JAXA Supercomputer System Annual Report April 2016-March 2017

Report Number: R16E0099

• Responsible Representative: Takeshi Hirabayashi(Space Technology Directorate I, GOSAT-2 Project Team)
• Contact Information: Tomoo Yamasaki(yamasaki.tomoo@jaxa.jp)
• Members: Yoko Ueda, Tomoo Yamasaki, Yoshiyuki Hirosaki, Kazuhiko Yotsumoto, Yoshiteru Kawai, Yuki Kobayashi, Taro Makino, Atsuhiko Murakami, Takaaki Nishino, Kouhei Sugimoto, Yoshikuni Shindo, Hiroshi Takao, Yugi Yata, Shin Ishida, Junichi Takaku, Kenji Kowata, Takehito Yoshida, Fumie Kataoka
• Subject Category: Space(Satellite operation)

Abstract

As the successor to the 'IBUKI' (GOSAT) mission, GOSAT-2 aims to obtain the greenhouse gases concentrations with higher levels of accuracy via even higher-performance onboard observation sensors. The project will serve to provide observation data to environmental administrations and contribute the international anti-global warming efforts.

Goal

GOSAT-2 project retrieve and estimate the global concentration distribution of major greenhouse gases including the sources and natural absorbers with high level of accuracy to contribute to environmental administration as follows.

- Improved precision of climate change predictions

- Early detection of changes in the Earth system

- Better understanding of emission reduction level of the anthropogenic greenhouse gases and changing natural sink

- Contribution to air pollution monitoring policies

Also, GOSAT-2 project researches and develops new earth observation technologies required for future earth observing satellites.

Objective

Goals in accuracy improvement of greenhouse gasses distribution map is as follows.

- Retrieve monthly average of carbon dioxide (CO2) and methane (CH4) at a accuracy levels of 0.5 ppm and 5 ppb respectively over an entire atmospheric column, at a 500km mesh over land and at 2,000 km mesh over ocean under condition of less interference of clods or aerosol in the atmosphere.

(ppm: parts per million, ppb: parts per billion)

Goals in accuracy improvement of greenhouse gasses net flux over land is as follows.

- Estimate the net flux with ±100% accuracy in the region where monthly CO2 net flux at about a 1000km mesh is ±0.2GtC/area/year or more.

GtC:Giga-tons of Carbon

References and Links

Please refer 'JAXA | Greenhouse gases Observing SATellite-2 (GOSAT-2)'.

Use of the Supercomputer

GOSAT-2 project utilizes JSS2 as one of the GOSAT-2 Mission Operation System which processes the observation data of GOSAT-2. When processing algorithm is updated, JSS2 reprocesses all data observed in the past. Also, JSS2 is used as a remote storage of all data required for its reprocessing.

Necessity of the Supercomputer

As the reprocessing targets of GOSAT-2 products extends to all data observed in the past, more computer resources (core, memory, storage, etc.) are required than in the real-time processing.

It is necessary to use JSS2 to shorten the reprocessing time and to provide the reprocessing products to GOSAT-2 users more quickly.

Achievements of the Year

In the process of manufacturing and testing of the GOSAT-2 Mission Data Processing Software, we completed the following evaluation work and decided to adopt each option.

(1) Evaluation of job scheduler options

We evaluated the effect of job scheduler options on SORA-PP that can run GOSAT-2 Mission Data Processing Software as fast as possible.

As a result, GOSAT-2 decided to adopt the following options.

① -P exec-policy=share (A physical node is shared with other jobs.)

② --no-quota-check (Submitting the job without checking quota usage.)

(2) Evaluation of Intel(R) compiler options

We evaluated compiler options that can control the semantics of floating-point calculations to match the level-1 product quality created on SORA-PP at JSS2 and created by the GOSAT-2 Mission Operation System computer at Tsukuba Space Center.

As a result, GOSAT-2 decided to adopt the following options.

① -fp-model precise (Disables optimizations that are not value-safe on floating-point data)

② -fp-model source (Rounds intermediate results to source-defined precision)

Publications

N/A

Computational Information

• Parallelization Methods: Serial
• Process Parallelization Methods: n/a
• Thread Parallelization Methods: n/a
• Number of Processes: 1
• Number of Threads per Process: 1
• Number of Nodes Used: 600
• Elapsed Time per Case (Hours): 0.0167, 0.1333
• Number of Cases: 9

Resources Used

Total Amount of Virtual Cost(Yen): 400,552

Breakdown List by Resources

Note: Virtual Cost=amount of cost, using the unit price list of JAXA Facility Utilization program(2016)

JAXA Supercomputer System Annual Report April 2016-March 2017