CSSEF: Community Atmopshere Model (CAM) testbed activities
How water moves through the atmosphere is a central issue in climate science with enormous implications for society as well as the carbon cycle. Furthermore, the poor simulation of the water cycle by today's atmospheric models is a large impediment to the development of policy-relevant p redictions for climate change mitigation and adaptation. The organizing goal of CSSEF atmospheric model development is the creation of a weather resolving global atmosphere model that better simulates the water cycle. To achieve this goal, we envision the following central components:
- An atmospheric testbed that uses inâsitu and satellite water cycle observations to rapidly calibrate the model in a systematic and reproducible way using techniques developed by the uncertainty quantification community
- Computational science research to facilitate the running of weather-resolving atmospheric models on Leadership Class Facility (LCF) computational resources and the development of models capable of mesh refinement and non-hydrostatic dynamics
- Focussed parameterization development as needed to address the fact that the target resolution of the model is approaching the regime where non-hydrostatic dynamics become important.
The CSSEF atmosphere testbed activity is lead by Steve Klein, LLNL with co-leads Mark Taylor, SNL and Sally McFarlane, PNNL.
Summary of FY11 Tasks and Deliverables
|Task||Lab.||Goals and Deliverables by... b> td>
||August 1, 2011
||September 1, 2011
||October 1, 2011
||November 1, 2011
||CAM5-SE variable resolution simulation
||Infrastructure for SGP nest built. Generate and test suite of regionally refined grids over SGP
||Complete grid generation and testing.
||Complete initial CAM5-SE simulations with variable grid. Provide simulation with 1/8 degree resolution over SGP.
||CAM5-SE 1/8 deg resolution simulation
||Complete CAM-SE, CLM-SE and AMIP modifications. Benchmark tesing and test-bed data requirements
||Complete short (1-month) spin-up runs. Start full (1-year) experiment.
||Full simulation complete. Initial diagnostic analysis results (mean climate, spectra and visualizations)
||CAM5-SE forecast capability
||Test simulation complete
||Dataset preparation and uncertainty
||Initial comparison completed
||Cloud regime classification and metrics development
||Metrics applied to model observation comparison
||Precipitation Dataset development
||Initial dataset delivered to testbed
||Parameterization and resolution issues
||Generate and test regionally refined grids in aqua-planet. Configure diagnostic analysis approach. Complete initial set of aqua-planet experiments.
||Complete full set of aqua-planet experiments. Provide initial diagnosis of parameterization-resolution dependencies
||Comprehensive analysis of variable and uniform resolution dependencies in aqua-planet simulations (visualizations/spectra/budgets)
Task 1. CAM5-SE in forecast mode (CSSEF Proposal Page 98 Calibration Platform)
Personnel: Jim Boyle, 20% for all four months
- Early Month 1 : Obtain and build ESMF regridding utility at LC
- Month 1 : Generate grid descriptor files needed by regridding application
- Month 2 : Test robustness of regridding utility between FV lat-lon grid and CAM-SE grid and generate initial condition files for the atmosphere on CAM-SE grid
- Late Month 2 : Generate files of weekly SST/sea ice observations for the boundary conditions in the CAM format on SE grid
- Early Month 3 : Generate initial condition for land model on CAM-SE grid
- Late Month 3 : Test code to insert the SE initial conditions into the CAM initial or restart files
- Early Month 4 : Test running CAM-SE forecasts with prescribed SSTs, observed initial conditions, spun-up land
- Late Month 4 : Superficial analysis of simulation quality of the forecast against ARM data with comparison to identical forecast made with CAM-FV
Task 1: Implementation of metric associated with CMBE in testbed, including addition of uncertainty estimates on several variables, implementation of CMBE within ISDE, and development of framework for creating ESG-CF compliant files within ISDE. Perform initial comparison to CAM-SE at SGP
Relevant Proposal Deliverable: Initial design and prototype of calibration testbed, including at least one dataset and metric; p. 103
Laura Riihimaki; scientist with data/measurement expertise; Laura will add uncertainty estimates to CMBE variables; work with UQ group to understand uncertainty requirements and lead CMBE-CAM comparison
Krista Gaustad; software engineer; Krista will port the CMBE code to ISDE
Sherman Beus; software engineer; Sherman will develop the capability to produce ESG-style netCDF files in ISDE
Sally McFarlane; atmospheric scientist and atmosphere co-lead; Sally will manage task and coordinate with LNNL and data/testbed group
- July 15 Determine ESG netCDF standards; determination of measurement errors for selected CMBE variables
- July 31 Guidelines from UQ group on dataset uncertainty needs for UQ
- Aug 15 Prototype of code for propagating errors through CMBE averaging
- Aug 31 Implementation of templates for ESG-CF data standards into the ISDE process configuration manager
- Sep 15 Design metrics for CAM-CMBE comparisons including guidance from atmosphere UQ group; Port CMBE code to ISDE
- Sep 30 Implementation of error propagation code in ISDE; prototype code for CAM-CMBE comparison; obtain initial CAM-SE variable resolution files
- Oct 15 Produce and load CMBE files with uncertainty estimates into ESG; prototype of CAM-CMBE comparison codes
- Oct 31 Comparison of CAM-SE variable resolution runs to CMBE at SGP
Task 2: Develop precipitation and upper tropospheric water vapor metrics for validation platform using TRMM and geostationary satellite data; put TRMM and geostationary data in ESG; initial comparison of satellite data to CAM-SE
Relevant Proposal Deliverable: Initial design and prototype of validation platform; p. 103
Duli Chand; atmospheric scientist with satellite measurement expertise; Duli will develop precipitation and water vapor metrics from satellite data and lead satellite-model comparisons
Krista Gaustad; software engineer; Krista will produce ESG-format satellite files and implement metric codes in ESG
Sherman Beus; software engineer; Sherman will coordinate with satellite ESG groups to understand file formats and design file headers; contribute to visualization of CAM/data in ESG
Sally McFarlane; manage task and coordination with NCAR/SNL on model/data comparison
- July 15 Obtain example TRMM and geostationary satellite files; contact NASA JPL group for ESG satellite format conventions; determine software available in ESG
- July 31 Design of headers for satellite ESG files; Identify storage location for multiple years of global satellite data
- Aug 15 Obtain sample CAM-SE global run from NCAR/SNL; design of precipitation and water vapor metrics with input from UQ group
- Aug 31 Initial code for converting TRMM and geostationary files to ESG format
- Sept 30 Prototype code to calculate and visualize metrics from satellite and model data; Put multiple years of processed TRMM and geostationary files in ESG.
- Oct 15 Implementation of codes to create metrics in ESG
- Oct 30 Initial comparison of TRMM/CAM-SE
Task 3: Provide guidance to PNNL atmospheric data group on addition of uncertainty estimates to CMBE dataset, propagation of uncertainty, and development of metrics useful for UQ. Analysis of uncertainty associated with comparing single-point CMBE variables to grid-box average from model
Relevant proposal deliverable: Initial design and prototype of calibration testbed, including at least one dataset and metric; p. 103
Yun Qian (scientist/modeler provide expertise and multiple resolution model runs)
Laura Riihimaki (scientist/analyze multiple resolution model runs to estimate uncertainty associated with comparing single-point measurements to grid-box averages from models)
- July 15 Guidance to atmosphere group on needs of UQ group
- July 30 Guidance to atmosphere group on propagation of uncertainty
- Aug 30 Initial analysis of model runs with range of horizontal resolution
- Sep 30 Guidance to atmosphere dataset group on uncertainty associated with averaging specific single-point variables over various scales
- Oct 30 Application of above uncertainty analysis to CMBE uncertainty estimates
Task 1: CAM5.1 1/8 degree simulation
Personnel: Mark Taylor: July,August,September,October 40%
This task is in support of the page 107 year 1 milestone, Deliver high and low resolution CAM5 configurations for testbed. Mark Taylor (in collaboration with Rich Neale/NCAR) will configure CAM5-SE for 1/8 degree and perform short simulations, with CAM5 physics with prescribed aerosols. Short test simulations will be followed by a longer 1/8 degree simulation. This initial simulation is designed to provide a baseline 1/8 degree data set for initial testing of test-bed infrastructure and for comparison for the variable resolution configurations. It will also serve to demonstrate the scalability and performance of CAM5 on the DOE LCFs.
- July 15: Upgrade to latest CAM version, bring in CLM modifications to support CLM on the cubed-sphere grid, bring in AMIP modifications
- Aug 1: Perform short benchmark runs to tune load balancing and I/O options in order to obtain reasonable performance on Intrepid BG/P and determine precise CPU-hour requirements. (these simulations will be very expensive and performance is very sensitive). Coordinate I/O requirements (data needed) with test-bed.
- Aug 15: Perform several month long simulations to spin up the model and check that the simulation is reasonable.
- Sep 1: Initiate longer simulation
- Sep 15: Deliverable: finish simulation
- Oct 1: Complete initial analysis: AMWG diagnostics, spectra, visualizations
Task 2: CAM5 Variable Resolution
Mark Taylor: July, August, September,October 50%
James Overfelt: July, August, September,October 25%
Mike Levy (Post-doc): July, August, September,October 75%
Oksana Guba (Post-doc): July, August, September,October 75%
This task is in support of the page 107, year 1 milestone, Deliver CAM5 variable resolution: initial configuration with refinement to 1/8 degree resolution. For this task we will develop initial CAM5 variable resolution configurations with 1/8 degree resolution over the SGP ARM site. These initial configurations will be used to determine the best types of grid refinement, smoothing and scale-aware dissipation (expected to be completed in year 2). Based on previous work in CAM and other models, we expect obtaining good results will require meshes with smooth transition regions (from low to high resolution) and a resolution dependent dissipation mechanism. Mark Taylor will supervise the project and perform the initial analysis and evaluation in CAM (in collaboration with Rich Neale/NCAR and Mike Levy (SNL post-doc). James Overfelt will lead the mesh generation and smoothing project - generating a suite of locally refined meshes with different refinement and smoothing algorithms, and associated software engineering (in collaboration with SNL post-docs Mike Levy and Oksana Guba)
- July 15: Testing CUBIT generated grids in CAM4 Aqua planet
- Aug 1: Complete generation of suite of grids using the CUBIT algorithm and CUBIT based smoothing operations
- Aug 15: Continue testing addtional grids in CAM4 Aqua planet. Complete generation of grids with the alternative refinement and smoothing techniques.
- Sep 1: Continued testing in CAM4 Aqua plaet
- Sep 15: Complete initial CAM5 simulations with variable resolution
- Oct 1: Deliverable: Complete CAM5 simulations with 1/8 degree resolution over the SGP ARM site.
NCAR's committment to the atmosphere component of this activity is split into the the two major tasks of (a) implementing CAM5-SE variable resolution capability centered over the ARM-SGP site in real-earth CESM simulations (with SNL) in support of CSSEF atmosphere model science goal 1 and (b) diagnosing, identifying and addressing resolution-related parameterization challenges in support of CSSEF atmosphere science goal 3. Tentative sub-tasks and deliverables for each major task in FY2011 are outlined below:
Task 1: Regionally Refined Capability over the ARM-SGP (with SNL)
Personnel: Jose Garcia, Rich Neale; Mark Taylor, SNL
This activity aims to implement the CAM-SE regional refinement capability into the CESM modeling framework to ultimately be available in CESM releases. Part of this task is to provide the required compopnent datasets to initial and run experiments on specified regionally refined grids.
- Jul 15 Initial implementation of CAM5-SE/CLM-SE infrastructure for variable resolution on CESM trunk
- Aug 1 Initial dataset preparation and testing for the selected ARM-SGP centered refined grid.
- Aug 15 Low-resolution Testing of selected CUBIT-generated grids in the CESM framework
- Sep 1 Support SNL high-resolution grid testing activities
- Oct 1 Provide a user-ready (available on the code trunk) CESM capability to run AMIP-type experiments with at least one regionally refined ARM-SGP centered grid
Task 2: Resolution Dependent Parameterization Challenges
Personnel: Brian Medeiros, Rich Neale, Jose Garcia
Diagnose tropical mean state and variability in low (1 or 2 deg) and high (0.25 deg or 0.125 deg) resolution aqua-planet simulations using CAM4 as it relates to the changing role of individual parameterizations with resolution. This activity aims to document and understand how the role of physical parameterizations depends on model resolution in both global and reginally refined simulations. The additional impact of the method of grid refinement will be documented.
- Jul 15 Complete diagnostic scoping activities for existing aqua-planet CAM-SE interpolated and native grid output
- Aug 1 Completed 3-year simulations of CAM4 aqua-planet (control SST profile) at selected low and high resolutions. Select CUBIT grid for regionally refined tropical region.
- Aug 15 Initial diagnosis of high frequency statistics (daily and higher) and wavenumber frequency characteristics of uniform simulation experiments. Start regionally refined (1deg to 0.25deg) CAM4 aqua-planet simulations
- Sep 1 Complete regionally refined aqua-planet simulations. Perform preliminary uniform versus regional refined simulation anaylsis
- Sep 15 Complete alternative refined grid CAM4 aqua-planet experiments
- Oct 1 Diagnostic analysis of the parameterized physics reponse to uniform and regionally refined resolution simulations (mean climate, spectra, budgets, phenomenon). Propose simple scale-aware augmentations to the CAM deep convection parameterization