Climate and Sustainability: Using Weather Satellites to Monitor CO2
Since the late 1950s, scientists have been keeping a close eye on carbon dioxide (CO2) levels thanks to regular observations from Hawaii’s Mauna Loa Observatory. This effort gave birth to the renowned Keeling Curve, a visual representation of the rising global CO2 concentrations in Earth’s atmosphere. To effectively tackle the challenge of human-induced greenhouse gas emissions, it’s crucial to track the variations of CO2 across different regions and periods. While current space-based CO2 sensors like NASA’s Orbiting Carbon Observatory-2 (OCO-2) offer high-precision data, their coverage is limited, mapping only a small portion of Earth’s surface and revisiting each location every 16 days.
In contrast, geostationary satellites, such as the GOES East satellite, provide a broader perspective by orbiting the Earth at higher altitudes and scanning an entire hemisphere every 10 minutes, primarily for weather forecasting. However, these satellites weren’t originally designed to map CO2 concentrations.
Innovative AI Solutions for CO2 Monitoring
In a groundbreaking development, Google researchers have leveraged the Empirical Research Assistance (ERA) to create a physics-guided single-pixel neural network. This sophisticated model extracts column-averaged CO2 signals from existing GOES East observations. By combining data from 16 wavelength bands of GOES East with lower tropospheric meteorology, solar angles, and the day of the year, the model can estimate CO2 concentrations with remarkable precision. After training on sparse observations from OCO-2 and OCO-3, it provides comprehensive CO2 estimates everywhere, every 10 minutes.
Validation and Impact
Research findings shared at the International Workshop on Greenhouse Gas Measurements from Space highlight the AI model’s ability to harness the high spatial and temporal density of GOES East observations. This achievement allows for the tracking of column-averaged CO2 with unprecedented spatial and temporal resolution. Comparisons with independent data from additional years of OCO-2 observations and the Ground Column Total Carbon Observation Network validate the model’s capability to accurately capture true CO2 variability.
These promising results underscore the potential of AI algorithms to extract additional value from existing observation instruments, especially in the context of resource-intensive satellite research missions. This project is just one of many initiatives Google researchers are pursuing to address climate and greenhouse gas-related questions using ERA.
For more information, visit the source link Here.
“`
