# Fog Monitor Overview

Document Sample

```					      Fog Monitor Overview

   Theory and Method
   Localization
   The Fog Monitor
   The Fog Monitor Display
Fog Monitor Overview
Theory and Method

separate presentation
by
Qin Zeng
Supporting theory

Fog monitoring is a kind of object-oriented processing
All we have to know is how to describe a fog object: its properties
and its behaviors
Spectral feature of fog in multiple satellite channels

(a) VIS Channel
(b) 10.7 (3.9) μm Channel                                   properties
(a) 3.9 μm Channel
(b) 10.7 μm Channel

Kinematic feature of fog
behaviors
Particle/channel          Fog        stratus        Other cloud        Long IR      Short IR     Vis

size vs wavelength        1~10       5~6            >10                10.7         3.9          0.4-0.7
μm         μm             μm                 μm           μm           μm

According to Mie scattering theory:
In the VIS channel, reflectivity mainly is related to the depth of the cloud/fog.
In channel 3.9µm, fog/stratus has maximum scatter rate with its size equivalent to 3.9 µm.

fog/stratus
cloud

VIS Channel             3 . 9 μm   IR             1 1 μm    IR

•Here two differences can be used to distinguish fog and cloud: 1) VIS difference; 2) 3.9 µm IR difference.
•The fog region in satellite image should be smooth, edge-clear&regular.
•Snow cover and sea ice are bad reflectors in channel 3.9 µm, while fog and stratus are good.
Data Normalization of the VIS satellite data
Normalization with fine tuning
Assumed that all objects are black bodies:
so TBemit3.9 = TB11 = Treal , and according to the Planck radiation law,

2hc 2
Remit3.9  R(TBemit3.9 )  R(TB11 )                                      ch
1
k3.9TB11
3.9 5 (e                1)
Rreflect3.9  Rdet ect3.9  Remit3.9
2hc 2                              2hc 2
                    ch
                  ch
k3.9TB 3.9                          k3.9TB11
3.9 5 (e                  1)       3.9 5 (e                1)
This is how reflective product of RAMSDIS ONLINE is generated . Because of the
assumption of black body and TBemit3.9 = TB11 , it is an approximated reflective product
and it does not take the angular correction into consideration
So in our algorithms, it is not adopted for fog detection but for sea ice/snow cover
exclusion.
http://www.cira.colostate.edu/RAMM/Rmsdsol/main.html
http://www.nrlmry.navy.mil/~turk/intro.html
Allen, R.C., P.A. Durkee, and C.H. Wash, 1994: Snow-cloud discrimination with
multispectral satellite images. J. Appl. Meteor
Channel           VIS (0.4 - 0.7μm)        3.9 μm IR                     10.7 μm IR

satellite                                  ( Mix together)

0 means “can be ignored”     1 means “major”

Without the presence of sun light, in fog region, a
positive value with
[10.7 μm channel - 3.9 μm channel]. This
difference is caused by the emissivity difference of
fog between 10.7 um and 3.9 μm.

Above will be good only during the night time:
Other water/ice clouds like AC/AS also look similar
to Fog in fog product [7]. But in 10.7 μm channel ,
the brightness temperature of fog is higher than
that of higher water/ice cloud (AC/AS). This can
in a degree help distinguish them.
R satellite) = R(ground) - R(absorbed) + R(cloud)
(

Different emissivities of different cloud types
Absorptivity is related to emissivity and the depth of the cloud/fog

Thin cloud        (night)    : TB(10.7 μm ) - TB(3.9 μm )   <0
Thick cloud       (night)    : TB(10.7 μm ) - TB(3.9 μm )   ~0
Fog               (night)    : TB(10.7 μm ) - TB(3.9 μm )   >0
ice cloud         (night)   : TB(10.7 μm ) - TB(3.9 μm )    <0
Water/ice cloud   (night)   : TB(10.7 μm ) - TB(3.9 μm )    >0

Comment: So fog product should be
checked to eliminate other higher
water/ice cloud (AC/AS) by using
TB derived from 10.7 μm channel.
Sea Fog Discrimination Chart from Gary Elrod
SAFESEAS Fog Monitor

 process only those parts of the images that cover the monitoring area.

Nighttime                                              Daytime
    use the visible image
   use 3.9 and 10.7 um infra-red images to        Based on thresholds, apply feature extraction
produce night time fog product.                 algorithm to identify fog areas.
   Apply feature extraction to identify fog       apply smoothness filter to eliminate noise. (optional)
area based on night time fog productTwilight
    Apply a filter to eliminate snow/ice (optional)
threshold                              No
Detection

 Filter out high cloud using channel 10.7 um
 Apply a filter to eliminate very tiny size area as noise. (optional)
   Apply a filter to eliminate edge-irregular detected area (Optional)
SAFESEAS Fog Monitor
Smoothness Definition:

Smoothness = (1- deviation/mean) * 100%

1 n
  ( xi  x ) 2
n i 1
smoothness  1                    100%
x
1 n
x    xi
n i 1

X denotes the gray value at the VIS pixel
and n is the number of pixels
SAFESEAS Fog Monitor
Fog Monitor D2D display
SAFESEAS Fog Monitor
Fog Monitor D2D Display (Twilight)
Limitations (1)

   When the sky is overcast with high level (thick) cloud, fog can not be seen through by
satellite.
   During the transition of daytime and nighttime, neither VIS product nor fog product is
good. And twilight is exactly the time when fog mostly affects marine interests.
    Lack of observations in marine zones causes problem for verification
    No effective approach to distinguish fog from stratus just using satellite data
Limitations (2)
Data normalization

At very low solar elevation, in the clear air
region , there will still be small
brightness values. After normalization,
they will be enhanced to look like fog
areas and will be detected by the Fog
Monitoring Processor.
Limitations (3)
Radiation from the ground goes through thin cirrus/high cloud ,which makes cirrus look like fog in VIS
Limitations (3) (continue)                            [14]

Use fractal dimension to try to reduce the detection of the irregular area
This is still not a good solution but supplied as an optional filter for user

FD = 2ln(P/4)/ln(A)
P= perimeter A= Area

•    Count the outside edges of the grid boxes as the perimeter
•    Amount of the grid boxes as the area.
Limitations (3) (continue)

Before                                                        After

Application of Fractal Dimension to filter out irregular area
Limitations (4)
Limitations (4) (continue)
References
   [1] Liu Jian, Xu Jianmin, Fang Zongyi Analysis of the particle sizes at the top of cloud and fog with
NOAA/AVHRR data Quarterly Journal of Meteorology
   [2] http://www.cira.colostate.edu/ramm/visit/fog.html fog product tutorial
   [3] http://meted.ucar.edu/topics_fog.php Fog/stratus tutorials
   [4] Chen Weimin, Satellite Meteorology (Chinese version) Beijing Meteorological publishing house
   [5] Ellrod G P. Advances in Detection and Analysis of Fog at night Using Advanced Very High
Resolution Radiometer(AVHRR) Imagery [J] Meteo. Magazine
   [6] http://meted.ucar.edu/topics_satellite.php satellite products on METED
   [7] http://meted.ucar.edu/satmet/goeschan/ GOES channel selection on METED
   [8] http://www.ssd.noaa.gov/ satellite service division homepage
   [9] http://aviationweather.gov/awc/aviation_weather_center.html Aviation Weather Center
   [10] http://www.cira.colostate.edu/ Cooperative Institute for Research in the Atmosphere, Colorado
State University
   [11] http://www.nrlmry.navy.mil/ Naval Research Laboratory Monterey
   [12] ALBERS S. 1992: Photometric correction of GOES visible satellite images. Preprints, Sixth conf.
on Satellite Meteorology and Oceanography, Atlanta, GA, Amer. Meteor. Soc., 223-225.
   [13] Allen, R.C., P.A. Durkee, and C.H. Wash, 1994: Snow-cloud discrimination with multispectral
satellite images. J. Appl. Meteor
   [14] Olsen, E.R., R.D. Ramsey and D.S.Winn. 1993. A modified fractal dimension as a measure of
landscape diversity.Photogram. Eng. Remote Sens. 59: 1517-1520
   [15] http://www.cira.colostate.edu/RAMM/Rmsdsol/main.html RAMSDIS ONLINE
Fog Monitor Overview
Key Terms
   Area of Responsibility
An office’s area of responsibility (AOR) will consist of all the zones and counties
for which it normally has the responsibility of issuing forecasts, watches, and
warnings.

   Monitoring Area
An office’s monitoring area (MA) will consist of its AOR plus the AOR of each

   Channel 2
Channel 2 refers to the 3.9 micrometer infra-red satellite image.

   Channel 4
Channel 4 refers to the 10.7 micrometer infra-red satellite image.

   Fog Product
The fog product is the image derived by subtracting 10.7 micrometer image
brightness temperatures from 3.9 micrometer image brightness temperatures.
Fog Monitor Overview
Localization

   will determine the default MA.
* which zones and counties are included in the MA.
* (OB-6) which fixed stations are in the MA.
* (OB-6) in which zone/county is each fixed station located.
For the alpha test, the default MA will be based on the same
“shape” files used for OB-4 SAFESEAS. It is the responsibility of
each office to make sure its systems have the current versions of
these files before localization is run.
   will generate the “dummy” fog threat image to display for
incomplete image sets.
   will generate the default (factory setting) monitor thresholds.
   (OB-6) will generate the default (factory setting) display
thresholds.
Fog Monitor Overview

The Fog Monitor will provide a GUI for customizing the MA.
Specifically, the office will be able to:

• add and delete zones and counties.
• (OB-6) add and delete fixed stations.
• (OB-6) customize the association of fixed stations with zones
and counties, that is, which stations are monitored to determine
each zone’s / county’s threat level.
Fog Monitor Overview

The Fog Monitor will provide a GUI that will enable offices to
customize the range of:
 normalized brightness values that should be considered fog in

visible images.
 temperature differences that should be considered fog in fog

product images.
Fog Monitor Overview

The customization GUI will also allow offices to enable/disable filters in the
fog recognition algorithm:
 in channel 2 images, each pixel colder than a customizable threshold

temperature should be considered snow/ice.
 in channel 2 images, each pixel having a brightness temperature colder

than a customizable threshold temperature should be considered cloud.
 in channel 2 images, each pixel having a brightness temperature warmer

than a customizable threshold temperature should be considered ground or
water surface.
 in visible images, a group of contiguous pixels having a brightness

variation greater than a customizable threshold cannot be considered fog.
 in visible images, a group of contiguous pixels having fewer than a

customizable number of pixels cannot be considered fog.
 a pixel within a configurable number of degrees from the terminator should

not be monitored.
Fog Monitor Overview

In OB-6, a GUI will also be provided that will enable offices
to customize the thresholds used for monitoring point
observations (METARs, C-MAN reports, ship reports, buoy
reports, and mesonet reports) for fog. We will “negotiate”
details for this tomorrow!
Fog Monitor Overview
The Fog Monitor
The fog monitor will be a persistent background process, much like the SAFESEAS, SCAN, and
FFMP processors. The alpha test version will monitor:
   visible imagery at those times and in those places that the sun is sufficiently above the horizon;
   the fog product at those times and in those places that the sun is sufficiently below the horizon.
In all cases, channel 2 images are used for certain kinds of filtering (unless the office has disabled
the filters).

The OB-6 fog monitor will also monitor point observations. The details of this will be “negotiated”
tomorrow.

The fog monitor will determine a fog threat level for each zone/county in the MA, and an overall
fog threat level for the entire MA. The threat level is the most severe of all the individual pixel
(OB-6: and point observation fog) threat levels for the past two hours.

The fog monitor generates a fog threat level image and a zone fog threat file for use by the Fog
Monitor display. Finally, the Fog Monitor threat level indicator and the Fog Monitor display are
signaled to update themselves.
Fog Monitor Overview
The Fog Monitor Display

   the Fog Monitor fog threat level image:
* black = this location is not within the MA.
* gray = insufficient data is available for determining
the fog threat level for this location.
* green = there probably is not fog at this location.
* yellow = there may be fog at this location.
* red    = there probably is fog at this location.

   (OB-6) the Fog Monitor station plot:
* only reports from within the MA are plotted.
* variables to be plotted will be “negotiated” tomorrow.
Fog Monitor Overview
The Fog Monitor Display (concluded)

   the alpha test Fog Monitor zone fog threat table:
* two columns: zone ID and fog threat level (color).
* click column header to sort by that column.

   the OB-6 Fog Monitor zone/station fog threat tables and
trends:
* except for variables, same as SAFESEAS.
* variables to be included will be “negotiated” tomorrow.

```
DOCUMENT INFO
Shared By:
Categories:
Stats:
 views: 7 posted: 9/1/2010 language: English pages: 30
How are you planning on using Docstoc?