A landscape-level analysis of yellow-cedar decline in coastal British Columbia

Abstract

Yellow-cedar (Chamaecyparis nootkatensis D. Don (Spach)) is currently undergoing a dramatic decline in western North America. Recent research suggests that site factors combined with a shift in climate have predisposed yellow-cedar trees to decline. We conducted the first landscape-level analysis of the decline in coastal British Columbia to assess relations between the decline and topographic variables. We used lasso-penalized logistic regression to model yellow-cedar decline presence and absence with topographic variables derived from a digital elevation model. Model results indicated that low elevation sites close to the coast, which are more exposed and have more variation in elevation, are more likely to show evidence of decline. The logistic model fit the data well (Nagelkerke R2= 0.846) and had high predictive accuracy (AUC = 0.98). The topographic variables identified by the model influence degree of soil saturation, temperatures, and snowpack presence in a forest stand, supporting the proposed associations in the current decline hypothesis. The analysis also highlighted the utility of the lasso logistic model for selecting significant variables and mapping areas at high risk for decline. Knowledge of the determinants of the spatial pattern of decline will improve predictability and provide critical information for conservation and management of yellow-cedar.