Overview of Weather Models for Transpac Sailors
There are a variety of weather and ocean forecast models whose data can be acquired for free. Below are descriptions of some you are likely to encounter, which will hopefully demystify them a bit.
1. Global weather models
GFS (Global Forecast System)
The Global Forecast System, or GFS, is the US’s primary global weather forecast model. It is global in the sense that it calculates the state of the atmosphere and how it changes everywhere on the planet at every time step, and in that it focuses on large scales. Its forecasts are used by weather forecasters at the National Weather Service and disseminated in many ways. It is an old-school model in that it represents the atmosphere in an abstract way, as sine and cosine waves, rather than on a grid as the modern global weather models do and as all regional models do, when it makes calculations. Of course this is obscured to the user, but it influences the forecasts made by the model.
By some measures, the GFS isn’t as good as some of the flagship weather forecast models from other countries, like Europe, the UK, and Canada (see Cliff Mass’s blog http://cliffmass.blogspot.com/2014/04/the-us-slips-to-fourth-place-in-global.html). Hopefully we can turn it around with more computing power and focus on improving the model.
The latest version, 12.0, has a horizontal resolution around 13 km (7 nmi, the exact resolution varies with latitude) out to the 10 day forecast, and then 35 km (20 nmi) from 10 to 16 days. In previous versions, the switch to lower resolution happened at a week instead of 10 days. It uses a much higher resolution SST observational dataset, 5 minutes instead of 1 degree, but because they are observations they are only available before the forecast time.
GFS Ensemble Forecast System (GEFS) and North American Ensemble Forecast System (NAEFS)
An ensemble forecast is a set of model integrations with the same model, each with slightly different initial conditions or model physics. The goal of an ensemble system is to provide a measure of how certain the forecast is. The more similar the simulations are, the more confidence we should put in the forecast. The range or spread across the model integrations is how you can accomplish this (see NCEP’s GEFS-MNSPRD). It can also visualized with spaghetti plots (see NCEP’s GEFS-SPAG), where one contour is chosen and plotted for each integration. The messier the spaghetti looks, the less certain the forecast is. It is a fun exercise to step through a loop of spaghetti plots. It will start out very clean and smooth and get messier into the future, as the ensemble members diverge and what we can know about the future state of the atmosphere diminishes because of the chaotic nature of atmospheric motions. This extra information about forecast uncertainty comes at a cost, and that cost is resolution, since NOAA has limited computational power.
The GEFS is an ensemble of 20 runs of the GFS model. The resolution of the GEFS is 55 km (30 nmi) for the first 8 days, and then reduced to about 80 km (40 nmi) out to 16 days.
The NAEFS is a joint venture between the meteorological services of the US, Canada, and Mexico which began in 2004. Two ensembles of 20 members each, the GEFS and an ensemble run by the Canadian model, are combined along with statistical adjustment incorporating observations to produce forecasts out to 14 days.
Non-US global models: ECMWF, UKMET, CMC, NOGAPS/NAVGEM
Nearly every national weather service runs its own global weather forecast model. Since it’s crucial to make accurate weather forecasts, most of these models ingest similar observations, but the way that they do it is slightly different, and each model’s physics and other details differ as well. One might be inclined to ask which is “the best” and rely on it, but this is probably not the best approach. All of these models can hold their own against the US models, at least in many situations, so it’s difficult to say which is the best in general. On any given day, it is worth comparing the analysis (which is in the past, so it can be verified) to see which model integration does the best job at capturing what we know has already happened. Aside from that, weather forecasters often treat the different models as an ensemble (see above) to get an idea of the range of possibilities of how the weather will unfold, and how much certainty to have in the forecast.
The US Navy runs its own global weather forecast system, called NAVGEM (US NAVy Global Environmental Model), formerly NOGAPS (Navy Operational Global Atmospheric Prediction System), at the Naval Research Laboratory. This model can be used just like the other global models. Since it’s run by the Navy, it also includes some ocean surface fields that are neglected by strictly atmospheric models like the GFS.
2. Regional weather models
NAM – North American Mesoscale Forecast System
The NAM is the flagship regional weather forecast model run by NCEP/NOAA. It is regional in the sense that the atmospheric state is only calculated on a subset of the whole globe, rather than for the whole global, like the GFS. Also in contrast to the GFS, the NAM is built so that it can explicitly calculate smaller-scale phenomena that produce large vertical motions (that is, it does not assume all motions are hydrostatic).
The NAM has nested grids. That means it has one coarse grid covering its entire domain, but then finer grids focusing on regions of interest. The coarsest grid, covering the whole domain, is at 35 km resolution. There are grids covering the Pacific and the continent at 12 km resolution, and higher resolution grids over some land regions. When hurricanes develop, nested grids focusing on the hurricanes also run.
There are other regional atmospheric models. In the US, the main regional model is WRF (Weather Research and Forecasting model), run mostly for research, rather than operationally at a variety of institutions. WRF replaced the MM5 a few years ago.
3. Models with more of an oceanic focus
COAMPS (Coupled Ocean/Atmosphere Mesoscale Prediction System)
Another regional model, which also incorporates the ocean state, is COAMPS, run by the US Navy. It is initialized by NOGAPS/NAVGEM.
WW3 – NOAA WaveWatch III model
Unlike the other models, which calculate the state of the atmosphere, the WaveWatch model calculates ocean waves. As an input, it takes the near-surface winds from the GFS model. The plain WW3 model is global.
WW3-ENP WaveWatch III regional Eastern North Pacific model
This is a regional implementation of the WW3 model focused on the eastern north Pacific waters.
GFS (and GEFS, NAEFS)
14.3 COMMUNICATION RESTRICTIONS
Competitors may only utilize weather information that is routinely available throughout the year to the general public without charge, and whose availability is publicly indexed. For example: Competitors may NOT arrange for routers or meteorologists to provide them with advice, custom data, or compilations of
public data during the race, no matter how that information is communicated. Competitors may receive regularly scheduled weather broadcasts or weather fax transmissions (e.g. from NOAA, USCG, WWV, NMC, KVM70). Competitors may receive imagery from satellites (e.g. NOAA, APT satellites).
Competitors may use any means to retrieve data from the Internet (e.g. from the web, from ftp sites, from email responders), provided that those data are intended for public use without charge, are routinely available for free throughout the year, and are publicly indexed (e.g. can be found via Google).
Prior to their preparatory signal, there is no limitation on private services or any other source of data or consulting, except that a competitor that has started may not provide weather information to another competitor that has started, or to a competitor that has not yet started except through the information provided to or from Transpac Race Communications.
This amends and clarifies RRS 41 (c), which states:
41 OUTSIDE HELP
A boat shall not receive help from any outside source, except
(c) help in the form of information freely available to all boats.