Is nuclear waste storage safe over geological timescales?

How do deep geological repositories keep waste isolated for hundreds of thousands of years?

Deep geological repositories (DGRs) use a multi-barrier system — a combination of engineered and natural barriers — to prevent radioactive materials from reaching the surface. The waste is first sealed in corrosion-resistant canisters, then surrounded by bentonite clay that swells when wet to self-seal any cracks, and finally placed in stable rock formations like clay, granite, or salt hundreds of meters underground [2][3]. This layered approach means that even if one barrier degrades over millennia, the others continue to contain the waste.

The near-field environment — the area immediately around the waste canisters — undergoes complex thermal, hydraulic, mechanical, and chemical changes over time, including canister corrosion and hydrogen gas generation. A 2025 review of DGR designs emphasizes that understanding these coupled processes is critical for predicting long-term safety, and that bentonite's self-sealing property is a key feature that maintains barrier integrity [3]. Essentially, the repository is designed to evolve in a predictable, slow way that keeps radionuclides trapped.

What do the safety models actually predict about radiation exposure?

Computer models that simulate radionuclide transport over hundreds of thousands of years consistently show that peak radiation doses to humans remain extremely low. A 2024 study using the GoldSim code modeled radionuclide release from two types of South Korean spent nuclear fuel (PLUS7 and ACE7) and found that the maximum annual dose was less than 1 mSv — well below South Korea's natural background radiation of 2.4 mSv per year [1]. The primary contributors to this dose were carbon-14 and iodine-129, which are long-lived but migrate very slowly through the engineered barriers.

Another modeling tool, TransPyREnd, was developed specifically to simulate radionuclide transport over geological timescales, accounting for processes like diffusion, sorption (binding to rock), and radioactive decay chains [6]. These models show that sorption in the buffer material (bentonite) is especially effective at reducing doses — the 2024 study found that the buffer's ability to sorb carbon-14 and iodine-129 significantly lowered the peak exposure [1]. In other words, the very materials chosen for the repository are designed to chemically bind radioactive elements and prevent them from moving.

What are the biggest uncertainties and challenges in proving long-term safety?

The main challenge is that we cannot directly test a repository for hundreds of thousands of years, so safety cases rely on models and indirect evidence. A 2021 analysis of Finland's Posiva project (the world's first operational DGR) shows that nuclear waste organizations use historical and cultural analogies — like comparing the stability of deep clay to ancient geological formations — to argue that the underground is more predictable and 'safer than' the surface environment [5]. This is a discursive strategy to bridge the gap between scientific models and public confidence.

Geoscientists also face epistemological uncertainty — they cannot know every detail of how a host rock will behave over geological time. A 2025 study of Switzerland's Opalinus clay repository project describes how scientists publicly stage the clay as 'stable' through maps, models, and materials, but acknowledges that this stability is partly a performative act to make the project politically and socially feasible [4]. The key point is that while the science is robust, there is always residual uncertainty about long-term geological changes (e.g., future glaciations or seismic events), which is why site selection is so rigorous and why multiple barriers are used.

Sources used in this answer

Related research questions