Recorded presentationJ2.2
Fast neural network emulations of long wave radiation for the NCEP Climate Forecast System Model: seasonal prediction and climate simulation
Vladimir M. Krasnopolsky, NCEP/NWS/NOAA (SAIC), Camp Springs, MD; and M. S. Fox-Rabinovitz, S. J. Lord, Y. T. Hou, and A. A. Belochitski
The approach to calculation of model physics using accurate and fast neural network (NN) emulations, has been previously proposed, developed and thoroughly tested by the authors for the National Center for Atmospheric Research (NCAR) Community Atmospheric Model (CAM) driven by climatological SSTs [Krasnopolsky et al. 2005, Krasnopolsky and Fox-Rabinovitz 2006]. In this study, the NN approach is being implemented for the NCEP climate forecast system (CFS) coupled model. NN emulations have been developed [Krasnopolsky et al. 2008] for one of the available CFS long wave radiation parameterizations, RRTM-LW2 (adapted from AER Inc. RRTM-LW version 3.0 with 256 k-values) [Mlawer et al, 1997]. The model radiation is the most time consuming component of model physics in CFS. A specific effort has been devoted to generalizing/extending the NN methodology/approach for the coupled CFS model. The LWR NN emulations are two orders of magnitude faster than the original LWR.
An extensive validation of LWR NN emulation in the CFS model integrations has been performed. The results of the 10-year (1995-2005) CFS model simulation performed with NN emulation of RRTM-LW, the original LWR parameterization, have been validated against the parallel control NCEP CFS model simulation using the original RRTM-LW. The comparison of seasonal predictions (for the first four seasons of model simulation) and 10-year simulation, in terms of time averaged model prognostic and diagnostic fields as well as their time series show a very close similarity for the parallel runs. The LWR NN emulation is approximately 100 times faster than the original LWR parameterization and the CFS model with the LWR NN emulation is approximately 20% faster than the original CFS.
ACKNOWLEDGMENTS. The authors would like to thank Drs. H.-L. Pan, S. Saha, S. Moorthi, and M. Iredell for their useful consultations and discussions. The research is supported by the NOAA CPO CDEP CTB grant NA06OAR4310047.
REFERENCES:
Krasnopolsky, V.M., M.S. Fox-Rabinovitz, and D.V. Chalikov, 2005: “Fast and Accurate Neural Network Approximation of Long Wave Radiation in a Climate Model”, Monthly Weather Review, vol. 133, No. 5, pp. 1370-1383.
Krasnopolsky, V.M., and M.S. Fox-Rabinovitz, 2006: “A New Synergetic Paradigm in Environmental Numerical Modeling: Hybrid Models Combining Deterministic and Machine Learning Components”, Ecological Modelling, 191, pp. 5–18.
Krasnopolsky V. M., M. S. Fox-Rabinovitz, Y. T. Hou, S. J. Lord, and A. A. Belochitski, 2008: "Accurate and Fast Neural Network Emulations of Long Wave Radiation for the NCEP Climate Forecast System Model", Proc.of 20th Conference on Climate Variability and Change, 88th AMS Annual Meeting, New Orleans, LA, 20-24 January 2008, CD-ROM, P3.10
Mlawer, E.J., S.J. Taubman, P.D. Brown, M.J. Iacono, and S.A. Clough, 1997: “Radiative transfer for inhomogeneous atmospheres: RRTM, a validated correlated-k model for the longwave, J. Geophys. Res., 102, D14, 16,663-16,682
Joint Session 2, Applications of artificial learning techniques in climate variability, especially as it relates to the urban environment
Wednesday, 14 January 2009, 8:30 AM-10:00 AM, Room 125A
Previous paper Next paper