In selecting sites for their forts, monasteries, and other buildings, the Romans were naturally drawn to eminences commanding views of the surrounding country. It was, of course, desirable, and in the case of forts absolutely necessary, to have a supply of water within the building, and it was the custom, therefore, to sink a deep bore within the precincts, generally in the centre of the courtyard. In many cases, however, owing to the elevation, the water would not rise to the surface, the supplies being drawn up by hand. In order to make the water of such a well available for irrigating the lands situated outside the fort, one or more gently inclined tunnels were excavated so as to tap the bore below the fort; the well was thus made to serve a double purpose. Excellent examples of such conduits are to be seen at Dêr el Ghennîma. Here the well is in the middle of the courtyard, and three underground tunnels converge on it from the low-lying and anciently cultivated lands to the north and north-west. Along those portions where the bottom of the channel was in soft ground, and not more than a metre or two below the surface, the conduit was made in the form of an open trench, the sides being carefully built in with stone. [Illustration: QASR LEBEKHA AND THE NORTHERN ESCARPMENT OF THE OASIS.] When the excavation of the trenches was completed, the open portions were covered over with large flat slabs of rock, so that the channels were not only well protected from blowing sand, but quite invisible on the surface. The underground conduits at Dêr el Ghennîma were indeed quite unsuspected before they were accidentally discovered a year or two ago. The tapping of the bore within the fort had, of course, the effect of still further lowering the water-level as regards the courtyard; it is quite likely, therefore, that the inhabitants had a means of blocking the underground channel when they desired the water to rise to its maximum level in the well. But in any case the few extra metres of distance from the surface to the water could not have been a matter of importance, as the comparatively small supplies needed for domestic purposes within the fort were doubtless raised by means of a bucket and rope attached to a windlass.