29.11.2014 Views

WATER & SOIL - These are not the droids you are looking for.

WATER & SOIL - These are not the droids you are looking for.

WATER & SOIL - These are not the droids you are looking for.

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

In <strong>the</strong> first test <strong>the</strong> GEV method was often affected in<br />

this manner, producing many curves that fitted <strong>the</strong> data<br />

very well but giving 100-year flood peak estimates that were<br />

<strong>not</strong> always sensible. For example, <strong>the</strong> trend in <strong>the</strong> lower<br />

half of <strong>the</strong> series would cause <strong>the</strong> GEV curve to flatten off<br />

at <strong>the</strong> top end, impþing that <strong>the</strong>re was a limit to flood<br />

peaks of about twice <strong>the</strong> mean annual flood. While <strong>the</strong>re<br />

may be an upper limit to flood magnitude, it is certainly<br />

more than <strong>the</strong> figure implied (see Tables 3.3 and 3.6).<br />

In comparison, <strong>the</strong> straight-line fits of <strong>the</strong> Gumbel and<br />

EVI methods were often a good approximation to <strong>the</strong> data<br />

and gave more sensible 100-year flood peak estimates.<br />

However, this better per<strong>for</strong>mance by <strong>the</strong> t$'o-pa¡ameter<br />

methods can be attributed to <strong>the</strong> small samples used<br />

(NERC 1975; pp. 159-60). The fact that <strong>the</strong> Gumbel<br />

method still per<strong>for</strong>med better than <strong>the</strong> GEV method in <strong>the</strong><br />

second test suggests that <strong>the</strong> samples in this test may also<br />

have been small. However, it is also likely that <strong>the</strong> leastsquÍues<br />

fitting technique of <strong>the</strong> Gumbel method influenced<br />

<strong>the</strong> evaluation test in <strong>the</strong> method's favour.<br />

The two-parameter log-Normal method also gave good<br />

approximations to <strong>the</strong> data at times, producing a reasonable<br />

proportion of good fits. However, <strong>the</strong> method assumes<br />

that <strong>the</strong>re is no skew in <strong>the</strong> logarithms of <strong>the</strong> series, an<br />

assumption which was r<strong>are</strong>ly true. Hence <strong>the</strong> method did<br />

<strong>not</strong> per<strong>for</strong>m as well as o<strong>the</strong>rs in <strong>the</strong> test, including its p<strong>are</strong>nt<br />

method, <strong>the</strong> three-parameter LP3.<br />

A.4.5 Gonclusions<br />

The evaluations in <strong>the</strong> two tests involved a good deal of<br />

subjective judgement, but this typifies <strong>the</strong> present situation,<br />

with <strong>the</strong> objective goodness-of-fit indices <strong>not</strong> providing rigorous<br />

enough criteria <strong>for</strong> discriminating between different<br />

frequency analysis methods.<br />

From <strong>the</strong> findings of <strong>the</strong> tests, <strong>the</strong> following conclusions<br />

were reached:<br />

(¡) <strong>the</strong> Jenkinson method was <strong>the</strong> superior method in both<br />

tests and should be used in flood frequency analysis,<br />

especially when <strong>the</strong> sample is small;<br />

(iÐ <strong>the</strong> Gumbel method improved in per<strong>for</strong>mance with increase<br />

in sample size, and should be a satisfactory alternative<br />

to <strong>the</strong> Jenkinson method on <strong>the</strong> larger sam-<br />

Ples;<br />

(ii¡) <strong>the</strong> GEV and LP3 methods were more flexible than <strong>the</strong><br />

two-parameter methods, with <strong>the</strong>ir frequency curves<br />

geneially following <strong>the</strong> trend in <strong>the</strong> data extremeiy<br />

well;<br />

(iv) on small samples, in particular, <strong>the</strong> straight-line fits<br />

from <strong>the</strong> two-parameter methods can give good approximations<br />

to <strong>the</strong> data and sometimes more sensible<br />

results than those obtained from <strong>the</strong>ir p<strong>are</strong>nt threeparameter<br />

methods;<br />

(v) on small samples <strong>the</strong> LP3 method appears to per<strong>for</strong>m<br />

better than <strong>the</strong> GEV method, being influenced less by<br />

<strong>the</strong> trend in <strong>the</strong> lower part of a series;<br />

(vi) on larger samples <strong>the</strong> GEV and LP3 methods can produce<br />

similar shaped curves and give a comparable per<strong>for</strong>mance.<br />

<strong>These</strong> conclusions apply <strong>for</strong> individual records and up to<br />

References<br />

Beard, L.R. 1974: Flood flow frequency techniques. Center<br />

tor Research in Water Rnources, Univercity of<br />

Texas, Austin, Technicøl Report CRVR-|19.<br />

1977: Guidelines <strong>for</strong> determining flood flow frequency.<br />

U.S. Water Resounaes Council, Bulletin No.<br />

I7A of <strong>the</strong> Hydrologlt Committee.<br />

Beard, L.R.; Frederick, A.J. 1975: Hydrologic frequency<br />

analysis. .In Hydrologic Engineering Methods <strong>for</strong><br />

Water Resources Development Vol. 3. The Hydrologic<br />

Engineering Center, U.S. Army Corps of Engineers,<br />

Davis, Cali<strong>for</strong>nia.<br />

Benson, M.A. 1968: Uni<strong>for</strong>m flood-frequency estimating<br />

methods <strong>for</strong> federal agencies. Water Resources Reæørch<br />

4(5):891-908.<br />

Bobée, B.B. Robitaille, R. 195: Correction of bias in <strong>the</strong><br />

estimation of <strong>the</strong> co-efficient of skewness. Wøter Resources<br />

Research I I (6): 841-4.<br />

1977: The use of <strong>the</strong> Pea¡son type 3 and Log-<br />

Pearson type 3 distributions revisited. lfoter Resoutces<br />

Research I 3(2): 427 -41.<br />

Chernoff, H.; Lieberman, G.J. 1954: Use of normal probability<br />

paper. Journøl of <strong>the</strong> Americøn Stotisticol<br />

Associotion 49: 778-85.<br />

1956: The use ofgeneralised probability paper <strong>for</strong><br />

continuous distributions. Annals of Ma<strong>the</strong>moticol<br />

Statistics 27:806-18.<br />

Chow. V.T. l95l: A general <strong>for</strong>mula <strong>for</strong> hydrologic frequency<br />

analysis. Tronsactions of <strong>the</strong> Americøn Geophysicol<br />

Union 32(2): 231-7.<br />

Foster, H.A. l9A: Theoretical frequency curves and <strong>the</strong>ir<br />

applications to engineering. Tronsactiotts of <strong>the</strong><br />

American Society ol Civil Engineen 872 142-73.<br />

Gumbel, E.J. l94l: The return period of flood flows. ánnols<br />

of Ma<strong>the</strong>motical Stotistics 12: 163-X).<br />

1954: Statistical <strong>the</strong>ory of extreme values and<br />

some practical applications. US Bureou of Slondards,<br />

Applied Ma<strong>the</strong>motics Series 33l. 15-16.<br />

l!Xó: Extreme value analysis of hydrologic data.<br />

In Statistical Methods of Hydrology. Proceedings of<br />

Hydrology Symposium No. 5, McGill University,<br />

Canada, Z,-25Februw. pp. 147-69.<br />

Harter, H.L. 1969: A new table of percentage points of <strong>the</strong><br />

Pearson Type III distribution. Technometrics Il(l):<br />

177-81.<br />

Hazen, A. l9l4: Storage to be provided in impounding reservoirs<br />

<strong>for</strong> municipal water supply. Transøctions of<br />

<strong>the</strong> American Society of Civil Engineers 78:<br />

1539-641.<br />

Irish, J.; Ashkanasy, N.M. 1977: Flood frequency analysis.<br />

I¿ Australian Rainfall and Runoff. Chapter 9' The<br />

Institution of Engineers, Australia.<br />

Jenkinson, A.F. 1955: The frequency distribution of <strong>the</strong><br />

annual maximum (or minimum) values of meteoro-.<br />

logical elements. Quorterly Journal of Royal Meteo'<br />

rological Society 87: 158-71.<br />

Statistics of extremes. Iz Estimation of<br />

Maximum Floods. Chapter 5. VMO Technicql Nole<br />

No.98. pp.193-227.<br />

-1969:<br />

Kite, G.W. 1976: Frequency and risk analyses in hydrology.<br />

Inland Waters Directorate, Water Resources Branch,<br />

Dept. of E<br />

Kopittke, R.A.;<br />

e, K.S. 1976: Fre"<br />

quency an<br />

in Queensland. Ia<br />

tion of Enginæn,<br />

Hydrology<br />

Austroliq, Notionsl Co4ference Publication' No.<br />

76/2. pp.2È4.<br />

Linsley, R.K.; Kohler, M.A.; Paulhus, L.H. 195: Hydrology<br />

<strong>for</strong> Engineers. McGraw-Hill, New York.<br />

Water & soil technical publication no. 20 (1982)<br />

9l

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!