Computation of reliable and precise long term estimates from the available short term hydrologic records is often a challenging task for the design engineers/hydrologists.
Peak flow magnitudes relating to ‘sixteen’ streams were utilised for deriving long term estimates using Maximum likelihood method (Gumbel 1941 and Panchang et al. 1962). The results revealed that most of peak flows were underestimated and their departures from the respective prototype magnitudes were of the order of 10 to 40 per cent.
These peak flow magnitudes were graduated by the Pearson type III distribution with the result that a perfect calibration (departure within 1 per cent only) was achieved for three streams. For remaining streams the departures were reduced and were within 20 per cent.
The precision of the estimates, deduced from the use of the Pearson type III distribution, was ascertained by evaluating the confidence intervals for these estimates. For three cases the confidence intervals (deduced from the latter distribution) were smaller than their counterparts deduced from the Gumbel's distribution. For the remaining streams' data, the confidence intervals were nearly double those obtained with the Gumbel's distribution. Thus to achieve conformity between prototype and model (using the Pearson type III distribution), one has to be content with even slightly less precise estimates, but which are, otherwise, realistic.