Synthetic satellite imagery
Louie Grasso
Cooperative Institute for Research in the Atmosphere
NOAA/NESDIS CoRP 15-16 August 2006
Introduction
• Motivation
• 32-bit
• GOESR and NPOESS.
• 64-bit cluster.
• Simulations and synthetic imagery
• Summary.
Motivation
1) Better understand satellite imagery. That is, synthetic satellite imagery can be used with model output to understand the imagery.
2) Synthetic satellite data can be used to aid in algorithm development.
3) Synthetic imagery can be used for NWP model verification.
32-bit
1) 1997: AIX operating system, 196 Mbytes of ram, few hundred Mbytes of disk space, RAMS-3b.
2) 1998: Linux operating system, 512 Mbytes of ram, 500 Mbytes of disk space, RAMS-429. Wow! Just think what I can do now.
3) Upgrades! 1 Gbyte of ram 2 Gbytes of ram 3.3 Gbytes of ram. Standard Linux Kernel Bigman Kernel Experimental Kernel. Disk space increased to 100 Gbytes. RAMS-43. Oh boy! Just think what I can do now.
4) 32-bit cluster, 20 dual processor nodes, 2 Gbytes of ram, few hundred Gbytes of disk space. Test RAMS-43. Gee-wee what a fast machine! Just think what I can do now.
GOESR and NPOESS
1) Run RAMS43 and use output to make synthetic GOESR-ABI and NPOESS VIIRS images for three mesoscale weather events.
2) 200 Gbytes of disk space should be plenty—I was in for a surprise!
3) 32-bit 4 Gbyte ram too restrictive. GOESR-ABI has 2 km footprint, but I had to use 4 km horizontal grid spacings. NPOESS VIIRS has a horizontal footprint near 400 m, but I had to use 1 km horizontal grid spacings.
4) Horizontal extent of domain too small to produce a realistic looking synthetic satellite image.
5) Must move to 64-bit cluster: A whole new world!
64-bit
1) Test RAMS43. (a) Reproduce a 32-bit run, (b) Exceed 4 GB of ram.
2) Repeat (6), this time do runs in parallel.
3) More ram added: 8GB16GB32GB. 32GB is hardware limit.
4) More disk space added: 500GB1TB2.5TB raid to 10TB raid.
5) Each time more ram was added, I re-ran the same job to take advantage of the extra memory by increasing the horizontal domain.
6) The first time I submitted a 25 GB job, a lot of time went by before the model wrote out the initial data. I thought something was wrong, so I did what any good scientist would do—kill the job!
7) Fifteen minutes passed before RAMS43 wrote out 15 GB of initial data. I had no idea the files would be this large.
Simulations
Grid 1 Grid 2 Grid 3 Grid 48 May 2003 90x66 192x16
2502x522 1027x852
12 Feb 2003
90x66 192x162
502x522 992x902
1 Oct 2002 90x66 192x162
932x782 1002x1002
Table 1: Number of x and y points for each grid in each simulation.
Start grids 1-3
Stop grids 1-3
Start grid 4
Stop grid 4
8 May 2003 12 z 20 z 17 z 20 z
12 Feb 2003
00 z 19 z 17 z 19 z
1 Oct 2002
12 z 00 z 3 Oct
18z 2 Oct 00z 3 OctTable 2: Start and stop times for each grid in each simulation.
Synthetic 2 km ABI Bands 7-168 May Severe Weather Case (19 UTC)
3.9 µm 6.2 µm 7.0 µm 7.3 µm
8.5 µm 9.6 µm 10.35 µm 11.2 µm
12.3 µm 13.3 µm
GOES-12 at 10.7µm GOES-R at 10.35 µm NPOESS VIIRS at 11.02 µm 4 km 2 km 400 m
Synthetic GOES-12, GOES-R, and NPOESS-VIIRS imagery.
2 Oct 2002
Hurricane Lili
12 February 2003
Lake effect snow
8 May 2003
Severe weather
Summary
• Motivation
• 32-bit
• GOESR and NPOESS.
• 64-bit cluster.
• Simulations and synthetic imagery
