1. Introduction

The Arctic is warming at a rate of 1.0°C per decade due to ice albedo feedback, exacerbating the pressure on the “Driving layer” of cascade risks (Soyez, 2020: 3) . Norilsk, the northernmost industrial city in the world, is built on permafrost and is highly susceptible to the effects of permafrost degradation caused by climate change.

On May 29, 2020, a diesel storage tank at Norilsk Nickel’s thermal power plant No. 3 collapsed, leaking 21,000 tons of diesel. Approximately 15,000 tons of diesel flowed into the Anbarnaya River, and 6,000 tons of diesel seeped into the soil, contaminating an area of 180,000 m² (Norilsk Nickel, 2021: 23). This accident was not only a technical malfunction, but a series of chain reactions triggered by climate change, technical and construction defects, and management risk. Permafrost degradation in the Arctic is one of the main reasons for this accident.

Existing research has examined the reasons for the accident, including permafrost degradation, building defects, and management failures (Ryzvanov et al., 2022: 344), as well as its consequences: contaminant diffusion (Bezmatemykh et al., 2021: 414), soil deposition (Androkhanov et al., 2021: 69), and groundwater contamination (Yakuteni and Solovyev, 2020: 52).

Schneider von Deimling et al. (2021:2451) revealed the “two-stage behavior” of permafrost degradation, explaining how the pressure from the “Driving layer” nonlinearly transformed into the risk of the “Failure layer”. However, these findings remain fragmented.

This paper, referencing the research of Ma Xiaoxue et al. (2022: 222), introduces the concept of cascading risk and integrates all risk factors into a unified cascading risk framework (Driving layer-Failure layer-Diffusion layer-Response layer) to reveal the complete risk propagation path of infrastructure on permafrost in the Arctic.

2. Correlation between permafrost degradation and infrastructure risk

2.1 Engineering mechanisms of permafrost-infrastructure interaction

Industrial facilities transfer heat downwards, accelerating the warming of permafrost in the Arctic. The industrial region in Norilsk has experienced accelerated warming in recent decades (Soyez, 2020: 3) . The continuous warming is the core driver of the “Driving layer”.

In the case of permafrost degradation, the bearing capacity of pile foundations in industrial infrastructures decreases, and permafrost creep rate and uneven settlement of pile foundations accelerate (Schneider von Deimling et al., 2021: 2455). These effects exhibit significant time lags: risks accumulate slowly until a critical point is reached, and then manifest in the form of structural failure. This is the key engineering mechanism of cascading risks.

2.2 Analytical framework for cascading risks

This paper constructs a four-stage model of infrastructure risk cascades in the permafrost region of the Arctic. This framework places the accident within a coupled «Climate-Technology-Management» system, emphasizing the risk transmission paths across four interrelated levels:

Driving layer: It is the systemic pressure source. Continuous climate warming leads to permafrost degradation.

Failure layer: It is the critical disturbance. Structural failures are caused by permafrost degradation and construction defects.

Diffusion layer: It amplifies the consequences. Pollutants diffuse through soil,  surface water, and groundwater.

Response layer: It belongs to external intervention. The effectiveness of emergency response and governance determines the final extent of damage.

3. Cascading risk analysis of the Norilsk accident

3.1 Accident overview and direct causes of the «Failure Layer»

On May 29, 2020, a diesel storage tank at Norilsk Nickel’s thermal power plant No. 3 collapsed, leaking 21,000 tons of diesel. This is the start point of the “Failure layer”.

According to the Federal Environmental, Industrial and Nuclear Supervision Service of Russia (Rostekhnadzor), the accident was caused by three interacting factors (Ryzvanov et al., 2022: 344):

(1) Technical factors: Pile displacement and uneven settlement are caused by permafrost degradation.

(2) Construction defects: The piles were shorter than designed and not embedded in bedrock, increasing the risk of loss of pile bearing capacity.

(3) Management failures: The tank had been in need of major repair since 2018, but was put into use without safety assessment. These failures create safety hazards for the “Response layer”.

3.2 Transition from the “Driving Layer” to the “Failure Layer”

Prior to the accident, the warming rate in the Arctic reached a rate of 1.0°C per decade (Soyez, 2020: 3), leading to permafrost degradation. As permafrost degraded, permafrost began to creep, and the bearing capacity of the pile foundations decreased, resulting in the settlement of pile foundations.

The process is difficult to detect because of the slow temperature changes and lack of monitoring. The uneven settlement of pile foundations leads to cracks in the bottom and wall of the storage tank, ultimately resulting in diesel spill. The accident was consistent with the predictions of the “Two-stage” model (Schneider von Deimling et al., 2021: 2455), and was the result of long-term interaction between permafrost degradation and construction defects.

3.3 Systemic collapse from the «Failure Layer» to the «Diffusion Layer»

(1) Dam collapse

The capacity of the dam was too small to bear the dynamic impact of diesel spill, leading to the collapse of the storage tank (Norilsk Nickel, 2021: 23). This marks that the accident transferred from the “Failure layer” to the “Diffusion layer”. An accident that could have been controlled began to spread outside.

(2) Diffusion path

6,000 tons of pollutants seeped into the soil, and 15,000 tons of pollutants entered the Anbarnaya River. Analysis confirmed that dissolved diesel migrated towards the Kara Sea (Bezmatemykh et al., 2021: 414). At the same time, pollutants also seeped into the ground: in addition to 880 tons of contaminated soil removed, estimated 5,120 tons of pollutants entered the groundwater system, forming a long-term pollution source (Yakuteni & Solovyev, 2020: 52).

Crucially, the “Diffusion layer” did not stop with the water. Soil surveys indicate that large-scale pollutant deposition was found at 25-30 km downstream of the Anbarnaya River; the total petroleum hydrocarbon (TPH) concentration in this area was as high as 133-373 mg/cm³, while only 10-12 mg/cm³ upstream (Androkhanov et al., 2021: 69). Therefore, most of the diesel spill did not flow into Lake Piasino, but rather accumulated in the abrupt slope area due to the unique downstream topography (Androkhanov et al., 2021: 69), thus altering the ultimate scope of impact.

3.4 «Response Layer» failure and disaster amplification

As the risk spread, the “Response layer” completely failed, exacerbating the disaster.

The risk assessment from the company estimated the probability of a catastrophic accident at 1.5×10⁻⁵, but ignored the dynamic risks (Schneider von Deimling et al., 2021: 2455). At the same time, the emergency response plans were inadequate, as evidenced by the failure of the oil boom, and the concentration of pollutants outside the boom exceeded the safety standard by several times (Yakuteni & Solovyev, 2020: 52). More seriously, the company delayed reporting the spill (Shushakova, 2020: 456), causing the “Diffusion layer” to spread uncontrollably.

4. Discussion on the reasons of cascading risks

4.1 Coupling characteristics of multi-source risks

The risk coupling characteristics of the Norilsk accident (shown in the table below) clearly reveal why individually controllable risks, when combined, lead to disaster.

In fact, each individual risk could have been effectively controlled from the “Driving layer”, “Failure layer”, and “Response layer”.

Proper construction of pile foundations could have prevented rapid failure; continuous permafrost monitoring could have helped to detect the pile settlement early; adequate emergency plans could have effectively controlled pollution. However, technical, geological, and managerial risks accumulate over time, ultimately converging into a systemic disaster at the «Diffusion layer».

4.2 Uniqueness of infrastructure risks in permafrost regions

First, permafrost is dynamic. The mechanical properties of permafrost will change with the temperature, thus requiring continuous monitoring.

Second, climate change is dynamic. Historical data cannot predict future changes; failure times may be underestimated if the models ignore thermal disturbances (Schneider von Deimling et al., 2021: 2455). This is precisely why the assessment of the probability of a catastrophic event (1.5×10⁻⁵) failed.

Third, a positive feedback loop exists. Heating infrastructure accelerates permafrost degradation, creating a vicious cycle. Post-accident monitoring systems aim to break this cycle (Nornickel, 2021: 23).

5. Conclusions and implications

5.1 Main conclusions

First, this accident was the result of the superposition of multiple risks related to climate change, engineering, and management: permafrost degradation created systemic pressure (Driving layer); construction defects laid safety hazards (Failure layer); and management errors exacerbated the consequences of the risks (Diffusion layer). This illustrates the cascading development process of risks from slow accumulation to sudden outbreak.

Second, traditional risk assessment methods will be ineffective under the nonlinear and insidious nature of permafrost degradation. The company considered that catastrophic failure is impossible based on the probability of 1.5 × 10⁻⁵, while ignoring the situation of permafrost degradation. Assessment methods based on static probability fail in non-stationary environments (Schneider von Deimling et al., 2021: 2455).

Third, cascading risk management requires systemic synergy among technologies, management, and institutions. It is insufficient to improve a single stage; instead, we should apply systematic measures, such as engineering remediation, permafrost monitoring, early warning, and standard revision.

5.2 Implications for risk governance

Based on the lessons learned from the Norilsk accident, the following recommendations are made:

(1) Technical aspects

It is necessary to promote thermal stabilization technologies and build a comprehensive monitoring system for permafrost in the Arctic. The monitoring systems built by the company after the accident strengthened monitoring of the “Driving layer” and provided early warning of risks to the “Failure layer” (Nornickel, 2021: 23). This accident highlighted the necessity of complying with environmental regulations and regular equipment monitoring.

(2) Management level

Risk assessment must be elevated from static compliance to dynamic analysis, incorporating climate change scenarios and cascading pathways. The risk assessment process must cover the entire chain from the “Driving layer” to the “Response layer”.

(3) Institutional aspects

Incorporate climate change factors into engineering design standards for permafrost in the Arctic. Regularly update relevant standards to ensure they accurately reflect the current situation. Establish a special fund for environmental protection and remediation, ensuring the availability of funds. Revise and improve risk emergency response plans to ensure they comply with dynamic principles.