CE 143: HYDROLOGY

CHAPTER 2: PRECIPITATION (Statistics of Rainfall Data)

Learning Outcomes:
1. Identify the climate in the Philippines and its components.
2. Define some terminologies in meteorology.
3. List existing hydrometeorological observing networks in the Philippines and their
uses.
4. Differentiate different areal precipitation methods and calculate sample problems.
5. Apply the knowledge in statistics in calculating design and extreme rainfall.

2.4 Statistics of Rainfall Data

• Rainfall measurements are seldom used directly in engineering design, but rather
the statistics of rainfall measurements are typically used.
• Rainfall statistics are most commonly presented in the form of intensity-duration-
frequency (IDF) curves, which express the relationship between the average intensity
in a rainstorm and the averaging time (duration), with the average intensity having a
given probability of occurrence.

IDF
The data required to calculate the IDF curve are a record of rainfall measurements in
the form of the depth of rainfall during fixed intervals of time, t, typically on the order of 5
minutes. For a rainfall record containing several years of data, the following computations
lead to the IDF curve:
Step 1. For a given duration of time (= averaging period), starting with t, determine
the annual maximum rainfall (AMR) for this duration in each year.
Step 2. The AMR values, one for each year, are rank-ordered, and the return period,
T, for each AMR value, is estimated using the Weibull formula, where n is the number of years
of data and m is the rank of the data corresponding to the event with return period T.
𝑛+1
𝑇=
𝑚
Step 3. Steps 1 and 2 are repeated, with the duration increased by t. An upper-limit
duration of interest needs to be specified, and for urban-drainage applications, the upper-
limit duration of interest is typically on the order of 1–2 h.

Prepared by: JG Solmerin, RCE

 Step 4. For each return period, T, the AMR versus duration can be plotted, and this
relationship is called the depth-duration-frequency curve. Dividing each AMR value by the
corresponding duration yields the average intensity, which is plotted versus the duration, for
each return period, to yield the IDF curve.

Sample Problems:
1. How many years of rainfall data are required to derive the IDF curve for a return
period of 10 years?
Answer: A minimum of 9 years of rainfall data are required.

2. A rainfall record contains 32 years of rainfall measurements at 5-minute intervals. The

annual maximum rainfall amounts for intervals of 5 min, 10 min, 15 min, 20 min, 25
min, and 30 min have been calculated and ranked. The top three annual maximum
rainfall amounts, in millimeters, for each time increment are given in the table below.
Calculate the IDF Curve for a return period of 20 years.

Return ∆𝑡 𝑖𝑛 𝑚𝑖𝑛𝑢𝑡𝑒𝑠
Period, T
5 10 15 20 25 30
(years)
33 12.1 18.5 24.2 28.3 29.5 31.5
16.5 11.0 17.9 22.1 26.0 28.4 30.2
11 10.7 17.5 21.9 25.2 27.6 29.9
Answer: IDF Curve for a return period of 20 years.

160

140

120
Intensity (mm/h)

100

80

60

40

20

0
5 10 15 20 25 30
Duration (minutes)

Prepared by: JG Solmerin, RCE

Partial-Duration Series
The previous example has illustrated the derivation of IDF curves from n annual
maxima of rainfall measurements, where the series of annual maxima is called the annual
series. As an alternative to using an n-year annual series to derive IDF curves, a partial-
duration series is sometimes used in which the largest n rainfall amounts in an n-year record
are selected for each duration, regardless of the year in which the rainfall amounts occur.
The return period, T, assigned to each rainfall amount is calculated using:
𝑛+1
𝑇=
𝑚
Table 1. Factors for Converting Partial-Duration Series to Annual Series
Return Period (years) Factor
2 0.88
5 0.96
10 0.99
25 1.00
>25 1.00
Source: Frederick, et. al. (1977)

• Analysis of the annual-maximum series produces estimates of the average period

between years when a particular value is exceeded, while analysis of the partial-
duration series produces estimates of the average period between values of a
particular magnitude.
• The frequency distribution of rainfall amounts derived using the between the return
period of an event derived from an annual series, Ta, and the return period of that
same event derived from a partial-duration series, Tp, can be estimated by (Chow,
1964)
1
𝑇𝑝 = −
ln (1 − 𝑇𝑎−1 )
• In general, Tp < Ta, with the difference between Tp and Ta decreasing as Ta increases.
A rainfall amount with a return period derived from a partial-duration series can be
converted to a corresponding rainfall amount for an annual series by using the
empirical factors given in Table 1.
• These factors are applicable for return periods greater than 2 years.
• Partial duration and annual series are approximately the same at the larger recurrence
intervals, but for smaller recurrence intervals the partial-duration series will normally

Prepared by: JG Solmerin, RCE

 indicate events of greater magnitude. For rainfall records shorter than 20-25 years it
is recommended that return periods be estimated using partial-duration series (NRC,
2009).
Sample Problem:
A 10-year rainfall record measures rainfall increments at 5-minute intervals. The top
six rainfall increments derived from the partial duration series, estimate the frequency
distribution of the annual maxima with return periods greater than 2 years.

Rank 1 2 3 4 5 6
5-min rainfall (mm) 22.1 21.9 21.4 20.7 20.3 19.8

DESIGN RAINFALL

• A hypothetical rainfall event corresponding to a specified return period is usually the

basis for the design of stormwater management systems. In contrast to the single-
event design storm approach, a continuous-simulation approach is sometimes used
where a historical rainfall record is used as input to a rainfall-runoff model; the
resulting runoff is analyzed to determine the peak runoff rate or hydrograph
corresponding to a given return period.
• The design-storm approach is more widely used in engineering practice than the
continuous simulation approach.
• Design storms can either be synthetic or actual (historic) design storms, with synthetic
storms defined from historical statistics.
o Return Period
▪ The return period of a design rainfall should be selected on the basis of
economic efficiency.
▪ Typical return periods are as follows according to ASCE (2006):
• 2-15 years, with 10 years most common, for storm sewers in
residential areas.
• 10-100 years, for storm sewer in commercial and high-value
districts.
▪ In selecting the return period for a particular project, local drainage
regulations should be followed if they exist.
o Rainfall Duration
▪ The duration of a design storm is usually selected based on the time
response characteristics of the catchment.
▪ The time response of a catchment is measured by the travel time of
surface runoff from the most remote point of the catchment to the
catchment outlet and is called the time of concentration.

Prepared by: JG Solmerin, RCE

 ▪On small urban catchments with areas less than 40 ha (100 ac), current
practice is to select the duration of the design storm as equal to the time
of concentration.
▪ This approach usually leads to the maximum peak runoff rate for a given
return period.
▪ For the design of detention basins, however, the duration requiring the
largest detention volume is most critical, and several different storm
durations may need to be tried to identify the most critical design-storm
duration.
o Rainfall Depth
▪ The design-rainfall depth for a selected return period and duration is
obtained directly from the intensity-duration-frequency (IDF) curve of
the catchment.
▪ The IDF curve can be obtained from regulatory manuals that govern
local drainage designs, estimated from local rainfall measurements, or
(in the United States) derived from National Weather Service (NWS)
publications such as TP-40 (Hershfield, 1961).
▪ Realistic temporal distributions of rainfall within design storms are best
determined from historical rainfall measurements. In many cases,
however, either the data is not available or such a detailed analysis
cannot be justified. Under these conditions, the designer must resort to
empirical distributions. Frequently used methods for estimating the
rainfall distribution in storms are the triangular method, alternating-
block method, and the NRCS 24-h hyetograph.

Reference/s:
Chow, V. T., Maidment, D. R., & Mays, L. W. (1988). Applied Hydrology. United States:
McGraw-Hill Book Company.
*Excerpts are taken from Instructional Materials of Engr. Karen Joyce Santiago

Prepared by: JG Solmerin, RCE