CloudSat L3 Product by cuiliqing

VIEWS: 36 PAGES: 28

									CloudSat L3 Product
     William B. Rossow
  City College of New York
        List of L2 Products
2B – GEOPROF
2B – CLDCLASS
2B – CWC-RO
                      ABOUT 10 GB/day
2B – TAU              In Pixel/Scanline—
2B – CWC-RVOD         Orbit Format
2B – FLXHR
2B – GEOPROF-LIDAR
2B – CLDCLASS-LIDAR
MODIS-AUX
ECMWF-AUX
               Accessing L2

       Current Data Access System NOT
         Suitable for Climate Studies

(E.g. Cannot Order Even 1 Global Month of 3 Products)



      Many Changes Required to Facilitate
              Climate Studies
RATIONALE FOR L3 PRODUCT
   BASED ON ISCCP EXPERIENCE


Can not use L2 for climate studies (currently)

   Most ―non-satellite‖ users will not use
       pixel/scanline—orbit format

    Even ―satellite‖ users tend not to use
         another satellite’s dataset
RATIONALE FOR L3 PRODUCT
  BASED ON ISCCP EXPERIENCE


      Most users are ―confused‖ by
      sampling & re-mapping issues



 Most users start with monthly mean maps
RATIONALE FOR L3 PRODUCT
 BASED ON ISCCP EXPERIENCE


     Diagnostic Studies facilitated by
  ―all variables at each location‖ format

     User community only just now
            (after 20+ years)
   realizing value of joint histograms
     CLOUDSAT L3 PRODUCT

        Part A – Basic Cross-Sections

Twice-daily, Reduced Resolution (25 km - 500 m)

     Gridded at 2o x 4o with Cloud Fraction

Merged, Averaged L2 Variables at Each Location
CLOUDSAT L3 PRODUCT
 Part B -- Statistical Histograms

    Reflectivity vs Particle Size

 Optical Thickness vs Particle Size

  Water Content vs Particle Size

 Water Content vs Precipitation ??
     CLOUDSAT L3 PRODUCT
         PART C – Monthly Statistics
   Cloud Layer (Type) Properties from Part A
        (including Vertical Structure ??)
     Accumulated Histograms from Part B

           Additional Histograms
 Water Content– Particle Size– Temperature
Water Content—Particle Size—Relative Humidity
       Cloudy Alpha & Beta Parameters
       Clear Alpha & Beta Parameters
   Approximate Volumes

     PART A & B = 10 GB/month
(about 30 times larger than ISCCP D1)



      PART C = 100 MB/month
(about 10 times larger than ISCCP D2)

Should probably reduce volume more
TEST RESULTS
Along Track Cloud Amount
Full Resolution Layer Cloud Amount
        Averaging Tests
  LAYER CLOUD AMOUNT FREQUENCY




5 bin averaging      15 bin averaging
             Averaging Tests
• Along-track
                          7/1/06 - 7/14/06




          1 day                2 weeks
 Vertical Structure

                     N layers      Cloud
                                   layer %
                          1           51.0
                          2          40.7
                          3           7.6
                          4           0.7


50% optimal for number of layers
           Vertical Structure
• Aggregated over 14 days (7/1/2006 - 7/13/2006)




   20% optimal for layer separation & thickness
BACKUP SLIDES
File Size per Granule
1B-CPR                    20 MB
2B-CLDCLASS               19 MB
2B-CWC-RO                 160 MB                          ******
2B-FLXHR                  34 MB
2B-GEOPROF                33 MB
2B-TAU                    9 MB
ECMWF-AUX                 90 MB                           ******
MODIX-AUX                 51 MB
 Source: CloudSat Standard Data Products Handbook, CIRA
                     File structure
•   0.16 s/profile * 36,383 profiles/gran = ~ 5800 s/gran
•   86400 s/day / 5800 s/gran = ~ 15 files/day per variable
•   1 granule/file
•   Geolocation data and data quality flags are repeated
    in 6 2B files
    – ~ 9 MB/granule * 15 files * 5 = 675 MB/day redundant
• ECMWF has lower vertical and horizontal resolution
  and is interpolated to vertical bin res of 240 m
    – Can reduce size from 80 MB/granule 20 MB/granule by
      storing at intrinsic resolution with a pointer
• Current format yields ~ 10 GB/day of data

              GOAL for L3 is 3-10 GB/month
              (10-30 Larger than ISCCP D1)
               Reduction Factor is 30-100
        Level 3 Science Product
• Stack variables at the pixel level
• Binary vs HDF format:
   – Time to read in 1 granule for 4 L2 HDF products (2B-CLDCLASS,
     2B-CWC-RO, 2B-GEOPROF, ECMWF-AUX) is ~3 s
   – … and write out as 1 binary file is ~8 s
   – … and read in 1 binary file is ~9 s (=> ~1 s to read in binary file)
• Tests performed using IDL
         Sampling/Averaging Data
   • Sample or average along track or vertically
       – GEOPROF CPR_Cloud_mask (cloudy if 20-40, clear if 0)
       – Accumulate difference between actual and sampled or
         averaged value and total number of detections (cloud
         amount)
       – Differences taken at pixel level
       – If sampling or averaging vertically, replicate values

    Summary of CloudSat data resolution tests
    rms variances

                                              1 day   2 days     14 days
    Sampling        Vertical                   0.72     0.64        0.75
                    Along Track           0.07-0.60            0.02-0.20

    Averaging       Vertical                   0.51    0.46         0.53
                    Along Track   1 deg   0.01-0.16            0.00-0.07
                                  2 deg   0.00-0.12
                    Both                  0.36-0.39


Vertical Factor = 2, Horizontal Factor = 10, Total = 20
              Averaging Tests
 • Along track averaging: RMS error increases with
   sampling interval
 • Averaged in 1 latitude bins




5 bin averaging : RMS=0.03    15 bin averaging: RMS=0.16
Averaging tests
             Difference Map




Trimmed off regions with no/low detections
              Vertical Structure
   • Aggregated over 14 days (7/1/2006 - 7/13/2006)




Between 0% & 25% optimal for layer thickness & separation
                            Basics
• One cloud profiling radar (CPR) profile every 1.1 km
  along track
• Each profile has 125 vertical bins
   – Each vertical bin ~ 240 m thick
• Footprint for a single profile is ~ 1.4 km (across-track)
  by 2.5 km (along-track)
• A granule is one orbit
       • A granule stars at the 1st profile that falls on or past the equator
         on the descending node
       • ~ 36,383 profiles per granule
Cloud Layer Thickness
   Thickness     N
      (m)
   240         477318
   480         287381
   720         254148
   960         268701
   1200        257948
   1440        232157
   1680        203949
   1920        177229
   2160        157046
   2400        138950
   2640        124770

								
To top