Skip to content

Ocean Carbon Cycle

Science Demonstration Case “Ocean”#

Authors: Julia Klima, Jannes Kruse, Jonas Neumann

A novel approach was explored to understanding the ocean carbon cycle through the utilization of a recently developed, uniformly structured data cube encompassing key ocean carbon cycle variables. The methodology involves a combination of dimensionality reduction and feature selection techniques, enabling to unravel the intricacies of the carbon cycle across various geographic locations. The study spans a period from 1998 to 2020, during which the complexity of the ocean's carbon cycle was analysed. We find that the originally eight-dimensional feature space can be effectively condensed into three to four dimensions. This reduction not only simplifies the representation but also enhances our understanding of the underlying processes. A significant aspect of our analysis is the identification of geographical patterns in the carbon cycle's complexity. These patterns are closely linked to the dynamics of thermohaline circulation and the movement of water masses. Specifically, areas of intense upwelling and the warm, surface-near currents of the Global Conveyor Belt are highlighted, along with their anomalies and regional extreme events.

This research pinpoints several key variables, such as Particulate Inorganic Carbon and Mixed Layer Depth, as critical in both a global and regional context for representing the simplified dimensions of the carbon cycle. These findings are pivotal in deepening our understanding of the carbon cycle's regional behaviours. We will now iterate the findings with the data providers in order to understand how to enhance the predictability and effectiveness of future research in carbon modelling and oceanic pathways.

Study design using a data cube of oceanic carbon indicators.
Steps performed to obtain the dimensionally reduced and clustered new data cubes
Results of the Principal Component Analysis per pixel.

The study presented here is still in preliminatry state. However, we find that using Principal Component Analysis (PCA) and Principal Feature Analysis (PFA) is very helpful to analyze ocean carbon cycle variables from a data cube. The key findings are summarized as follows:

  • Dimensionality Reduction: The PCA results indicated that most of the ocean's carbon cycle can be described by three to four dimensions, significantly reducing the original eight-dimensional feature space. This suggests that many ocean carbon pump variables are interrelated and exhibit coordinated variations over time.
  • Geographical Patterns: The study revealed distinct geographical patterns in the number of dimensions per pixel. These patterns are associated with ocean currents, upwelling regions, wind zones, and differences between coastal and open-sea areas. In open-sea regions, the data cube variables often require four dimensions for description, while near coastlines, three dimensions are generally sufficient.
  • Regional Variations: The analysis identified specific regions with unique dimensional patterns. For example, large areas above 30° N and 30° S in the westerly wind zone predominantly show three dimensions. Notable regions like the equatorial-subtropical zone, the Pacific-Indian Ocean transition, and areas around the Global Conveyor Belt also exhibit distinct patterns.
  • El Niño-Southern Oscillation Area: This area is mostly represented by four dimensions, with patterns influenced by trade winds, upwelling regions, and warm ocean currents. The equatorial-subtropical zone is the only area where five dimensions are occasionally present.
  • Principal Component Variability: The first principal component (PC1) explains a significant portion of the variability in many regions, especially around the equator. However, in areas like the East Pacific Rise, the variability is spread across multiple dimensions.
  • Principal Feature Analysis (PFA) Results: PFA identified key variables that describe the reduced dimensionality of the ocean carbon cycle. Particulate Inorganic Carbon (PIC) is a major variable, present in almost all pixels, indicating its importance in describing the variability of the ocean carbon pump.
  • Phytoplankton Variability: Different types of phytoplankton (micro, pico, and nano) show varying presence across different regions, indicating their role in the ocean carbon cycle's variability.
  • Mixed Layer Depth (MLD) and Other Variables: MLD is generally important for describing variability, but its presence varies geographically. Other variables like Primary Production (PP) and Particulate Organic Carbon (POC) also show varying importance across different locations.
  • Cluster Analysis: The clustered PFA results reveal patterns in variable importance across different ocean regions. These clusters help in understanding the regional differences in the ocean carbon cycle
  • Implications and Limitations: The study provides insights into the interconnected nature of oceanic processes affecting the carbon cycle. However, it also acknowledges limitations such as the exclusion of certain variables, the lack of a comprehensive theoretical framework, and the absence of data in certain latitudes.

Overall, the study offers a nuanced understanding of the ocean carbon cycle's complexity and its geographical variability, highlighting the interconnectedness of various oceanic processes and their impact on carbon cycling.