WORLD METEOROLOGICAL ORGANIZATION ________________ CBS/ICT/DPFS/Doc. 5(2) _______ (31.V.2002) CBS IMPLEMENTATION/COORDINATION TEAM ON DATA-PROCESSING AND FORECASTING SYSTEMS ITEM: 5 MOSCOW, RUSSIAN FEDERATION, 3-7 JUNE 2002 ENGLISH ONLY APPLICATION OF NWP TO SEVERE WEATHER FORECASTING (Submitted by J. Coiffier, Météo-France) Summary and purpose of document This document outlines what is currently termed as "severe weather events". Action proposed The meeting is invited to take into account the recommendations outlined in this document and to make proposals. CBS/ICT/DPFS/Doc. 5(2), p.2 SEVERE WEATHER FORECASTING 1 – Introduction The aim of this preliminary paper is to give a rapid survey about what are currently called “severe weather events.” In paragraph 2 we point out that this denomination of severe weather is often the result of a subjective assessment linked to the consequences of the weather rather than to the meteorological phenomenon itself. In paragraph 3 we try to classify severe weather events with their main characteristics and to give an idea about the current possibilities of the NWP models to forecast them. In paragraph 4 we propose an approach, which seems appropriate to take advantage of the existing meteorological centres to build a system able to deliver early warnings in case of severe weather. 2 - About the definition of severe weather The inspection of the web sites elaborated by numerous national meteorological services around the world shows that in spite of different appreciations from one country to the other one, the same meteorological phenomena are quoted as being responsible for severe weather. It is important to note that the qualification of severe weather often depends on the observed or experienced consequences; these consequences themselves depend on parameters external to the atmospheric machinery; thus heavy precipitation over a large inhabited land surface can remain unobserved while a lower quantity of precipitating water concentrated over a small basin can produce disastrous floods. Nevertheless one can try to classify the meteorological phenomena leading to severe weather as follows: tropical storms and tropical cyclones (hurricanes, typhoons); severe extratropical storms over ocean or over land; large scale heavy precipitations (rainfall or snowfall); active convective events with associated phenomena (heavy precipitations, hail, lightnings, gusts, tornadoes); persistence of extreme temperatures (cold wave or heat wave) phenomena lowering dramatically the visibility or holding up the circulation (fog, duststorm, blackice). A report by the National Climatic Data Centrer/NESDIS/ NOAA listing the significant global weather and climate events for the year 2000 shows that most of these phenomena and their consequences (floods, mudslides, drought conditions, fires) are responsible for number of fatalities, perishment of many livestock and disastrous effects on the property and way of life of millions of people. 3 - The main characteristics of the various phenomena 3.1 - Tropical storms and cyclones Tropical storms and cyclones are always responsible for many fatalities and economic loss every year and for this reason WMO established a global system in order to early detect such events and to issue forecasts about their trajectory and intensity. Although tropical cyclone forecasting remains out of the scope of this study it is interesting to examine the characteristics of this phenomenon in order to understand why the forecasting systems works rather satisfactorily and allow people to take appropriate protection measures. The birthplace of the tropical storms and the favourable period is well identified (low latitudes, SST beyond a given threshold). So each forecasting centre responsible for a specific geographical area tries to detect as soon as possible the tropical storms on the images provided by the stationary satellites and to follow their evolution. The tracking of the cyclones is made easier because their displacement is relatively slow. It is difficult to accurately forecast the trajectory of the tropical cyclones even with available fine mesh numerical CBS/ICT/DPFS/Doc. 5(2), p.3 models and a small location error can have important consequences (for a small size island for example). Nevertheless the combined use of numerical models, statistical adaptation and empirical approach allows to forecast a set of trajectories. The main difficulty arises from the lack of accurate measurements to define the initial state at the convenient scale and to the deficiencies of the parameterization of all the processes working inside a tropical cyclone. The implementation by WMO of Regional Specialized Meteorological Centres for tropical cyclone forecasting allows organizing the watch and insuring a valuable support to the National Meteorological Services. 3.2 - Severe extratropical storms over ocean and over land The main mechanism, which explains the cyclogenesis is the interaction between potential vorticity anomalies and baroclinic areas linked to the jet-streams. The FASTEX experiment, which took place in 1997 over the northeastern part of the Atlantic was especially devoted to study the formation and the evolution of the storms in this region. Severe storms are associated with the cyclogenesis phenomena at the midlatitudes and are characterized by very strong winds due to the intense pressure gradients around the low pressure system. Over the ocean these intense surface wind induce high waves, which are a real hazard for the ships. Over land very strong gusts are responsible for important damages to the forests, to the urban infrastructures and to aerial power transportation lines. Note also that, depending on the season, intense low pressure systems can be accompanied by with important precipitation of rain or snow. Typical cyclogenesis belong to synoptic meteorological phenomena, which can be detected by means of the available satellite images and meteorological observations, when they are located at the right place. Global or limited area models operated by many forecast centres of the National Meteorological Services with some success currently forecast them. Nevertheless it has to be stressed that the behaviour of the real atmosphere is often different of the simulated atmosphere, even when using the “best models”. On one hand several models very dynamically active have a worrying tendency to deepen some lows too much. On the other hand there are number of examples of important storms misforecast by numerical models (the 1999’s storms over northeast Atlantic after Christmas, for instance). One can say that the extratropical storms can be forecast with some success by using numerical models but it remains difficult to forecast accurately the time when the deepening becomes intense as well as the exact trajectory. Every important storm gives the opportunity to compare various models operated by the forecast centres in order to identify the possible sources of error. Recent studies carried out with the 1999’s storms case, show that the ability of the numerical models to forecast accurately the cyclogenesis does not depend on the size of the mesh (when using mesh size of the order of about 50 km) nor on the system of equations (hydrostatic versus non-hydrostatic). In contrast there is an important dependence on the initial state and the way to obtain it (selection of the observations, data assimilation procedures). These conclusions are also confirmed by number of experiments consisting in modifying model initial conditions in order to better fit several crucial observations or to improve the agreement between dynamical fields (potential vorticity) and satellite WV images. In order to obtain accurate forecasts it is necessary to define accurately the initial state, especially in the region where precursors of the storm can be identified. Unfortunately these regions are often far from dense observation networks. Improvement of the initial state comprises the development of new measurements in connection with efficient data assimilation systems. It is also important to mention the new concept of targeted observations, which consist in making measurements at the most sensible locations (which can be obtained from recent “adjoint techniques”). The early detection of storms is an important part of the current activity of the National Meteorological Services, which make use of numerical models. Nevertheless due to the uncertainties about the CBS/ICT/DPFS/Doc. 5(2), p.4 initial state, accurate deterministic prediction of the intense storms remains far of the perfection. 3.3 - Large scale heavy precipitations The precipitations are artificially divided into two categories according to the process that is responsible for the lifting of the moist air particles. One distinguishes the large-scale precipitations, which are associated to synoptic vertical velocities, and the convective precipitations, which take place within unstable layers of the atmosphere. Nevertheless the conceptual models, which describe the structure of the atmospheric perturbations, show that this separation is not always very clear. In the numerical models, as and when the resolved scales become smaller, convective circulations tend to be taken into account by the dynamics so that it is more appropriate to speak about resolved scale precipitation (which are taken into account by the dynamics) and unresolved scale precipitation resulting from the various parameterized moist physical processes. For sake of clarity one speaks here about heavy large-scale precipitations associated mainly with synoptic perturbations. Among the various ways to parametrize the processes responsible for the large scale precipitations, there are several levels of sophistication. Simple models only use one characteristic variable for the atmospheric water vapour and determine the precipitation amount by removing moisture to avoid supersaturation. The effects of condensation/evaporation are taken into account to calculate the corresponding heating/cooling; a simple relationship allows computing the cloud cover from the relative humidity. More sophisticated models deal with the three phases of the water in the atmosphere and describe the various interactions between them, according to microphysics processes; they are able to discriminate between precipitating liquid water and the small droplets remaining within the clouds. As the network of raingauges measures the total precipitation it is not possible to assess the ability of the numerical models to forecast large-scale precipitations only. The comparison between accumulated precipitation forecast by the models with the estimates deduced from the observation network exhibit generally similar patterns. Nevertheless day to day controls show that there can be many discrepancies, which can be caused by deficiencies in the evaluation of the various forcings on the resolved vertical velocity and in the calculation of the unresolved scale precipitation resulting mainly from the release of convective instability. Improvement of the quality of precipitation forecasts is expected to result from the use of a finer mesh allowing to better represent the orographic forcing and from a more accurate description of the microphysics processes. A correct initialization of the moisture content in the atmosphere is also essential for avoiding the spin-up effect, which deteriorates the quality of the very short-range forecasts. At present time forecasters cannot completely rely on precipitation fields given by the model to evaluate the expected amounts. They have to examine them in parallel with the dynamical fields to recognize the structures responsible for the forcing on the vertical velocity. It is important to note that the consequences of heavy precipitation in terms of floods do not only depend on the accumulated amount of water reaching the surface but also on the properties of the basin collecting the water (shape, soil water content). So it is clear that atmospheric numerical model have to be coupled with hydrological models in order to provide useful information. From the forecaster’s point of view it is necessary to be informed about the current state of the surface. In several cases a small increase of the precipitation can result in disastrous floods ; this is the reason why it is difficult to base a warning system on fixes threshold values. 3.4 - Convective events Convection is the main physical mechanism leading to the development of the severe thunderstorms observed in many places around the world, which are responsible of many fatalities and economic loss. Typical hazardous phenomena are associated with thunderstorms. CBS/ICT/DPFS/Doc. 5(2), p.5 Hail is at the origin of important damages not only for agricultural areas but also for urban areas when the diameter of the hailstones is larger than 2 cm. The formation of hail is very complex and depends on many facts such as the updraft strength, the wind structure and the cloud microphysics. This explains why accurate prediction of hail is difficult by using numerical models. Strong wind gusts are often associated with the thunderstorm can induce fall of trees and cause important damages to the habitations. They can occur in a large variety of convective situations and are generally associated with the cold air outflow as the downdraft reaches the surface. Flash floods are frequently associated with severe thunderstorms as a consequence of intense rainfall occurring at the same place during several hours and strongly depend on the nature of the surface (especially over saturated soils). Thus relatively minor rainfall can cause damaging flash flood. Flash flood is the phenomenon, which is responsible for important loss of life induced by strong convective events. Tornadoes are phenomena, which are associated with convective events, especially over several parts of the world (the Great Plain in Northern America for example). The very intense wind that takes place inside these small scale vortices and their sucking effect is the reason why the tornadoes are listed as the meteorological phenomena having the most destructive effect on life and property. They are often associated with supercell thunderstorms but can also occur in other convective events. The hazardous convective events are mesoscale phenomena, which are difficult to properly simulate with numerical models. Nevertheless numerical models provide relevant information about the synoptic environment propitious to the development of severe thunderstorms. Low tropopause anomalies, wet bulb potential temperature and wind convergence in the lowest layers, wind shear, helicity, convective available potential energy (CAPE) allow to know the regions where intense thunderstorms are likely to occur and to alert the services in charge of the public safety. The National Meteorological Services which implemented nowcasting systems based on satellite images, radar networks, lightning detection systems are the able to detect the location of growing up convective cells and to try to extrapolate their displacement one or two hours ahead and to calculate estimated water accumulation at the surface. It can be expected that the progresses in numerical modelling (with horizontal resolution of the order of 1 kilometer) and the improvements in the simulation of cloud microphysics will allow better simulating the life cycle of the thunderstorms. Nevertheless it is important to point out that numerical models start from an initial state, which has to be carefully analyzed especially for the small-scale distribution of humidity and liquid water.ppp 3.5 - Extended periods of extreme temperature Among the climate events cold waves and heat waves in many places around the world are regularly reported. Cold waves and heat waves are characterized by the persistence during several days of extreme temperatures going beyond the climatological values. When such events take place in regions where the population, do not use to cope with such conditions, they can cause many fatalities. It is important to note also that prolonged periods of heat wave are propitious to the increase of tropospheric ozone concentration in urban areas and to propagation fire over forested areas. The origin of such events is linked with the global atmospheric circulation and the distribution of the large-scale pressure centres around the Earth. It is recognized that the interannual perturbations of the atmospheric circulation like the ENSO in southern Pacific or the NAO for the northern Atlantic have an important impact on the occurrence of such events in several parts of the globe (So negative values of the NAO index are associated with rather cold winters over Europe, while active El Niño induces heat waves over south-east Asia and northern America. These large-scale phenomena are rather well predicted by the current global models several days in advance. Even seasonal forecast have some skill for predicting these kinds of events (the last El Niño forecast in 1998, for example). Nevertheless despite of good meteorological forecasts, it seems that is difficult to determine precisely the time CBS/ICT/DPFS/Doc. 5(2), p.6 when the situation becomes really dangerous for the human life if protection means are lacking. Forecasting such events clearly lies in the realm of the possibilities of meteorological centres operating global models for medium range or seasonal outlook purposes. 3.6 - Other events affecting the economical life There are other meteorological phenomena, which are able to disturb the economical life of a country: reduced visibility due to sand storms or dense generalized fog can considerably hinder the communication by air and on the roads. Similarly the occurrence of blackice over large parts of a country can be responsible of serious accidents. This kind of events are not always easy to forecast with help of numerical models because their formation and their disappearing depend on threshold effect. Nevertheless if such events can have temporarily important consequences for the economical life they represent a less threatening danger than severe thunderstorms. 4 - How to use numerical models for severe weather forecasting. After this rapid survey of the severe weather phenomena events with their main characteristics and the efficiency of the numerical models to forecast them one can formulate the following points. NWP global models are generally able to forecast large-scale circulation of the atmosphere and the position of the pressure centres responsible for the establishment during several days of airflows leading to cold waves during the winter (resp. heat wave during the winter). NWP models are able to produce realistic lifecycle of extratropical cyclones leading to strong wind or/and heavy precipitations but in the details (trajectory, chronology) the reality often differ from the numerical simulation because of the incomplete knowledge of the initial state. So it is risky to want to directly use only one model output in order to issue warnings. NWP gives important information about the synoptic environment favourable for the development of severe thunderstorms but are unable to give with accuracy the right location and intensity of the most intense phenomena. The development of warning systems for severe convective storms is essentially based on nowcasting techniques involving radar networks and lightning detection systems. The example of the United States where such systems have been developed show that they allow to better inform the public thus to reduce the number of fatalities. If mesoscale models are able to simulate the behaviour of severe local storms, it is not possible at the moment to use them as operational tools for several reasons. efficient mesoscale models require a very important computing powers (they are likely to deal with the same number of degrees of freedom as larger scale global models) ; dynamic and moisture variable have to be carefully initialised by using mesoscale measurements collected as rapidly as possible ; finally, one can expect that mesoscale model integrations will be very sensitive to the initial conditions so that it will be too restrictive to trust in purely deterministic forecasts. Taking into account the fact that accurate forecasts of severe weather events are difficult to achieve with help of individual models it is essential to try to take advantage of the variety of NWP models, which are currently operated by number of National Meteorological Centres. The realization of ensemble forecasts and the production of probabilistic forecasts is the best way to take into account the uncertainty about the specification of the initial state and to reap the benefits from the differences between the numerical models. It is important to note that ECMWF developed by means of CBS/ICT/DPFS/Doc. 5(2), p.7 the Ensemble Prediction System (EFS) a “severe weather forecast index” (by measuring a distance between the distribution function of the forecasts with the climatological) which is currently evaluated by forecasters. Such an approach would be quite convenient in the view of the implementation of a coordinated system aiming to issue pre-warnings for severe weather events. ________________________