Exploring Deep Sea

1 view

Skip to first unread message

Edward

unread,

Aug 3, 2024, 4:44:13 PM8/3/24

to gazaligtimb

Why We Care
Deep-sea corals, also known as cold-water corals, are corals that do not rely on sunlight, and therefore are typically found in deeper waters (>50 m). Deep-sea corals occur in all oceans, and under favorable conditions, form dense aggregations that create habitat for a myriad of species. These ecosystems are some of the richest and most diverse in the deep sea, yet they are often overlooked by managers, scientists and the general public, who may be unaware that rich and diverse communities can thrive in deep waters under the right conditions. While deep-sea coral ecosystems are out of sight and out of mind to most, they are not immune to anthropogenic impacts, as deep-sea trawling, oil and gas exploration, submarine cable laying and ocean acidification threaten these systems much more than their shallow-water counterparts. Deep-water corals are particularly vulnerable to such impacts, because they are typically slow growing and long lived (on the orders of centuries to millennia), and as a result require very long timescales to recover from disturbances. The objective of SEDCI is to study deep-sea coral ecosystems throughout the Southeast U.S. Region in order to further the management, conservation and protection of these important deep-water ecosystems.

What We Are Doing
SEDCI conducts multiple research expeditions each year, which survey deep-sea coral ecosystems using ships, submersibles, and other equipment. Additionally, SEDCI conducts complementary research projects focused on seafloor mapping, species identification, habitat suitability monitoring, environmental monitoring and data mining. These efforts will provide important baseline information that is needed to support management efforts of deep-sea coral ecosystems.

SEDCI focuses on studying deep-sea (>50 m) coral ecosystems off the Southeast U.S., a region including the U.S. federal waters of the Caribbean, Gulf of Mexico and South Atlantic Bight. This area corresponds to the cumulative extent of the jurisdictions of three fishery management councils, including the Caribbean, Gulf of Mexico and South Atlantic Fishery Management Councils. Furthermore, this area includes numerous areas where deep-water corals are protected, including three national marine sanctuaries, and various other marine protected areas. SEDCI focuses on surveying deep waters in and around these areas in order to support management efforts.

Website Owner: National Centers for Coastal Ocean Science
USA.gov Department of Commerce National Oceanic and Atmospheric Administration National Ocean Service
Copyright 2023 Privacy Policy Accessibility Disclaimer Survey Freedom of Information Act

This ten-day expedition will start and end in Honolulu and utilize the deep-water mapping capabilities of E/V Nautilus to survey previously unmapped seafloor in the U.S. EEZ south of the Main Hawaiian Islands.

The site is secure.
The ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

We present a Galaxy based web server for processing and visualizing deeply sequenced data. The web server's core functionality consists of a suite of newly developed tools, called deepTools, that enable users with little bioinformatic background to explore the results of their sequencing experiments in a standardized setting. Users can upload pre-processed files with continuous data in standard formats and generate heatmaps and summary plots in a straight-forward, yet highly customizable manner. In addition, we offer several tools for the analysis of files containing aligned reads and enable efficient and reproducible generation of normalized coverage files. As a modular and open-source platform, deepTools can easily be expanded and customized to future demands and developments. The deepTools webserver is freely available at -freiburg.mpg.de and is accompanied by extensive documentation and tutorials aimed at conveying the principles of deep-sequencing data analysis. The web server can be used without registration. deepTools can be installed locally either stand-alone or as part of Galaxy.

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

To improve coastal adaptation and management, it is critical to better understand and predict the characteristics of sea levels. Here, we explore the capabilities of artificial intelligence, from four deep learning methods to predict the surge component of sea-level variability based on local atmospheric conditions. We use an Artificial Neural Networks, Convolutional Neural Network, Long Short-Term Memory layer (LSTM) and a combination of the latter two (ConvLSTM), to construct ensembles of Neural Network (NN) models at 736 tide stations globally. The NN models show similar patterns of performance, with much higher skill in the mid-latitudes. Using our global model settings, the LSTM generally outperforms the other NN models. Furthermore, for 15 stations we assess the influence of adding complexity more predictor variables. This generally improves model performance but leads to substantial increases in computation time. The improvement in performance remains insufficient to fully capture observed dynamics in some regions. For example, in the tropics only modelling surges is insufficient to capture intra-annual sea level variability. While we focus on minimising mean absolute error for the full time series, the NN models presented here could be adapted for use in forecasting extreme sea levels or emergency response.

In order to improve coastal adaptation, it is critical to better understand and predict the characteristics of sea levels, such as their temporal variation and magnitude over long time periods, typically multiple decades1,2. Coastal sea level variability results from a combination of multiple processes, including mean sea level variations, tides, waves and storm surges resulting from the passing of low pressure systems and strong winds3,4. We focus here on the non-tidal residual, also referred to as surge or surge residual, i.e. the water level after removal of the tide and mean sea level. At hourly time scales or shorter, temporal variation in surge residuals depends on the direction and variations in wind and pressure gradients, as well as local topographic characteristics such as the bathymetry and complexity of the coastline5,6,7. Surge levels superimposed on high tides can exceed land thresholds and contribute to nuisance flooding or extreme impacts when caused by tropical or extratropical cyclones. At the global scale, studies have used hydrodynamic modelling8 or data-driven approaches to reconstruct surge time series9,10,11,12. The advantage of hydrodynamic models is that with adequate model resolution and meteorological forcing, they can resolve physical coastal processes and their interactions. They are also valuable for understanding epistemic uncertainties and the relative contributions of different oceanographic and coastal processes in total water levels. However, these models are computationally demanding and take a long time to set up13,14,15. This limits their ability to be used in the simulation of large ensembles of events16,17.

In this paper, we explore the capability of different deep learning approaches to predict surge levels at the global scale. To do so, we predict hourly surge using four types of NN and evaluate their predictive skill. We train, validate and test a CNN, LSTM and combined CNN-LSTM (ConvLSTM) model to capture spatial, temporal, and spatio-temporal dependencies for surge level observations from 736 tide stations. We benchmark our NN models with a simple probabilistic reference model based on climatology. Next, for 15 selected locations with diverse surge characteristics, we examine the NN skill gained from increasing the spatial extent considered around each location and the number of different variables used as predictors. Additionally, we show the capability of the four NN types to gain skill when adding complexity to their respective network architecture.

We predict hourly surge at tide stations from the Global Extreme Sea-Level Analysis Version 2 database (GESLA-2)37 using four different deep learning models following the main steps described in this section. In brief, we extract the predictand from the GESLA tide stations and predictor variables from the atmospheric reanalysis ERA5 from ECMWF28. For each station, we construct and run four NN model types and compare their performance with observed surge levels and with a simple probabilistic model as benchmark. Finally, for fifteen stations, we analyse the influence of the number of predictor variables, the spatial extent considered around each location (from hereon called spatial footprint), and the architecture of the NN on its performance.

Four different types of NN models are set up and trained to predict surges: an Artificial Neural Network (ANN), a Long Term Short Term Neural Network (LSTM), a Convolutional Neural Network (CNN), and a Convolutional LSTM (ConvLSTM), which is a combination of the latter two. In this section, we provide an overview of their specific features and selected architecture. For more detailed information, the reader is referred to the accompanying references shown in the following section.

The ANN, the most general form of NN, has been extensively applied in various fields of science to capture nonlinear processes31. The LSTM is a derivation of the Recurrent Neural Network (RNN) in that it captures sequence-to-sequence patterns in their internal state as memory, but has advantages over the conventional RNN as they can selectively store long term information31,36,50,51. The CNN is a class of NN models that works well in capturing spatial features, shapes and texture due to its shared weight architecture and is thus often applied for image recognition purposes31,33,52,53. We also apply a ConvLSTM in order to capture spatiotemporal information. The ConvLSTM combines sequence-to-sequence learning with convolutional layers and emerges in current studies with promising applications in predicting spatiotemporal information54.