Oceans are changing fast, and we do not always understand the how or why. Yet, oceans play a crucial role in our global ecosystem, and these changes impact us all. The problem is, we have less insight into our ocean’s inner operating mechanisms in many respects than we have on our galaxy. So what’s the solution? Well, Capgemini and the Norwegian Institute of Marine Research have taken on the challenge to use machine learning and AI to read, analyze and interpret vast amounts of data collected hundreds of meters below sea level, thus gaining a better understanding of events and inner workings of the ocean’s mechanisms.
Further, what works for the oceans indeed might work for businesses too. Despite the significance of our oceans, vast parts of the marine ecosystems remain unexplored. There are tremendous unknowns in understanding biophysical processes, recruitment processes, and the vulnerability of these systems. Blanket darkness in deep waters is challenging as it makes it impossible to use optical techniques for long-range observations. Nonetheless, the 3D underwater world is highly dynamic with tidal currents, mixing water masses and nutrients, and large vertical and horizontal movements of plankton and fish.
Here’s the big secret. In contrast to light, sound may propagate long distances in water, and several fishes and, in particular, whales are using sound for communication, navigation, breeding, and identification of prey. Humans can mimic, however still poorly, the sound expertise of whales by developing tools to observe the underwater world.
Sensing the LoVe Ocean Observatory
LoVe (Lofoten-Vesterålen) Ocean Observatory situated in the ecological hotspot off the Northern Coast of Norway has placed sensor platforms across the continental shelf, the dominant currents and whale and fish migration routes, e.g., the world largest cod spawning migration to and from spawning grounds are passing the transect to explore the unknown.
Equally important, the observatory uses subsea sensor equipment to perform accurate time recording of many advanced underwater sensor data. Broadband echo-sounders are used to identify organisms in the water column, while hydrophones are used for listening to natural and manufactured sounds. In addition, a hydroacoustic current meter monitors water currents. Other sensors at LoVe Ocean Observatory measure CO2, chlorophyll, oxygen, temperatures, salinity, pressure, and more. Moreover, up-close images are taken of deep-water corals and other seabed habitats.
These measurements provide fundamental new insights into marine processes, including the behavioral dynamics of organisms and variation concerning physical and chemical properties. It gets better. They enhance understanding of relationships between physical drivers and biological response, a prerequisite for models that reflect actual dynamics in the marine ecosystem. Real-time streaming of data supports better modeling, a fundamental tool for handling hazards like oil spills and evaluating impacts of human activity and exploitation of marine ecosystems. Invariably, detailed information helps to understand the consequences of climate variation and associated trends.
Help from machine learning and AI
The growing ability to measure an Increasing number of dynamic characteristics within marine ecosystems in real-time drives the need for ever more intelligent data analysis methods and techniques. Issues ranging from vast amounts of multi-dimensional, real-time data streams to sensor-drift, incomplete data, replaced sensor output shifts or addressing re-calibrated equipment elegantly.
Thus, there is a crucial need to harness the power of machine learning and AI techniques to analyze sensor data quality, detect sensors, identify an occurrence of ecosystem events and provide alerts for rare and harmful events.
Understanding our oceans is vital. Despite the complexity and differences between marine ecosystems worldwide, global solutions are needed to identify abrupt events or, even more importantly, to recognize gradual changes in marine ecosystems affected by worldwide warnings and the cumulative effect of man-made stressors on the ecosystem.
Tackling data science challenges
Capgemini’s yearly Global Data Science Challenge has gathered over thousands of consultants and data scientists to contribute to the Norwegian Institute of Marine Research (NIMR), helping them to create a step forward towards the goal of an improved understanding of our Oceans and contributing to the United Nations decade for Ocean science for sustainable development.
Together we build solutions for accessing, interpreting, analyzing, and reporting on information contained in terabytes of data collected hundreds of meters below sea level.
Case in point, insights gained range from ways of processing data collection under extreme circumstances, analysis of vast amounts of time-series data on many diverse groups of granularity, unraveling patterns of bio-marine life migration related to many different features of underwater life, and more. The result? Insights, data, and methods are shared with the community for the greater good.
Learning about the oceans by inducing events and theories in a data-driven manner is challenging. Not only is monitoring a complex ecosystem in change a problematic task, inducting system behavior from vast amounts of sensor data is not trivial either. Observable and provable through multi-dimensional data are events such as algae-blooming due to seasonal effects or cod migration patterns in the northern Atlantic. You can correlate these local events with expected seasonal influences and locations for events to support conclusions on the absence or presence of climate change, the status of the gulf stream, or other super events.
Those events are published and eventually correlated with other analyzed events across the globe. Also, mechanisms to (semi-)automate such event publishing and cross correlation are yet to be implemented. Currently, we experience a global uptake of monitoring oceans. UNESCO, the Global Ocean Observing System, and LoVe observatory and private-owned foundations like C4IR Ocean show the way.
A beneficial choice for global and open publishing of analyzed events using semantic web standards would make event reports machine-readable and allow for automated alignment and merging of found trends and facts. Opening and sharing all our data on our oceans and co-operating on their analysis and interpretation globally is crucial for sustainable co-inhabiting species on our planet. Indeed, it’s an effective form of guidance for many business endeavors as well.
The big unknown
Oceans remain largely unexplored and still hide many secrets, but increasing amounts of relevant data are becoming available.
Understanding is critical
Without deeper insights into the mechanisms of balancing and keeping oceans as healthy and self-maintaining ecosystems, we will not be sustainable over the long term.
The Lofoten-Vesterålen Ocean Observatory generates tons of data. Capgemini and the Maritime Research Institute co-operate using state-of-the-art ML and AI techniques to unravel raw data for information modeling.
Robert Engels, CTO, Insights & Data, Capgemini
Espen Johnsen, Project Leader, LoVe – Ecosystem Acoustic