A verification framework for interannual-to-decadal predictions experiments

Decadal predictions have a high profile in the climate science community and beyond, yet very little is known about their skill. Nor is there any agreed protocol for estimating their skill. This paper proposes a sound and coordinated framework for verification of decadal hindcast experiments. The framework is illustrated for decadal hindcasts tailored to meet the requirements and specifications of CMIP5 (Coupled Model Intercomparison Project phase 5). The chosen metrics address key questions about the information content in initialized decadal hindcasts. These questions are: (1) Do the initial conditions in the hindcasts lead to more accurate predictions of the climate, compared to un-initialized climate change projections? and (2) Is the prediction model’s ensemble spread an appropriate representation of forecast uncertainty on average? The first question is addressed through deterministic metrics that compare the initialized and uninitialized hindcasts. The second question is addressed through a probabilistic metric applied to the initialized hindcasts and comparing different ways to ascribe forecast uncertainty. Verification is advocated at smoothed regional scales that can illuminate broad areas of predictability, as well as at the grid scale, since many users of the decadal prediction experiments who feed the climate data into applications or decision models will use the data at grid scale, or downscale it to even higher resolution. An overall statement on skill of CMIP5 decadal hindcasts is not the aim of this paper. The results presented are only illustrative of the framework, which would enable such studies. However, broad conclusions that are beginning to emerge from the CMIP5 results include (1) Most predictability at the interannual-to-decadal scale, relative to climatological averages, comes from external forcing, particularly for temperature; (2) though moderate, additional skill is added by the initial conditions over what is imparted by external forcing alone; however, the impact of initialization may result in overall worse predictions in some regions than provided by uninitialized climate change projections; (3) limited hindcast records and the dearth of climate-quality observational data impede our ability to quantify expected skill as well as model biases; and (4) as is common to seasonal-to-interannual model predictions, the spread of the ensemble members is not necessarily a good representation of forecast uncertainty. The authors recommend that this framework be adopted to serve as a starting point to compare prediction quality across prediction systems. The framework can provide a baseline against which future improvements can be quantified. The framework also provides guidance on the use of these model predictions, which differ in fundamental ways from the climate change projections that much of the community has become familiar with, including adjustment of mean and conditional biases, and consideration of how to best approach forecast uncertainty.

Scale-selective verification of rainfall accumulations from high-resolution forecasts of convective events

The development of NWP models with grid spacing down to 1 km should produce more realistic forecasts of convective storms. However, greater realism does not necessarily mean more accurate precipitation forecasts. The rapid growth of errors on small scales in conjunction with preexisting errors on larger scales may limit the usefulness of such models. The purpose of this paper is to examine whether improved model resolution alone is able to produce more skillful precipitation forecasts on useful scales, and how the skill varies with spatial scale. A verification method will be described in which skill is determined from a comparison of rainfall forecasts with radar using fractional coverage over different sized areas. The Met Office Unified Model was run with grid spacings of 12, 4, and 1 km for 10 days in which convection occurred during the summers of 2003 and 2004. All forecasts were run from 12-km initial states for a clean comparison. The results show that the 1-km model was the most skillful over all but the smallest scales (approximately <10–15 km). A measure of acceptable skill was defined; this was attained by the 1-km model at scales around 40–70 km, some 10–20 km less than that of the 12-km model. The biggest improvement occurred for heavier, more localized rain, despite it being more difficult to predict. The 4-km model did not improve much on the 12-km model because of the difficulties of representing convection at that resolution, which was accentuated by the spinup from 12-km fields.

Some numerical experiments on the effect of the variation of static stability in two-layer quasi-geostrophic models

A method to solve a quasi-geostrophic two-layer model including the variation of static stability is presented. The divergent part of the wind is incorporated by means of an iterative procedure. The procedure is rather fast and the time of computation is only 60–70% longer than for the usual two-layer model. The method of solution is justified by the conservation of the difference between the gross static stability and the kinetic energy. To eliminate the side-boundary conditions the experiments have been performed on a zonal channel model. The investigation falls mainly into three parts: The first part (section 5) contains a discussion of the significance of some physically inconsistent approximations. It is shown that physical inconsistencies are rather serious and for these inconsistent models which were studied the total kinetic energy increased faster than the gross static stability. In the next part (section 6) we are studying the effect of a Jacobian difference operator which conserves the total kinetic energy. The use of this operator in two-layer models will give a slight improvement but probably does not have any practical use in short periodic forecasts. It is also shown that the energy-conservative operator will change the wave-speed in an erroneous way if the wave-number or the grid-length is large in the meridional direction. In the final part (section 7) we investigate the behaviour of baroclinic waves for some different initial states and for two energy-consistent models, one with constant and one with variable static stability. According to the linear theory the waves adjust rather rapidly in such a way that the temperature wave will lag behind the pressure wave independent of the initial configuration. Thus, both models give rise to a baroclinic development even if the initial state is quasi-barotropic. The effect of the variation of static stability is very small, qualitative differences in the development are only observed during the first 12 hours. For an amplifying wave we will get a stabilization over the troughs and an instabilization over the ridges.

Verification of European subseasonal wind speed forecasts

Analysis of the forecasts and hindcasts from the ECMWF 32-day forecast model reveals that there is statistically significant skill in predicting weekly mean wind speeds over areas of Europe at lead times of at least 14–20 days. Previous research on wind speed predictability has focused on the short- to medium-range time scales, typically finding that forecasts lose all skill by the later part of the medium-range forecast. To the authors’ knowledge, this research is the first to look beyond the medium-range time scale by taking weekly mean wind speeds, instead of averages at hourly or daily resolution, for the ECMWF monthly forecasting system. It is shown that the operational forecasts have high levels of correlation (~0.6) between the forecasts and observations over the winters of 2008–12 for some areas of Europe. Hindcasts covering 20 winters show a more modest level of correlation but are still skillful. Additional analysis examines the probabilistic skill for the United Kingdom with the application of wind power forecasting in mind. It is also shown that there is forecast “value” for end users (operating in a simple cost/loss ratio decision-making framework). End users that are sensitive to winter wind speed variability over the United Kingdom, Germany, and some other areas of Europe should therefore consider forecasts beyond the medium-range time scale as it is clear there is useful information contained within the forecast.

A new method for the characterisation and verification of local spatial predictability for convective scale ensembles

The use of kilometre-scale ensembles in operational forecasting provides new challenges for forecast interpretation and evaluation to account for uncertainty on the convective scale. A new neighbourhood based method is presented for evaluating and characterising the local predictability variations from convective scale ensembles. Spatial scales over which ensemble forecasts agree (agreement scales, S^A) are calculated at each grid point ij, providing a map of the spatial agreement between forecasts. By comparing the average agreement scale obtained from ensemble member pairs (S^A(mm)_ij), with that between members and radar observations (S^A(mo)_ij), this approach allows the location-dependent spatial spread-skill relationship of the ensemble to be assessed. The properties of the agreement scales are demonstrated using an idealised experiment. To demonstrate the methods in an operational context the S^A(mm)_ij and S^A(mo)_ij are calculated for six convective cases run with the Met Office UK Ensemble Prediction System. The S^A(mm)_ij highlight predictability differences between cases, which can be linked to physical processes. Maps of S^A(mm)_ij are found to summarise the spatial predictability in a compact and physically meaningful manner that is useful for forecasting and for model interpretation. Comparison of S^A(mm)_ij and S^A(mo)_ij demonstrates the case-by-case and temporal variability of the spatial spread-skill, which can again be linked to physical processes.