Environmental Engineering Reference
In-Depth Information
3.4.7 Missing Data and Other Issues for Using Archived
Weather Data
It is not uncommon to have gaps in precipitation records due to malfunctioning
equipment, temporary closure of a station, and various other factors. Missing data
are often estimated using data from other precipitation gages, preferably located
close to the site. This estimation can be made using the weighted average method
(Eq.
3.6
), or using the normal-ratio method (Dingman
2002
:115) that estimates
missing values (
p
m
) from long-term average precipitation at the station with missing
record (
P
0
), the long-term average precipitation at each station (
P
i
), and the data for
the missing time interval (
p
i
) at other stations
N
X
N
1
P
0
P
i
p
i
p
m
¼
(3.7)
i¼
1
where
N
is the number of nearby stations with valid data. The long-term average
may be annual or monthly depending on the site climatic condition and the nature of
analysis (e.g., weekly or monthly water budget).
Archived precipitation data from other sources should to be examined for consis-
tency, particularly if the user does not have first-hand knowledge of the station history
and conditions. Changes in measurement method, gage location, or the surrounding
environment can induce artificial offsets or trends in the data (Dingman
2002
:117).
It is important to review the station history, if available, and also use a double-mass
curve technique to identify any suspicious data. A double-mass curve is a graph
showing cumulative monthly or annual precipitation from a reference station on the
horizontal axis and cumulative precipitation from the station of interest on the vertical
axis (Searcy and Hardison
1960
). The slope of a double-mass curve should be
constant if there has been no change in the station of interest. If there is a statistically
significant change in slope, the data from the station of interest can be multiplied by a
correction factor to compensate for the change. The reference station should have
consistent data and be located reasonably close to the station of interest.
In addition to checking the consistency, attention should be paid to gage
calibration and correction procedures. For example, some weather stations operated
by governmental and municipal agencies may not apply corrections for gage-catch
deficiency, resulting in a negative bias in measured precipitation, snowfall in
particular. In recent years, gridded precipitation data have become available from
government agencies such as the U.S. National Oceanic and Atmospheric Admin-
istration (NOAA). These data are generated over a large region (e.g., North
America) typically by interpolating observation data and/or refining numerical
weather model outputs. While these data sets offer convenient means to estimate
precipitation for a given region, the data are not intended as a surrogate for local
precipitation measurements. Therefore, gridded precipitation data should be
validated using observational data from local stations.
