age stratified spectral based procedure.  Forest composition was assessed with respect to the relative amount of evergreen and deciduous forest present within a pixel.

Using the information from the latest FIA plots (1993-1994) as a preliminary sample, it was determined that approximately 100 plots per county were needed to achieve a ± 10% sampling error for total cubic foot volume at the 95% level of confidence for each county. Plots were randomly chosen from contiguous areas of at least 8 acres in size. Non-forested areas were excluded. Route files were made of the coordinates of the plots.  Using real time differential GPS receivers (accurate to ± one meter), two-member field crews navigated to each randomly located plot center.  A circular fixed area plot of 0.20 acres was established and attributes of trees within this area were measured and recorded using data recorders. Measurements were used to determine volume, growth and site attributes. All measurements taken were designed to be compatible with FIA plot data.

Evaluation of the original 100 plots per county indicated the desired sampling error had not been achieved. Additional plots were randomly allocated to each county as needed using the procedures described above to achieve the target precision levels.

In an effort to assure quality field data, error checks were conducted in the field and at database creation time. A computer algorithm was used to screen the data collected by the field crews. The algorithm examined the field data, applied error checking rules set to detect, and correct errors, wrote the corrected data to the database, and printed the original and corrected record for manual correction in the field data files.