AREAL PRECIPITATION ESTIMATES FROM POINT MEASUREMENTS

During the 8-year period; 1968 through 1975, a program was undertaken on the 42 square mile Bad Lake Watershed "Dense Precipitation Network" located in the semi-arid area of Southwestern Saskatchewan to study the spatial and tem­poral character of precipitation events. The "Dense Precipi­tation Network" included 41 recording precipitation gauges which provided a gauging density of approximately 1 per square mile. The primary purpose of this research was to obtain information on the character of rainfall over small areas on the Canadian Prairies which would broaden the data base for future hydrologic research and assist in the design and evaluation of other precipitation gauge networks in the Prairie region. Frequency distributions of maximum storm amounts and intensity-duration-frequency studies indicated that the pre­dominant rain storm activity on the prairies is associated with low intensity, small rainfall events. The majority of storms produce maximum point rainfall amounts of less than 0.11 inches, and account for only approximately 4.6% of the total summer rainfall. Storms having a maximum point rain­ fall depth greater than 0.50 inches produce the major part of the total seasonal (May-September) rainfall. For storms covering the entire basin the statistic, coefficient of variation (CV), was used to differentiate between uniform and non-uniform storms. Storms having a value of CV ≤ 30% were considered uniform. Approximately 54% of the total number of storms that occurred over the entire network had values of CV < 30%. The majority of the monthly or seasonal rainfall totals at a particular location were provided by uniform storms. Integration of storm rainfall amounts over extended time periods leads to a decrease in the spatial variability of rainfall. The spatial variability for monthly rainfall totals are generally lower than individual storms occurring during that month. During the five-year period; 1971-1975, values of CV for monthly totals varied from 9.7% to 41.3%. A further decrease in the spatial variability of rainfall occurs whenever data is integrated over time periods greater than one month. Values of CV for seasonal rainfall ranged from 5.1% to 16.3% while the value of CV for the 5-year sea­sonal total was 3.1%. Correlation coefficients (r) calculated between gauges on the network using data from uniform and non-uniform storms indicated that the rate of correlation decay was more rapid for non-uniform storms than for uniform storms. Assuming that a correlation coefficient (r) of 0.90 between gauges is adequate for describing storm rainfall detail, a single gauge could be used to provide acceptable estimates of the precipitation at all other stations within the network. For non-uniform storms a minimum of 15 gauges were required. The spatial character of a precipitation event as determined from various networks having different gauging densities may be statistically different. Using standard statistical tests, it is shown that; the positioning of gauges, the network gauge density, the spatial variability of the event and the length of period over which precipitation is totalled affect the representativeness of the precipita­tion network. The minimum gauge densities (gauges/mile2) required to preserve the "dense" network statistics; xd and sd2, on the 42 square mile watershed for storm, monthly, seasonal and 5-year seasonal rainfall amounts were found to be; 0.43, 0.38, 0.10 and 0.05 respectively