Precipitation Data Analysis

How to analyze rainfall data?

• In order to avoid erroneous conclusions it is important
to give the proper interpretation to rainfall data.
• Checking data homogeneity
• For statistical analysis rainfall data from a single series
should ideally possess property of homogeneity - i.e.
properties or characteristics of different portion of the
data series do not vary significantly.
• Rainfall data for multiple series at neighboring stations
should ideally possess spatial homogeneity
• Tests of homogeneity is required for validation
purposes and there is a shared need for such tests with
other climatic variables
• Secondary validation of rainfall data
• Spatial homogeneity testing
• Consistency tests using double mass curves
• Correcting and completing rainfall data
• Adjusting rainfall data for long-term systematic shifts –
double mass curves
• Basic statistics are widely required for validation and
reporting. The following are commonly used:
• arithmetic mean
• median - the median value of a ranked series Xi
• mode - the value of X which occurs with greatest
frequency or the middle value of the class with greatest
frequency
• standard deviation - the root mean squared deviation Sx :
• skewness and kurtosis
• Rain is well described in terms of a depth of
rainfall over a certain duration.
• Duration - period of time over which the rain
is measured e.g. one year - in the case of
annual rainfall; one month (for many climate
purposes); or so many days, hours or minutes
• Rainfall rate or intensity rate (in mm per hour
is calculated by dividing the depth by the
duration and is simply a measure of the
'heaviness' of the rainfall.
• To compare the severity of different rainfall
events, we might compare several rainfalls
measured over one hour. For instance, we
might be interested in the highest one-hour
rainfall during a calendar year at a certain
location. And going one step further, a civil
engineer designing a drain intended to fail (no
more frequently than) once every ten years
(on average) might ask the question, "What is
the rainfall depth over one hour exceeded, on
average, once in ten years?". This "once in ten
years" is a FREQUENCY.
• Another might ask about the 72-hour rainfall
exceeded on average once every 100 years. In
fact the number of possible combinations of
duration and frequency is many - take any one of
13 standard durations from 5 minutes to 72
hours, and any one of 7 frequencies from 1 per
year to 1 per 100 years, and you have 91
combinations. For this reason, the results are
best tabulated, or represented as a graph of
intensity versus duration for 7 different
frequencies.
• For convenience, frequencies (e.g. 1 per 100
years) are replaced by "Average Recurrence
Intervals" or "Return Periods" (e.g. 100 years).]
Intensity-Frequency-Duration analysis
• Also known as Frequency Analysis
• The main use of these data is in engineering
design of the following:
- design and risk assessment of dams and bridges,
- design of roof and stormwater drainage
systems,
- flood plain management,
- soil conservation studies,
- to express the "severity" of a single rainfall
event (in terms of its rarity).
 Estimating Missing Precipitation
Data
• Many precipitation stations have short breaks in their
records because of absences of the observer or
because of instrumental failures.
• It is often necessary to estimate this missing record.
• Precipitation amounts are estimated from observations
of three stations as close to and as evenly spaced
around the station with the missing record as possible.
• If the normal annual precipitation at any of the index
stations is within 10% of that for the station with the
missing record, a simple arithmetic average of the
precipitation at the index stations provides the
estimated amount.
• If the normal annual precipitation at any index
stations differs from that at the station in
question by more than 10%, the normal-ratio
method is used. The amounts at the index
stations are weighted by the ratios of the
normal-annual-precipitation values.

Where: Px= missing monthly precipitation at station x

PAx= normal annual precipitation for station x
PAi= normal annual precipitation for station i
Pi= monthly precipitation for station I
n = number of station
Example:
• There are 4 precipitation gauging stations in a
basin. The precipitation amounts for the months
of June, July, and August are missing from the
record for one gauging station in a basin. This
station belongs to a network of four in that basin.
For those three months, the other three stations
recorded the following:
Station 1 Station 2 Station 3

June 55 65 75

July 47 50 45

August 45 40 55
• Estimate the missing precipitation values if the
long-term annual average precipitation at the
four stations is:
Station 1 Station 2 Station 3 Station 4
June 60 65 70 67
July 50 55 65 60
August 45 47 60 55

Using normal-ratio method:

June: Px = ((67/60)55+(67/65)65+(67/70)75)/3 = 66.73
July: Px = ((60/50)47+(60/55)50+(60/65)45)/3 = 50.83
August: Px = ((55/45)45+(55/47)40+(55/60)55)/3 = 50.74
 Double-Mass Analysis
• Changes in gage location, exposure,
instrumentation, or observational procedure may
cause relative change in the precipitation catch.
The double-mass analysis tests the consistency of
the record at a station by comparing its
accumulated annual or seasonal precipitation
with the concurrent accumulated values of mean
precipitation for a group of surrounding stations.
• Accumulated values of mean precipitation are
plotted in a graph to assess its consistency.