COS 15-4 - Assessing the impact of detrending methods on black and white spruce tree rings response to climate change

Abstract: In the boreal forests of the northern high latitudes, tree ring widths have been historically positively correlated with temperature; however, the strength of this correlation has not been universal, and is shifting with climate change. Trees are responding to not only warming temperatures, but the myriad consequences of warming, such as thawing permafrost, shifting hydrology and changing nutrient cycles. These co-occurring environmental shifts are creating site-specific conditions to which trees must adapt. Using black and white spruce tree ring chronologies from NOAA’s International Tree Ring Database that spanned Alaska and Canada, we divided our chronologies by species and tree age. We compared nine different detrending methods (dplR) to determine the most representative chronology. Once the best three detrending methods were confirmed, we applied linear mixed effects models to determine the strength of different climate, topographic, and edaphic predictor variables. In order to use the CRU climate data, we truncated the results to be from 1901 onwards. Most of our sites (45) were sampled only until 2000, but we were able to look at seven additional sites with chronologies going into the 2010s.

Our results show that for different age groups, the best method of detrending varies. Older sites tended to be better represented by the basal area increment detrending, while younger sites were better matched with the regional curve standardization method. The linear mixed effect models had low predictive power, but suggested that site level characteristics were important to consider when assessing a stand’s projected response to climate change. Soil characteristics (soil organic carbon and sand content) and topography (slope and aspect) were strong predictors; while many other variables were correlated with latitude and longitude and require further analysis. By analyzing the tree rings through a variety of detrending methods, we were better able to suit the method to the site, so that residual effects of annual radial growth and lifetime changes were better captured. Our findings suggest that the detrending methods will leave residual effects on the chronology used to determine drivers of growth; but by using a suite of detrending methods we have greater confidence in our results.