Map Accuracy Specifications, Part 2
Where did they come from, what are they and what do they really mean?
By Robert A. Fowler

Reality and politics
Accuracy and precision should depend much on what the user plans to do with the data (although far too often they do not). If you are planning a new road through some uncharted territory, you might consider a 2 meter pixel resolution accurate to 2, 3 or 10 meters sufficient to produce a rough location. However, if you are engineering that road, you may want map information accurate to 10 centimeters so that you can calculate rough cuts and fills to determine how much money is going to be spent excavating or earth moving.
     At this point all of the engineers out there are saying, "but we want one inch accuracy." Well just about anything is possible for a price, but you don't really want to know what the price is going to be for that, and anyway a ground survey will provide a greater degree of accuracy certainty and probably be less expensive at this point. On the other hand, is that level of accuracy really necessary? Once a bulldozer gets working on the ground, the tracks of the machine will probably sink into the ground more than 1 inch so the whole construction accuracy issue is subject to debate. Even in bridges the coefficient of expansion of concrete (10-5/¡) can be greater than the computerized output accuracy achieved in a survey.
     If we look at municipal mapping, the accuracy of the parcel fabric has some intrinsic value (as people pay taxes on the amount of land they own), but in many cases this is based on real ground measurements-not fence lines or whatever was collected from an aerial photo, at least in high value city areas. Even more to the point, the positional accuracy requirements for buildings etc. that sit on the lots is an even more arguable point. Who really cares if the positional accuracy of a building is better than a meter or not, as long as it sits within the lot boundaries? The fact that the approximate dimensions of the building indicate its size and likely what it is (a single dwelling bungalow, an apartment building or a gas station) is probably really all that matters. The same for other detail. The precise location of a fire hydrant is not (or should not be) a critical issue. If it is there on the map within 5 or 10 feet of its real position (at least whoever needs to know knows that somewhere in the locale there is a fire hydrant) and on the ground it will be pretty obvious once you get within a 100 feet of it (it's not likely you will need to trip over it to find it!). We used to spend hundreds of hours (and thousands of dollars) collecting driveways on large scale maps. I often wondered what the purpose was and, apart from some esoteric drainage calculations, have never found out. If a new pipeline is going to be laid and you need to dig up people's driveways, it has to be a lot less expensive to get on a bicycle and ride along the road in question counting how many people you will need to compensate, than paying for driveways to be collected for an entire city on the chance you may need them some day. ("But we always did it this way" or "It doesn't look finished without them.") Today, of course, with digital orthophotos that illustration is academic, but you get the point.
     However, these observations are political and the fact they are real questions, almost never enters the picture when it comes to the accepted standards for mapping.
     Since we have just mentioned detail, we should consider that this, too, is a sub set of map accuracy. What gets shown at any particular scale and how accurately it is positioned enters the equation, and this again is a hangover from the old days of printed maps.
     You can only fit so much onto a piece of paper (compared to layers in a computerized system), so the amount of detail which appears on any specific scale is due to the limitations of what could be reasonably squeezed onto the paper, and this affects what is considered "normal for any scale of map." As a result, you are expected to show sidewalks on 1:1,000 scale mapping, but drop them from 1:2,000 scale mapping. The width of a river or stream is shown as a double line at any scale where it is wide enough to be drawn that way. The limiting factor is the distance you can physically draw or print two lines apart before they run together.

So what are the normal map accuracy standards?
Surprisingly, while there are a multitude of specifications used around the world, they do not appear to differ very much from one another, indicating the engineers and surveyors of yore had a pretty good communications system. The ones that affect us in North America are the United States Geological Survey specifications, the American Society of Photogrammetry and Remote Sensing, the various state and provincial government regulations, most of which are clones of one of the former, and the United States Department of Transportation, and its clones in the state governments and an extremely similar set of specifications used by provincial Departments of Transportation in Canada and the Ministry of Natural Resources Canada mapping specifications.
     Overall, these specifications are very similar and vary only slightly, usually in some minor numerical value. To my knowledge, as far as accuracy specifications go, overseas specifications, whether you deliver maps in the United Kingdom or Burundi, vary again only marginally. For example, accuracy specifications in aerial triangulation may specify an RMS of 37 microns in one set of specs and 35 microns in another. A difference of 2 microns isn't worth losing sleep over.
     Indeed, the photogrammetry training manual for the U.S. Corps of Engineers, which is probably one of the best compendiums of knowledge on the subject written in words that anybody can understand, states the five main (U.S.) sources for standards used for specifying spatial mapping products and resultant accuracy compliance criteria as:
a) The office of Management and Budget (OMB) "United States National Map Accuracy Standards" (Bureau of the Budget, 1947)
b) Photogrammetry for Highways Committee "Reference Guide Outline: Specifications for Aerial Surveys and Mapping by Photogrammetric Methods for Highways" (1968)
c) U.S. Department of Transportation (DOT) "Surveying and Mapping Manual" (1985)
d) American Society for Photogrammetry and Remote Sensing (ASPRS) "ASPRS Accuracy Standards for Large Scale Maps" (ASPRS 1990)
e) U.S. National Cartographic Standards for Spatial Accuracy (No date)
     Each of the above has been slanted toward a specific type of mapping, but as the Corps of Engineers training manual says, "Éuse of any of these standardsÉwill result in a quality product." In other words, any of them will do, and notice the dates which tells us something. Technology may change but the results people expect haven't.

So what do these standards say?
Almost everyone of these standards says something like:
     Compilation will be carried out using methods that ensure the following horizontal accuracy. Ninety percent of all well defined features except those unavoidably displaced by symbolization will be within 0.5mm of their true position (when plotted to scale). So for example 0.5mm at 1:50,000 scale is 0.5x50,000 = 25 meters.
     In vertical the elevation accuracy of contours is: 90 percent of all contours will be within half the contour interval in open areas. The elevation accuracy of spot elevations will be 1/4 of the contour interval. So for example, a 10 meter contour will be expected to be within +/-5 meters, or a spot elevation on a map with 10 meter contours will be expected to be +/-2.5 meters.
     In the U.S. there are three levels of mapping standards. They are called simply enough, Class 1, Class 2 and Class 3. Class 1 maps are the most accurate. Class 2 maps have an allowance of twice the RMSE of a Class 1 map and Class 3 maps allow three times the error of a Class 1 map. For the purposes of this article (and for most map users' requirements) Class 1 maps are the norm and are the subject of this discussion.
     RMSE is, as probably everyone knows, the Root Mean Square Error. This is derived from the square root of the average of the squared discrepancies when compared to a higher level independent survey. The RMSE is normally defined in terms of ground scale errors. The RMSE is the cumulative result of all errors of all the processes included in the making of the map. Table 1 indicates the maximum allowable errors at various scales for well defined points.
     In the Corps of Engineers manual there are pages of limitations for what constitutes accuracy which can be achieved using analog versus analytical stereoplotting instruments. But what it all comes down to is the standard specifications already quoted.
     To put that in perspective, however, you should know a little about scales of mapping, compared to scales of photo and what is commonly accepted here and around the world.
     To obtain 1 meter contours it is accepted that you fly photo at a scale no greater than 1:10,000 for analog instruments and 1:12,000 for analytical instruments. Because of the limitation on what you can see in comparison to what is normally shown at a specific scale, you would normally use these scales of photo to produce 1:2,000 scale maps.
     So 1:2,000 scale maps are generally accepted to require 1 meter contours. 1:1,000 maps get a 0.5 meter contour. 1:10,000 scale maps usually have a 5 meter contour, 1:20,000 and 1:25,000 scale maps a 10 meter contour and 1:50,000 scale maps a 20 meter contour. 1:100,000 scale maps usually have a 50 meter contour and 1:250,000 and up often have a 100 meter contour.
     Index contours (contours drawn at twice the normal contour thickness) have no bearing on accuracy; they are there simply to make the mapping easy to read. These are usually (though not always) every fifth contour. For example if you are using 1 meter contours, the index contour will be every fifth contour. However, for convenience, if you are using 25 meter contours, some mapping agencies use every fifth contour and some every fourth contour. So you could have index contours every 125 meters or every 100 meters.
     While all of these mapping specifications have been and are based on survey and photogrammetric techniques, these techniques have become the norm and the de facto standard expected, regardless of any production method used. So if you went into the field using a total station to produce 1:25,000 scale maps, you would be expected to produce 10 meter contours to go with them. If you wanted to produce 1:25,000 scale maps from SPOT imagery or RADARSAT you would still be expected to supply a 10 meter contour. Now, it would be pushing the limits to expect to get 10 meter contours accurate to plus or minus half their interval (5 meters) from SPOT images or RADARSAT data - so the short answer is simply you should not be using SPOT images to make 1:25,000 scale maps. If you do, you have to state on the map something to the effect of, "This map does not meet National Map Accuracy Standards."
     One of the major problems in dealing with digital maps is the fact that you can enlarge and reduce them at will and create scales at which the specifications are totally irrelevant. You can easily make a graphical plot of a 1:50,000 scale map at 1:2,000 scale but it is not a conventional 1:2,000 scale map and it will never have the accuracy of anything but a 1:50,000 scale map. Conversely you could plot a 1:2,000 scale map at 1:50,000. You may not be able to read anything on the paper, but it would still be at 1:2,000 scale accuracy (assuming the plotter is capable of plotting that accurately at such a reduced scale). To retain any accuracy controls, you have to assume map accuracy relates only to the original scale at which the map was produced.
     I have yet to see it, but presumably in the near future electronic maps will all have an undeletable rider which pops up every time you access the map and says something to the effect of, "This map product meets National Map Accuracy standards for 1:50,000 scale mapping," to safeguard the producers, agents and sellers from lawsuits and to remind users of what they are actually getting.
     So in summary, mapping specifications (whether you agree with it or not) are based on surveying and photogrammetry criteria. And they can be roughly stated as:
     A map is never expected to be produced at more than six times original data acquisition scale. (I will go out on a limb here and state categorically, the preceding is rarely put in specifications, is often abused by unscrupulous map producers, but is just plain common sense.)
     Contour intervals are pretty well fixed at a maximum of 1/10,000 (1/12,000) of the photo scale and/or if photo methods are not being used at the interval at the map scale with which they have become associated. They should be accurate to half of the interval.
     Spot elevations should be accurate to 1/4 of the contour interval.
     Positional accuracies are reckoned to be to 0.5mm or between 1/30th and 1/50th of an inch at map scale plotted out on paper (or mylar).
     And while these standards apply to aerial photo mapping, they have become and are the generally accepted standards used for maps made from any source.
     Is it time to change the rules?
     I don't think so. Is it time to change our attitude about what is needed? Probably. In my opinion, a lot of the perceived accuracy clients think they want is not really required. Many clients ask for far more than they will practically ever use or need, or the cost/benefit warrants. The section of this article on the politics of accuracy probably should be read with more emphasis than any other part. On the other hand, clients do have a right to expect that what they are sold is what they are going to get. If there are valid reasons for not meeting normal map accuracy standards-and sometimes there are-these should be prominently stated both in the written contract and on every file or map sheet produced. Pushing the limits of any technology should only be done with the greatest of caution and full disclosure to the unsuspecting, paying public.
     Technology is great. I love it, especially when it makes my work easier. But most things are still constrained by the laws of physics and we shouldn't forget that.

About the Author:
Robert Fowler, O.L.S., C.S.T., C.E.T., is proposals manager for Intermap Technologies in Ontario, Canada. He has written on a number of subjects previously for EOM. He has more than 30 years experience in surveying and mapping, and has written mapping specifications for a number of clients, including the Canadian Department of National Defense and contributed to the mapping specifications for a number of other countries' mapping agencies. He may be reached at 613-226-5442.

Back