Session of the wmo-ioc joint Technical Commission for Oceanography and Marine Meteorology (jcomm) agreed that it would be logical to transform the wmo wave Programme into the jcomm wind Wave and Storm Surge Programme
Operational Two-Dimensional Storm Surge Prediction Models
Download 1.37 Mb.
|
- Navigate this page:
- 5.2 Setting the Initial State for the Forecast
- 5.3 Nesting Higher Resolution Storm Surge Models
- 5.4 Operational Products and Requirements
Operational Two-Dimensional Storm Surge Prediction ModelsThe implementation of a storm surge prediction model implies a careful combination of physical aspects, numerical methods and relevant contributing processes, as described in section 4. An overview of operational practice regarding storm surge prediction from results of a worldwide survey is included in the Dynamical Part of this Guide (http://www.jcomm-services.org). There, about 75 % of the reported operational / pre-operational applications are 2-D models, not considering finer nested grids. Due to the essentially two-dimensional nature of physical processes involved, as described in section 2, the application of depth-averaged hydrodynamical models detailed in section 3, can produce accurate forecasts of storm surges for most practical purposes. They are transportable, adaptable and valid for a wide range of scales. They have proved to represent accurately the tides, the storm surge and the interaction between them. Most of the operational applications referred to in the dynamical part of the guide are forced by tides at the open boundaries. Surge estimation at the boundary is used for 2-D, while seasonal conditions are reported for 3-D models. Information on the storm intensity, track and landfall is provided to tropical cyclone applications. The models' resolution ranges from 10 / 20 km for regional models and 3-D applications to 1 km for finer nested grids. The 2-D models are especially suitable for ensemble forecasting. The updated assessment of the state of the art on operational storm surge numerical models is detailed in Tables 3 to 6 of the Dynamical Part of the Guide. International co-operation, data and models availability provide a favourable scenario for the development of such applications. At the same time, they provide a frame for local products quality requirements. 5.1.1 The Rationale of the Forecasting SystemLocal / regional applications main challenge is to be accurate while practicable, that is, to keep in mind the final objective that useful forecasts reach the public in a timely fashion. The developer identifies local or regional phenomena features and consequently chooses the approach and tools that better represent them, always fitting real capabilities for achieving sustained services. The accuracy of the meteorological forcing required for the storm surge modelling limits the forecast range in practice, compared to the available meteorological forecasts. Experimental tests are recommended in this sense, according to the accuracy required to the application. On the contrary, for phenomena requiring very short range forecasts, the whole process time needed for a given forecasting time should be taken into account, as well as the frequency of successive runs. The former is composed by meteorological fields delay plus the acquisition process plus the processing and dissemination time. The latter is usually called data assimilation cycle in NWP and will be referred to in next section. The life-time of the phenomena and/or the propagation speed of the waves in a coastal sea or into an estuary define the possibility of real-time use of data. On the other hand, the dynamics of propagation of the waves in the basin to be considered determines the potential benefits of data assimilation for certain locations. A wide variety of uses of sea level observations in real time in conjunction with the numerical prediction is evidenced in operational / pre-operational applications, as described in Table 9.3 that is: forecast bias correction, initial conditions (assimilation), blending in bulletins with the forecasts, application of regression and empirical methods, model validation. Forcing winds used belong mostly to high-resolution national NWP models and / or hurricane predictions. The forecast ranges of most of the operational applications are from 36 to 72 hours, although a forecast range as long as 120 hours has been reported. The predictions of surges generated by tropical cyclones have shorter ranges, usually within 12 hours. 5.2 Setting the Initial State for the ForecastThe evolution of long gravity waves for short and medium range periods is highly dependent on the initial state. The period of time for which the regime is achieved is dependant on the local specific characteristics, and is referred to as spin-up period, particularly applied when starting from equilibrium. For every individual forecast the initial fields should be as accurate as possible. That is why the model is re-run from the starting time of the previous run (restart fields), up to the initial state (time 0) of the current forecast. This is called the hindcast period of the run. In the hindcast, the driving meteorological fields are diagnoses or analyses from observations. Figure 5.1 outlines the procedure for the storm surge (including tide) numerical model operation. Its frequency is related to the analysed meteorological fields’ availability, that is, the data assimilation cycle, usually 6 or 12 hours. On the other hand, the update of forecasted meteorological information depends on the meteorological models´ output, usually at a 1 or 3-hour interval. The model run corresponding to the isolated tidal part completes the forecasting procedure for the storm surge. Elsewhere in the guide, how to deal with the surge contribution to the total water level has been discussed. 
Figure 5.1: Handling the meteorological information for storm surge numerical forecasting. The direct assimilation of sea level data may be used in the hindcast period to improve the definition of the initial state of the forecast. Elsewhere in this section, methods for data assimilation are discussed.
The storm surge arriving to the coast has in general been generated and propagated in varied conditions, according to particular local situations. A very general case is the generation of the surge over shelf seas and its modification in estuaries, islands, bays and other restricted areas. As it was seen in previous chapters, this scenario may require different parameterisations of subscale processes, such as bottom friction, evidenced in the model calibration process. The representation of local features of bottom topography requires different resolutions as well. Local modification of tides, such as amplification, may be poorly represented in coarse grid models. The latter is true also for the surge part, which is modified locally, even if it is not locally produced. The nesting technique is also appropriate to better represent the effects of remote perturbations in the area of interest, as modified by complex topography and local processes, such as flooding and drying or tide-surge interaction in either extremely shallow water or strong currents scenarios. The nesting technique allows the consideration of the storm surge entering the domain of the limited area model. The value of the storm surge given by the coarse model at the boundaries of the fine mesh model is extracted at time intervals to obtain the storm surge, i.e. the difference of the water level and current given by runs with and without atmospheric effect. These values are introduced in the open boundaries condition ... as the external surge, when applied to the run of the model of fine mesh with the atmospheric forcing. 5.4 Operational Products and RequirementsProducts derived from the 2-D operational models are diverse: time varying sea level (usually surge only) forecasts at specified locations and/or charts, local peaks and maxima charts, outputs for flooded areas and currents. There was one report of the application of a statistically derived scale of risk degree for set up (floods) as well as for abatement (navigation risk). On the enquiry about additional requirements received from community and not provided by the models, flooded areas, oil spill evolution and surface currents were mentioned. The performance of operational / pre-operational storm surge models is monitored, in most cases, on continuous basis. The sea level products considered for the validation are either the full time series, the peaks or levels at selected times, such as high and low waters. The statistical parameters obtained, usually for different forecast ranges, are varied. The bias, RMS, standard deviation, average percentage error, linear regression (correlation coefficient) and the relation of standard error to mean square deviation are chosen by the different services. Statistics are provided either with a monthly or yearly frequency or may be related to the occurrence of major storms. One case was reported of documented normative for quality tests on the operational models. Practices for operational model verification are summarized later in Table 9.2. It also refers to models´ performance and some verification results, as reported to the worldwide survey by several countries.
|