Science Systems & Applications, Inc/Climate & Radiation Laboratory

Affiliation: Science Systems & Applications, Inc/Climate & Radiation Laboratory
Event Date: Wednesday, October 16, 2019

Location: Building 33, Room H114
Time: 1:30 PM

The MARA System and Its Applications

The heritage operational methodology for generating gridded statistical aggregations of Atmosphere Product retrievals (e.g., cloud, aerosol, atmospheric water vapor, etc.) from the MODerate resolution Imaging Spectroradiometer (MODIS) is generally quite static. Adding or modifying statistical capabilities is a time and labor-intensive task. The current algorithm and aggregated products (MOD08_x3/MYD08+x3 for Terra/Aqua MODIS, respectively) are also tightly coupled to the MODIS instrument and only support serial execution. Because of these limitations, it has not been possible to utilize the existing aggregation algorithm for research-level global analyses of the MODIS products. For example, variable or geographic data sub-setting and/or aggregation resolution changes have not been an option. The existing algorithm has also suffered from various portability issues due to its dependence on custom toolkits. Other aggregation systems such as YORI depend on per-granule precomputation of desired statistics, which may or may not be desirable under specific conditions and for a large aggregation leads to significant disk usage for temporary file storage.
The research-level Multi-instrument Atmospheric Retrieval Aggregation (MARA) system has been developed to address these challenges. MARA is a flexible, modular, cross-platform code written in a mix of Fortran90 and C without any custom toolkits. It uses a fully external instrument, aggregation type, variable and statistic specification, and fully automates output variable name and type generation removing the heavy labor component of the MOD08/MYD08 file specification process. The MARA system also externalizes specification of aggregation resolution and any temporal constraints, and features customizable geographical sampling options via, e.g., pre-defined regular boxes or irregular coordinate sets such as the shape of a section of a country. MARA gives users an unprecedented level of detail during aggregation as it is able to create videos of aggregation as it is happening thus giving insights as how orbital overlap and timing may influence the final gridded product. Furthermore, the MARA system supports multiple parallelization levels determined by the problem size in order to optimize computing resource use.
In this presentation we will show the current implementation of the MARA system and show analyses that previously would have been prohibitively expensive to perform.
613 Seminar Series Coordinators