Integrating Contemporary and Historical Climate Records to Predict - PowerPoint PPT Presentation

Integrating Contemporary and Historical Climate Records to Predict Future Climate-Species Suitability for Big Sagebrush in WNA Kyle Taylor University of Wyoming

Overview • Paleo-Climate Data Compilations • NOAA Labs • Methods for Reconstructing Climate Conditions • Tree-ring Records • Lake-Sediment Records • Building a Statistical Model Using Climate Data • Some Initial Predictions for Climate-Species Suitability in Western North America • Summary

Dataset Compilations • NOAA Paleo-climate Models

Tree-Ring Records Tree-ring records can be used to build linear models of correlated climate factors, particularly precipitation and temperature. 1. Fit a Regression Model to Ring Widths Using Recent Records of Precipitation and Temperature Ring width = β + f(precipitation + temperature ) + ε 2. Solve for Temperature / Precipitation Using Ring Widths for Records that Extend Beyond the Range of Instrument Data.

Linear Approximation and Interpolation of Tree-Ring Data Ring width = β + f(precipitation + temperature + …) + ε

Lake Sediment Records • Abundance of plant species inferred from quantity of pollen grain • Nearby fire occurrence inferred from charcoal Species Abundance = Suitability

Current (observed) Model Overview Paleo-climate Record

Hierarchical Regression =precipitation, =temperature, =fire intensity.

Making Predictions 1. Train our model with observations of precipitation, temperature, and fire at each grid cell in the paleo-climate and current-climate record. 2. Take , coefficients fit to paleo-climate and current-climate and apply them to projections of temperature, precipitation, and projected fire-risk in the future.

High suitability (p>=0.7) Moderate suitability (p>=0.2)

Summary • Using observations of species abundance and climate-effects from the Paleontological record has the potential to greatly improve climate-species modelling. • Climate datasets derived from paleo-ecological tree-ring and lake- sediment records are improving and becoming more useful for reliable species modelling. • My model cannot be evaluated well, so performance cannot be quantified as easily as traditional species distribution models. • Over-prediction of suitability may be a problem, as well as extrapolation of predictions beyond the range of data used in training models