1/28/2021 – Development of Global Aerosol Forecast Model (GEFS-Aerosols) into NOAA’s Unified Forecast System (UFS)
Presenter: Li (Kate) Zhang; NOAA GSL & CIRES CU Boulder
Abstract: NOAA’s National Weather Service (NWS) is on its way to deploy various operational prediction applications using the Unified Forecast System, a community-based coupled, comprehensive Earth modeling system. A chemical component developed in collaboration among the Global Systems Laboratory (GSL), Chemical Science Laboratory (CSL), and Air Resource Laboratory (ARL) was coupled online with the FV3 Global Forecast System (FV3GFS) using the National Unified Operational Prediction Capability (NUOPC)-based NOAA Environmental Modeling System (NEMS) software. It replaced the previous operational global aerosol prediction NEMS GFS Aerosol Component (NGAC) system on September 24th, 2020 as an ensemble member in the operational Global Ensemble Forecast System version 12 (GEFSv12), named GEFS-Aerosols. This chemical component of atmospheric composition in GEFS-Aerosols is based on WRF-Chem with aerosol modules based on the Goddard Chemistry Aerosol Radiation and Transport model (GOCART). Compared to WRF-Chem and the GOCART aerosol modules, the major updates in GEFS-Aerosols include the FENGSHA dust scheme implemented and developed by ARL, the Blended Global Biomass Burning Emissions Product (GBBEPx V3) from NESDIS, which provides biomass burning emission and Fire Radiative Power (FRP) data with the biomass burning plume rise modules from WRF-Chem, the global anthropogenic emission inventories derived from the Community Emissions Data System (CEDS). Sub-grid scale tracer transport and deposition are handled inside the physics routines, including consistent implementation of positive definite tracer transport and wet scavenging in the Simplified Arakawa-Schubert (SAS) scheme. This study describes the details of GEFS-Aerosols model development and corresponding evaluation of the real-time and retrospective experiments using various observations from in situ measurement, satellite and aircraft data. GEFS-Aerosols have the capability to forecast the aerosols and the hazardous air quality caused by fire and dust events. The predictions demonstrate a substantial improvement for both composition and variability of aerosol distributions over those from the currently operational NGAC global aerosol prediction system.
2/11/2021 – Improving weather forecast skill and rainfall climatology of FV3GFS using machine learning
Presenter: Chris Bretherton, Senior Director, Vulcan Climate Modeling and Professor of Atmospheric Sciences, University of Washington
Abstract: Vulcan Climate Modeling (a small philanthropically-supported group in Seattle) and NOAA/GFDL are collaborating on a pilot project to use machine learning to develop a skillful corrective parameterization for a full-complexity global atmospheric model that helps it evolve more like a reference data set, which could be a reanalysis or a finer-grid global model. We have applied this approach to climate-oriented versions of FV3GFS with 100-200 km grids. Encouragingly, it significantly improves their 0-7 day weather forecasts and time-mean precipitation distribution, both in present and SST-perturbed climates.
Presenter: Guillaume Vernieres JCSDA
Abstract: The Joint Effort for Data Assimilation Integration (JEDI) is a generic data assimilation (DA) software infrastructure developed at the Joint Center for Satellite Data Assimilation (JCSDA) to facilitate and eventually expedite transfer between research and operation. The JCSDA is also spearheading a JEDI based marine DA effort for NOAA-EMC and NASA-GMAO, for the initialization of the ocean and sea ice component of the Unified Forecast System (UFS) and the Goddard Earth Observing System (GEOS). This presentation will briefly describe the scope of the marine DA effort at the JCSDA and focus on the specific application of the JEDI to the initialization of the marine components of the UFS for NOAA’s Next Generation Global Ocean Data Assimilation System (NG-GODAS). The NG-GODAS is a joint effort between NOAA-EMC, NOAA-CPC and the JCSDA, the first objective is to replace the aging GODAS based ocean sea ice reanalysis.
Presenter: Michael Barlage NOAA/NWS/NCEP/EMC, Land Team Lead
Abstract: A collaborative effort is currently underway to develop NOAA’s next generation Unified Forecast System (UFS) framework. Within the UFS, there are multiple major earth system components, including atmosphere, oceans, and land. UFS applications span local to global domains and predictive time scales from sub-hourly to seasonal predictions. These pose challenges and provide opportunities for the development and evaluation of its land components. This presentation will discuss on-going efforts in addressing and coordinating a land evaluation framework and land model physics advances for UFS. Models lacking proper coupled land-atmosphere behavior will underperform at all temporal and spatial scales, and will be prone to systematic biases in temperature, humidity and precipitation. The overarching initial UFS Land goal is to ensure equal, and achieve superior, model performance compared to the baseline UFS Land model. Progress will be presented on the community activities such as the UFS Land Working Group and the UFS Land Workshop. To facilitate UFS community engagement and accelerate R2O transition, a hierarchical testing approach is being developed that involves a spectrum of LSM-only simulations, a single-column coupled land-atmosphere modeling system, and coupled simulations both without and with a prognostic ocean. This approach is used to isolate and quantify the impacts of individual components before systematically increasing complexity and inherently introducing non-linear, difficult to track interactions. This provides a direct pathway for candidate models to diagnose problem areas in the model process chain, which enables identification of specific parameterizations, that are the source of poor model performance.
*The is no webinar recording available for this session
Presenter: Mark Govet, NOAA/OAR Global Systems Division
Abstract: Graphics Processing Units (GPUs) have become the dominant type of processor on the TOP500 list of most powerful HPC systems. NOAA’s Global Systems Laboratory has been exploring GPUs since 2010. As part of the work, GSL developed the Non-hydrostatic Icosahedral Model (NIM) and demonstrated performance-portability of the model on CPUs, GPUs and MIC processors with a single source code. Performance results reported in 2015 showed GPUs to be about 3 times faster than CPUs with minimal changes to the code. The design and parallelization strategy was successfully adopted by the MPAS model, achieving similar performance characteristics. The successful work with the NIM led the GSL team to apply the same approach to the FV3 dynamical core used in the UFS today. The two year effort (2016 – 2018) to port the FV3 to GPUs was not ultimately successful however. This presentation will describe and compare the NIM and FV3 parallelization efforts, and share lessons learned in developing performance-portable codes.
Presenter: Lucas Harris, NOAA/GFDL
Abstract: One of the major strengths of the GFDL Finite-Volume Cubed-Sphere Dynamical Core (FV3) is its efficiency and scalability on traditional CPUs. FV3 was designed to fit the MPI-OpenMP paradigm which has served HPC well for three decades. But this may be changing as emerging exascale systems are using manycore processors, especially GPUs. Even newer architectures like ARMs and FPGAs are on the horizon. Previous results to re-write FV3 for GPUs have shown 10x or more speedups in the entire core. More recent results have shown 50x GPU speedups to some modules within FV3. However, each new architecture requires re-writing of scientific codes for best performance. We are collaborating with Vulcan Inc, NASA Goddard, and university partners to port FV3 and the UFS into the GT4py Domain-Specific Language. By doing so the model can be re-compiled for target processors in a way so the code and data can be laid out for optimum performance. Progress and prospects for both traditional and DSL ports will be discussed.
Presenter: Tara Jensen, NCAR/RAL – DTC
Abstract: The Developmental Testbed Center (DTC), in collaboration with the National Oceanic and Atmospheric Administration (NOAA) and the Unified Forecast System’s Verification and Validation Cross-Cutting Team (UFS-V&V), a three-day workshop to identify key verification and validation metrics for UFS applications. The workshop was held remotely 22-24 February, 2021. Approximately 300 participants registered for this event from across the research and operational community.
The goal of this workshop was to identify and prioritize key metrics to apply during the evaluation of UFS research products and guiding their transition from research-to-operations (R2O). Because all UFS evaluation decisions affect a diverse set of users, workshop organizers encouraged members from government, academic, and private-sector organizations to participate in the workshop. The organizing committee used the outcome of the 2018 DTC Community Unified Forecast System Test Plan and Metrics Workshop to form the foundation of the workshop and to prepare and disseminate a series of three pre-workshop surveys to interested parties. The results of the surveys were used to prepare the discussion points of the breakout groups to streamline the metrics prioritization process.
During the workshop, the opening plenary was focused on providing background information to the participants. The participants then joined breakout groups to discuss how to apply the prioritized metrics to the full R2O development stages and gates. The breakout groups were stratified by time scales (Short Range, Medium Range, Sub-Seasonal and Seasonal). There was also a break-out group focused on how to define the research-to-operation gates and assign metrics to them.
This presentation will provide a preliminary report out of the findings from the workshop and discuss what the committee perceives are the next steps.
Presenter: Enrique Curchitser, Rutgers University
Abstract: Nearly 40% of the US population lives in what is considered coastal regions. The economic
services of the coastal regions such as shipping, tourism, fisheries, an industry constitute a
significant portion of the economy. US waters and exclusive economic zones are characterized by their diversity: from sub-tropical islands (e.g., Hawaii), eastern and western boundary currents (west and east coasts of the continental US), sub-Arctic (e.g., Bering Sea) to the Arctic shelf (e.g., Chukchi Sea). At the same time many of the coastal regions are susceptible to extreme weather events. Providing accurate forecasts to these regions is paramount for the safety of the population and for the economic vitality of coastal areas.
In coastal regions, improving the ocean component of the forecast system can lead both to
improved weather forecasts and estimates of impacts such as the ones resulting from storm surge and precipitation events. In this talk we describe the development and implementation of regional ocean modeling capabilities using the NOAA GFDL MOM6 ocean circulation model. We describe a strategy for a robust and holistic coastal and regional modeling capacity that leverages sustained NOAA investments in ocean model development at the Geophysical Fluid Dynamics Laboratory. The strategy is designed to provide a lasting and improved capability for fundamental process studies and weather and subseasonal-to-seasonal predictions in coastal systems.
Presenter: Glen Romine, NCAR
Abstract: NOAA’s Rapid Refresh Forecast System development includes plans to utilize high-resolution ensemble analysis methods to initialize forecasts. Traditionally, the design process for convection-permitting ensemble forecast systems has been largely ad hoc. More recently, teams that include collaborations between NCAR and NOAA have identified a simplified forecast system design approach that can use a single-physics modeling system based on the UFS short-range weather application. A key benefit of this approach includes a simplified code repository and workflow, but challenges remain to achieve both skillful and reliable predictions. Here, we present highlights from several activities contributing toward convection-permitting ensemble forecast system design, including: i) tools to trace sources of systematic model error to individual model components, such as specific physics packages; ii) approaches to reduce systematic errors in continuous cycling by increasing analysis system resolution along with large-scale analysis blending approaches; iii) diagnosing scale dependent perturbation growth rates within ensemble forecasts; and iv) post-processing methods to enhance the use and usability of ensemble forecasts. Collectively, these activities help guide forecasts system development within NOAA. During the presentation we will also touch on plans to extend systematic model error detection toward situation dependent errors within the global UFS medium range weather application.
Presenter: Joseph Olson, NOAA /GSL
Abstract: The Mellor–Yamada–Nakanishi–Niino (MYNN) Eddy Diffusivity-Mass Flux (EDMF) planetary boundary layer (PBL) scheme has been used in the operational High-Resolution Rapid Refresh (HRRR) since 2014. It has been developed primarily to help provide operational short-range convection-allowing forecasts for the contiguous United States (CONUS). This scheme has been selected for inclusion within the set of advanced experimental physical parameterizations in support of the Unified Forecast System (UFS). This expanded scope requires testing the MYNN-EDMF at much longer forecast lead times, at much longer time steps, over more regions of the Earth, and within a new dynamical core. The characterization of model errors specific to the MYNN-EDMF has been accomplished since its implementation into the FV3/CCPP framework, and subsequent development has progressed for improving clouds, solar radiation, and many other sensible weather variables for this expanded application scope. This presentation will overview key features of the MYNN-EDMF and highlight some recent development activities and associated performance examples. Successes and remaining challenges are identified for further research.
*There is no webinar recording for this session
Presenter: Amal EL Akkraoui; NASA Global Modeling and Assimilation Office (GMAO)
Abstract: Building on the success of the Modern-Era Retrospective analysis for Research and Applications, MERRA, and its successor MERRA-2 (released in 2009 and 2015 respectively), the NASA Global Modeling and Assimilation Office (GMAO) continues its incremental effort towards a decadal goal of an Integrated Earth System retrospective analysis, MERRA-3 (~2025), coupling components of the atmosphere, ocean, chemistry, land, and ice. While aspects of the coupled Goddard Earth Observing System (GEOS) model and data assimilation are currently under active development, an intermediate reanalysis featuring recent advances in the GEOS atmospheric component is planned as a stepping-stone towards MERRA-3. The GEOS-5 Retrospective analysis for the 21st Century, GEOS-R21C, is a hybrid 4D-EnVar atmospheric reanalysis that will cover the period 2000-onwards and feature the NASA’s Earth Observing System EOS and post-EOS satellite observations. Analysis fields from GEOS-R21C are planned to drive two other retrospective products, an off-line chemistry reanalysis (GEOS-R21C-Chem) and a high-resolution downscaled product for the polar regions (PolarMERRA). Both are also expected to aid in the ongoing coupling effort in advance of MERRA-3.
This talk will present an overview of GMAO retrospective analysis products and focus on the strategy for the two upcoming reanalyses GEOS-R21C and MERRA-3. Configuration details and preliminary results from prototype-R21C will be presented along with an overall review of new and revised features from MERRA-2.
6/17/2021 – Advancing Probabilistic Prediction of High-Impact Weather using Ensemble Reforecasts and Machine Learning
Presenter: Russ Schumacher, Colorado State University
Abstract: Convective weather hazards—excessive rainfall, tornadoes, large hail, and damaging winds—occur on spatial and temporal scales that are not well represented in numerical weather prediction model output. The predictability limit for these hazards is short, so reliable probabilistic forecasts are needed rather than deterministic predictions that will inevitably have large errors. To address these challenges, over the past several years we have developed a suite of probabilistic forecast systems, referred to as Colorado State University-Machine Learning Probabilities (CSU-MLP), that use the Global Ensemble Forecast System (GEFS) Reforecast datasets, historical observations of hazardous weather, and machine learning algorithms to generate skillful, reliable guidance that operational forecasters can use as a “first guess” when generating outlooks. Through close collaboration, testbed evaluations, and iterative improvements, CSU-MLP excessive rainfall guidance has been transitioned into operational use at the Weather Prediction Center. We have recently established a similar collaboration with the Storm Prediction Center with the goal of a similar operational transition for probabilistic severe weather guidance. This presentation will summarize some of the challenges and successes in operations-to-research-to-operations for incorporating machine learning into the operational forecast process, and plans for future work using the Unified Forecast System.
Presenter: Michael Toy, NOAA/GSL & CIRES/CU Boulder
Abstract:The GFS model physics has long included the parameterization of subgrid-scale drag forces and dissipative heating due to the vertical propagation and breaking of gravity waves forced by topography and non-stationary sources such as convection. Two additional parameterizations of orographic sources of subgrid drag have recently been implemented in the RAP/HRRR regional forecasting models developed by the Global Systems Laboratory (GSL). These schemes, which represent gravity wave breaking in highly stable boundary layers and turbulent orographic form drag, account for drag forces generated by horizontal topographic variations down to the ~1km scale. All of these schemes are now available in one “unified” physics module within the FV3GFS/CCPP framework. This presentation will provide a physical overview of the parameterizations. Preliminary test results with the global FV3GFS will be presented, and some possible future directions will be described.
8/12/2021 – Clouds in the Cloud: Developing NOAA’s Next Generation Convection-Allowing Prediction System with Cloud HPC
Presenter: Jacob Carley, NOAA/EMC
Abstract: The Rapid Refresh Forecast System (RRFS) is NOAA’s next generation convection-allowing ensemble and is currently planned for implementation in the last quarter of 2023. The planned configuration for RRFS is ambitious and its development requires substantial computational resources. The system will cover North America at 3 km grid-spacing with 65 vertical layers having 36 members deployed for its data assimilation system and a subset of 9 members for the free forecast. The RRFS will be hourly-updating and feature forecasts out to 18 hours every hour with extensions to 60 hours 4 times per day. To develop such a system requires computational resources beyond those available on currently deployed on-prem systems.
Cloud-based high performance computing (HPC) capabilities have grown over the past several years and it is apparent that cloud HPC has a role in research and development of our numerical weather prediction systems, and perhaps in operations. In this talk we will describe our recent and ongoing work in developing and testing, in real-time at NOAA testbeds, a prototype RRFS ensemble forecast system using cloud infrastructure.
Presenter: Paul Dirmeyer, George Mason University
Abstract: The Unified Forecast System (UFS) is a fully coupled Earth modeling system. It will be the system for NOAA‘s operational numerical weather prediction applications. UFS has been progressing through several prototype simulations (lately P5, P6, and P7) by improving model configurations on the way toward an operational version. The land surface model (LSM) beginning with P7 is the multi-parameterization version of Noah (Noah-MP), replacing the older Noah LSM. The main differences between these LSMs include the number of land tiles and snowpack layers, and the simulation of an aquifer below the bottom layer.
This study investigates land-atmosphere interactions in the UFS prototypes and their influence on model mean state bias, with particular attention to the impact of Noah-MP on the coupled model simulation. For the model evaluations, the modeled states related to land-atmosphere interactions for two July seasons (2012-2013) simulated in sub-prototype P7a are assessed against in situ observations as well as satellite-based products with attention to consistency in vegetated land cover and coupling regimes (energy vs. moisture limit controls on surface fluxes). There is an improvement in the simulation of net radiation and the terrestrial coupling index at the land surface without a large degradation on the other state variables (e.g., surface soil moisture). A remarked improvement in surface air temperature over the central US is related to the increased soil moisture sourced by precipitation. This suppresses sensible heat flux, corresponding to weakened land-atmosphere interaction, which corrects the warm bias by improving the coupling regime.
10/14/2021 – The Common Community Physics Page and its Role as an Enabler of Hierarchical System Development
Presenters: Ligia Bernardet (NOAA/GSL) & Mike Ek (NCAR)
The Common Community Physics Package (CCPP) is a collection of physical parameterizations and a framework that enables their use by host Earth System Models. The CCPP is now an integral part of the Unified Forecast System (UFS) and is used in developmental versions of its Subseasonal-to-Seasonal, Medium-Range Weather (MRW), Short-Range Weather (SRW), Hurricane, and Atmospheric Composition applications. The CCPP is designed to lower the bar for community involvement in physics testing and development through increased interoperability, improved documentation, and continuous support to the developers and the users.
The CCPP is used with a variety of host models, which increases the breadth of innovations that can benefit the UFS. The CCPP has been adopted in developmental versions of NCAR and Navy models, and is distributed with the CCPP Single Column Model (SCM), which offers a simpler and computationally inexpensive avenue for testing and developing atmospheric physics compared to a full three-dimensional model. Given its capabilities to be initialized and forced both by observational field campaign data and previous UFS simulations, the CCPP SCM is a key element in Hierarchical System Development (HSD), which includes testing small elements (e.g. physics schemes) of an Earth System Model (ESM) first in isolation and then with progressive coupling, all the way up to fully-coupled global system. Scientists can utilize the CCPP package to rapidly develop and test prototype code, as well as to tune and explore the parameter space of their schemes.
The CCPP has been developed as open source code, with public releases and support provided by the Developmental Testbed Center (DTC). It is also embedded in public releases of the CCPP SCM and of the UFS MRW and SRW applications. Other resources for CCPP users and developers are an online tutorial, archived materials from UFS training sessions, scientific and technical documentations, a catalog of case studies that highlights UFS MRW biases, and a community forum. In this seminar, we will provide an update of the CCPP and SCM latest developments, and review the plans for CCPP future development and transition to operations.
12/9/2021 – Model Diagnostics Task Force – A Walkthrough of the Technical Vision and the Diagnostics Package
Presenters: Aparna Radhakrishnan; Princeton University/NOAA/GFDL & Wenhao Dong; UCAR/GFDL
Climate and weather model development requires ongoing improvements in the representation of a growing list of physical processes. Process-oriented diagnostics (PODs) seek to give insight into the physical mechanisms needed to guide model development. The Model Diagnostics Task Force package (MDTF Diagnostics) is an open-source Python-based unified framework that runs process-oriented diagnostics (PODs) on weather and climate model data. The software package promotes the development and integration of diagnostics by subject matter experts across government, academia, and the private sector to improve the understanding of underlying processes in models under development by NOAA-GFDL and NCAR.
In this talk, we will provide an overview of the MDTF framework, encompassing both the technical and the scientific aspects. The technical overview will lay out the design goals and vision, encompassing key aspects such as Continuous Integration (CI), cloud computing and containerization to further strengthen collaborative development and foster community engagement. A science blurb centered around the Madden–Julian oscillation (MJO) process diagnostics and mesoscale convective systems based on GFDL simulations will be highlighted.