The planetary boundary layer (PBL), the lowest layer of the troposphere, significantly influences weather near the Earth’s surface and holds the potential to enhance storm forecasting and improve climate projections. A research team from Lincoln Laboratory’s Applied Space Systems Group has been studying the PBL with a focus on deploying machine learning for creating 3-D profiles of the atmosphere and resolving its vertical structure to aids drought predictions.
Their work builds upon a series of fast and operational neural network algorithms which the Laboratory developed for NASA missions, including the Time-Resolved Observations of Precipitation structure and storm Intensity with a Constellation of Smallsats (TROPICS) mission, and Aqua, a satellite collecting vital data about Earth’s water cycles. The algorithms, based on “shallow” neural networks, are instrumental in retrieving temperature and humidity data from the satellites, improving the accuracy and expansive global coverage of these observations. The aim is to shift towards “deep” learning techniques to capture a comprehensive three-dimensional image of the atmosphere.
This transition to deep learning and artificial intelligence techniques is expected to deliver a better statistical representation of the 3-D temperature and humidity imagery of the atmosphere. The team collaborated with NASA Goddard Space Flight Center and demonstrated these retrieval algorithms could significantly improve the PBL detail, including the accurate determination of the PBL height.
A key application of this improved understanding of the PBL is in the prediction of droughts. Lack of humidity near the surface, specifically at the PBL level, is a leading indicator of drought. The team at Lincoln Laboratory’s Climate Change Initiative are now working in collaboration with NASA’s Jet Propulsion Laboratory to use these neural networks to enhance drought prediction over the continental United States.
The next phase of the project is to cross-check the deep learning results with direct measurements collected in the PBL using radiosondes, instruments flown on weather balloons. There is an expectation that this improved neural network approach holds promise for superior drought predictions compared to existing indicators and will become a reliable tool for climate scientists in the coming decades.