Sampling intensity and normalizations: Exploring cost-driving factors in nationwide mapping of tree canopy coverThis article is part of a larger document. View the larger document here.
There are many factors that will determine the final cost of modeling and mapping tree canopy cover nationwide. For example, applying a normalization process to Landsat data used in the models is important in standardizing reflectance values among scenes and eliminating visual seams in the final map product. However, normalization at the national scale is expensive and logistically challenging, and its importance to model fit is unknown. Cost also increases with each location sampled, yet appropriate photo sampling intensity relative to the FIA grid has yet to be explored. In addition, cost is also affected by how intensively the photo plots themselves are sampled with a dot count, and the effect of reducing the number of dots on predictive models is also unknown. Using intensively sampled photo plot data in 5 pilot areas across the United States, we address these three cost factors by exploring the effect of a normalization process of Landsat TM data on model fits of tree canopy cover using Random Forests regression, the relationship between the sampling intensity of photo interpreted plots and model fit, and the relationship between the number of dots for each photo interpreted location and model fit.