Skip to main content

Preparing Gumbel probability paper

Short answer

Use ordinary arithmetic axes: vertical for discharge, horizontal uniformly scaled for reduced variate y_T = –ln[ln(T/(T–1))]. Mark T values at their y_T positions (e.g., T=5 at 1.50) to create a non-uniform T scale; Gumbel data plot as a straight line.

Prepare a Gumbel probability paper from an ordinary graph paper provided to you.

Gumbel probability paper is a plotting sheet on which data that follow Gumbel's distribution plot as a straight line. It can be prepared from ordinary arithmetic graph paper because Gumbel's equation xT=xˉ+Kσx_T=\bar x+K\sigma is linear in the reduced variate yTy_T.

Steps

  1. Choose the vertical axis as an ordinary arithmetic scale for the variable (flood discharge or rainfall depth), covering the range of data and predictions.
  2. For selected return periods compute the reduced variate
yT=−ln⁡[ln⁡TT−1]y_T=-\ln\left[\ln\frac{T}{T-1}\right]
T (years)251020501002005001000
yTy_T0.3671.5002.2502.9703.9024.6005.2966.2146.907
  1. Draw the horizontal axis with a uniform scale for yTy_T (for example 1 cm = 1 unit of yTy_T from −1 to 7).
  2. At the positions of the computed yTy_T values mark the return periods T (and the probability of exceedance P=1/TP=1/T) as labels. This gives the non-uniform return period scale along the top or bottom of the sheet.
  3. For plotting data: arrange in descending order, assign T=(N+1)/mT=(N+1)/m, and plot the flood against T (that is, against its yTy_T position).
  4. Draw the best-fit straight line and extend it to read the flood for any desired T.
 Q
  |                          o
  |                    o
  |              o
  |        o
  |  o
  +--+----+----+----+----+---- y_T
    0.37  1.50 2.25 3.90 4.60 6.21
    T=2   T=5  T=10 T=50 T=100 T=500

Two points are enough to fix the straight line, since for Gumbel's distribution the line is Q=xˉ+σ (yT−yˉn)/SnQ = \bar x + \sigma\,(y_T-\bar y_n)/S_n.

✓ Answer checked by Claude Sonnet 5.5

All Engineering Hydrology chapter 6 questions →