A study of reduced numerical precision to make superparameterization more competitive using a hardware emulator in the OpenIFS model. (28th February 2017)

• Record Type: Journal Article
• Authors: Düben, Peter D.; Subramanian, Aneesh; Dawson, Andrew; Palmer, T. N.
• Is Part Of: Journal of advances in modeling earth systems. Volume 9:Number 1(2017)
• Page Start: 566
• Page End: 584
• Publication Date: 2017-02-28
• Subjects: superparameterization -- reduced precision -- inexact hardware -- cloud resolving model -- high‐performance computing -- model uncertainty; Geological modeling -- Periodicals; Climatology -- Periodicals; Geochemical modeling -- Periodicals; 551.5011
• DOI: 10.1002/2016MS000862
• Languages: English
• ISSNs: 1942-2466

Abstract

Abstract: The use of reduced numerical precision to reduce computing costs for the cloud resolving model of superparameterized simulations of the atmosphere is investigated. An approach to identify the optimal level of precision for many different model components is presented, and a detailed analysis of precision is performed. This is nontrivial for a complex model that shows chaotic behavior such as the cloud resolving model in this paper. It is shown not only that numerical precision can be reduced significantly but also that the results of the reduced precision analysis provide valuable information for the quantification of model uncertainty for individual model components. The precision analysis is also used to identify model parts that are of less importance thus enabling a reduction of model complexity. It is shown that the precision analysis can be used to improve model efficiency for both simulations in double precision and in reduced precision. Model simulations are performed with a superparameterized single‐column model version of the OpenIFS model that is forced by observational data sets. A software emulator was used to mimic the use of reduced precision floating point arithmetic in simulations. Plain Language Summary: Weather and climate models cannot represent physical processes of the Earth System explicitly that are smaller than the distance between model grid points. Due to limitations in computing power, this distance is typically larger than 10 km in simulations of the global atmosphere. However, the spatial scale for many important physical processes, such as clouds, is much smaller than this and large errors are generated for predictions of both weather and climate due to limited resolution. To approximate the behavior of subgrid‐scale processes within atmosphere models, superparameterization was developed that is running a two‐dimensional small‐scale model within each grid column of the global model. Superparameterization can improve model simulations but it causes a very large increase in computational cost in comparison to standard simulations. To reduce computational cost, this paper investigates whether it is possible to reduce numerical precision when running the small‐scale model. It is shown that precision can indeed be reduced such that computing costs can potentially be reduced significantly. It is also shown that results of an investigation of reduced numerical precision provide valuable information for the quantification of model uncertainty and model development. Key Points: To reduce computing cost for superparameterized atmosphere models, a reduction in precision is studied A detailed precision analysis is performed that relates numerical precision to model uncertainty Results of the precision analysis are used to improve the model for both high and low‐precision runs