Computing in Science & Engineering

            

RXUQDO &RPSXWLQ LQ 6FLHQFH DQG QLQHHULQ

Fo

!DQXVFULSW ," &L6(#$%&$#%&#%%%'()& !DQXVFULSW 7\SH )HJXODU

rR

"DWH 6X*PLWWHG *\ W+H Q,D $XW+RU

ev

&RPSOHWH -LVW RI $XW+RUV 6+HQ. /R#:HQ0 1$6$. 26)&0 3!&4. (66,& 1HOVRQ. /URQ0 1$6$,$)&. 7DR. :HL#5XR0 1$6$,26)&. -LQ. 6X+#-DQJ0 1RUW+ &DUROLQD $77 6WDWH 3QLYHUVLW\.

w

ie

((%(H 5QR8OHGJH DQG GDWD HQJLQHHULQJ WRROV DQG WHF+QL9XHV : ((% 2HQHUDO : ( "DWD. "(&(; &RQFXUUHQW 4URJUDPPLQJ : "(& 5H\8RUGV 4URJUDPPLQJ 7HF+QL9XHV : " 6RIW8DUH,6RIW8DUH (QJLQHHULQJ. YLVXDOL]DWLRQ. +XUULFDQH PRGHOLQJ. FOLPDWH PRGHOLQJ

ly

On

Page 15 of 62

Advanced Visualizations of Scale Interactions of Tropical Cyclone Formation and Tropical Waves Bo-Wen Shen1,2([email protected]), Bron Nelson3 , Wei-Kuo Tao2, and Yuh-Lang Lin4 1

University of Maryland, College Park; 2NASA Goddard Space Flight Center; 3

NASA Ames Research Center; 4North Carolina A&T University

(Submitted to the IEEE CiSE for publication in Jan. 2012; Revisied in Mar. 2012)

Fo

Abstract

rR

One of the current challenges in tropical cyclone (TC)1 research is to determine ways to improve our understanding of TC formation and intensification, which involves the TC’s interactions with environmental flows. The newly developed Coupled NASA Advanced Global

ev

Multiscale Modeling and Concurrent Visualization System (CAMVis) shows potential for such studies. With the goal of improving short-term hurricane forecasts and hurricane climate

ie

simulations, a 3D streamline package was recently developed and integrated into CAMVis to

w

provide insight into the multi-scale interactions between TC formation and tropical waves. We discuss how 3D visualizations can help illustrate the predictive relationship between TC

On

formation and an African Easterly Wave or a mixed Rossby gravity wave. We then suggest that accurate simulations of large-scale waves can help narrow down the uncertainties of TC genesis prediction, and that accurate representation of both large-scale (tropical wave) processes and

ly

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Computing in Science & Engineering

small scale (e.g., precipitation) processes are crucial for extending the lead time of hurricane genesis

prediction.

1

Depending on their location, TCs are referred to by other names, such as hurricanes (in the Atlantic region), typhoons (in the West Pacific region), tropical storms, cyclonic storms, and tropical depressions.

1

Computing in Science & Engineering

1. Introduction During the past ten years, statistics have shown that hurricane is the deadliest weather event in the U.S. (Figure 1). Consequently, there is an urgent need to improve both short- and longterm hurricane forecasts. At the same time, studies in TC inter-annual variability and the impact of climate change (such as global warming) on TCs have received increasing attention (e.g., Bengtsson et. al., 2007), particularly due to the fact that 2004 and 2005 were the most active hurricane seasons in the Atlantic, while 2006 was not as active as predicted. However, while TC track forecasts have been steadily improving over the past several decades, intensity and genesis

Fo

forecasts have lagged behind (Figure 2). One of the major challenges in TC genesis prediction is the accurate simulation of complex

rR

interactions across a wide range of scales, from the large-scale environment (deterministic), to mesoscale flows, down to convective-scale motions (stochastic). Therefore, improving intensity prediction relies on the accurate representation of a TC’s structure and its interactions with both

ev

large-scale environmental processes and small-scale moist processes (such as convection and surface fluxes exchanges). Figure 3 shows the major features of a TC, which include (1) an eye,

ie

(2) an eyewall, (3) an elevated warm-core, (4) low-level inflow with counterclockwise

w

circulation and (5) upper-level outflow with clockwise circulation. The TC “eye” refers to a region in the center of the TC where winds are light and skies are clear to partly cloudy; and the

On

“eyewall” refers to a wall of dense thunderstorms that surrounds the eye of a hurricane. An elevated warm-core usually appears at upper levels (about 200–300 hPa) where the cyclone’s temperature is warmer at its center than at its periphery.

ly

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 16 of 62

General circulation models (GCMs) have been used to study TC genesis statistics and interannual variability, but their insufficient grid spacing and unsophisticated physics parameterizations are known limiting factors in simulating a TC’s structure and its interaction with environmental flows. Recent advances in high-resolution global modeling and supercomputing have made it possible to mitigate some of the aforementioned issues.

2

Page 17 of 62

rR

Fo Figure 1: Statistics of fatalities caused by extreme weather events during the past 10 years (light blue) and 30 years (yellow). The 10-year statistics show that the hurricane is the deadliest

ev

event. Data source: http://www.nws.noaa.gov/os/hazstats.shtml

w

ie ly

On

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Computing in Science & Engineering

Figure 2: The progress of hurricane forecasts by the National Hurricane Center. Horizontal axis indicates the years from 1989 to 2010; the vertical axis represents forecast errors. Lines with different color show different forecast intervals, from 24 hours (red) to 120 hours (blue). During the past 20 years, track forecasts have been steadily improving (left panel), but intensity forecasts

have

lagged

behind

(right

http://www.nhc.noaa.gov/verification/verify5.shtml

3

panel).

Data

source:

Computing in Science & Engineering

Fo

Figure 3: Major characteristics of a hurricane, including (1) the eye: a region in the center of a TC where the winds are light and skies are clear to partly cloudy; (2) the eyewall: a wall of

rR

dense thunderstorms that surrounds the TC’s eye; (3) an elevated warm-core (not shown) where the cyclone’s temperature is warmer at its center than at its periphery; (4) low-level counterclockwise circulation; (5) upper-level clockwise circulation. (Courtesy of the COMET

ev

program at the National Center for Atmospheric Research.)

ie

When NASA’s Columbia supercomputer became operational in late 2004, its computing power enabled the deployment of the global mesoscale model (GMM) at very high resolution

w

(i.e., 0.25o, 0.125o, and 0.08o), which resulted in remarkable hurricane forecasts during the very active 2004 and 2005 Atlantic hurricane seasons (Atlas et al., 2005; Shen et al., 2006). The

On

GMM’s ability to forecast hurricane intensity was first demonstrated with Hurricane Katrina (2005), which was the sixth most intense hurricane in the Atlantic, and which devastated New

ly

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 18 of 62

Orleans and the surrounding Gulf Coast region. While both 0.25o and 0.125o runs showed remarkable track forecasts, the higher-resolution runs produced more realistic intensity forecasts. This suggested that the better intensity forecasts are due to the finer resolution, which became sufficient to resolve the near-eye wind distribution. Further analysis of the 96h 0.125o simulations for Katrina showed realistic vertical structures of the storm, including maximum winds near the top of the boundary layer, a narrow eyewall, and an elevated warm core. The performance of the 0.08o model for Hurricane Rita (2005) was documented in Biswas et al. (2007), which showed improved track and intensity forecasts with increasing resolution (i.e., from 0.25o to 0.125o to 0.08o).

4

Page 19 of 62

Since 2007, in addition to the simulations of intense hurricane structures, we have investigated the interaction of TC formation and tropical waves to understand the model’s performance in predicting the evolution of the TC intensity. Tropical wave theory was first developed by Matsuno (1966) using the linearized equations of motion valid for the tropical regions. He obtained five different wave modes which include eastward and westward propagating inertio-gravity waves, eastward propagating Kelvin waves, westward propagating Rossby waves, and mixed Rossby gravity (MRG) waves. The last one is a special wave mode that behaves like a Rossby wave for low zonal wave numbers and an inertio-gravity wave for high zonal wavenumbers. Since then, the role of theses idealized waves in daily weather and their

Fo

impact on TC activities have been been examined. For example, Frank and Roundy (2006) conducted a comprehensive observationally-based study to search for the predictive relationship

rR

between tropical waves and TC formation. They found a strong relationship between TC formation and enhanced activities in equatorial Rossby waves, tropical depression (TD)-type disturbances (or easterly waves), MRG waves, or Madden-Julian Oscillation (MJO), suggesting

ev

possibilities for extending the lead time of TC genesis prediction with numerical models. Note that both easterly waves and MJOs do not belong to normal mode solutions of Matsuno’s

ie

equations. Easterly waves may be viewed as off-equatorial westward propagating Rossby gyres

w

(Kiladis et al., 2009). The distinction between an equatorial Rossby wave and a MRG wave and their impact on TC formation will be discussed in section 4.3.

On

Although it has been suggested that the aforementioned tropical waves may appear as a precursor to TC genesis, it only becomes feasible until recently to simulate the complicated

ly

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Computing in Science & Engineering

multiscale interactions of TC formation and tropical waves with advanced global modeling and supercomputing technologies. Here, the term multiscale interaction is loosely defined as the nonlinear processes that involve flows with different scales, including wavelength reduction of large-scale waves, TC genesis associated with the appearance of a mesoscale vortex, and/or their interaction with resolved or parameterized convective-scale processes. Since 2007, the association of TC formation with the three types of tropical waves shown in Figure 4 has been examined in three case studies by Shen et al. (2010a, 2011), Shen et al. (2010b), and Shen et al. (2012a), respectively. In Shen et al. (2010a), we have shown that accurate representation of multiscale flows can contribute to the 5

Computing in Science & Engineering

ie

ev

rR

Fo Figure 4: A schematic view of three different types of tropical waves. (a) An equatorial

w

Rossby wave, which appears as a pair of vortex circulations and low pressure centers (labeled by L) symmetric with respect to the equator, one in the Northern Hemisphere and the other in the

On

Southern Hemisphere; (b) African Easterly Waves (AEWs), which are identified by the inverted trough in white dashed lines and have their origins over North Africa (courtesy of Dr. Chris Landsea, http://www.aoml.noaa.gov/hrd/tcfaq/A4.html); (c) a mixed Rossby gravity (MRG)

ly

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 20 of 62

wave, which appears asymmetric with respect to the equator as shown by the zonal winds (red) and low pressures (labeled by L). Thin arrow lines indicate the direction of total winds. (reproduced from Shen et al. 2012a).

predictability of TC Nargis formation, including northward movement of westerly wind belt, enhanced monsoonal circulation, formation of a pre-TC mesoscale vortex, and convective-scale precipitation system during the intensification stage. It was also shown that the pre-TC vortex was indeed in association with the northern vortex of an equatorial Rossby wave. However, 6

Page 21 of 62

because of the complexities in scale interactions, insightful visualizations have become crucial for improving the understanding of these transient processes and/or illustrating these processes in a simple way, which could in turn provide an efficient tool to systematically verify (or monitor) a model’s performance. To achieve this goal, we have successfully deployed the coupled advanced global multiscale modeling and concurrent visualization system (CAMVis) on NASA supercomputers (Shen et al., 2011); and recently developed the quasi-3D streamline package (StreamPack) and integrated it into the CAMVis system. In section 2, we introduce the NASA CAMVis. We then discuss in section 3 how a quasi-3D StreamPack is developed. In section 4, we first present a visualization of Hurricane Katrina (2005) to show its interaction

Fo

with an approaching upper-level jetstream, and hypothesize that this interaction might have led to Katrina’s intensification before making landfall. We then discuss two 3D visualizations to

rR

illustrate (i) the formation of Hurricane Helene (2006) and its association with an African easterly wave (AEW), and (ii) the formation of twin TCs and their interaction with an MRG wave. We conclude with a summary and discussion of future plans in section 5.

2. The NASA CAMVis

w

ie

ev

To improve high-impact tropical weather prediction, the CAMVis system (Shen et al., 2011) has been successfully deployed on NASA’s Columbia and Pleiades supercomputers, showing

On

promise in pursuing the related TC studies. CAMVis consists of the state-of-the-art GMM (Shen et al. 2006; Shen et al. 2010a,b) and global multiscale modeling framework (MMF; Tao et al.,

ly

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Computing in Science & Engineering

2009) running on Pleiades (as of November 2011, ranked 7th on the TOP500 list) and employing a concurrent visualization (CV) framework on the 128-screen hyperwall-2 (Ellsworth et al., 2006). The hyperwall-2 has modern graphics cards, InfiniBand interconnects, 1,024 cores and 475 terabytes of fast disk, and thus is capable of rendering one-quarter-billion pixel graphics. Recent developments and improvements to CAMVis include: (1) a revised parallel implementation to improve the MMF’s performance and parallel scalability; (2) deployment of the 1/8 degree GMM on the Pleiades supercomputer; (3) improvement of the parallel “M-on-N” data transfer model in the CV system version 2.0, which enables parallel data transfer between “M” computing nodes on Pleiades and “N” visualization nodes on the hyperwall-2; (4)

7

Computing in Science & Engineering

development of data modules to fuse NASA satellite data such as QuikSCAT sea winds and TRMM precipitation for inter-comparisons with model simulations at comparable resolutions; (5) development of the CV to Web (CV2Web) that enables real-time access to CV products via web browsers; and (6) development of the quasi 3D streamline packages (StreamPack) to provide insightful understanding of the hurricane’s multiscale interactions and transient dynamics. The first and second items are being documented in an article for publication. Items (3) and (4) were presented in Shen et al. (2011). The developments in items 5 and 6 are discussed below.

Fo

3. Recent Development

3.1 A Web Interface: Concurrent Visualization to Web (CV2Web)

rR

In order to maximize the results from a single simulation run, multiple products are usually generated, representing various fields and regions of interest as well as numerous feature-

ev

extraction and visualization techniques. When time-stamped outputs arrive from the computing nodes, each visualization node sequentially computes all of the requested visualizations,

ie

producing one image per visualization request. As part of the CV pipeline, the resulting

w

animations are streamed, as they are being generated, to remote displays at the investigators’ facilities. Although these features provide unprecedented opportunities for researchers to

On

conduct research activities, the dependence on supercomputer-scale visualization facilities seems to be an issue for many end users. Recently, in order to support more end users, such as managers, CV2Web has been implemented to provide a simple means for accessing real-time

ly

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 22 of 62

visualization results produced by CAMVis. This new capability is built on a hyperwall CV pipeline, enabling real-time web access to CV products. The CV2Web interface leverages the InfiniBand fabric between Pleiades and hyperwall-2 to give users with instrumented codes the ability to see multiple visualizations of their simulation in progress via a web browser, from any location. The current version supports real-time image update for each visualization. 3.2 Quasi-3D Streamline Package (StreamPack) In order to seamlessly visualize all different fields and their predictive relationship among different scale flows from the simulations with CAMVis, we have developed and integrated 8

Page 23 of 62

different visualization packages. As mentioned in the introduction, it is important to improve the representation of a TC’s structure and its interaction with environmental flows. As a TC’s structure displays altitude dependence, such as low-level counterclockwise circulation and upper-level clockwise circulation, visualizations of the TC’s circulations at different heights can be very helpful for illustrating both its evolution and interactions with environmental flows. To achieve this, we developed the 3D StreamPack, which is used to generate the streamlines at different heights in different colors. However, the current version does not really use the information on vertical wind velocity in the “Z” dimension (altitude), partly because the atmosphere is basically hydrostatic with a horizontal scale on the order of tens or hundreds of

Fo

kilometers but a vertical extent up to 10-40 km. Consequently, the streamlines produced in the visualizations are not true 3D lines. Rather, each pressure level is treated independently: 2D

rR

streamlines are produced within a pressure level, and then the levels are stacked to produce a quasi-3D image. To facilitate the discussions, we conceptually divide the levels into three “layers”, including low, middle and upper layers. Each of these layers, which are shown mainly

ev

in blue, green, and pink, respectively, contains several (e.g., 3~5) contiguous levels. Opacity is used to control the “transparency” of the streamlines at different heights to

ie

illustrate both the TC’s structure and its relationship with surrounding flows. The goal is to

w

clearly display major features that appear in all of the three layers and/or a specific layer. In the StreamPack, the degree of opacity is determined by a parameter called “alpha”, which is a

On

function of wind speeds. The larger the alpha value is, the more opaque streamlines are. In other words, if there exist more streamlines at different levels (layers),

they would look less

transparent. Figure 5 shows the impact of opacity on the quasi-3D streamlines generated from

ly

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Computing in Science & Engineering

exactly the same model data. Each panel uses a different “alpha” (i.e., opacity) value. The symbol ‘Z’ in the color bar indicates altitudes (heights). As mentioned, red lines show highaltitude (upper-level) winds, green lines show middle-altitude (middle-level) winds, and blue lines show low-altitude (low-level) winds. With the alpha set too high (Figure 5a), we see too much information about winds that are not important. Namely, streamlines at different layers appear to be overlapped together. With the alpha set too low (Figure 5c), we see the main feature, but not enough detail about surrounding (environmental) winds. In the visualizations that are discussed in section 4, we chose the alpha value in Figure 5b. The StreamPack displays streamlines at a specific level only when their speeds are “faster” so they are assigned with 9

Computing in Science & Engineering

larger alpha values. Alternatively, we may say that when wind speeds (w) exceed the critical wind speed (wc), larger alpha values are assigned to the corresponding streamlines, so they become more visible. Therefore, given the same data set (e.g., one specific frame), a larger alpha means a smaller wc. Therefore, given the same alpha (i.e., wc) in a time series of data, the appearance (or disappearance) of streamlines in a moving frame (e.g., following the TC) represents a change of wind speeds from wwc) to w>wc (w
wind speeds. In other words, the evolution of streamline density may

qualitatively indicate the evolution of average wind speeds, i.e., the denser streamlines, the stronger the average wind speeds.

ie

ev

rR

Fo Figure 5: Impact of opacity on the 3D streamline representation of a TC and its surrounding

w

flows: (a) high opacity, (b) medium opacity, (c) low opacity. Low-level winds are in blue, midlevel winds in green and yellow, and upper-level winds in red and pink.

On

4. Visualizations of Tropical Waves and Tropical Cyclone Formation

ly

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 24 of 62

As stated previously, improving the understanding of hurricane formation is one of the most challenging hurricane research activities. During the past several years, the performance of the GMM in simulating TC formation has been verified against global analyses and NASA satellite data such as QuikSCAT sea winds and Tropical Rainfall Measuring Mission (TRMM) precipitations. The selected TC cases include the severe cyclonic storm Nargis (2008) by Shen et al. (2010a), Hurricane Helene (2006) by Shen et al. (2010b), and twin TCs in May 2002 by Shen et al. (2012a). The formation mechanisms of these TCs are closely related with an equatorial Rossby wave, an Easterly wave and a mixed Rossby gravity wave, respectively. Except for the Nargis which was previously discussed in Shen et al. (2011), we demonstrate the capabilities of

10

Page 25 of 62

StreamPack and CAMVis in examining the scale interactions of the TC formation and tropical waves in the following. We begin with the visualizations of scale interactions between Katrina and an approaching upper-level jet stream in section 4.1. Then, we provide two visualizations to (1) examine the predictive relationship between an intensifying AEW and the formation of Hurricane Helene (2006) in section 4.2; and (2) address the scale interaction of twin TC formations and an MRG wave in section 4.3. These animations have been uploaded as Google documents which can be accessed via the corresponding URLs that are provided in the captions of Figures 6, 8 and 9. By clicking on each of these URLs, a “preview” version of the corresponding animation (at a

Fo

coarser-resolution) is playing. An original version of the animation (at full resolution) can also be downloaded by clicking “File” and then choose “download” in the preview mode. With these

rR

advanced 3D visualization techniques, we provide an integrative view of TC genesis to emphasize the importance of understanding the relative roles of large-scale and small-scale forcing in TC genesis and intensification in section 4.4.

ev

4.1 Visualizations of Scale Interactions for Katrina

w

ie ly

On

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Computing in Science & Engineering

11

Computing in Science & Engineering

Figure 6: Streamline visualization of multiscale interaction between the outflows of Hurricane Katrina and an approaching upper-level jet stream from a 5-day 1/8-degree run, which is associated with the intensification of Katrina before landfall. Low-level streamlines are in blue, upper-level streamlines are in pink and red. Katrina’s intensification is indicated by the appearance of dense, red streamlines at the upper levels (panels c and d), which are in association with strong vertical motion. The symbol ‘J’ indicates the jet stream. The corresponding animations can be found: http://tiny.cc/ns208

Fo

Enabled by NASA supercomputing technologies, we first deployed the high-resolution global model at a resolution of 1/8 degree on Columbia supercomputer and obtained a realistic simulation of Katrina’s movement, intensity, and near-eye wind distribution from a 5-day run

rR

(e.g., Shen et al., 2006). In Figure 2 of Shen et al. (2006), it was shown that the 1/8 degree model was able to capture intensity evolution during the second phase of intensification (e.g., between

ev

0000 UTC August 28 and 1200 UTC August 29, 2005.) Recently, the model has been ported and tested successfully on the Pleiades supercomputer, which produced similar results with

ie

insignificant differences from the version on Columbia. With the newly developed streamline package, it becomes feasible to examine the role of the interaction between Katrina’s outflow

w

and its environmental flows on the storm’s intensification. Figure 6 shows the evolution of a simulated Katrina from a 5-day run. All of the frames are derived from a high-resolution

On

visualization (http://tiny.cc/ns208). As time progresses, stronger upper-level anti-cyclonic flows (in red) developed in association of the initial intensification of Katrina (e.g., Figure 6b). Later

ly

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 26 of 62

on, Katrina experienced rapid intensification. The animation helps lead to a hypothesis that the (horizontal) phasing of an approaching jet stream and the southwesterly outflow of Katrina (to the northwest of Katrina) may have further strengthened the upper-level anticyclonic flow over the Katrina, as indicated by the appearance of denser upper-level streamlines in expanding areas in pink, and thus enhanced Katrina’s development along with strong deep convections (see also Figures 6c, 6d).

12

Page 27 of 62

w

ie

ev

rR

Fo ly

On

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Computing in Science & Engineering

Figure 7: Evolution of the upper-level jet steam and the upper-level circulation of Katrina at 200 hPa from the NCEP reanalysis (a-c) and the model run (d-f). Panels from top to bottoms are validated at 1200 UTC May 28, 1200 UTC May 29, and 0000 UTC May 30, respectively. While ‘J’ roughly indicates the location of the jet stream, a grey circle indicates the location of the simulated Katrina.

While a systematic study for verifying this hypothesis is being planned (e.g., submission of a proposal), in this paper we provide a preliminary analysis of 200 hPa winds from the NCEP analysis (as “observations”) and the model run. As NCEP analysis data are available only at a 13

Computing in Science & Engineering

time interval of 6 hours, it is challenging to apply these data to take advantage of the 3D StreamPack with high temporal-resolution capability. A different graphic package was used to generate vector and wind speeds (shaded) plots to examine the evolution of velocities at 200 hPa (Figure 7), which can be used to verify our methodology in generating the visualization in Figure 6. It is shown that as the jet stream and the Katrina moved closer from 1200 UTC 28 AUG to 1200 UTC 29 AUG, i.e., a possible horizontal phasing, both of the jet streaks and upperlevel flow of Katrina were enhanced. Shaded areas in Figure 7 indicate the existence of jet streaks that are localized regions of very fast winds embedded within the jet stream. This kind of “mutual” and “cooperative” interaction that may potentially lead to the intensification of

Fo

Katrina before its landfall is being investigated in a separate study. During the period of 1200 UTC 29 to 0000 UTC 30, the simulated Katrina continued to intensify, which is different from

rR

the observed Katrina. Plausible reasons for the false intensification includes that: (1) As the 5day run predicts a slow speed of Katrina, the simulated Katrina stayed over the ocean for a longer time so that it can extract more energy from the ocean; (2) Because the simulated jet

ev

streaks are about 5 degree south as compared to the observation (e.g., Figures 7c and 7f), simulated Katrina’s outflow can still interact with the jet streaks, leading to Katrina’s

ie

intensification. Note that although the NCEP analysis data revealed the jetstreaks with stronger

w

intensity (about 5~10 m/s larger) than the simulated ones, it resolved a weaker Katrina than that of the best track and the simulated Katrina as well. Note that Figure 7 displays vector and wind

On

speeds at a specific level while Figure 6 shows streamlines in different “layers” which contain multiple levels. The quasi-3D streamlines seem to be very effective in representing the scale interaction and linking it to the Katrina’s intensification. However, more rigorous verification is still needed and is the subject for a future study.

ly

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

4.2 Visualizations of Helene’s (2006) Formation and an Intensifying AEW The second visualization is used to examine the predictive relationship between an intensifying AEW and the formation of Hurricane Helene (2006) (Shen et al., 2010b). Previous studies have shown that nearly 85% of intense hurricanes over eastern North Atlantic (Landsea, 1993) have their origins as AEWs, and the initiation of an AEW has been found to be related to the release of instability associated with an African easterly jet (AEJ). 14

Page 28 of 62

Page 29 of 62

Therefore, to extend the lead time for predicting hurricanes that originate near the Cape Verde Islands, it was important to accurately simulate the multiscale interactions between hurricanes, AEWs, and the AEJ, as well as the impact of surface processes. During the NASA African Monsoon Multidisciplinary Analyses (NAMMA) period between late August and late September 2006, six AEWs appeared over Africa, propagated westward, and then passed by the Cape Verde Islands. In early September, an observed AEW developed into a Cape Verde storm—Hurricane Helene. With the CAMVis GMM, Shen et al. (2010b) conducted extendedǦrange (30Ǧday), highǦresolution global numerical experiments to simulate the initiation and propagation of these six consecutive (AEWs) and their association with hurricane formation.

Fo

It is shown that the statistical characteristics of individual AEWs are realistically simulated with larger errors in the 5th and 6th AEWs. Remarkable simulations of a mean African easterly jet

rR

(AEJ) are also obtained. Of interest is the potential to extend the lead time for predicting hurricane formation as the 4th AEW is realistically simulated (Figure 4 of Shen et al. 2010b), which is briefly illustrated with a 3D streamline visualization below.

w

ie

ev

ly

On

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Computing in Science & Engineering

15

Computing in Science & Engineering

Figure 8: Formation of Hurricane Helene (2006) and its association with the intensification of an African Easterly Wave (AEW) in a 30-day run initialized at 0000 UTC August 22, 2006. Upper-level winds are in pink, middle-level winds in green and low-level winds in blue. (a) Initial formation of a closed circulation associated with the 4th AEW that moves to be over the ocean, validated at 0000 UTC Sep. 13 (day 22); (b) initial formation of Helene associated with enhanced low-level inflow with counterclockwise circulation, validated at 2100 UTC Sep. 14; (c) further intensification of Helene with an enhanced outflow with clockwise circulation (indicated in pink), validated at 2200 UTC Sep. 16. An animation can be found at: http://tiny.cc/j9ul9

Fo

The visualization in Figure 8 shows: (a) the initial formation of a closed circulation

rR

associated with the 4th AEW that moves into the ocean, validated at 0000 UTC Sep. 13 (day 22); (b) TC (Helene) formation and initial intensification associated with intensified, low-level inflow with counterclockwise circulation which is indicated by the appearance of low-level

ev

streamlines in blue, validated at 2100 UTC Sep. 14; and (c) further intensification of the TC with an enhanced outflow with clockwise circulation (indicated in pink), validated at 2200 UTC Sep.

ie

16. This experiment, with other parallel experiments documented in Shen et al. 2010b, suggested

w

the potential to extend the lead time (up to 20 days in this case) for predicting hurricane formation.

ly

4.3 Visualizations of Twin TCs and an MRG Wave

On

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 30 of 62

Previous studies suggest that a pair of twin TCs, symmetric with respect to the equator, may occur associated with a large-scale Madden-Julian Oscillation (MJO). On 1 May 2002, MJOorganized convection appeared over the Indian Ocean and moved eastward. Five days later, two TCs, Kesiny and 01A, formed successively and then turned into a twin TC. In a recent study (Shen et al. 2012b), the formation of these twin TCs is hypothesized as resulted by the scale interactions of three gyres associated with an MRG wave during an active MJO phase. Because of its asymmetric features, a MRG wave may be a precursor to the successive formation of TCs at different longitudes with time lags in different hemispheres, one in the Northern Hemisphere (NH) and the other in the Southern Hemisphere (SH). 16

A pair of TCs might move at different

Page 31 of 62

phase speeds and eventually turn into twins symmetric with respect to the equator, appearing as a transition from an MRG wave to an equatorial Rossby wave. However, from a modeling perspective, predicting these TCs is very challenging, because a time lag of 3-5 days in their formation would require a high level of performance in extended-range (beyond 5 days) simulations.

High-resolution simulations suggested that our model is capable of reproducing the evolution of the MRG wave and thus the formation of TCs Kesiny and 01A about two and five days in advance as well as their subsequent intensity evolution and movement within an 8-10 day period.

Fo

Here, we discuss how advanced animations can help illustrate the transient processes of successive formation of TCs associated with the intensified MRG wave. Figure 9 shows the

rR

formation of these twin TCs from a 10-day simulation initialized at 0000 UTC 1 May 2002. In each panel, the NH and SH are on the left and right sides of the panel, respectively. Labels “E” and “W” indicate easterly and westerly winds, respectively. Panel (a) shows the formation of TC

ev

Kesiny, the southern companion of the twin TCs. As time proceeds, the interaction between the westerly and easterly winds (Figure 9b) and the intensified MRG wave (Figure 9c) may have led

ie

to the formation of TC 01A, the northern companion of the twin TCs (Figure 9d). The animation

w

(http://tiny.cc/19r48) clearly shows that the successive formation of TCs Kensiny and 01A is associated with the intensification (appearance) of an MRG wave along the Equator (e.g., during

On

the period of May 5~7), which can be identified in Figure 9c by rotating this panel 90 degree clockwise and comparing the embedded while box with

Figure 4c. Note that due to its

intensification after twin TC formed and started moving off the Equator, the MRG wave becomes

ly

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Computing in Science & Engineering

more visible in the 3D visualization (Figure 9c). The white arrow in each of the zoomed-in panels (middle panels of Figure 9) is referred to as the spinning axis, and roughly indicates the location of vortex centers at different heights. Through the high temporal resolution visualization, it can be shown that the spinning axis of a mature TC points vertically (in the z direction), while the direction of the spinning axis changes with time during the formation stage. The latter suggests that the vortex centers at different heights are not coherent, which in turn suggests that it is challenging to improve the initialization of a weak vortex because of the absence of the vertical coherence.

17

Computing in Science & Engineering

ev

rR

Fo Figure 9: Visualization of the formation of the twin TC in May 2002. In each panel, the NH

ie

and SH are on the left and right sides of the panel, respectively. Labels “E” and “W” indicate

w

easterly and westerly winds, respectively. Panel (a) shows the formation of TC Kesiny in the SH. Panels (b-d) indicate the formation of TC 01A, the counter companion of the twin TCs in the NH.

On

The white arrow (the so-called spinning axis) in each of the zoomed-in panels (middle) indicates the location of the vortex centers at different heights. The equator is labeled ‘EQ’ and indicated by a white line in panel (a). The appearance of the intensified mixed Rossby gravity wave is

ly

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 32 of 62

shown in the white box in panel (c) which includes a side panel with a sketched wave pattern (bottom left). An animation can be found at: http://tiny.cc/19r48 4.4 A View of TC Genesis In this study, we have showed the potential of the 3D StreamPack with the capabilities of high temporal resolution visualizations for illustrating the transient dynamics of TC formation. Based on this study and earlier studies (Shen et al., 2010a,b; 2011; 2012a), we propose the following view of TC genesis in a triple-scale (i.e., large-, medium- and small-scales) system: (1) Both large-scale (tropical wave) systems and small-scale (e.g., precipitation) systems are 18

Page 33 of 62

important for the intensification and formation of a TC (or vortex) at the mesoscale. Because of the asymmetry in spatial and temporal scales and the strengths among these systems, it is important to understand the relative importance of large-scale and small-scale forcing at different stages of a TC lifecycle. Here the large-scale and small-scale forcing can be viewed as the external and internal processes, respectively, in the conceptual model of the TC genesis proposed by Zehr (1992). (2) Large-scale tropical waves such as an AEW or MRG wave may provide determinism in regards to the timing and location of TC genesis. Namely, an accurate representation of the evolution of the large-scale tropical waves may help “narrow” down the uncertainties in the area and period for TC formation; (3) Accurate simulations of small-scale

Fo

processes remain challenging, but their aggregate feedbacks to the TC genesis, especially under strong “forcing” such as with an MRG wave or an equatorial Rossby wave, can be simulated

rR

with some degree of satisfaction (e.g., Shen et al. 2010; 2012a). However, it would require further improvement in the visualization package (such as inclusion of strong vertical motion associated with heavy precipitation) to illustrate these feedback processes. In reality, there is no

ev

clear separation between the external and internal processes. In deference to this view on TC genesis and intensification, large-scale forcing may still continue to impact the strength of the

ie

TC after its genesis, and thus should be considered properly to predict the intensity of the TC.

w On

5. Concluding Remarks

ly

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Computing in Science & Engineering

To save lives and reduce the costs associated with storm damage, it is crucial to improve the short-term forecasts of hurricane intensity and formation, and to improve our understanding of TC inter-annual variability and the impact of climate change (e.g., doubling CO2 and/or global warming) on TC activities. As TC dynamics involve multiscale interactions among large-scale flows, mesoscale vortices, and small-scale cloud motions, an ideal numerical model suitable for TC studies should possess capabilities in simulating and displaying (e.g., visualizing) these multi-scale processes and their cross-scale interactions. The newly developed CAMVis, consisting of advanced, global multiscale modeling and concurrent visualization systems, has previously shown a potential for simulating the predictive relationship between TC formation and tropical waves. 19

Computing in Science & Engineering

To illustrate the related transient dynamics, we recently developed the 3D streamline package (StreamPack) to generate streamlines at different heights. We have provided three visualizations to illustrate (1) the interaction of Hurricane Katrina and an upper-level jet stream, which is hypothesized to explain the intensification of Katrina before its landfall; (2) the association of Hurricane Helene’s formation and intensification with an intensifying AEW; and (3) scale interactions of twin TC formation and an MRG wave. With these results, we emphasize that it is important to understand the relative role of large-scale and small-scale processes, which are not completely separable, in the evolution of a TC at different stages. The quasi-3D StreamPack is a powerful tool for displaying TC cross-scale interactions.

Fo

However, the current version does not use the information about vertical wind velocity in the “Z” dimension (altitude). Ultimately, we would like to do true 3D real-time simulation and

rR

visualization, which will take considerable work to accomplish the goal of greater realism; this is a subject for future study. A future version of StreamPack, coupled with the enhanced scalability of CAMVis—the latter of which is now in preparation for publication,—will be used with

ev

experimental, real-time simulations to systematically monitor the realism of TC formation and its association with different tropical waves.

w

ie ly

On

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 34 of 62

20

Page 35 of 62

Acknowledgements: We are grateful to the following organizations for their support: the NASA Earth Science Technology Office, the Advanced Information Systems Technology Program, the National Science Foundation’s Science and Technology Center, and the NASA Modeling, Analysis Prediction Program. We also appreciate the support of the computing time and facilities by the the NASA High-End Computing Program, and the NASA Advanced Supercomputing facility at Ames Research Center (ARC). We thank Ms. Jill Dunbar of NASA ARC for proofreading this manuscript.

References

rR

Fo

Atlas, R., O. Reale, B.-W. Shen, S.-J. Lin, J.-D. Chern, W. Putman, T. Lee, K.-S. Yeh, M. Bosilovich, and J. Radakovich, 2005: Hurricane forecasting with the high-resolution NASA finite

volume

general

ev

circulation

doi:10.1029/2004GL021513.

model.

Geophys.

Res.

Lett.,

32,

L03801,

ie

Bengtsson, L., K., I. Hodges, and M. Esch, 2007: Tropical cyclones in a T159 resolution global climate model: comparison with observations and re-analyses. Tellus A 59 (4), 396.416 doi:10.1111/j.1600-0870.2007.00236.x

w

Biswas, R., M.J. Aftosmis, C. Kiris, and B.-W. Shen, 2007: Petascale Computing: Impact on

On

Future NASA Missions. Petascale Computing: Architectures and Algorithms, 29-46 (D. Bader, ed.), Chapman and Hall / CRC Press, Boca Raton, FL.

ly

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Computing in Science & Engineering

Ellsworth D., B. Green, C. Henze, P. Moran, T. Sandstrom, 2006: Concurrent Visualization in a Production Supercomputing Environment. IEEE Trans. on Visualization and Computer Graphics,. 12, 5, September/October 2006. Frank, W. M. and P. E. Roundy, 2006: The Role of Tropical Waves in Tropical Cyclogenesis. Mon. Wea. Rev. 134, 2397-2417.

Kiladis, G. N., M. C. Wheeler, P. T. Haertel, K. H. Straub, and P. E. Roundy (2009), Convectively coupled equatorial waves, Rev. Geophys., 47, RG2003, doi:10.1029/2008RG000266. 21

Computing in Science & Engineering

Landsea, C. W., 1993: A climatology of intense (or major) Atlantic hurricanes. Mon. Wea. Rev., 121, 1703-1713. Matsuno, T. 1966: Quasi-Geostrophic Motions in the Equatorial Area. J. Meteor. Soc. Japan, 14, 25-42. Shen, B.-W., R. Atlas, O. Reale, S.-J. Lin, J.-D. Chern, J. Chang, C. Henze, and J.-L. Li, 2006: Hurricane Forecasts with a Global Mesoscale-Resolving Model: Preliminary Results with Hurricane Katrina (2005). Geophys. Res. Lett., 33, L13813, doi:10.1029/2006GL026143. Shen, B.-W., W.-K. Tao, W. K. Lau, R. Atlas, 2010a: Predicting Tropical Cyclogenesis with a Global Mesoscale Model: Hierarchical Multiscale Interactions During the Formation of

Fo

Tropical Cyclone Nargis (2008). J. Geophys. Res.,115, D14102, doi:10.1029/2009JD013140. Shen, B.-W. W.-K. Tao, and M.-L. C. Wu, 2010b: African Easterly Waves in 30-day High-

rR

resolution Global Simulations: A Case Study during the 2006 NAMMA Period. Geophys. Res. Lett., L18803, doi:10.1029/2010GL044355. Shen, B.-W., W.-K. Tao, and B. Green, 2011: Coupling Advanced Modeling and Visualization

ev

to Improve High-Impact Tropical Weather Prediction. IEEE Computing in Science and Engineering (CiSE), vol. 13, no. 5, pp. 56-67, Sep./Oct. 2011, doi:10.1109/MCSE.2010.141

ie

Shen, B.-W., W.-K. Tao, and Y.-L. Lin, 2012a: Genesis of Twin Tropical Cyclones Revealed by

w

a Global Mesoscale Model: the Role of Mixed Rossby Gravity Wave. (submitted to AGU J. Geophys. Res.; under revision)

On

Silva-Dias, P. L., W. H. Schubert and M. DeMaria, 1983: Large- scale response of the tropical atmosphere to transient convection. J. Atmos. Sci., 40, 2689-2707.

ly

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 36 of 62

Tao, W.-K., D. Anderson, J. Chern, J. Entin, A. Hou, P. Houser, R. Kakar, S. Lang, W. Lau, C. Peters-Lidard, X. Li, T. Matsui, M. Rienecker, M. R. Schoeberl, B.-W. Shen, J. J. Shi, and X. Zeng, 2009: The Goddard Multi-Scale Modeling System with Unified Physics. Ann. Geophys., 27, 3055-3064. Zehr, R.M., 1992: Tropical cyclogenesis in the Western North Pacific. NOAA Technical Report NESDIS 61, U. S. Department of Commerce, Washington, DC, 181 pp.

22