ASSESSING POTENTIAL CLIMATE-CHANGE EFFECTS ON LOBLOLLY-PINE GROWTH - A PROBABILISTIC REGIONAL MODELING APPROACH

Most models of the potential effects of climate change on forest growt h have produced deterministic predictions. However, there are large un certainties in data on regional forest condition, estimates of future climate, and quantitative relationships between environmental conditio ns and forest growth rate. We constructed a new model to analyze these uncertainties along with available experimental results to make proba bilistic estimates of climate change effects on the growth of loblolly pine (Pinus taeda L.) throughout its range in the USA. Complete regio nal data sets were created by means of spatial interpolation, and unce rtainties in these data were estimated. A geographic information syste m (GIS) was created to integrate current and predicted climate data wi th regional data including forest distribution, growth rate, and stand characteristics derived from USDA Forest Service data. A probabilisti c climate change scenario was derived from the results of four differe nt general circulation models (GCM). Probabilistic estimates of forest growth were produced by linking the GIS to a Latin Hypercube carbon ( C) budget model of forest growth. The model estimated a greater than 5 0% chance of a decrease in loblolly pine growth throughout most of its range. The model also estimated a 10% chance that the total regional basal area growth will decrease by more than 24 X 10(6) m(2) yr(-1) (a 92% decrease), and a 10% chance that basal area growth will increase by more than 62 X 10(6) m(2) yr(-1) (a 142% increase above current rat es). The most influential factor at all locations was the relative cha nge in C assimilation. Of climatic factors, CO2 concentration was foun d to be the most influential factor at all locations. Substantial regi onal variation in estimated growth was observed, and probably was due primarily to variation in historical growth rates and to the importanc e of historical growth in the model structure. (C) 1998 Elsevier Scien ce B.V.