GPS Chart discrepancy article

[Roger Long]

The issue of "Points East" just now hitting the stands has an article about the significant chart discrepancy discovered at or near the spot where the Portland, Maine fireboat ran aground. If you aren't in New England or want to read it on line, it's now available at http://www.pointseast.com
If you want to see the charts, you'll have to read the full version of the issue available in the link at the bottom of the page instead of the feature story version.

I've had some interesting email exchanges on this with the owner of a navigation software equipment company who is on an industry/government charting standards advisory group and the fellow who wrote the article sidebar. I just got an email from a product manager at West Marine is found it very interesting and said is going to talk to their GPS suppliers.

There is going to be more discussion about this soon on the "Points East" forum:

http://www.pointseast.com/forum/index.php

[StevenBauer]

I was in Hamilton Marine this afternoon and Points East wasn't on display yet. :(



Steven

[paladin]

What did they use to make the chart.
Charts are normally generated from old maps made using the theodalite or similar devices....GPS coordinates will vary by as much as a half mile in some locations. To do a proper chart, every sq. foot of real estate would have to be re-mapped using GPS. Never use a gps unless you have been to that precise location and previously recorded the coordinates and made the corrections on your chart.

[George Ray]

This is a very interesting post. To me, the title implies an issue related to GPS accuracy or datum mis-match, etc while the underlying issue seems more one of cartography both digital and otherwise that lacks Quality Checks. I am not aware that GPS is subject to geographic anomalous badness inaccuracy like loran so if there is error is seems it's the cartographer or user datum error. I have been anchored in the Canaries and looked at an old C-Map chart on a PC and it showed us on the land but other wise have not noticed error in plotting.

Wonderful wakeup call you have given us and on a similar note it is wonderful that NOAA is distributing their digital charts (raster & vector) at no charge. One of the challenges that comes out of that however is keeping them updated. I have been using Coastal Explorer charting software for a while and I just love it when I ask it to check for chart updates and it finds a bunch. Now if they would only tell me what the updates were. I have corresponded with the company and was told that currently NOAA does not provide much useful info with the electronic updates to inform users but it is not lost on them that it would be useful and important information.
How many of the charts plotting programs are doing such updates I wonder. Doing it by hand is not a job I would want.

GPS chart plotter cartography such as Garmin Blue Chart just to name one company (I have some blue charts). Assuming they have this same error I wonder if you can get Garmin to issue updates for free to anyone that has an older version with such an error?

It is way too easy to put undeserved trust in electronic gizmos.

[KAIROS]

As the NOAA 'straight dope' sidebar on the last page of Roger's article says, the vector versions of charts which allow for some beneficial features in chartplotters, are derived from raster charts. In other words, these digital vector charts are not the same as the digital raster charts.

Although the kind of accidents this might cause are no good, neither version of the chart can be said to have an error. The difference between the two is just a result of their being two different presentations of the same information.

Even a digital 'copy' of a paper chart is not quite the same as the original. But, when you convert that raster copy into vector information for use with a chartplotter, watch out.....as is indicated in this case.

This is a perfect example of how these two formats (raster vs. vector) can be different even starting from the same source. The raster version is like an identical twin (slight difference), while the vector version is like a fraternal twin (different).

[KAIROS]

Hey Roger, do you have the text from that NOAA sidebar to post here? It is a pretty good blurb. I can't seem to copy it from the digital online version of the article.

[Roger Long]

The difference between the two is just a result of their being two different presentations of the same information.

Not in this case. Take a close look at the charts. The vector chart shows the beacon sitting squarely on top of the 18 ft MLW contour. The 6 ft MLW contour is actually 120 feet out from it in the narrow passage. The recent relocation of the beacon was inconsequential.

It's hardly the same information.

[KAIROS]

Not in this case. Take a close look at the charts. The vector chart shows the beacon sitting squarely on top of the 18 ft MLW contour. The 6 ft MLW contour is actually 120 feet out from it in the narrow passage. The recent relocation of the beacon was inconsequential.

It's hardly the same information.

The presentation.....the way it looks...is definitely different. Maybe a better way for me to say it is that by definition the vector data is a generalized version of what is seen in the raster version. Instead of that contour line being defined by 300 points per inch in the raster version, the line is represented by several points in the vector version (you can see the points connected by the straight lines in your zoomed image). In this case, with the sharp bends in the original contours, more points are needed to define a reasonable shape.

The vector data is generalized so that it can be stored in a reasonable (disk) space and can be displayed quickly using low-end graphics equipment (a chart plotter). Scrolling through the original raster version would be slow and the data would cost more (and many of the features used to market chart plotters would not be possible).

As I mentioned, the NOAA sidebar on the last page of your article summarizes the issue pretty well. If you go into it much deeper it gets ugly.

The groundings you point out are one reason why I don't use a chartplotter or vector data for navigating. You can't have the detail of the original chart in vector form in a chartplotter with current technology at reasonable expense.

This has nothing to do with GPS error either [meaning the error inherent in a position from a gps]. It's all about the difference between the chart and reality.

[KAIROS]

Here's what that NOAA sidebar from Roger's article looks like.....maybe legible.

http://www.ducksoupinn.com/dope.gif

Thanks for this, Roger. It reminds me to keep using paper charts....

[Bruce Hooke]

The presentation.....the way it looks...is definitely different. Maybe a better way for me to say it is that by definition the vector data is a generalized version of what is seen in the raster version. Instead of that contour line being defined by 300 points per inch in the raster version, the line is represented by several points in the vector version (you can see the points connected by the straight lines in your zoomed image). In this case, with the sharp bends in the original contours, more points are needed to define a reasonable shape.

The vector data is generalized so that it can be stored in a reasonable (disk) space and can be displayed quickly using low-end graphics equipment (a chart plotter). Scrolling through the original raster version would be slow and the data would cost more (and many of the features used to market chart plotters would not be possible).

It looks to me like there is something more going on here. If you look NE of the spire the vector data shows quite a bit more detail than the raster in terms of the depth contours. So, if it was just a matter of how many points are being used to record the position of a line, I don't think we would have the problem we are seeing here. To me this points to an issue with how the raster data are being converted to vector data or some other actual discrepancy in the compilation of the data (I am not sure what format the actual original source data are in).

However, where things get really interesting is when we overlay the vector onto the raster and make the vector layer partly transparent. See the image below (Roger, if you have any copyright concerns about what I have created here from your work, let me know and I will delete it). If you look closely you will see that the nun and the red dayboard match up exactly, as do various other things, but the green spire on the vector chart is closer to the middle of the channel than on the raster chart. This overlay also shows how bad the mistmatch is between the depth contours. This, to me, confirms that this is not just a matter of simplification to make the raster chart data reasonably compact, but an issue with how the conversion was made.

I just did some scaling and it looks to me like there is a difference of about 60 feet in where the two charts place the spire, and this cannot be explained by the relocation of the spire as the spire was reportedly moved ENE and the difference we are seeing here is pretty much perpendicular to that. There is still a difference in the depth contours, but if nothing else, if the spire was shown on the vector chart where it is on the raster chart, that depth contour well outside the spire might have warned some folks to look out for shallow water on the channel side of the spire.

You may notice that the shoreline to the south of the spire does not match up either, which caused me to suspect that maybe the vector version had been compressed vertically. However, I tried stretching it until the shoreline matched up and the spire was still about 45 feet or so from where it is on the raster chart.

http://www.bghooke.com/brucehooke/images/chartcomparison.jpg

[KAIROS]

Excellent comparison....

After checking the chart scales and some other literature, I think the main problem here is that we're expecting too much from charts having an original scale of 1:40,000. A 50-foot error is not unreasonable at this scale when comparing a charted position with the actual position. However, this discrepancy noted in this thread is between two charted locations. I don't know what the acceptable error is in this case for NOAA ENCs. Whatever it is, add that onto that 50-or-so feet to give a good idea of how accurate a vector chart is.

Roger points out an excellent example of when the best source of information is essential. And that would not be the vector chart in a chart plotter. Keep the original paper chart handy. It's the most likely to give the best information. Especially in tight little passes like this where the acceptable error range become so significant.

[sorry if i seem to be pontificating here; it's because i know too well how that data gets on the original chart and in the digital vector version...i've made hundreds of digital charts....i use paper charts when navigating]

[George Ray]

Summarize: ( one inch line and a single ATON )
RASTER: 18,000pts
A picture made up by points printed on page and the points in themselves have no meaning and very little size. So, it takes many point to create something. For example at 600dpi (a) it takes about 6,000 points (dots) to give 1" long line on the chart 1/60" width and (b) an ATON may take 12,000 points (swag) in varying colors for the little image and the text. So now we have a one inch line and one ATON (Aid to Navigation) and have used about 18,000 points. So a single point has little meaning, it takes hundreds before you can even see something.

VECTOR: 31pts
A database of points, but with a big difference. Each point has meaning. A single point can be an ATON with all it's color, text, visibility regions, etc. The one inch line only requires about 30 points to appear nearly as detailed as the raster and to add to the goodness the database entry for each point has information about is location and elevation (depth) and bottom composition. The points and lines can then be displayed on the screen of your choice. ( I choose an iPad in a ziplock bag.)

How we got here:
I think most early vectors were created by a draftsman (not a cartographer ) sitting at a huge digitizing tablet with an overlaid raster chart and he-she is click-click-clicking along contour lines and on the ATON's. Easy to imagine that relationship troubles or indigestion or first thing Monday morning or end of the day on Friday could lead to a mistake. the question is not as much why is there a mistake rather why so few?

Where we go next:
I suspect that survey work electronically/digitally integrates GPS position and measured data so that it just sort of flows right onto the database and hence onto the plotter screen. Far fewer problems with errant clicks of the digitizer.

Question(s):
Is the issue/error under discussion here one of fundamental data error or final display error?
Is the MFD multi function display or the underlying software that massages the vector data on the way to the display faulty in some way?
Has anyone or any company been identified as the source of the bad data or bad display and then notified?

[KAIROS]

I suspect that survey work electronically/digitally integrates GPS position and measured data so that it just sort of flows right onto the database and hence onto the plotter screen. Far fewer problems with errant clicks of the digitizer.


I have dreamt this many times. The reality is that it does sometimes happen like this, but sometimes not, sometimes really not, and sometimes the data are ugly and recalcitrant. It's sometimes more a black art than science. There are SO many variables and potential errors to account for.

When you see airplanes successfully navigated (and land and take off) only by computers using digital aeronautical charts, then maybe imagine navigating with only gps and digital nautical charts being really failsafe.

NOAA does outstanding quality control though. And, on this 13290 chart the sources of data are surveys less than 10 years old for Whitehead Passage. This is not typical.....probably a very high priority region near Portland. This is one of the better cases.

Maybe NOAA would fix this 'error' in the vector data, or maybe it falls within acceptable limits for chart accuracy at this scale. It looks like a spot that needs an inset map on the paper chart, and the more detailed data would generate a more representative vector chart.

Clearly a vote for keeping the real paper or raster chart (digital copy) handy and referring to it regularly. And don't zoom in too far on those digital charts. Maybe you can set a zoom limit scale on your chart plotter so you will less likely be deceived by that mesmerizing technology.