GRIB Data Warning...a good example

[weatherman]

Hi,

A deep area of low pressure (~980mb) is going to be crossing Scotland this weekend. By 6am Sunday morning it is expected to be off northern Scotland...

http://www.wetterzentrale.de/pics/Rtavn481.png


Now, given this area of low pressure, it's depth, development and track, wind speeds across outhern and western Scotland as the feature passes through, can be expected to be around 35-45kt (F 8 -F9) gusting in excess of 60kt (F10-F11) over exposed coasts and hills.

Many of you will have access to GRIB data and can check the speeds predicted by the models. This data will be the same as the chart below...

http://www.wetterzentrale.de/pics/Rmgfs485.gif

Notice how the mean wind speeds from the GRIB data shown here are of the order of 20-25kt (although the orange colours in the far NW of the chart indicated mean speeds of 30kt).

This is a good example of how GRIB data will smooth out wind speeds and, although very useful, GRIB is only a tool in the forecasting box, especially in potentially dangerous situations.

Thought you might be interested.

Simon

[Nordic]

Very interesting!

Have you got any advice, on how to avoid being caught out by this averaging that gets done?
Can one get a standard deviation as well?

Alan

[2hullsgood1hullbad]

Thanks for that. We'd noticed that this seemed to happen, and since then we always add 5kts to whatever the Grib data predicts.

[jkd]

Outstanding! Thank you for the lesson Doc.:)

I would be interested in seeing a weekly series of this type of thing if you are up to it. :cool::)

John

[ForumAdmin]

Weatherman - Simon offers a paid for weather forecast service for individual boats and a number of other services.

I am trying to set up a weather forum with a range of features helpful to members.

[2MT]

IMO this is the problem with GRIB data all over the place. Fishing off the west coast of the US the fishermen observe a a wind that steadily blows north right up the coast. its not a very wide band but its there all the time. it does not show on the wind charts until there is a storm running up from the south and it intensifies.

In alaska the fishermen forced NOAA to keep Peggy our best weather forecaster on the Job even though she was never trained. Why? because Peggy knows the area and weather in detail so she corrects for the smoothing effects in many models.

An example is the area around false pass and sand point in the aleutians.
If you look at the weather data you see that the wind speeds are often pretty close to the same. this is due to the smoothing. Reality is far diffrent. the winds at sand point are very often 30 to 40kts higher.

I dont know why the models are made to smooth the data but its sure nice that there are weather folks around that look past the smoothing.

[Maxingout]

I refer to grib files as grib fantasies.

I use them to orient myself to possible wind direction and speed, but I don't depend on them. I have seen cruisers sit for days in port because the grib fantasies are not favorable. The gribs are not reliable in areas of tropical depressions and hurricanes, and when you are a trade wind sailor, that can limit the benefit of the grib fantasies.

I like the grib files. I just wish they worked better in the trade winds where I like to sail.

[TYRNTLZRDKING]

I refer to grib files as grib fantasies.

I use them to orient myself to possible wind direction and speed, but I don't depend on them.

Dave,
What do you depend on for your weather forcasts?

[Maxingout]

We would pull down the grib fantasies to see what they said, and sometimes they actually worked.

I would pull down weather fax a couple of times a day for the section of ocean that we were located in. I would compare the weather fax to the gribs and make my own opinion of what was going to happen, and then make my contingency plans based on what I saw. Many times the gribs were helpful, but I found them to be of more psychological assistance rather than helping me make routing decisions on long passages. If gribs were favorable, that made me feel good, and that was worth something. It also made crew members feel good to see favorable grib fantasies in the tropics.

When I look at weather information, I do contingency planning. I look at highs and lows and their probable direction of movement. I look at the best case scenario, and the worst case scenario every day. I then look at wind direction, speed, sea state, and fuel supplies so that I know what I will do if the worst case scenario develops.

When I sailed across the Atlantic, there were tropical storms to the north of our track, and yachts that cut the corner from the Canaries got beat up by the bad weather. We sailed south to the Cape Verdes and even further south as we crossed the Atlantic. My plan was to head south toward the doldrums if anything threatened, and I had plenty of fuel to activate my plan.

Weather information is excellent for 24 to 48 hours, and frequently good out to three days, but after that it's anybody's guess. That time frame gives me plenty of time to activate my contingencies if things are headed the wrong way. When you are out cruising, you aren't on a schedule to get across the Atlantic. I carried enough fuel for 1400 miles, and with a seventy-two hour window of opportunity, I felt that I could deal with any contingency, even if I had to travel 500 miles in the wrong direction.

Weather fax was my most important tool, and the grib fantasies were often helpful.

[Bvimatelot]

www.passageweather.com (http://www.passageweather.com) is an excellent site for gribs plus a lot more. I noticed the last time I looked at it that they have a button to "de-smooth" in the area you are interested in but I didnt have time to explore the feature. Sounded useful though.....

[lhsmith]

The smoothing that folks complain about is an inherent feature of the weather models. Constructing such models (of water instead of air) occupied much of my professional career, so I will say a few words about the models with the hope that people can make more informed use of the gribs.

The points at which the models make predictions are really averages for each latitude-longitude grid box during a period of an hour. Obviously, the observed wind can differ a lot from place to place within each grid box during an hour.

The way the models work is to take weather conditions “now” in all grid boxes and predict changes an hour later (called a time step). Then pretending the prediction is “now”, they make a prediction an additional hour later. Repeating this process over and over they produce hourly predictions for several days into the future. Given that we don’t have perfect knowledge of present conditions and that the models don’t represent weather processes perfectly, it is easy to see how imperfections magnify in the predictions, making them meaningless more than a few days hence.

Model averages are pretty good estimates over open ocean compared to along the coast because, of course, land alters surface wind patterns. The models only know land in terms of its average elevation for an entire grid box. With such large grid boxes, coastlines and mountain ranges are poorly represented relative to how humans experience them, and coastal predictions can be unsatisfactory.

The models would be able to predict more detailed variations if they had a finer grid spacing and shorter time step. The models are such computational monsters, however, that they require a few hours on the largest computers in the world to run 5-day predictions. Halfing the horizontal grid size and halfing the time-step multiplies the computation time by a factor of 8. It's easy to see that you wouldn’t have to increase the level of detail much before predicting an hour later would take an hour of computer time, which is not very useful.

Wind is likely to be gusty any time that there is a large atmospheric pressure difference over a relatively short distance (a few 100 km horizontally or a few km vertically). Such conditions produce large differences in wind speeds within an area that creates turbulence, the source of gusts. The space and time scales of turbulence and gusts are much smaller than can be represented explicitly by the models. Even the scales of thunderstorms are too small for the models.

This may have been said elsewhere already but I have not seen it. If anyone is aware of such descriptions, please provide a reference.

[therapy]

lhsmith,

I have been told that a hurricane can be there and not shown on a grib map.

That makes it pretty worthless as far as I can tell.