Offshore Yachting : February March 2008
number of lucky vessels linked to the internet. Navigators can download forecast GRIB (Gridded Binary data files) winds into their favourite routeing software (such as MaxSea, Deckman, Raymarine, etc.). Users should understand that these GRIB winds or files mostly come from models running at low (coarse) resolutions, anywhere from 20 kilometres to 120 kilometres. Yachtsmen and women can have more confidence in the low-resolution models when they're well away from the effects of land, but should be very careful when they draw near the coast. Low-resolution models have a lot of trouble resolving the effects of land on winds generally, particularly where very localised effects (headlands, estuaries, etc.) and sea/land breezes come into play. These difficulties are mostly due to the smoothing of topographical fields and the assumptions used in the maths and physics packages that drive these beautiful 'things'. In these situations, navigators will have to 'massage' the model output to take into account the complex and not so complex effects of topography, sea and land breezes, etc. Local knowledge often plays a big role in this 'massage process'. One rule of thumb says that model resolution lower than five kilometres (that is, numbers bigger than five kilometres) will not resolve the sea/land breeze circulation and the effect of coastlines on the local wind flow. Here are some reasons why the computer model forecast can go wrong: • The accuracy of mathematics and physics packages (due mostly to an incomplete physical understanding of atmospheric processes) • Density of the grid spacing. A model solves about seven complex equations for grid-points. These grid-points not only cover the Earth's surface, but extend upwards for some 30-plus levels. The further apart the grid points (that is, the lower the resolution), the less accurate the model • Accuracy of the input feld (the observations and hence the computer model's starting analysis has to be correct). Garbage In will normally imply Garbage Out (GIGO) • Model topography • Size of the model domain (the area over which the model runs; the smaller, the better) • Initial errors tend to increase over time One of the more popular GRIB download sites is the Global Marine Network (GMN) at www.globalmarinenet.net/grib.htm. Another great site is via Frank Singleton at www.franksingleton.clara.net/gribus.html 'Portable' high-resolution model versions can be tuned for any yacht race/cruise track around the world and hence the user can obtain detailed forecast wind velocity data three to four days ahead. These types of model can also be run, for limited areas, onboard a vessel utilising a high-powered laptop and either a mobile telephone or a satellite link to download data. As a matter of interest for you forecasters out there, professional meteorologists normally work through the following list in order to arrive at a forecast. NWP is just a part of the overall process. Anyone wishing to forecast the weather for themselves would be well advised to do the same: • Perform a comprehensive analysis (as best you can at sea) • Forecast funnel: work from large to small scale • Large scale: global NWP analyses • Synoptic scale: regional NWP analyses and manual charts • Local (meso/micro) scale: high-resolution NWP plus any local analyses and observations From the above, formulate a conceptual model for what is happening. Construct a 3-D picture of the current atmospheric situation that explains as far as possible all the current weather observations and changes. Once you're happy with this, then: • Use NWP guidance (forecast product) to predict larger scale controls on local weather • Take a critical approach to NWP: assess how well it has performed over the past few days; if you can, compare the output of different NWP models • Take diurnal (24-hour) variations of the various weather elements and local topographic effects into account • Note short-term trends in conditions from your weather log (for example, for the last six hours) • Apply any local knowledge • Lastly, produce the forecast Remember that NWP guidance and hence your weather forecast can fail, so YOU have to weather watch utilising observations. Be prepared to AMEND your forecast if and when the need arises. Philosophically, numerical weather prediction must represent one of the great triumphs of human achievement, combining as it does the physics of the atmosphere and oceans, mathematics, and the high technology of the supercomputer. It is one area where high technology is being applied every day to provide information on that one thing that affects us all -- the weather. How did we do it before NWP? Good forecasting. Reference: Whitaker, R, 1990: Numerical Weather Prediction, e Institute of Mathematics and its Applications. October/November, Volume 26. WEATHER WATCH 68 | offshore O Figure 1. A forecast chart from a very high resolution NWP model of Tasmania, valid for 2000hr AEDST 23/1/2008, showing forecast winds in knots and surface temperature in degrees Celsius.
December January 2008
April May 2008