We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
Capturing the Diversity of Mesoscale Trade Wind Cumuli Using Complementary Approaches From Self‐Supervised Deep Learning.
- Authors
Chatterjee, Dwaipayan; Schnitt, Sabrina; Bigalke, Paula; Acquistapace, Claudia; Crewell, Susanne
- Abstract
At mesoscale, trade wind clouds organize with various spatial arrangements, shaping their effect on Earth's energy budget. Representing their fine‐scale dynamics even at 1 km scale climate simulations remains challenging. However, geostationary satellites (GS) offer high‐resolution cloud observation for gaining insights into trade wind cumuli from long‐term records. To capture the observed organizational variability, this work proposes an integrated framework using a continuous followed by discrete self‐supervised deep learning approach, which exploits cloud optical depth from GS measurements. We aim to simplify the entire mesoscale cloud spectrum by reducing the image complexity in the feature space and meaningfully partitioning it into seven classes whose connection to environmental conditions is illustrated with reanalysis data. Our framework facilitates comparing human‐labeled mesoscale classes with machine‐identified ones, addressing uncertainties in both methods. It advances previous methods by exploring transitions between regimes, a challenge for physical simulations, and illustrates a case study of sugar‐to‐flower transitions. Plain Language Summary: Clouds are a fundamental player affecting our planet's energy balance, making their accurate representation crucial in climate models. One open question is how they organize on a scale of a few 100 km (mesoscale) in the tropical northern Atlantic region east of Barbados. Satellite observations can help to categorize these clouds, but previous methods had limitations in capturing the full range of cloud arrangements and transitions between different cloud forms. We have introduced a novel approach that utilizes machine learning and geostationary satellite data to address this issue. Our machine learning model autonomously learns to recognize various cloud patterns and distributions. We conducted a comparative analysis between the categories generated by the machine and those identified by human experts to understand the strengths and weaknesses of both methods. Additionally, we explore a case study where clouds undergo a transformation, changing from a structure resembling sugar to one resembling flowers. This particular transformation was found difficult to capture with physical simulation before. The clear signatures of the transition identified by our machine learning approach can help to better understand cloud evolution, which is crucial for improving climate models and predicting how cloud behavior may change in a changing climate. Key Points: Mesoscale cloud organization can be taxonomized by a two‐step deep learning approach in the feature space continuumComparing seven machine‐identified classes with humans' four recognized categories underlines the significance of uncertainty estimatesNew diagnostic is provided to analyze the temporal transition between regimes, as illustrated for human‐labeled sugar‐to‐flower regimes
- Subjects
BARBADOS; DEEP learning; TRADE winds; MACHINE learning; GEOSTATIONARY satellites; ATMOSPHERIC models; SPATIAL arrangement
- Publication
Geophysical Research Letters, 2024, Vol 51, Issue 12, p1
- ISSN
0094-8276
- Publication type
Article
- DOI
10.1029/2024GL108889