Abnormal rainfalls of long duration and high intensity are common in the semi-arid region of Mexico. They come at irregular intervals, though tending to coincide with the early fall. The floods of August, 1909, were a repetition of similar occurrences in the past; and, though there are no numerical records of previous cases, local traditions and historical state documents describe them as having occurred since the foundation of the city, at intervals of from 15 to 40 years. The graphic descriptions of the places flooded are in accord with the character of the floods of August, 1909, and September, 1910. The diagram, Fig. 21, is a record of the rainfall during the latter flood, and was plotted from intermittent readings of standard gauges. It demonstrates that the intensity increased toward the mountains on the south, which form the tributary water-shed of the Santa Catarina River, showing a difference of 10.54 in. between the city and the Estanzuela Dam, which is not quite 12 miles to the southeast. [Illustration: FIG. 21.--RAINFALL DURING FLOODS OF SEPTEMBER 14TH-16TH, 1910, IN MONTERREY.] An estimate of the volume of discharge of the river at the time of maximum flood is only a reasonable conjecture which (without special reference to accuracy) aims to impress those who have not witnessed such occurrences with the tremendous volume coming from barren steep surfaces previously saturated. The original computation, referred to by the author, was obtained from the average of two different methods which gave results close to each other. In one method the extent and nature of the water-shed were considered, together with the maximum period of precipitation that occurred, sufficient to gather a maximum volume of water in the river. In the other method the volume was derived from a cross-section of the wetted perimeter of the river at the time of maximum flow, in combination with velocity approximations obtained by using rough floats. This gave 271,500 cu. ft. per sec. The figure submitted by the author, 235,000 cu, ft. per sec., is in accord with the proposed formula[9] for impervious surfaces by C. E. Gregory, M. Am. Soc. C. E. In the first and last methods, the intensity, a governing factor, is more or less of an assumption, and the cross-sectional method is also unreliable, as the river-bed was greatly disturbed, due to the high velocity of the water, which deepens the channel to a considerable extent at times of maximum flood, the gravels being redeposited during the period of subsidence. Such was the case during the flood of September, 1910, when the depth of gravel above the roof of the San Geronimo Infiltration Gallery was diminished to such an extent that it was so inefficient as a filter for the flood as to permit the percolation of turbid water into the underground supply. [9] _Transactions_. Am. Soc. C. E., Vol. LVIII. p. 458.