Modelling Radionuclide Concentrations in Fucus on the Swedish West Coast
(2026) FYSP10 20261Department of Physics
Particle and nuclear physics
- Abstract
- Past and present measurements of radionuclide activity concentrations in the seaweed species Fucus spp. on the Swedish west coast have shown significant temporal variation. The current understanding is that marine discharges originating from remote nuclear facilities contribute to the observed activity levels. Initial statistical exploration has been done, but no detailed modelling has been performed thus far. In this project, an attempt is made to use a Bayesian dynamic generalized additive model (dGAM) to model observed activity concentrations at Särdal, Sweden as a response to remote discharges from facilities that are suspected to contribute to observed levels. The isotopes 14C, 137Cs, 129I and 60Co are studied due to their the highest... (More)
- Past and present measurements of radionuclide activity concentrations in the seaweed species Fucus spp. on the Swedish west coast have shown significant temporal variation. The current understanding is that marine discharges originating from remote nuclear facilities contribute to the observed activity levels. Initial statistical exploration has been done, but no detailed modelling has been performed thus far. In this project, an attempt is made to use a Bayesian dynamic generalized additive model (dGAM) to model observed activity concentrations at Särdal, Sweden as a response to remote discharges from facilities that are suspected to contribute to observed levels. The isotopes 14C, 137Cs, 129I and 60Co are studied due to their the highest data availability for this sampling location. For 129I, variance is allowed between summer and winter samples in an effort to study seasonal effects on measured activity concentrations.
The models for 14C do not follow literature findings of transport time or structure, hinting that the signal may be too weak, the model too complex, or key predictors (local NPP effluent) are missing from the model. The model for 137Cs agrees reasonably well with literature, showing a dominant contribution from Baltic Sea outflow related to Chernobyl depositions, and initial effect from Sellafield, UK found after 4-5 years. The 129I model also agrees reasonably with current understanding of transport time, with estimates for Sellafield to Särdal at around 4-5 years and La Hague to Särdal at 3-4 years. The 129I model also hints at structurally higher activity concentration in the samples collected in winter, which corroborates recent findings. Finally, across the different isotope models we observe that the models tend to have very broad and dispersed effect well beyond the initial arrival of radioactive material, highlighting a non-trivial temporal pattern which the distributed lag approach is able to capture well.
We conclude that dGAMs are a powerful tool, but despite strong regularization, these models appear to be too flexible in the case where physics restrictions can reasonably be imposed. Future work could include more rigorous model validation, exploring explicit modelling of dilution factors, increasing the information in the model by sourcing higher frequency discharge data or exploring mixed-frequency models, and exploring distributed-lag approaches that allow physics restrictions to be imposed more directly than is possible in the current dGAM formulation. (Less)
Please use this url to cite or link to this publication:
https://lup.lub.lu.se/student-papers/record/9226168
- author
- Nelissen, Pim
- supervisor
- organization
- course
- FYSP10 20261
- year
- 2026
- type
- L3 - Miscellaneous, Projetcs etc.
- subject
- language
- English
- id
- 9226168
- date added to LUP
- 2026-05-13 11:57:00
- date last changed
- 2026-05-13 11:57:00
@misc{9226168,
abstract = {{Past and present measurements of radionuclide activity concentrations in the seaweed species Fucus spp. on the Swedish west coast have shown significant temporal variation. The current understanding is that marine discharges originating from remote nuclear facilities contribute to the observed activity levels. Initial statistical exploration has been done, but no detailed modelling has been performed thus far. In this project, an attempt is made to use a Bayesian dynamic generalized additive model (dGAM) to model observed activity concentrations at Särdal, Sweden as a response to remote discharges from facilities that are suspected to contribute to observed levels. The isotopes 14C, 137Cs, 129I and 60Co are studied due to their the highest data availability for this sampling location. For 129I, variance is allowed between summer and winter samples in an effort to study seasonal effects on measured activity concentrations.
The models for 14C do not follow literature findings of transport time or structure, hinting that the signal may be too weak, the model too complex, or key predictors (local NPP effluent) are missing from the model. The model for 137Cs agrees reasonably well with literature, showing a dominant contribution from Baltic Sea outflow related to Chernobyl depositions, and initial effect from Sellafield, UK found after 4-5 years. The 129I model also agrees reasonably with current understanding of transport time, with estimates for Sellafield to Särdal at around 4-5 years and La Hague to Särdal at 3-4 years. The 129I model also hints at structurally higher activity concentration in the samples collected in winter, which corroborates recent findings. Finally, across the different isotope models we observe that the models tend to have very broad and dispersed effect well beyond the initial arrival of radioactive material, highlighting a non-trivial temporal pattern which the distributed lag approach is able to capture well.
We conclude that dGAMs are a powerful tool, but despite strong regularization, these models appear to be too flexible in the case where physics restrictions can reasonably be imposed. Future work could include more rigorous model validation, exploring explicit modelling of dilution factors, increasing the information in the model by sourcing higher frequency discharge data or exploring mixed-frequency models, and exploring distributed-lag approaches that allow physics restrictions to be imposed more directly than is possible in the current dGAM formulation.}},
author = {{Nelissen, Pim}},
language = {{eng}},
note = {{Student Paper}},
title = {{Modelling Radionuclide Concentrations in Fucus on the Swedish West Coast}},
year = {{2026}},
}