When do ship tracks not form?

Abstract:

Studied for many decades now, ship-tracks are a classic example of aerosol-cloud interactions. These perturbations to cloud albedo by aerosol emitted from ship exhaust are however very local and attempts to understand their relevance in terms of estimating global cloud droplet number and liquid water sensitivities flounder on the question of their representativeness. Here we describe our efforts to determine how accurately ship-tracks, which form under specific environmental conditions, reflect the broader forcing due to aerosol cloud interactions.

Machine learning is transforming many areas of science by providing new tools to analyse and understand the huge volumes of data that new instruments and models can provide. One particular area where machine learning has made rapid progress is in object detection. Robust techniques are now available which can quickly identify objects in images without any need for the thresholding or edge detection which have been used in the past and often struggles with inhomogeneous features.

Using a deep convolutional neural network trained on several existing hand-logged datasets we have created a dataset of ship-tracks from across the East Pacific using Geostationary Operational Environmental Satellite (GOES)-16 for the year of 2018. Combining this high temporal resolution satellite data with an Automatic Identification System (AIS) shipping inventory we can estimate for the first time the sensitivity of clouds to shipping emissions, both where (geographically and under which environmental conditions) the sensitivity is apparent and, crucially, where it is not. Implications for estimates of ERFaci and other applications of these novel machine learning techniques are also discussed.