But, to continue the story of a thunder-cloud, we always find that after a time the cirriform top flattens out and gradually subsides, and this is usually accompanied by a descent of the cloud base to a lower level. Meanwhile, it frequently happens that the whole series of phenomena is repeated in one of the attendant cumulus. Plates 51 and 52 are also two views of the same cloud at different times. In Plate 51 we have the main part of the storm on the right, while on the extreme left a lower part of the cloud is rising rapidly into a tall dome. In Plate 52 the central top has lost its cirriform margin and has distinctly flattened, while the left-hand dome has risen much higher and is beginning to throw out the projecting bits. [Illustration: Plate 51. THUNDER-CLOUD. (_Cumulo-nimbus._)] [Illustration: Plate 52. THUNDER-CLOUD. (_Cumulo-nimbus._)] The hard-topped cumulus which fringe the lower disc, and the vast pile of cirriform and hazy cloud which forms the centre of a cumulo-nimbus, are shown in Plate 53, which represents part of the side of a great thunder-cloud. In this case the diameter of the lower disc was about 15 miles, and the upper disc was rather larger. The uprising column in the middle was about 7 miles across, and the height from base to summit about 3 miles. The whole system contained between 100 and 150 cubic miles of cloud. When photographed it was over the northern part of Salisbury Plain. Lightning played repeatedly between the back of the white cumulus and the hazy mass behind it, and the rumble of thunder was all but continuous for nearly half an hour as the great cloud passed by. [Illustration: Plate 53. THE FLANK OF A GREAT STORM.] This was an unusually large cloud for this country, but specimens of 10, 20, or 30 cubic miles are quite common. Now for the explanation of the series of events. To begin with, we have the production of an ordinary cumulus, but the equilibrium is unstable, the growth of the cloud, therefore, becomes more and more rapid, and the rapid condensation adds to the instability until the rising column is so much lighter than an equal column outside that a powerful updraught is created, strong enough for a time to hold up the raindrops or even hailstones. At length the condensation is complete, the upper part of the cloud consisting of snow crystals exactly like those of any other cirrus. In the mean time, the rapid ascending current necessarily involves an indraught from around, and consequent descending currents to supply it. The result is to set up a circulating system, moving inwards along the ground, upwards in the central column, and outwards in the upper disc. The downward currents are sometimes shown by a curling over of the edges of the upper disc, but the phenomenon is not often seen, as the descending movement is generally enough to dry up the cloud particles.