Beyond ENC

S Salter and A Watkin

Simon Salter, CherSoft Ltd, The Point, Bradmarsh, Rotherham, UKS60

Andrew Watkin, CherSoft Ltd, The Point, Bradmarsh, Rotherham, UKS60

This paper discusses contemporary issues with ENC data and display systems. It identifies several areas where experience has exposed deficiencies in the use of ENC. We then go on to suggest several ways of improving cartographic data handling in the marine environment.

Some of what we argue may seem counter-cultural since our experience of software development in the ENC context has given us an awareness of its limitations as well as its advantages. We want to see ENC continue to develop and for this reason we feel that it may be time to take a fresh look at some issues before they become needlessly ingrained in our collective consciousness.

There are a number of issues that fall into this category. In this paper we discuss just two: the importance of cartographic input to a charting system, and the potential advantages of a more dynamic and situation aware approach to symbology and standards.

  1. Drawing Issues
  2. Vector and Raster

Vector charts (ENCs) are generally considered to be an advance over raster charts (such as ARCS) and possibly even a successor to paper charts. However as they come into more common use several generic deficiencies are becoming apparent. Much has been said about the relative merits of raster and vector. Possibly too much since a large proportion of the voices have been driven by political and commercial concerns arguably to the detriment of the end user. However there remains an interesting technological issue which is seldom dwelled on; raster charts look better.

Fig 1: Typical raster chart. Clear understandable information which has carefully been compiled so as to be fit for purpose

In our experience vector evangelists rattle out a litany of raster failings at this point, but the fact remains - raster looks better. It is clearer, suffers less from clutter, is better balanced and is totally familiar[1].

So what’s going on? It seems obvious to us that the fundamental difference between raster and vector as implemented by ARCS and ENC is the contribution of the cartographer. While a good raster chart is an excellent facsimile of the chart a cartographer creates ENCs disregard much of the essential layout information since they try to recreate the chart dynamically, redrawing it from scratch (or very nearly). If we take a step back here it should be no surprise that this automated process is less effective than a cartographer’s direct output. Good cartographers are highly skilled people who exercise professional judgement and experience in what they do. They can spend a day looking at an amended chart trying to assess whether it still retains the same value in terms of balance, content and clarity – and even scale and coverage. Everything we know about computer automation should have warned us that the deep human complexity of this process would be difficult to reproduce. Computers are dumb, don’t learn from experience and have very little time in which to do the drawing. Substituting algorithmic complexity for cartographic expertise for the most part fails miserably and our first thought for the future is that chart formats will need to retain more input from the cartographers.

1.2.Rendering

Rendering of ENC is comparatively slow and rarely gives satisfactory results when compared to a paper chart. Creating good quality charts from vector data will require adding cartographic information to the data set. There are some hints at this in ENC and a few more in VPF. The notion can be taken a lot further. Basically chart data should be designed for display at a particular scale.

All good map or chart data is designed for display at a particular scale, the scale in turn chosen by the needs of the end-user. This is as true for hill walkers and car drivers as it is for mariners. Yet with ENC, a combination of the limitations of display technology and inside-out, feature driven development, has meant that the balance, usability, and at times safety of an ENC has been compromised. We cannot put a standard size chart on one inexpensive screen so we work around the problem by panning and zooming. Good cartography has been sacrificed in the temple of zoom.

Fig 2. Vector data. Without the ability to zoom displays can easily become cluttered

Moreover, zoom is not necessarily the great advantage of vector charts that some would have us believe. It is closer to the truth to acknowledge that ENC actually needs zoom to help manage clutter. At anything like the information density of paper charts the ENC becomes dangerously cluttered.

Zoom is the obvious way out, but it does then dump unnecessary responsibility on the mariner. Instead of being presented with the right data in the right form and at the right scale, the mariner has to make additional decisions about what he wants to see displayed. Some may welcome this, although the safety implications are obvious.

The theoretical basis of S-57 is in the separation of geometry and attribution. This produces a (simple) model of the earth. Other mapping schemes use a database driven representation of a map. This has substantial cartographic information in addition to the model of the earth. Land map sales are more competitive than marine charts and so have maybe been pushed that bit harder. The lack of cartographic information in S-57 compounds the rendering problem so that the information density has to be kept down.

At an imaginative stretch you might find a use for 1:53,146 scale, or some other arbitrary scale, chart but for the most part knowing and understanding the scale of a chart is essential. Given a series of fixed scale charts the cartographers (remember them?) can be rolled out to turn the raw survey data into useful chart information. Their task involves aspects such as deciding what should be visible, how labels should be placed, what symbols should be used and so on. The chart has a purpose. Manipulating the data towards this end is highly skilled job. At the culmination of this we have another cartographic data set which tells us how to layout the chart at a certain scale. Great. Now the geeks can get involved. They will pre-process the data for display, optimise the rendering engines, cache off-screen images and support dynamic anti-aliasing. What ever all that means the end result will be fast, clear, reproducible chart images with all the benefits of an active data layer and the clarity or real paper charts.

  1. Making full use of the available data
  2. Data Quality

ENC predicates better quality data than is actually available. Work is underway to re-survey many areas of the world but it takes time. Currently much ENC data is produced by tracing over existing raster charts. Because ENC cannot capture all the information and knowledge that goes into a paper chart it cannot make full use of the information actually available today. Paper charts used to be produced, in many respects, in isolation from other paper charts. The essential requirement being that the data was consistent on the chart – not that it should be consistent with adjoining charts. This data is still being used for ENC which is why you will often see discontinuities such as contours jumping at cell boundaries. On paper charts the Source Data Diagram (SDD) is an essential consideration for the experienced mariner. ENC has the CATZOC attribute but this is of limited use. Rather than trying to hide issues of data quality it should be made an important aspect of the chart enabling the cartographer to make full use of the data to create a chart without misleading the navigator.

2.2.Projections

Although ENC claims to be projection independent this is not quite true. The path of a line between two points is not defined in ENC, although it is always taken to be straight line on a cylindrical projection. Co-ordinates are therefore defined on a two dimension grid (the surface of the cylinder) and so the ENC model is a flat, square world. Although navigators are typically comfortable with the distortion associated with cylindrical projections, these effects are really a hang over from paper based charts[2] and this is an odd juxtaposition with the notion that it is a good model for marine data. It is possible, but computationally intensive, to re-project ENC data, but other approaches to representing vector information might be more conducive to supporting general purpose systems such as would be essential for high latitude navigation. This is important since with the advent of global warming trans-polar routes are becoming increasingly feasible and we cannot ignore high latitudes to anything like the extent we once could.

Once we free ourselves from the limitations of paper charts the most obvious way to view the world is to look at it like a space man. Or a bird. This is what the world looks like if I were flying over it. Technically we call this a tangential projection. It is a natural way to look at the surface of the planet. By actually defining a straight line and using 3D coordinates a successor to ENC could power a new generation of chart viewers.

Fig 3. Google earth – a tangential projection which displays high latitudes as easily and naturally as equatorial regions

Another type of useful projection is a variation on the bird’s eye view where the direction of view is angled up towards the horizon. Again this is a very natural view in that it mimics common human perception. Objects in the distance, where they are less important, become compressed where as closer objects appear larger. Just like real life. This type of view is already commonly used in automobile satellite navigation systems – which of course have to be readily useable and understandable to untrained navigators.

  1. Dynamic Symbology

The potential advantages are obvious if we think it through. While the main drive behind symbology in ENC is one of standardisation (one object, one symbol, everyone knows what it means), this makes much less sense when one considers the role of a chart in terms of situational awareness. For example, on a bridge system a nearby ARPA target heading straight towards you at 40 knots should slap you in the face. It is important, you need to be very aware of it, and quickly. Meanwhile another target at the limit of radar range doing 4 knots away from you is less of an immediate concern. A standard symbol cannot make this differentiation. Alarms on the bridge are almost useless because of the competition from all the other alarms, so an event like crossing a Safety Contour (a second very real example) needs to be signalled to the navigator in a more connected way. By this we mean that the symbology should help to show immediately what the problem is rather than just indicating that there is a problem somewhere.

The appropriate approach to situation awareness depends on purpose. In a passage planning exercise an interactive list of potential hazards is quite appropriate. However in a front-of-bridge voyage monitoring situation a more automated display that concentrated on immediate issues would be more useful.

Facilitating a rapid appraisal of complex scenarios requires a much more dynamic approach to symbology. This does not just mean being able to switch layers on and off; it means variable size, colouring, transparency and even animation within the symbology. It means being able to manually mark objects for specific purposes, allowing dynamic input data (GPS, ARPA, RADAR), modifying the display and allowing the navigator to be able to interact by adapting symbols to suit his purpose. (The ENC safety contour is a step in this direction albeit a very small one).

To create real dynamic displays will require a re-think over the way standards are managed. Criteria will need to be phrased in terms such as ‘is this situation clearly visible?’ rather than the prescriptive testing for the appropriate static symbol. At the same time information such as display scale and source data information need to modify the drawing so as to create a feel for the quality of the chart information (back to the cartographic point). Contour lines and other symbols should be drawn using less precise edges where appropriate to offset the usual human tendency to assume that computer data is completely accurate.

Fig 4. Simple dynamic symbology. The safety contour is an immediate concern to the vessel and has been high-lighted in red

Full situational awareness requires access to an open ended set of data and hence symbols. The full range of possibilities is large and growing: Radar, ARPA, AIS, tides and so on. To try and regulate all the possibilities would require an extraordinarily capable crystal ball. So if type approval is not intended to inhibit the development of these technologies then the approval criteria will need to be coined in terms of subjective usability.

There is a lack of stability in ENC caused by the number of parameters required to specify a view. For paper charts the only required parameters are chart number and update week number. This defines completely and reproducibly exactly what can be seen. However for ENC we need a list of contributory cells, zoom level, rotation, position, screen size, screen resolution and a hundred or more user configurable display options.The combination of these produces a chart display which can be disorientating and unpredictable. An incident investigation, for example, would want to ascertain exactly what chart information was being displayed. Easy with paper charts but very uncertain with ENCs. Much of this instability is an attempt to deal with clutter. The ability to switch layers off is really a de-cluttering control which then dumps responsibility as to what should be displayed onto the mariner. However the chart display should really be a tool not a process. Cluttering in charts is a combination of static symbology and lack of cartographic input. If these are dealt with appropriately then there will be far less need for operator intervention in creating a useable display. Charts will be more stable, more familiar, more useable and safer.

  1. Summary

Computers cannot build good charts from raw survey data. Future chart information needs to include cartographic expertise.

Rapid situation awareness is a key issue in terms of the advantages of dynamic symbology and more intuitive projections.

Neither of these issues are new, but we should be examining them more closely in the context of ENCs. It is a mistake to think that the way ENCs are produced and regulated now will be fixed into the future.

  1. CherSoft

CherSoft is a small UK based company that has been specialising in navigation systems with high performance raster and vector display kernels for the last 12 years.

[1]Neither is this article designed to be too one-sided in this particular aspect of the debate. CherSoft has worked very effectively with both raster and vector charts for 12 years.

[2] ENC is still tied up with paper charts and so inherits a whole raft of issues including apparently strange behaviour at high latitudes.