Desktop Visualization of Meteorological Data Using Paraview
Everyone is interested in the weather, yet few are aware of how complex atmospheric processes actually are. Simple questions such as “Will it rain today?” or “Will I need a jacket tonight?” may seem elementary, but to a professional meteorologist it is often difficult to accurately predict when and how weather patterns will evolve. The atmosphere changes through time-varying, non-linear physical relationships between energy, mass, and momentum in all spatial dimensions, and despite numerous surface and remote sensing data platforms, observations are few and far between. This makes the job of an operational meteorologist one of statistician, physicist, data analyst, and to some extent, gambler.
To improve the understanding of meteorological processes and features, it is necessary to more fully analyze weather patterns as they form and change. Physically-based numerical weather prediction (NWP) models have come a long way in diagnosing and predicting atmospheric patterns, and although they are not perfect, they do provide an excellent source of guidance for forecasters. However, despite having access to 3D gridded data sources from the NWP models, forecasters must continue to rely on 2D maps and cross sections at a limited number of locations. These 2D analysis procedures, normally based on constant pressure (isobaric) or constant entropy (isentropic) formulations, require extensive training and experience to use correctly.
It is logical that the inclusion of 3D NWP model data in the forecast process should entail the development of robust 3D visualization processes for meteorological analysis. However, although various development efforts have produced 3D visualization software for viewing gridded meteorological data, there remains a distinct lack of usable software for operational meteorologists who are limited by both time and computer resources.
With the advent of more powerful CPUs and GPUs in desktop computers, there is no reason that a robust meteorological forecasting procedure should not include 3D visualization capabilities, especially given the general availability of NWP model output. This does not mean that 3D analysis procedures are meant to replace existing 2D techniques, only that the current forecasting processes could be significantly improved through the inclusion of advanced 3D visualization and diagnostics. This article outlines some of the early results of a research project that brings 3D visualization of meteorological data to the desktop computer using ParaView, including some of the more common features that are used for meteorological analysis.
Initial Applications of Visualization – Student Seminar Course
After some early discussions regarding motivation and approaches to meteorological visualization using ParaView, especially regarding ease-of-use and training, a senior-level / graduate seminar course was organized in the Department of Geosciences at Mississippi State University in the Spring of 2009. This course involved training students how to set up and use the Weather Research and Forecasting (WRF) model for case study applications, and how to use ParaView for subsequent visualization and analysis. WRF is used in a variety of missions, from operational forecasting to research applications, and has quickly become an accepted tool for the research and operational meteorology community.
The students used both MS Windows and Mac laptops to run ParaView, while WRF was run on a small Linux cluster and the output data files were transferred to the students' laptops. There was a noticeable learning curve at the beginning of the course, especially with those having minimal computer experience. However, each of the groups quickly learned how to use ParaView to produce useful graphics and perform analysis on the data volume. As they became more experienced and comfortable with ParaView their enthusiasm grew and by the end of the semester there was substantial interest within the class and the meteorological faculty within the Department of Geosciences in the usefulness and future applications of 3D visualization in meteorology.
Upon completion of the course, three students obtained jobs or internships at National Weather Service (NWS) offices in Memphis, TN; Birmingham, AL; and Jackson, MS. Several of these offices have now become interested in using ParaView to visualize NWP model output, providing motivation for continued research into the use of ParaView for 3D visualization in meteorological applications.
Visualization Case Studies
The seminar class was a definite success and served as a spring board for continued work. Upon successful completion of the seminar course, the authors were convinced that 3D visualization of NWP model data using ParaView was a viable option for operational meteorologists. In response, several case studies were selected to provide a more detailed look at the use of WRF and ParaView for analysis of atmospheric processes. This step was necessary in showing that the visualization and analysis tools within ParaView were applicable to meteorological diagnostics of large-scale, complex weather systems.
The first case study focuses on a tropical cyclone, Hurricane Wilma, while the second case study deals with an extratropical cyclone, the 1993 “Storm of the Century”. Each of these cases provides an opportunity to asses various features related to storm structure, development, and evolution, which is critical in diagnosing and forecasting meteorological processes associated with the different types of events. Results give indications regarding the usefulness and strength of 3D visualization using ParaView in an operational setting.
Extra-tropical Cyclone – 1993 “Storm of the Century”
Extra-tropical (mid-latitude) cyclones are dynamically driven systems that arise due to horizontal temperature gradients between the tropics and the poles. They are largely driven by velocity patterns in the upper levels of the troposphere; specifically, the upper-level feature called the jet stream. Extra-tropical cyclones serve the purpose of transporting excess heat energy to the higher latitudes. They are normally associated with warm and cold frontal processes, and are often termed “cold-core” systems due to the existence of a pool of cold air wrapping around the center of circulation.
An extra-tropical cyclone traversed a large portion of the U.S. during March 12-13, 1993, aptly called the “Storm of the Century” because of its extraordinary size and intensity as well as its widespread impact. The storm initially developed off of a closed low pressure system of the Gulf of Mexico, where it rapidly intensified as it moved northward along the Atlantic coast. Record low temperatures and pressures were set along the east cost of the U.S., along with extreme winds and snowfall. Additionally, the emergence of a squall line caused the genesis of severe thunderstorms and tornadoes in Florida.
The gridded data for use in ParaView were obtained from a WRF simulation of the 1993 extra-tropical cyclone on a computational grid with 12 km spatial resolution and 31 vertical levels. This resolution allowed for assessment of major synoptic-scale processes, such as warm and cold fronts, while maintaining an acceptable run-time. The simulation time was March 12, 1993, at 00 UTC to March 13, 1993, at 23 UTC, producing a 48-hour forecast time with hourly output. Initial and laterally boundary conditions were provided by the North American Regional Reanalysis (NARR) dataset, which provided 32-km spatial and 3-hourly temporal resolution input data.
Perhaps the most important and critical step in analyzing extratropical cyclones is frontal analysis, which involves defining the location and extent of the warm and cold fronts. Although many atmospheric variables such as temperature, humidity, and wind are all related to frontal passage, fronts are commonly defined based on temperature boundaries. Using the contour filter within ParaView, it is easy to display an isothermal surface (surface of constant temperature) to define the location and vertical extent of these boundaries. Coloring the surfaces by relative humidity adds an additional dimension, allowing for the assessment of moisture transport within the circulation (Figure 1a). Additionally, isosurfaces of relative humidity allow for a rapid assessment of cloud extent and structure within the system (Figure 1b), although these features are limited by the spatial resolution of the model domain. Diagnosis of frontal features is often difficult using 2D analysis techniques; however, 3D analysis greatly simplifies this issue by showing the physical shape of the front, substantially decreasing the time needed to identify associated processes in an operational environment.
Figure 1. (a) 260K isothermal surface colored by relative humidity (%), looking south over Minnesota towards the Gulf of Mexico and (b) 95% relative humidity isosurface colored by height (meters) for March 13, 1993 at 12 UTC.
ParaView’s stream tracer filter provides a critical analysis tool for diagnosing a system’s circulation patterns. Figure 2 provides an example of the circulation around the 1993 extratropical cyclone, indicating both the shape and extent of the rotation. Based on a seed point just above the Earth’s surface at the apex of the maximum moisture transport, indicated by the isosurface of constant dew points coming off the Gulf of Mexico, a northwesterly tilted circulation core can clearly be seen. This characteristic, along with the strong counter-clockwise rotation to the east of the center of circulation, provides rapid and effective details regarding the intensity and future propagation of the system without the need for various 2D maps at select horizontal layers.
Figure 2: 278K dew point surface colored by height (meters) with stream tracers colored by vorticity ( 10-5 rad s-1) for March 13, 1993 at 12 UTC. The view is to the south, looking over Wisconsin towards the Gulf of Mexico.
Tropical Cyclone – Hurricane Wilma
Tropical cyclones are similar to their extratropical counterparts in that they’re cyclonically rotating systems based on a temperature gradient. However, the basic difference is that the tropical systems feed off of a vertical temperature gradient between the surface and the upper atmosphere, meaning that they are maintained by surface heat fluxes, vertical mass and energy transport. Since the systems arise due to energy building up at the surface, they are referred to as warm-core systems and are not associated with frontal features. Instead, they are characterized as having a large cloud shield with a distinct eye and eye wall at the center of the circulation; therefore, the methods of analysis are focused on temperature and moisture patterns around the storm’s core.
Despite the easy identification of tropical cyclones using satellite platforms, they are extremely difficult to diagnose and predict due to the complex vertical nature of the storm structure. This is compounded by the lack of observations within the storm environment, especially in the middle and upper troposphere where the energy and momentum transfers are most intense. As a result, NWP model data and 3D analysis potentially provide an invaluable tool in assessing storm structure and evolution, as well as identifying potential threats associated with extreme winds and precipitation.
The 2005 hurricane season in the Atlantic basin was record breaking, producing several of the strongest and costliest storms ever recorded in the region, including Hurricanes Katrina, Rita, and Wilma. Hurricane Wilma had the lowest recorded pressure (882 hectopascals) and highest 1-minute sustained winds (295 km hr-1) of any Atlantic hurricane and was especially noteworthy because of its intensity and eventual transition into an extratropical cyclone.
For the Hurricane Wilma case study, a 48-hour WRF simulation was done at a 12 km horizontal resolution with 40 vertical levels, providing hourly 3D grids from 00 UTC on October 25 to 00 UTC on October 27, 2005. As with the extra-tropical case study, the initial and lateral boundary conditions were obtained from the NARR dataset.
Figure 3: 260 (top) – 300 K (bottom) isothermal surfaces (surfaces are in 10° increments), colored by relative humidity (%). Image represents data from October 25, 2005 at 12 UTC.
As with extra-tropical circulations, use of the contour filter and scalar shading provides rapid assessment of the temperature and moisture conditions within the data volume. However, since tropical cyclones are based on vertical processes, analysis includes the generation of numerous isothermal layers to define the structure through the atmospheric column. Figure 3 shows how these layers, colored by relative humidity, can be used to quickly identify the shape and structure of the circulation. Specifically, the warm core east of the Carolinas is shown by the vertical bulge of the isothermal surfaces, while the circulation and transport of moisture in the middle and upper levels is shown by the wrapping of higher relative humidity values to the north and west of the storm’s core. Additionally, the influence of the southern Appalachian Mountains can readily be seen by the vertical enhancement of the isothermal layers across western North Carolina.
Figure 4: Vertical slice of Hurricane Wilma on October 25, 2005 at 12 UTC, colored by temperature, showing (a) temperature contours (K) and (b) wind vector glyphs colored by wind speed magnitude (m s-1).
One of the most important points of analysis for tropical cyclones is the identification of the location and extent of the warm core. Using the slice filter within ParaView, it is relatively straightforward to locate the area of maximum energy transport from the surface. Figure 4a shows the location of the warm core at 12 UTC on October 25, 2005, with temperature contours and shading used to more clearly indicate the vertical temperature gradients. From this figure it is easy to quantify the depth and horizontal dimensions of the warm core, as well as the shape and modification relative to other atmospheric and surface features. Figure 4b, which shows wind vectors on the same vertical slice, provides a rapid identification of the vertical motion within the warm core, allowing for analysis of the vertical extent of the convection within the system. Additionally, by shading and shaping the vectors based on wind magnitude, the location and magnitude of large-scale atmospheric features, such as the jet stream (indicated by the high wind velocities in the upper levels) is easily identified.
Discussion
The availability of 3D gridded datasets from NWP models provides an invaluable asset to operational meteorologists, and 3D visualization techniques can maximize the usefulness of these data for diagnostic and forecasting applications. Although 3D visualization of NWP model output is not a new concept, it has usually been limited to higher-end workstations and PCs. This article provides a quick overview of how ParaView can be used to perform basic 3D analysis of gridded weather data on a desktop computer, significantly improving the effectiveness and efficiency of the assessment and quantification of storm structures and processes.
Two case studies involving an extra-tropical cyclone (1993 “Storm of the Century”) and a tropical cyclone (Hurricane Wilma) were presented here, along with examples of how 3D visualization techniques and ParaView filters can be used to analyze event-specific features. For the extra-tropical cyclone analysis, ParaView was able to allow for efficient assessment of velocity profiles within the data volume using stream tracer filters, providing information on the three dimensional dynamic structure of the 1993 cyclone. Additionally, frontal analysis was more effectively performed by looking at the shape and pattern of temperature surfaces with moisture included as a scalar shading. These simple visualization techniques could potentially enhance the understanding and prediction of extra-tropical cyclone genesis and evolution by providing direct visual interpretation of middle- and upper-level dynamics and related processes.
Analysis of Hurricane Wilma using 3D visualization techniques allowed for rapid quantification of the location and extent of the warm core and associated vertical motion. This gives information regarding the vertical and horizontal energy gradients and fluxes within the storm’s center, greatly improving diagnostics of storm intensity and structure. Using the slice and contour filters within ParaView it was relatively straightforward to identify features related to convection and large-scale wind fields, which would be useful for tropical cyclone forecasting and estimation of storm impacts. With higher resolution simulations, it is entirely possible to obtain detail regarding eyewall processes within the atmospheric column, enhancing our understanding of the role of energy and mass fluxes in processes such as eyewall circulation and eyewall replacement cycles.
It should be noted that the 3D visualization environment within ParaView allows for easy manipulation of the users viewpoint; therefore, the 2D nature of the figures presented in this article here are just a hint of what is actually possible. In addition, beyond the cases presented here, there are numerous applications of 3D visualization techniques in meteorology, from the scale of a thunderstorm to global wave patterns. By providing meteorologists the tools to visualize and analyze weather patterns and processes quickly and easily at their desktop computers, the future of weather analysis and forecasting may quickly change.
Conclusion
The 12 students in the seminar class were a great group. They worked through their initial frustration and the learning curve, and we were delighted to see what they were able to accomplish in one semester. Additionally, this experience encouraged us to continue visualizing WRF output with ParaView, and that is a principle area of research we are currently pursuing.
Jamie Dyer is an Assistant Professor in the Department of Geosciences at Mississippi State University. He received his Ph.D. from the University of Georgia where he specialized in cryospheric hydroclimatology and statistical analysis of large-scale climate datasets. His current research interests include analysis of local and regional weather processes using numerical weather prediction models and 3D visualization methods, including land surface/atmosphere interactions and related precipitation processes.
Philip Amburn is a Research Associate Professor in the Geosystems Research Institute at Mississippi State University. His research work focuses on scientific visualization of the output of numerical models and simulations. He is a retired Lt Col from the USAF and received his Ph.D. is Computer Science from the University of North Carolina at Chapel Hill.
This work is licensed under a Creative Commons Attribution-NoDerivs 3.0 Unported License. Additional information on this license can be found here.