You can bet the nuclear submarine USS San Francisco had a full range of electronic charts and plenty of watchstanders on duty when it slammed into a Pacific seamount in January. This was not a navigational error in the traditional sense of not knowing where a boat is or believing it is somewhere it isn’t. I’m sure the officers knew exactly where they were and were probably quite relaxed running at 500 feet in waters charted at 6,000 feet. It was their rotten luck the surveyors had missed an underwater mountain!
This is not the first time something like this has happened. In 1973 MV Muirfield suffered extensive damage to its keel when it struck bottom in an area of the Indian Ocean charted at 5,000 feet. A subsequent survey revealed an uncharted seamount 1.5 miles wide. Closer to home, in 1992 RMS Queen Elizabeth 2 hit an uncharted rock off Block Island.
The extent of bottom coverage is one issue. The level of accuracy with which data is collected is another. Pre-GPS inshore surveying had a customary positional accuracy within 1.5 millimeters times the scale of the survey. Offshore (out of sight of reference points on land), positional accuracy was far less.
What this means is that given a survey at 1:20,000, the level of inshore accuracy in charting soundings and features would be within 1.5 x 20,000, which is 30,000 mm or 30 meters. It was common practice to survey at twice the scale at which the data was used — e.g., a 1:20,000 survey would be used to construct a 1:40,000 or smaller-scale chart. So now we have a level of accuracy of +/-30 meters on our 1:40,000 chart. Thirty meters at this scale is 0.75 mm on the chart. Given traditional plotting techniques, no navigator could plot with a precision greater than this, so all the pieces fit together.
With the introduction of GPS and electronic charts, the pieces started to fall apart. Any modern off-the-shelf GPS has a positional accuracy of 10 to 15 meters. If you add wide-area augmentation system (WAAS) corrections, this goes to 2 meters. So long as plotting is done on paper charts, the fact that the GPS is more accurate than the placement of details on the chart is more or less irrelevant because the accuracy of a fix would continue to be determined by the accuracy with which it could be plotted. However, electronic charting with electronic position plotting provides a degree of precision that the underlying chart data simply does not support.
Current International Hydrographic Organization survey standards call for a positional accuracy of +/-2 meters for all critical soundings (harbors and channels), and +/-5 meters for most other inshore soundings, with less accuracy required offshore. Although these standards are no better than a WAAS-corrected GPS, the pieces fit once again. Unfortunately, this is somewhat irrelevant, because hydrographic offices worldwide have shifted resources into digitizing existing databases without much new survey work!
of a paper chart on a screen at the same scale as they are shown on the chart, you won’t be able to read them. At the electronic chart display and information system (ECDIS) level, which sets the standards for electronic charting on big ships, the industry seems to have settled on a zoom level of 1:1.7 for making electronic chart displays readable. This is to say a paper chart at 1:20,000 is displayed electronically at a scale of 1:20,000/1.7, or 1:11,765. At the recreational level, for reasons I do not have the space to explore in this article, the relationship between a paper chart and its electronic display is often 1:6.25. That means a 1:20,000 paper chart is now displayed at a scale of 1:3,200.
Let’s consider an area of an original survey that was done at 1:10,000, with a positional accuracy within +/-15 meters. This area was found to be foul with rocks. The chartmaker, working at a scale of 1:20,000, showed a couple of these as close together as possible using traditional drafting techniques, which is to say about 2 mm apart (representing 40 meters in the real world). No sane navigator plotting on this paper chart would try to take a vessel between these displayed rocks. But now this chart gets digitized and displayed at 1:3,200. The space between the rock symbols on the display has just increased to 2 x 6.25, or 12.5 mm. All of a sudden, it looks like I can take my boat between these rocks, especially when its position, based on my WAAS-corrected GPS, is displayed with pinpoint accuracy. Unfortunately for me, there is no gap between these rocks — just another rock the chartmaker could not show at the original chartmaking scale of 1:20,000.
Overzooming a chart — i.e., using it at a scale for which it was not designed — is one of the cardinal sins of navigation. All electronic charts are overzoomed to some extent, with those found in the recreational marketplace typically grossly overzoomed. To compensate for this, in theory, any navigator using electronic plotting should place an imaginary circle of possible error around the boat’s plotted position. This circle should have a radius equal to the allowable error that was used in plotting the features on the chart. Note that this circle of error does not represent errors in plotting the position of the boat, which will be phenomenally accurate, but instead represents the extent to which the features around the boat may be out of place on the chart. All potential hazards should be kept outside the circumference of this circle, or, put another way, if any comes inside this circle, we are clearly into the realm of eyeball navigation.
So how do we determine the radius of this circle? We need to know two things:
Unfortunately, neither is readily available for most electronic charts supplied to the recreational marketplace!
Many (but by no means all) paper charts have the first part of this information included in a source diagram, and the scale is, of course, clearly stated beneath the chart title. A source diagram is typically a small inset chart that breaks up the area covered by the chart into sections according to the age of the data on which each section is based, the scale at which this data was collected, and the method used to collect it. When navigating in any of the waters covered by the chart, a quick glance at the source diagram will provide a good sense of the accuracy and reliability of the underlying data for that section of the chart.
present. This gives us some sense of the accuracy of the source data, which is the first bit of information we need to develop our circle of possible error. Now we need to know the scale at which the chart is being displayed.
Unfortunately, the relationship between electronic charts and display hardware in the non-ECDIS marketplace (i.e., recreational boats) is such that though there may be a display scale given in the chart header or elsewhere (e.g., 1:20,000), the system does not in fact know at what scale it is displaying a chart. The displayed scale is purely nominal and is generally far removed from reality.
As such, even if you find a source diagram and can determine what would be an appropriate radius for a possible error circle, it will be very hard to create the circle of possible error unless there is a tool that allows you to specify a circle with this radius around the displayed position of your boat. (For example, if the source chart were based on a survey at 1:20,000, have the software place a circle around the boat with a radius representing 30 meters.)
The difficulty of determining a suitable circle of possible error gets ratcheted up a notch with quilted charts. Quilting is the process of electronically merging charts so the navigator can move seamlessly from one chart to another. But the source charts may have been made at different scales, based on surveys with differing levels of accuracy. In order to quilt them, the electronic chart manufacturer brings them to a common scale, generally by enlarging the scale of smaller-scale (and inherently less accurate) charts to that of the larger-scale (and inherently more accurate) charts. The only visual clue for the navigator is a change in font sizes (the labeling and soundings derived from smaller-scale charts will be magnified compared with those from larger-scale charts).
The difficulty of determining a suitable circle of possible error goes up a couple more notches with all vector-based charts. Any time you see enlarged labels or soundings on a raster chart, this is a visual clue that the underlying data may be less accurate than data on other parts of the chart, but you get no such clues on vector charts.
Vectorizing is the technology that underlies all cockpit chart plotters and many other electronic charting systems. It starts with a raster image, and then all the data — outlines of landmasses, individual soundings and features, etc. — is converted to a mass of points that are stored in the electronic file on the basis of their position, with an attached file that describes the feature represented by that point. When the computer displays the chart, it places all the points on the display screen. Continuous lines (e.g., those around landmasses) are created by connecting the relevant dots. Features (e.g., a buoy) are created by placing the appropriate symbol and label at the relevant point. And so on.
The benefits of vector technology (see sidebar above) have a significant negative consequence on a navigator’s ability to get a sense of the accuracy of the displayed data. For example, there might be information from a 1:80,000 chart (with a likely survey accuracy within +/-60 meters) quilted with information from a 1:20,000 chart (+/-15 meters), but because all fonts are the same size, as the navigator moves from one survey area to the other, there are no visual clues as to the changing scale — and probable accuracy — of the chart or survey from which the data was derived. What is more, the font size on a vector chart remains constant at all zoom levels, providing no visual clues as to when the chart is being grossly overzoomed, unlike a raster chart, on which all fonts steadily enlarge the more the operator zooms in until they start to pixilate.
At the ECDIS level, all information on a vector chart is required to be tagged with a metadata file that can be queried to determine the source scale of the information, and there are various safeguards against overzooming. Many vector charts found in the recreational marketplace have no such metadata files and safeguards.
Electronic chart datums
In order to convert survey data into a chart, you need an accurate mathematical model of the shape of the world in the region of the survey. Over time, such models, known as geodetic datums, were developed for all the continents and subcontinents. Virtually all electronic charts have been converted to the World Geodetic System 1984 (WGS 84) datum, in which case a GPS should always be set to this datum when used in conjunction with these charts.
However, the British Admiralty’s reluctance to convert some charts to WGS 84 points out a problem. In some parts of the world no matter what algorithms have been used to make the conversion from existing chart datums to WGS 84, they are not wholly accurate. The conversion process introduces a new layer of errors of unknown magnitude with no mechanism for alerting navigators.
And sometimes there are outright mistakes. On a recent cruise in the West Indies from St. Lucia to Bequia, using latest-generation electronic charts in a chart plotter, we found fixes in St. Lucia were pretty accurate; those along the coast of St. Vincent were consistently off by a half mile in terms of their longitude; and those in Bequia were once again pretty accurate. The track of our chartered boat took us over the hills of St. Vincent and through a church and fort! My guess is that when the chartmaker converted the chart datum of the source paper chart into WGS 84, a simple mistake threw everything off by half a mile. This kind of thing is a bit shocking.
The electronic dream
The current reliance on old chart data, and algorithms that convert from multiple datums to WGS 84, is a far cry from the potential that existing technology offers us. Surveys can now be carried out at a far greater pace, with a far higher level of detail and accuracy, and at a much lower cost than could ever be done before. Experimental surveys of extensive areas have already been done to a 2-centimeter level of detail and accuracy.
The vast amounts of data that can be collected can then be stored in an electronic database that creates a nominal 1:1 relationship with the world. Because the data is stored electronically using the fixes attached to each piece of data when it was collected, no further drafting errors are introduced &mdash the level of accuracy with which the data was collected (e.g., 2 cm) is retained in the database. In theory, all navigators could be given access to all this data with software that selects appropriate details for any desired display scale (i.e., the more you zoom in, the greater the level of detail that gets displayed &mdash NOAA funds an experimental unit at the University of New Hampshire that has developed algorithms that can do this). The underlying accuracy at any zoom level then remains that of the original survey.
This is the ultimate dream of electronic charting &mdash what was once thought of as a Worldwide Electronic Navigational Database. Experimentally, the UNH folks have gone one step farther, introducing real-time tide corrections into electronic charts: all the soundings on a chart change as the state of the tide changes. If you input the boat’s speed and project its position to some other point on the chart, the software calculates how long it will take to get there and displays the actual depths you will find, adjusted for the state of the tide that will prevail when you arrive!
The WEND and associated developments require a massive worldwide surveying effort and a willingness on the part of national hydrographic offices to share their databases and other information. Unfortunately, in the rush to get into electronic charts, hydrographic offices have shifted resources out of surveying into digitizing existing databases, while disagreements over copyright issues and mechanisms for sharing data are as bad as ever. The WEND is a long way off!
Living in the present
As frustrating as it is to think about the unrealized potential in electronic charting, what we have is still spectacular technology. Most of the time the existing databases are remarkably accurate considering the tools with which much of the data was collected, and even low-end electronic chart plotters give us a level of navigational precision that was more or less unthinkable in the age of paper charts.
However, it is precisely this level of accuracy that can lull us into a false sense of security concerning the underlying data on electronic charts and the technologies used to create them. Then there are changes that humans or nature (the Indian Ocean tsunami, for example) may have wrought in the years since the data was collected.
The greater the degree of precision with which a boat’s position can be plotted &mdash and today, with WAAS-corrected GPS and electronic plotting, this is generally on the order of +/-2 meters &mdash the more important it is to maintain a healthy skepticism of the accuracy of electronic charts. Rather than a single point, always think in terms of a circle of possible error around the displayed position of your boat.
USS San Francisco, MV Muirfield and RMS Queen Elizabeth 2 unfortunately learned this vital navigational lesson the hard way.