Comprehensive review and evaluation of whole stand volume yield models for the aspen forest type in Minnesota, USA

Abstract

Abstract Aspen (Populus spp.) exists as a forest type with circumboreal significance. In Minnesota, aspen is the most abundant and utilized timber species. Dating from 1929 to as recently as 2019, many whole stand volume yield models have been developed for the aspen forest type in Minnesota. With many model forms to choose from, this work sought to narrow down the best performing models across a myriad of forest conditions and locations. Data from recent cycles of the U.S. Forest Service, Forest Inventory and Analysis program were utilized for model fitting and assessment. Existing model forms were assessed, as well as 423 new model forms. Models were cross validated and compared with predictive accuracy statistics including root mean square error, mean absolute deviation and an agreement index (pseudo-R2). Analyses were conducted to evaluate model performance in relevant subsets, including: statewide, ecological section, age class and site class. To reflect model complexity and differences in available inventory data, models were also assessed by the number of required input variables. A newly postulated model form, the GZ model, provided improved accuracy across all subset scenarios, whereas a recently developed model form provided simplicity and applicability across many subsets. This work provides users of aspen whole stand volume yield models with a method to review model forms fit under a variety of conditions and select the most appropriate model.