by Bruce Amlicke
With all due apologies to W. Shakespeare.
Hello, I am Bruce and I GRIB. We all do. As sailors, our one primary need (other than Rum) is to get the best possible weather information. This used to mean printed weather charts available by Weather FAX or voice weather forecasts broadcast on the Single Sideband (SSB) radio. With the advent of computers and the Internet, we can now get more detailed weather information. In 1985, the WMO (World Meteorological Organization) met to establish a format to exchange weather data. GRIB (GRIdded Binary or General Regularly-distributed Information in Binary form) resulted. From a sailor’s point of view, it is used to distribute the products of Global Weather Models, chiefly the GFS (Global Forecast System) model.
We get GRIB weather data directly in the form of plots from such programs as zyGrib, Passage Weather, OCENS, or Buoy Weather or indirectly in the form of forecasts from such as Chris Parker and Weather Underground. It is readily downloaded and can be easily plotted to give a graphic image of the current or future weather pattern in your area. Because of easy availability and display, it has often become the primary source of weather information for many sailors. BUT, should it?
So first off, what is a Computer Weather Model?
It is a computer program or APP that calculates the future weather from known current weather current conditions (initial conditions) and known constraints (boundary conditions). It differs from the programs that run on your PC or tablet in size (there are smaller models used for limited areas that do run on PCs). These programs run on massive computers and take hours to complete a single run. There are several major weather forecast models: the North American Mesoscale Model or NAM, the Global Forecast System or GFS, the long standing Nested Grid Model or NGM and the European Model, ECMWF. Typically we most often encounter the GFS and less often the European Model.
How good are these models?
In the early days of programming, a term GIGO (Garbage in Garbage out) originated. The GRIB data is only as good as the data gathered from the underlying weather model, which in turn is only as good as the data the model uses to begin with.
Predicting the weather on a global scale is one of the toughest problems out there. The models depend on understanding the weather at the time the model is run (initial conditions) and surface and upper atmosphere conditions as the model moves forward in time (boundary conditions). Any errors in these conditions work their way into the model and make the results less accurate as time goes on. We are all familiar with the error cones given for hurricane tracks. Similar, though not as easily plotted, errors exist for the global numerical weather models.
A Real limitation is the lack and quality of input data. Raw atmospheric data is collected from a whole variety of sources. The traditional means of collection were fixed stations (often near airports) and weather balloon data. Now satellite and other data go into the mix. One would hope that more data would produce better results.
But, one problem that still remains is bad data. The collection process is much more automated than it once was, but it still depends on humans who can (and do) make mistakes. Also instruments are known malfunction. Some years ago, I was asked to examine a dataset that had been used for weather forecasting to determine if it could be used to test weather models. I was surprised at the amount of errors present. I found wind directions greater than 360 degrees, wind velocities that were negative (some places it does suck), and many other somewhat subtler errors. While the filtering process has improved, there is still bad data getting into the models.
Besides bad data, you have missing or incomplete data. If you look at a map of were weather data is collected you notice that most are in populated areas. Rural areas and the oceans have much fewer collection points. Satellites have filled in some of these areas, but there are still great stretches of the world that are not sampled. Also as budgets get tighter it takes longer to repair or replace stations that have failed.
Finally there is un-representative data. Historically many weather observation stations were set up near airports to meet the needs of aviation. Pilots need up to the minute data on the landing and takeoff conditions. As urban sprawl has enveloped many airports, the heat islands around major cities have distorted the data collected at the airports. There are many more of these conditions.
Limits and Biases of the Computer Weather Model themselves
The biggest limitation of the world wide weather models is their grid. The current GFS model uses a grid of 15 nautical miles! So any phenomena with a resolution less than 15 nautical miles (.25 degree) will not be resolved. That is why the winds between islands in the Eastern Caribbean are not shown accurately. The Euro model uses a similar .25 degree grid. However, things are not quite so bad. All model use sub-grid modeling. This is an attempt to model features that are too small to resolve by the model grid. So looking at a typical wind plot of the Eastern Caribbean you see some effect of the islands presence, but not a lot.
Each model is built by different organizations and uses different computer algorism, so there is a natural bias to the output. Experts in the field have long said that the Euro model seems to be more accurate. However, since the Euro model output is not as easy to acquire (it cost more money to buy the data), the GFS model predominates. On good paper on this subject is http://www.accuweather.com/en/weather-blogs/weathermatrix/why-are-the-models-so-inaccurate/18097
From the Model to the GRIB
Once the model is run GRIB files are extracted from the solution data set. GRIB files are in a standardized format that makes them easy to use by the community. There are three GRIB data formats: GRIB 0 which is no longer in use, GRIB 1 which is use most commonly today, and a new GRIB 2 or second edition that is being introduced. The newest format allows much more background data to be passed along with the required values (wind speed, temperature, pressure, etc.). This should result in better information to the consumer (us).
Presenting the GRIB Picture
When we open a GRIB application (Passage Weather, for example) we see a plot of the value we are interested in (wind speed and direction, for example) overlaid on a map. You notice that there are many more wind arrows displayed than the 4 per degree that the model calculates. What the display routine does in interpolate the wind arrows from the 4 or more surrounding points where the model produces an output value. Any discontinuity is smoothed out. Weather fronts are discontinuities in wind speed and direction and are blurred, particularly strong fronts. Very localized effects suffer the same fate. Interpolation exists in time as well. Any results presented for less than 6 hours have been interpolated.
What can be done?
To take full advantage of the GRIB data and get a complete picture of the weather for your location or passage, consider the following?
- Remember that GRIB data is untouched by human hands unlike weather charts that are signed by the forecaster. A large grain of salt is needed some times. If it does not look right, perhaps it is not right.
- Don’t rely on a single GRIB model run – I often start looking at the GRIBs several days in advance or more. You will notice whether the predictions for a given day or weather window are consistent or vary from model run to run. I am much more comfortable with a forecast that consistently predicts the same conditions as the date approaches, than those which look good one day and bad the next.
- Do your own error analysis – pick a point or several points along your intended route and try to get current conditions for that point. Weather buoys are good (real ones not buoy weather ones). How does the model predict the weather for that point compare with the actual conditions?
- Get a complete picture – besides looking at the GRIBs for your passage, look at the weather charts and listen to forecasts. Look well beyond your route to see what conditions might affect your area. In the Eastern Caribbean, cold fronts coming off the US and High Pressure areas in the North Atlantic will compress or expand the pressure gradient and thus the wind velocity and direction. The persistent low pressure area in Northern South America will do the same. Tropical waves approaching from Africa will often signal squalls and a wind shift.
- Have a plan – When making a passage, I look at the weather at the destination and for each leg of the route. I also look at the weather at places I might duck into if conditions are not right or some something goes wrong.
- Weather Window or Sucker Hole – I used to be a fighter pilot. The commander would often look out his window, notice a patch of blue sky, and launch the fleet. This patch of blue would often disappear about the time we wanted to return. In sailing, we see the same thing. I look for a consistent weather window that is longer than my intended passage and well within my personal limits. What I try to avoid is one that looks good today and gets smaller with each passing day and were the conditions within the window get worse with each model run. I try not to get suckered in by the “have to get there syndrome.”
GRIB weather data is a very useful tool to the cruising sailor. The quality of the data and the presentations drawn from that data continue to improve. The real limitation in the Eastern Caribbean is the effects of the islands or more accurately the gaps between the islands. A rule-of-thumb I use is that as I get within about 3 miles of the gap (sailing in the lea of the island), I expect the wind to begin to increase toward the forecast value and the wind direction to swing toward my nose (headed). How much depends on how close I am to the island. As I reach the end of the island I expect the wind to increase to around 5 knots above the forecast value. Only when I have reached a point about 3 miles from the end of the island does the wind approach the forecast values and the direction nears the forecast direction. The reverse happens when I get to about 3 miles of the next island. The wind begins to go aft and the velocity increases around 5 knots.
The other effect is that of the current that often flows either north or south along the back side of the islands. It can make the seas near the ends of the islands a real mess.
Lastly, “Hey, let’s be careful out there”, Sergeant Phil Esterhaus, Hill Street Blues.