Probabilistic Machine Learning Estimation of Ocean Mixed Layer Depth From Dense Satellite and Sparse In Situ Observations

Abstract

The ocean mixed layer plays an important role in the coupling between the upper ocean and atmosphere across a wide range of time scales. Estimation of the variability of the ocean mixed layer is therefore important for atmosphere‐ocean prediction and analysis. The increasing coverage of in situ Argo profile data allows for an increasingly accurate analysis of the mixed layer depth (MLD) variability associated with deviations from the seasonal climatology. However, sampling rates are not sufficient to fully resolve subseasonal (<90 $< 90$ day) MLD variability. Yet, many multivariate observations‐based analyses include implicit modeled subseasonal MLD variability. One analysis method is optimal interpolation of in situ data, but the interior analysis can be improved by leveraging surface data with regression or variational approaches. Here, we demonstrate how machine learning methods and satellite sea surface temperature, salinity, and height facilitate MLD estimation in a pilot study of two regions: the mid‐latitude southern Indian and the eastern equatorial Pacific Oceans. We construct multiple machine learning architectures to produce weekly 1/2° gridded MLD anomaly fields (relative to a monthly climatology) with uncertainty estimates. We test multiple traditional and probabilistic machine learning techniques to compare both accuracy and probabilistic calibration. We validate our methodology by applying it to ocean model simulations. We find that incorporating sea surface data through a machine learning model improves the performance of spatiotemporal MLD variability estimation compared to optimal interpolation of Argo observations alone. These preliminary results are a promising first step for the application of machine learning to MLD prediction.

Incorporating Uncertainty Into a Regression Neural Network Enables Identification of Decadal State‐Dependent Predictability in CESM2

“…This study introduces the identification of state‐dependent predictability on decadal timescales using a regression‐based neural network to predict sea surface temperatures (SSTs) across the globe within the Community Earth System Model, version 2 (CESM2, Danabasoglu et al., 2020) pre‐industrial control simulation. We demonstrate a powerful technique for incorporating uncertainty into the prediction of regression neural networks which has previously only been used a handful of times in climate science (Barnes & Barnes, 2021; Foster et al., 2021; Guillaumin & Zanna, 2021). We further leverage this uncertainty output to identify which initial states are associated with the lower uncertainty predictions.…”

Data-driven Modelling of Engineering Systems with Small Data, a Comparative Study

“…Having dense data does not contradict with small size of the data. In data-driven modelling, the inputs of a dense data set are distributed in the input space rather uniformly [22]. Such a data set is small, if its size (number of samples) is small compared to the number of parameters of an appropriate model for such a problem.…”

Sinh-arcsinh-normal distributions to add uncertainty to neural network regression tasks: Applications to tropical cyclone intensity forecasts

“…Although the approach was introduced decades ago (Nix and Weigend, 1994, 1995) and is a standard in the computer science literature (e.g., Sountsov et al, 2019; Duerr et al, 2020), it is much less known in the earth science community. A few applications of the method appear in the recent literature (e.g., Barnes and Barnes, 2021; Foster et al, 2021; Guillaumin and Zanna, 2021; Gordon and Barnes, 2022), but all of these focus strictly on the Gaussian distribution.…”