Now it is easy to conceive how the changes of density which we can thus produce artificially may be produced in nature. If, in serene weather, the surface of the sea is much colder than the air of the atmosphere, as it frequently is, and as it was to a very great degree during the phenomena described by Mr. Scoresby, the air next the sea will gradually become colder and colder, by giving out its heat to the water; and the air immediately above will give out its heat to the cooler air immediately below it, so that the air from the surface of the sea, to a considerable height upwards, will gradually diminish in density, and therefore must produce the very phenomena we have described. [Illustration: _Fig. 36._] The phenomenon of Dover Castle seen on the Ramsgate side of the hill, was produced by the air being more dense near the ground and above the sea than at greater heights, and hence the rays proceeding from the castle reached the eye in curve lines, and the cause of its occupying its natural position on the hill, and not being seen in the air, was that the top of the hill itself, in consequence of being so near the castle, suffered the same change from the varying density of the air, and therefore the castle and the hill were equally elevated and retained their relative positions. The reason why the image of the castle and hill appeared erect was, that the rays from the top and bottom of the castle had not crossed before they reached Ramsgate; but as they met at Ramsgate, an eye at a greater distance from the castle, and in the path of the rays, would have seen the image inverted. This will be better understood from the preceding diagram, which represents the actual progress of the rays, from a ship SP, concealed from the observer at E by the convexity of the earth PQE. A ray proceeding from the keel of the ship P is refracted into the curve line P _c x c_ E, and a ray proceeding from the top-mast S, is refracted in the direction S _d x d_ E, the two rays crossing at _x_, and proceeding to the eye E with the ray from the keel P uppermost; hence the ship must appear inverted as at _s p_. Now if the eye E of the observer had been placed nearer the ship as at _x_, before the rays crossed, as was the case at Ramsgate, it would have seen an erect image of the ship raised a little above the real ship SP. Rays S _m_, S _n_, proceeding higher up in the air, are refracted in the directions S _m m_ E, S _n n_ E, but do not cross before they reach the eye, and therefore they afford the erect image of the ship shown at _s´ p´_.