Integration of a statistical time series forecasting model with GCMs to provide location-specific regional temperature and precipitation projections

Abstract:

Location-specific future temperature and precipitation information has been increasingly utilized in engineering practice such as updating design values and risk assessments for existing infrastructure. Experience to date with use of GCM projections and downscaling techniques in practical applications has indicated a range of limitations and uncertainties. An alternative approach involving integration of GCMs with statistical forecasting of long-term historical observations was developed to obtain improved projections at individual cities for engineering use. A commonly used statistical time series forecasting model – Autoregressive Integrated Moving Average Model (ARIMA) ­– was identified and utilized to perform statistical analyses and extrapolation of the historical climate trend and variability in regional observations. ARIMA modeling results were incorporated with the climate change projections obtained from an ensemble of GCMs. Additional statistical techniques were included to provide the projected city-level temperature and precipitation in daily resolution and to estimate return periods of annual extremes. By bringing together climate models with location-specific long-term historical observations, this integrated GCM-ARIMA technique aims to provide an option for acquiring temperature and precipitation projections at particular locations for infrastructure engineering.