The CoASTS program was conceived by the Marine Optical Laboratory of the European Commission’s Joint Research Centre (JRC) in collaboration with several European research institutions. Beginning in 1995 with monthly sampling, later relaxed to bi-monthly or less, CoASTS built a time series of bio-optical measurements at the “Acqua Alta” Oceanographic Tower (AAOT) in the northern Adriatic Sea, a site whose waters are moderately optically complex, shaped by moderate concentrations of sediments and colored dissolved organic matter and occasionally by phytoplankton and its degradation products. The program ran continuously until 2016, producing 176 field campaigns in total, of which the 125 performed from December 1998 onward, covering 634 stations, were selected for the current dataset because they fulfilled the standardized methods and data reduction processes required for consistency.

BiOMaP, the companion program, took the same measurement philosophy on the road. From 2000 to 2022, ship-based campaigns carried equivalent bio-optical measurements across the major European seas: the Baltic Sea with its CDOM-dominated waters; the Adriatic, Black, Greenland, Ligurian and North Seas with diverse concentrations of colored dissolved organic matter and suspended particulate matter; and the Eastern and Western Mediterranean Seas with their oligotrophic and mesotrophic Case-1 waters. All 36 BiOMaP campaigns, totaling 1915 stations, were included in the dataset. While CoASTS sampled water from near-surface, 8 and 14 meter depths along with optical, hydrographic and meteorological profiles, BiOMaP stations limited water sampling to the near-surface layer, yet applied the same instruments, community measurement methods, extended quality control schemes and consolidated processing codes throughout.

The heart of the new release is a set of absorption coefficients measured on discrete water samples collected at approximately one meter depth. For pigmented and non-pigmented particles, the coefficients a_ph and a_dt cover the 400 to 750 nanometer interval, while the coefficient a_ys for colored dissolved organic matter spans 350 to 680 nanometers. All are delivered with 3 nanometer spectral resolution at 2 nanometer increments, a dramatic step beyond the multi-spectral data products centered on nominal wavelengths that the same authors released in a 2024 companion paper. That spectral richness matters because the next generation of hyperspectral satellite sensors, such as those anticipated in the NASA Plankton, Aerosol, Cloud, ocean Ecosystem mission lineage, will observe the ocean in hundreds of narrow bands, and algorithms exploiting those bands need reference spectra at matching resolution.

The measurement methodology is a story of meticulous standardization. Particulate absorption was determined with the Transmittance-Reflectance (T-R) method of Tassan and Ferrari, which combines light transmittance and reflectance measurements using an integrating sphere coupled to a dual-beam spectrophotometer. The approach captures both forward-scattered and backward-scattered light from particles retained on Whatman GF/F glass fiber filters with 0.7 micrometer pore size, and a theoretical model minimizes the impact of differences in backscattering. The method is considered appropriate for any particle type, including highly scattering minerals and highly absorbing sediments, and was initially run on a Perkin Elmer Lambda-19 spectrometer, replaced by a Lambda-950 from 2004 onward. To separate the non-pigmented fraction from the total, the team applied a sodium hypochlorite bleaching solution that rapidly attacks pigment molecules while leaving detritus largely untouched, with the pigmented fraction then obtained by difference.

Since 2003, the depigmentation step was itself upgraded in a way that illustrates the programs’ obsession with reproducibility. The improved methodology combines chemical bleaching with photo-bleaching and applies real-time control of the process by monitoring sample reflectance in the red spectral region. The original approach required successive additions of oxidizing agent and rinsing, which incrementally perturbed the distribution of particles on the filter and degraded accuracy. The controlled version accelerates bleaching, prevents repeated measurement trials, and standardized the entire process, yielding better accuracy and reproducibility of the derived coefficients. Quality control was systematic: every absorption spectrum was visually inspected for residual pigment absorption, evidenced by shoulders in the blue-green and slight peaks in the red, and when poor bleaching was evident, the duplicate sample collected at every station was analysed instead.

CDOM absorption followed its own rigorous protocol. Water was filtered through Millipore 0.22 micrometer cellulose filters, treated with sodium azide to prevent bacterial growth, and measured in 10 centimeter quartz cuvettes against a milli-Q water reference on Perkin Elmer Lambda-12 and later Lambda-35 dual-beam spectrometers. The authors acknowledge that the 10 centimeter path length can reduce accuracy for samples with very low CDOM absorption, but standardization across geographically diverse programs took priority. Assuming CDOM absorption is null in the red, the final coefficients were corrected by subtracting the mean of spectral values in the 670 to 680 nanometer window. The resulting spectra reveal striking regional contrasts: mean CDOM absorption at 350 nanometers ranges from a minimum of 0.13 per meter in the oligotrophic Eastern Mediterranean to a maximum of 2.2 per meter in the highly absorbing Baltic Sea, while mean particulate absorption at 400 nanometers climbs from roughly 0.01 per meter in the Mediterranean to 0.11 per meter in the Baltic.

Beyond the raw spectra, the authors fitted exponential decay functions to the non-pigmented particle and CDOM coefficients, extracting slopes that are widely used in bio-optical modelling. The detrital slopes vary from 0.009 per nanometer in the Eastern Mediterranean up to 0.013 per nanometer in the North Sea, and remain essentially unchanged when the fit is restricted to spectral windows that exclude residual pigment absorption. CDOM slopes range from 0.015 per nanometer in the Eastern Mediterranean to 0.021 per nanometer in the Baltic, rising slightly when the fit is confined to 350 to 550 nanometers. Comparisons with published studies from the Mediterranean, Baltic, Black Sea and Arctic show general agreement, confirming that the dataset sits comfortably within the known range of natural variability while offering something most compilations cannot: identical instrumentation, sampling and analysis protocols applied across different times and regions.

One of the most intriguing findings emerges from ternary plots of the three absorption components at 442 nanometers. While the majority of samples show predominance of absorption by CDOM, as expected for the Black Sea and Baltic, the pronounced clustering toward CDOM dominance observed even in the oligotrophic Eastern Mediterranean and some Western Mediterranean waters conflicts with the classic Case-1 water classification, in which phytoplankton governs optical properties. The finding echoes earlier independent work in the northwest Mediterranean, further confirming a significant CDOM background in Mediterranean waters. Meanwhile, ratios of phytoplankton absorption at 442 versus 676 nanometers, with correlations between 0.96 and 0.99, decrease from oligotrophic to eutrophic waters, a pattern the authors attribute to larger mean phytoplankton cell sizes in nutrient-rich environments, which increase the pigment packaging effect and reduce the proportion of photoprotective pigments.

The full CoASTS-BiOMaP hyperspectral dataset is available in tabular form through the PANGAEA repository, and the authors expect it to fulfil application needs for bio-optical investigations supporting the new generation of hyperspectral satellite ocean color sensors. In an era when satellite ocean color underpins everything from fisheries management to carbon cycle science, the value of a quarter-century of consistent, quality-controlled, spectrally resolved field measurements is difficult to overstate. What began as monthly trips to an offshore tower in the Adriatic has matured into one of the most comprehensive hyperspectral absorption archives ever assembled for European seas, a quiet monument to the power of long-term, standardized observation.

Subject of Research: Hyperspectral absorption coefficients of optically significant seawater constituents from the CoASTS and BiOMaP field programs

Article Title: Coastal Atmosphere & Sea Time Series (CoASTS) and Bio-Optical mapping of Marine optical Properties (BiOMaP): the hyperspectral absorption coefficients by optically significant constituents

Article References: Berthon, J.-F., & Zibordi, G. (2026). Coastal Atmosphere & Sea Time Series (CoASTS) and Bio-Optical mapping of Marine optical Properties (BiOMaP): the hyperspectral absorption coefficients by optically significant constituents. Earth System Science Data, 18(10), 7499-7514. https://doi.org/10.5194/essd-18-7499-2026

Image Credits: AI Generated

DOI: 10.5194/essd-18-7499-2026

Keywords: ocean color, hyperspectral absorption, CoASTS, BiOMaP, CDOM, phytoplankton, Adriatic Sea, Baltic Sea, Mediterranean Sea, satellite validation, bio-optics, Joint Research Centre

The study, conducted by Antonio Puertas Gallardo and Magdalena Stepien of the European Commission’s Joint Research Centre in Ispra, Italy, together with Antonino Brunetto of ARHS Developments Italia in Milan, examined 102 projects funded under Horizon Europe, the EU4Health programme, and the Digital Europe Programme. Together, these initiatives represent a combined investment of more than 840 million euros dedicated to understanding, detecting, treating and preventing cancer. The work is significant not merely for its headline figures but for its methodology, which demonstrates how modern artificial intelligence tools can bring order to the sprawling, fragmented record of publicly funded research.

The researchers employed a combination of quantitative and qualitative methods to characterise the funding landscape and identify patterns within it. The centrepiece of their approach was the application of large language model methodology to assign standardised keywords to each project in the portfolio. This was no trivial task. EU-funded cancer projects span an enormous range of activities, from laboratory-based investigations of tumour biology to public health campaigns and the construction of digital infrastructure for sharing clinical data. Without a common vocabulary, comparing projects across different programmes and directorates-general becomes nearly impossible, and opportunities for synergy go unnoticed.

By using large language models to read and classify project descriptions, the team developed a set of 56 standardised keywords capable of clustering the 102 projects into coherent thematic groups. This keyword framework allowed the researchers to identify synergies and thematic interconnections among cancer-related initiatives that would otherwise remain buried in separate databases and reporting formats. The most frequent terms to emerge from this analysis related to healthcare, policy and guidance, and digital aspects, reflecting the fact that EU cancer funding is not confined to the bench but extends deep into health systems, regulation and technology.

The thematic findings offer a revealing snapshot of European priorities. Most of the funded projects concentrated on diagnosis, treatment and prevention, the three pillars of clinical oncology. When the analysis turned to specific cancer types, colorectal and breast cancers emerged as the most common areas of focus, a pattern consistent with their status as two of the most frequently diagnosed malignancies in Europe and with the existence of dedicated screening programmes across many member states. The prominence of digital keywords underscores how thoroughly data science, artificial intelligence and e-health technologies have penetrated the cancer research agenda, from computational pathology platforms to federated networks for sharing sensitive patient information.

Equally striking were the geographic patterns the analysis uncovered. Institutions from larger western and southern European countries proved to be the most active participants in EU cancer funding, dominating both the coordination of projects and participation more broadly. Distinct collaboration networks between specific institutions were identified, revealing clusters of organisations that repeatedly join forces across successive funding calls. These networks represent a form of institutional capital that is invisible in simple funding statistics but crucial to understanding how European research capacity is actually organised and reproduced over time.

Perhaps the most politically consequential finding concerns the relationship between national characteristics and funding success. The study found that countries with larger populations were involved in a greater number of projects, but that countries with a higher gross domestic product per capita were not. In other words, wealth per citizen does not predict engagement with EU cancer funding, but sheer demographic weight does. This distinction matters for ongoing debates about the equity of EU research funding. Smaller member states have long voiced concerns that framework programmes disproportionately reward large, well-established research systems, and this analysis provides granular evidence with which such concerns can now be assessed and, potentially, addressed through targeted policy interventions.

The technical achievement underlying these findings deserves attention in its own right. Assigning standardised keywords to research projects has traditionally been a labour-intensive exercise requiring expert panels to read and code hundreds of project abstracts, a process that is slow, expensive and difficult to reproduce. Large language models, by contrast, can process project descriptions at scale and apply a consistent classification scheme, making it feasible to update the mapping continuously as new projects are funded. The authors’ use of this technology within an official Commission analysis signals a broader shift in how public administrations may approach evidence-based policymaking, treating the text of research proposals and reports not as administrative residue but as analysable data.

The practical implications extend beyond academic interest. The European Commission’s cancer projects tool, which provided the data for this study, is designed to give policymakers, researchers and the public a transparent view of EU cancer funding. By demonstrating how the tool’s contents can be systematically analysed, the authors have laid what they describe as the foundation for more strategic planning of EU cancer funding. Identifying thematic clusters makes it possible to spot both overcrowded areas, where additional investment may yield diminishing returns, and neglected ones, where gaps in the portfolio may warrant attention in future work programmes. Mapping collaboration networks similarly allows funders to see which institutions are consistently included and which remain on the margins of European cancer research.

The study arrives at a moment when Europe’s ambition to reduce the cancer burden is codified in policy initiatives spanning research, health systems and digital infrastructure. What this analysis makes clear is that ambition is backed by substantial and traceable resources: more than 840 million euros distributed across 102 projects spanning three complementary programmes. What it also makes clear is that the distribution of those resources follows identifiable structural patterns, shaped by population size, institutional networks and thematic fashion. By making those patterns visible, the Joint Research Centre team has transformed a collection of individual project records into a strategic instrument, one that could help ensure that the next generation of European cancer funding is planned not by habit or accident, but by design.

Subject of Research: Analysis of European Union cancer research funding across Horizon Europe, EU4Health and the Digital Europe Programme

Article Title: Unveiling European cancer funding: insights from the European Commission’s cancer projects tool

Article References: Puertas Gallardo, A., Brunetto, A., & Stepien, M. (2026). Unveiling European cancer funding: insights from the European Commission’s cancer projects tool. Health Research Policy and Systems. https://doi.org/10.1186/s12961-026-01535-9

Image Credits: AI Generated

DOI: 10.1186/s12961-026-01535-9

Keywords: cancer funding, Horizon Europe, EU4Health, Digital Europe Programme, large language models, European Commission, Joint Research Centre, cancer policy, research collaboration, colorectal cancer, breast cancer, health research policy