Nor is it the absolute elevation alone that is to be taken into the account; in any given locality, we may assume that the higher the orchard is situated above the water levels, the safer it will be, and that the lowest depressions are the most unsafe or frosty. It is not always the mere elevation, but rather the relative elevation of the site, that renders it more desirable than another in the same region. There are many orchards that are situated upon a moderate bluff, with a rapid descent of only a few feet or yards, into a swale or valley of moderate extent; these we find to be uninjured, when another at a greater elevation, but in a depressed basin surrounded by higher lands, will be found to have suffered from the influence of frost. In the one case, the cold air could flow off rapidly into the adjoining depression, while in the other, the cold air from adjoining slopes would collect, and accumulate in the situations described. In the great plateaus of the world, we often find immense tracts of land so nearly of the same level as scarcely to afford sufficient drainage for the surplus water; of course, we should expect to find, in such places, little variation of temperature arising from difference in elevation. But even in such situations, whether we examine the table-lands of our timbered regions, or the extended areas of the prairie country, we shall find that the drift formation which covers these vast tracts, has not been distributed evenly, but that there are successive rolls or swells frequently recurring, which give, in some instances, considerable variations of level. A bold ridge, of fifty feet or more in hight, rises abruptly from the level prairie, stretching along for miles, and affording admirable exposures for orchard sites. Such places are observed to be free from late and early frosts. In other places, there is an abrupt depression of the surface, answering the same purpose--drawing off the cold air. These may be very moderate in their extent, as the prairie sloughs, or they may be small vales, the courses of the minor streams, or of larger extent, the valleys of rivers, or the depressions of lakes. In these latter cases, the modifying influences of considerable bodies of water enter into the frost problem as an element of no mean value.