Satellite remote sensing for enhancing national forest inventory

Many countries have established national forest inventories (NFIs) to assess and monitor their forest resources. NFIs typically are based on statistically designed samples of field data, from which design-based estimates are produced that describe the amount, condition, health, and change in forest resources. This thesis surveys the use of remote sensing activities within NFIs and assesses satellite remote sensing approaches for mapping forest attributes and for increasing the statistical precision of design-based estimates. One NFI is addressed in detail, the Forest Inventory and Analysis (FIA) program of the United States Department of Agriculture (USDA), Forest Service. Forest land area estimates produced by FIA and by the USDA Natural Resources Conservation Service's National Resources Inventory are compared with each other and with estimates derived from four satellite image-based land cover products. An image stratification approach is proposed for calibrating satellite image-based sub-pixel estimates of continuous forest attributes, such that resulting per-area estimates match estimates from inventory sources. The calibration technique is tested on the Moderate Resolution Imaging Spectroradiometer (MODIS) Vegetation Continuous Fields (VCF) dataset. Satellite image classifications provide a means for stratifying design-based estimates even after field data are collected. Several image classification and estimation approaches are tested to determine their effect on precision of subsequent stratified estimates. Satellite imaging sensors have a range of radiometric, spectral, temporal, and spatial resolution characteristics. A 30 m spatial resolution dataset is aggregated to multiple coarser resolutions to determine the effects of aggregation technique and resulting spatial resolution on image estimates and design-based stratified estimates of forest land area.