i
Fast NN emulation of the super-parameterization in the multi-scale modeling framework
-
2014
-
Details:
-
Personal Author:
-
Corporate Authors:
-
NOAA Program & Office:
-
Description:Extension of the NN emulation framework for superparameterization (SP) has been developed. The current PNNL-MMF has been used to provide the training set for a NN emulator. The large scale fields from the MMF that provide the SP inputs are used as the NN input vector and the aggregated results of the SP are used as the output vector. The limited training dataset was drawn selectively from four days of MMF global simulations, spanning diurnal and latitudinal variability. Experiments with different NN architectures have been performed and assessed. These experiments include comparisons of a single emulating NN with many outputs vs. a battery of simpler emulating NNs with different outputs. Also experiments with different output normalizations and different emulating NN complexities have been conducted. Training an ensemble of emulating NNs or developing NN-SP has been performed. Performance (accuracy and speedup) of the NN-SP has been estimated on an independent simulated dataset. Multiple statistics have been calculated for NN-SP and the differences between SP and NN-SP. The results obtained in this work can be summarized as follows: (1) The SP can be emulated by NN with a satisfactory accuracy (based on validation on an independent data set). (2) The NN-SP provides an impressive, two orders of magnitude, speedup as compared with the original SP. It provides a practical opportunity to use NN-MMF for decadal and longer climate simulations.
-
Content Notes:V.M. Krasnopolsky, Michael Fox-Rabinovitz, Philip J.Rasch, and Minghuai Wang.
"MMAB Contribution No. 316."
"January, 2014."
System requirements: Adobe Acrobat Reader.
Includes bibliographical references (pages 67-70).
-
Keywords:
-
Series:
-
Document Type:
-
Place as Subject:
-
Rights Information:Public Domain
-
Compliance:LIbrary
-
Main Document Checksum:
-
Download URL:
-
File Type: