Herding DGVM cats: A fuzzy logic modeling approach to incorporate the implications and uncertainty of DGVM results into resource management decisions

Future projections from dynamic global vegetation models (DGVMs) are often touted as valuable data to aid resource managers in decision making. However, interpreting model results in a meaningful way to support landscape level and policy decisions is challenging. The sheer volume of data can be overwhelming. Results must be interpreted in light of model assumptions and limitations in the model as well those built into the climate models providing DGVM inputs. Furthermore, results vary due to a number of factors including: 1) assumptions about atmospheric CO₂ levels; 2) climate projections used as DGVM inputs; 3) climate data downscaling methods; 4) DGVM parameterization; 5) DGVM algorithm selection.

The Environmental Evaluation Modeling System (EEMS) is a spatial fuzzy logic decision support system used to answer questions and provide guidance to resource managers. EEMS models use a node-based, bottom up tree structure and are transparent and easy-to-understand. Through EEMS Online, users can explore models along with associated map layers to understand how inputs and operators contribute to the answer underlying a decision support question.

For Oregon and Washington west of the Cascade crest, we have created a series of EEMS models to inform decisions surrounding carbon sequestration and forest resilience. These models incorporate outputs from over 40 runs of the MC2 DGVM. These results vary due to climate inputs, downscaling methods, fire suppression assumptions, and algorithm selection. These EEMS models incorporate interpretations of model results (for instance the implication of model vegetation type change for tree mortality). Also these models depict the uncertainty across the suite of DGVM model runs by producing the "best case," "worst case," and average results for the management question being asked. This allows managers to constrain projected future conditions without having to interpret gigabytes of raw DGVM outputs and helps bridges the gap between science and decision making.