Laserfiche WebLink
<br />001451 <br /> <br />SOURCES OF ERROR IN HYDROLOGIC RECORDS <br /> <br />GENERAL CONSIDERATIONS <br /> <br />The preceding part of this manual has considered limitations that <br />are inherent to hydrologic records and data, even when those records <br />comprise accurate measurements or determinations. Here considered are <br />the more common sources of error in the records--error in the sense <br />that the qpantities recorded differ from those that actually pertained. <br /> <br />Errors have been, and always will be present in hydrologic records, <br />largely because the instrumentation and procedures of measurement are of <br />necessity !l compromise between the ideal and the practiCal. Over the years, <br />instruments and procedures have improved so that certain sources of error <br />have diminished; further improvement will ensue but error-less records <br />probably never will be realized. <br /> <br />Certain errors are random in time and in magnitude; and may compensate <br />within a period shorter than the term of study. If so, conclusions derived <br />from the study may not be influenced substantially. Other errors may be <br />systematically plus or aystematically minus, or, even if' random in magni- <br />tude and algebraic sign, they may be much more prevalent in a particular <br />part of the record. In this event, conclusions may be distorted if' the <br />errors are not discriminated. <br /> <br />Further, most hydrologic records have been tabulated, totaled or <br />averaged, typed, and proofread by human beings--of whom all are prone <br />to error now and then. Misplaced decimal points, transposed or extra <br />digits, and typographic errors occasionally survive meticulous checking <br />and proofreading, and so become published. Some such errors can be <br />discovered:readily and the cause may be deduced by a diligent user of the. <br />records--for example, in a series that defines a relatively steady <br />hydrologic state, an erroneous value commonly can be discriminated with <br />assurance.. On the other hand, a similar error in a series that spans an <br />unsteady state could be isolated only by going back to the original data. <br /> <br />In a presumably continuous record it is obvious that "the maximum <br />today must be at least as great as the minimum of yesterday, today, and <br />tomorrow; and the minimum today must be at least as small as the maximum <br />of yesterday, today, and tomorrow." Philosophically that principle is <br />inviolate, yet some violations that have been published are too numerous <br />to be considered inadvertent or typographic. The philosophically impossible <br />record must be verified or discarded. <br /> <br />In ev~ry hydrologic study, therefore, an early step should be to test <br />the pertinent data for internal consistency and, as has been stated, to <br />discriminate data that appear abnormal and may be susceptible to corrective <br />adjustment. Tests must be devised to suit objectives of the particular <br />study. Co~on procedures are double-mass plotting of records for a given <br />variable from various stations, pairing mass diagrams of unlike but related <br />variables, ,"routing" stream.flows through successive stations, and arraying <br />data by magnitude and examining the environmental aspects of extreme values. <br />Other procedures will be suggested by the kind of study. <br /> <br />6 <br /> <br /> <br />- :,i.-- < <br />