The Evolution of an Object-oriented Big Data Tool for Analysis of Climate Extremes
Abstract
The importance of dynamic data mining tools in the era of big data is invaluable to researchers across the geophysics spectrum and beyond. This study showcases the usefulness of the CONNected-objECT (CONNECT) algorithm for information gathering and tracking across hydroclimate extremes, namely hurricanes, atmospheric rivers (ARs), and heatwaves.
The CONNECT algorithm is an object-oriented "tracking" algorithm designed for use with large datasets. CONNECT's primary strength comes from its ability to extract multi-dimensional information at voxel and object scales, making CONNECT a useful tool to examine the patterns of climate extremes across spatial scales. CONNECT's initial usage was for quantifying and tracking extreme rainfall events, where it proved to be a valuable source of information regarding the most notable hurricanes of the 2017 hurricane season, namely Harvey, Irma, and Jose. However, owing to the fluctuating nature of storm-related precipitation in lower intensity storms, CONNECT suffered issues of over-segmentation when rainfall was selected as the target variable. To overcome the issue of over-segmentation, CONNECT was adapted to the integrated water vapor transport (IVT) variable—a well-established variable for the tracking of ARs. Using IVT as the target variable allowed for the continuous tracking of storm objects impossible to attain with rainfall. However, CONNECT's usefulness as an AR tracking method suffered from its permissive criteria for classifying an AR, along with issues of temporal under-segmentation. To combat these shortcomings, CONNECT was further adapted for AR indentification by including more restrictive criteria and a region growing technique (AR-CONNECT). AR-CONNECT introduced variable mapping capabilities, which allowed for AR objects to have all storm-related rainfall (among other Earth climate variables) mapped to the object. The most recent work with CONNECT has been for the extraction of heatwaves. Heatwaves and ARs share similarities in their difficulty to predict at sub-seasonal to seasonal (S2S) timescales. This presents CONNECT with a useful opportunity to extract object characteristics and climate fluctuations present at the time of events, which are potentially useful for S2S predictions of heatwaves.- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2019
- Bibcode:
- 2019AGUFMNH11A..07S
- Keywords:
-
- 4313 Extreme events;
- NATURAL HAZARDS;
- 4314 Mathematical and computer modeling;
- NATURAL HAZARDS;
- 4318 Statistical analysis;
- NATURAL HAZARDS;
- 4339 Disaster mitigation;
- NATURAL HAZARDS