Sunday, April 5, 2020

Stargazing for Mariners: Northern Sky

We have a new Regiment of the North Star that can be used for finding latitude from Polaris without the official USNO annual nautical almanac. It can be used anytime you see Polaris, on any date or year. The method relies on identifying one of two stars, Alkaid at the tip of the handle of the Big Dipper and Segin, the trailing star of Cassiopeia on the other side of Polaris.

Both stars are trailing stars, meaning as the sky rotates, these are the stern of the constellation thought of as a boat sailing around the pole, once a day. They are also about equal distance from the pole and opposite to each other.

They are not bright stars. Alkaid at is the brightest of the Big Dipper stars, which as a group represent typical Magnitude 2 stars.  The leading 3 stars of Cassiopeia are magnitude 2 and the trailing 3 are magnitude 3.  Segin is the weakest of the lot at 3.3.

Because these stars are faint, we have to rely a lot on the pattern of the constellation. Alkaid is very easy, at the tip of the handle of the Big Dipper, but we need more help with segin.

Cassiopeia looks like an "M" from the Middle... meaning standing at the pole of the sky where Polaris is. Keeping that in mind, you can find this group, located about the same distance from the Pole as Polaris is on the other side.

Also the trailing loop of the "M" Cassiopeia is lazy, meaning not as tall, which also helps find this weak member of the group.

Other tips that help are these:

(1) Often the entire sky between Cassiopeia and the Big Dipper is blank except for two stars, and they will be Polaris and Kochab, one the Guards on the tip of the cup of the Little Dipper.  Kochab, Polaris, and Alkaid are about the same brightness.

(2) The line between Segin and Alkaid passes very closely through the the pole of the sky, and if you zoomed in you would see that Polaris is on the Cassiopeia side.

(3) We are talking about a large span of the sky between Alkaid and Segin. At middle latitudes they span from about NW to NE in bearing.

(4) Since the stars are faint, when any part of either constellation is within some 10ยบ of the horizon, they are likely to be extinguished by the (clear weather) atmosphere.

Wednesday, March 18, 2020

SCATSat — The Indian Scatterometer

At present we have 6 satellites providing scatterometer data, led by Metop-A, B, and C with their ASCAT Scatterometers. There is also the USN satellite Coriolis with the WindSAT instrument. Here I want to note the one from the Indian Space Research Organization (ISRO) called Oceansat 2, with its scatterometer called ScatSat. Valuable wind data from all five instruments are available in graphic format from the STAR Center  (Google "ASCAT"; it will be the first link.) Graphic scatterometer data are also presented by KNMI at the EUMETSAT's Ocean and Sea Ice Satellite Application Facility (OSI SAF).

Figure 1.  SCAT stats with satellite id info needed for tracking. Data from China's National Satellite Ocean Application Service (NSOAS) instrument HY-2B is only available at KNMI/OSI-SAF.

The data index pages from the STAR Center for five instruments are shown below.

The ASCAT-A, B, and C instruments have the same pattern; they just come by at different times.

We see immediately that the wide swaths of the scatsat data are very attractive, but we need some confirmation that the data are dependable.  One easy check is to see if they agree with the ASCAT data where there is overlap. The NWS highly regards the ASCAT data, which is used as input to model forecasts. A couple samples are below.

These two passes are about an hour apart, which is fairly close for this type of comparison. We see here very good agreement—also keep in mind the ScatSat is 12.5 km resolution compared to 25 for the ASCAT, so the details we see in ScatSat SW of Islas Marias could be real.

Another example:

These passes are at 0424 and 0451 respectively, which we read from the small purple times at the base of the diagrams.

Again we see good agreement and indeed we see the great value of the much larger data swath. Check out the Low in the bottom right, just missed in the ASCAT pass. We might even guess that this Low is moving north because of the stronger winds on the right-hand side.

And one more:

These passes are 0604 and 0451 respectively, 1 hr and 13 min apart. Not too bad for a rough comparison. We see wind directions good throughout and wind speeds seem consistent. The lower resolution ASCAT wind speeds are easier to read.

We can do a much broader comparison using a unique display tool at OSI SAF called Multi-platform Product viewer.  This tool does exactly what we want. We choose a location of interest where there is ScatSat data and then with the tool we can look at all passes at about that time from the ASCAT instruments. Below is just one comparison off the coast of Perth, Australia, taken about 30 min apart. I will try to make a short video demo of the tool and post it here.

Wind speeds are a bit harder to read in the KNMI data as they include a cloud overlay and the color codes are tied to the Beaufort Force scale. The green arrows are the ECMWF model forecasts for the times of the satellite data, which is another neat feature of this display. We should not be too surprised by the general agreement observed as the ASCAT winds are used to seed the model runs.  

In summary, this ScatSat data seems a boon to ocean navigation. You can get it the same way we get the ASCAT data underway with an email request to Saildocs. The key is knowing the file name to ask for. We have a graphic index to the file names in our textbook Modern Marine Weather.  You can figure these on your own from the main index page at STAR Center.  Just hover over a location to read the link. Then make a list.

The data images that cover the first 700 mi on the way from San Francisco to Hawaii would be this request to


This gets you the ascending and descending passes. When the images are full of data, the email is about 100 kB each.  When empty of data, about 15 kB.

A good way to learn the value of this data is compare these images to the GFS or other models multiple times a day for several days.  You will then be wanting to make your file list so you can ask for them at any time.

You can predict where the satellite will be using this link:

There are mobile apps and even screensavers that will track satellites that can be helpful.

Thank you India Space Research. Recall that we benefited from their OSCAT program for several years (2009 to 2014) when the amazing US QuikSCAT failed after a long and successful career.

[And a note of thanks to navigator Jeff Feehan, for directing me to the new Multi-Platform Product Viewer at KNMI, which was added here after the original was posted.]


Note added Apr 7.

Since publication of this note, we have had two emails like the one below, indicating that the data are offline, and sure enough when this notice comes we do not see data at the STAR Center.  In each case, we get another email the same day or next informing that the data are back online.  So this is just a heads up that this is not an official source of data.

For what it is worth, the ASCAT data itself as obtained from the STAR Center is also described after all these years as still Experimental.  But we have not seen outages of that data.

OSI SAF service message #2041
Sent on Tue, 07/04/2020 - 06:39 UTC

Title : ScatSat-1 winds unavailable

Message :
We are currently not receiving ScatSat-1 input data. The OSI SAF ScatSat-1 
winds (OSI-112-a, OSI-112-b) are therefore not available. Note that you can
check the current processing status on [1].

This service message has been published on OSI SAF web site :

This email has been sent to people who chose to subscribe to wind service
messages on OSI SAF web site.
You can update your subscriptions in your profile on


Saturday, February 29, 2020

Evaluating a Weather Forecast — Slides and Notes Only

with very abbreviated slide notes and references at the end.

• This talk is largely highlighting topics whose details are important, which are covered in     
  the book.
• Link to

• I used to say “but they are not marked good or bad," but this is no longer true as we now  
   have probabilistic forecasts for marine weather, as we shall see.
• We have lots of ways to check the timing of a system, mainly with pressure.
• If a forecast is in doubt then just do "half of what you want" until the next forecast in 6 hr
• We are not going to just download a GFS grib and make a decision.

• Print out each of the maps viewed side by side for quick overview 
• Both available underway by email request to Saildocs
• Learn something about the map AND something about the text 
• Here we learn the speed of the high moving E 
• We sail around Highs, so this is crucial info 
• First 3 days of GFS is usually good, but we need routes farther, so need extended forecasts. 
• Text forecasts (Storm Advisories) are especially important when dealing with tropical 
   storms. They are available from Saildocs.

• GFS overlaid onto OPC analysis viewed in Expedition compared to official Forecast  
• We want to use the digital data, so we use OPC maps to check them. 
• Compare at several forecasts. ie compare 06h forecast made at 12z with the 0h 
  forecast made at 18z
• Timing is everything. GFS 0-96h is about 4h, longer forecasts add an hour 
• Discussion tells which models were used and why 
• Both available underway by email request to saildocs 
• Auto load maps in Expedition and OpenCPN. One button click to get georeferenced maps 
   right on your nav screen 
• Let’s look at the zoomed area 

• Two GFS isobar parameters — needs a talk of its own! 
• Viewed here in LuckGrib, which has a tutorial with references on MSLET
• Saildocs switched to MSLET as has LuckGrib. Both still call it PRMSL to minimize  
  confusion in display options. WRF models also use MSLET.
• With MSLET we can better understand wind flow relative to isobars (gradient and 
• Standard PRMSL is effectively an average over ~80 nmi 
• Especially valuable around tropical stormsbut GFS is still not dependable there. 

• Accounts for patches of model wind not consistent with PRMSL isobars 
• Set isobar spacing to 0.25 mb and often see the trof of a front, or some indication of a trof       preceding a front (squall line?) 
• We can set the GRIB viewer or nav software to interpolate the model forecasts to
   the time at hand.
• This wind check process will be illustrated as we proceed

• New automated way to compare measurements to models and buoys
• Many videos and info online at
• Calibrated instruments are crucial, as is corrections for wind heights
• Designed for analyzing Expedition log files… but could be other sources.
• N2k gateways provide ways to log wind data from any system
• Our example only with a buoy 
• We added the error bars and redid the base figures
• Automated for buoys with internet connection. We have a note on how to use the buoy
   function underway. Buoy reports by email.

• Models and forecasters do not use all ship reports. Many subtle tests applied to the reports 
• Can get either map by email request. Only BW version can we get at 50% size reduction
   from saildocs.
• This bottom right corner or ridge is crucial to Hawaii races, rare to have reports here 
• Either one can be downloaded automatically, georeferenced in opencpn or Expedition. 
• We look at the ship reports on the surface analysis for consistency.  

• Recall nominal wind barbs are ± 2.5 kts
• SST > 80F for hurricane.
• Stable air would make it lower; very unstable only up 10% or so, not 14 to 25.
• Curvature could raise is about 10%. But not in this case
• Ship reports can distort the isobars... ie the forecast adjusts to account for the report
• Something to think about when not in agreement.
• Cannot rely on “wind&waves” maps. Sometimes lots on barbs; more often just
   those from the reports.
• Strange format to emphasize we use 80% and not the more standard 60%
• Practice with GFS wind and pressure to see how it works out.

• One click download of all buoy data in Expedition... super nice feature
• Mouse over shows what you would read at NDBC
• Procedure, download the buoys then set model time to match the buoy times,
   which may vary from one to the other.
• Expedition brings in latest set of buoy data but saves it when you re-load
• This will be a feature of the S-412 weather overlay to S-100 ENCs.

• NDBC stopped sending them to NWS. No reason given. No time for comments.
• No other NOAA source, hence moving weather  data delivery to commercial third parties
• Still in FTP instructions
• We can get the data from Saildocs,  one line per buoy
• Or subscribe specifying days to send, starting time UTC, and hourly interval
• Saildocs has a shortcut: send buoy51000.cur, which works for most buoys but not all. 
  The above link covers all sources as well as land stations such as wpow1.txt

• All reports within past 6hr within 300 nmi.
• Server checked on the whole minute and mail sent.

• Provides text report plus an attached GPX file of the data
• Includes range and bearing to report.
• Pressure in inches but we convert to mb in the GPX file

• Inset shows ship report compared to GFS at the same time
• Can load the GPX reports into any nav program. This example is OpenCPN
• Need to adjust GFS model time to match each report
• We can ID ships to projects when we get later reports
• Can spot isobar shifts. Use red data to figure where that isobar should be.
• Ships with full reports likely have better data… i.e., they are more careful.
• Isobars are moving west at this time.

• GRIB ASCAT available from Expedition and LuckGrib.
• Timing the main issue: GRIB versions are 15+ hours old; graphic index is updated
   hourly, and latest pass to show up will be ~ 2.5 hr old
• Latest useful pass could be 6 to 10 hr old.
• Generally get useful data about 3 times a day, since it does not have to be where
   you are to be useful.
• Bottom right is overlay with GFS in Expedition. ASCAT red, GFS black. Shows good  
   agreement east and west of Cabo, but very poor to the SE.

• File name index in Modern Marine Weather, the main reference for ASCAT
• Also ASCAT B and C… A plus B makes a somewhat broader swath. Mixing data
   about an hour apart.
• This is a most powerful tool, but it comes with a learning curve. See textbook.
• There is also another set of sat data from Navy WindSat, but it is not often used by NWS.

• Text discusses satellite passage apps to help predict when you get the next useful data
• This is rough sketch
• Added so you can read from the slides.
• Can plan out coverage before the race or voyage, so you know specifically when new data
   will be available and you can add it to your sources timetable.

• In the NH we sail around Highs
• Omega block can last 10 days
• If a blocking High, the map will look like this tomorrow.
• Else it will likely change
• We can get these 500 mb maps from Saildocs 

• Main reference on role of 500 mb maps is from Joe and Lee article. Online at the Mariners
   Weather Log and the book by Chen and Chesneau Heavy Weather Avoidance
• Centers of Action discussed in Modern Marine Weather

• See text for more discussion
• These general rules discerned from reading a lot of forecast discussions.
• Operative rule: there is no dependable rule; just guidelines.
• BUT now we have much more to work with… as we shall see.

• We will follow up on this example
• Let’s see what happens at about 72 hr out along a Pacific sailing route
• SD shows the variations between the ensemble members at each grid point
• A small variation means forecast not as sensitive to input variations and thus it is more
• Large SD means forecast  is not as dependable. Small input changes lead to large forecast changes.
• Now back to last slide on 500 mb indicating maybe weak forecast about 72 hr out

• SD of 2 kts is more or less consistent, ie ± 2 kts or so is agreement.
• At 44h, 68% of all winds will be within 12 to 16 kts
• Compared to 68% within 0 and 28 kts at 90h with divergence starting about 72h
• Out 84 hours out the SD blows up. Average is about same as control, but member 
   runs diverge a lot, meaning forecast is fragile.
• In short, after about 80 hr the forecast is likely in question.

• In stable winds we go out 5 days and still have small SD

• Another way to look at forecast dependability
• NBM Oceanic covers most of the globe to 20S
• Probably best extended forecast, meaning more than 3 days or so
• Maybe best on the ocean for all times. This is not clear yet. Could be GFS is best for first 3 days.
• Now look at another example first GFS only, then the blend, then the GEFS

• With GFS we see the wind and pressure over time. This is a determinist model. We learn the values, 
  but no uncertainties.
• We do see a veer when the trof  the the nW crosses at 66h
 BUT that is all we see!
• Now let's look at this same forecast with NBM

• From h0 to h41 wind is ± 2 kts, which is as good as it gets
• Then it blows up
• Forecast is weak after this point
• Now let's compare this same forecast with GEFS (ensemble)

• Here we see mean and control winds along with SD values
• When we see both GEFS and NBM indicate weakness or strength at about the same time
   and place, we are likely safe  in believing it.
• Next we look are some really new resource, just one week old

• This forecast (a blend of model outputs) added standard deviation (SD) just last week.
• Extends about 400 to 600 nmi offshore.
• We see notable light patch running NE of Guadalupe Island… but no more details.
• Could be best regional model… includes HRRR and NDFD. We need to study this one in
  comparison to HRRR, which we know works well most of the time to learn more about
  local  forecasting.  We have a note online on this topic. Use of Regional Models

• Another example of very new data, i.e., SD In NBM conus just last week
• SD in NBM v3.2 new as of Feb 20
• The light patch has good SD
• But before and after are very uncertain ± 5 kts of wind
• Next add lower valley marker

• Let’s look at this point on Google Earth
• Here we see what likely leads to the uncertainty in the forecast
• About 1,200 ft each side of 150 to 200 ft
• Channeling is very sensitive to upstream wind direction
• In short, SD has shown us where the forecast is weak… which we do not see from
  any deterministic model

• New resources that help us evaluate a forecast.
• Note.  In this study we found forecasts that differ quite notably can still yield about the same optimum routes, with different finish times.  So an eye to the polar diagrams can show how sensitive the route is to the actual wind direction. But as stressed many places, the polars have to be right, or the routing has little meaning, if not a distraction. Also we obviously do not want to take risks for small gains, even though the optimum route will do so, namely go for whatever makes it faster. Thus again the value of the statistical forecasts.

Speaker's related sites:

Frank Bohlen's articles on the Gulf Stream

Ocean Prediction Center (, Jos Sienkiewicz, Branch Chief)

Locus Weather (Ken Mckinley's weather services) (Stan Honey and Sally Lindsay Honey's 's articles and videos)

Seattle Forecast Office (Kirby Cook, Science and Operations Officer) (David Burch HQ)

References cited in discussions:

Use of Regional Models

Inverse Barometer Effect

Free Marine Barometer app (iOS and Android)






NOMADS (NCEP source of model data)

How to Obtain Custom Grib Files

How to Combine Grib Files