Positional Accuracy Part 2
So how do you define accuracy? In the GIS world there are a few different types of accuracy to consider. When some people hear the word “accuracy” they think about the bullseye on a dart board and how close someone can stick the dart near that bullseye. Since spring training just started others may think about a baseball pitcher and how close he can get the ball to the catcher’s mitt or strike zone. As a GPS/GIS person I usually define accuracy as how close the GIS feature is to its true location on the planet, with separate accuracy values for horizontal and vertical location. But accuracy can be defined a few ways in a GIS.
First a few questions; what is the true location of a feature and how can someone find the true location of a feature to tell how accurate their data is? (To be answered later) The second question; does the true location really matter? I have found over the many years I have been working with GIS data that people often don’t want “accurate” data, they want data that looks good. By looking good I mean that the data agrees with the base mapping that is currently being used. So even if the base map is of poor quality as long as the other data overlays nicely everything else doesn’t matter.
That leads me to my multiple definitions of accuracy. In the GIS world we have what I often refer to as “absolute accuracy”, which I use to describe how close a feature is to its true location, as defined by the geodetic datum and adjustment being used. And then we have what I describe as “relative accuracy”, which I use to describe how close to the true distance one feature is relative to another shown on the same map. In this part of the country most land surveys are still done with relative accuracy as the only consideration and I find in the GIS world that digitizing is still very popular because of the relative accuracy visual benefits.
In the early stages of GIS development the difference between relative and absolute accuracy were a big issue. People were trying to use poor quality base maps that sometimes had no real-world coordinates at all. Since then base map quality has improved tremendously. Good thing because as more people wanted to use GPS to add data to their GIS absolute accuracy became very important. Even still I find that the base map and relative accuracy to the base map is the most important concern when displaying data to people that don’t understand accuracy issues.
In my area of the U.S most people are now using orthophotos as their base map. But if I go out and locate the position of a well defined feature using survey-grade GPS techniques and I establish the “absolute accuracy” of that feature to +/-1cm but that location doesn’t overlay onto the base map nicely I find that people want to believe the base map before the feature location. It seems hard for people to understand the idea that everything we look at on a map has an error associated with it and that it's ok for a +/-1cm point to be "off" 80cm when the base map itself is +/-1m. Some features may have more error or lower accuracy than others, even the base map itself.
It seems that relative accuracy to a base map is still most important when viewing the data but absolute accuracy becomes more important when the GIS matures and people want to start uploading the GIS data back onto ever more sophisticated GPS equipment and portable computers for maintenance and/or feature recovery, especially is the feature is buried or difficult to find. The bottom line, use the most accurate base map available, if possible include estimated accuracy values in all databases, make sure all datum references are properly identified, and if you need to “adjust” feature locations to make the data overlay onto a base map better please make a copy of that layer before editing if the original layer has a higher absolute accuracy than the base map.
Next Post: Where do GIS and GPS positional errors come from?
Categories: Data,accuracy,GIS,Mapco,Maps,GPS
Tags: Data accuracy GIS Mapco Maps GPS
I've had customers who would complain that the building outlines we mapped was incorrect because they didn't match their orthophotos, completely ignoring the fact that buildings are often "leaning" to the side, and the outlines matched perfectly at the foot of the buildings. Of course we used it as an opportunity to upsell to "true orthophoto".
Reply to this
Another great article. Quite coincidentally, Wired publishes a story underscoring (in my opinion) the relevance of relative accuracy:
Chile Earthquake Moved Entire City 10 Feet to the West
http://www.wired.com/wiredscience/2010/03/chile-earthquake-moved-entire-city-10-feet-to-the-west/
Reply to this
Thanks for the complement Atanas! I had not heard of the 10ft shift in Chile. California has to deal with the same issues, though at a much slower rate . Years ago they installed a high accuracy GPS on both sides of the faults out there so they could monitor and update the datum info. Currently all the CORS have published velocity data and I am told that in the future all U.S. geodetic control will have velocity data published with the geodetic control datasheet. Even now when I do a high accuracy control project with CORS I include the epoch date along with the datum info i.e. NAD-83 CORS 2002.
On another note I was wondering if I should be including categories or tags when I publish a blog?
Reply to this
Whether or not to include categories and/or tags is up to you. They do make the articles easier to find by humans and computers, so I use them now (although I did not in my first year of blogging).
Reply to this
I also think Wired may have gotten their unit conversions wrong, and the actual shift is 1/10 of a foot.
Reply to this
I shared my suspicion with the Wired article's author, Betsy Mason. She promptly responded with a link to the source. 10 feet is correct.
http://www.eurekalert.org/pub_releases/2010-03/osu-rsh030810.php
Reply to this
I enjoyed reading the discussion on accuracy, both relative and absolute, as it brought back long ago memories. The discussion and often arguments about which type of accuracy is necessary for effective GIS or perhaps preferably "Geo-Spatial Information Processing" goes back before we even had the word GIS. At the Keystone III AM/FM conference around 1980 I was engaged in a panel dialog with an eminent photogrammetrist about the appropriate type of accuracy needed for a utility AM/FM projects. After the session and much more conversation I concluded that the relative accuracy argument was being pushed by many vendors in the industry for several reasons. most utilities did not have PROFESSIONAL mappers on staff, either surveyor/civil engineers or cartographers; photogrammetric base mapping was expensive and difficult with very large data files, Computers were slow, memory starved and often built fot business/monetary calculations rather than engineering precision. Most software only supported 16 or 32 bit data for a long time, A major vendor only got on board with 64 bit data relatively recently.
I was surprise that the author refers to most surveyors still using relative accuracy, since my experience is that all of the successful surveyors I know are using GPS enabled processes even if the final field work is done with total stations or robotic stations. You can't stay in business with transit and tape approaches any longer. It may very well be that the survey is done in real GPS coordinates , yet published to the client in what appears to be a relative manner. The failure of clients to require metadata in the survey notes hides the methods used.
Another factor that has influenced the discussion is the failure of many end users of GIS data acquisition projects to understand the difference between accuracy and precision.
I am also convinced that most GIS projects are intended to be automated cartography vs. automated spatial models sufficient to engineer solutions from. From my own experience I know that cartographic license is employed to place overhead pole lines at an offset from the streets that places the poles in the front yards when viewed on a high resolution ortho base. This phenomenon occurs because of a software requirement to have a one to one relationship for each phase of the conductor and the graphic occurrence of it.
Thanks for an interesting topic and one that still needs to be discussed and better understood after all these years.
Reply to this