Grid references of old records
Posted: Sat Mar 03, 2018 4:44 pm
Grid references of old records
The Ordnance Survey grid did not appear on maps of Great Britain until post 1945, with the publication of the Popular Edition of the One inch to 1 mile scale maps. In Ireland, the Irish Grid did not appear on generally available maps until the 1970s, when it was used on the 1:126,720 (half inch to the mile) series. (For recording for the first Atlas of the British Flora (Perring & Walters 1962) the OS grid for GB had been extended to Ireland by David Webb).
The DDb has >443,000 pre-1930 records (using end year for records with date ranges), of which 98% have grid references: 47% at tetrad or better, 35% at monad or better and 2% at 100m precision or better. Additionally, the DDb has 432,000 records from 1930 - 1945 (using end year for records with date ranges), of which 99% have grid references: 80% at tetrad or better, 78% at monad or better and 68% at 100m precision or better. (The very high % of 100m precision records is mainly due to Professor Ronald Good's detailed mapped records in Dorset).
While many (most?) of these grid references will be perfectly accurate, it is important to remember that the given grid reference was not part of the original record but was added later, most probably by someone other than the original recorder, and possibly without local knowledge. This has a number of implications for validation of such historic records.
• If there is a discrepancy between the site name and the grid reference of a record, and this discrepancy cannot be resolved, for modern records the record would need to be treated as doubtful or rejected. But for old records (pre OS and OSI grids) only the site name is a valid (though possibly vague) element of the record; an incompatible grid reference can be discounted (edited or removed).
• The grid reference may be over precise. The local VCR is best able to make this judgement. In densely populated lowland areas, with a multitude of place names, it will generally be acceptable to apply quite precise grid references to old records (eg. tetrad or even monad resolution). But in upland areas especially, location names may refer to large and ill-defined areas, where at best a hectad grid reference is the best that can be assigned. Large upland parishes may cover multiple hectads and it may be best to not give a grid reference, only a site name and vc.
• Over precise grid references may distort inferred altitudinal ranges. This may be a particular issue in upland areas where there has been a tendency to assign a summit grid reference to records that use that hill or mountain as a site name.
• Details of the same record, or duplicate herbarium specimens, may have been entered into computer databases more than once and different grid references assigned to each version. At worst grid references in different hectads might have been used, hence inflating the apparent distribution of a species. This is especially apparent where a major geographic feature runs north – south, or east – west.
• The law of parsimony suggests that we should aim for consistency, assigning the minimum number of grid references to groups of duplicate or probable duplicate records. The DDb (on the 'my county' - Data validation page) already flags records with site names that may be inconsistently applied to grid-references (i.e. the site name is associated with two or more unrelated locations). A VCR can usefully investigate this potential issue in more detail by running a query that lists all site names (for old records) with the number of tetrads (or monads) that each site name is used in. Adapt this query as appropriate:
https://database.bsbi.org/search.php#re ... cff5e5320a
Records where a site name is used in multiple grid squares would be worth investigating; are the different grid references justified?
None of this is to suggest that all grid references of old records are wrong, but they should be re-examined carefully when validating records on the DDb.
This post was edited / updated on 7/3/2018.
Andy Amphlett
The Ordnance Survey grid did not appear on maps of Great Britain until post 1945, with the publication of the Popular Edition of the One inch to 1 mile scale maps. In Ireland, the Irish Grid did not appear on generally available maps until the 1970s, when it was used on the 1:126,720 (half inch to the mile) series. (For recording for the first Atlas of the British Flora (Perring & Walters 1962) the OS grid for GB had been extended to Ireland by David Webb).
The DDb has >443,000 pre-1930 records (using end year for records with date ranges), of which 98% have grid references: 47% at tetrad or better, 35% at monad or better and 2% at 100m precision or better. Additionally, the DDb has 432,000 records from 1930 - 1945 (using end year for records with date ranges), of which 99% have grid references: 80% at tetrad or better, 78% at monad or better and 68% at 100m precision or better. (The very high % of 100m precision records is mainly due to Professor Ronald Good's detailed mapped records in Dorset).
While many (most?) of these grid references will be perfectly accurate, it is important to remember that the given grid reference was not part of the original record but was added later, most probably by someone other than the original recorder, and possibly without local knowledge. This has a number of implications for validation of such historic records.
• If there is a discrepancy between the site name and the grid reference of a record, and this discrepancy cannot be resolved, for modern records the record would need to be treated as doubtful or rejected. But for old records (pre OS and OSI grids) only the site name is a valid (though possibly vague) element of the record; an incompatible grid reference can be discounted (edited or removed).
• The grid reference may be over precise. The local VCR is best able to make this judgement. In densely populated lowland areas, with a multitude of place names, it will generally be acceptable to apply quite precise grid references to old records (eg. tetrad or even monad resolution). But in upland areas especially, location names may refer to large and ill-defined areas, where at best a hectad grid reference is the best that can be assigned. Large upland parishes may cover multiple hectads and it may be best to not give a grid reference, only a site name and vc.
• Over precise grid references may distort inferred altitudinal ranges. This may be a particular issue in upland areas where there has been a tendency to assign a summit grid reference to records that use that hill or mountain as a site name.
• Details of the same record, or duplicate herbarium specimens, may have been entered into computer databases more than once and different grid references assigned to each version. At worst grid references in different hectads might have been used, hence inflating the apparent distribution of a species. This is especially apparent where a major geographic feature runs north – south, or east – west.
• The law of parsimony suggests that we should aim for consistency, assigning the minimum number of grid references to groups of duplicate or probable duplicate records. The DDb (on the 'my county' - Data validation page) already flags records with site names that may be inconsistently applied to grid-references (i.e. the site name is associated with two or more unrelated locations). A VCR can usefully investigate this potential issue in more detail by running a query that lists all site names (for old records) with the number of tetrads (or monads) that each site name is used in. Adapt this query as appropriate:
https://database.bsbi.org/search.php#re ... cff5e5320a
Records where a site name is used in multiple grid squares would be worth investigating; are the different grid references justified?
None of this is to suggest that all grid references of old records are wrong, but they should be re-examined carefully when validating records on the DDb.
This post was edited / updated on 7/3/2018.
Andy Amphlett