2017 STAR Seminars

Starting 2014, ECMWF has embarked on a research program on HPC Scalability, aiming to achieve Exascale numerical weather prediction systems by 2025.

ECMWF operational forecast generates massive amounts of I/O in short bursts, accumulating to tens of TB in hourly windows. From this output, millions of user-defined daily products are generated by a complex chain of transformations and regridding operators and finally disseminated to member states and commercial clients.

These products are processed from the raw output of the IFS model, within the time critical path and under strict delivery schedule. Upcoming resolution increases and growing popularity will increase both the size and number of these products. Based on expected model resolution increases, by 2020 we estimate the operational model will output over 100 TB/day and need to archive over 400 TB/day. Given that the I/O workload is already one of the strongest bottlenecks in ECMWF's workflow, this is one of the main challenges to reach Exascale NWP.

We present a new software stack that ECMWF is developing to tackle these future challenges in the scalability of model I/O and product generation, and reworking its operational workflows to adapt to forthcoming I/O technologies.

In particular, we will present the adaptation of IFS I/O server to the use of NVRAM technologies as a way to buffer large amounts of forecast outputs en route to the product generation and archival systems, thus minimizing file-system I/O within the operational critical path and collocating post-processing with model computation.

About the Speaker:

Dr. Tiago Quintino is the Team Leader for Scalability at the ECMWF's Development Section, in Reading, United Kingdom. His career spans 17 years researching numerical algorithms and developing high performance scientific software in the areas of Aerospace and Numerical Weather Prediction. Lately he is focusing on scalable data handling algorithms for generation of meteorological forecast products, optimizing their workloads and I/O of massive data-sets.

Operational since October 2004, the National Snow Analysis (NSA) will complete its 13th year of operations in 2017. The NSA is a collection of operational products and services derived primarily from the Snow Data Assimilation System (SNODAS). SNODAS combines a mass and energy balance model of the surface snowpack over the CONUS and southern Canada, driven by numerical weather prediction (NWP) model analyses and forecasts, with an assimilation system that updates SNODAS states using observations collected by surface stations and surveyors, satellites, and aircraft (via NOAA’s Airborne Snow Survey program). Clients of the NSA include NWS River Forecast Centers and other government agencies, emergency managers, policymakers, and the general public. The NSA provides clients with near-real-time raster data sets, imagery, basin averaged snowpack information, and a wide variety of other products available via an interactive web interface. Given its years of operations, the NSA now performs routine comparisons of SNODAS states with period-of- record (currently consisting of water years 2005-2016) normals, providing valuable context for real-time analyses. This presentation will provide an overview of the NSA and SNODAS, with some highlights from the winter of 2016-17.

About Mr. Fall

Greg Fall joined the National Hydrologic Remote Sensing Center (now the OWP-Chanhassen, MN) in 1999 and contributed to the design, development, and implementation of the Snow Data Assimilation System (SNODAS). He is currently the lead for the Office of Water Prediction's (OWP) National Snow Analysis function, which encompasses SNODAS and related products and services. Greg also serves as lead for the National Water Model Forcing Data Improvement Project and the Experimental Gridded Snowfall Analysis Project at OWP.

This presentation gives an overview of a unique multi-year set of ship-based atmospheric data acquired over open oceans as part of ongoing National Oceanic and Atmospheric Administration (NOAA) Aerosols and Ocean Science Expedition (AEROSE) field campaigns. Following the original 2004 campaign onboard the NOAA Ship Ronald H. Brown, AEROSE has operated on a near-yearly basis since 2006 in collaboration with the Howard University NOAA Center for Atmospheric Sciences (NCAS) and the NOAA Prediction and Research Moored Array in the Tropical Atlantic (PIRATA) Northeast Extension (PNE). In this presentation, attention is given to atmospheric soundings of ozone, temperature and water vapor obtained from dedicated ozonesondes and radiosondes launched to coincide with low earth orbit environmental satellite overpasses (viz., the Suomi National Polar-orbiting Partnership (SNPP), MetOp-A,-B and the NASA A-Train). Data from the AEROSE campaigns are unique in their range of marine meteorological phenomena germane to the satellite missions in question, including dust and smoke outflows from Africa, the Saharan air layer (SAL), atmospheric rivers (ARs), Hadley cells and distribution of tropical water vapor, and atmospheric ozone. The multi-year AEROSE sounding data have been invaluable as correlative data for validation of environmental data records (EDRs) derived from the Joint Polar Satellite System (JPSS) NOAA-Unique Combined Atmospheric Processing System (NUCAPS) as well as the NOAA Geosynchronous Operational Environmental Satellite (GOES-R/16), as well as numerous other science applications. A summary of these data, along with an overview of some important science research highlights, including meteorological phenomena of general interest, will be presented.

Highly migratory species are inherently difficult to manage as they cross human-imposed jurisdictional boundaries in the open seas. Top predators face multiple human-induced threats such as ship- strike risk and non-target catch (bycatch) in fisheries. Current management approaches use large-scale seasonal closures to avoid bycatch of highly migratory predators, but here we explore a dynamic ocean management approach that tracks ocean features in space and time. Such targeted management approaches require an understanding of how distribution and abundance varies with the oceanic environment through time. Given these data are often sparse and are collected using multiple platforms, e.g. fisheries catch, fisheries independent surveys, and telemetry studies, an approach that synthesizes across data type would provide a more holistic understanding than a single approach alone. Here we explore the California Drift Gillnet fishery that targets swordfish, thresher shark, and mako shark, but also can catch a number of species as bycatch including sea lions, sea turtles, and blue sharks. While still in the formative stage, this tool uses habitat models and risk weightings to estimate catch / bycatch ratios as a function of management concern in near time. We have explored the tool in two years, 2012 and 2015 an average year and an El Nino year respectively, to examine how predicted patterns in catch and bycatch change. These approaches could be applied to other migratory species for which telemetry, catch, or survey data are available, and emphasizes the utility in integrating multiple data types for marine conservation and management.

Seasonal forecasting has great scope for use in marine applications, particularly those with a management focus. Seasonal forecasts from dynamical ocean-atmosphere models of high risk conditions in marine ecosystems can be very useful tools for managers, allowing for proactive management responses. The Australian Bureau of Meteorology's seasonal forecast model POAMA currently produces operational real-time global forecasts of sea surface temperatures, with tailored outlooks produced for coral reef, aquaculture and wild fisheries management in Australian waters.

Operational realtime seasonal forecasts for coral bleaching risk on the Great Barrier Reef predict warm conditions that may lead to coral bleaching several months in advance, and play an important role in the Great Barrier Reef Marine Park Authority's Early Warning System. Early warnings of potential bleaching risk can assist reef managers to prepare for the likelihood of an event, focusing resources, briefing stakeholders and increasing awareness of bleaching onset. In marine farming and fishing operations in Australia, seasonal forecasting is being used to reduce uncertainty and manage business risks. Further, habitat distribution forecasts can be generated by combining these environmental forecasts with biological habitat preference data, providing industry with species- specific information. POAMA will be upgraded to the new higher resolution ACCESS-S seasonal prediction system in 2017, in collaboration with the UK Met Office.

Dynamical forecasts potentially offer improved performance relative to statistical forecasts, particularly given baseline shifts in the environment due to climate change. Seasonal forecasts are most useful when management options are available for implementation in response to the forecasts. Improved management of marine resources, with the assistance of such forecast tools, is likely to enhance future planning, industry resilience and adaptive capacity under climate change.

About the Speaker: Dr. Claire Spillman holds a PhD in Environmental Engineering and joint BEng/BSc degrees in Environmental Engineering (Hons) and Chemistry from the University of Western Australia. Her postgraduate work investigated impacts of estuarine circulation and oceanic inputs on aquaculture production using high resolution hydrodynamic-ecological modelling.

Dr. Spillman is a senior research scientist at the Bureau of Meteorology, Australia. Her current research is primarily focused on dynamical seasonal forecasting in marine applications, particularly for coral reef and fisheries management. Applications include predictions for Great Barrier Reef coral bleaching risk, Australian commercial fisheries and aquaculture on multiweek to seasonal timescales.

The attached PDF contains single page summaries of all the talks presented by NOAA/NESDIS/STAR researchers at the 2017 AMS Meetings.