Skip to content. Skip to navigation
CIM Menus
 

ClimateNet: Bringing the power of Deep Learning to the climate community via open datasets and architectures


Karthik Kashinath


May 10, 2019 at  2:30 PM
McConnell Engineering Room 437

Abstract:

In this talk we briefly discuss how deep learning (DL) has had some remarkable successes in pattern recognition and pattern discovery in climate science. However, access to large amounts of reliable, expert-labeled ground truth data remains the biggest challenge to: (i) improving the accuracy and performance of these supervised DL models; (ii) successfully deploying these models at scale; and (iii) building unified models for all classes of weather and climate patterns.

In an effort to address this major challenge we have launched ClimateNet, a project whose mission is to create community-sourced open-access expert-labeled datasets and architectures for improved accuracy and performance of DL models on a range of supervised learning problems: classification, detection, segmentation and tracking. The motivation stems from the fact that applying DL to pattern recognition tasks in the weather and climate sciences remains challenging due to scarcity of reliable labelled data. While there exist many heuristics and algorithms for detecting weather or climate patterns, including extreme events, the disparities between the output of these different methods within a single class of event are huge and often impossible to reconcile (see, for example, the recent ARTMIP initiative: Atmospheric River Tracking Method Intercomparison Project (ARTMIP): project goals and experimental design). This project attempts to circumvent this problem by drawing a leaf from the book of success stories in the computer vision community. ( The ClimateNet webpage)

We present our preliminary efforts in the development and organization of this project, how it relates a larger concerted effort called EnviroNet.

Bio:

Karthik Kashinath leads various climate informatics projects at the Big Data Center @ NERSC (Lawrence Berkeley Lab). He received his Bachelors from the Indian Institute of Technology, Madras in 2007, Masters from Stanford University in 2009 and PhD from the University of Cambridge, U. K. in 2013. His background is in engineering and applied physics. He has worked on various projects spanning a wide range of disciplines from supersonic aircraft engines to battery technologies to complex chaotic systems and turbulence. He joined Lawrence Berkeley Lab in 2013 as a post-doc in climate science with Bill Collins and NERSC in 2017 as a member of the Data & Analytics Services group. His current research interests lie in novel data analytics and pattern discovery methods for large complex systems such as Earth’s climate. When he is not in front of the computer he runs up mountains, swims in lakes and cooks exotic global cuisines.