Expedition Expedition Navigation Software

You are not logged in. Would you like to login or register?

5/19/2019 9:38 am  #1

How does Model Accuracy actually work?

Model Accuracy allows a navigator and crew to instantly understand the strengths and weaknesses of particular GRIB sources (ex: GFS, EC, COAMPS, Predict Wind, ect…) and make a more educated and confident decision on which GRIB source to trust the most. The utility of tracking macro and micro shifts in TWD and watching the shifts inside frontal systems is invaluable in understanding the power of each GRIB source. The ability to have absolute confidence in the primary GRIB which you are routing your boat off of is in fact the best GRIB available is truly intangible. It allows the navigator to spend less time in the nav station and more time resting or on deck.

It allows the crew to know the heading and tactics are based off the best information available and nothing was left to just a “gut feeling” or chance.Inside the graphical plots and results pages the user can identify the “tendencies” of each GRIB (forecasting under/over speed for TWS or TWD left or right of actual captured data). Understanding the EXACT tendencies over a selected time period will allow the navigator to correctly twist TWD and raise/lower TWS inside the optimal routing functions of other sailing software. This way the projected optimal course can always be based on not only the most accurate weather file, but the most accurate weather file that is correctly “calibrated” for the current and past forecast errors. 

There are a plethora of GRIB sources, all claiming they are the best source for you and your team to trust as you cross an ocean. Finally there is an application which provides a metric for the consumer, the absolute end user, which ranks the  accuracy of different GRIB sources and exposes their successes and shortcomings. This software was made strictly for you, the sailor.

Professional sailing teams spend massive amounts of time and money constantly refining the VPPs (polars) of the boat to ensure the optimal routing software has the most accurate projected performance. Yet every “optimal course” has two variables in the solution, the boat’s VPPs and a weather GRIB. Until now there has been no way to formally analyze the accuracy while on the water of the second 50% of the optimal solution, the GRIB. Without reservation a team is now able to know they are pairing their best VPPs with the best GRIB possible for the optimized route. Every sailing team always wants to know they have the absolute best VPPs loaded into their routing software, now they can know they have the absolute best GRIB as well. 

The Model Accuracy software provides levels of confidence in GRIB predictions by using statistical and other analytic techniques to evaluate GRIBs in the context of TWS & TWD forecast performance. The analyses are time and location sensitive, and take into account variations in GRIB data granularity. Mathematical techniques are used to clean the "noisy" data that is often encountered in ocean deployment, so as to provide consistent predictions that can be relied on for tactical decision making. The predictions are delivered to the user in numerical and graphical format, thus providing both very precise prediction information as well as easy-to-comprehend overviews that raise the user's understanding of the prediction.https://static1.squarespace.com/static/542eec32e4b0fc2c609ef58b/t/543142dce4b0e71434177889/1412514560551/?format=1000w
TWS Graphical Plot:The graphical output shows how the predictions from the various GRIB sources correlate with the observations from the boat's instrumentation.The GRIB sources are color coded, and the tagged with markers to indicate GRIB deliveries (green squares) and GRIB prediction time points (red circles). This example illustrates the output for True Wind Speed, and a similar graph is provided for True Wind Direction.https://static1.squarespace.com/static/542eec32e4b0fc2c609ef58b/t/5431445de4b0fb15d647725f/1412515031061/?format=1000w
RMS: Root Mean SquareThe numerical output from the analyses provides several quantitative indicators of the reliability of the GRIB predictions, for each GRIB source. These numbers are summarized into a final "ERROR" value, to allow navigators make a quick decision of which GRIB source to trust, when there is no time to consider the component values.


5/19/2019 9:38 am  #2

Re: How does Model Accuracy actually work?

Check out the new tutorials for Model Accuracy!


     Thread Starter

6/30/2019 12:59 am  #3

Re: How does Model Accuracy actually work?

Interesting development I have been trying to get from other developers... without success! Which makes me very curious about your software I just discovered via the Expedition forum. Do you know whether you are (still!) the only ones to do this so far?

A few technical question:
1. Why use TWS and TWD and not Ground Wind data, which is what the forecasts provide?
2. What corrections do you apply to TWD and TWS before comparing to forecasts?
3. Is your correlation "static", simply cross-correlating (adjusted) boat measurements and forecasts along the boat course, at the time she was there? Or do you allow for time and space shifts as forecast may be accurate wind-wise but just offset in time by a few hours or in space by a few (tens of ) NM?
4. Do you have the ability to work with Ensemble data sets?

Last edited by JL (6/30/2019 3:12 am)


6/30/2019 10:08 pm  #4

Re: How does Model Accuracy actually work?

Good questions JL!
Looking forward to the answers.


Board footera


Powered by Boardhost. Create a Free Forum

Zapfware North University. Racing Tactical Software - Expedition.
Online courses