Hi

I am using Indicia with a national grid reference system and sample locations are stored as squares with a size of the square indicating the accuracy of the record. When a user indicates the sample location on the map a web service call returns the square with size varying according to how much the map is zoomed in.

On the other hand, where latitude and longitude are used as the spatial reference system then  a point location is stored with no indication of precision.

In recent correspondence with John van Breda about this he replied to me "We ought to add a precision field. The geom stored is indeed a point which is probably preferable to trying to store a circle with radius."

However, if a point and precision is stored then I guess for every spatial query it would have to be expanded to a polygon to see if it intersects with the query area which may be more trouble than storing a polygon in the first place.

Jim Bacon.

Hi,

thanks for the hint how to use our precision attribute.
It's good to hear that you plan to support this directly in the database schema. I wonder if creating the buffer on the fly at spatial queries is not too expensive? Check If a buffer is needed, creating it and the testing if its in the area is in the area I'm searching for. I think it would be more economic saving the buffered polygon in an extra column since spatial queries are quite common.

Regards

Daniel

Would another option be to create an sref module, as we have for other coordinate systems,  that has an sref_to_wkt function that returns a polygon which is more of less circular, centred on the selected point? When setting up the client website you could choose this for fuzzy lat/long data entry. This would not need any change to the core code or data structure.

If this were created Daniel could then run an update on his dataset to convert it to this form. The entered_sref and the fuzzy fields would retain the original information but the geom field would now contain a polygon rather than a point. We would need to invent a value for sref_system.

Jim Bacon.

My thought was that the fuzziness could be derived from the size of the geom polygon rather than adding a new field to the data structure - indeed I have already had to do exactly this. It wasn't very pretty.

I can see the merit of doing this once rather than many times for each  point-based system too.

It is a required field in the NBN exchange format even when using a grid. (See Precision field in http://www.nbn.org.uk/Guidebooks/Data/A-beginners-guide-to-sharing-data/The-NBN-Exchange-format.aspx)

So I think you are right to add a Precision field to samples but do you need a second geom field? You have stored the entered_sref and, if the Precision is 0 then the geom could be a point and if the Presision is non-zero then the geom could be a polygon.

Jim Bacon