A method to aid in the determination of the sampling size of AGCM ensemble simulations

Abstract

Historically, the high computational expense of performing lengthy climate simulations has limited the number of possible realizations. Here, we exploit the computational advantage of a recently developed parallel atmospheric general circulation model (AGCM) to determine the number of realizations required to calculate the model's statistics to a specified degree of certainty. Using standard statistical analysis techniques, this minimum ensemble size is found to be highly dependent on which output field is under examination. A strong dependence on location, season and averaging is also revealed.