times bring about unexpected results. Observers with experience are alert to these possibilities; the inexperienced observer must learn them. In general, if the shelter is distant twice the height of an object there will be no errors caused by reflection or by absorption from that object.
Anomalies of Temperature.—As a rule, minimum temperatures—and they usually occur just before sunrise—are less apt to be affected by unusual conditions than are maximum temperatures. The minimum temperatures on the south side of a building are usually the same as those on the north side. In prolonged hot spells, however, this does not always hold true. The walls of the southerly exposure may absorb so much warmth during the day that not all of it is radiated at night. As a result, a minimum registered under such conditions will be too high.
The prevalence of a stiff wind, especially the northwest wind of cold waves, equalizes temperatures to a remarkable extent; the minima of stations covering considerable areas rarely vary more than 1 or 2 degrees. On the other hand, on very still nights the minima of stations only a few miles apart may vary several degrees.
On very cold, still nights cold air tends to settle by gravity into low spots. This condition is so marked that the minima of localities only a few rods apart may vary as much as 2 or 3 degrees. This difference is very noticeable in mountain valleys where the cold air is apt to flow down the valleys at night. Frosts occur much more frequently along valley floors than in the foothills and the benches higher up.
City and suburban temperatures usually have about the same daily means, and their monthly averages should not vary more than a degree. The daily maxima and minima not infrequently vary several degrees. This is due chiefly to the fact that the less amount of smoke and dust in suburban localities favors absorption of heat in the day time and permits radiation at night.
In early fall and also in late spring, frost may be observed in sheltered places on the ground when the thermometer registers several degrees above freezing. An observer may therefore conclude that his minimum thermometer is not registering correctly. It is not likely than an error has occurred; ground sur-