Strange coordinate shift between small-scale and large-scale Ordnance Survey vector data sets (Strategi vs. OpenMap Local/VectorMap District)

Question

I'm trying to combine data from Ordnance Survey's small-scale Strategi data set with some of their large-scale OpenData data sets (in particular OS Open Map Local and OS VectorMap District) to fill in local detail. But even though the small-scale data sets are clearly derived from the same underlying Ordnance data, there is a systemic coordinate shift between the data sets, as you can see in the screenshot below. The area shown is a closeup of the West coast of Shetland, British National Grid reference square HP, the layers with the HP_ prefix are from the OS VectorMap District data set, the remaining layers whose data is shifted around 120 meters to the West are from Strategi.

Apart from the obvious simplifications and omissions of the small-scale data set I have not been able to find any information on how or why the data from the two data sets should differ in terms of projection or position, as all Ordnance Survey data sets come with the identical EPSG:27700 projection.

Does anyone have any idea where the coordinate shift might come from? Ideally I would like to be able to correct for it in order to combine both types of data sets in my map. It seems highly unlikely that geometry simplification alone could cause an accidental 120 meter shift of the data.

Answer 1

Strategi has been discontinued by the OS. It looks, to me, more like the 1:1,000,000 data. At this scale, each pixel is 1000 feet so matching it with small scale data will always be 'challenging'

Answer 2

Kevin, in cartography there is the notion of cartographic generalization and you have come across it voohuyu. In fact, this is a problem of creating, updating, visualizing and storing data. Previously, 30-40 years ago, cartographers did not have tools to organize data storage using modern tools such as Oracle or Postgres and therefore,they came up with such a cunning, namely, to change the configuration and shapes of objects, in order to display them on maps of different scales, and when this ideology was replaced by automation, it revealed all the shortcomings. As a result, in order to avoid a double (triple, quadruple, etc.) divergence, you need to use data from a larger scale or create from a multi-scale data one, as I call it a hybrid data scale. I hope that the translation did not greatly distort my reasoning, with respect