earth system modeling framework status cecelia deluca noaa cooperative institute for research in...

12
Earth System Modeling Framework Status Cecelia DeLuca NOAA Cooperative Institute for Research in Environmental Sciences University of Colorado, Boulder [email protected] ESMF Executive Board December 13, 2011

Upload: everett-richard

Post on 04-Jan-2016

214 views

Category:

Documents


1 download

TRANSCRIPT

Earth System Modeling Framework Status

Cecelia DeLucaNOAA Cooperative Institute for Research in Environmental Sciences

University of Colorado, Boulder [email protected]

ESMF Executive BoardDecember 13, 2011

Release Update• Delivered ESMF 5.2.0r in July 2011

– Mature, portable, fully featured component framework– Low overhead (typically less than 3%)– Parallel regridding with many options (bilinear, higher order,

conservative methods; logically rectangular grids and unstructured meshes; pole options; 2D or 3D; from file or during model run)

– Regridding, metadata handling, calendars, and other utilities can be used separately from components

– Compliance checker helps developers understand coding conventions

HOMME Cubed Sphere Grid with PentagonsCourtesy Mark Taylor of Sandia

FIM Unstructured Grid Regional Grid

ESMF supported grids

Fact Sheet

• Open source• 4000 downloads• 600,000 lines of source code• 4600+ unit and system tests in regression test suite• Tested on ~24 platforms nightly • 387 public methods, all methods are tested• Comprehensive user documentation and examples• 75% of interfaces are “frozen” and backward compatible• Complete Fortran interface; C and Python as needed

• About 7 FTE developers• In third funding cycle: sponsored by NOAA NWS, NOAA CPO,

Department of Defense and NASA• Development priorities set quarterly by multi-agency Change Review

Board

Support Metrics

2.2.2r 3.1.0r 4.0.0r 5.2.0r

Bug Metrics

2.2.2r 3.1.0r 4.0.0r 5.2.0r

About ¼ of current bugs are behavioral problems; other categories include gaps in documentation and tests, and consistency issues

Feature Request Metrics

2.2.2r 3.1.0r 4.0.0r 5.2.0r

A Common Model Architecture for Interoperability

• ESMF component interfaces alone do not guarantee technical interoperability – ESMF can be implemented in multiple ways

• Also need:– A common physical architecture – the scope and relationships of physical

components (e.g. what processes are wrapped as components? land surface as subroutine or component?)

– Metadata conventions and usage conventions (e.g. who can modify component data?)

– The next steps in model interoperability involve encoding these conventions in software tools and templates

• The National Unified Operational Prediction Capability (NUOPC) is developing conventions and code templates that target a more advanced level of interoperability than ESMF alone

• A “NUOPC Layer” prototype is bundled with ESMF 5.2.0r

Functions of NUOPC Layer Prototype

• Establish an architecture in which major components are siblings that can operate concurrently or sequentially.

• Allow inheritance from generic component templates.• Couple components with pre-fabricated connectors. • Standardize the number and function of phases of initialize, creating a

standard setup pattern. • Constrain the data that may be sent between components with

standardized field data structures and a CF-based field dictionary.• Implement a compliance checker to provide feedback during

component development.• Use compatibility checking to determine if required import fields for a

component were supplied. Other run-time reporting alerts users to any issues with compliance.

Recent Impacts of ESMF• Interpolation of cubed sphere (unstructured) and lat/lon ocean

– User: Community Earth System Model– Grids: HOMME cubed sphere atmosphere to lat/lon ocean grid– Impact: ESMF parallel conservative regridding enabled integration of a high resolution

dynamical core into CAM, reduced distortion near the pole. • Enables CLM land model to run on cubed sphere

– User: Community Earth System Model– Grids: Land lat/lon to HOMME cubed sphere– Impact: ESMF parallel bilinear mapping from lat/lon to HOMME cubed sphere allowed

high resolution land model to move forward for CESM.• Better values at poles for unstructured to lat/lon remapping

– User: Community Earth System Model / DOE collaborators– Grids: NCAR MPAS unstructured grid to POP ocean grid– Impact: ESMF conservative interpolation solved problems with bad weights at the pole.

• Allows fast interpolation of large topography data set– User: NASA Global Modeling and Assimilation Office– Grids: 4km global lat/lon grid to 7km cubed sphere grid– Impact: ESMF conservative regridding interpolated topography data in 1.5 minutes,

which otherwise would take hours. • Allows fast interpolation of data between very large meshes

– User: Community Surface Dynamics Modeling System– Grids: 16 million triangle mesh to 16 million triangle mesh– Impact: ESMF bilinear regridding allows interpolation of data between two large

meshes that other packages could not handle for the surface dynamics community.

Summary

ESMF tools are increasingly critical infrastructure because:• Complex physical problems require multi-scale simulations, which in turn require

coupling of components on different kinds of grids• ESMF regridding has the reliability that production systems need, the flexibility and

features that research systems need• Parallel and scalable algorithms enable ESMF to handle new grids that are

increasingly huge, unstructured, and adaptive• ESMF has enabled groups like the Community Earth System Model to support

cutting edge science and take advantage of massively parallel computing architectures

From a user perspective:“use of the parallel ESMF offline regridding capability has reduced the time it takes to create CLM surface datasets from hours to minutes” - Mariana Vertenstein, NCAR Community Earth System Model

ESMF Strategic Plan

The previous Plan (2008-2010) had three main goals:1. To complete the ESMF product2. To develop partners and technologies that enable ESMF to

operate in heterogeneous environments3. To support the adoption of ESMF by applications

The draft Plan updates the first goal and adds two more:1. To maintain and evolve the ESMF product2. To develop partners and technologies to enable ESMF to

operate in heterogeneous environments3. To support the adoption of ESMF by application4. To develop a common model architecture in collaboration with

NUOPC5. To enable self-describing models and end-to-end workflows

Questions for the Board

• Are the high level goals appropriate?• Are there any major gaps?• Any errors, typos, omissions?• Any questions?