Combining ground-based Forest Inventory and Analysis (FIA) data and map-based topographic and MODIS data to map landscape attributes in Nevada

Background/Question/Methods

Ground-based forest inventories often tabulate population statistics for geographic areas, land ownerships, or ecological classifications. Frequently, users also desire maps of inventory information. However, map-based technologies generally start with remote-sensing information independent of land-based inventories or use model-based methods to map selected variables from them. Our objective was to map land-based inventory statistics for all attribute variables (e.g., forest area, basal area, volume) in a common sample design, without developing separate models for each variable. The Forest Inventory and Analysis (FIA) grid network of plots covers all forest land in the United States. Double sampling for stratification is used: phase-1 information is from multispectral data and phase-2 information is from field plots. In our study, we combined ground-based FIA data with remote-sensing information to produce cover maps of commonly calculated inventory statistics for Nevada, a sparsely forested state that features mountainous dryland and conifer forests between valleys of rangeland and desert. A logistic regression model was built to estimate the probability of the predominant basal area cover from inventory and auxiliary topographic variables. Then, using this classification of the Nevada landscape, carbon and other FIA variables were mapped by linking FIA analysis to map categories.
Results/Conclusions Modeling results allowed for simple assembly of map pixels for estimating statistics from FIA data for an area of interest. We mapped carbon estimation for Nevada land ownerships as an example. Carbon ranged as high as 173 Mg/ha for Nevada’s conifer-aspen forests, but averaged 30 Mg/ha overall for the state. Carbon averages for Nevada’s forest types dominated by the indicated trees included: 19 Mg/ha for juniper, 33 Mg/ha for pinyon, 37 Mg/ha for mountain mahogany, and 65 Mg/ha for conifer-aspen. Using our method, many forest attributes (in addition to carbon) can be easily mapped to produce objective information from field inventories. This is particularly valuable in an area such as Nevada, where much of the land is publicly owned and there has been a contentious history of tree cutting and livestock grazing. Combining ground-based and map-based data into the same sample design holds promise for improved forest landscape planning and management and for monitoring forest ecology in response to human activities.