By AI Trends Staff
Researchers at the Norwegian Meteorological Institute are studying how to apply AI to predictions of the spread of sea ice, to make warning to vessels in the polar seas cheaper, faster and more widely available.
The effort is indicative of a trend to apply AI to the management of the environment, with the aim of gaining accuracy and potentially lowering costs.
Efforts to predict sea ice today are resource intensive. “As of now, large resources are needed to create these ice warnings, and most of them are made by The Norwegian Meteorological Institute and similar centers,” stated Sindre Markus Fritzner, a doctoral research fellow at UiT The Arctic University of Norway, in a recent account in Phys.org.
The ice warnings used today are based on dynamic computer models fed with satellite observations of the ice cover, and whatever updated data can be gathered about ice thickness and snow depth. This generates an enormous volume of data, which then needs to be processed by powerful supercomputers.
Fritzner is working on training a machine learning model by loading in data for one week, then data for how it will look one week later. In this way, the machine learns and eventually will be able to make predictions. Once developed, the algorithm will consume less computing power than the processing of today’s physical models.
“If you use artificial intelligence and have a fully trained model, you can run such a calculation on a regular laptop,” Fritzner stated.
In the current system, the predictive models need to be run on supercomputers at the Norwegian Meteorological Institute, which then needs to run the models on a supercomputer and transmit the data back to the vessel. “If equipped with the right program and artificial intelligence, this can be done from the vessel itself, with nearly no computing power required at all,” Fritzner says.
While the outlook is promising, more development work remains. “As long as the changes in the ice were small, the machine learning functioned quite well. When the changes were greater, with a lot of melting, the models struggled more than the physical models,” Fritzner stated. The physical models are constantly adapted to large geophysical changes like increased melting and rapid changes to the weather.
Smart Sensors to Span Continents
A network of smart sensors spanning continents is being constructed by the Sage project, with the help of a $9 million grant from the National Science Foundation awarded in September 2019 to a team led by Northwestern University.
The Sage project calls for advanced machine learning algorithms to be moved to the edge of an Internet of Things (IoT) network, to bring the data analysis very near to the site where data is gathered, according to an account on the website of Sage. The organization is directed by Pete Beckman of Northwestern University, and includes participation by 10 to 15 other academics. The site describes the organization as wanting to build a software-defined sensor network, cyberinfrastructure for AI at the edge.
Linking small, powerful, computers directly to high-resolution cameras, air quality and weather sensors, and experimental Light Detection and Ranging (Lidar) systems, this new distributed infrastructure will enable researchers to analyze and respond to data almost instantly. From early detection of wildfire smoke plumes in California to identifying ultrasonic calls of bats or the patterns of pedestrians in a busy crosswalk, Sage’s AI-enabled sensors will give scientists a new tool to understand our planet, the site states.
The distributed, intelligent sensor network will work to understand data on the impacts of global urbanization, natural disasters such as flooding and wildfires, and climate change on natural ecosystems and city infrastructure. Sage will embed computers directly into the sensor network and rely on advancements in edge computing to analyze the torrent of sensor data as it streams past.
Project partner instruments include the NSF-funded Array of Things, the NSF’s National Ecological Observatory Network (NEON), Atmospheric Radiation Measurement (ARM), the High-Performance Wireless Research and Education Network (WIFIRE). Sage will work to integrate measurements from these multiple instruments in the “software-defined” sensor network.
Sage test nodes will be deployed in California, Colorado, and Kansas and in urban settings in Illinois and Texas. The project will build on the open source technology platform used in the Array of Things project, which has deployed more than 100 sensors with edge computing capabilities within Chicago.
(During an interview with AI Trends, Pete Beckman elaborated on some details of the Sage project.
Can you describe the project in a nutshell?
In a nutshell, for the Sage project, scientific instruments are spread around the country and the instruments are improved with the addition of AI and computation directly into the instrument in what is now called edge computing.
At Argonne, we started working on edge computing before it was called edge computing. To make an instrument smart, by adding a GPU or an AI chip, makes for an interesting research project.
How is it going and where are you in the project?
The project started just before Oct. 1, 2019, about a year ago. Just last week we released the first version of the software stack, Sage 0.1. We have some demo nodes up and running. In the next few months, we plan to have nodes running in California and Colorado. Covid has slowed some of the physical deployments, but the software has been accelerated. It is a three-year project.
What are the challenges?
One of them is that AI is a new area. Applying AI and new algorithms requires a lot of new talented students and faculty. The technology is moving very rapidly; we have to work hard to stay ahead or at least at the curve.
The other challenges are conceptual in nature. This idea of running AI at the edge is not a model everyone is yet comfortable with; it is a new space. We have to write the code and integrate the data streams. It’s a challenge being able to overcome the software limitations we have in computer science and programming.
What is the message for students?
Students need to be multi disciplinary. You can do AI for computer science, then it gets applied. Students need to be conversant in other domains. One example is a snowflake camera. It’s a set of cameras and a flash that takes pictures of snowflakes; they can be characterized by the type of snow coming down. The student working on that project has to know AI, computer vision, computer science, algorithms and the patterns of different snowflakes and atmospheric conditions.)
University of Arizona Researchers to Help Malta with Water Planning
Researchers at the University of Arizona Center for Innovation have announced that a subsidiary company using technology spun out of the university has been awarded a contract by the European Union to study the water needs of the island of Malta, according to a press release from the university.
Malta, the most densely populated European nation, is an archipelago consisting of one main island and several smaller ones located in the Mediterranean Sea between Sicily and Africa. It ranks among the most water scarce nations in the world. Today, with a booming economy and thriving as a popular tourist destination, Malta is experiencing a “perfect storm” that is becoming all too typical across the globe: too little natural freshwater for its growing demand, made worse by more frequent droughts from climate change.
NOAH Arizona LLC will study the feasibility of implementing its patented water management decision support system on the island, working in partnership with Malta’s Energy and Water Agency and Water Services Corporation, the water utility in Malta.
NOAH’s patented water management system combines real-time data streams with AI and optimization models. The system helps identify optimal water management strategies that minimize costs while maximizing water sustainability and quality. NOAH worked with the startup incubation team at the University of Arizona Center for Innovation to identify markets to prove how AI can be applied to water management challenges that nations like Malta are facing.
“This is a wonderful opportunity for NOAH to further advance our system by collaborating with water experts in Malta. They are on the front-line of some of the world’s most serious and difficult water problems,” stated Emery Coppola, NOAH LLC co-founder and president.
Currently, Malta obtains approximately 60n percent of its water supply from three plants using a reverse osmosis (RO) process for water purification. The remaining water demand is met through extraction of diminishing groundwater via 100 production wells.
A major challenge will be to identify the optimal trade-off between RO and groundwater sources among multiple conflicting objectives given that the water quality from RO is superior to groundwater, whereas the production costs required to produce freshwater by RO are higher than groundwater.