This post is dedicated to a dear colleague of mine, who I have only worked with for a year but has had a profound effect on me and the way I look at my work. His name is Stuart Leather and he’s off to work for another company but he has left a phrase in my head that will never leave:
“Always check you balls”
Wise words, I think as we have become more digital we have forgotten (or at least not double checking) our “balls” or as I prefer to call them, ellipsoids & spheroids.
This has really come to light for me over the last few months where I have been helping people with the new QGIS 2.0 and also working on some web map development. Those of you lucky enough to follow me on Twitter will have seen my constant tweets about NTV2 support for QGIS 2.0 and also questions over how this can be remedied…..what was more surprising to me was that I have over 2500 followers and only a couple people had a clue what I was talking about.
More worryingly, I recently posted a question about NTV2 & grid transformations on GIS StackExchange and although I got a couple of half sensible answers, I was shocked to read the suggestion that I didn’t need a grid transformation from WGS84 to NAD83 as they were the same spheroid……
So….by this I am guessing that the GIS world has gone digital and has forgotten that they have balls, let alone check them.
Okay, for the sake of Stu, I am going to try and explain in very simple terms why you need to go get your balls sorted.
When using a single geodetic system (ball) it is relatively easy to switch between zones or different transformations as it only requires calculation based on a constant, hence why changing between WGS84 & WGS84 UTM Zone 30N is simple.
So why is it so much harder between WGS84 and OSGB36 or WGS84 and NAD83? Yes, you got it! It’s all to do with your balls, WGS84, OSGB36 & NAD83 all use different spheroids and these are not not even similar to each other, I could write a nice chapter here on geodetics but I am hoping that you know the reasons and issues here….essentially to get the best calculation between the balls you need a GRID transformation, this will allow for the variances between the balls, providing, in most cases, accuracy between balls of about 2-4cm.
You don’t always need this level of accuracy, if you are working on a 1:50k map, you really wouldn’t need to worry that much, so in this instance you could get away with a much simpler Helmert transformation (also called a 7 parameter transformation). A Helmert transformation, you may have seen it in your ArcGIS under something like “WGS84_to_OSGB36_Petroleum” will normally give an accuracy of about 7m between data.
So, will the world end if I DON’T check my balls?….well it won’t end but your career might….you see, depending on where in the world you are working, you could see differences between 50m and above (200m in the further extents of the UK between WGS8 and OSG36), so designing that coffee shop to sit next to the local cake shop might all go dramatically wrong…..
What I would like to see is for GIS developers to see this and realise that there is not enough provision made for those of us who need to provide a level of accuracy in our work or at least work between land & sea in the same map. OpenGeo have made a great start and allow the use of NTV2 grids and NADCOM grids within their web system. I have struggled to get PROJ4 and GDAL to work correctly with grid or helmert transformations, this means that transformations need to be applied to data (usually using ESRI ArcGIS) BEFORE inclusion to data before inclusion to most web map systems.
Thanks Stu for all your help over the last year, making me love my balls again. Good luck helping others with their balls.