IN51B-1811
The Hurricane Problem – The Three Faces of the Big Data Challenges

Friday, 18 December 2015
Poster Hall (Moscone South)
Svetla M Hristova-Veleva1, Mark Boothe2, Sundararaman Gopalakrishnan3, Ziad S Haddad1, Brian Knosp4, Bjorn Lambrigtsen1, Peggy Li1, Michael t Montgomery2, Noppasin Niamsuwan1, Tsae-Pyng J Shen1, Vijay Tallapragada5, Simone Tanelli4, Samuel Trahan5, Francis J Turk1 and Quoc A Vu4, (1)NASA Jet Propulsion Laboratory, Pasadena, CA, United States, (2)Naval Postgraduate School, Monterey, CA, United States, (3)Atlantic Oceanographic and Meteorological Laboratory, Hurricane Research Division, Miami, FL, United States, (4)Jet Propulsion Laboratory, Pasadena, CA, United States, (5)NOAA College Park, College Park, MD, United States
Abstract:
Despite recent progress in hurricane track forecasts, we still lack understanding of the multi-scale interactions that sometimes lead to cylogenesis or rapid intesification and other times do not. To improve hurricane forecasts we need to understand the physical processes that control hurricane evolution and to evaluate whether the models represent them properly. This is where we face Big Data challenges in three different ways: dealing with a multitude of observations; extracting relevant information from voluminous model forecast; and performing carefully designed diagnostics to evaluate the models.

Satellite observations provide invaluable information. However, needed are long-term observations of multiple parameters, from a multitude of instruments. These data come from disparate sources, in different formats, with varying latency. Bringing all these observations to bear on addressing the hurricane problem presents the observational side of the Big Data challenges.

Hurricane evolution is sensitive to the storm internal dynamics as well as environmental characteristics. This is why accurate forecasting requires the use of regional models, with higher resolution and better parameterizations, as well as the use of global models that better depict the large-scale environment, necessary for properly capturing the important scale interactions. Extracting relevant information from the extremely voluminous model forecasts, we face the model-related side of the Big Data challenges.

To properly evaluate the models we need to go beyond the comparison of the geophysical fields and use instrument simulators to compute synthetic observations from the model fields for a more direct comparison. Producing realistic synthetic data requires the use of complex, computationally intensive, instrument simulators. This demand, in addition to developing on-line analytics to support model evaluation, is where we face the analytics side of the Big Data challenges.

We are now developing a system – http://tropicalcyclone.jpl.nasa.gov- that addresses the issues of model evaluation and processes understanding with the goal of improving the accuracy of hurricane forecasts.

In this presentation we will use our framework to describe, define and illustrate the three faces of the Big Data challenges we confront today.

<< Previous Abstract | Next Abstract >>