HomeMachine LearningMachine Learning NewsUsing ML to map Methane Emissions in California

Using ML to map Methane Emissions in California

Audio version of the article

Using precision instruments and new mapping and machine-learning tools, a research team has been pinpointing sources of the greenhouse gas.

Using ML to map Methane Emissions in California 4

In October 2016, an aircraft equipped with NASA’s Airborne Visible Infrared Imaging Spectrometer – Next Generation (AVIRIS-NG) detected multiple methane clouds emerging from the Sunshine Canyon landfill near Santa Clarita, California. The pillars were large enough for the Jet Propulsion Laboratory (JPL) researchers to inform plant operators and local enforcement agencies, an important step in a process to better track local gas emissions.

Methane is a short-lived but powerful greenhouse gas that has been responsible for around 20 percent of global warming since the industrial revolution. Dairy cows and beef cattle produce methane through their intestines and give it off in burps. Their manure also produces methane, and when stored in manure ponds it can be a major source of emissions. In oil and natural gas production, methane is released from the ground and the infrastructure for storage and transport can leak. And landfills are a source of methane when organic materials are broken down by bacteria in anaerobic conditions.

The condition of California aims to reduce such methane emissions and tries to reduce 40 percent below the 2013 level at the end of this decade. However, to reduce emissions, the state must receive better management in the sources. California Air Resource Board (CARB): The State Agency, supervised air pollution control efforts, traditionally estimates estimated greenhouse gas emissions in the inventory of known emission activities. However, this approach can lose leaks or other volatile emissions, so that carb employees were interested in measuring air emissions to improve greenhouse gas billing and identify reduction options.

The images above show the methane readings from the AVIRISNG instrument during its October 2016 and 2017 flights over Santa Clarita, California. The methane emissions from the Sunshine Canyon landfill are shown in a gradient from yellow to red, with red representing the highest concentrations. The reduction in methane concentrations after the landfill improvements were implemented. The flights were part of the California Methane Survey, an ongoing project to map the sources of methane emissions across the state. But before the flights took off, climate scientist Francesca Hopkins from the University of California, Riverside, and Riley Duren from JPL (now at the University of Arizona) set out to map every potential methane source across the state to better focus on time to be able to limited flight and prioritize observations.

They opted for a GIS-based approach that processed many publicly available spatial data sets to develop a map that could help them quickly match methane plumes with likely sources. The research team organized the potential methane-emitting infrastructure in California into three sectors: energy, agriculture, and waste. The dataset, called Sources of Methane Emissions (VistaCA), contains more than 900,000 entries and is available from the Distributed Active Archives Center of NASA Oak Ridge National Laboratory (ORNL DAAC).

From August 2016 to November 2017, a JPL-based team flew aircraft equipped with the AVIRIS-NG instrument over 22,000 square miles of the state. “Currently there is no methane observing system that can efficiently survey the entire land surface at high resolution,” said Duren. “We had to focus on high-priority areas.” The flight paths were planned so that they would cover at least 60 percent of methane point-source infrastructure in California.

To expedite data analysis, Duren and his colleagues used machine learning techniques (e.g. neural networks) to automatically identify the gaps that were detected during flight. At the same time, Talha Rafiq, a PhD student at UC Riverside, developed an algorithm to remedy the speed of the observations of methane clouds. The technologies allowed the team to share their findings within weeks with facility operators and regulators in California to alert them of fugitive methane emissions and to help accelerate remediation. More than 272,000 individual systems and equipment components were inspected. Emissions less than 0.2 percent of this infrastructure accounted for at least a third of California’s methane inventory, landfills and composting facilities accounted for 41 percent of measured emissions. Duren, Hopkins, and others published their results in Nature in 2019.

In the case of Sunshine Canyon, the landfill operator confirmed the methane emissions and found that they were due to problems with surface coverage and gas separation systems. In the following year, the operator introduced a number of changes that drastically reduced emissions. AVIRISNG confirmed a methane reduction. These results were documented in 2020 by Duren, Daniel Cusworth (project scientist at the University of Arizona) and others in Environmental Research Letters. The survey data can be viewed on the Methane Source Finder portal. The research was funded by NASA’s Advancing Collaborative Connections for Earth System Science program and the Prototype Methane Monitoring System for California at NASA’s Carbon Monitoring System.

Source link

- Advertisment -Using ML to map Methane Emissions in California 5Using ML to map Methane Emissions in California 6

Most Popular

- Advertisment -Using ML to map Methane Emissions in California 7Using ML to map Methane Emissions in California 8