Stochastic parametrizations and model uncertainty in the Lorenz ’96 system

Abstract

Simple chaotic systems are useful tools for testing methods for use in numerical weather simulations owing to their transparency and computational cheapness. The Lorenz system was used here; the full system was defined as ‘truth’, whereas a truncated version was used as a testbed for parametrization schemes. Several stochastic parametrization schemes were investigated, including additive and multiplicative noise. The forecasts were started from perfect initial conditions, eliminating initial condition uncertainty. The stochastically generated ensembles were compared with perturbed parameter ensembles and deterministic schemes. The stochastic parametrizations showed an improvement in weather and climate forecasting skill over deterministic parametrizations. Including a temporal autocorrelation resulted in a significant improvement over white noise, challenging the standard idea that a parametrization should only represent sub-gridscale variability. The skill of the ensemble at representing model uncertainty was tested; the stochastic ensembles gave better estimates of model uncertainty than the perturbed parameter ensembles. The forecasting skill of the parametrizations was found to be linked to their ability to reproduce the climatology of the full model. This is important in a seamless prediction system, allowing the reliability of short-term forecasts to provide a quantitative constraint on the accuracy of climate predictions from the same system.

Accounting for Model Error from Unresolved Scales in Ensemble Kalman Filters by Stochastic Parameterization

“…It has been shown to have a mean close to zero (see e.g. (Arnold et al 2013;Chorin and Lu 2015;Mitchell and Carrassi 2015)), hence the biases in the means of forecast ensembles are negligible. This removes a concern about ensemble bias in the use of covariance inflation and localization to account for such model error (Dee and Da Silva 1998;Li et al 2009).…”

“…We take K = 18, J = 20, F = 10, h x = −1 and h y = 1. Here one model time unit is approximately equal to five atmospheric days, deduced by comparing the error doubling time of the model to that observed in the atmosphere (Lorenz 1996;Arnold et al 2013;Mitchell and Carrassi 2015).…”

Reliable, robust and realistic: the three R's of next-generation land-surface modelling

“…Palmer, 2012;Arnold et al, 2013). Stochastic parameterizations represent one or more model parameters as a statistical distribution of values.…”

Climate SPHINX: evaluating the impact of resolution and stochastic physics parameterisations in the EC-Earth global climate model

“…Neelin, 2000, 2003;Arnold et al, 2013). Berner et al (2012) showed that including stochastic physics can reduce systematic biases in the model's mean climate, comparable to improvements gained by increasing the model resolution.…”