1. Ground and satellite-based night-time cloud identification techniques
at the Pierre Auger Observatory
Johana Chirinos
Michigan Technological University
Workshop
Cloud pic

2. Identifying Clouds over the Pierre Auger Observatory
As part of atmospheric studies for Pierre Auger Observatory:
Identifying night-time clouds in the field of view of the Fluorescence
Detectors(FD) is important for getting correct shower profiles.
Ground-based night-time cloud identification techniques:
XLF and CLF laser shots seen by FDs
Infrared Cloud Cameras
LIDARs (previous talk by Aurelio)
Satellite-based night-time cloud identification technique:
GOES12/13 satellite images over Southern Hemisphere
04/18/11
Workshop

3. Shower profile from cosmic rays
Shower profile:particles along shower path
correct?
Shower development viewed by 4FD
FD pixels seeing shower profile
04/18/11
Workshop

4. Clouds affecting shower profile
Attenuation of light by cloud
Cosmic Ray
Shower
Decrease or absence of fluorescence light
Hacer un mejor dibujo FD
04/18/11
Workshop

5. Clouds affecting shower profile
Forward directed light cone
Forward directed light scattered towards FD
Cosmic Ray
Shower
Hacer un mejor dibujo FD
04/18/11
Workshop

6. CLF/XLF FD
04/18/11
Workshop

7. Cloud over CLF Cloud between CLF&EyeP r o f i l e
04/18/11
Workshop

8. IR Clouds Cameras 4 IR cameras: one at each FD site.
During FD operation:
-Full sky mosaic
27 images every 15'
04/18/11
Workshop

9. 5 images across FDs field-of-view every 5'
IR Clouds Cameras
5 images across FDs field-of-view every 5'
Cloud
Camera
Images
Image
Processing
FD Pixels
Cloud
Index
04/18/11
Workshop

10. IR Clouds Cameras Cloud-affected event Cloud Camera Image
Image Processed
04/18/11
Workshop

11. Satellite GOES12/13
Extensive atmospheric monitoring task for many topics over 3000km2:
Great effort for constructing, running, maintaining equipment/software.
For cloud identification: Satellite data will eliminate these worries Better resolution in space and time, especially to identify exotic events.
Double bang
Cloud
04/18/11
Workshop

12. Satellite GOES12/13
GOES (Geostationary Operational Environmental Satellite) at 35,800 km.
Full-disc view of Earth: 1 visible band: 1km by 1km.
4 IR bands: km by 4km.
GOES-12/13 Imager at 75º W: Southern Hemisphere Scan Sector
68-70º W longitude, 34-36º S latitude each 30' (hh09ss, hh39ss) at night
04/18/11
Workshop

13. 4 Satellite IR Bands
Emission spectrum for a 280 K black-body at Earth's surface with absorption effects of atmospheric water vapor:
Cloud identification during FD shifts.
Satellite dark time: 00:09am / 03:09am / 06:09am / 09:09am
- Band 2 & 4: unaffected, none of earth's atmospheric gases absorb very well Able to sense earth's surface and clouds.
- Bigger effect in band 3:responds to water vapor at middle and upper layers.
04/18/11
Workshop

14. Ground-truthing cloud identification technique
To ground-truth algorithm: comparison with instruments at Auger Observatory. :
IR Cloud Camera and LIDARs:
Different field of view/timing: not easy/perfect comparison.
CLF:
Cloudy/clear state of CLF pixel is regularly monitored by CLF.
Every 15' during FD nights: CLF produces 50 laser shots seen by FD.
For 2007 data, we plot all 50 laser profiles in one plot:
If peak over laser profile: Cloud over CLF
04/18/11
Workshop

15. CLF(1 hour data) 07:09 07:00 Clear 07:15 Clear Satellite 07:39
Clear/Cloudy tag of CLF pixel: If CLF shots 9' before & 6' after both clear/cloudy.
CLF pixel clear
07:30 Cloudy over CLF :45 Cloudy between CLF & Eye
Satellite
07:39
CLF pixel undefined
04/18/11
Workshop

16. Ground Temperature Correlation
T2 & T4: sensitive to temperature of emitting surface.
For clear pixels: T2 & T4 should be correlated with ground T.
For CLF pixel: T4 vs. T of ground from weather station at CLF(Tclf).
Correlation for clear nights:
T4 ~ Tclf
T4 < Tclf when cloudy.
Linear fit for clear nights:
Small residual: CLF pixel clear.
Large residual: CLF pixel cloudy.
04/18/11
Workshop

17. Residuals animated gif
Assuming a relative uniform T of region: T of all pixels from region ~ Tclf
For every pixel: When residual small, pixel is clear(clear blue).
When residual large: pixel is cloudy(dark blue).
Better model of T over array: WS at CLF, at 4 FD, at BS and nearby towns.
04/18/11
Workshop

18. Satellite info only: T3 vs T2-T4
T3 & T2-T4: independent of current ground T.
Clear by CLF : condensed blob
Cloudy by CLF: anti-correlated line
Greatest separation between clear and cloudy.
We project data onto this line.
We fit to an empirical function of cloud probability.
04/18/11
Workshop 18

19. Satellite info only: maps
Pixels with cloud probability < 20%: light green, clear. Pixels with higher cloud probabilities in grey scale. ATMON 2010
04/18/11
Workshop

20. Satellite info only: maps
Cloudiness since 2007
04/18/11
Workshop

21. Ideas for a collaborative work :
We have more than 5 years of different types of atmospheric data. These could be useful to other scientists.
For example:
Ground-truthing MODIS satellite results for a CTA site search.
04/18/11
Workshop

22. Cloud Identification Principles
Clouds are typically colder than Earth's surface:
T2&T4(non-absorbing): Sense unattenuated radiation from emitting surface.
Lower T2 or T4 than Earth T: marker for clouds.
Wavelength dependence in emissivity of clouds, but not for Earth.
Depends on relationship between cloud droplet size and wavelength.
T2-T4:sensitive to emissivity differences for clouds between the bands.
T2-T4:  for clouds. T2 ~T4: Earth surface.
Clouds: mixture of water vapor and liquid water droplets.
T3 varies with water vapor: fraction of cloud in a pixel.
Cloud identification algorithms with T2, T4, T2-T4, and T3 appear promising.
04/18/11
Workshop