Hello, A yearly allocation for the LCRC cluster has been requested with the following updated information: Submitter/PI: Scott Collis Project Name: rainfall Division: EVS Project title: Developing radar based observations for climate model development Associated funding: DoE-CESD Climate Model Development and Validation (CMDV) Other Systems: Assorted ANL systems (my own 32 core machine). Systems at OLCF used for formal ARM work (open cluster with no quota system..) Science: The primary goal of the CMDV LLNL/ANL/BNL consortium project is to investigate if a very high resolution climate model (The DoE ACME model with regionally refined meshes at 10km) can reproduce mesoscale flows and features (~50km) that we observe in nature. Argonne's role is to create tailored retrievals and analyses to feed into a metrics based validation system. We will be focusing on the decadal record of radar data from the Australian Bureau of Meteorology's C-Band Radar in Darwin Australia. Project description: The project will use the Python-ARM Radar Toolkit (Py-ART) and NASA’s MultiDop package to ingest, correct and retrieve vertical velocities from dual polarimetric radars. Since time steps (~10 minutes) are independent the methodology is pleasantly parallel.  We will map the radar problem using dask. Dask manages execution at the nodes and Py-ART and Multidop manage the saving and reducing of the data.  During the past Rainfall project, vertical velocities from 4 years of data had been retrieved using Py-ART and Multidop. However, there are potential errors in the generated data that are caused by errors in Py-ART’s ability to dealias Doppler velocities. Therefore, the allocation will be used to reprocess these data using an improved dealiasing algorithms from codes developed by the Australian Bureau of Meteorology. The Py-ART codebase is fairly efficient with higher demand components written in C and FORTRAN. It is, however, memory intensive with 2-4GB per job needed. Fortunately this is perfect for the configuration of Bebop. Most importantly Py-ART and multidop are thread-safe.  We have already mapped the radar problem to Bebop using 1500 cores effectively, radically reducing the time to retrieve the vertical velocities. We will be using this allocation to work through many years (10+) of radar data, roughly 350,000 radar "Volumes".  The overwhelming majority of core hours will be used by jobs with greater than 100 cores.  The team consists of myself and my supervisor, Scott Collis.  Industry partnership: Project URL: http://climatemodeling.science.energy.gov/news/new-funding-opportunity-annou... Current FY Hours Used: undetermined amount New FY Requested allocation: 120000 Q1: 30000 Q2: 30000 Q3: 30000 Q4: 30000 Justification: Storage requirements: None, we have paid for a large storage system Thank You, The LCRC Accounts System