Publications


Stochastic modelling and energy-efficient computing for weather and climate prediction.

Philosophical transactions. Series A, Mathematical, physical, and engineering sciences 372 (2014) 20140118-

T Palmer, P Düben, H McNamara


Climate forecasting: build high-resolution global climate models.

Nature 515 (2014) 338-339

T Palmer


Modes of climate variability and heat waves in Victoria, southeastern Australia

Geophysical Research Letters 41 (2014) 6926-6934

TJ Parker, GJ Berry, MJ Reeder, N Nicholls


The Structure and Evolution of Heat Waves in Southeastern Australia

Journal of Climate 27 (2014) 5768-5785

TJ Parker, GJ Berry, MJ Reeder


Addressing model error through atmospheric stochastic physical parametrizations: impact on the coupled ECMWF seasonal forecasting system.

Philosophical transactions. Series A, Mathematical, physical, and engineering sciences 372 (2014) 20130290-

A Weisheimer, S Corti, T Palmer, F Vitart

The finite resolution of general circulation models of the coupled atmosphere-ocean system and the effects of sub-grid-scale variability present a major source of uncertainty in model simulations on all time scales. The European Centre for Medium-Range Weather Forecasts has been at the forefront of developing new approaches to account for these uncertainties. In particular, the stochastically perturbed physical tendency scheme and the stochastically perturbed backscatter algorithm for the atmosphere are now used routinely for global numerical weather prediction. The European Centre also performs long-range predictions of the coupled atmosphere-ocean climate system in operational forecast mode, and the latest seasonal forecasting system--System 4--has the stochastically perturbed tendency and backscatter schemes implemented in a similar way to that for the medium-range weather forecasts. Here, we present results of the impact of these schemes in System 4 by contrasting the operational performance on seasonal time scales during the retrospective forecast period 1981-2010 with comparable simulations that do not account for the representation of model uncertainty. We find that the stochastic tendency perturbation schemes helped to reduce excessively strong convective activity especially over the Maritime Continent and the tropical Western Pacific, leading to reduced biases of the outgoing longwave radiation (OLR), cloud cover, precipitation and near-surface winds. Positive impact was also found for the statistics of the Madden-Julian oscillation (MJO), showing an increase in the frequencies and amplitudes of MJO events. Further, the errors of El Niño southern oscillation forecasts become smaller, whereas increases in ensemble spread lead to a better calibrated system if the stochastic tendency is activated. The backscatter scheme has overall neutral impact. Finally, evidence for noise-activated regime transitions has been found in a cluster analysis of mid-latitude circulation regimes over the Pacific-North America region.


Benchmark Tests for Numerical Weather Forecasts on Inexact Hardware

MONTHLY WEATHER REVIEW 142 (2014) 3809-3829

PD Dueben, TN Palmer


How Does the Quasi-Biennial Oscillation Affect the Stratospheric Polar Vortex?

JOURNAL OF THE ATMOSPHERIC SCIENCES 71 (2014) 391-409

PAG Watson, LJ Gray


On the use of inexact, pruned hardware in atmospheric modelling.

Philosophical transactions. Series A, Mathematical, physical, and engineering sciences 372 (2014) 20130276-

PD Düben, J Joven, A Lingamneni, H McNamara, G De Micheli, KV Palem, TN Palmer

Inexact hardware design, which advocates trading the accuracy of computations in exchange for significant savings in area, power and/or performance of computing hardware, has received increasing prominence in several error-tolerant application domains, particularly those involving perceptual or statistical end-users. In this paper, we evaluate inexact hardware for its applicability in weather and climate modelling. We expand previous studies on inexact techniques, in particular probabilistic pruning, to floating point arithmetic units and derive several simulated set-ups of pruned hardware with reasonable levels of error for applications in atmospheric modelling. The set-up is tested on the Lorenz '96 model, a toy model for atmospheric dynamics, using software emulation for the proposed hardware. The results show that large parts of the computation tolerate the use of pruned hardware blocks without major changes in the quality of short- and long-time diagnostics, such as forecast errors and probability density functions. This could open the door to significant savings in computational cost and to higher resolution simulations with weather and climate models.


The real butterfly effect

NONLINEARITY 27 (2014) R123-R141

TN Palmer, A Doering, G Seregin


The use of imprecise processing to improve accuracy in weather & climate prediction

Journal of Computational Physics (2013)

PD Düben, TN Palmer, H McNamara

The use of stochastic processing hardware and low precision arithmetic in atmospheric models is investigated. Stochastic processors allow hardware-induced faults in calculations, sacrificing bit-reproducibility and precision in exchange for improvements in performance and potentially accuracy of forecasts, due to a reduction in power consumption that could allow higher resolution. A similar trade-off is achieved using low precision arithmetic, with improvements in computation and communication speed and savings in storage and memory requirements. As high-performance computing becomes more massively parallel and power intensive, these two approaches may be important stepping stones in the pursuit of global cloud-resolving atmospheric modelling. The impact of both hardware induced faults and low precision arithmetic is tested using the Lorenz '96 model and the dynamical core of a global atmosphere model. In the Lorenz '96 model there is a natural scale separation; the spectral discretisation used in the dynamical core also allows large and small scale dynamics to be treated separately within the code. Such scale separation allows the impact of lower-accuracy arithmetic to be restricted to components close to the truncation scales and hence close to the necessarily inexact parametrised representations of unresolved processes. By contrast, the larger scales are calculated using high precision deterministic arithmetic. Hardware faults from stochastic processors are emulated using a bit-flip model with different fault rates. Our simulations show that both approaches to inexact calculations do not substantially affect the large scale behaviour, provided they are restricted to act only on smaller scales. By contrast, results from the Lorenz '96 simulations are superior when small scales are calculated on an emulated stochastic processor than when those small scales are parametrised. This suggests that inexact calculations at the small scale could reduce computation and power costs without adversely affecting the quality of the simulations. This would allow higher resolution models to be run at the same computational cost. © 2013 The Authors.


Influence of stochastic sea ice parametrization on climate and the role of atmosphere-sea ice-ocean interaction

Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences 372 (2014) 20130283-20130283

S Juricke, T Jung


Studying edge geometry in transiently turbulent shear flows

Journal of Fluid Mechanics 747 (2014) 506-517

M Chantry, TM Schneider

© 2014 Cambridge University Press.In linearly stable shear flows at moderate Reynolds number, turbulence spontaneously decays despite the existence of a codimension-one manifold, termed the edge, which separates decaying perturbations from those triggering turbulence. We statistically analyse the decay in plane Couette flow, quantify the breaking of self-sustaining feedback loops and demonstrate the existence of a whole continuum of possible decay paths. Drawing parallels with low-dimensional models and monitoring the location of the edge relative to decaying trajectories, we provide evidence that the edge of chaos does not separate state space globally. It is instead wrapped around the turbulence generating structures and not an independent dynamical structure but part of the chaotic saddle. Thereby, decaying trajectories need not cross the edge, but circumnavigate it while unwrapping from the turbulent saddle.


Genesis of streamwise-localized solutions from globally periodic traveling waves in pipe flow.

Phys Rev Lett 112 (2014) 164501-

M Chantry, AP Willis, RR Kerswell

The aim in the dynamical systems approach to transitional turbulence is to construct a scaffold in phase space for the dynamics using simple invariant sets (exact solutions) and their stable and unstable manifolds. In large (realistic) domains where turbulence can coexist with laminar flow, this requires identifying exact localized solutions. In wall-bounded shear flows, the first of these has recently been found in pipe flow, but questions remain as to how they are connected to the many known streamwise-periodic solutions. Here we demonstrate that the origin of the first localized solution is in a modulational symmetry-breaking Hopf bifurcation from a known global traveling wave that has twofold rotational symmetry about the pipe axis. Similar behavior is found for a global wave of threefold rotational symmetry, this time leading to two localized relative periodic orbits. The clear implication is that many global solutions should be expected to lead to more realistic localized counterparts through such bifurcations, which provides a constructive route for their generation.


More reliable forecasts with less precise computations: a fast-track route to cloud-resolved weather and climate simulators?

Philosophical transactions. Series A, Mathematical, physical, and engineering sciences 372 (2014) 20130391-

TN Palmer

This paper sets out a new methodological approach to solving the equations for simulating and predicting weather and climate. In this approach, the conventionally hard boundary between the dynamical core and the sub-grid parametrizations is blurred. This approach is motivated by the relatively shallow power-law spectrum for atmospheric energy on scales of hundreds of kilometres and less. It is first argued that, because of this, the closure schemes for weather and climate simulators should be based on stochastic-dynamic systems rather than deterministic formulae. Second, as high-wavenumber elements of the dynamical core will necessarily inherit this stochasticity during time integration, it is argued that the dynamical core will be significantly over-engineered if all computations, regardless of scale, are performed completely deterministically and if all variables are represented with maximum numerical precision (in practice using double-precision floating-point numbers). As the era of exascale computing is approached, an energy- and computationally efficient approach to cloud-resolved weather and climate simulation is described where determinism and numerical precision are focused on the largest scales only.


Potential sea ice predictability and the role of stochastic sea ice strength perturbations

Geophysical Research Letters 41 (2014) 8396-8403

S Juricke, HF Goessling, T Jung

©2014. The Authors.Ensemble experiments with a climate model are carried out in order to explore how incorporating a stochastic ice strength parameterization to account for model uncertainty affects estimates of potential sea ice predictability on time scales from days to seasons. The impact of this new parameterization depends strongly on the spatial scale, lead time and the hemisphere being considered: Whereas the representation of model uncertainty increases the ensemble spread of Arctic sea ice thickness predictions generated by atmospheric initial perturbations up to about 4 weeks into the forecast, rather small changes are found for longer lead times as well as integrated quantities such as total sea ice area. The regions where initial condition uncertainty generates spread in sea ice thickness on subseasonal time scales (primarily along the ice edge) differ from that of the stochastic sea ice strength parameterization (along the coast lines and in the interior of the Arctic). For the Antarctic the influence of the stochastic sea ice strength parameterization is much weaker due to the predominance of thinner first year ice. These results suggest that sea ice data assimilation and prediction on subseasonal time scales could benefit from taking model uncertainty into account, especially in the Arctic. Key PointsSea ice model uncertainty estimates increase spread for subseasonal predictionsSeasonal prediction estimates not affected by sea ice model uncertainties


Build high-resolution global climate models

NATURE 515 (2014) 338-339

T Palmer


It's all just physics

PHYSICS WORLD 27 (2014) 18-19

T Palmer


The influence of tropical cyclones on heat waves in Southeastern Australia

Geophysical Research Letters 40 (2013) 6264-6270

TJ Parker, GJ Berry, MJ Reeder


Estimation of the local response to a forcing in a high dimensional system using the fluctuation-dissipation theorem

Nonlinear Processes in Geophysics 20 (2013) 239-248

FC Cooper, JG Esler, PH Haynes

The fluctuation-dissipation theorem (FDT) has been proposed as a method of calculating the response of the earth's atmosphere to a forcing. For this problem the high dimensionality of the relevant data sets makes truncation necessary. Here we propose a method of truncation based upon the assumption that the response to a localised forcing is spatially localised, as an alternative to the standard method of choosing a number of the leading empirical orthogonal functions. For systems where this assumption holds, the response to any sufficiently small non-localised forcing may be estimated using a set of truncations that are chosen algorithmically. We test our algorithm using 36 and 72 variable versions of a stochastic Lorenz 95 system of ordinary differential equations. We find that, for long integrations, the bias in the response estimated by the FDT is reduced from ∼75% of the true response to ∼30%. © 2013 Author(s).


Impact of snow initialization on sub-seasonal forecasts

Climate Dynamics (2013) 1-14

YJ Orsolini, R Senan, G Balsamo, F Vitart, A Weisheimer, FJ Doblas-Reyes, A Carrasco, RE Benestad

The influence of the snowpack on wintertime atmospheric teleconnections has received renewed attention in recent years, partially for its potential impact on seasonal predictability. Many observational and model studies have indicated that the autumn Eurasian snow cover in particular, influences circulation patterns over the North Pacific and North Atlantic. We have performed a suite of coupled atmosphere-ocean simulations with the European Centre for Medium-Range Weather Forecasts (ECMWF) ensemble forecast system to investigate the impact of accurate snow initialisation. Pairs of 2-month ensemble forecasts were started every 15 days from the 15th of October through the 1st of December in the years 2004-2009, with either realistic initialization of snow variables based on re-analyses, or else with "scrambled" snow initial conditions from an alternate autumn date and year. Initially, in the first 15 days, the presence of a thicker snowpack cools surface temperature over the continental land masses of Eurasia and North America. At a longer lead of 30-day, it causes a warming over the Arctic and the high latitudes of Eurasia due to an intensification and westward expansion of the Siberian High. It also causes a cooling over the mid-latitudes of Eurasia, and lowers sea level pressures over the Arctic. This "warm Arctic-cold continent" difference means that the forecasts of near-surface temperature with the more realistic snow initialization are in closer agreement with re-analyses, reducing a cold model bias over the Arctic and a warm model bias over mid-latitudes. The impact of realistic snow initialization upon the forecast skill in snow depth and near-surface temperature is estimated for various lead times. Following a modest skill improvement in the first 15 days over snow-covered land, we also find a forecast skill improvement up to the 30-day lead time over parts of the Arctic and the Northern Pacific, which can be attributed to the realistic snow initialization over the land masses. © 2013 Springer-Verlag Berlin Heidelberg.