1

Re: Missing records from polygon search

This concerns the use of the Report Wizard to return taxon records within a specified polygon.
A typical request is for such records within a 500m buffer zone around a linear object (road, rail, canal)
On the NBN Gateway I have seen demonstrated on a couple of occasions, the use of an algorithm which selects records based upon the proportion of a tile which falls within such a zone.
Recorder appears not to use such an algorithm and only returns records for which the SW origin of the grid reference falls within the buffer zone.
Thus, if you request records from a polygon which has a diagonal SW edge, no records are returned from the SW if their SW origin is outside the polygon.
The more precise the grid reference of the records, the less the error but records from 4 figure grid references are largely omitted from such a region (1Km square) and a sufficient number of 6 figure (100m sq) records whose origin is just outside such a zone are omitted to be of concern.

You can check this yourself using just Recorder.
Drag and drop a very busy taxon (e.g. all birds) onto the map.
Select a polygon with a NW to SE diagonal and use it to select the same taxon.
Drop these onto the map via the Report output option & then display at appropriate dot scale
How many dots are mostly within your polygon on the SW edge but are omitted from the polygon search?

So I've failed to report a critically important bat roost because I only have a 6 figure grid reference for it.

2

Re: Missing records from polygon search

When we realised Recorder behaved this way we stopped using it for data searches and exported all the records into GIS (MapInfo). There we plot records at the centre of the record's resolution square (e.g. 1km records are plotted in the centre of the 1km square) and colour them according to resolution (red=10m, green=100m, blue=1km, pink=2km, yellow=10km). That way when we draw a search polygon we can easily see if there are any records just outside it that ought to be considered. e.g. if the search polygon clips the corner of a 1km square then we include the blue records too by adding a search polygon around the blue dots (and possibly the pink and yellow ones too). If time permits we may look at such records to see if there is anything in the location's descriptions that clearly locates them outside the search area so that they can be excluded from the report.

Regards,
Keith Balmer
Beds and Luton LRC

3

Re: Missing records from polygon search

I too am happy enough with plotting onto MapInfo afterwards and I very much like your idea of characterising them wrt resolution and colour-coding accordingly. I've been creating square tiles of the correct size and position.

The big problem arises with the phrase "exported all". This really has to mean "all" taxon occurrences within Recorder, not just the ones that you select using geospatial routines as these clearly don't work.

So in effect we're having to run another database alongside Recorder in order to obtain the services we need - which rather defeats the objective of Recorder.

If the current "does what it says on the tin" version of Recorder is now stable, perhaps these outstanding GIS and other geospatial issues need to be rather high up the list of priorities for the next tranche of Recorder development.

4

Re: Missing records from polygon search

We have some home-grown code written by a volunteer to export all records from R2002 (over half a million) into a MapInfo layer. We didn't use any inbuilt R2002 functionality, but may do once we eventually migrate to R6 assuming we can make it work in a similar way. (If it still plots things at the SW corner we won't use it).

Regards, Keith

5

Re: Missing records from polygon search

I would love to see Recorder's spatial queries take into account the whole square rather than the SW corner - this has always been a real limitation.

Charles Roper
Digital Development Manager | Field Studies Council
http://www.field-studies-council.org | https://twitter.com/charlesroper | https://twitter.com/fsc_digital

6

Re: Missing records from polygon search

I recently downloaded a bit of VB code from the ArcGIS website that you can use in a "Calculate field" expression to convert point data into polygon grid squares of varying sizes, so points with grid references of 2, 4, 6, 8 or 10 characters can be converted in polygons of the relevant size from 10m to 10km (or any other size if required).  It is a relatively simple bit of code that can be adapted as required if you have some VB knowledge.

At the moment we do spatial in GIS using point data, but because all of our data happens to have 8 character grid refences we are fortunate we don't "miss" any records.  However, in the future we may get data from MapMate that may be less precise.  We therefore plan to use this code to convert all our notable species points into polygons so that GIS spatial queries using point, lines or polygon search criteria will select all "relevant" notable records regardless of their recorded precision.

Does anyone else use a similar approach?

Andy Foy
Systems Manager
Greenspace Information for Greater London (GiGL) CIC
www.gigl.org.uk

7

Re: Missing records from polygon search

Yes, Andy, I do that. It was what I was referring to in my posting for "Feature requests" under the titleMIF export and is based upon records selected through the Wizard.
The coding required to do it is quite straightforward, it works fine for rectilinear polygons.

8

Re: Missing records from polygon search

I know this is a bit of an old thread, but now I've noticed it, I feel I need to comment.

If we're talking about exporting all taxon occurrences to polygonised GIS sets, we need to remember that many occurrences in recorder are associated with a location (site) and that the spatial ref of that location is nominal (a centroid?). The only way to represent these records correctly would be to export them as a polygon the same shape as the site itself, unless they have a sample spatial ref which differs from the location spatial ref.

Of course Recorder is currently ignorant of such notions and the various spatial reports will only consider the spatial ref of an occurrence, thus they return only a fraction of the available data, especially where sites are either large, or linear.

At present nearly half of our data (several 100,000 records) is site based.

Rob Large
Wildlife Sites Officer
Wiltshire & Swindon Biological Records Centre