Flow-dependent versus flow-independent initial perturbations for ensemble prediction

Ensemble prediction relies on a faithful representation of initial uncertainties in a forecasting system. Early research on initial perturbation methods tested random perturbations by adding 'white noise' to the analysis. Here, an alternative kind of random perturbations is introduced by using the difference between two randomly chosen atmospheric states (i.e. analyses). It yields perturbations (random field, RF, perturbations) in approximate flow balance.
The RF method is compared with the operational singular vector based ensemble at European Centre for Medium Range Weather Forecasts (ECMWF) and the ensemble transform (ET) method. All three methods have been implemented on the ECMWF IFS-model with resolution T _L 255L40. The properties of the different perturbation methods have been investigated both by comparing the dynamical properties and the quality of the ensembles in terms of different skill scores. The results show that the RF perturbations initially have the same dynamical properties as the natural variability of the atmosphere. After a day of integration, the perturbations from all three methods converge. The skill scores indicate a statistically significant advantage for the RF method for the first 2–3 d for the most of the evaluated parameters. For the medium range (3–8 d), the differences are very small.     

Predicting the East Australian Current

Results are presented from an ensemble prediction study (EPS) of the East Australian Current (EAC) with a specific focus on the examination of the role of dynamical instabilities and flow dependent growing errors. The region where the EAC separates from the coast, is characterized by significant mesoscale eddy variability, meandering and is dominated by nonlinear dynamics thereby representing a severe challenge for operational forecasting. Using analyses from OceanMAPS, the Australian operational ocean forecast system, we explore the structures of flow dependent forecast errors over 7 days and examine the role of dynamical instabilities. Forecast ensemble perturbations are generated using the method of bred vectors allowing the identification of those perturbations to a given initial state that grow most rapidly. We consider a 6 month period spanning the Austral summer that corresponds to the season of maximum eddy variability. We find that the bred vector (BV) structures occur in areas of instability where forecast errors are large and in particular in regions associated with the Tasman Front and EAC extension. We also find that very few BVs are required to identify these regions of large forecast error and on that basis we expect that even a small BV ensemble would prove useful for adaptive sampling and targeted observations. The results presented also suggest that it may be beneficial to supplement the static background error covariances typically used in operational ocean data assimilation systems with flow dependent background errors calculated using a relatively cheap EPS.     

On ensemble prediction of ocean waves

The numerical ensemble prediction is a well accepted method for improving the performance of atmospheric models. In the context of ocean wave modeling little has been researched or documented about this technique. An essential study of the method of ensemble prediction applied to deep water waves has been carried out. A framework is defined for obtaining perturbations of the directional wave spectra and for employing an ensemble of wind fields generated by an atmospheric model. The third-generation global wave model WAM is used with real atmospheric conditions to investigate the effect on wave predictions of perturbed initial conditions and atmospheric forcing. Due to spectral shape stabilisation, perturbing wave initial conditions has limited utility in ensemble prediction. However, the members could be used in wave data assimilation schemes in an interactive way. Using ensembles of the atmospheric condition can generate diverging solutions, justifying the ensemble procedure by itself. In the cases studied, it is observed that the ensemble mean outperformed the other members. The solution behaviour suggests using a lower-order approximation of the model to generate ensemble members with less computational cost.     

Temperature and precipitation extremes in the second half of the twentieth century from numerical modeling results and observational data

The procedure and results of an objective comparison of climatologies and historical trends of temperature and precipitation extremes are considered using observations and numerical experiments of 20th-century climate simulations. The experimental climatic realizations of five coupled atmosphere-ocean models represented in the Fourth Assessment Report of the Intergovernmental Panel on Climate Change are used. The effect of using ensembles of climatic realizations (including multimodel ensembles) in the problems of simulating climatology and climate changes in annual meteorological extremes is estimated. The positive effect of the ensemble approach is evident when individual ensemble members show significant success.     

Examination of targeting methods in a simplified setting

The effectiveness of 2 methods for targeting observations is examined using a T21 L3 QG model in a perfect model context. Target gridpoints are chosen using the pseudo‐inverse (the inverse composed of the first three singular vectors only) and the quasi‐inverse or backward integration (running the tangent equations with a negative time‐step). The effectiveness of a target is measured by setting the analysis error to zero in a region surrounding the target and noting the impact on the forecast error in the verification region. In a post‐time setting, when the targets are based on forecast errors that are known exactly, both methods provide targets that are significantly better than targets chosen at random within a broad region upstream of the verification region. When uncertainty is added to the verifying analysis such that the forecast error is known inexactly, the pseudo‐inverse targets still perform very well, while the backward integration targets are degraded. This degradation due to forecast uncertainty is especially significant when the targets are a function of height as well as horizontal position. When an ensemble‐forecast difference is used in place of the inexact forecast error, the backward integration targets may be improved considerably. However, this significant improvement depends on the characteristics of the initial‐time ensemble perturbation. Pseudo‐inverse targets based on ensemble forecast differences are comparable to pseudo‐inverse targets based on exact forecast errors. Targets based on the largest analysis error are also found to be considerably more effective than random targets. The collocation of the backward integration and pseudo‐inverse targets appears to be a good indicator of target skill.     

Sensitivity analysis of the climate of a chaotic system

This paper addresses some fundamental methodological issues concerning the sensitivity analysis of chaotic geophysical systems. We show, using the Lorenz system as an example, that a naïve approach to variational ("adjoint") sensitivity analysis is of limited utility. Applied to trajectories which are long relative to the predictability time scales of the system, cumulative error growth means that adjoint results diverge exponentially from the "macroscopic climate sensitivity"(that is, the sensitivity of time‐averaged properties of the system to finite‐amplitude perturbations). This problem occurs even for time‐averaged quantities and given infinite computing resources. Alternatively, applied to very short trajectories, the adjoint provides an incorrect estimate of the sensitivity, even if averaged over large numbers of initial conditions, because a finite time scale is required for the model climate to respond fully to certain perturbations. In the Lorenz (1963) system, an intermediate time scale is found on which an ensemble of adjoint gradients can give a reasonably accurate (O(10%)) estimate of the macroscopic climate sensitivity. While this ensemble‐adjoint approach is unlikely to be reliable for more complex systems, it may provide useful guidance in identifying important parameter‐combinations to be explored further through direct finite‐amplitude perturbations.     

Using a mesoscale ensemble to predict forecast error and perform targeted observation

Using NCEP short range ensemble forecast（SREF） system,demonstrated two fundamental on-going evolutions in numerical weather prediction（NWP） are through ensemble methodology.One evolution is the shift from traditional single-value deterministic forecast to flow-dependent（not statistical） probabilistic forecast to address forecast uncertainty.Another is from a one-way observation-prediction system shifting to an interactive two-way observation-prediction system to increase predictability of a weather system.In the first part,how ensemble spread from NCEP SREF predicting ensemble-mean forecast error was evaluated over a period of about a month.The result shows that the current capability of predicting forecast error by the 21-member NCEP SREF has reached to a similar or even higher level than that of current state-of-the-art NWP models in predicting precipitation,e.g.,the spatial correlation between ensemble spread and absolute forecast error has reached 0.5 or higher at 87 h（3.5 d） lead time on average for some meteorological variables.This demonstrates that the current operational ensemble system has already had preliminary capability of predicting the forecast error with usable skill,which is a remarkable achievement as of today.Given the good spread-skill relation,the probability derived from the ensemble was also statistically reliable,which is the most important feature a useful probabilistic forecast should have.The second part of this research tested an ensemble-based interactive targeting（E-BIT） method.Unlike other mathematically-calculated objective approaches,this method is subjective or human interactive based on information from an ensemble of forecasts.A numerical simulation study was performed to eight real atmospheric cases with a 10-member,bred vector-based mesoscale ensemble using the NCEP regional spectral model（RSM,a sub-component of NCEP SREF） to prove the concept of this E-BIT method.The method seems to work most effective for basic atmospheric state variables,moderately effective for convective instabilities and least effective for precipitations.Precipitation is a complex result of many factors and,therefore,a more challenging field to be improved by targeted observation.     

Deep North Atlantic freshening simulated in a coupled climate model

The observed recent freshening trend in the deep North Atlantic and the Labrador Sea is investigated in three forced ensembles and a long control simulations using the HadCM3 coupled ocean–atmosphere–sea-ice climate model. The 40 yr freshening trend during the late half of the 20th century is captured in the all forcings ensemble that applies all major external (natural and anthropogenic) forcing factors. Each ensemble has four members with different initial conditions taking from the control run at a 100 yr interval. No similar freshening trend is found in each of the four corresponding periods of the control simulation. However, there are five large freshening events in a 1640 yr period of the control run, each following a sudden salinity increase. A process analysis revealed that the increase in salinity in the Labrador Sea is closely linked to deep convections while the following freshening trend is accompanied by a period of very weak convective activities.The fact that none of the five large freshening events appears in the four corresponding periods following the initial conditions of the four members of the all forcings ensemble suggest that external forcings may have contributed to triggering the events. Further analyses of two other ensemble simulations (natural forcings only and anthropogenic forcings only) have shown that natural rather than anthropogenic factors are responsible. Based on our model results, we can not attribute the simulated freshening to anthropogenic climate change.     

Sea surface temperature predictability in the North Pacific from multi-model seasonal forecast

Sea surface temperature (SST) prediction based on the multi-model seasonal forecast with numerous ensemble members have more useful skills to estimate the possibility of climate events than individual models. Hence, we assessed SST predictability in the North Pacific (NP) from multi-model seasonal forecasts. We used 23 years of hindcast data from three seasonal forecasting systems in the Copernicus Climate Change Service to estimate the prediction skill based on temporal correlation. We evaluated the predictability of the SST from the ensemble members' width spread, and co-variability between the ensemble mean and observation. Our analysis revealed that areas with low prediction skills were related to either the large spread of ensemble members or the ensemble members not capturing the observation within their spread. The large spread of ensemble members reflected the high forecast uncertainty, as exemplified in the Kuroshio–Oyashio Extension region in July. The ensemble members not capturing the observation indicates the model bias; thus, there is room for improvements in model prediction. On the other hand, the high prediction skills of the multi-model were related to the small spread of ensemble members that captures the observation, as in the central NP in January. Such high predictability is linked to El Niño Southern Oscillation (ENSO) via teleconnection.

     

GCM experiments on changes in atmospheric predictability associated with the PNA pattern and tropical SST anomalies

Based on results from a simple three-level quasi-geostrophic model, Lin and Derome suggested that atmospheric predictability is influenced by the Pacific/North American (PNA) pattern. In the present study, predictability experiments are conducted with the Canadian Centre for Climate Modelling and Analysis general circulation model (CCCma GCM). A 47-yr integration of the GCM with specified sea surface temperature (SST) for the years 1948–94 is first performed. Forecasts are initiated whenever the PNA pattern is in a strong positive or strong negative phase during this simulation. For each forecast, an ensemble of six initial conditions is generated with small random perturbations. Forecasts initiated when the PNA is in its positive phase have smaller growth rates of ensemble standard deviation than forecasts initiated when the PNA is in its negative phase. Regional characteristics of the prediction spread are also examined. Similar experiments are conducted to determine the relationship between atmospheric predictability and SST anomalies in the tropical Pacific. Forecasts initiated when tropical SST anomalies are positive have smaller growth rates of ensemble standard deviation than forecasts initiated when tropical SST anomalies are negative. However, cases with positive tropical SST anomalies but without a strong PNA pattern show a similar prediction spread to cases with negative SST anomalies. The results suggest that, in comparison to the PNA pattern, the influence of tropical SST anomalies is only secondary. A set of three-layer diagnostic equations is used to analyze the GCM results. It is speculated that the transient eddies have a stronger influence on the circulation anomalies (and therefore reduce the atmospheric predictability more) in the negative PNA phase than in the positive PNA phase.     

The spectra of singular values in a regional model

Large portions of the spectra of singular values are determined for both moist and dry versions of a tangent linear, regional model for 4 different synoptic cases. Norms considered include the usual energy norm and versions of a norm measuring only the energy in some set of rotational mode perturbations. At most, only a few percent of the singular vectors possible with any of the norms are growing ones. Inclusion of moist physics in the tangent linear model greatly affects the leading singular vectors but does not increase the number of growing singular vectors much. Most singular vectors are damping ones, and therefore random perturbations drawn from a white‐noise distribution will likely damp during the 24‐h forecast periods considered. Only a few singular vectors are required to explain a significant portion of the final‐time variance of such perturbations, however, because the leading singular values are so large compared with the rest. The truncated rotational mode norm is shown to be very useful for investigating these properties.     

Some remarks on the westward propagation of the monsoon depression

Several westward propagation properties of the Indian monsoon depression were neglected by previous studies. They include:(1) the slower propagation speed of the depression depicted by a quasi‐geostrophic model, (2) the initiation of the asymmetric secondary circulation with respect to the depression center, and (3) the absence of the depression perturbation in the upper troposphere. Some further insights into these neglected propagation properties of the depression are obtained from the streamfunction budget analysis with the ECMWF (European Centre for Medium Range Weather Forecasts) reanalysis data. (1) The inclusion of relative vorticity stretching, which is neglected in a quasi‐geostrophic model, increases the depression's westward propagation speed. (2) Within the large‐scale environment of the summer monsoon, the coupling of the east‐west differentiation of the meridional absolute vorticity advection with the CISK mechanism is conducive to the initiation and development of the asymmetric secondary circulation associated with the depression. (3) The Tibetan high is formed by summertime global‐scale stationary waves which are maintained by a Sverdrup balance. The positive streamfunction tendency induced by the upper‐tropospheric vortex stretching over the monsoon region suppresses the development of the monsoon depression in the upper troposphere.     

北极气旋的季节、年际变化及其与北极海冰、大气遥相关的关系

The seasonal and inter-annual variations of Arctic cyclone are investigated. An automatic cyclone tracking algorithm developed by University of Reading was applied on the basis of European Center for Medium-range Weather Forecasts(ECMWF) ERA-interim mean sea level pressure field with 6 h interval for 34 a period. The maximum number of the Arctic cyclones is counted in winter, and the minimum is in spring not in summer.About 50% of Arctic cyclones in summer generated from south of 70°N, moving into the Arctic. The number of Arctic cyclones has large inter-annual and seasonal variabilities, but no significant linear trend is detected for the period 1979–2012. The spatial distribution and linear trends of the Arctic cyclones track density show that the cyclone activity extent is the widest in summer with significant increasing trend in CRU(central Russia)subregion, and the largest track density is in winter with decreasing trend in the same subregion. The linear regressions between the cyclone track density and large-scale indices for the same period and pre-period sea ice area indices show that Arctic cyclone activities are closely linked to large-scale atmospheric circulations, such as Arctic Oscillation(AO), North Atlantic Oscillation(NAO) and Pacific-North American Pattern(PNA). Moreover,the pre-period sea ice area is significantly associated with the cyclone activities in some regions.     

Downward longwave radiation in general circulation models: a case study at a semi‐arid continental site

The present case study evaluates the downward longwave radiation at the surface (DLR) in several high‐resolution (≈1°) general circulation models (GCMs) using surface observations from a semiarid continental site in New South Wales, Australia (Uardry, 34.39°S, 142.30°E). This site is located on a large grassland plain uniform in both its land use and landcover type, and is therefore particularly well suited for a comparison with GCM grid mean values. Monthly averages of newly constructed clear‐sky and all‐sky DLR climatologies and the resulting cloud‐radiative forcing are compared. It is shown that the GCMs exceed the observed DLR under cloud‐free conditions by 10–20 W m⁻² at this semiarid site on an annual basis, with a strong seasonal dependence. The calculated clear‐sky fluxes are overestimated during the warmer summer season, with large absolute values of DLR, while the biases are reduced in the colder and dryer winter season with smaller fluxes. This gives direct support for recent evidence that the DLR model biases depend systematically on the thermal and humidity structure of the cloudless atmosphere. Fluxes from strongly emitting atmospheres tend to be overestimated, but may be underestimated from atmospheres with smaller emission. This points to common problems inherent in the simulation of the emission from the cloudless atmosphere in current longwave radiation codes.
The comparisons of the all‐sky climatologies at Uardry show that the clear‐sky biases are partly masked in the models with an insufficient cloud‐radiative forcing, thereby counterbalancing the excessive DLR of the cloud‐free atmosphere. On the other hand, when the cloudradiative forcing is improved, the biases in the cloud‐free atmosphere become fully apparent in the all‐sky fluxes.     

基于GPU技术的风暴潮集合预报

台风预报的准确性在风暴潮预报中起着重要作用。台风强度和路径的不确定性意味着使用集合模式来预报风暴潮。本文利用中央气象台的最优路径台风参数驱动国家海洋环境预报中心业务化的水动力学模型，开展华南沿海的风暴潮模拟，模式模拟结果与实测吻合较好。为了改进计算效率，采用CUDA Fortran 语言对模型进行了改造，改造后的模型在计算结果与原模型基本一致的基础上，计算时间缩短了99%以上。通过融合欧洲中期天气预报中心（ECWMF）的50条路径与3种可能台风强度构造出了150个台风事件，并用150个台风事件驱动改进的风暴潮数值模型，计算结果可以提供集合预报产品和概率预报产品。通过“山竹”台风风暴潮过程可以发现集合平均预报结果和概率预报结果与实测吻合较好。改进的数值模型可以运行普通工作站上，非常适合风暴潮集合预报，并且可以提供更好的决策产品。     

基于多源数据的台风风暴潮概率预报研究:台风集合的构建

建立了一套用于台风风暴潮集合预报的台风集合构建方案。首先基于中国中央气象台、中国香港天文台、中国台湾中央气象局、美国联合台风预警中心、日本气象厅和韩国气象台6家预报中心的预报数据,构建一个误差更小的24 h、48 h和72 h预报时效的台风分析数据;然后基于分析数据构建9个路径样本(1条分析路径+2个概率圆上的8条概率路径)和3个台风最大风速(概率偏弱、居中和偏强)样本,形成27个台风样本集合,并根据分析风场的误差分布合理确定不同台风样本的发生概率。通过对台风"利奇马"的应用,证实该集合方案可以覆盖大部分可能的情景,集合样本具有较强的代表性,可用于台风风暴潮的集合预报。     

The long-term variability of sea surface temperature in the seas east of China in the past 40 a

The global surface temperature change since the mid-19th century has caused general concern and intensive study. However, long-term changes in the marginal seas, including the seas east of China, are not well understood because long-term observations are sparse and, even when they exist, they are over limited areas. Preliminary results on the long-term variability of sea surface temperature (SST) in summer and winter in the seas east of China during the period of 1957-2001 are reported using the Ocean Science Database of Institute of Oceanology, Chinese Academy of Sciences, the coastal hydrological station in situ and satellite data. The results show well-defined warming trends in the study area. However warming and cooling trends vary from decade to decade, with steady and rapid warming trends after the 1980s and complicated spatial patterns. The distribution of SST variation is intricate and more blurred in the areas far away from the Kuroshio system. Both historical and satellite data sets show significant warming trends after 1985. The warming trends are larger and spread to wider areas in winter than in summer, which means decrease in the seasonal cycle of SST probably linked with recently observed increase of the tropical zooplankton species in the region. Spatial structures of the SST trends are roughly consistent with the circulation pattern especially in winter when the meridional SST gradients are larger, suggesting that a horizontal advection may play an important role in the long-term SST variability in winter.     

A stochastic operational forecasting system of the Black Sea: Technique and validation

In this article, we present the latest version of an ensemble forecasting system of the hydrodynamics of the Black Sea, based on the GHER model. The system includes the Weakly Constrained Ensembles algorithm to generate random, but physically balanced perturbations to initialize members of the ensemble. On top of initial conditions, the ensemble accounts also for uncertainty on the atmospheric forcing fields, and on some scalar parameters such as river flows or model diffusion coefficients. The forecasting system also includes the Ocean Assimilation Kit, a sequential data assimilation package implementing the SEEK and Ensemble Kalman filters. A novel aspect of the forecasting system is that not only our best estimate of the future ocean state is provided, but also the associated error estimated from the ensemble of models. The primary goal of this paper is to quantitatively show that the ensemble variability is a good estimation of the model error, regardless of the magnitude of the forecast errors themselves. In order for this estimation to be meaningful, the model itself should also be well validated. Therefore, we describe the model validation against general circulation patterns. Some particular aspects critical for the Black Sea circulation are validated as well: the mixed layer depth and the shelfopen sea exchanges. The model forecasts are also compared with observed sea surface temperature, and errors are compared to those of another operational model as well.     

集合成员个数对集合预报技巧影响的试验分析

利用全球中期预报模式T63L9,选取2004年6月4日至13日10d作为试验个例进行了集合预报试验,分析了不同集合成员个数对于预报结果的影响。结果表明,集合预报的技巧都明显高于单个控制预报。在集合成员较少时,随集合成员教的增加,集合预报的技巧提高明显,当集合成员数多于11个时,集合预报的效果提高缓慢。在中期预报时段内。集合成员数11为集合预报效果随集合成员教趋于饱和的临界值,如果继续增加成员数.预报效果提高较少,但计算量却大大的增加。本文只是单个试验个例的分析结果。为验证结论的普适性,还需要进行更多的试验。     

Urban and ocean ensembles for improved meteorological and dispersion modelling of the coastal zone

A high-resolution (1.67 km) ensemble transform (ET)-based meso-scale modelling system utilizing urbanization and sea surface temperature (SST) perturbations is used to examine characteristics of sea breeze/heat island interactions and atmospheric transport and dispersion for Tokyo. The ensemble displays a positive spread–skill relationship, with the addition of urban perturbations enabling the ensemble variance to distinguish a larger range of forecast error variances. Two synoptic regimes are simulated. For a pre-frontal period (stronger synoptic flow), there is less variability among ensemble members in the strength of the urban heat island and its interaction with the sea breeze front. During the post-frontal time period, the sea breeze frontal position is very sensitive to the details of the urban representation, with horizontal frontal variation covering the width of the urban centre (∼30 km) and displaying significant impacts on the development and strength of the heat island. Moreover, the dosage values of a tracer released at offshore and urban sites have considerable variability among ensemble members in response to small-scale features such as coastally upwelled water, enhanced anthropogenic heating and variations in building heights. Realistic variations in SST (i.e. warm Tokyo Bay or local upwelling) produce subtle sea breeze variations that dramatically impact tracer distributions.