The Effects of Spatial Uncertainties of Ground Plots and Corresponding Satellite Pixels on Mapping Forest Communities

摘要

The USDA-Forest Service's Forest Inventory and Analysis program maintains an array of over 300,000 permanent ground plots systematically spread over the entire United States for use in estimating forest attributes and many other applications. The forest inventory plots are geo-referenced with uncorrected GPS. The USGS - EROS Data Center has responsibility for providing land cover and vegetation cover maps of the entire country as well as derived fire risk maps using Landsat TM satellite imagery. Recently the two agencies have statistically combined the two efforts in order to create continuous maps of forest characteristics. Assessments of accuracy of the resulting maps of forest volume, size, canopy height and species type indicate that a considerable amount of error is due to positional uncertainties and inaccuracies of both the plots and the image pixels. An evaluation of the area of positional uncertainty of the input data as well as an evaluation of how accuracy changes based on how the data sets are combined provides a method for making the most accurate map. Furthermore, it leads to increased understanding of how ground based data interacts with satellite pixel brightness values. This paper describes the issue and suggests solutions.