Interdisciplinary research on applications of big-data and AI in Earth sciences

In recent years, I am fortunate to team up with a group in the department of Information Systems of UMBC, led by Dr. Jianwu Wang, to work together on some highly interesting and important interdisciplinary research. Dr. Wang is an expert in the areas of “Big Data” and artificial intelligence (AI). Our collaborative research aims to use the latest advances of technology to tackle some long-lasting challenges in the areas of satellite remote sensing, climate modeling and data analytics. Another important objective of this collaboration for me is to introduce the “Big Data” and AI, which are the most vibrant areas of research in recent years, to our physics students. In 2018, we won a NSF grant to launch “Multidisciplinary Research and Education on Big Data + High-Performance Computing + Atmospheric Sciences” (known as the “UMBC Cyber-training”). With this NSF support, we have successfully finished three years of cyber-training which attracted ~50 graduate students from UMBC (5 from the ATPH) and other universities from the U.S. This program has been highlighted many times in UMBC website. Dr. Wang and I were invited by the UMBC IAAC (Interdisciplinary Activities Advisory Committee) to give a talk on our interdisciplinary collaboration experiences. In addition to education and outreach, the cyber-training program has also inspired many interesting research projects. In my view, a few projects could even lead to breakthroughs in satellite remote sensing. For example, one of the cyber-training student teams has been working on dust plume detection from satellite images using AI-based algorithm. The preliminary results are highly encouraging and could later be used in many applications, such as real-time dust event detection and tracking. Another example of our interdisciplinary collaboration is a NASA funded project to develop a novel and user-customizable data processing and analytic infrastructure based on the big-data techniques to aggregate irregular discontinuous pixel-level satellite observations to regular grid-level continuous products. This research could provide the roadmap for the data processing system of future NASA satellite missions. Recently, our interdisciplinary team has won a major grant to “Develop Passive Satellite Cloud Remote Sensing Algorithms using Collocated Observations, Numerical Simulation and Deep Learning” from NASA’s Advancing Collaborative Connections for Earth System Science (ACCESS) Program. One objective of this research is to use the combination of AI and collocated passive-active observations to develop a novel satellite cloud remote sensing algorithm that is hopefully free of the impacts of 3-D radiative transfer effects. This grant will bring over 1.2 million research funding to UMBC for the next three years.