User Guide to ECMWF forecast products - [PDF Document] (2024)

  • 7/23/2019 User Guide to ECMWF forecast products

    1/129

    Location: Livelink 4320059 Owner: Erik Andersson

    Date: 23/07/2013 Version 1.1

    User guide to ECMWF forecast

    products

    Copyright 2011, 2013 ECMWF

  • 7/23/2019 User Guide to ECMWF forecast products

    2/129

    2Location: Livelink 4320059 Owner: Erik Andersson

    Date: 23/07/2013 Version 1.1, Page 2of 129

    Date of original issue: October 2011

    Author: Anders Persson

    Version number Date Changed by Change description

    1.1 23/07/2013 Erik Andersson New terminology, ENS

    initial perturbations

  • 7/23/2019 User Guide to ECMWF forecast products

    3/129

    iLocation: Livelink 4320059 Owner: Erik Andersson

    Date: 23/07/2013 Version 1.1, Page iof 129

    1.Introduction..............................................................................................................................1

    2. The ECMWF forecasting and assimilation system...............................................................2

    2.1. The ECMWF global atmospheric model.......................................................................2

    2.1.1. The modelequations.....................................................................................2

    2.1.2. The numerical formulation..........................................................................2

    2.1.3. The rationale for high resolution.................................................................3

    2.1.4. Topographical and climatological fields...................................................... 3

    2.1.5. The formulation of physical processes......................................................... 4

    2.1.6. The land surface model................................................................................6

    2.1.7. The ocean wave model..................................................................................6

    2.2. The dynamic ocean model..............................................................................................7

    2.3.

    Data assimilation.............................................................................................................7

    2.3.1. The four-dimensional data assimilation (4D-Var)....................................... 8

    2.3.2. The ECMWF early delivery system............................................................. 8

    2.4. Retrieving ECMWF forecasts......................................................................................10

    2.4.1. Temporal retrieval.....................................................................................10

    2.4.2. Spatial retrieval..........................................................................................10

    2.4.3. Orography..................................................................................................10

    2.4.4. The bi-linear interpolation.........................................................................10

    2.4.5. The subsampling procedure.......................................................................12

    2.4.6. Interpolating land and seapoints...............................................................13

    2.5. The relation between grid point values andobservations........................................... 15

    2.6. Some characteristics of NWP output...........................................................................16

    2.6.1. Forecast error growth................................................................................16

    2.6.2. Downstream spread of influence................................................................16

    2.6.3. The relation between scale and predictive skill......................................... 17

    2.6.4. Forecast jumpiness.................................................................................20

    2.6.5. Flip-flopping forecasts................................................................................21

    2.6.6.

    Jumpiness and forecastskill.......................................................................22

    2.6.7. Forecast trends cannot be extrapolated..................................................... 22

    2.6.8. Other state-of-the-art deterministic models.............................................. 22

    3. The forecast ensemble...........................................................................................................25

    3.1. The rationale behind theensemble...............................................................................25

    3.1.1. Qualitative use of the ensemble..................................................................25

    3.1.2. Quantitative use of the ENS.......................................................................25

    3.1.3. Characteristics of a good ensemble............................................................ 26

    3.2.

    Generation of the ENS..................................................................................................26

  • 7/23/2019 User Guide to ECMWF forecast products

    4/129

    iiLocation: Livelink 4320059 Owner: Erik Andersson

    Date: 23/07/2013 Version 1.1, Page iiof 129

    3.2.1. Different perturbation techniques.............................................................26

    3.2.2. Quality of the individual perturbed analyses............................................ 29

    3.2.3. Quality of the individual perturbedforecasts............................................ 31

    3.3.

    The ensemble at different lead times...........................................................................32

    3.3.1. The 10-day range........................................................................................32

    3.3.2. The day 9 to 10 overlap..............................................................................33

    3.3.3. The 10 to 15 day range...............................................................................33

    3.3.4. Forecasts from 15 to 32 days......................................................................33

    3.3.5. Seasonal forecast........................................................................................33

    3.4. Basic forecast products.................................................................................................33

    3.4.1. Postage stampmaps................................................................................33

    3.4.2. Spaghetti diagrams.................................................................................34

    3.4.3.

    Plumes.....................................................................................................34

    3.4.4. Ensemble mean andmedian.......................................................................35

    3.4.5. Ensemblespread.........................................................................................35

    3.4.6. Probabilities................................................................................................36

    3.4.7. Forecast expressed in terms of intervals.................................................... 36

    4. Recommendations on categorical and probabilistic medium-rangeforecasting ............... 37

    4.1. Relation between deterministic and probabilistic forecasts....................................... 37

    4.2. Differences between short- and medium-range operational useof NWP ................. 37

    4.3.

    Medium-range forecasting without theensemble.......................................................38

    4.3.1. Assessment based on the latest forecasts................................................... 38

    4.3.2. Assessment based on the two latestforecasts............................................. 38

    4.3.3. Assessment based on the last three or more forecasts............................... 38

    4.3.4. Is it possible to compare manual and computer-generateddeterministic

    forecasts?...................................................................................................................39

    4.4. Medium-range forecasting based only on theENS..................................................... 40

    4.4.1. Use of the ensemble mean (EM).................................................................40

    4.4.2. Criticism of the EM....................................................................................40

    4.4.3.

    A synoptic example of combining EM and probabilities........................... 40

    4.4.4. Use of probabilities.....................................................................................42

    4.4.5. Probabilities over time intervals................................................................43

    4.4.6. Probabilities over areas..............................................................................44

    4.4.7. Probabilities of combinedevents................................................................44

    4.4.8. Modification of the probabilities................................................................45

    4.4.9. Calibration of probabilities........................................................................45

    4.4.10. Ensemble jumpiness...............................................................................45

    4.5.

    Medium-range forecasting with the ENS andHRES................................................. 46

  • 7/23/2019 User Guide to ECMWF forecast products

    5/129

    iiiLocation: Livelink 4320059 Owner: Erik Andersson

    Date: 23/07/2013 Version 1.1, Page iiiof 129

    4.5.1. Weather situations with good agreement between ENS andHRES ......... 46

    4.5.2. When the ENS and HRES differ with respect to spread only.................. 47

    4.5.3. Weather situations where agreement between the ENS andHRES is poor

    48

    4.5.4. Forecaster intervention with theENS........................................................ 51

    4.6. Forecasting high-impact weather in the medium range.............................................. 52

    4.6.1. The forecasters role...................................................................................52

    4.6.2. Probabilities or categorical forecasts?....................................................... 52

    4.7. Summary: do the opposite to the computer!...............................................................53

    5. Derived products based on the ENS.....................................................................................55

    5.1.1. Ensemble mean and spread charts............................................................ 55

    5.2.EPSgrams.......................................................................................................................55

    5.2.1.

    Overview.....................................................................................................55

    5.2.2. Ten-day EPSgrams.....................................................................................56

    5.2.3. Fifteen-day EPSgrams................................................................................57

    5.2.4. The weather parameters in EPSgrams...................................................... 57

    5.2.5. Interpreting EPSgrams..............................................................................59

    5.3. Wave EPSgrams............................................................................................................61

    5.4. The Extreme Forecast Index(EFI)...............................................................................63

    5.4.1. The EFI reference climate..........................................................................63

    5.4.2. The cumulative distribution function........................................................ 63

    5.4.3. Calculating the EFI....................................................................................65

    5.4.4. The interpretation of the EFI.....................................................................66

    5.4.5. EFI maps....................................................................................................67

    5.5. Tropical cyclone diagrams...........................................................................................67

    5.6. Cyclone track maps.......................................................................................................70

    5.7.Clustering.......................................................................................................................71

    5.7.1. Weather scenario clustering.......................................................................72

    5.7.2. Climatological weather regimes.................................................................73

    5.7.3.

    Tubing.........................................................................................................74

    6. Epilogue: how to increase the publics trust in medium-rangeweather forecasts ............ 75

    6.1. How can trust in medium-range forecasts be increased?........................................... 75

    6.1.1. Improving the forecast system...................................................................75

    6.1.2. Trust in individual forecasts......................................................................75

    6.1.3. When the deterministic forecast cannot be trusted................................... 75

    6.2. The role of the forecaster in the medium-range.......................................................... 76

    6.3. How the forecaster can addvalue.............................................................................76

    Appendix A

    Some statistical concepts to facilitate the use andinterpretation of deterministicmedium-range forecasts................................................................................................................77

  • 7/23/2019 User Guide to ECMWF forecast products

    6/129

    ivLocation: Livelink 4320059 Owner: Erik Andersson

    Date: 23/07/2013 Version 1.1, Page ivof 129

    Introduction................................................................................................................................77

    A-1 Forecastvalidation......................................................................................................77

    A-1.1 The mean error...........................................................................................77

    A-1.2

    Forecastvariability.....................................................................................78

    A-1.3 False systematic errors...............................................................................79

    A-1.4 False model climate drift............................................................................80

    A-2 Forecastverification...................................................................................................81

    A-2.1 Measures of accuracy.................................................................................81

    A-2.2 The effect of mean, analysis and observation errors on theRMSE .......... 81

    A-2.3 The decomposition of MSE........................................................................82

    A-2.4 Forecast error baseline...............................................................................82

    A-2.5 Error saturation level.................................................................................83

    A-2.6

    Measure of skill - the anomaly correlation coefficient.............................. 83

    A-3 Interpretation of verificationstatistics....................................................................84

    A-3.1 Interpretation of RMSE and ACC.............................................................84

    A-3.2 Effect of flow dependency..........................................................................84

    A-3.3 The double penalty effect.......................................................................85

    A-3.4 Subjective evaluations................................................................................85

    A-4 Graphicalrepresentation...........................................................................................85

    A-4.1 Forecast errors...........................................................................................86

    A-4.2

    Flow dependence........................................................................................87

    A-4.3 Damping of forecast anomalies..................................................................88

    A-4.4 Forecast error correlation..........................................................................88

    A-4.5 Forecast jumpiness and forecast skill........................................................ 89

    A-4.6 Combining forecasts...................................................................................89

    A-5 The usefulness of statisticalknow-how....................................................................90

    A-6 Utilityverification.......................................................................................................91

    A-6.1 The contingency table.................................................................................91

    A-6.2 The expected expenses............................................................................91

    A-7 Practicalexamples......................................................................................................92

    A-7.1 A situation with no weather forecast service............................................. 92

    A-7.2 The benefit of a local weather service........................................................ 93

    A-7.3 The establishment of two new weather services........................................ 93

    A-8 An introduction to probabilistic weatherforecasting........................................... 95

    A-8.1 Uncertainty - how to turn a disadvantage into an advantage................... 95

    A-8.2 Making even more use of uncertainty - probabilities................................ 96

    A-8.3 Towards more useful weather forecasts.................................................... 98

    A-8.4

    Quality of probabilistic forecasts...............................................................98

  • 7/23/2019 User Guide to ECMWF forecast products

    7/129

    vLocation: Livelink 4320059 Owner: Erik Andersson

    Date: 23/07/2013 Version 1.1, Page vof 129

    A-8.5 When probabilities are not required......................................................... 98

    A-8.6 An extension of the contingency table the SEEPS score..................... 99

    Appendix B Some statistical concepts to facilitate the use andinterpretation of ensemble

    forecasts 101

    Introduction..............................................................................................................................101

    B-1 The reliabilitydiagram............................................................................................101

    B-1.1 Reliability.................................................................................................103

    B-1.2 Sharpness..................................................................................................103

    B-1.3 Under- and overconfident probability forecasts...................................... 104

    B-2 Rank histogram (Talagranddiagram)..................................................................105

    B-3 Verificationmeasures...............................................................................................107

    B-3.1 The Brier score - the MSE of probability forecasts................................. 107

    B-3.2

    Decomposition of the Brier score.............................................................107

    B-3.3 The Brier score is a proper score......................................................... 109

    B-3.4 The Brier skill score.................................................................................109

    B-3.5 The rank probability score (RPS)............................................................109

    B-4 The relative operating characteristics (ROC)diagram......................................109

    B-5 Calibration ofprobabilities.....................................................................................111

    B-6 Statistical post-processing model outputstatistics............................................113

    B-6.1 The MOS equation...................................................................................113

    B-6.2 Simultaneous corrections of mean error and variability........................ 113

    B-6.3 Short-range MOS.....................................................................................114

    B-6.4 Medium-range MOS................................................................................114

    B-6.5 Adaptive MOS methods...........................................................................115

    References and furtherliterature.................................................................................................119

  • 7/23/2019 User Guide to ECMWF forecast products

    8/129

  • 7/23/2019 User Guide to ECMWF forecast products

    9/129

    1. Introduction

    1Location: Livelink 4320059 Owner: Erik Andersson

    Date: 23/07/2013 Version 1.1, Page 1of 129

    1. Introduction

    Behind good forecast practices are often hidden goodtheories;

    equally, good theories should provide a basis for goodforecastpractices. Professor Tor Bergeron, personal communication1974

    The aim of this User Guide is to help meteorologists makeoptimal use of the forecast

    products from ECMWF, develop new products and reach new sectorsof society and thereby

    satisfy new demands. This is done by presenting the forecastsystem and advising on how best

    to use the output, not least how to build up trust in theforecast information. The emphasis is

    on the medium-range forecast products, since the way forecastersdeal with medium-range

    NWP output differs in many ways from how they deal withshort-range NWP on the one hand

    and monthly and seasonal NWP on the other. The main outline:

    1. The ECMWF forecasting system, i.e. the dynamical model, thedata assimilation and

    the product delivery system, are described in broad andnon-technical terms.

    2. The interpretation of the NWP output is complicated by itsoften counter-intuitive,

    non-linear behaviour. The high-resolution forecast (HRES) shouldtherefore not be

    over-interpreted, in particular not in the medium-range or whenextreme weather is

    likely. Then the use of probabilities or other risk assessmentsare needed.

    3. A good forecast that is not trusted is a worthless forecast.The ECMWF forecast

    ensemble (ENS), which is given extensive coverage, provides abasis for formulating

    the most accurate categorical forecasts and the probabilities ofalternative

    developments. Methods to combine HRES and ENS are suggested.

    4. In the medium-range the use of statistical know-how counts asmuch as synoptic

    experience, since daily operational work is to a large extent amatter of assessing,

    combining and correcting NWP information. In two appendicesstatistical concepts

    for validating and verifying deterministic and probabilisticforecasts and for making

    the best use of NWP information are presented.

    5. The forecaster is not a computer. Throughout the User Guideforecasters are advised

    not to try to imitate NWP, but to perform quite differently,with fewer details, more

    uncertainty and no U-turns.

    This User Guide is the fruit of several years of discussionswith scientists, forecasters and

    meteorologists who are interested in statistics, both fromEurope and elsewhere. The

    interaction between these three specialized groups has been themain driving force and

    inspiration for this publication.

    The User Guide gives only an introduction to the forecastinformation provided by ECMWF.

    Users are advised to keep themselves updated about the productsthrough the ECMWF

    Newsletter and web site.

  • 7/23/2019 User Guide to ECMWF forecast products

    10/129

    2. The ECMWF deterministic forecasting system

    2Location: Livelink 4320059 Owner: Erik Andersson

    Date: 23/07/2013 Version 1.1, Page 2of 129

    2. The ECMWF forecasting and assimilation system

    The ECMWF forecasting system (the IFS) consists of severalcomponents: an atmosphericgeneral circulation model, an ocean wavemodel, a land surface model, an ocean general

    circulation model and perturbation models for the dataassimilation (EDA) and forecast (ENS)

    ensembles (see Chapter3), producing forecasts from days to weeksand months ahead.

    2.1.

    The ECMWF global atmospheric model

    The atmospheric general circulation model describes thedynamical evolution on the resolved

    scale and is augmented by the physical parameterisation,describing the mean effect of sub-

    grid processes and the land-surface model. Coupled to this is anocean wave model (Bechtold

    et al, 2008).

    2.1.1. The model equations

    The model formulation is based on a set of basic equations, ofwhich some are diagnostic and

    describe the static relationship between pressure, density,temperature and height, and some

    are prognostic and describe the time evolution of the horizontalwind components, surface

    pressure, temperature and the water vapour contents of an airparcel.

    Additional equations describe changes in the hydrometeors (rain,snow, liquid water, cloud

    ice content etc). There are options for passive tracers such asozone. The processes of

    radiation, gravity wave drag, vertical turbulence, convection,clouds and surface interaction

    are, due to their relatively small scales (unresolved by themodels resolution), described in a

    statistical way asparametrization processes (arranged inentirely vertical columns).

    2.1.2. The numerical formulation

    The model equations are discretized in space and time and solvednumerically by a semi-

    Lagrangian advection scheme. It ensures stability and accuracy,while using as large time-

    steps as possible to progress the computation of the forecastwithin an acceptable time.

    For the horizontal representation a dual representation ofspectral components and grid

    points is used. All fields are described in grid point space.Due to the convergence of the

    meridians, computational time can be saved by applying a reducedGaussian grid. This

    keeps the east-west separation between points almost constant bygradually decreasing thenumber of grid points towards the poles atevery latitude in the extra-tropics. For the

    convenience of computing horizontal derivatives and tofacilitate the time-stepping scheme, a

    spectral representation, based on a series expansion ofspherical harmonics, is used for a

    subset of the prognostic variables.

    The vertical resolution is finest in geometric height in theplanetary boundary layer and

    coarsest near the model top. The -levels follow the earthssurface in the lower-most

    troposphere, where the Earths orography displays largevariations. In the upper stratosphere

    and lower mesosphere they are surfaces of constant pressure witha smooth transition in

    between.

  • 7/23/2019 User Guide to ECMWF forecast products

    11/129

    2. The ECMWF deterministic forecasting system

    3Location: Livelink 4320059 Owner: Erik Andersson

    Date: 23/07/2013 Version 1.1, Page 3of 129

    2.1.3. The rationale for high resolution

    The higher the numerical resolution, the more accurate thecalculations become. A high

    spatial resolution also enables a better representation oftopographical fields, such as

    mountains and coastlines, and the effect they have on thelarge-scale flow. It also produces amore accurate description ofhorizontal and vertical structures, which facilitates the

    assimilation of observations.

    The smallest atmospheric features which can be resolved byhigh-resolution forecasts have

    wave lengths four or five times the numerical resolution.Although these atmospheric systems

    have a predictability of only some hours, which is about thetime it takes to disseminate the

    forecasts, their representation is nevertheless important forenergetic exchanges between

    different atmospheric scales.

    Increasing the resolution not only benefits the analyses andforecasts of the small-scale

    systems associated with severe weather but also those oflarge-scale systems. The abilityaccurately to forecast theformation of large-scale blocking omega anticyclones and cut-

    off lows depends crucially on increasing the resolution tokilometres (Miller et al, 2010).

    The interpolation technique used when forecasts are retrieved ispresented in section2.4.

    2.1.4. Topographical and climatological fields

    The model orography is derived from a data set with a resolutionof about 1 km which

    contains values of the mean elevation above the mean sea level,the fraction of land and the

    fractional cover of different vegetation types. This detaileddata is aggregated (upscaled) to

    the coarser model resolution.

    The resulting mean orographycontains the values of the meanelevation above the mean sea

    level. In mountainous areas it is supplemented by sub-gridorographic fields, to enable the

    parametrization of the effects of gravity waves and provideflow-dependent blocking of the

    air flow. For example, cold air drainage in valleys makes thecold air effectively lift the

    orography.

    The land-sea mask is a geographical field that contains thepercentage of land and water

    between 0 (100% sea) and 1 (100% land) for every grid point. Agrid point is defined as a

    land point if its value indicates that more than 50% of the areawithin the grid-box is covered

    by land, see section2.4.6.

    The albedo is determined by a combination of background monthlyclimate fields and

    forecast surface fields (e.g. from snow depth). Continental,maritime, urban and desert

    aerosols are provided as monthly means from data bases derivedfrom transport models

    covering both the troposphere and the stratosphere.

    Soil temperatures and moisture in the groundare prognosticvariables. There is a lack of

    observational data, so observed 2m temperature and relativehumidity act as very efficient

    proxy data for the analysis.

    The snow coveragedepthis analysed every six hours fromsnow-depth observations, satellite

    snow extent and a snow-depth background field. The snowtemperature is also analysed from

    satellite observations. They are forecast variables.

  • 7/23/2019 User Guide to ECMWF forecast products

    12/129

    2. The ECMWF deterministic forecasting system

    4Location: Livelink 4320059 Owner: Erik Andersson

    Date: 23/07/2013 Version 1.1, Page 4of 129

    Sea surface temperature (SST) and ice concentrationare based onanalyses received daily

    from the Met Office (OSTIA, 5 km). It is updated during themodel integration, according to

    the tendency obtained from climatology.

    The temperature at the ice surface is variable and calculatedaccording to a simple energy

    balance/heat budget scheme, where the SST of the underlyingocean is assumed to be -1.7C.

    The sea-ice cover, which is kept constant in the 10-day forecastintegration, is relaxed

    towards climatology between days 10 and 30, with a linearregression. Beyond day 30 the

    sea-ice concentration is based on climatological values only(from the ERA 1979-2001 data).

    2.1.5. The formulation of physical processes

    The effect of sub-scale physical processes on weather systems isexpressed in terms of

    resolved model variables in a technique called parametrization.It involves both statistical

    methods and simplified mathematical-physical models, such asadjustment processes. So, forexample, the air closest to the earthssurface exchanges heat with the surface through

    turbulent diffusion or convection, which adjusts unstable airtowards neutral stability (Jung et

    al, 2010).

    The convection scheme does not predict individual convectiveclouds, only their physical

    effect on the surrounding atmosphere, in terms of latent heatrelease, precipitation and the

    associated transport of moisture and momentum. The schemedifferentiates between deep,

    shallow and mid-level convection. Only one type of convectioncan occur at any given grid

    point at one time (seeFigure 1).

    Figure 1: The ECMWF total convective rainfall forecast from 28November 2010 12 UTC + 30h. The

    convection scheme has difficulty in advecting wintery showersinland over Scotland and northernEngland from the relatively warmNorth Sea. The convection scheme is diagnostic and works on a

    model column, so cannot produce large amounts of precipitationover the relatively dry and cold

  • 7/23/2019 User Guide to ECMWF forecast products

    13/129

    2. The ECMWF deterministic forecasting system

    5Location: Livelink 4320059 Owner: Erik Andersson

    Date: 23/07/2013 Version 1.1, Page 5of 129

    (stable) wintery land areas. In nature these showers succeed inpenetrating inland through a

    convectively induced upper-level warm anomaly leading tolarge-scale lifting and saturation.

    Clouds, both convective and non-convective, are handled byexplicit equations for cloud

    water, ice and cloud cover. Liquid and frozen precipitation arestrongly coupled to other

    parametrized processes, in particular the convective scheme andthe radiation. The scheme

    also takes into account important cloud processes, such ascloud-top entrainment and the

    evaporation of water. Fog is represented in the scheme as cloudsthat form in the lowest

    model level.

    The radiation spectrum is divided into a long-wave part(thermal) and a short-wave part

    (solar radiation). Since it has to take the cloud-radiationinteraction into account in

    considerable detail, it makes use of a cloud-overlap algorithm,which calculates the relative

    placement of clouds across levels. For the sake of computationalefficiency, the radiation

    scheme is called less frequently than the model time step on areduced grid. Nevertheless, it

    accounts for a considerable fraction of the total computationaltime.

    For the precipitation and hydrological cycles both convectiveand stratiform precipitation

    are included in the ECMWF model. Evaporation of theprecipitation, before it reaches the

    ground, is assumed not to take place within the cloud, only inthe cloud-free, non-saturated air

    beside or below the model clouds. The melting of falling snowoccurs in a thin layer of a few

    hundred metres below the freezing level. It is assumed that snowcan melt in each layer,

    whenever the temperature exceeds 0C. The cloud-overlap algorithmis also important for the

    life history of falling precipitation: from level-with-cloud tolevel-with-clear-sky and vice

    versa.

    The near-surface wind forecastdisplays severe weaknesses in somemountain areas, due to

    the difficulty in parametrizing the interaction between the airflow and the highly varying sub-

    grid orography (seeFigure 2). As with many other sub-grid-scalephysical processes that need

    to be treated in simplified ways, this problem will ultimatelybe reduced when the air-surface

    interaction can be described explicitly, thanks to a higher andappropriate resolution. The

    system also produces wind-gust forecasts as part ofpost-processing (Balsamo et al, 2011).

  • 7/23/2019 User Guide to ECMWF forecast products

    14/129

    2. The ECMWF deterministic forecasting system

    6Location: Livelink 4320059 Owner: Erik Andersson

    Date: 23/07/2013 Version 1.1, Page 6of 129

    Figure 2: MSLP and 10 m wind forecast from 2 March 2011 12 UTC +12 h. The 10 m winds are

    unrealistically weak over the rugged Norwegian mountains. Valuesof 10 m/s might be realistic in

    sheltered valleys, but not on exposed mountain ranges.

    2.1.6. The land surface model

    In the H-TESSEL scheme (Hydrology-Tiled ECMWF Scheme for SurfaceExchange overLand) the main types of natural surfaces found overland are represented by a "mosaic"

    approach. In other words, each atmospheric model grid-box is incontact and exchanges

    energy and water with up to 6 different types of parcel or"tile" on the ground. These are: bare

    soil, low and high vegetation, water intercepted by leaves, andshaded and exposed snow.

    Each land-surface tile has its own properties, describing theheat, water and momentum

    exchanges with the atmosphere; particular attention is paid toevaporation, as near-surface

    temperature and humidity are very closely related to thisprocess.

    The soil (with its four layers) and the snow-pack (with onelayer) have dedicated physical

    parametrizations, since they represent the main land reservoirsthat can store water and energy

    and release them into the atmosphere in lagged mode.

    Finally, the vegetation seasonality is described by the leafarea index (LAI) from

    climatalogical data. The LAI describes the growing, mature,senescent and dormant phases of

    several vegetation types in H-TESSEL (four types of forests andten types of low vegetation).

    2.1.7. The ocean wave model

    The wave model at ECMWF is called the WAM. It describes the rateof change of the 2-

    dimensional wave spectrum, in any water depth, caused byadvection, wind input, dissipation

    due to white capping and bottom friction and non-linear waveinteractions. It is set up so as

    to allow the two-way interaction of wind and waves with theatmospheric model. It is also

    incorporated in the medium-range, monthly and seasonalensembles.

  • 7/23/2019 User Guide to ECMWF forecast products

    15/129

    2. The ECMWF deterministic forecasting system

    7Location: Livelink 4320059 Owner: Erik Andersson

    Date: 23/07/2013 Version 1.1, Page 7of 129

    Radar altimeter wave-height data are assimilated fromsatellites. Buoy wave data are not

    assimilated; instead, they serve as an independent check on thequality of modelled wave

    parameters. The propagation of swell in the wave model ishandled by a simple scheme that

    gives rise to a smoothing of the wave field. At present theeffects of surface currents on the

    sea state are not taken into account. In particular areas, suchas the Gulf Stream or Agulhas

    current, the current effect may give rise to localised changesof up to one metre in the wave

    height.

    The representation of the sea-ice fields is not as accurate aswould be needed to handle waves

    near the ice edge. Due to the present model resolution, waveproducts near the coasts and, to a

    lesser extent, in small enclosed basins (e.g. the Baltic Sea)may be of lower quality than the

    open-ocean products.

    2.2. The dynamic ocean model

    The three-dimensional general circulation ocean model canreproduce the general features ofthe circulation and the thermalstructure of the upper layers of the ocean and its seasonal and

    inter-annual variations. It has, however, systematic errors,some of which are caused by the

    coarse vertical and horizontal resolution: the model thermoclineis too diffuse; the Gulf

    Stream does not separate at the right location.

    The ocean analysis is performed every 10 days, down to a depthof 2000 m. Observational

    input comes from all around the globe, but mostly from thetropical Pacific, the tropical

    Atlantic and, to an increasing degree, from the Indian Ocean. Inplaces where the ocean floor

    is below 2000 m the information from above 2000 m is propagateddownwards by

    statistical vertical influence functions, similar to those inthe atmospheric data assimilation.The ocean-atmosphere couplingisachieved by a two-way interaction: the atmosphere affects

    the ocean through its wind, heat and net precipitation(precipitation-evaporation), whilst the

    ocean affects the atmosphere through its SST.

    For the seasonal forecasts the interaction is once a day, whilefor the ENS it is every hour.

    This high-frequency coupling may have some positive impact onthe development of some

    synoptic-scale systems, such as tropical cyclones.

    2.3. Data assimilation

    The observations used for the analysis of the atmosphere can bedivided roughly into

    conventional, in-situ observations and non-conventional,remote-sensing observations.

    The conventional observations consist of direct observationsfrom surface weather stations,

    ships, buoys, radiosonde stations and aircraft, both at synopticand, increasingly, at asynoptic

    hours. All surface and mean sea-level-pressure observations areused, with the exception of

    cloud cover, 2 m temperature and wind speed (over land). 2 mtemperature and dew point

    observations are used in the analysis of soil moisture. Observedwinds are used from ships

    and buoys but not from land stations, not even from islands orcoastal stations.

    The non-conventional observationsare achieved in two differentways: passive technologies

    sense natural radiation emitted by the earth and atmosphere orsolar radiation reflected by the

    earth and atmosphere; active technologies transmit radiation andthen sense how much is

  • 7/23/2019 User Guide to ECMWF forecast products

    16/129

    2. The ECMWF deterministic forecasting system

    8Location: Livelink 4320059 Owner: Erik Andersson

    Date: 23/07/2013 Version 1.1, Page 8of 129

    reflected or scattered back. In this way surface-wind vectorinformation is, for example,

    derived from the influence of the ocean capillary waves on theback-scattered radar signal of

    scatterometer instruments (Hersbach and Janssen, 2007).

    2.3.1. The four-dimensional data assimilation (4D-Var)

    The increasing availability of asynoptic data andnon-conventional observations has

    necessitated the use of advanced analysis procedures, such asfour-dimensional variational

    data assimilation (4D-Var), where the concept of a continuousfeedback between observations

    and model data is put on a firm mathematical foundation(Andersson and Thpaut, 2008).

    The 4D-Var analysis uses the model dynamics and physics tocreate, over a time window,

    (currently 12 hours), a sequence of model states that fits asclosely as possible with the

    available observations and background, i.e. a short-rangeforecast that serves to bring the

    information forward from the previous cycle. These states areconsistent with the dynamics

    and physics of the atmosphere, as expressed by the equations ofthe model. The correction of

    one model variable generates physically and dynamicallyconsistent corrections of other

    variables. For instance, a sequence of observations of humidityfrom a satellite infrared

    instrument that shows a displacement of atmospheric structureswill entail a correction not

    only of the moisture field but also of the wind and temperaturefields.

    The impact of the observations is determined by the assumedaccuracy of the observations

    and the model short-range forecasts. While the former can beregarded as more or less static,

    the latter are flow-dependent; the uncertainty may be larger ina developing baroclinic low

    than in a subtropical high-pressure system. The background-erroraccuracy is also dependent

    on the local observation density. To estimate the flow-dependentuncertainty, a set of 3-hourforecasts, valid at the start of the4D-Var time window, is computed from ten perturbed,

    equally likely analyses. They differ because of small variationsimposed on the observations,

    the sea surface temperature and model error parameterization.These variations reflect the

    uncertainties in the observations, the SST and the forecastevolution. The perturbations

    produced using this ensemble of data assimilations (EDA) arealso used for the construction

    of the perturbations in the forecast ensemble (see chapter3,inparticular Section 3.2.1.For

    further detail on the EDA see Isaksen et al, 2010).

    The 4D-Var system handles all observational data similarly,including radiances from

    satellites. It compares the actual observations with what wouldbe expected, given the model

    fields. For satellite radiances the variational scheme modifiesthe model fields of temperature,

    wind, moisture and ozone in such a way that the simulatedobservations are brought closer to

    the observed values.

    2.3.2. The ECMWF early delivery system

    The 4D-Var analysis uses observations from a 12-hour timewindow, either 21 - 09 UTC (for

    the 00 and 06 UTC analyses) or 09 - 21 UTC (for the 12 and 18UTC analyses). To provide

    the best initial condition for the next analysis a fullresolution 3-hour forecast is run, based on

    the previous 4D-Var analysis (seeFigure 3).

  • 7/23/2019 User Guide to ECMWF forecast products

    17/129

    2. The ECMWF deterministic forecasting system

    9Location: Livelink 4320059 Owner: Erik Andersson

    Date: 23/07/2013 Version 1.1, Page 9of 129

    Figure 3: The 00 UTC cycle of the delayed cut-off 4D-Varanalysis covering 21 09 UTC starts with a

    15-hour forecast from the previous 18 UTC 4D-Var delayed cut-offanalysis (the 09 - 21UTC

    assimilation). Waiting for most of the available observations toarrive, the delayed cut-off analysis

    starts at 14:00 UTC using the 3-hour forecast as initialcondition. The rest of the 15-hour forecast isused as background(first guess) for the 12-hour delayed cut-off 4D-Var analysis (andsimilarly for

    the next 12 UTC analysis cycle).

    To ensure the most comprehensive global data coverage, includingsouthern hemisphere

    surface data and global satellite-sounding data, the 4D-Varanalysis waits about 5 hours to

    ensure that almost all available observations have arrived.

    Figure 4: The 12 UTC cycle of the early delivery analysis alsostarts from a 3-hour forecast, now from

    06 UTC, which is used as the background (first guess) for a6-hour early delivery 4D-Var analysis

    covering the time interval 09 - 15 UTC. The operational ten-dayforecast then starts from the 12 UTC

    analysis at about 16:30 UTC. The early delivery cut-off 12 UTCanalysis starts at 16:00 UTC.

    Although waiting for later data benefits the quality of theanalysis and its subsequent forecast,

    it adversely affects product timeliness. To overcome thisproblem, ECMWF has introduced its

    early delivery system,which allows the 00 and 12 UTC operationalanalyses to be produced

    significantly earlier, without compromising the operationalquality of the forecast products

    (seeFigure 4).

    To achieve this, an early cut-off analysis is made, relying onobservations arriving during the

    first four hours, which accounts for about 85% of the availableglobal observations. Since 80 -

    85% of the value of each 4D-Var analysis stems from thebackground (first guess)

    information and only 15-20% from the latest observations, notmaking use of the remaining

    15% of the observations reduces the predictive skill by a fewhours. Since this enables

    ECMWF to disseminate its forecasts 10 hours earlier, there is anoperational gain of 4 - 6

    hours in effective predictability. It is important to note thatthe background information

    always comes from the 12-hour 4D-Var where, thanks to the latecut-off, almost all available

    observations have been used.

  • 7/23/2019 User Guide to ECMWF forecast products

    18/129

    2. The ECMWF deterministic forecasting system

    10Location: Livelink 4320059 Owner: Erik Andersson

    Date: 23/07/2013 Version 1.1, Page 10of 129

    2.4. Retrieving ECMWF forecasts

    The exact value of the forecast parameters can be affected bythe way the data is retrieved,

    interpolated and presented.

    2.4.1. Temporal retrieval

    All forecast parameters, both surface and upper air, based on 00and 12 UTC HRES and ENS,

    are available at 3-hourly intervals up to +144 hours and at6-hourly intervals from +150 to

    +240 hours. The parameters are available hourly up to +90 hoursto members of the Boundary

    Conditions (BC) optional programme. Also available to BCprogramme members are two

    additional cycles, at 06 and 18 UTC, with all forecastparameters, both surface and upper air

    available hourly up to +90h.

    Precipitation forecasts are provided as values accumulated fromthe start of the forecast

    integration. The range of the daily variation of the forecast 2m temperature and wind gust is

    best estimated by retrieving the forecast maximum and minimumvalues. In both cases the

    valid time is defined as the time at the end of the period. Thecombination of accumulated and

    instantaneous forecast information can occasionally lead toinconsistencies, for instance,

    during the passage of a cold front: whereas there might bealmost cloud-free conditions at the

    end of the interval, they will be timed together withsignificant precipitation amounts

    accumulated over the wholetime interval.

    2.4.2. Spatial retrieval

    ECMWF forecast products can be retrieved at a wide range ofspatial resolutions, from

    regular and rotated lat-lon grids to the original regular andreduced Gaussian grid. The data

    can be retrieved from model, pressure, isentropic oriso-potential vorticity (PV) levels,

    depending on the parameter.

    Temperature, wind and geopotential forecast information isstored in spectral components but

    can be interpolated to any specified latitude-longitude grid.This interpolation can also be

    applied to near-surface parameters, although direct use of theoriginal reduced Gaussian grid

    point values is strongly recommended, especially forprecipitation and other surface fluxes to

    avoid the undesired effects of interpolation.

    2.4.3. Orography

    Because valleys and mountain peaks are smoothed out by the modelorography the directmodel output of 2 m temperature may representan altitude significantly different from the

    real one. A more representative height might be found at one ofthe nearby grid points. Any

    remaining discrepancy can be overcome by a correction using theStandard Atmosphere lapse

    rate or statistical adaptation (see AppendixB-6).

    2.4.4. The bi-linear interpolation

    Since 1979 ECMWF has used a bi-linear interpolation techniquebecause of its efficiency. It

    uses the 2 x 2 grid points closest to the selected interpolationlocation and takes a weighted

    average to arrive at the interpolated value (seeFigure 5).

  • 7/23/2019 User Guide to ECMWF forecast products

    19/129

    2. The ECMWF deterministic forecasting system

    11Location: Livelink 4320059 Owner: Erik Andersson

    Date: 23/07/2013 Version 1.1, Page 11of 129

    Figure 5: Bi-linear interpolation of four model grid points(black crosses) starts by linear interpolation

    between each pair of grid points (red circles). These two, ineither the latitudinal or longitudinal

    direction, are then used for interpolation to the requestedlocation (filled circle), weighted according totheir distance fromeach of the model grid points.

    The weights are based on the distance of the interpolatedlocation from each of the model grid

    points. Although linear in both directions, the bi-linearinterpolation is not linear but

    quadratic, except along lines which connect the model gridpoints (seeFigure 6).

    Figure 6: Example of bilinear interpolation for any locationwithin the grid box. The interpolated value

    for the centre is 1.25, but may take any value between 0 and 2elsewhere.

    Every interpolation technique has its advantages anddisadvantages. When the interpolation

    grid length is significantly coarser than the model grid,small-scale variability might

    misleadingly appear to represent larger scales. If, for example,the interpolation is made to a

    location close to one of the grid points, it will more or lesstake this value, even if it happens

    to represent a small-scale extreme. Only if the interpolationpoint is in the centre, does an

    interpolated value represent the mean over the grid-boxarea.

    For the dissemination, all fields are bi-linearly interpolatedto a 0.125 lat/lon grid,

    corresponding to a 13.5 km resolution in the meridionaldirection.

  • 7/23/2019 User Guide to ECMWF forecast products

    20/129

    2. The ECMWF deterministic forecasting system

    12Location: Livelink 4320059 Owner: Erik Andersson

    Date: 23/07/2013 Version 1.1, Page 12of 129

    Figure 7: For a given interpolation grid (red circles) theproportion of model information taken into

    account depends on the model grid resolution (black crosses).When the interpolated grid is twice the

    model grid length (or less), all model grid values will be usedin the interpolation.

    When the model grid is lessthan half the interpolation gridlength, the proportion of used grid

    points decreases (see Figure 7). If, for example, the model gridlength is a quarter of the

    interpolation grid, only a quarter of the model grid points aretaken into account. This has the

    undesired effect of not conserving area totals, which makes itunacceptable for use for surface

    fields, such as precipitation.

    2.4.5. The subsampling procedure

    Data may be requested on grids much coarser than 0.125 or 13.5km. Then a subsample of

    the 0.125 resolution grid points is selected. If, for example,interpolated values in 0.5, 1.0 or

    1.5 resolutions are requested, every 4th, 8th or 12thinterpolated value will be selected and

    disseminated (seeFigure 8).

    Figure 8: When the requested interpolation grid length is, forexample 0.5, every 4 thof the bi-linearly

    interpolated values (red circles) in a 0.125 resolution isselected.

    This will have the undesired effect that model grid point valueswhich, essentially represent

    small scales, may by chance appear to represent much largerscales (seeFigure 9).

  • 7/23/2019 User Guide to ECMWF forecast products

    21/129

    2. The ECMWF deterministic forecasting system

    13Location: Livelink 4320059 Owner: Erik Andersson

    Date: 23/07/2013 Version 1.1, Page 13of 129

    Figure 9: Example of the effect of inappropriate interpolationof precipitation fields. To the left the

    forecast is interpolated in a 0.25 x 0.25 grid, to the right ina 1.5 x 1.5 grid.

    2.4.6. Interpolating land and sea points

    When the interpolation of the 2 m temperature or 10 m wind takesplace over or near a coast

    line, the interpolation makes use of the land-sea mask (seeSection2.1.4) to decide whether

    the four grid points are land or sea points (see Figure 10).This determines whether the

    interpolated value should be regarded as a land or sea point. Inthis way there will be no

    undesired smoothing of gradients along coast lines.

    Figure 10: A rather detailed coast line (grey line) is definedby the model high-resolution grid

    (crosses). In the interpolation to a coarser grid, only the fournearest grid points to the interpolated

    position are used. Depending on whether they are predominantlyof land or sea character, they will

    unambiguously define an interpolated land (green) or sea (blue)point.

  • 7/23/2019 User Guide to ECMWF forecast products

    22/129

    2. The ECMWF deterministic forecasting system

    14Location: Livelink 4320059 Owner: Erik Andersson

    Date: 23/07/2013 Version 1.1, Page 14of 129

    Systematic differences between HRES and ENS (see chapter 3) can,for example, occur in

    connection with strong gradients along coasts, small islands orin mountainous regions. Any

    such discrepancy is most clearly apparent during the first fewdays, when the spread is

    normally small (seeFigure 11 andFigure 12).

    Figure 11: EPSgram for Kontiolahti in eastern Finland, 22 April2011 12 UTC. The systematicdifference between the HRES (blue line)and the ensemble Control forecast (red line) is around 5C.

    Kontiolahti lies close to a small lake resolved in HRES but notin the ENS (for more details about

    EPSgrams see Section 5.2).

    Figure 12: The difference in 2 m temperature between the HRESand the ensemble Control forecast for

    23 April 2011 00 UTC + 12h. The maximum and minimum differencesare indicated as integers. The

    interval is 2C. The differences are largest where thediscrepancy between the HRES and ENS land-sea

    mask or orography is largest.

    At grid points along coastlines the marine influence may beoverestimated and statistical

    interpretation schemes can be beneficial, in particular fortemperature forecasts (see Appendix

    B-6).

  • 7/23/2019 User Guide to ECMWF forecast products

    23/129

    2. The ECMWF deterministic forecasting system

    15Location: Livelink 4320059 Owner: Erik Andersson

    Date: 23/07/2013 Version 1.1, Page 15of 129

    2.5. The relation between grid point values and observations

    The reduced Gaussian grid values, like all other grid values,should not be considered as

    representing the weather conditions at the exact location of thegrid point, but as a time-space

    average within a two- or threedimensional grid box (Gber et al,2008). The discrepancybetween the grid-point value and theverifying observed average can be both systematic and

    non-systematic. The systematic errors reflect the limitations ofthe models ability to simulate

    the physical and dynamic properties of the system; thenon-systematic errors reflect synoptic

    phase and intensity errors (seeFigure 13).

    Figure 13:The comparison between NWP model output andobservations ought ideally to follow a two-

    step procedure: first from grid-point average to observationarea average. The systematic errors are

    then due to model shortcomings; the non-systematic stem fromsynoptic phase and intensity errors. In

    the next step, the systematic errors between observation averageand point observation result fromstation representativeness and thenon-systematic from sub-grid scale variability.

    When the NWP model output is compared with point observations,additional systematic and

    non-systematic errors are introduced, due to theunrepresentativeness of the location and the

    observations sub-grid variability (seeFigure 14).

    Figure 14: In reality, the comparison between NWP andobservations must for simplicity bypass thearea average stage. Thisresults in the systematic and non-systematic errors emanating fromdistinctly

    different sources.

  • 7/23/2019 User Guide to ECMWF forecast products

    24/129

    2. The ECMWF deterministic forecasting system

    16Location: Livelink 4320059 Owner: Erik Andersson

    Date: 23/07/2013 Version 1.1, Page 16of 129

    Systematic errors due to model deficiencies and/or observationalrepresentativeness can be

    partly corrected by statistical means (see AppendixB-6).Non-systematic synoptic errors can

    be dampened by different ensemble approaches (see Chapter4), butthe errors due to sub-grid

    variability can only be remedied by new model versions withhigher numerical resolutions. Amodel-independent estimate of thesub-grid noise can be made by verifying the

    observations from one observing station as forecasts for aneighbouring observing station.

    A typical value for hom*ogenous terrain is about 1C with typicaldistances of 50-150 km.

    2.6. Some characteristics of NWP output

    Output from NWP models does not behave in a simple or regularway due to the non-linear

    nature of the forecast system.

    2.6.1. Forecast error growth

    Forecast error growth is, on average, largest at the beginningof the forecast. At longerforecast ranges it levels offasymptotically towards the error level of persistenceforecasts,

    pure guesses or the difference between two randomly chosenatmospheric states (see Figure

    15). This error level is significantly higher than the averageerror level for a simple

    climatological average used as a forecast (see AppendixA-2 fordetails).

    Figure 15: A schematic illustration of the forecast errordevelopment of a state-of-the-art NWP (full

    curve), persistence and guesses (dotted curve), whose errorsconverge to a higher error saturation

    level than modified forecasts, which converge at a lower level(dashed curve).

    2.6.2. Downstream spread of influence

    Influences in the forecasts, both good and bad, often travelfaster downstream than the

    synoptic systems themselves. A two-day forecast over Europe maybe affected by the initial

    conditions over most of the North Atlantic, a five-day forecastalso by the initial conditions

    over the North American continent and easternmost North Pacific.There is also an ever

  • 7/23/2019 User Guide to ECMWF forecast products

    25/129

    2. The ECMWF deterministic forecasting system

    17Location: Livelink 4320059 Owner: Erik Andersson

    Date: 23/07/2013 Version 1.1, Page 17of 129

    present influence from the subtropical and tropical latitudes,particularly when a subtropical

    depression, tropical storm or hurricane enters the westerlies(seeFigure 16).

    Figure 16: Schematic illustration of the typical propagation offorecast errors over the northern

    hemisphere towards Europe in situations with generally zonalflow. The errors propagate mainly along

    the storm track, which during the warm season is displacedpolewards. Forecast errors or jumpiness

    at D+3 typically have their origin over the eastern part of theNorth American continent, at D+5 over

    the western part or the eastern part of the North Pacific. Inrare cases, forecast failures at D+7 have

    been traced back even further. During all seasons, but inparticular during the summer and autumn,

    forecast errors associated with disturbances in the tropics orsubtropics can move into the zonal

    westerlies.

    Hence, if the short-range forecast is initially poor (good) overthe area of interest, this does

    not mean that the medium-range forecast for the same area isnecessarily poor (good). Any

    attempt to judge the medium-range performance a priori from theshort-range performance

    ought to be made over large upstream areas and also involve theupper-air flow (Bright and

    Nutter, 2004).

    2.6.3. The relation between scale and predictive skill

    It is known from theory and synoptic experience that the largerthe scale of an atmospheric

    system, the longer its timescale and the more predictable itnormally is (seeFigure 17).

  • 7/23/2019 User Guide to ECMWF forecast products

    26/129

    2. The ECMWF deterministic forecasting system

    18Location: Livelink 4320059 Owner: Erik Andersson

    Date: 23/07/2013 Version 1.1, Page 18of 129

    Figure 17: A schematic illustration of the relationship betweenatmospheric scale and timescale. The

    typical predictability is currently approximately twice thetimescale, but might ultimately be three times

    the timescale

    Small baroclinic systems or fronts are well forecast to aroundD+2, cyclonic systems to

    around D+4 and the long planetary waves defining weather regimesto around D+8.

    Exceptions are features that are coupled to the orography, suchas lee-troughs, or to the

    underlying surface, such as heat lows. The predictable scalesalso show the largest

    consistency from one run to the next.

    Figure 18 shows 1000 hPa forecasts from the operational model.The forecast details differ

    between the forecasts but large-scale systems, such as a lowclose to Ireland, a high over

    central Europe and a trough over the Baltic States are commonfeatures.

    The +144 h forecast from 14 August predicted a south-westerlygale over the British Isles six

    days later. It would, however, have been unwise to make such adetailed interpretation of the

    forecast, considering the typical skill at that range. Only astatement of windy, unsettled and

    cyclonic conditions would have been justified. Such a cautiousinterpretation would have

    avoided any embarrassing forecast jump, when the subsequent +132h and +120 h runs

    showed a weaker circulation. The same cautious approach wouldhave minimized the forecast

    jump with the arrival of the +108 h forecast.

  • 7/23/2019 User Guide to ECMWF forecast products

    27/129

    2. The ECMWF deterministic forecasting system

    19Location: Livelink 4320059 Owner: Erik Andersson

    Date: 23/07/2013 Version 1.1, Page 19of 129

    Figure 18: A sequence of 1000 hPa forecast maps ranging from+156 h to +96 h, all verify on 20

    August 2010, 00 UTC.

    Smoothing out or, more correctly, filtering away small-scaledetails, in order to highlight the

    predictable scale, does not necessarily have to be donesubjectively (by eye). There are also

    various convenient ways to do it objectively. For example,retaining only the first 20 spectral

    components filters away all scales smaller than 1000 km andbrings out the more predictable

    large-scale pattern (seeFigure 19).

  • 7/23/2019 User Guide to ECMWF forecast products

    28/129

    2. The ECMWF deterministic forecasting system

    20Location: Livelink 4320059 Owner: Erik Andersson

    Date: 23/07/2013 Version 1.1, Page 20of 129

    Figure 19: Same as Figure 18 but based on the 20 largestspectral components.

    Five of the six forecasts now show much larger coherence, with acyclonic feature

    approaching the British Isles and a stationary high pressuresystem over central Europe.

    Spectral filtering does not take into account how thepredictability varies due its flow

    dependency: a small-scale feature near Portugal might be lesspredictable than an equally

    sized feature over Finland. Section4.4.3 shows how the ensembleforecasting system would

    treat the same synoptic situation in a more consistent andoptimal way.

    2.6.4. Forecast jumpiness

    Since every new forecast run is, on average, better than theprevious one, it is also different.

    These differences occur because of new observations that modifyprevious analyses of the

    atmospheric state and thereby the subsequent forecasts generatedfrom these analyses.

    Usually, the differences in the forecasts are small or moderatebut can occasionally be quitelarge and appear as forecast jumps.This jumpiness is an unavoidable consequence of a

    non-perfect dynamical forecast system and not a problemper se.Only when the forecasts are

    perfect will there never be any jumpiness (Persson and Strauss,1995).

    Just because the most recent forecast is, on average, betterthan the previous one, does not

    mean that it is alwaysbetter. A more recent forecast can, asshown inFigure 20,frequently be

    worse than a previous one; with increasing forecast range itbecomes increasingly likely that

    the 12 or 24 hours older forecast is the better one. Chapter 4describes how forecasters can

    handle forecast jumpiness by combining previous forecasts withthe most recent one.

  • 7/23/2019 User Guide to ECMWF forecast products

    29/129

    2. The ECMWF deterministic forecasting system

    21Location: Livelink 4320059 Owner: Erik Andersson

    Date: 23/07/2013 Version 1.1, Page 21of 129

    Figure 20: The likelihood that a 12-h or 24-h forecast is better(in terms of RMSE) than todays

    forecast. The parameter is the MSLP for N Europe and the periodOctober 2009-September 2010. Theresult is almost identical, if ACCis used as the verification measure.

    2.6.5. Flip-flopping forecasts

    The order in which the jumpiness occurs can provide additionalinsights. According to

    Table 1 the likelihood that precipitation occurs seems to beabout equal for the last two

    forecasts being consistent (R R -) or the last threeflip-flopping (R - R).

    Last 3forecasts

    84,96,108h

    Numericalprobability

    Observedfrequency

    Numberof

    forecasts

    - - - 0% 6% 598

    - - R 33% 15% 66

    - R - 33% 22% 46

    R - - 33% 36% 59

    - R R 67% 30% 43

    R - R 67% 44% 27

    R R - 67% 47% 43

    R R R 100% 74% 157

    Table 1: The percentage of cases when > 2mm/24 h has beenobserved, when up to threeconsecutive ECMWF runs (+84, +96 and+108h) have forecast >2mm/24 h for Volkel,

    Netherlands October 2007-September 2010. Similar results arefound for other west and

    north European locations and for other NWP medium-rangemodels.

  • 7/23/2019 User Guide to ECMWF forecast products

    30/129

    2. The ECMWF deterministic forecasting system

    22Location: Livelink 4320059 Owner: Erik Andersson

    Date: 23/07/2013 Version 1.1, Page 22of 129

    This might be because, although the last two forecasts are moreskilful than the earliest

    forecast, they are also, on average, more correlated. What theearliest forecast might lack in

    forecast skill, it compensates for by being less correlated withthe most recent forecast. The

    agreement between two on average less correlated forecastscarries more weight than two onaverage more correlated.

    2.6.6. Jumpiness and forecast skill

    It is intuitively appealing to assume that a forecast is morereliable, if it has not changed

    substantially from the previous run. Objective verifications,however, show a very small

    correlation between forecast jumpiness and the quality of thelatestforecast (seeFigure 21).

    The jumpiness relates rather to the skill of the average of theforecasts (see AppendixA-5).

    Figure 21: The correlation between 24 hour forecast jumpinessand forecast error for 2 m temperature

    forecasts for Heathrow at 12 UTC, October 2006 - March 2007.While the relationship between

    jumpiness and error is low in the short range, it increases withforecast range and asymptotically

    approaches the 0.50 correlation.

    2.6.7. Forecast trends cannot be extrapolated

    Trends in the development of individual synoptic systems oversuccessive forecasts do not

    provide any indication of their future development. If, duringits last runs, the NWP has

    systematically changed the position and/or intensity of asynoptic feature, it does not mean

    that the behaviour of the next forecast can be deduced by simpleextrapolation of previous

    forecasts (Hamill, 2003).

    2.6.8. Other state-of-the-art NWP models

    What has been said so far applies, in principle, to all majorstate-of-the-art NWP models,

    spectral- or grid-point-based, global or limited area,hydrostatic or non-hydrostatic. The

    differences in their average forecast quality are lesssignificant than the daily variability of the

    scores. Hence, the best NWP model, on average, is notnecessarily the best on a particular

  • 7/23/2019 User Guide to ECMWF forecast products

    31/129

    2. The ECMWF deterministic forecasting system

    23Location: Livelink 4320059 Owner: Erik Andersson

    Date: 23/07/2013 Version 1.1, Page 23of 129

    day. An NWP model that has recently performed significantlybetter (or worse) than other

    models (of about the same average skill), is not likely tocontinue to do so.

    However, as mentioned in Section2.6.2 and further discussed inSection4.2,it is as difficult

    to determine the model of the day from one of several NWP modelforecasts as it is fromconsecutive forecasts from the same model.Forecasters are advised to treat forecasts from

    different NWP models as a multi-model ensemble whose membersdiffer slightly in their

    initial conditions and model characteristics. The betterforecasters learn to handle the NWP

    output in this way, the better they will be able to manage theensemble forecasts, where these

    problems are more consistently addressed (see Chapter 4).

  • 7/23/2019 User Guide to ECMWF forecast products

    32/129

  • 7/23/2019 User Guide to ECMWF forecast products

    33/129

    3. The forecast ensemble

    25Location: Livelink 4320059 Owner: Erik Andersson

    Date: 23/07/2013 Version 1.1, Page 25of 129

    3. The forecast ensemble

    The value of NWP forecasts would be greatly enhanced if thequality of the forecasts could be

    assessed a priori; consequently, in parallel with improving theobservational network, the data

    assimilation system and the models, methods of providing advanceknowledge on how certain(or uncertain) a particular forecast is andwhat possible alternative developments might occur

    are being developed.

    3.1. The rationale behind the ensemble

    The ECMWF forecast ensemble is based upon the notion thaterroneous forecasts result from

    a combination of initial analysis errors and model deficiencies,the former dominating during

    the first five days or so. Analysis errors amplify most easilyin the sensitive parts of the

    atmosphere, in particular where strong baroclinic systemsdevelop. These errors then move

    downstream and amplify and thereby affect the large-scale flow.To estimate the effect of

    possible initial analysis errors and the consequent uncertaintyof the forecasts, small changesto the 4D-Var analysis are made,creating an ensemble of many (currently 50) different,

    perturbed, initial states. Model deficiencies are represented bya stochastic process. In order

    to save computational time, the ensemble members are run with alower resolution version of

    the IFS.

    3.1.1. Qualitative use of the ensemble

    If forecasts starting from these perturbed analyses agree moreor less with the forecast from

    the non-perturbed analysis (the ensemble Control forecast), thenthe atmosphere can be

    considered to be in a predictable state and any unknown analysiserrors would not have a

    significant impact. In such a case, it would be possible toissue a categorical forecast withgreat certainty.

    If, on the other hand, the perturbed forecasts (the ENS) deviatesignificantly from the Control

    forecast and from each other, it can be concluded that theatmosphere is in a rather

    unpredictable state. In this case, it would not be possible toissue a categorical forecast with

    great certainty. However, the way in which the perturbedforecasts differ from each other may

    provide valuable indications of which weather patterns arelikely to develop or, often equally

    importantly, not develop.

    3.1.2. Quantitative use of the ENS

    The ENS provides the ensemble mean (EM) forecast (or theensemble median) where the less

    predictable atmospheric scales tend to be averaged out. Theaccuracy of the EM can be

    estimated a priori by the spread of the ensemble: the larger thespread, the larger the expected

    EM error, on average.

    More importantly, the ENS provides information from which theprobability of alternative

    developments is calculated, in particular those related to riskof extreme or high-impact

    weather.

  • 7/23/2019 User Guide to ECMWF forecast products

    34/129

    3. The forecast ensemble

    26Location: Livelink 4320059 Owner: Erik Andersson

    Date: 23/07/2013 Version 1.1, Page 26of 129

    3.1.3. Characteristics of a good ensemble

    a) The ensemble forecasts should display no mean errors (bias),otherwise the

    probabilities will be biased as well.

    b) The forecasts should have the ability to span the fullclimatological range, otherwise the

    probabilities will either over- or under-forecast the risks ofanomalous or extreme weather

    events.

    c) Any systematic errors with respect to mean error orvariability can be detected by the

    deterministic verification methods discussed in Appendix A. Theycan, however, also be

    measured through the probabilistic verification methods outlinedinAppendix B.

    3.2. Generation of the ENS

    3.2.1.

    Different perturbation techniquesThe small perturbations addedto the Control analysis to create 50 perturbed initialconditions

    are computed by a combination of three methods:

    a) A singular vector (SV) technique seeks perturbations on wind,temperature and pressure

    that will maximizetheir impact on a 48 hour forecast, measuredby the total energy over

    the hemisphere outside the tropics. The maximization does notmean that the SV only

    intensifies weather systems; equally often it weakens them. Inaddition, the systems can

    be slightly displaced. Since SV calculations are quite costly,they have to be run at a low,

    T42, resolution corresponding to almost 500 km.

    To specifically address uncertainties in the moisture processestypical of low latitudes, inparticular of tropical cyclones, aspecial version of the SV is created using a linearised

    diabatic version of the model. These tropical SVs may alsoinfluence forecasts of extra-

    tropical developments, when, for example, tropical cyclonesenter the mid-latitude some

    days into the forecast and interact with the baroclinicdevelopments in the westerlies.

    b) The perturbations are modified by using differences betweenthe members of an ensemble

    of data assimilations (EDA). The EDA is an ensemble ofindependent 4D-Var data

    assimilations where the main analysis error sources(observation, model and boundary

    conditions errors) are represented by perturbing the relativequantities (observations,

    forecast model and sea surface temperature, respectively)according to their estimated

    accuracy.

    c) Model uncertainty is represented by two different stochasticperturbation techniques. One,

    stochastic physics, randomly perturbs the tendencies in thephysical parametrization

    schemes. The other, stochastic backscatter, models the kineticenergy in the unresolved

    scales by randomly perturbing the vorticity tendencies. Thewhole globe is perturbed,

    including the tropics. The Control forecast is run withoutstochastic physics.

    In the idealized schematics Figure 22, one can see how the4D-Var 12-hour assimilation

    window (left part of the diagram) modifies the initialtrajectories of the EDA members (in

    yellow) to reflect the information from the assimilatedobservations (black dots with error

    bars). The analysis trajectories (in green) show the impact ofthe new observations on the

  • 7/23/2019 User Guide to ECMWF forecast products

    35/129

    3. The forecast ensemble

    27Location: Livelink 4320059 Owner: Erik Andersson

    Date: 23/07/2013 Version 1.1, Page 27of 129

    ensemble: the spread of the ensemble has been reduced and a biashas been corrected by

    reducing the magnitude of some of the highest values.

    At the end of the assimilation window the EDA is used to provide(a) background error

    information for the successive deterministic analysis update and(b) the initial perturbations of

    the ensemble around the control analysis.

    Figure 22 Schematic representation of the ensemble ofassimilations (EDA, green) and its link with the

    forecast ensemble (ENS, blue). The yellow area represents thespread of forecasts starting from the

    previous EDA analyses 12 hours earlier.

    Once the different sets of SVs have been separately calculatedover the northern and southern

    hemispheres and over the tropics between 30 N and 30 S, they arelinearly combined (using

    coefficients randomly sampled from a Gaussian distribution) andadded to the EDA

    perturbations to make a set of 25 global perturbations. Thesigns of these 25 global

    perturbations are then reversed to obtain another set of 25mirrored global perturbations.

    This yields a total of 50 global perturbations for 50alternative analyses and forecasts.

    Consecutive members therefore have, pair-wise anti-symmetricperturbations. The anti

    symmetry may, depending on the synoptic situation and thedistribution of the perturbations,

    disappear after one day or so, but can occasionally be noticed3-4 days into the perturbed

    forecasts (seeFigure 23 andFigure 24).

  • 7/23/2019 User Guide to ECMWF forecast products

    36/129

    3. The forecast ensemble

    28Location: Livelink 4320059 Owner: Erik Andersson

    Date: 23/07/2013 Version 1.1, Page 28of 129

    Figure 23:1000 hPa perturbed analyses and forecasts of members 1and 2 from 15 August 2010, 12

    UTC; the positive and negative perturbations in red and bluedashed lines respectively. At initial time

    the perturbations are pair-wise anti symmetric, weakening ordeepening a shallow low-pressure system

    on the westernmost Atlantic (upper images). 24 hours into theforecast, the perturbations in member 1

    have led to the low splitting into two cyclonic pressuresystems, in member 2 to a significant deepening

    of the single low pressure system.

  • 7/23/2019 User Guide to ECMWF forecast products

    37/129

    3. The forecast ensemble

    29Location: Livelink 4320059 Owner: Erik Andersson

    Date: 23/07/2013 Version 1.1, Page 29of 129

    Figure 24: Same asFigure 23 but for 15 August 2010 00 UTC. Inthis case the anti symmetry is stillclearly seen 24 hours into theforecast, member 1 having the low deepened and displaced into aslightlymore westerly position, member 2 having the low weakenedand displaced into a slightly more easterly

    position.

    3.2.2. Quality of the individual perturbed analyses

    An unavoidable consequence of modifying the initial conditionsaround the most likely

    estimate of the truth, the 4D-Var analyses, is that theperturbed analysis is on average slightly

    degraded. The RMS distance from truth for a perturbed analysisis, in the ideal case, on

    average 2 times the RMS distance of the unperturbed analysisfrom the truth (see Figure25).

  • 7/23/2019 User Guide to ECMWF forecast products

    38/129

    3. The forecast ensemble

    30Location: Livelink 4320059 Owner: Erik Andersson

    Date: 23/07/2013 Version 1.1, Page 30of 129

    Figure 25: A schematic illustration of why the perturbed initialconditions will, on average, be further

    from the true state than the Control analysis is. The analysisis known, as well as its average error, but

    not the true state of the atmosphere (which can be anywhere onthe circle). Any perturbed analysis can

    be very close to the truth, but is in a majority of the casesmuch further away: in the ideal case the

    average distance is the analysis error times Consequently, theproportion of the perturbed analyses that are better than theControl

    analysis for a specific locationand for a specific parameter,such as the 2 m temperature or

    MSLP, is only 35% (see Figure 26); considering more than onegrid point lowers the

    proportion even further.

    Figure 26: Although the perturbed analyses differ on averagefrom the Control analysis as much as

    Control from the truth, for a specific gridpoint only 35% of theperturbed analyses are closer to the

    truth than the Control analysis.

    If an ensemble member is closer to the truth than to the Controlin, for example, Paris, it

    might not be so in Berlin. Indeed, the larger the area, the lesslikely that any of the perturbed

  • 7/23/2019 User Guide to ECMWF forecast products

    39/129

    3. The forecast ensemble

    31Location: Livelink 4320059 Owner: Erik Andersson

    Date: 23/07/2013 Version 1.1, Page 31of 129

    members are better than the non-perturbed Control analysis(Palmer et al, 2006). For a region

    the size of a small ECMWF Member State, only about 7% of theperturbed analyses are, on

    average, better than the Control analysis, for the larger MemberStates this decreases to only

    2% (seeFigure 27).

    With respect to the forecasts, in the short range only a smallnumber of the perturbed forecasts

    are, on average, more skilful than the Control forecast.However, with increasing forecast

    range the average proportion of perturbed forecasts that arebetter than the Control forecast

    increases, eventually asymptoting to 50%.

    Figure 27: Schematic representation of the percentage ofperturbed forecasts with lower RMS error

    than the Control forecast for regions of different sizes:northern hemisphere, Europe, a typical small

    Member State and a specific location. With increasing forecastrange, fewer and fewer perturbed

    members are worse than the Control (from Palmer et al 2006).

    3.2.3. Quality of the individual perturbed forecasts

    Since the perturbed analyses have, ideally on average, 41%larger analysis errors than

    Control, this makes the individual ensemble forecasts on averageless skilful than the

    unperturbed Control forecast. The difference in predictive skillvaries with season and

    geographical location, but is about one day (seeFigure 28).

  • 7/23/2019 User Guide to ECMWF forecast products

    40/129

    3. The forecast ensemble

    32Location: Livelink 4320059 Owner: Erik Andersson

    Date: 23/07/2013 Version 1.1, Page 32of 129

    Figure 28: Schematic image of the RMS error of the ensemblemembers, ensemble mean and Control

    forecast as a function of lead-time. The asymptoticpredictability limit is defined as the average

    difference between two randomly chosen atmospheric states. In aperfect ensemble system the RMS

    error of an average ensemble member is times the error of theensemble mean.However, what the perturbed forecasts may lack inindividual skill, they compensate for by

    their large number, their ability to form good median orensemble mean values and reliable

    probability estimations. The information should therefore beused in its totality, i.e. from all

    the members in the ensemble. The low proportion of perturbedforecast members better

    than the Control in the short range makes the task of trying toselect the member of the day

    very difficult and, perhaps, impossible. There are no knownmethods to a priori identify the

    best ensemble member beyond the first day or so (see Sections2.6.2 and 4.2).

    3.3.

    The ensemble at different lead times

    To use computer resources cost-efficiently, the horizontalresolution of the ensemble is

    reduced at day 10, and the remainder of the forecast (out to 15or 32 days) is run at half the

    horizontal resolution of the first 10 days.

    3.3.1. The 10-day range

    In spite of its coarser resolution, which is half that of HRES,the ensemble Control forecast

    performs very similarly to HRES with respect to synopticpatterns. Differences are most

    noticeable for small-scale extreme weather events, where HRES isable to generate,

User Guide to ECMWF forecast products - [PDF Document] (2024)

FAQs

Is ECMWF forecast free? ›

A subset of ECMWF real-time forecast data are made available to the public free of charge. Their use is governed by the Creative Commons CC-4.0-BY licence and the ECMWF Terms of Use.

How accurate is the ECMWF weather forecast? ›

Also to increase your confidence: both models also works with general accuracy of 95–96% for up to 12 hours, 85–95% for three days, and 65–80% for 10 days.

What is the difference between Ukmo and ECMWF? ›

The UKMO model, unlike the MRF and ECMWF, is a grid model rather than a spectral model. This makes it more susceptible to computational instability for long-range forecasts. The horizontal grid is 1.5° latitude x 1.875° longitude (or 150 x 180 km at the equator), and there are 15 sigma levels in the vertical.

How often does the ECMWF model run? ›

The ECMWF updates frequency is 2 times a day.

The update frequency of the forecast is the regular time interval after which new forecast data is received from the supercomputers.

Does AccuWeather use ECMWF? ›

Nearly ten years ago, I helped AccuWeather.com launch our AccuWeather.com Professional service. I am now proud to announce the biggest upgrade since then: The full ECMWF model! That's right, it's the most accurate worldwide computer forecast model that you can't get anywhere else.

What is the difference between NOAA and ECMWF? ›

ECMWF forecasts are found to be of higher quality than the NOAA forecasts. The normalized RMSE of ECMWF's High Resolution forecasts ranges from 14% to 34%.

What is the most reliable weather forecast model? ›

Global models with worldwide weather forecasts

The ECMWF is generally considered to be the most accurate global model, with the US's GFS slightly behind.

How far out does ECMWF go? ›

Our long-range (seasonal) forecasts provide information about atmospheric and oceanic conditions up to seven months into the future.

What is the most accurate weather forecaster in the world? ›

Powered by proprietary GRAF technology (Global High-Resolution Atmospheric Forecasting), The Weather Company is The World's Most Accurate Forecaster1 and a trusted weather partner for people and businesses everywhere.

What is the difference between Eumetsat and Ecmwf? ›

EUMETSAT is contributing to the Data Curation and Products component, focusing on satellite data and a comprehensive set of meteorological observations. Meanwhile, the ECMWF Pilot Project is pivotal for the Analysis, Modelling, and Ensembles component.

What is the difference between SRTM 30 and 90? ›

First of all, the main difference is the spatial resolution 30 m and 90 m for SRTM 30 and SRTM 90 respectively. In addition to, SRTM 30 is considered SRTM data set enhanced with U.S. Geological Survey's GTOPO30.

What is the difference between Nexrad and TDWR? ›

The TDWR radar stations are maintained and operated by the Federal Aviation Administration. Both radar's are pulsed Doppler types that measure reflectivity out to 460 km, and radial velocity and spectrum width out to 300 km for NEXRAD and 90 km for TDWR.

What country is ECMWF? ›

The European Centre for Medium-Range Weather Forecasts (ECMWF) is an independent intergovernmental organisation supported by most of the nations of Europe. It is based at three sites: Shinfield Park, Reading, United Kingdom; Bologna, Italy; and Bonn, Germany.

What is the best long range weather forecast website? ›

AccuWeather gathers the best and most comprehensive weather data to deliver forecasts with Superior Accuracy. Forecasts are pinpointed for every location on Earth and extend further ahead than any other source.

What are the hours for ECMWF? ›

European wave model

Analyses are available for the four main synoptic hours 00, 06, 12 and 18 UTC; forecasts are produced out to 5 days based on the 00/12 UTC analysis.

How to get ECMWF forecast data? ›

Real-time data can be delivered to authorised or licensed users either ​directly by ECMWF or via the chosen Member or Co-operating State National Meteorological Service. This choice is specified during the ordering process. Open real-time data can be accessed free of charge via the public FTP.

Which is better GFS or ECMWF? ›

The ECMWF generally outperforms the GFS statistically, but there are instances where the GFS may perform better for specific events. Let's summarize the main differences between both leading weather forecasting models: Resolution: GFS runs at a lower resolution than the ECMWF model.

What weather app uses ECMWF? ›

Pflotsh ECMWF is our pro app: With a subscription to it, all other Pflotsh apps are unlocked. Professional ECMWF weather forecasts worldwide. Starting with temperature and precipitation forecasts up to highly specialized parameters. For the local forecasts, other models than ECMWF are also provided.

Which weather API is free? ›

Free API. Open-Meteo offers free access to its APIs for non-commercial use, making it convenient for individuals and developers to explore and integrate weather data into their projects.

Top Articles
Latest Posts
Article information

Author: Msgr. Refugio Daniel

Last Updated:

Views: 5893

Rating: 4.3 / 5 (54 voted)

Reviews: 93% of readers found this page helpful

Author information

Name: Msgr. Refugio Daniel

Birthday: 1999-09-15

Address: 8416 Beatty Center, Derekfort, VA 72092-0500

Phone: +6838967160603

Job: Mining Executive

Hobby: Woodworking, Knitting, Fishing, Coffee roasting, Kayaking, Horseback riding, Kite flying

Introduction: My name is Msgr. Refugio Daniel, I am a fine, precious, encouraging, calm, glamorous, vivacious, friendly person who loves writing and wants to share my knowledge and understanding with you.