Detecting Spatial Patterns of Peatland Greenhouse Gas Sinks and Sources with Geospatial Environmental and Remote...

Area of Science:

• Environmental Science
• Remote Sensing
• Climate Change Research

Background:

• Peatlands are critical regulators of major greenhouse gases (GHG): methane (CH4), carbon dioxide (CO2), and nitrous oxide (N2O).
• Understanding the spatial distribution of GHG sinks and sources in peatlands is essential for effective land use-based climate change mitigation strategies.
• Geospatial environmental data offer insights, but the utility of satellite remote sensing data for predicting GHG patterns remains underexplored.

Purpose of the Study:

• To predict the spatial distribution of methane, carbon dioxide, and nitrous oxide sinks and sources across Finland.
• To compare the predictive accuracy of geospatial environmental data, remote sensing data, and a combination of both using machine learning.

Main Methods:

• Utilized 143 field measurements for model training and validation.
• Employed MaxEnt machine-learning modeling to predict GHG spatial patterns.
• Compared three datasets: geospatial environmental data (climate, topography, habitat), remote sensing data (Sentinel-1, Sentinel-2), and a combined dataset.

Main Results:

• The combined dataset achieved the highest predictive accuracy (average test AUC 0.845, stability AUC 0.928).
• Geospatial environmental data alone showed slightly lower accuracy (test AUC 0.810, stability AUC 0.924).
• Remote sensing data alone resulted in reduced predictive accuracy (test AUC 0.763, stability AUC 0.927).

Conclusions:

• Reliable estimation of GHG sinks and sources in peatlands cannot be achieved using only remote sensing data.
• Integrating multiple data sources, particularly geospatial environmental and remote sensing data, is crucial for accurate GHG spatial pattern predictions.
• This approach supports informed climate change mitigation strategies in the land use sector.