How accurate are the lat-lon refs that Overpass produces? (I mean in notation terms: obviously the real-world accuracy depends on 10,000 things.)

They seem to be always to 10 decimal places. My maths is terrible, but I think that means the 10th decimal place is indicating 0.01mm of distance (assuming 111km per degree of latitude).

Is that right?? It seems unrealistically accurate, and maybe I've missed something.

The reason for asking is to ask Calc/Excel for '±100m from x/y'.

Thanks

lat                                         
52.5802291381

52  111km                                       
    .5  11.1km                                  
        8   1.11km                              
            0   111m                            
                2   11.1m                       
                    2   1.11m                   
                        9   11.1cm              
                            1   1.11cm          
                                3   0.111cm     
                                    8   0.011cm 
                                        1   0.001cm

asked 15 Sep, 15:59

eteb3's gravatar image

eteb3
28091323
accept rate: 6%


"How accurate something is" depends not on how accurately the data is stored, but how accurate the data was when it was first recorded. This applies whether you're looking at data in OSM or (say) how many hospitals Boris Johnson is actually going to build before 2030.

In the case of OSM you have to ask yourself "how accurately might someone have recorded something". I'd hope within the UK that someone adding something now would get the building it's in to within 10m either way, though of course they might simply pick the wrong building and be a few extra 10s of m out on top of that (hence SK53's 50-100 metres being in the right ballpark).

If you're comparing OSM with an external dataset then obviously you'll have to consider how accurate that data is too - if it contains "old" data (15 years or so?) then some of it in the UK might have been geocoded to the most significant part of the postcode. Even if not, any rural areas anything geocoded from just the postcode will be quite a way off.

permanent link

answered 15 Sep, 21:34

SomeoneElse's gravatar image

SomeoneElse ♦
34.6k69359829
accept rate: 15%

geocoded to the most significant part of the postcode That's my understanding of how the lat-lon were produced on the external dataset.

(16 Sep, 07:33) eteb3

According to the wiki, nodes are stored in the OSM database with 7 decimal places. I just tested and I only get 7 decimal places from Overpass (Overpass Turbo or overpass-api.de). A different Overpass instance could theoretically be configured to return more decimal places, but any beyond the first 7 should always be 0 for nodes coming from the main OSM database (and those would be false precision).

permanent link

answered 15 Sep, 17:33

alester's gravatar image

alester
6.1k25992
accept rate: 30%

My bad: I have two datasets from two sources, and I mixed them up. One is from Overpass, the other is from somewhere else.

You're right, my Overpass output gives 7 decimal places. Am I right that - in theory - that is still 1cm of precision?

Does that mean I can ask Excel to round both datasets to 3 decimal places, and I can then see if two lat-lons, which are dissimilar only in the right-most decimal places, are likely to represent the same on-the-ground object?

(15 Sep, 17:57) eteb3
1

I don't think that's a good strategy; there are plenty of much more sophisticated geospatial operators around & available for instance in QGIS. A reasonable buffer for conflation in the UK in urban areas is 50-100 metres (i.e, roughly your 3 dp), but that may be too small in rural areas, particularly if geolocation of data is based on postcode centroid. I would translate datasets to British National Grid (guessing this is your PoW dataset) because you can do direct comparisons in metres (and in Excel if you wish).

(15 Sep, 21:12) SK53 ♦

Thank you, I will check out QGIS.

For the conversion to British National Grid, is there an algorithm that I can run through Excel, or a lookup table, or is it a question of going line by line? (I'm guessing 'surely not!' but possibly the solution is beyond my technical ability.)

(16 Sep, 07:31) eteb3
1

At one point the OS provided an Excel addin, but the easiest way is with QGIS which will allow you to save data in another projection (I usually use Geojson & havent experimented with CSV). You can also use the ogr2ogr command line tool. The key things are OSGB is identified as EPSG:27700 and WGS84 is EPSG:4326. (The minor complication is QGIS now provides a rather over rich range of reprojecting which can be confusing, but I think for your purposes the default should be fine).

(16 Sep, 15:02) SK53 ♦

Great, thank you. Look forward to a long weekend learning a new computer program some time soon!

(16 Sep, 19:30) eteb3
Your answer
toggle preview

Follow this question

By Email:

Once you sign in you will be able to subscribe for any updates here

By RSS:

Answers

Answers and Comments

Markdown Basics

  • *italic* or _italic_
  • **bold** or __bold__
  • link:[text](http://url.com/ "title")
  • image?![alt text](/path/img.jpg "title")
  • numbered list: 1. Foo 2. Bar
  • to add a line break simply add two spaces to where you would like the new line to be.
  • basic HTML tags are also supported

Question tags:

×49
×39
×1
×1

question asked: 15 Sep, 15:59

question was seen: 228 times

last updated: 16 Sep, 19:30

powered by OSQA