Estimating Forest Productivity Using Site Characteristics, Multipoint Measures, and a Nonparametric Approach

Abstract

Abstract Understanding the productivity of forestland is essential in sustainable management of forest ecosystems. The most common measure of site productivity is breast height–age site index (BHASI). BHASI has limitations as a productivity measure and can compound error in predictive models. We explored the accuracy of productivity predictions using an alternative productivity measure (10-meter site index) and a nonparametric approach. An orthogonal sampling design ensured samples were collected across the range of conditions known to influence Douglas-fir (Pseudotsuga menziesii var. glauca) height-growth rates. Using climate, soil, and topographic data along with 10-meter site index measurements, we compared five possible models to estimate forest productivity. Model parameters, performance, and predictions were compared. Twelve validation sites were used to test the accuracy of model predictions. Model performance was significantly improved when smoothing span values were optimized and elevation was added as a predictor. A four-predictor nonparametric model with a bias-corrected Akaike information criterion–optimized smoothing span value produced the most accurate results and was used to produce forest productivity maps for the study area. The low resolution of currently available climatic data and the complex nature of the study area landscape necessitate a topographic variable for accurate productivity predictions. Study Implications Defining and understanding forest productivity is of interest to a wide variety of natural resource professionals including ecologists, climate change experts, forest biometricians, and forest managers. A new method of defining forest productivity using multipoint height-age pairs at 10 and 20 meters and calculated growth rates combined with an appropriate landscape-scale stratification and a nonparametric approach provides accurate productivity estimates. This method is more widely applicable and more precise for specific locations than previous productivity estimation methods. Better productivity and tree growth information will provide more accurate estimates of future forest condition and structure.