Predictability of weather and climate

Abstract:

© Author(s) 2020. This work is distributed under This study examines methods of calibrating projections of future regional climate for the next 40–50 years using large single-model ensembles (the Community Earth System Model (CESM) Large Ensemble and Max Planck Institute (MPI) Grand Ensemble), applied over Europe. The three calibration methods tested here are more commonly used for initialised forecasts from weeks up to seasonal timescales. The calibration techniques are applied to ensemble climate projections, fitting seasonal ensemble data to observations over a reference period (1920–2016). The calibration methods were tested and verified using an “imperfect model” approach using the historical/representative concentration pathway 8.5 (RCP8.5) simulations from the Coupled Model Intercomparison Project 5 (CMIP5) archive. All the calibration methods exhibit a similar performance, generally improving the out-of-sample projections in comparison to the uncalibrated (bias-corrected) ensemble. The calibration methods give results that are largely indistinguishable from one another, so the simplest of these methods, namely homogeneous Gaussian regression (HGR), is used for the subsequent analysis. As an extension to the HGR calibration method it is applied to dynamically decomposed data, in which the underlying data are separated into dynamical and residual components (HGR-decomp). Based on the verification results obtained using the imperfect model approach, the HGR-decomp method is found to produce more reliable and accurate projections than the uncalibrated ensemble for future climate over Europe. The calibrated projections for temperature demonstrate a particular improvement, whereas the projections for changes in precipitation generally remain fairly unreliable. When the two large ensembles are calibrated using observational data, the climate projections for Europe are far more consistent between the two ensembles, with both projecting a reduction in warming but a general increase in the uncertainty of the projected changes.

Abstract:

The need for high‐precision calculations with 64‐bit or 32‐bit floating‐point arithmetic for weather and climate models is questioned. Lower‐precision numbers can accelerate simulations and are increasingly supported by modern computing hardware. This paper investigates the potential of 16‐bit arithmetic when applied within a shallow water model that serves as a medium complexity weather or climate application. There are several 16‐bit number formats that can potentially be used (IEEE half precision, BFloat16, posits, integer, and fixed‐point). It is evident that a simple change to 16‐bit arithmetic will not be possible for complex weather and climate applications as it will degrade model results by intolerable rounding errors that cause a stalling of model dynamics or model instabilities. However, if the posit number format is used as an alternative to the standard floating‐point numbers, the model degradation can be significantly reduced. Furthermore, mitigation methods, such as rescaling, reordering, and mixed precision, are available to make model simulations resilient against a precision reduction. If mitigation methods are applied, 16‐bit floating‐point arithmetic can be used successfully within the shallow water model. The results show the potential of 16‐bit formats for at least parts of complex weather and climate models where rounding errors would be entirely masked by initial condition, model, or discretization error.