Estimation of Missing Data

Double-Mass Analysis
Mean Areal Precipitation
Depth-Area Duration
Analysis
Frequency Analysis
Estimation of Missing Data
 Some precipitation stations may have short breaks in the
records because of absence of the observer or because of
instrumental failures. It is often necessary to estimate this
missing record.
 In the procedure used by the U.S. Weather Bureau, the
missing precipitation of a station is estimated from the
observations of precipitation at some other stations as
close to and as evenly spaced around the station with the
missing record as possible.
 The station whose data is missing is called interpolation
station and gauging stations whose data are used to
calculate the missing station data are called index stations.
 Methods of Estimating
Missing Data

1. Simple Arithmetic Method

2. Normal Ratio Method
3. Quadrant /Inverse Distance Weighing
Method
4. Regression Method
1. Simple Arithmetic Method
In this method simultaneous rainfall records of three
close-by stations are made use of. The stations should
however be evenly spaced around the station with
missing records. A simple arithmetic average of the
rainfall of the three selected stations gives the estimate
of the missing value. This method can be used to
calculate monthly as well as annual missing rainfall
values. This method should be used only when normal
annual precipitation at each of the selected stations is
within 10% of that station for which records are missing.
According to the arithmetic mean method the missing
precipitation 'Px' is given as:
𝒊=𝒏
𝟏
𝑷𝒙 = ෍ 𝑷𝒊
𝒏
𝒊=𝟏
Where: 'n' is the number of nearby stations, 'Pi' is
precipitation at ith station and 'Px' is missing
precipitation.
In case of three stations 1, 2 and 3, Px = (P1 + P2 + P3)/3
Example:
Find out the missing storm precipitation data of station
X given in the following table.
Station 1 2 x 3 4

Storm Precipitation (in) 3.8 3.25 ? 4.6 3.15

Annual Precipitation (in) 39.5 43.1 36.8 49.5 46.2

𝟏
𝑷𝒙 = σ𝒊=𝒏
𝒊=𝟏 𝑷𝒊 = ¼(3.8 + 3.25 + 4.6 + 3.15) = 3.7 in
𝒏
Test the normal annual precipitation at station X.
10% of 36.8 = 3.68 in
±3.68 + 3.68 = 40.48 in and 33.12 in
Since annual precipitation of adjacent station does not
lie with 10% so arithmetic mean method cannot be used
and instead normal ratio method will be used for better
accuracy.
Example:
Rain gauge X was out of operation for a month during
which there was a storm. The total rainfall at three
adjacent stations A, B and C are 37 mm, 42 mm and 49
mm, respectively. The average annual rainfall for all the
stations are X = 694 mm, A = 726 mm, B = 752 mm and
C = 760 mm. Using the arithmetic method, estimate the
missing total rainfall for gauge X.
Stations Precipitation (mm) Annual Precipitation (mm)
A 37 726
B 42 752
C 49 760
X ? 694
 𝟏
𝑷𝒙 = σ𝒊=𝒏
𝒊=𝟏 𝑷𝒊 = 1/3(PA + PB + PC) = 1/3(37 + 42 + 49)
𝒏
PX = 42.7 mm

Test the normal annual precipitation at station X.

10% of 694 = 69.4 mm
±69.4 + 694 = 624.6 mm and 763.4 mm
Since the annual precipitation of the stations are within
the given range, the precipitation at station is 42.7 mm.
2. Normal Ratio Method
Normal ratio method (NRM) is used when the normal
annual precipitation at any of the index station differs
from that of the interpolation station by more than 10%.
In this method, the precipitation amounts at the index
stations are weighted by the ratios of their normal
annual precipitation data in a relationship of the form.
According to the normal ratio method the missing
precipitation is given as:
𝒊=𝒏
𝟏 𝑵𝒙
𝑷𝒙 = ෍ 𝑷𝒊
𝒏 𝑵𝒊
𝒊=𝟏
Where: Px is the missing precipitation for any
storm at the interpolation station 'x', Pi is the
precipitation for the same period for the same
storm at the "ith" station of a group of index
stations, Nx the normal annual precipitation value
for the 'x' station and Ni the normal annual
precipitation value for 'ith' station.
Example:
Precipitation station X was inoperative for part of a
month during which a storm occurred. The respective
storm totals at three surrounding stations A, B and C
were 98, 80 and 110 mm. The normal annual
precipitation amounts at station X, A, B and C are,
respectively, 800, 1008, 842 and 1080 mm. Estimate the
storm precipitation for station X.
Stations Precipitation (mm) Annual Precipitation (mm)
A 98 1008
B 80 842
C 110 1080
X ? 800
 𝒊=𝒏
𝟏 𝑵𝒙
𝑷𝒙 = ෍ 𝑷𝒊
𝒏 𝑵𝒊
𝒊=𝟏

PX = 800/3(PA/NA + PB/NB + PC/NC)

PX = 800/3(98/1008 + 80/842 + 110/1080)
PX = 78.42 mm
3. Quadrant/Inverse Distance
Weighing Method
Establish a set of axes running through X and determine
the absolute coordinates of the nearest surrounding
points P, Q, R, S, T and U (choose one station in each
quadrant).
The estimated precipitation at X is determined as a
weighted average of the other six points. The weights are
reciprocals of the sums of the squares of distance X and
Y; that is:
σ𝒏
𝒊=𝟏 𝑷𝒊 𝑾𝒊 𝟏 𝟏
𝑷𝒙 = σ𝒏
Where: 𝑾𝒊 = =
𝒊=𝟏 𝑾𝒊 𝑫𝒊 𝟐 𝑿𝒊 𝟐 + 𝒀𝒊 𝟐
Where: Pi is the rainfall of the surrounding
stations, (Xi, Yi) is the coordinate of each
surrounding station, (0,0) is the coordinate of the
station with missing data.
Note: This method considers the distance of the
stations.
Example:
Stations A, B, C, D, E, F and G are rain gauge stations.
During a period of time, station A was out of operation.
During the same period, the total rainfall for other
stations were 40 m, 45 mm, 37.5 mm, 50 mm, 47.5 mm,
and 42.5 mm, respectively. Compute the missing total
rainfall of station A using quadrant method.
Quadr Precipitation, P Coordinate
Station Di2 Wi WiPi
ant (mm) X Y
B 40 4 2 20
I
G 42.5 2 1 5 0.2 8.5
II F 47.5 -5 2 29 0.0345 1.638
D 37.5 -3 -2 13 0.0769 2.885
III
E 50 -3 -3 18
IV C 45 1 -6 37 0.027 1.216
Total 0.338 14.239

σ𝒏
𝒊=𝟏 𝑷𝒊 𝑾𝒊
𝑷𝒙 = σ𝒏
= 14.239/0.338 = 42.127 mm
𝒊=𝟏 𝑾𝒊
Therefore, the rainfall at station A, PA = 42.127 mm.
4. Regression Method
Regression analysis is a procedure for fitting an equation to a
set of data. Specifically, given a set of measurements on two
random variables, y and x, regression provides a means for
finding the values of the coefficients a and b for the straight
line (y = a+ bx) that best fits the data.
The coefficients a and b can be found using least square
method using the following two equations simultaneously:
𝒏𝒂 + 𝒃 ෍ 𝒙 = ෍ 𝒚

𝒂 ෍ 𝒙 + 𝒃 ෍ 𝒙𝟐 = ෍ 𝒙𝒚
Example:

The rainfall data of the table below are the annual

rainfall (mm) and monthly rainfall (mm) of a
meteorological station from 1990-1995 E.C. Estimate the
monthly rainfall value of the station for the year 1996
E.C/ using the regression method.
 Data of Station

Year Annual RF (mm) Monthly RF (mm)

1990 1100 200

1991 1200 220

1992 1000 180

1993 950 170

1994 1150 210

1995 1250 230

1996 1300 x
 Develop an equation in the form of y = a + bx, that is
monthly RF = a + b*annual RF.
 The coefficients a and b can be found using the
equations
𝒏𝒂 + 𝒃 ෍ 𝒙 = ෍ 𝒚

𝒂 ෍ 𝒙 + 𝒃 ෍ 𝒙𝟐 = ෍ 𝒙𝒚
 Data of Station
Year Annual RF (mm) Monthly RF (mm) x2 xy

1990 1100 200 1210000 220000

1991 1200 220 1440000 264000

1992 1000 180 1000000 180000

1993 950 170 902500 161500

1994 1150 210 1322500 241500

1995 1250 230 1562500 287500

1996 1300 x

Sum 6650 1210 7437500 1354500

Therefore, the equations become:

6𝑎 + 6650𝑏 = 1210 → 1
6650𝑎 + 7437500𝑏 = 1354500 → (2)
Solving the two equations simultaneously:
a = -20 and b = 0.2 in which y = -20 + 0.2x or monthly
rainfall = -20 + 0.2*annual rainfall
To estimate the monthly rainfall for the year 1996,
substitute the annual rainfall of the year in to the
equation and is estimated as:
Monthly rainfall (mm) = -20 + 0.2(1300 )
= 240 mm
 Double Mass Analysis
 Double-mass-curve analysis is the method that is used
to check for an inconsistency in a gaged record.
 A double-mass curve is a graph of the cumulative catch
at the rain gage of interest versus the cumulative catch
of one or more gages in the regions that have been
subjected to similar hydro meteorological occurrences
and are known to be consistent
 If a double- mass curve has a constant slope, the
record is consistent.
 If a double –mass curve has not a constant slope, the
record is not consistent and need to be adjusted.
 Steps to check and adjust the consistency of rainfall
data of a station X:
I. The doubtful station, say A, is marked and the group
of stations surrounding it are identified.
II. Determine the cumulative rainfall of the station A
III. Sum the data of the neighbouring stations and
determine the cumulative
IV. Plot the sum cumulative rainfall of stations with
cumulative station A to determine the double mass
curve.
V. If there is a changing slope in the double-mass
curve, adjust it correct the data of station A by the
adjustment factor
 Adjustment of the Upper Section of a Double-
Mass Curve

∆𝒀𝒊
𝑺𝒊 =
∆𝑿𝒊

Say the slopes of the two sections, S1, and S2, can be
computed from the cumulative catches:
Example:
Data for gages A, B, and C and D are given the table
below. Check the consistence of data for gage D and
adjust it using double-mass curve method.
Annual Catch (in.) at Gauge
Year A B C D
1977 30 31 27 29
1978 33 36 32 32
1979 26 27 24 28
1980 27 26 27 29
1981 34 34 30 30
1982 31 33 31 29
1983 28 30 24 28
1984 35 34 33 39
1985 37 39 36 41
1986 34 35 35 37
 Cumulative Catch (in.) for
Annual Catch (in.) at Gauge
TOTAL (A + B + C) Gage
Year A B C D A+B+C D d2
1977 30 31 27 29 88 88 29
1978 33 36 32 32 101 189 61
1979 26 27 24 28 77 266 89
1980 27 26 27 29 80 346 118
1981 34 34 30 30 98 444 148
1982 31 33 31 29 95 539 177
1983 28 30 24 28 82 621 205
1984 35 34 33 39 102 723 244 35
1985 37 39 36 41 112 835 285 36.8
1986 34 35 35 37 104 939 322 33.2
The equation for adjusting the upper section of the curve is:
0.3301
𝑑2 = 𝐷2 = 0.8973𝐷2
0.3679
 Adjustment of the Lower Section of a Double-
Mass Curve

𝑺𝟐
𝒚𝟏 = 𝒀𝟏
𝑺𝟏

𝑆2/𝑆1 is the adjustment factor and multiplying each value

in the Y1 series by the adjustment factor.
Example:
In the table below annual rainfalls of gage E, F, G and H
are given. Gage H was permanently relocated after a
period of 3 years (at the end of 1981); thus adjust the
recorded values from 1979 through 1981 of gage H using
double mass curve assume the data for the other gages
are consistent.
Annual Catch (in.) at Gauge
Year E F G H
1979 22 26 23 28
1980 21 26 25 33
1981 27 31 28 31
1982 25 29 29 31
1983 19 22 23 24
1984 24 25 26 28
1985 17 19 20 22
1986 21 22 23 26
 Cumulative Catch (in.) for
Annual Catch (in.) at Gage
TOTAL (E + F + G) Gage
Year E F G H E+F+G H h1
1979 22 26 23 28 71 71 28 24.7
1980 21 26 25 33 72 143 61 29.1
1981 27 31 28 31 86 229 99 33.5
1982 25 29 29 31 83 312 130
1983 19 22 23 24 64 376 154
1984 24 25 26 28 75 451 182
1985 17 19 20 22 56 507 204
1986 21 22 23 26 66 573 230

The adjusted values from 1979 through 1981 can be computed

using:
0.3803
ℎ1 = 𝐻1 = 0.8809𝐻1
0.4323