A Time Series Regression Analysis of Future Climate

Abstract

Current approaches to climate modeling, including environmental simulation, may not be able to generate actionable results for a few decades yet. Over the last 50 years, methods attempting to capture and predict states of the climate system have flourished and diversified. However, many such models are subject to errors and uncertainty arising from parameterization problems, the obligate characterization of poorly understood phenomena, and high capacity requirements stemming from the incredible computing power needed. As the window for meaningful actions towards altering the climate change trajectory closes, we should consider the use of simple methods that generally predict the conditions of the future climate. For my analysis, I developed a time-series regression analysis of land surface trends in precipitation and near-surface temperature. For each global 0.5º land surface grid, values for 1901-2009 baseline means were calculated, and 2050 values were predicted using time series regression models for each of four historical data subsets. Average predicted warming across the subsets range from 0.89 ºC to 5.8 ºC above the baseline, with high northern latitudes predicted to experience the most warming. Precipitation is predicted to follow the “wet getting wetter, dry getting dryer” paradigm, with average predicted changes across the subsets ranging from 3.2% to 26% above the baseline.