|
Chapter
4 will cover the subject of errors that occur during pilot balloon
observations. We will review the importance of knowing the types
of errors, then go through observational and processing errors,
then about how to detect the errors using CORRIGE and then how
to correct specific errors. Finally we will go through some examples
that may help you understand the process of correcting errors
with CORRIGE.
4.1
Importance of knowing all sorts of possible errors.
The
process of gathering research data is a long one, unfortunately,
and it is exposed to different sources of error. Whenever the
dataset is finally archived in a computer the quality may have
been reduced by a large sum of errors of different kinds. To make
the best out of the dataset you will need to clean it by going
through quality control procedures. To do this it will be therefore
important to know all sources of error, understand which ones
can be corrected, learn how to correct them, and furthermore,
correct them.
4.2
Observational errors
This
is the broadest group of possible errors that may take place during
the process of gathering pilot balloon data. They occur during
the process of observation. Observational errors tend to be related
mainly to an bad reading mistaken by the observer, but also due
to other causes such as a lost of adjustment in the optical theodolite,
a lack of precision, and mistaken setting of the equipment. This
section of the tutorial will discuss the most common types of
observational errors starting with the most frequent ones. Even
though some examples will be included, most of the examples will
be presented in section 4.6 of this tutorial.
4.2.1. Angle read off by 1°
This
is probably the most frequent error found in pilot balloon observations.
It occurs when the observer reads an angle higher or lower than
the correct one in 1°. It will tend to happen particularly when
the decimals lay between ~.8 and .0, especially if the calibrated
drum is not well aligned (it should read exactly zero when the
mark on the azimuth scale corresponds exactly with a degree graduation
on this scale). The observer may get confused when reading the
decimals, and commit a 1 degree error. This error will be obvious
within CORRIGE and can be easily corrected. An example can be
observed in figure 15.
Figure 15. Angle read off by 1°.
4.2.2. Scale read in reversed order
This
error is also common among observers with little experience and
it is produced when the observer gets confused with the direction
of the scales. This error typically produces readings, which are
off by 2°, 4°, 6°, 8°, etc. Sometimes the error can be as high
as 18°, but this is does not occur very often.
The
confusion occurs because of two facts. The first one is that the
elevation angle scale increases towards the right (as any conventional
scale in the western hemisphere), whereas the azimuth angle scale
increases towards the left, this is, rolls in inverse order. A
second fact is that sometimes when the observer reads the angle
scale, he/she may find many small degree marks surrounding a simple
numerical value, which makes it harder to read than other situations
when two numerical values are visible. This may confuse him/her,
and since the readings have to be carried out quickly, he/she
might unwillingly reverse the scale. Figure 16 illustrates a sounding
with a 10 degree error, which may have been caused by this problem.
Some illustrative examples include reading 41 instead of 39, 231.5
instead of 228.5 or 347 instead of 353. If you notice, the correct
value is a mirror of the error with respect to the closest product
of 10. Analyzing the same examples respectively, the closest product
of 10° will be 40° for the first case, 230° for the second case
and 350° for the last one. See figure 17.
4.2.3. Balloon crossing through the center of the sky
This
error can be common in some locations where persistent wind circulations
may cause the balloon to pass near the center of the sky, this
is, right above the theodolite. During the observation the proper
thing to do is to rotate the theodolite about 180° in order to
avoid the elevation angle to become larger than 90°, which would
be unrealistic. Sometimes when this situation occurs, however,
the observer does not have enough time to rotate the theodolite
180° and he/she may remain following the balloon with the elevation
angle higher than 90°. Figures can be found in the example section.
CORRIGE
does not accept elevation angles larger than 90°. Any values larger
than 90 degrees will be converted by the program to 90.0. If this
is the case the program will not plot the sounding correctly.
CORRIGE can not help with this type of error and one way to correct
it involves the use of a spread sheet or any other program as
you need to modify both the elevation and the azimuth angles using
formulas.
The
procedure to correct this error follows. First you need to depict
all the sector of the sounding that has elevation angles equal
or larger than 90°. Then you must apply the following equation
to the elevation angles:
Final
elevation = 180° - Actual elevation
And
the following equation to the corresponding azimuth angles:
New
azimuth = 180° - Actual azimuth
If
the value of any new azimuth angle equals or exceeds 360° you
must apply only to those azimuth angles the following relationship:
Final
azimuth = Actual Azimuth - 360°
When
the file is corrected you should then run CORRIGE in order to
correct other possible errors.
4.3
Processing errors.
Processing
errors occur when dictating or typing the data. They are less
common than the observational errors and all of them can be corrected
using CORRIGE, unlike the observational errors, which in some
cases will not be correctable using CORRIGE.
4.3.1. Ten degree errors
This
is a fairly common error, but it is easy to correct within CORRIGE.
Mappens more frequently as a processing error but it can also
occur as an observation error. It basically happens when the person
typing or digitizing the data could not hear or read well or read
well the tens of the angle. The same applies to one, two or even
three hundred degree errors.
4.4 Finding errors with CORRIGE.
Input
files you process with CORRIGE for the first time will most probably
contain both observing and processing errors. There are two ways
to start looking for errors with CORRIGE. One of them is to start
the detection using the angle curves and the other is start looking
at the wind profiles. It is easier to start with the angles and
then get some help from the wind profiles for reasons we will
describe next. We will also discuss in this section about some
criteria proven useful to correct the data.
4.4.1.
Detecting errors with the angle profiles.
The
easiest way to start depicting errors is using the elevation and
azimuth curves. We recommend, however, start using the wind profiles
instead as there are some criteria developed for this and the
results cab be more reliable. Most of the angle curves tend to
be smooth, which eases the detection of errors. Figure 16 is a
clear example of this. There is an obvious error in about minute
7 (see arrow), when the azimuth angle profile jumps from a value
of about 140° to an angle of about 150° and back to 140°. This
seems to be a 10° error. This type of error is very easy to recognize
even without zooming on it. In most other cases zooming in on
the error becomes necessary.
Figure
16. A 10-degree error example.
Many
errors are smaller in magnitude and the zoom-in feature will be
very useful to depict them. It is the case of the one degree errors
which are relatively common. Sometimes, though, more than one
one-degree error may occur very close to each other and complicate
things a little. If this is the case, the recommendation is to
use the wind profile as a guidance to decide which observations
are erroneous and which ones are correct. A good sounding will
produce a wind profile with smooth changes in wind direction and
speed. Also, values of vertical shear of ~50kt per km or higher
should be looked into as possible errors. Only in regions where
jets or other wind-related phenomena occur, sharp changes in the
wind profile may happen. If this is the case, though, the new
wind speed should be consistent in at least 3 observations. Abrupt
changes that occur in one observation and disappear in the next
one can be taken as errors.
Some
legitimate situations in which the profile may appear too sharp
are the following. One common situation is when the balloon crosses
the north, this is, when the azimuth angle changes from 359° to
0°. The only way for CORRIGE to graph this change is to generate
a sharp jump in the azimuth angle that at first seems to be an
error. With practice you will find out that this situation is
not hard to solve. A 360° jump on the screen will immediately
tell you that the jump is justified as long as the rest of the
sounding is consistent with the jump. Figure 17 illustrates a
situation when this occurred. In this example the jump is a valid
situation. Just visualize the first part of the sounding being
placed over the 360° mark and the sounding will become smooth.
Figure
17. The balloon crossing the north is reflected as a jump in CORRIGE's
azimuth angle profile.
Another realistic situation may cause the angle profiles to look
sharp is when the balloon crosses close to the center of the sky.
When this happens the elevation angle reaches a maximum, and the
azimuth changes very quickly sometimes veering or backing almost
180°. This will not generate jumps as sharp as in the prior case
but may generate large and continuous changes in the azimuth angle
accompanied by a maximum in the elevation angle that must coincide
with the sharpest change in the azimuth. The characteristic signature
of this case is that the elevation angle must be high, generally
higher than 70 degrees.
A
final situation that may be misleading in CORRIGE is when there
are sharp wind direction and/or wind speed. These are frequently
associated to low-level jets, strong diurnal breezes opposed to
the synoptic flow and other particular situations as was mentioned
before. The rule of thumb is to check if the change is consistent
along the sounding, this is, if the wind speed is consistently
higher (or lower) in at least three consecutive levels (readings).
After
gathering some practice you will be able to notice when abrupt
changes correspond to an error and when not.
4.4.2.
Using the wind profile to detect the errors.
Once
in a while you may be fooled by the angle profiles. This can happen
for example when an error occurs near a change in the slope of
the curve or when the elevation angle jumps from a value close
to 360 to a value close to zero or viceversa (i.e. the pilot balloon
crosses through the north). While the user may be thinking that
the jump and the nearby observations may be correct, sometimes
there may be hidden errors that are only visible in the wind profile
screen. In other cases both the elevation angle profile and the
azimuth angle profile may appear smooth enough as to be correct,
however some errors may be present that may be more visible when
analyzing the wind profile. Even if you find yourself more comfortable
by starting with the angle profiles, you should always go through
the wind profile as it is the final output, and may show features
that the angle profiles will not.
In
some situations it may be easier to identify an error through
the wind profile screen than through the angle screens. Some useful
criteria to identify errors in the wind profile have arisen from
knowledge in meteorology. Developing sharp deterministic criteria
for the identification of errors, though, is a hard task. Even
when applying established criteria, the correction of errors should
always be performed carefully. There are cases in which the errors
are so small that they become hard to depict, particularly when
the scale of the natural variability of the atmospheric flow has
an order close to the order of magnitude of the errors.
The
criteria used to detect errors in the PACS-SONET pilot balloon
data were the following. An error is assumed to be present when
either the wind component "u" or "v" changes at a rate of 50 m/s
in 1 km, or more. Another criteria more applicable to CORRIGE
is to check for possible errors whenever there is a change of
5 knots in the wind speed between one level and the next. An error
will almost always exist when the change of speed between two
consecutive levels is in the order of 10 knots. It is recommended
to use some guidance from the wind direction as well. Strange
changes in the wind direction may be related to errors.
Figure
18. Evident errors in the wind profile.
Figure 18 shows some evident errors on the wind profile. In this
case the change is of 9 kt in 217 meters, then 15 kt in the next
216 meters, then 21 kt in the next 217-meter layer and finally
14 kt from 7008 to 7225 m ASL. There are many indicators here
that tell you that an error is present. Firstly, there are too
many abrupt changes in a very small portion of the sounding. In
fact, there is an abrupt change in every level through four consecutive
levels. Secondly, the magnitude of the changes is in the order
of 10 knots and sometimes even 20 knots. The last indicator is
that there is also a sharp change in the wind direction. In this
case there are probably two errors in the angle data since one
error in the angles creates two bad data points in the wind profile.
This is due to the particular design of the program and also due
to the fact that each angle value is used twice in determining
the horizontal displacement of the balloon.
Other
criteria that should be considered is that large changes in wind
direction can occur, but only when the wind is very weak, this
is, close to calm. If you see that the wind direction in your
profile is changing significantly in a certain layer and that
calls your attention, the first thing you should check is the
wind speed. If you find that winds are weaker than 3 knots your
data may be correct. If your wind speed, though, is of about 5
kt or larger, you should definitely seek for errors in the angle
profiles.
4.4.3.
Evident errors that cannot be corrected using CORRIGE
.
There are some evident errors that cannot be corrected using CORRIGE. Some of them are listed as follows .
1. When the elevation angle is larger than 90 degrees. This errors shows up as a flat section in the angle curve in color yellow. This error, described priorly in this document, occurs when the balloon crosses the center of the sky and does not give enough time to the observer to turn the theodolite 180 degrees. The way to correct this error is by using a spread sheet or any other program that can apply the equations described in section 4.2.3 .
Elevation angles larger than 90 degrees can also occur when the elevation angle column is flipped with the azimuth angle column. Before continuing we should flip the order of the columns using a spread sheet or a text editor .
2. When commas are used instead of dots, the angle curves take the shape of stairs. This happens because CORRIGE ignores the decimals and all the values are assumed to be integers. The solution is to change the commas for dots with a spread sheet or text editor before continuing with CORRIGE .
3. Another error occurs when the azimuth angle is larger than 360 degrees. CORRIGE does nor accept values equal or larger than 360 degrees. If this occurs, the program obtains a fake value by interpolation of neighboring values. The new value will appear as a yellow dot. This error can be corrected using CORRIGE, but it is recommended to be corrected in the DAT file.
4. In some occasions error in the time column might occur. In some soundings a value of 05 was digitized instead of a value of 0.5. Even though this error is very evident in CORRIGE it can not be corrected with the program . Few people detect this error and if they do they do not know how to correct it .
4.4.4. Subtle errors .
This section discusses the errors that are the hardest to detect. These errors can not be corrected using CORRIGE but what is worse is that they can not be detected using CORRIGE either. The most common errors are described as follows .
1.
The most common error occurs when the heading has less than two lines. In this case CORRIGE does not read the first observation or the first two observations. In this case the time in the WIN file starts in minute 1.0 or 1.5
.
2.
Another frequent error occurs when the time is digitized in sexagesimal format (mm:ss). This error can be easily identified by carefully observing the sounding section, not the zoom section. Most of the times this error will be almost undetectable. When this happens CORRIGE ignores the seconds and the WIN file results in values of 0.0, 1.0, 1.0, 2.0, 2.0, etc in the time column
.
4.5. Examples.
In
this section we will present a short and very diverse set of examples
about how to correct some errors. Please notice that we will discuss
the correction of specific errors instead of going though the
correction of the entire profile.
4.5.1.
Obvious errors.
Obvious
errors can be considered to be the several types of degree errors
such as 1, 10, 100, 200 and 300 degree errors. These are easy
to detect with CORRIGE and also easy to correct. The most common
way to proceed is to detect obvious errors in the azimuth angle
curve and correct them. The same can be done using the elevation
angle curve. At a first glance errors are not obvious so we use
zoom-in window to find them. We notice the presence of an error
located near observation #7 (figure 19). It is a 1-degree error.
Small degree errors may not be obvious unless you use the zoom-in
window to detect them. As we are dealing with an obvious error
we can correct it and proceed to revise the elevation angle curve.
To correct them we only need to subtract 1 degree using the "down
arrow" key. Figure 20 shows the corrected angle.
Figure 19. 1-degree error as displayed by the zoom-in window.
Figure
20. The error in the azimuth angle profile is corrected using
the down arrow.
After
correcting all the obvious errors in the azimuth angle curve we
proceed to the elevation angle curve. We find some obvious errors
in observations 7 and 8 as shown in the zoom window of figure
21. Observation #7 needs to be added 1 degree and observation
#8 needs to be subtracted 1 degree.
Figure
21. Errors in the elevation angle curves.
Figure 22 shows the elevation angle curve after the corrections.
Figure 23 shows the wind profile before the corrections and figure
24 after the corrections. When we complete the correction of all
the obvious angles we can start correcting the harder ones.
Figure
22. Now we smooth the errors in the elevation angle profile.
Figure
23. Wind profile. The cursor is placed in observation #7.
Figure
24. This time the corrections result on minor changes in the wind
profile.
In
this case the impact of the corrections have been small. The higher
the location of the errors the larger the impact they will have
on the sounding.
Example
2: One degree error.
Figure
25 shows a one degree error in the elevation wind profile. This
is another simple example. After loading the file we did not find
any suspicious observation on the azimuth angle profile, therefore,
we switched into the elevation angle profile. The way to correct
this error is to add 1 degree.
Figure 25. Elevation angle curve showing a 1-degree error.
Figure
26. Azimuth angle profile.
In
this case the azimuth angle curve (figure 26) shows no obvious
errors in the observation that corresponds to the error in the
elevation angle curve. Figure 27 shows the wind profiles before
and after the correction. The impact of this one single correction
is small but noticeable. There is a change in direction and wind
speed from a northwesterly wind at 4 knots to a west-northwesterly
at 7 knots.
Figure
27. Wind profile before correction (A) and after correction (B)..
Example
3: Two degree error.
Figure
28 shows a 11-degree error located at minute 23. It is an obvious
error and this one can be detected even without zooming-in on
it.
Figure
28. Azimuth angle screen.
Figure
29 shows the error as viewed in the zoom-in window using different
vertical resolutions. When large errors like this occur it may
be easier to use smaller vertical resolutions when zooming-in.
In this case the 2-degree per mark vertical resolutions better
displays the error.
Figure
29. Playing with the vertical scale of the zoom screen.
The impact of this error is large and can be easily detected in
the wind profile. Figure 30 shows the elevation angle curve in
the left panel and the wind profile curve in the right panel.
As you can see, there was no obvious error in the elevation angle
curve at the observation that corresponds to the error in the
azimuth angle curve. The large impact can be attributed to the
11-degree error in the azimuth angle profile.
Figure
30. Always revise elevation angle profile and the wind profile.
Example 4: 100-degree error.
This
error is pretty obvious. In this case the observer(s) was(were)
distracted and committed the same error in two consecutive observations.
It is still an obvious error because the jump was of about 100°,
a round number, and after two observations the curve jumped again
to follow the initial curve's tendency. The error is even more
obvious when analyzing the wind profile. Figure 31 includes the
azimuth angle curve, the elevation angle curve and the wind profile.
The elevation angle curve was displayed to show that the errors
in the wind profile were caused solely by the errors in the azimuth
angle profile.
Figure 31.
Effects of a 100° error.
Figure 32 shows the corrected errors in the azimuth angle curve
and the impact this has in the wind profile.
Figure
32. Correction of a 100° error.
Example
5: Correcting all the obvious errors in an entire profile.
This
profile has several obvious errors which can be easily corrected.
Figure 33 shows the obvious errors in the elevation angle curve
and in the azimuth angle curve and their corrections.
Figure
33. Correction of several obvious errors.
Figure 34 illustrates the wind profile with errors and figure
35 the same profile after the corrections.
Figure
34. Wind profile after initial correction.
Figure
35. Wind profile after all corrections.
4.5.5. Balloon crossing the north.
A
peculiar situation occurs when the balloon crosses the north,
as described previously in this document. This results in a large
jump from values just below 360 degrees to values just above 0
degrees. This jump may initially seem to be an obvious error but
it is not. When the jump takes place only once and the curves
are smooth in both sides of it, the jump will tend to be valid.
To
make sure the jump is valid it is recommended to revise the elevation
angle curve and the wind profile. Figure 33 shows an example of
a correct sounding. There is large jump in the azimuth angle but
both the elevation angle and the wind profile look coherent. Figure
34 also shows a good example of this situation.
Figure
35. Balloon crossing the north.
Figure
36. Balloon crossing the north, another example.
Example
2: Error close to valid jump.
Figure
38 shows an azimuth angle curve. In this case the balloon has
crossed the north as suggested by the large jump from near zero
to near 340 degrees. There is another jump, however, right after
the 340-degree observation. For people with not much practice
it may seem hard to detect where the error is, or how to correct
it.
Figure
38. What is wrong in this azimuth angle curve?
In this case
it is recommended to seek guidance in the wind profile. Before
doing this, though, we check the elevation angle curve for errors.
Figure 39 shows this curve.
Figure 39. Elevation angle seems fine.
In
this case the elevation angle curve appeared to be correct, so
any clear errors in the wind profile will be related to the error
in the azimuth angle. Figure 40 shows the wind profile with a
clear error near 3001 m ASL.
Figure 40. Initial wind profile.
If
you have experience with CORRIGE, you will find that there is
a 50-degree error in the observation right after the jump. If
you are not sure about the magnitude of the error, you can make
an initial correction and revise the results using the wind profile
until the winds are smooth in direction and speed. Figure 41 displays
the corrections done to this profile. The solution was to add
50 degrees into the 230 degree observation.
Figure
41. Profiles before and after the correction.
This
is a good example of a situation in which utilizing the wind profile
may greatly ease the depiction of errors. You should always revise
the wind profile before making any corrections as some sharp features
in the angle profiles may be correct.
|
