Laserfiche WebLink
<br /> <br />-30,. <br /> <br />F~equency'analyses <br /> <br />Several techniques are, used ,for frequency analyses of precipitation <br />data, according to the particular type of study. Care must be taken that <br />the data used in the frequency analysis represent the same time interval. <br />For example, hourly precipitationrebordsare publishep for clock hours <br />and do not necessarily representthe'maximum precipitation in any one- <br />hour period. Likewise, reports from daily precipitation stations may <br />not represent maximum 24-hour precip~tation amounts. Various technical <br />papers that have been published by the Weather Bureau will, in many' cases, <br />furnish the required frequency infoI'1flation. <br /> <br />Estimating missing records <br /> <br />. <br /> <br />Missing records should be estimated only when it is reasoned that <br />the estimated data will be of definite advantage. Simple ratio methods <br />of normals or averages (based on a common periOd), or comparison of <br />slopes on double-mass plottingsmaJ be used for estimating yearly or <br />monthly amounts from adjacent stations. The Weather Bureau estimates <br />missing recordsbythenormal,.ratio method and, when they are available <br />from daily snowfall amounts according to the conventional density of 0.10, <br />these estimates are included in the monthly totals published in the <br />~nnual climatological bulletins. It, should be noted, however, that the <br />~atio between stations may change frpm "season to season. Missingdaily <br />or storm amounts generally are best estimated by constructingisohyetal <br />maps., For shorter intervals the most desirable procedure would be to <br />compare mass curves with those of adjacent stations. Hour of measure- <br />ments should be reviewed to assure that the records used are comparable. <br /> <br />Statistical analyses <br /> <br />In using precipitation records for "statistical analyses the greatest <br />pitfall is non-consistent records. Accordingly, a check should be made <br />on the consistency of all records prior to such analyses. Inconsistency <br />in a record, for example, tends to increase the apparent auto or serial <br />correlation. <br /> <br />Early published data <br /> <br />Early published data, especially those'in "Bulletin W," include marry <br />estimates of monthly and yearly precipitation. Certain'of these estimates <br />~ave been shown to be in errorJ depending upon the intended use of the <br />records, adequate checks shciuld be milde of these estimates. In tabu- <br />lating the records in "Bulletin W," no effort was made to check their <br />consistency. For some stations the data actually were composed from two <br />or more sites fairly well separated in distance or in altitude. Footnotes <br />were used to indicate when records were so combined. <br />