Scientists from across disciplines came together to discuss the future of water resource management using artificial intelligence during a seminar on Saturday at this year’s American Association for the Advancement of Science meeting. The program was organized by Suzanne Pierce of UT’s Texas Advanced Computing Center.
The first speaker was Yolanda Gil, a professor of computer science and spatial sciences at the University of Southern California. Gil spoke about the potential for using artificial intelligence to automate decision-making. This is possible using the concept of “thoughtful AI,” in which machines acquire thoughtful learning patterns.
A variety of disciplines face problems that can be solved using AI, she said. Water resource management involves the interplay between water usage, land cover changes and food insecurity, among many other issues. Researchers, and the machines working on these problems, need to understand the interplay between these areas, she said.
In order to accomplish this, machine learning must integrate across disciplines. In particular, attention should be paid to the geosciences; the food, energy and water nexus; and the United Nations Development Programme, or UNDP, which sets goals for sustainable development.
Additionally, work in machine learning must pay attention to both human and natural systems. Especially when creating models for water management, the fields of economics, agriculture, sociology, natural sciences and infrastructure must be integrated. In general, the challenge is that many of these areas have different approaches to modeling. Gil said she hopes that machine learning can be a way to overcome these challenges.
The next speaker was Scott Peckham, a researcher in hydrologic sciences at the University of Colorado Boulder and the Institute of Arctic and Alpine Research, or INSTAAR. Peckham spoke of creating a standardized system of scientific quantities for better machine learning. Without such a system, there is a lack of interoperability, or the ability to exchange information across scientific disciplines.
The INSTAAR created a system known as the Community Surface Dynamics Modeling System, or CSDMS, which models the dynamic changes of Earth’s surface. This system uses standard names with base quantities to map variables, according to Peckham. This creates a uniformity in modeling. The system is based on seven “neat quantities” which are conserved throughout physics. These quantities, which are used across geosciences in models and data sets, are the base quantities to which all other quantities are related. Thus, Peckham and INSTAAR are able to organize variables and their synonyms into one unified system.
The last speaker was Vipin Kumar, a professor of computer science and engineering at the University of Minnesota. Kumar spoke of modeling global surface dynamics and using satellite data to create a global surface water monitoring system. Such a system would monitor water dynamics, including examining areas changing from water to land, shrinking islands and the construction of dams.
Other applications of a monitoring system include examining land-water interactions, hybrid physics models and supporting the Internet of Things for water. Such systems could use AI for visualization, decision support systems and connecting scientists with non-scientists.
The scientists ended the session with a message about the role of AI in sciences. The scientific community has embraced AI to support science, Pierce said. AI can be a way to deal with the often messy data scientists use, and it has the potential to create an open knowledge network between research scientists.