Mapping U.S. forest biomass using nationwide forest inventory data and moderate resolution information

Abstract

A spatially explicit dataset of aboveground live forest biomass was made from ground measured inventory plots for the conterminous U.S., Alaska and Puerto Rico. The plot data are from the USDA Forest Service Forest Inventory and Analysis (FIA) program. To scale these plot data to maps, we developed models relating field-measured response variables to plot attributes serving as the predictor variables. The plot attributes came from intersecting plot coordinates with geospatial datasets. Consequently, these models serve as mapping models. The geospatial predictor variables included Moderate Resolution Imaging Spectrometer (MODIS)-derived image composites and percent tree cover; land cover proportions and other data from the National Land Cover Dataset (NLCD); topographic variables; monthly and annual climate parameters; and other ancillary variables. We segmented the mapping models for the U.S. into 65 ecologically similar mapping zones, plus Alaska and Puerto Rico. First, we developed a forest mask by modeling the forest vs. nonforest assignment of field plots as functions of the predictor layers using classification trees in See5©. Secondly, forest biomass models were built within the predicted forest areas using tree-based algorithms in Cubist©. To validate the models, we compared field-measured with model-predicted forest/nonforest classification and biomass from an independent test set, randomly selected from available plot data for each mapping zone. The estimated proportion of correctly classified pixels for the forest mask ranged from 0.79 in Puerto Rico to 0.94 in Alaska. For biomass, model correlation coefficients ranged from a high of 0.73 in the Pacific Northwest, to a low of 0.31 in the Southern region. There was a tendency in all regions for these models to over-predict areas of small biomass and under-predict areas of large biomass, not capturing the full range in variability. Map-based estimates of forest area and forest biomass compared well with traditional plot-based estimates for individual states and for four scales of spatial aggregation. Variable importance analyses revealed that MODIS-derived information could contribute more predictive power than other classes of information when used in isolation. However, the true contribution of each variable is confounded by high correlations. Consequently, excluding any one class of variables resulted in only small effects on overall map accuracy. An estimate of total C pools in live forest biomass of U.S. forests, derived from the nationwide biomass map, also compared well with previously published estimates.

Introduction

The Forest Inventory and Analysis (FIA) program of the USDA Forest Service collects data annually on the status and trends in forested ecosystems nationwide. These inventory data support estimates of forest population totals over large geographic areas, (Scott et al., 2005). Regional maps of forest characteristics would make these extensive forest resource data more accessible and useful to a larger and more diverse audience. Important applications of such maps include broad-scale mapping and assessment of wildlife habitat; documenting forest resources affected by fire, fragmentation, and urbanization; identifying land suitable for timber production; and locating areas at high risk for plant invasions, or insect or disease outbreaks. Thus, there is a need to produce and distribute geospatial data of forest attributes, complementing FIA inventory data.

Total aboveground live biomass is a forest characteristic of particular interest. Forest soils and woody biomass hold most of the carbon in Earth's terrestrial biomes (Houghton, 1999). Land-use change, mainly forest burning, harvest, or clearing for agriculture, may compose 15 to 40% of annual human-caused emissions of carbon to the atmosphere, and terrestrial ecosystems, mainly through forest growth and expansion, absorb nearly as much carbon annually. However, estimates of land-atmosphere carbon fluxes, and the net of expected future ones, have the largest uncertainties in the global atmospheric carbon budget, which adds to uncertainties about future levels and impacts of greenhouse gasses (GHGs) in the atmosphere (Houghton, 2003, Prentice et al., 2001).

Consequently, the levels, mechanisms and spatial distribution of forest land-atmosphere C fluxes are an important focus for reducing uncertainties in the global C budget (Fan et al., 1998, Holland et al., 1999, Pacala et al., 2001, Schimel et al., 2001). Ecosystem process models that are physiologically-based, and that use satellite image-derived indices of photosynthesis, have permitted unprecedented global assessments of ecosystem productivity and carbon sinks at a spatial resolution of 0.5° (Nemani et al., 2003, Potter et al., 2003). The mechanistic nature of these models identifies how observed patterns in ecosystem productivity may relate to climate and atmospheric changes (Nemani et al., 2003). However, validating atmospheric and ecosystem model estimates of net forest C fluxes, and quantifying the C fluxes associated with changes in land use, which dominate these fluxes over longer time periods, requires spatially extensive data on forest C pools and net fluxes. Maps of forest biomass permit spatially explicit estimates of forest carbon storage and net fluxes from land-use change.

Our objectives here are to 1) produce a spatially explicit dataset of aboveground live forest biomass from ground measured inventory plots, at a 250-m cell size, for the conterminous U.S., Alaska and Puerto Rico; 2) evaluate model performance and spatially depict uncertainty in the dataset; 3) explore the relative contribution of the many predictor layers to the biomass models; and 4) use the resulting dataset to estimate aboveground live forest biomass and implied carbon storage for this area. We also describe a national geospatial predictor database that supported the mapping and how we standardized national FIA data, developed predictive models, and assessed model error.