Use of integrated global climate model simulations and statistical time series forecasting to project regional temperature and precipitation

Abstract

An integrated technique combining global climate model (GCM) simulation results and a statistical time series forecasting model [the autoregressive integrated moving average (ARIMA) model] was developed to bring together the climate change signal from GCMs to city-level historical observations as an approach to obtain location-specific temperature and precipitation projections. This approach assumes that regional temperature and precipitation time series reflect a combination of an underlying climate change signal series and a regional-deviation-from-the-signal series. An ensemble of GCMs is used to describe and provide the climate change signal, and the ARIMA model is used to model and project the regional deviation. Qualitative and quantitative assessments were conducted for evaluating the projection performance of the hybrid GCM-ARIMA (G-ARIMA) model. The results indicate that the G-ARIMA model can provide projected city-specific daily temperature and precipitation series comparable to historical observations and can have im-proved projection accuracy for several assessed annual indices compared to a commonly used downscaled projection product. The G-ARIMA model is subject to some limitations and uncertainties from the GCM-provided climate change signal. A notable feature of the G-ARIMA model is the efficiency with which projections can be updated when new observations become available, thus facilitating updating of regional temperature and precipitations projections. Given the increasing need for and use of location-specific climate projections in practical engineering applications, the G-ARIMA model is an option for regional temperature and precipitation projection for such applications.