Background

In Britain in old photographs: Luton, the author Stuart Smith, reproduces on page 126, a photograph of an old house which he states that he had been unable to determine the location of. Stuart Smith goes on to say that he believed that the house was demolished around 1970, and was possibly known as Bedford House.

During the early 2000’s, after seeing the photo, I did have a cursory scan over an old 1922 Luton map that I had, and identified Highfield House as a building that clearly was not marked on the modern Ordnance Survey map. From the historical map, Highfield had approximately the same geometry as the one in the historical photograph. That being said, I decided at the time to shelve the project for the time being.

Over Christmas 2023, I thought that I would take a closer look at identifying the house again, this time applying a more structured approach to the matter.

Some initial clues

Taking a fresh look at the problem, I annotated the photograph with possible clues that might be of importance in geolocating the building.

From the above exercise, two important features immediately presented themselves:

• The left hand side of the house was illuminated by the Sun, suggesting that it faced a Southerly direction. Thus I had an approximate orientation of the building.

• The second feature was that the building appeared to be on the side of a hill. There is a slight ground incline rising towards the left, and a tree line moving downwards from left to right.

The hill gradient

Luton is a town that sits mostly in the River Lee valley and is flanked on the West and Eastern side by a section of the Chiltern Hills. Given that the house was illuminated by the Sun on the left hand side, with the apparent ground gradient falling off to the right hand side of the photograph, I hypothesised that the building would probably be located on the Western hill side of the town. If the building was located on the Eastern hills it would probably be unlikely to be illuminated on the left hand side.

The tree line

Now if I were able to compare the gradient of the tree line (using it as a proxy for the ground) in the historical photograph with that of the location of Highfield House on the 1922 Luton map, it might be possible by comparing each gradient in order to assess whether they are the same or different, i.e. a bit like using the gradient as a kind of fingerprint for the location. Clearly there is lens and perspective distortion in the historical photograph, although in situations such as this one often has to work with incomplete and fuzzy information, thus making best use of the available information. None of this falls into the category of designing a standard science experiment! Also, I made the assumption that the trees were about the same age and thus had grown at much the same rate.

I imported the historical photograph into a CAD package and marked a series of points (the yellow dots) along the tree line.

From the above, the x/y coordinates of the points were exported from CAD, giving me a small data set from which I could compute the linear gradient of the tree line as it moved down the side of the apparent hill side.

LiDAR elevation gradient

The next stage was to extract the LiDAR elevation profile from the area that the tree line was located, which is marked on the historical map. I scanned the historical map, georeferenced it, then downloaded the appropriate 1m resolution LiDAR tile for that area, created a section line and extracted the elevation data along it.

Next the elevation gradient for the LiDAR dataset was computed.

The computed elevation gradient for the tree line being -0.12593805 and for the LiDAR -0.14752863, fairly close with the difference between them of 0.02308826.

Comparing the two gradients

The next step was to make a statistical comparison between the two gradients. Are they significantly different from each other, or are they approximately the same? To answer this question I developed a Bayesian model and computed it using Markov Chain Monte Carlo simulation in order to find the error distribution of the differences between the two gradients.

The point estimate of the difference of the two gradients (as above) was -0.02308826, this value falls nicely inside the 95% confidence limits, the yellow region in the plot to the side, so from a statistical perspective it may be concluded that the two gradients are about the same. From this result there is a good chance that the hill side in the historical photo is the same as the on-the-ground elevation of that section of the modern map. The two “fingerprints” match!

A conclusion

So to conclude, I will state from the analysis above, it looks like the “Unknown house”, stands a good chance of being Highfield House that was located on London Road in Luton. The OS grid reference of where Highfield House once stood is TL 08830 20470.

Of course there are other pieces of evidence that really need following up on. I had hoped to obtain some lower altitude historical imagery of the area so that I could perhaps check the house structure further. There was an RAF image that was taken of Luton in the 1950’s, when Highfield was still standing, but unfortunately this was taken from 25,000ft and thus too low resolution to be able to see any fine details.

One snippet of evidence that did crop-up, I managed to obtain copies of planning applications concerning Highfield from Luton Council. The most recent application in that list mentioned that 6.356 acres of the Highfield site had been made available for residential development at around the time period that Stuart Smith had stated in his book that the original photograph was taken. Certainly the houses on the site today look to date from the early 1970’s.