A chance trick of the light has provided proof that the town of Al-Qasr in the Dakhla Oasis was once a Roman fortress. Jenny Jobbins witnessed the evidence

The town of Al-Qasr, otherwise known as Qasr Dakhla, lies in Dakhla Oasis deep in the Western Desert 450kms due west of Luxor. Despite its remote setting it has had a colourful history: Romans exploited the oasis for agricultural produce; Libyans, including the Sanusi, made conquering raids; and it was not far from the infamous Darb Al-Arbain slave route. In the picturesque mediaeval section of the town narrow, partly covered streets wind past heavy ancient doors topped with elaborate lintels, and here and there through an open doorway can be glimpsed old grinding stones or a staircase leading to a crumbling roof.

Al-Qasr is the older of the two towns in Dakhla -- the other being Mut -- and was built on top of a tell, that mound of crumbled debris that marks the site of an ancient structure or settlement and which over time, since any collapsed buildings are composed largely of mud brick, settles into the natural landscape.

Archaeologists have long supposed that beneath the foundations of Al-Qasr are the remains of a Roman citadel. Fred Leemhuis, professor of Islamic Studies at Groningen University and field director of the Qasr Dakhla Project -- part of the Dakhla Oasis Project (DOP) -- told this author two years ago: "Undoubtedly there was a fortress there in Roman times, or even a Ptolemaic one. The Romans probably built a structure to surround the well, and I would be surprised if there was nothing Roman. But we have simply not found any evidence."

Leemhuis said then that a good quantity of datable potsherds or coins at a lower level would be enough to determine a Roman provenance. "Al-Qasr is the only tell in Dakhla. It has at least three levels of occupation," he said. "There might have been something before that, but we can only hazard a guess. Some time I'm going to excavate, but right now I'm too busy."

In the end, by a quirk of fate, he didn't need to. It happens to most of us: the mislaid glasses you find have been wearing all along, the lost car keys which were under your nose. Like those keys, the evidence was right before his eyes. "Archaeologists had been walking past it all the time," Leemhuis said this week. "They just didn't notice it."

What caught his eye was an outcrop of what had always been thought -- if any thought was given to it at all -- to be an outcrop of dried mud beneath a disused mosque on the edge of the old town. One morning this February Leemhuis was walking past the "rock" when he noticed that the sun caught a distinct line that appeared to be a course of brickwork. He called in the project's chief restorer, Rizq Abdel-Hay Ahmed, and local inspector Affaf Saad Hussein, and together they examined it more closely. Under the veneer of sun-baked mud they could distinguish several such courses. Far from being hardened earth this was mud-brick, and, moreover, the size of the bricks -- each 8x16x33 cms -- corresponded exactly to bricks in other Roman fortresses in the Western Desert. Since then other experts, including Roger Bagnold of Columbia University -- who has also walked past it many times -- have agreed the wall is Roman.

The brickwork continues on the other side of an open street which was at some point driven through it, and it went on again behind the mosque. This last piece of the wall, which still stands more than four metres high, was evidently a gateway and abuts what appears to be a circular or hexagonal tower. The wall has proved to be six metres thick and the stone foundations to have been dug down one metre. "Now we know it's there we can't think how we missed it," Leemhuis says. "And we didn't have to do anything."

Further excavations will have to be carried out before the wall can be dated with accuracy, but one historical conundrum now appears to be solved. Agricultural accounts from the Roman town of Kellis, the site of which lies between Al-Qasr and Mut, show records of grain and wine being sent to a place named, in Greek, "Takastra". Up to now no one has known where this might be, but now it can be surmised that Takastra, "the camp" from the Latin castra (military camp), later became Qasr, making its etymological link with the Arabic qasr (fortified town) obscure.

The discovery sets the seal of what has been a successful season for the Qasr Dakhla Project, which has also seen the completed restoration of the second of two adjoining houses at the centre of the town: the Beit Al-Qureishi, which belonged to a prominent local family. Unlike its neighbour the Beit Al-Qadi (the house of the judge), which was restored two years ago, the Beit Al-Qureishi was a ruin with only the façade left standing. Both houses have now been rebuilt using the traditional methods of the original construction: mud bricks with wooden supports and with roofs of palm logs and fibres. The window frames and doors are made of acacia wood. The houses have steep staircases -- two in the case of the Qureishi house -- leading to the upper floors. These stairwells have niches for oil lamps to compensate for what might be a gloomy interior, the main source of light being the small upper windows.