Ocean Mover's Distance: Using Optimal Transport for Analyzing Oceanographic Data

by   Sangwon Hyun, et al.

Modern ocean datasets are large, multi-dimensional, and inherently spatiotemporal. A common oceanographic analysis task is the comparison of such datasets along one or several dimensions of latitude, longitude, depth, time as well as across different data modalities. Here, we show that the Wasserstein distance, also known as earth mover's distance, provides a promising optimal transport metric for quantifying differences in ocean spatiotemporal data. The Wasserstein distance complements commonly used point-wise difference methods such as, e.g., the root mean squared error, by quantifying deviations in terms of apparent displacements (in distance units of space or time) rather than magnitudes of a measured quantity. Using large-scale gridded remote sensing and ocean simulation data of Chlorophyll concentration, a proxy for phytoplankton biomass, in the North Pacific, we show that the Wasserstein distance enables meaningful low-dimensional embeddings of marine seasonal cycles, provides oceanographically relevant summaries of Chlorophyll depth profiles and captures hitherto overlooked trends in the temporal variability of Chlorophyll in a warming climate. We also illustrate how the optimal transport vectors underlying the Wasserstein distance calculation can serve as a novel interpretable visual aid in other exploratory ocean data analysis tasks, e.g., in tracking ocean province boundaries across space and time.


page 8

page 10

page 12

page 13

page 16

page 17

page 23


Multilevel Optimal Transport: a Fast Approximation of Wasserstein-1 distances

We propose a fast algorithm for the calculation of the Wasserstein-1 dis...

The GenCol algorithm for high-dimensional optimal transport: general formulation and application to barycenters and Wasserstein splines

We extend the recently introduced genetic column generation algorithm fo...

A Novel Loss Function Utilizing Wasserstein Distance to Reduce Subject-Dependent Noise for Generalizable Models in Affective Computing

Emotions are an essential part of human behavior that can impact thinkin...

A gradual, semi-discrete approach to generative network training via explicit wasserstein minimization

This paper provides a simple procedure to fit generative networks to tar...

Interpretable Model Summaries Using the Wasserstein Distance

In the current computing age, models can have hundreds or even thousands...

Approximation of Wasserstein distance with Transshipment

An algorithm for approximating the p-Wasserstein distance between histog...

Pushing the right boundaries matters! Wasserstein Adversarial Training for Label Noise

Noisy labels often occur in vision datasets, especially when they are is...

Please sign up or login with your details

Forgot password? Click here to reset