Coastal Buffers

[SK53]

One issue with using OpenStreetMap is that the coastline is, broadly speaking, the MHWM (high tide line), but in many cases vice county recorders will have data from both the littoral and sub-littoral zones (e.g, Eelgrasses, Brown Algae, salt marsh plants, snails & other molluscs).

Also there are no good sources for either the low or high water mark, and OSM is only as good as available aerial imagery and perhaps GSGS3906 mapping. In places with high cliffs, such as the W end of Achill Island aerial imagery is more-or-less useless for even remotely accurate coastlines. In other areas the coastlines have not been re-visited since original creation from landsat or PGS. Therefore the accuracy of the coastline is not of the same order as boundaries based on inland features (mainly mapped as Townland boundaries).

Therefore there are two related problems:

• Use of the high water mark coastline excludes certain areas where biological recording is likely to take place (the sea shore, tidal rocks, mud flats etc).
• The coastline is subject to inaccuracies simply because of how it is derived.

There are two potential ways of dealing with this:

• Cease to use the coastline as the seaward boundary of the vice counties. Replace this with a marine boundary a fixed distance off-shore (NBN provide at least low water, 3 miles).
• Keep the existing boundary based on the high water coastline and add one or more additional polygons which can be merged together to create suitable VC boundaries for different purposes. Such polygons might include the littoral zone, and successive distances offshore.

In general I have decided in favour of the latter approach, for the following reasons:

• Experience with NBN files based on low water mark caused some cartographical problems, Ready separation of dry land from tidal areas offers more possibilities in terms of different use cases (validation and cartography being the foremost ones).
• It should be possible to generate various coastal buffers automagically without having to store them in OpenStreetMap directly. (Although this is not my approach in the initial implementation).
• As OSM data is continually being refined, one wants to reduce the direct dependencies and make it easier to pull data from OSM to this repository.
• Should low water data become available there is a clear mechanism for incorporating that and allowing additional polygons to be added.

As a starting point I have created a single example coastal buffer for County Down. This is based on the OSM coastline with a 250 m buffer and manual inclusion of large inshore areas such as Strangford Lough. The polygon has been added to OSM. This should meet current needs regarding any potentially missing off-shore islets or drying rocks.

[burkmarr]

I work at BRC and we have a requirement to create a set of buffered Irish VCs to facilitate coastal and marine record searching on iRecord. I think the current Irish VC boundaries we use are sourced from you. I could buffer these to three nautical miles on the seaward side (to be consistent with the GB VCs we have in our VC layer) and the only thing that would require a bit of work is delineating adjacent extensions, but it shouldn't be too onerous to do this manually.

But I saw this issue and I wondered if there scope for me to contribute this work to your project or if we could work together in some other way, using whatever GitHub workflow you'd prefer.

Also on another separate by related matter, BRC will likely take on the hosting and maintenance of the GB Vice Counties and I like the idea of doing this via an iRecord GitHub project - much as you've done with the Irish VCs. I see you've noted some geometry problems with the currently available data (https://forums.nbn.org.uk/viewtopic.php?id=4319). I came across the same problems in QGIS (selecting some polygons cased crashing) and did some work on fixing them. It would be great to have your input on this project too.

If you want to email me, I am rburkmar@ceh.ac.uk

[SK53]

Hi Rich,

Very interesting.

I think, will have to check, that I created buffers (250m) for all VCs when I created H40 Co Down. There were particular validation issues with islands on Carlingford Lough on the BSBI site which I wanted to help resolve. I also wanted to keep the existing (notional high water) polygons in step with all the other OSM boundaries.

I see no reason not to work together, but I'll need a couple of days to check what I did. The workflow is still a little more complex than I like:

• Check boundary status on OSM (in case any have been broken, altered significantly)
• Pull the boundaries down and convert to shape file (all in EPSG:4326)
• Create buffering in QGIS as shape file
• Pull buffer polygons into JOSM
• Cut buffer perimeter & merge with existing OSM ways (this is probably easiest place to deal with naïve buffering with overlaps)
• Push data back to OSM
• Pull all data back & make 27702 shape files
• Load to GitHub

I had hopes that GitHub would show diffs between each version (hence using Geojson) too, but as the Geojson diffs are just naïve textual diffs this doesn't work. I now prepare a difference in QGIS to help inform people using the data.

Adding another set of buffers to take the distance out to 3 nm in OSM would be rather simpler than the complexity of merging with coastline. It may also be possible to keep these solely in GitHub, which might be better than adding them to OSM (which is needed for the inshore buffers because of coastline).

Another specific issue with Ireland is that at present, unlike in UK, we don't have good high quality sources of high & low water. It's a mix of digitisation from aerial imagery and old GSGS 3906 maps. My 250 m buffers are basically a proxy for low water. Therefore the actual perimeter is likely to change as (and when) we get better forms of data for coastlines (both OSNI & OSI have some openish data but as yet not clearly licenced in a form we can use on OSM. When I compared OSM data with OSNI townland data it was the coastline which was most likely to be out of kilter. I have actually done some preliminary planning for an internal OSM project for enhancing the coastline, so this might be the time to get it going.

Lastly, I just want to check that you are aware of he work of Tom Humphrey & Jim MacIntosh on Scottish VCs for BSBI (I think mainly adding missing islands, but perhaps fixing the original broken polys for some of the Island VCs).

Will email you with any further thoughts.

[burkmarr]

Thanks that's very useful. Is the work done by Tom and Jim publicly available?

[japonicus]

Credit for the BSBI work that @SK53 referred to should go to Jim MacIntosh and Andy Amphlett (I was only marginally involved).

I'll check what's available - probably several fixes to invalid geographies and restoration of some missing islands. Would be happy to make available anything that might be useful (will contact you by email, so as not to derail this issue thread).

[SK53]

@japonicus thanks for this. Should have guessed Andy was involved too.

[burkmarr]

Yes many thanks @japonicus.