Advance data-centric research

CISL has a large portfolio of data-centric research activities. This section presents some highlights from FY2019 that demonstrate the diversity of these activities. They include advancements in the development of data assimilation tools that combine observational and model forecasts to produce large ensemble analysis and forecast data sets; advances in visualization and augmented reality for geophysical data; and methods and science for translating the influence of global processes that affect our climate into specific regional and local impacts.

Data assimilation research

The Data Assimilation Research Testbed (DART) is a community framework for ensemble data assimilation research and applications. In addition to developing advanced methods for data assimilation, the DART team collaborates with modelers and observationalists to develop data assimilation capabilities for new models and observations.

Ensemble assimilation methods are well-established for quantities with Gaussian distributed errors, but non-Gaussian assimilation remains a research frontier. Challenges arise for bounded variables like the concentration of an atmospheric pollutant or the streamflow in a hydrologic model. Two novel approaches are in development for use with DART. The first uses anamorphosis, a mapping of a quantity to make it more Gaussian; a simple example is a log-transform. A second example extends a method called the rank histogram filter to multivariate applications in which an ensemble is approximated by a piecewise continuous probability distribution function. Single-threaded implementations of both algorithms have been developed in DART and work is under way to complete scalable versions that can be applied to large models. These algorithms should be particularly effective in improving analyses and forecasts of bounded quantities by eliminating infeasible ensemble members.

A multi-decade atmospheric reanalysis using the new CAM6 model also is under way. The DART algorithms were carefully tuned to produce both high-quality ensemble reanalyses of the atmosphere and ensemble forcing files that are needed for assimilation with other CESM components such as POP, CTSM, and CICE.

Visualization research – VAPOR

CISL continued its efforts to devise new techniques for advanced scientific visualization. This year a novel GPU-based volume rendering engine was developed and incorporated into VAPOR – the CISL Visualization and Analysis Platform for Ocean, Atmosphere, and Solar Researchers. The engine is capable of supporting layered, curvilinear grids that are found widely in the geosciences, and capable of achieving interactive rendering rates on those grids using commodity hardware, a feat that was not possible previously.

direct volume rendering
Figure 1. A direct volume rendering of the vorticity field of a numerically simulated tornado. The volume rendering engine is believed to be the first of its kind to support curvilinear grids using a platform-portable GPU implementation.

The implementation is based on OpenGL, making it platform-portable. This work was performed in collaboration with the University of Wisconsin. 

Visualization research – augmented reality

CISL continued to explore novel means for communicating Earth system science to a wide variety of audiences. In FY2019, a mobile augmented reality (AR) application was developed to help educate students about the complex topic of map projections. The application scans a student's hand-colored picture of a global data set and converts the scanned image to an AR interactive map that can be viewed from a cell phone (see Figure 2). The work was presented at the 2018 AGU Fall Meeting and has been used in a variety of other education and outreach activities.

MapWarpAR application
Figure 2. A screenshot of the MapWarpAR application demonstrating the interactive mapping activity in augmented reality.

CISL also continued research efforts aimed at improving scientists’ ability to explore and understand complex data. A newly developed scientific animation playback system facilitates comparisons between multiple visualizations, helping to identify relationships and compare patterns. An article describing this work is being prepared for submission to the Bulletin of the American Meteorological Society.

CISL research staff also collaborated with university faculty to investigate the application of advanced flow visualization techniques to numerical simulations of cloud physics and tornado genesis. A paper describing preliminary efforts with the former was accepted by SC19.

Regional and local climate impact

CISL's Regional Integrated Science Collective (RISC) made further progress in FY2019 toward several major scientific advances in simulated climate model data production, archiving of data, bias correction, and regional process-level analysis through grants from NSF, the Department of Energy, and the Department of Defense. Having completed a set of simulations for the North American Coordinated Regional Climate Downscaling Experiment (NA-CORDEX) program the previous year, RISC extended the NA-CORDEX Data Archive in FY2019 to include important variables beyond temperature and precipitation, adding surface pressure, specific humidity, winds, incoming solar radiation, and orography.

RISC also published a version of the data set using a quantile-mapping, bias-correction approach (Cannon’s MBCn algorithm) that adjusts the statistical distribution of the model output to match that of observations. Research comparing multiple bias-correction methods has found that quantile-mapping approaches have the best all-around performance for a variety of uses, which makes them well-suited for the broad community of researchers and decision makers the archive is intended to support. The MBCn algorithm is a multivariate method that adjusts the joint distributions of the variables as well as their individual marginal distributions.

In addition, RISC published on the NA-CORDEX web site a collection of figures to accompany the data set. The NA-CORDEX Visualization Collection is a set of maps that summarize average temperature and precipitation data in the data set. It includes maps of seasonal and annual average climatology for different 30- and 50-year periods as well as maps of the change between periods. The figures were published via a publicly shared Google Drive folder linked from the project web site. This proved to be a low-cost solution that provides robust, reliable access to the files using a familiar, full-featured user interface and powerful search capabilities.

NA-CORDEX image
Figure 3: A sample image from the NA-CORDEX Visualization Collection, showing change in the 30-year average winter temperature between the end of the 20th and 21st centuries for the RegCM4 RCM driven by the HadGEM2-ES GCM.

To promote the treatment of data products as citable research objects, DOIs have been assigned to both the NA-CORDEX Data Archive (https://doi.org/10.5065/D6SJ1JCH) and the NA-CORDEX Visualization Collections (https://doi.org/10.5065/90ZF-H771). 

These efforts illustrate CISL’s approach to the grand challenge for Earth system science: translating the influence of global processes that affect our climate into specific regional and local impacts. CISL’s research combines knowledge of Earth system models, downscaling methods, scientific workflows for large data sets, statistics, and the needs and constraints of regional and local stakeholders. This effort integrates CISL expertise in data science and impact assessment with the goal of transferring climate science into useful products for decision making in adaptation research and risk analysis.