Areal Basis Research Articles

Cluster Analysis in Urban Demography

IN A previous paper in Social Forces the author suggested a method for deriving of phenomena based on similarity of locational index patterns among census tracts.' However, the method was limited to deriving these patterns or clusters when the several phenomena formed an exhaustive breakdown of some general class of events. The example employed was a breakdown of total employment by census tracts into nine broad occupational classes. This paper presents a technique applicable even when the variables under consideration are not an exhausitve subclassification or when they are entirely separate categories of events. The previous paper also pointed out what was termed a problem in studies of association based on small areal units comprising a contiguous social and economic area. This was a recognition that the relative emphasis accorded two principal aspects of geographical association varies among studies. Some studies seek to determine the extent of areal association of variables of interest in the entire urban area. Thus the degree of association between the incidence of various kinds of crimes by census tracts might be studied. Other studies place greater emphasis on combinations of events found locally associated. Thus an investigator might seek to classify individual tracts into several categories representing commonly associated crimes. This second approach leads frequently to attempts to delineate larger areas based on criteria of similarity of characterization and often of contiguity. Our investigator might establish belts or sectors in which different combinations of specific crimes predominate. In this approach, measurement of the over-all association of the various identifying events is frequently by-passed or made secondary, although it is implicit in the whole analysis, especially when attempts are made to delineate the boundaries of areas having combinations of characteristics in common. The method suggested in employing this approach in the paper previously cited recognized this dual aspect in prividing a test of predominant patterns in which one would seek to reject a hypothesis of a random coincidence of events characterizing individual areas with the suspected pattern. To reject such a hypothesis is of course to conclude that association does exist in the locational distributions of the characteristics comprising a pattern, although no measure of this association was then provided. The technique presented here employs the alternative approach. Just as one can begin with a characterization of areal units and end with an identification of sets of variables which are intercorrelated on an areal basis, he can alternatively begin with the variables themselves with their areal intercorrelations and proceed to separate out directly any sets of highly intercorrelated variables. It is evident that the circle can be completed by then identifying those areal units whose characteristics correspond in varying degree to any of the revealed sets of associated phenomena.

Appendix A—Flood‐ and drought‐events

Forty‐four more or less distinct flood‐events are listed chronologically in Table 1. A flood‐event as used herein is a flood of such magnitude as to be newsworthy. Any event may or may not have been associated with a single meteorological disturbance. It will be noted on Figure 1 that these events occurred during three more or less distinct periods and frequently are grouped on an areal basis. Twelve of the events occur between March and November, 1941, 28 between March, 1942, and August, 1943, and three during January, February, and March, 1944. There was almost a complete absence of flood‐events during 18 of the 40 months. Nation‐wide, December, 1942, was the wettest with respect to departure of stream‐flow from normal. Seven of the 44 flood‐events were associated with high temperatures, melting snow, and rain. Fourteen were widespread and record‐breaking and confined most generally to the central and southeastern states. The remaining 23 events were generally of a local nature but many broke existing records. Geographically the largest number of events were in the central part of the United States embracing Oklahoma, Kansas, Missouri, and Iowa and adjacent areas, and another is now in progress (May, 1944). Oregon was an exception so far as the West was concerned, with continuous flood‐events from November, 1942, to April, 1943. Events No. 36 (May, 1943) and 44 (April, 1944) are probably of greatest National importance and Nos. 19 and 20 (July, 1942) and No. 41 (August, 1943) of greatest local importance and are of special significance in that precipitation‐rates may have set new limits for storm‐magnitudes.

The Yield of Streams as a Measure of Climatic Fluctuations

M A UCH of our national economy depends on the extent to which fluctuations in the yield of streams can be overcome and stable water supplies for power, irrigation, industry, and municipal use be assured. To the water user directly and to the whole nation indirectly, the fluctuations in stream flow are of profound significance. Stream flow is dominantly the integrated result of many diverse climatic factors. Its volume and rate are sensitive to climatic variations, responding to daily changes in weather and also, in higher degree, to seasonal and yearly changes. The many records of stream flow now available provide not only information on water supplies but a register of the related climatic factors also. To the meteorologist interested in climatic fluctuations or the geographer interested in defining relations between climatic elements and dependent geographic factors, analysis of the stream-flow records may furnish a valuable side light. To meet such interests, the results of analyses of more than four thousand station-year records at selected gauging stations are here presented in the form of a series of maps showing for the United States and Canada the percentage of normal stream flow year by year I9 II to I942 inclusive, and in tabular form wet and dry indices on an areal basis for the five wettest and five driest years.