English version below
Für viele Menschen war es schwer zu verstehen, warum man im Frühjahr so weit nach Norden in die Labradorsee reist, um dort Forschung zu betreiben. Das Leben an Bord ist anstrengend und wird durch die harschen und schnell wechselnden Wetterbedingungen zusätzlich erschwert, was besonders den Einsatz unserer Forschungsgeräte kompliziert macht.
Worum geht es also bei unserer Expedition?
Diese Forschungsreise verfolgt zwei Hauptziele: die Untersuchung kleinräumiger Strukturen im Ozean und die Beobachtung großräumiger Meeresströmungen.
In den letzten Jahren haben Wissenschaftler erkannt, dass kleinräumige Strukturen im Ozean, wie Wirbel und Fronten, eine sehr wichtige Rolle spielen. Sie können starke Veränderungen in Temperatur und Salzgehalt verursachen, aber auch in anderen Eigenschaften wie Chlorophyll und dem Export von Kohlenstoff. Während dieser Fahrt nutzen wir eine Reihe von Instrumenten, um diese Strukturen detailliert zu beobachten und besser zu verstehen, wie sie funktionieren.
Das zweite Ziel besteht darin zu untersuchen, wie sich die Meeresströmungen im Atlantik im Laufe der Zeit verändern. Ein zentraler Bestandteil davon sind Langzeitbeobachtungen am 53°N-Observatorium. Dort betreiben wir sieben Verankerungen, lange Kabel, die am Meeresboden befestigt und mit Instrumenten ausgestattet sind, welche Temperatur, Salzgehalt, Sauerstoff und Strömungsgeschwindigkeit messen. Alle zwei Jahre bergen wir diese Verankerungen, sammeln die Daten ein und setzen sie anschließend erneut aus, damit die Messungen fortgeführt werden können.
Eine solche Expedition benötigt lange Vorbereitungszeit und einiges an Organisation auf See, deshalb habe ich unserem Chef-Wissenschaftler einige Fragen gestellt:
Wann haben die Vorbereitungen für die Expedition begonnen? Und wie lief das ab?
Der Antrag für diese Forschungsreise wurde bereits 2023 eingereicht. Darin wurden die Motivation, die Forschungsfragen und der Plan beschrieben, die Reise 2025 durchzuführen. Letztendlich wurde sie dann für 2026 angesetzt.
Die detaillierten Vorbereitungen begannen ungefähr ein Jahr im Voraus, also etwa im April 2025. Die Planung einer Forschungsreise umfasst mehrere Schritte. Wir müssen die Logistik organisieren, entscheiden, wer Teil des wissenschaftlichen Teams sein wird, und die wissenschaftlichen Arbeiten planen, die wir durchführen möchten. Und natürlich gehört auch eine ganze Menge Papierkram dazu 
Wann entstand die Idee, die Expedition im Frühling und nicht wie üblich im Sommer durchzuführen?
Die Idee entstand bei einem Projektantrag, den ich 2022 geschrieben habe. Er beinhaltete eine Forschungsfahrt zur Untersuchung kleinräumiger Ozeanstrukturen und ihrer Verbindung zur Frühjahrsblüte in der Labradorsee. Da wir speziell an der Frühjahrsblüte interessiert waren, war es wichtig, zu dieser Jahreszeit hier zu sein.
Das bedeutete, die Reise im Frühling zu planen, obwohl uns bewusst war, dass die Bedingungen schwieriger sein können als im Sommer. Aber wenn ich es noch einmal machen müsste, würde ich vorher einen Wetterbericht suchen, der einen deutlich ruhigeren März und April verspricht.
Was findest Du daran am interessantesten?
Hier draußen mitten im Ozean zu sein und die Daten, die wir sammeln, in Echtzeit zu betrachten. Es hat etwas ganz Besonderes, wenn die Messungen hereinkommen und man weiß, dass man den Ozean genau in diesem Moment beobachtet.
Wie werden die Entscheidungen zwischen Dir und dem Kapitän getroffen – in Bezug auf Wetter, Forschung und Sicherheit? Wann treffen Ihr euch? Und wie oft?
Von Anfang an haben wir vereinbart, etwa 36 Stunden im Voraus zu planen, angesichts des Umfangs der Arbeiten und der oft schwierigen Wetterbedingungen. Falls nötig, passen wir den Plan anschließend an.
Wir treffen uns jeden Morgen, ohne feste Uhrzeit, um gemeinsam den Wetterbericht anzuschauen und zu entscheiden, was machbar ist und was nicht. Bisher hat dieses Vorgehen sehr gut funktioniert. Gelegentlich mussten wir Arbeiten kurzfristig abbrechen, aber wir konnten uns immer anpassen, ohne viel wertvolle Forschungszeit zu verlieren.
Läuft die Expedition bisher wie geplant? Falls nicht, worin unterscheidet sie sich?
Die Daten, die wir bisher gesammelt haben, haben meine Erwartungen bereits übertroffen besonders angesichts der schwierigen Wetterbedingungen. Wir konnten sehr viel erreichen, und das liegt vor allem an der hervorragenden Zusammenarbeit zwischen der Schiffscrew und dem wissenschaftlichen Team an Bord.
Alle waren sehr flexibel und unterstützend, was es uns ermöglicht hat, uns schnell anzupassen und die verfügbare Zeit bestmöglich zu nutzen.
Tipp Nummer 1 für die Arbeit bei 10 Beaufort und 6 Meter hohen Wellen?
Immer eine Hand fürs Schiff und eine für die Wissenschaft
MSM142 – Who are we and why are we here in spring
For many people, it was difficult to understand why one would travel so far north to the Labrador Sea in spring to conduct research. Life on board is exhausting and made more challenging by harsh and rapidly changing weather conditions, which especially complicate the deployment of our research equipment.
So what is our cruise about?
This research cruise has two main goals: studying small-scale ocean features and monitoring large-scale ocean currents.
In recent years, scientists have realised that small-scale features in the ocean such as eddies and fronts play a very important role. They can create strong changes in temperature, salinity, and also in other properties like chlorophyll and carbon export. During this cruise, we use a range of instruments to observe these features in detail so we can better understand how they work.
The second goal is to study how ocean currents in the Atlantic are changing over time. A key part of this is long-term observations at the 53°N observatory. There, we maintain seven moorings long cables anchored to the seafloor and equipped with instruments that measure temperature, salinity, oxygen, and current velocity. Every two years, we recover these moorings to collect the data and then redeploy them to continue the measurements.
Such a cruise needs a long time of preparation and organisation during the cruise, so I asked our Chef Scientist a few questions:
When did you start preparing for the cruise? And how was that going?
The proposal for this cruise was submitted in 2023, which includes motivation and the research questions, with the plan to carry it out in 2025. In the end, it was scheduled for 2026. The detailed preparation really started about a year in advance, around April 2025.
Planning a research cruise involves several steps. We have to organise the logistics, decide who will be part of the science team, and plan the scientific work we want to carry out. And, of course… quite a bit of paperwork
When did you come up with the idea to have the cruise in spring, and not as usually in summer?
The idea goes back to a proposal I was writing in 2022. It included a cruise to study small-scale ocean features and how they are connected to the spring bloom in the Labrador Sea.
Since we were specifically interested in the spring bloom, it was important to be here at that time of year. That meant planning the cruise in spring, even though we knew that the conditions can be more challenging than in summer. But if I would have to do it again, I would look in the weather forecast in advance for a much calmer March and April.
What do you find the most interesting about it?
Being here, in the middle of the ocean, and looking in real-time at the data we are collecting. There is something quite special about the measurements coming in and knowing you are observing the ocean as it happens.
How are the decisions made between you and the captain, in terms of weather, research and safety? When do you meet? And how often?
From the beginning, we agreed to plan about 36 hours ahead, given the scope of the work and the often-challenging weather conditions. We then adjust the plan if needed.
We meet every morning, without a fixed time, to look at the weather forecast and decide together what can be done or not.
So far, this approach has worked very well. We have occasionally had to stop operations at short notice, but we have always managed to adapt without losing much valuable science time.
Is the cruise as you have planned it so far? If not, how does it differ?
The data we’ve collected so far has already exceeded my expectations, especially given the challenging weather conditions. We’ve been able to achieve a lot, and this is mainly thanks to the excellent collaboration between the ship’s crew and the scientific team on board.
Everyone has been very flexible and supportive, which has allowed us to adapt quickly and make the most of the time available.
Number 1 Tipp for working at 10bft and 6 meters waves?
Always keep one hand for the ship, and one for the science
By Qi-Fan Wu (Niels Bohr Institutet, University of Copenhagen)
In 1943, when Warren McCulloch and Walter Pitts showed that neurons could be represented by simple electrical circuits, they laid the first foundation for machines that could learn, adapt, and predict. In 2023, when ChatGPT became widely used, my Introduction to Python professor found that it could answer every question correctly on his course exam. In the history of machine learning, there has been a repeated oscillation between “extremely high expectations” and “deep skepticism.” What is machine learning? What should we expect from machine learning, and when should we be skeptical about it? Should the same principle also be applied to other numerical models?
The goal of machine learning is to make computers “learn” from “data”. From an end user’s perspective, it is about understanding your data, making predictions and decisions. Intellectually, it is a collection of models, methods and algorithms that have evolved over more than a half-century now [e]. Just as the human brain, neural networks, as one of the most popular machine learning methods, are theoretically capable of learning complex relationships from data. Theoretically, Neural Networks can compute any function in the world. No matter what the function is, there is guaranteed to be a neural network so that for every possible input x, the output value f(x) (or some close approximations) is output from the network (Figure 1). This result holds even if the function has many inputs and many outputs [a]. However, universal approximation only describes what neural networks are capable of, while the actual goal of machine learning is to fit an unknown function from a finite set of samples, ideally faster than traditional numerical methods.
In this blog, however, I do not want to focus on large language models that help with writing, coding, and basic background research. Instead, I want to discuss the training and use of special-purpose AI models, such as neural networks, for solving problems in physics, which is also the main topic of my PhD project. Nowadays, an increasing number of scientists are working on AI-related topics, including climate physics. If we think of the physical world as a forward dynamics model, then given the current state and the action to be taken, machine learning aims to predict the next state, while the entire world can be viewed as a huge digital database.
However, after the initial “extremely high expectations,” machine learning has also raised “deep skepticism”. In physics, especially climate physics, the “close approximations” mentioned earlier, together with the lack of standardized workflows, are often the source of trouble. The figure below shows rather discouraging results from reproducing ML-for-PDE-solving studies using stronger baselines.
As the quote attributed to von Neumann goes, “With four parameters I can fit an elephant, and with five I can make him wiggle his trunk.” All models are wrong, including physics-based numerical methods and climate models, but many are useful because different well-performing models can still reveal different aspects of the same physical system [c]. The physicist Paul Dirac reached a similar conclusion long ago: due to the limitations of human cognitive ability, scientific theories cannot be both closed and complete at the same time [d]. This means that we cannot have perfect, exact theories. He saw that theories based on approximations could sometimes have a considerable amount of beauty in them, and he began to infer that perhaps all theories of nature are, ultimately, only approximations [d]. Personally, I think the same rule could apply to machine learning models, and indeed to all models.
My journey on the METEOR made me appreciate the importance of data even more from a modeler’s perspective, and it deepened my belief that the people who create datasets deserve more applause and respect from the entire scientific community. Because of model uncertainty, data, especially observational data, become extremely important for understanding reality. Nature itself is the ultimate database, and its ocean of data is too vast to be compressed into a single dataset.
Machine learning models and climate reanalysis systems require these high-quality data to be reliable in real-world applications. Traditional numerical weather prediction and climate models, including general circulation models, have comprehensive physical foundations but require enormous computational resources, have limited spatial resolution, and struggle to integrate multi-source observations such as station, satellite, and radar data [e]. Although AI-based weather models have developed rapidly in recent years, they still suffer from inconsistent training datasets, time periods, and regions, varying evaluation metrics, and a lack of standardized code and experimental workflows – issues similar to those previously mentioned for AI-based approaches to solving PDEs in fluid mechanics [e]. Under these circumstances, data collected by METEOR, along with all observational data, are necessary for accurately modeling weather and climate, as well as for developing model architectures for them (Figure 3). A good model should embody a trinity of observational results, physical insight, and mathematical formalism. These three aspects should correspond perfectly, with no redundancy.
But what exactly can we do to make progress under data-limited conditions? And which scientifically important problems can be clearly formulated and addressed within an analytical modeling framework? These questions remain like dark clouds hanging over scientists working in related fields. A model that is mathematically beautiful and physically simple may still be inconsistent with observations. Some models considered correct may be mathematically unattractive, and their physical mechanisms may not be clearly explained either. My personal opinion is that, when dealing with this kind of situation in the age of AI, we may still need to rely on our own intuition (and even guessing), trying to understand reality with the help of many scientists who use observations and models to sail by night and expand the boundaries of human knowledge through tiny steps.
Note: Artificial Intelligence’s (AI) stated goal is to mimic human behavior in an intelligent manner, and to do what humans can do, which includes artificial “creativity” like driving cars, playing games, responding to consumer questions, etc. In that sense, AI seeks to create muscle and mind of humans, and mind requires learning from data, i.e. Machine Learning. However, Machine Learning helps learn from data beyond mimicking humans. Having said that, the boundaries between AI and ML are getting blurry day-by-day.
References:
[a] Charniak, E. An Introduction to Deep Learning. Cambridge, MA: MIT Press, 2019; 192.
[b] Nick McGreivy. I got fooled by AI-for-science hype—here’s what it taught me. 2025. https://www.understandingai.org/p/i-got-fooled-by-ai-for-science-hypeheres
[c] Fisher, A.; Rudin, C.; Dominici, F. All Models Are Wrong, but Many Are Useful: Learning a Variable’s Importance by Studying an Entire Class of Prediction Models Simultaneously. Journal of Machine Learning Research 2019, 20(177), 1–81.
[d] Dirac, P. A. M. The Principles of Quantum Mechanics. Oxford University Press: Oxford, 1930.
[e] Bansal, H.; Grover, A.; Jewik, J.; Nguyen, T.; Sharma, P. ClimateLearn: Benchmarking Machine Learning for Weather and Climate Modeling. In Advances in Neural Information Processing Systems 36; 2023; pp 75009–75025. https://doi.org/10.52202/075280-3279.
[f] Global Observing System (GOS). World Meteorological Organization. https://community.wmo.int/site/knowledge-hub/programmes-and-initiatives/global-observing-system-gos
By Nathalie Rodríguez Lara (GEOMAR), Federico Scarscelli (GEOMAR), Ajit Subramanian (LDEO), Qi-fan Wu (University of Copenhagen), Eduardo Lima (UFPE), Herbert Barbosa (UFPE), Joelle Habib (LOV) and Zengchao Xu (GEOMAR)
So, you have been invited to participate in an oceanographic research vessel? Congrats! Oh, it’s German… well. Here are some tips that will be especially useful in the following days.
We shall start with food. Everyone needs it after all.
As with all ships, the Meteor has some hard rules on mealtimes, so be aware of the hour and get ready to enjoy some German delicacies. Keeping the mealtimes is important because the staff must clean up and prepare for the next meal. However, if you are working, you can ask for your food to be put aside for you to eat later.
First of all, breakfast. For the first bite after getting up (or before going to bed, in case you had the night shift), you can always expect a large variety of eggs, cheese, ham, and bread. Moreover, in the Meteor breakfast menu, you even get a little surprise dish each day, featuring a specialty Hausmannskost, or traditional homemade recipe, usually chosen among north Germany’s delicacies. These recipes are made for people that work hard all day, so of course the hearty meals are there to sustain hours on hours of manual labour, with strong flavour as well. Some of our favourites are the Schlemmerschnitte, Bremen Knipp, Wurstschnitzel and, of course, one of Germany’s greatest, the Zwiebelmettbrötchen. Yes, they are quite heavy compared to a nice yogurt, but they will provide you with enough energy for any scientific endeavour or challenge you will face.
As you walk towards the Messe be ready to say Moin, hello, Morgen, morning or any variety of greeting for the day to anyone who passes you by. Even when this is not customary on German land, it is most imperative that you do it on a German ship. Otherwise, be prepared to be greeted by a very loud “Morgen” at 7 am and a disapproving look.
“Mahlzeit” you say as you enter lunch time, a greeting for mealtimes meaning quite literally “mealtime”. A selection of food is served buffet style, so you may choose your desired quantity and meal type. Vegetarian options are included at every meal for those who prefer it, and you can never forget bit of potato on the side. On Sunday, the cooks usually prepare some traditional “family lunch” menu, as a way to enjoy a small moment of home feeling, even when it is far away on land. Even in the hottest day in the Tropical Atlantic, you will immediately feel like celebrating Christmas at your German friend’s home.
For dinner, a similar situation can be expected: a selection of food awaits you at the Messe including everything you need for a good Abendbrot, or evening bread, to those initiating on the German lifestyle.
A lot of attention shall be put on shoes and clothes for all mealtimes. The Messe is to be enjoyed by everyone, so no dirty safety shoes or dirty work clothes shall cross the door (leave the shoes outside, if you must). Sit down with your fellow scientist, eat with people, talk to them, ponder about all the life choices that led you to this point in time and the contributions your days at sea will do to science. Let yourself enjoy the company of other humans and DO NOT use your phone!
Oh, look at the clock, its 15:00! Well, dear reader, I shall introduce you to the great German culture of “Kaffee und Kuchen” or coffee and cake. Every afternoon, the kitchen lays out a variety of sweet little treats for you to enjoy, only a piece though, as everyone should enjoy this lovely tradition. Those in not so friendly terms with caffeine, may also find various selections of teas or even a glass of milk. Perhaps some chocolate or snacks (at your own cost), if you prefer. What’s more, as you see your fellow men work tirelessly under the sun, bring some cake to them, offer a little break from science, and enjoy the long-lived tradition of pastries.
Regarding the work on board, although there is a predefined shift schedule, some activities require most people to be available to help, such as mooring recovery and deployment. Offer a hand with those who need it, using the ancient spells “Do you need some help?” and “Do you want to switch?”. It is very much appreciated by those on deck, who have been bearing the load of the sun and hours of manual labour, and who probably haven’t had a nice meal yet.
For the ones not so familiar with the German language, we offer two very useful work-related words: “stop” and “weiter”, meaning stop and continue. If you want to become a real professional and show off some advanced German skills, you can also add “schneller” and “langsamer”: faster and slower. Keep them in your heart and in your mind, and when the moment comes, you will find them more useful than not.
And, of course, after a tiring day of work, there are smiles, conversations, and music, along with the good feeling that comes from a job well done. Walk into the common areas for the evening: talk with people, maybe play a good round (or two) of Kniffel; perhaps join the kicker or pingpong tournament, a movie night sounds good, or just looking at the stars and the vastness of the night sky would suffice for today.
Another thing to keep in mind is house-keeping rules.
Follow your steward rules. Your cabin is your personal space, keeping it tidy and clean helps you and the housekeeping crew. Remember to leave your door open if you are not sleeping and take care of your shower curtain so it dries properly, we do not want mold on a ship!
Since we are already here, some general advice for life at sea.
An important aspect of life at sea is, sadly, sea sickness. It’s as common as it’s normal, everyone has tips and tricks. Here is ours: bring your pills, also known as Reisetabletten, take them every day and see how it works for you, maybe you are lucky and don’t need them after a few days. But never be too careless, the sea is as vast as it is treacherous, and big waves can change your internal balance easily. So, keep your pills close and keep them safe.
Be aware of the time changes, crossing the Atlantic may feel like a timeless void but time zones still exist to the rest of the world. The captain will announce the changes in time zone, moving an hour forward or backwards as needed, so be aware of this, keep an eye on the clocks and be prepared for a bit of a longer, or shorter, day ahead.
We thought of giving advice on science, but that’s what you are here for. Do not fret. Do your work and do it well. Trust that the knowledge that has brought you to this vessel can contribute, and be contributed, by those around you. Both the scientists and the crew are welcoming (but you have to follow the rules both for safety reasons and to work well in a limited shared space) and make a good effort to communicate in English, even if it is not so comfortable for all of them to so. So, dear reader, may your equipment be efficient, your samples uncontaminated and your results significant. Have a good trip, enjoy the camaraderie of new friends, and remember to enjoy these crazy days ahead.
Gute Reise und viel Spaß!
By Tina Hans (GEOMAR)
One main objective of the cruise is to investigate the large-scale ocean currents in the tropical Atlantic. For that purpose, we are maintaining several long-term observatories at the seafloor and in the water column. Additional to the moorings which have been described in the previous blog “Keeping the record alive”, we deployed and recovered close to the Brazilian coast so-called PIES. They have – as some might say unfortunately – nothing to do with pastries but are oceanographic instruments that measure the pressure at the seafloor as well as the time an acoustic signal takes to travel from the instrument to surface, where the signal is reflected, and back. We deployed six of those instruments across the continental shelf off Brazil at depth ranging from 150 metres to 3000 metres. These deployments are the result of a collaboration with the University of Bremen. We also successfully recovered one PIES that spent just over three years at the seafloor at a depth of 500 metres. With the data of the recovered PIES, we could extend our time series of seafloor pressure measurements at 500 metres depth. This time series, which goes back until 2013, spans now 13 years.
This still leaves the question of what the pressure at the seafloor can tell us about ocean currents. To answer this, one needs to know that the ocean dynamics are largely governed by a balance of two physical forces: the pressure gradient force and the Coriolis force. Essentially, when water ‘piles up’ somewhere, a current is created which attempts to even out the differences, and the direction of this current is deflected due to the Earth’s rotation. This force balance can also be used to directly relate the difference in pressure between two locations to the mean velocity in between these locations. We make use of this relation by measuring the seafloor pressure not just off Brazil but also off Angola at a similar latitude. With the combination of these measurements, we can calculate the mean north-/southward velocities across the Atlantic between Brazil and Angola. From this velocity we can then derive the strength of the Atlantic Meridional Overturning Circulation (AMOC).
However, there is one caveat: the pressure sensors are drifting over time. This makes it impossible to make statements about long-term trends, but we can still make statements about the seasonal to interannual variability of the AMOC. Therefore, the measurements of the PIES can be used to better understand the large-scale currents in the tropical Atlantic. In a next step, we are now using these measurements to better understand the linkage of the AMOC to climate variability in the tropics.
-
Greenhouse Gases11 months ago
Guest post: Why China is still building new coal – and when it might stop
-
Climate Change11 months ago
Guest post: Why China is still building new coal – and when it might stop
-
Greenhouse Gases2 years ago嘉宾来稿：满足中国增长的用电需求 光伏加储能“比新建煤电更实惠”
-
Climate Change2 years ago嘉宾来稿：满足中国增长的用电需求 光伏加储能“比新建煤电更实惠”
-
Renewable Energy8 months agoSending Progressive Philanthropist George Soros to Prison?
-
Climate Change2 years ago
Bill Discounting Climate Change in Florida’s Energy Policy Awaits DeSantis’ Approval
-
Carbon Footprint2 years agoUS SEC’s Climate Disclosure Rules Spur Renewed Interest in Carbon Credits
-
Greenhouse Gases12 months ago
嘉宾来稿：探究火山喷发如何影响气候预测