The Des Moines Sky Inits

jeanscricketInternet και Εφαρμογές Web

3 Νοε 2013 (πριν από 3 χρόνια και 7 μήνες)

89 εμφανίσεις

The Des Moines Sky Inits


Chuck Myers

Senior Forecaster

WFO DMX

May 2003

Updated
March
200
5


The
Des Moines Sky Inits
developed for the GFE Sky smart initialization are based on the work
of Walcek, C. J., 1994: Cloud Cover and Its Relationship to Relative
Humidity during a
Springtime Midlatitude Cyclone.
Monthly Weather Review
,
122
, 1021
-
1035.


Historical Note…
This work has been incorporated into all the standard Sky smart initializations
in GFE by the Forecast Systems Laboratory. However, the ability to e
asily tweak the
computations is not available. The
Des Moines Sky Inits

provide that versatility.



From the abstract:


Vertical distributions of fractional cloud coverage derived from the U.S. Air Force 3DNEPH
satellite, aircraft, and surface
-
based analys
is are compared with related standard meteorological
observations over the eastern United States.


During the period analyzed, cloud cover maximizes near 900 mb at 35% cloud cover and
decreases to near
-
zero cover at the surface. Above 900 mb, fractional cl
oudiness gradually
decreases to 10%
-
20% cover at 200mb. Cloud cover is positively correlated with relative
humidity and large
-
scale vertical velocity, and negatively correlated with wind shear and
temperature lapse rate, except in the lowest 100 mb, where
cloud cover is weakly correlated with
relative humidity, vertical velocity, wind shear, and temperature lapse rate. Mean fractional cloud
coverage observed at various humidities and pressures is derived from these observations, and
resolution
-
dependent alg
orithms for estimating cloud coverage from relative humidity are
suggested.


Using relative humidity alone as an indicator of cloud coverage, cloud amount can be assessed
only to within a root
-
mean
-
square difference of 15%
-
30% from the 3DNEPH cloud cover,
depending on the resolution at which calculations are performed.


Selected excerpts:


At a particular relative humidity, cloud amounts are greatest in the 800
-
500 mb layer of the
troposphere...


The highest cloud amounts occur under high humidities at 800
-
700 mb...

...10%
-
20% cloud coverage occurs at humidities as low as 20%...


...fractional area of cloud coverage decreases exponentially as the relative humidity falls below
100%, and that there is no clear critical relative humidity

where cloud coverage is

always zero.
Also, these trends suggest that at 100% humidity, cloud cover can be significantly lower than
100%.


Equation (1) is an approximation for cloud amount
f

as a function of relative humidity RH

(RH < 1):


f
(%) = min [
f
100

exp (RH
-
1 / 1
-
RH
e
)

, 100] ,



(1)


where
f
100

is the cloud cover extrapolated to or at 100% relative humidity, and

RH
e

(
e
-
folding
relative humidity) is qualitatively similar to the critical humidity

used in previous cloud cover
formulations, although here it represents
the relative humidity depression below 100% where
cloud amount decreases to 37% (
e
-
1
) of its value at 100% humidity.


Figure 9 shows the best values for
f
100

(Fig. 9a) and (1
-
RH
e
) (Fig. 9b), along with the
corresponding root
-
mean
-
square difference between

calculated and observed cloudiness.


Pressure level parameters and equations


From the optimized linear solutions in Figure 9a, and Figure 9b, linear pressure level equations
to calculate fractional cloud cover from model RH were deduced by plotting 1)
cloud cover at

RH=100% from Fig. 9a, and 2) , the calculated RH and cloud cover from the relative humidity
depression (RHdep) where cloud cover =
f(
100)
/
e

from Fig. 9b. Parameters and equations for
various levels are listed below, along with an extrapolate
d zero fractional cloud cover. The zero
fractional cloud cover is necessary to prevent a negative

fractional cloud cover during the smart
initialization run.




Pres


Fractional Cloud

RHdep @ 1/e


Calculated cloud cover @ RH

Level



@100% RH


cloud cover


from RHdep @
f(
100)
/
e
cloud cover

(mb)



(%)



(%)




(%)


950



42



23




16 @ 77


900



58



27




22 @ 73


850



70



31




26 @ 69


800



90



35




33 @
65


700



118



42




43 @ 58


600



96



50




36 @ 50


500



79



58




29 @ 42


400



58



65




22 @ 35


300



38



73




14 @ 27



Pres


Zero Fractional

Derived Linear

Level



Cloud Cover


Equation

(mb)



(%RH)




950



64


1.13(RH)
-

71


900



51


1.33(RH)
-

66


850



59


1.65(RH)
-

95


800



46


1.63(RH)
-

73


700



35


1.79(RH)
-

61


600



20


1.20(RH)
-

24


500



10


0.86(RH)
-

7


400



None


0.55(RH) + 3


300



None


0.33(RH) + 5


Initial Testing


Initial testing at WFO Des Moines discover
ed that the 300 mb and 400 mb pressure levels
produced too much fractional cloud cover. During the initial test, the smart initialization
accumulated fractional cloud cover through the entire column. A significant improvement was
noted in subjective compar
ison to satellite imagery after removal of the calculations for 300 mb
and 400 mb since these layers had a positive influence

to total fractional cloud cover.


After several weeks of more subjective comparisons to imagery, it became apparent that too
much
fractional cloud cover was still being consistently produced. This was especially true at the
mid levels since 100 percent relative humidity at 700mb can produce 118 percent fractional
cloud cover. A simple averaging correction was introduced, based on the

number of fractional
cloud layers calculated. This averaging lowered fractional cloud cover too much, so much so,
that it was very difficult to produce an overcast sky. To maintain the underlying scientific basis,
the number of fractional cloud layers use
d to produce the average fractional cloud cover was
reduced by one. This had an immediate effect of increasing the cloud cover where relatively
humidity was high, with minor effects where the relative humidity was low. The
DGEX
use this
correction

since 85
0, 700, and 500 mb relative humidity data are all that is available
.


More subjective comparisons showed that the short
er

range models (
various Eta(s),
RUC,
and
NGM

and GFS
were still
light
on fractional cloud cover, even after reintroduction of the 400mb
and 300mb levels. The layer divisor in the shorter range models has been reduced to 5, (from a
total of 9 layers) which produces acceptable results.


Installation


These instructions assume you are logged into
px1f
as 'ifps'.


1. Copy the DMX_Sky_Inits.ta
r file to a directory of your choice.


2. Untar the file,
tar xvf DMX_Sky_Inits.tar
. A file listing should include My
GFS80
.py,
MyEta
12
.py, MyRUC
80
.py
, and MyDGEX.py
.


(The GFSlr
,
NGM
80, and Eta80

have been
removed since processing these models is rather r
edundant.)


3. In your /awips/GFESuite/
primary/
etc/SITE directory, determine if you have local SmartInit
files. If you do, then cut and paste sections of the DMX_Sky_Inits as necessary into your local
files. If you do not have local SmartInit files, simply

copy the files into the directory.


4. In the same directory, open your localConfig.py and add the smart initialization files for each
model as exampled below for the Eta
12

model.


del serverConfig.INITMODULES[“Eta12”]

serverConfig.INITMODULES[

MyEta
12”
]

= [E
ta12
]


5.
Change directory to /awips/GFESuite/
primary/
bin and test the syntax of your updated
localConfig file by ./runIFPServer
-
n. If there are no errors, move on. Otherwise, fix your
localConfig.py file until there are no Python errors.


6.
Stop

and restart your ifpServer.


7. Test your
Des Moines Sky Inits

as suggested below.


Suggestions


To increase efficiency in the use of GFE in forecasting, WFO Des Moines displays forecast
elements as GFE weather groups under the WeatherElement drop
-
down me
nu. Current groups
are MaxT, MinT, T, Td, Wind, Sky, SnowAmt, QPF, PoP_Wx, and Public. Once a weather
element group is loaded into Grid Manager, it is toggled to the vertical display. All model, MOS,
and HPC initializations of that specific group are loade
d into the Grid Manager for inspection.
Each group is a single element except for the PoP_Wx groups which display both the PoP and
Wx weather elements.


At the very least, a Sky group will have to be created to test the smart initializations for your
are
a. It is recommended that you start with the default smart initialization, then make adjusts to
the number of layers to fine tune fractional cloud cover to subjective cloud cover from satellite
imagery. Keep in mind that, using relative humidity alone as

an indicator of cloud coverage,
cloud amount can be assessed only to within a root
-
mean
-
square difference of 15%
-
30%.


Testing of the
Des Moines Sky Inits

can be accomplished following this procedure after
installation:


1. Load your Sky group with the
Grid Manager in the vertical in GFE.


2. Open a Linux xterm window as user IFPS.


3. In /awips/GFESuite/
primary/
bin do ./ifpInit
-
a My
RUC80

(the .py file extension is not
required).


4. Monitor how your Sky weather element changes in the GFE Spatial Editor
, making sure to
stay ahead of the smart initialization processing by stepping ahead in the Grid Manager.


5. It would be advantageous to have another xterm open to tweak the appropriate SmartInit
python file by changing the layer divisor. You may try less

than whole numbers, such as 2.4 for a
divisor. Make your changes, and run the script again. It will be obvious when the python script
has errors. Continue to monitor the Sky inits over a period of time and make adjustments.


6. If you desire, you can deci
de NOT to install a specific model Sky Init, such as the NGM. By
doing that, you can get a feel for the differences between the default Sky init and your newly
installed inits.


7
. Coordinate your changes with the surrounding WFOs that share your ISC grids
.


WARNING

A topographic check is part of each model Sky SmartInit. This prevents cloud development
below ground level. At the same time, NO testing of the impact of this feature has been
completed for high terrain. It is unknown what the impact of adjust
ing the layer divisor will be on

layers that disappear and reappear around high terrain.


ACKNOWLEDGEMENT

Thanks go out to Mr. Mike Romberg of the Forecast Systems Laboratory in Boulder, Colorado
for python assistance and the forecasters at WFO Des Moines
for valuable input.