How ISS’s new AI-powered program will help real-time monitoring of the climate crisis
- Entrepreneurs are leveraging new technologies such as MI to predict outbreaks as the commercial space industry grows.
- Metaspectral has developed a platform that relies on real-time AI analysis of hyperspectral satellite imagery.
- Ample data and the ability to analyze and use it to our advantage is one of the most important tools we have at our disposal.
The world is in a climate crisis. With average global temperatures increasing every year, the threat of seasonal forest fires is becoming increasingly worse. In places like the Pacific Northwest, wildfire season causes extensive damage to woodlands, rural communities, and townships, destroying farmlands and infrastructure and forcing hundreds of thousands of residents to flee their homes.
These fires also lead to terrible air quality in cities located hundreds (or even thousands) of miles away. For instance, in September of 2022, the city of Vancouver (British Columbia) was ranked as having the worst air quality in the world - per the Air Quality Index (AQI). This resulted from wildfires raging in the interior, plus smoke from wildfires in the U.S. moving north.
In October, Seattle was similarly ranked as having the worst air quality in the world. Once again, the culprit was local wildfires and those happening further south. And cities like San Francisco and Los Angeles, which have made significant gains in air quality in recent years, are seeing those gains offset by record-setting wildfires raging across California.
To mitigate this problem, government agencies are turning to Earth observation satellites to monitor wildfires so they can prevent them from becoming uncontrollable. With the rise of the commercial space industry, entrepreneurs are leveraging new technologies (like machine learning) to predict outbreaks.
This includes Metaspectral, a Vancouver-based software company that has created a platform based on real-time AI analysis of hyperspectral satellite imagery. It enables the platform to anticipate potential forest fires and designate high-risk areas. This will allow firefighters and civil authorities to prevent outbreaks and focus their resources on the most affected areas.
The technology is planned for deployment on the International Space Station (ISS). It is also being adopted by the Canadian Space Agency (CSA) to analyze hyperspectral image data from the ISS and Earth observations satellites in Low Earth Orbit (LEO). As Metaspectral co-founder and CEO Francis Doumet shared with Interesting Engineering via email:
"Metaspectral is advancing the field of AI and has already been awarded a patent for using AI to compress hyperspectral imagery. In the recycling industry, the company has made significant advances in better identifying and sorting hard-to-recycle materials with AI. In the commercial space industry, Metaspectral is enabling space companies to maximize the operational capacity of space assets by allowing them to capture and transmit more data within the same bandwidth while also saving on operational costs by minimizing the data that needs to be transmitted to the ground."
A warming planet
According to the National Oceanic and Atmospheric Administration (NOAA), atmospheric carbon dioxide (CO2) levels exceeded 400 parts per million (ppm) in May 2013 for the first time since the Pliocene Era (ca. 5.4 - 2.4 million years ago). In fact, before then, CO2 levels had not exceeded 300 ppm for the past 300,000 years or so.
According to Sixth Assessment Report (AR6) prepared by the United Nation's Intergovernmental Panel on Climate Change (IPCC), the average surface temperature in 2021 was 1.51 °F (0.84 °Celsius) warmer than the twentieth-century average of 57.0 °F (13.9 °C), and 1.87 ˚F (1.04 ˚C) warmer than the pre-industrial period (1880-1900). The last time global temperatures reached these levels was 125,000 years ago, before the most recent glacial period. Moreover, the nine years from 2013 through 2021 rank among the ten hottest years on record.
Under normal circumstances, fires are part of the natural cycle of rainfall, dryness, and lightning in forested regions, grassland, and prairies. But with rising temperatures worldwide, this cycle has been disrupted, resulting in less precipitation during the summer, drier conditions, and droughts. Along with high winds, this trend drastically increases the risk of wildfires.
There's also the issue of intentional fires, which are part of land clearance efforts in many parts of the world. In the Amazon rainforest and the Cerrado grassland/savanna region of South America, farmers engage in "agricultural burning." The same is true in Southeast Asia and across Africa (especially in equatorial Africa), where farmers set fires in the late winter and early spring to clear the land of underbrush ahead of the growing season.
These fires generate large amounts of smoke pollution and greenhouse gases and can lead to the destruction of ecosystems. They can also disrupt transportation and communications, damage critical infrastructure (electricity, gas, and water), and cause the loss of property, crops, animals, and people.
Consequently, seasonal wildfires have become yet another environmental crisis plaguing communities worldwide. Just as oceanfront communities are subject to increased storm activity and flooding and places like the Midwestern U.S. ("Tornado Alley") are threatened by more frequent extreme weather events, communities within or near wooded areas are threatened by periodic wildfires.
EOS and thermal imaging
For decades, NASA and the Canadian Space Agency (CSA) have relied on Earth Observation Satellites (EOS) to monitor wildfires using infrared imaging cameras. Examples include Landsat-7 and its Enhanced Thematic Mapping Plus (ETM+) instrument, the Aura satellite and its Tropospheric Emission Spectrometer (TES), and Terra, which relies on its Moderate Resolution Imaging Spectroradiometer (MODIS). As Doumet explained:
"Earth Observation is vital for battling wildfires because it can analyze areas that are already burning and areas that have a high probability of succumbing to wildfires. Entire regions can be analyzed by remote sensing technology with hyperspectral sensors to quantify and characterize the level of "fuel" on the ground, including flammable materials such as dry or dead trees and plants."
"This allows authorities to identify the areas with the highest probability of a fire breaking out with greater accuracy than what could be determined using other methods like historical data. This becomes even more useful when the data is transmitted in real-time because earth observation can serve as an early warning system to alert ground forces as soon as a small fire erupts."
Over time, advances in infrared astronomy, machine learning, sensor design, hardware, software applications, and data science are transforming the field of satellite imaging and remote sensing technology. A key advancement is the development of "hyperspectral" imaging technology. Whereas conventional thermal cameras are generally restricted to taking infrared images in three spectral bands, "hyperspectral" imaging collects and processes information from across the electromagnetic spectrum (in over 300 bands).
In astronomy, this technique is commonly referred to as integral field spectroscopy, which is used by two key instruments on the Very Large Telescope (VLT) in Chile - the Fibre Large Array Multi Element Spectrograph (FLAMES) and the Multi-Unit Spectroscopic Explorer (MUSE) - as well as the Advanced CCD Imaging Spectrometer (ACIS) on NASA's Chandra X-ray Observatory.
Data compression and AI
Gathering information in over 300 spectral bands results in a tremendous volume of data that requires sophisticated analysis. This is where Metaspectral's unique approach comes into play, a combination of advanced data compression algorithms and novel machine learning models. These come together in the Metaspectral Fusion platform that analyzes hyperspectral data and streams it in real-time.
Machine Learning Operations (MLOps) agents also monitor and continuously improve the AI models, which run inference (or predictions) at the edge of (or in) the Cloud in real-time. Said Doumet:
"Hyperspectral images and their analysis produce an enormous amount of data - gigabits per second. Metaspectral's proprietary compression and deep learning algorithms allow the platform to process and analyze this data in real-time while performing pixel-by-pixel level analysis. By creating access to real-time data, the platform allows businesses, governments, or other users to benefit from insights within seconds of the data being captured instead of the days it has historically taken."
This allows high-risk areas to be designated based on their spectral signatures, quantifying an area's healthy trees and healthy grass versus its dead trees and dead grass. Users can then quantify the available fuel for forest fires and preemptively deploy resources or mitigation efforts.
Another key development has been the emergence of the commercial space industry (aka. NewSpace), which has led to lower costs and increased access to space. In addition to reusable rockets, cheaper launch services, and rideshare programs, this trend is allowing private companies to use technologies previously reserved for government entities.
In 2023, several NewSpace companies will commence operations with their own hyperspectral sensors, providing a wealth of information for various applications ranging from mining and defense to agriculture. These latter applications will play a vital role in the coming decades as climate change forces governments and humanitarian organizations to support more people with fewer resources. Said Doumet:
"In the areas of climate change, hyperspectral data from space will allow farmers to precisely track plant health and soil moisture to only use the exact amount of fertilizer and water required to maximize yield. It will also allow farmers to precisely measure the carbon sequestration capacity of their fields, which will empower them to tie into, and benefit from, emerging carbon credit systems."
Battling climate change
In addition to monitoring and anticipating wildfires, real-time EOS monitoring has a range of applications that could help mitigate. All over the world, the impact of rising temperatures and feedback mechanisms are triggering ecological and humanitarian crises.
"Earth Observation with hyperspectral imagery can be used to monitor environmental hazards, such as methane leaks or oil spills, and even qualify plastic levels on the surface of the ocean," said Doumet. "Metaspectral has also successfully used Earth Observation data to quantify the level of CO2 and other greenhouse gasses at the Earth's surface, which allows farmers to measure the carbon sequestration capability of their farms."
In the near future, an EOS payload developed by Metaspectral and Miami-based remote sensing developer HySpeed Computing will be sent to the International Space Station (ISS). Known as the Onboard Programmable Technology for Image Classification and Analysis (OPTICA,) this payload will demonstrate real-time compression, streaming, and analysis of hyperspectral data from Low Earth Orbit (LEO).
OPTICA is scheduled for launch in early 2023 as part of the SpaceX CRS-27 mission and will spend the next six months deployed aboard the ISS. The mission is sponsored by the ISS National Laboratory, which collaborates with NASA to promote research and development activities aboard the ISS.
"The objective of this mission is to demonstrate the real-time streaming and analysis capability of hyperspectral data," said Doumet. "This has historically been very challenging, taking hours if not days to complete. We hope to change this. Once this has been successfully demonstrated on the ISS, it will enable similar deployments in free-flying earth observation and deep-space missions."
In June, Metaspectral won funding from the Canadian Space Agency (CSA) as part of their smartEarth initiative to develop a method for systematically and methodically quantifying ground-level carbon dioxide (CO2) levels. The ability to provide accurate, real-time data on greenhouse gas (GHG) levels has several potential applications, running from environmental monitoring to measuring emissions from oil and gas pipelines, leaks from deepwater drilling activity, and other hazards.
Earth observation and satellite imaging have advanced considerably in recent decades, paralleling advancements in optics, spectroscopy, software, and data analysis with AI. The benefits are new applications for monitoring Earth's changing climate, coordinating responses, anticipating crises, and mitigating them.
In a rapidly-changing world, ample data and the ability to analyze and use it to our advantage is one of the most important tools we have at our disposal. As Doumet put it:
"Metaspectral's mission is to solve the world's most complex problems using advanced sensing technology and deep learning. For the last decade, computer vision has focused on recreating a human's ability to see, learn, and sense. Metaspectral's purpose is to take this a step further and create computer vision that far exceeds our own capabilities. This superhuman vision sees what conventional cameras and humans cannot, enabling us to advance humanity with better decision-making based on more accurate and plentiful data."
Can AI rescue lab rats and guinea pigs? Science’s long-debated ethical question may finally have some answers.