How Melting Arctic Ice Leads to Methane-Belching Arctic Lakes
Methane is a highly potent greenhouse gas with 80 times the warming power of carbon dioxide over twenty years even though it breaks down faster than carbon dioxide. Scientists estimate that 60% of methane emissions are from human activities (e.g., agriculture—including belching cows, fossil fuels, and decomposition of landfill waste). That leaves 40% of methane emissions from natural processes.
Climate change has brought a new source of methane emissions—melting permafrost and arctic lakes. Many new lakes are being created because of permafrost thawing; as the ice in the soil melts, the downward thrust in the landscape creates lakes called thermokarsts. When the lakes are covered with ice, the methane bubbles are trapped by the ice, but when the ice melts, these lakes become a point source for methane. Climate change is bringing longer ice-free periods and increased methane emissions.
But how can we quantify the amount of methane released? CEGE graduate student Divya Kumawat and MD Zuber have been working on this problem with CEGE Associate Professor Ardeshir Ebtehaj. CSE DSI Director Vipin Kumar and CSE DSI Assistant Director Michael Steinbach have provided machine learning guidance.
First, researchers need to understand the seasonal cycle of freezing and thawing—the phenology—of lake ice. Once the freeze-thaw cycle is better understood, climate models can estimate the amount of methane released from those bubbles trapped by ice.
To tackle the problem, this team explored using L-band passive radiometry observations from the Soil Moisture Active Passive (SMAP) satellite. Despite its advantages, these observations have not been used to study Arctic lakes. The L-band microwave frequency is highly sensitive to water, and its low frequency offers greater penetration depth, enabling observation of the entire water-ice column. Moreover, data can be collected day or night, unaffected by weather or cloud cover, since the atmosphere is nearly transparent to lower-frequency microwaves.
Kumawat developed an algorithm based on autoencoder architecture and statistical change detection method to take the radiometer data to characterize the freeze-thaw cycle of these thermokarsts. In their first study, the team showed that using L-band passive radiometry in retrieving lake-ice status (i.e., frozen, melting, clear water, and freezing condition) is promising. Comparisons between SMAP retrievals and MODIS images of Lake Baikal in Russia and the Great Bear Lake in Canada provided initial visual validation of the method. They showed that the L-band retrievals could reveal the presence of frozen, freezing, melting, and clear water states while MODIS images showed either frozen or clear water (see Figure 1).
The team determined that L-band data weren’t suitable for smaller lakes, but could be used to monitor lakes more significant than 50 km2. New higher resolution L-band data may make it possible to study the phenology of smaller lakes. According to Kumawat, they are now trying to design a multimodal time series algorithm using the radiometry data and satellite images to resolve lake phenology for small lakes.
The team plans to use SMAP satellite observations to quantify methane emissions during ice melt. Although methane bubbles in the ice generate a detectable signal, unresolved radiometric complexities in modeling the interaction between methane bubbles in snow and electromagnetic waves currently hinder accurate estimates. To address this, a newly purchased radiometer will be deployed over Alaskan lakes for experimental studies with team member Katey Walter Anthony, Professor at the Water and Environmental Research Center at the University of Alaska Fairbanks. Our team will mount the radiometer on a drone to collect high-resolution data. This approach will pave the way for current and future L-band satellite missions to estimate global methane emissions from space.
Portions of this project have been funded by a university-wide DSI seed grant and a $.5M NASA grant.