New Insights from Phosphogypsum Tailings Dam Analysis

Tailings dam failures can devastate communities and environments and are often triggered by liquefaction during seismic events. But what if traditional liquefaction assessment criteria misclassify material behavior, leading engineers to apply overly conservative—or worse, inadequate—safety measures?  

My colleagues Bowen Fan and Professor Tao Wang from Wuhan University and I recently published a technical closure (Closure to “Dynamic Response Characteristics and Liquefaction Analysis of a Phosphogypsum Tailings Dam”) in the International Journal of Geomechanics published by the American Society of Civil Engineers that addresses critical questions about liquefaction prediction in phosphogypsum tailings dams. This work matters because accurate characterization of soil behavior under cyclic loading directly influences the safety and economics of tailings storage facilities worldwide.​ 

The Challenge: Distinguishing Flow Liquefaction from Cyclic Mobility

We received thoughtful questions from discussers about whether ITASCA’s P2PSand constitutive model could accurately capture liquefaction behavior under various loading conditions, particularly with initial shear stress bias, and whether our chosen excess pore pressure ratio (Ru) threshold of 0.8 was appropriate for triggering liquefaction in stability analyses. These aren’t academic quibbles—they’re fundamental questions that affect how we assess dam safety and assign post-liquefaction strengths in our models.​​ 

Traditional practice often treats all materials reaching high pore pressure ratios as having “liquefied,” but this oversimplifies the actual physics. Flow liquefaction involves catastrophic loss of shear resistance, while cyclic mobility produces large deformations without complete strength loss. Applying the wrong criterion can either overestimate risk (unnecessarily increasing costs) or underestimate risk (compromising safety).​ 

Our Method: Parametric Single-Element Tests

To investigate model performance, our team conducted extensive single-element simulations using the P2PSand model under undrained cyclic direct simple shear (DSS) loading. We tested multiple scenarios varying relative density (Dr), initial vertical effective stress (σ’v0), static bias ratio (α), and cyclic stress ratio (CSR).​ 

Our simulations revealed critical distinctions. Under very loose conditions (Dr = 25%), the model successfully captured flow liquefaction behavior with pronounced strain softening—exactly matching recent experimental findings by Reid et al. (2024) and Kim et al. (2024) for reconstituted loose sand samples. The model even simulated the quasi-steady state, a local minimum in shear resistance, during the flow liquefaction phase, a phenomenon that sometimes cannot be captured in laboratory tests. 

However, when we tested conditions representative of the actual phosphogypsum tailings in our original study (Dr = 39%, σ’v0 = 500-1,500 kPa), the results told a different story. Even with high static bias ratios up to α = 0.2 and elevated cyclic stress ratios of CSR = 0.08, our simulations consistently showed cyclic mobility rather than flow liquefaction with the designated initial stress level. The material exhibited large deformations without significant shear resistance loss, and most elements in the liquefiable zone had α values below 0.05.

The Results: Practical Implications for Dam Safety

We conducted sensitivity analyses varying the Ru threshold from 0.8 down to 0.7 and 0.6, which demonstrated that lowering the threshold increased the number of elements classified as liquefied, reduced factors of safety (from 1.19 to 1.16 and ultimately 1.02), and shifted potential slip surfaces downslope. While this might seem to suggest the use of lower thresholds for conservatism, we caution against blanket application.​ 

Our key insight: applying a uniformly low Ru threshold may misclassify cyclic mobility as flow liquefaction, inappropriately including non-liquefied elements with negligible shear bias in post-liquefaction residual strength calculations. This matters because cyclic mobility elements with high α but low Ru don’t require being labelled “liquefied” as true flow liquefaction zones because their shear resistances do not lose.​ 

What’s Next? 

Our closure underscores the need for more refined post-liquefaction assessment criteria that incorporate multiple parameters, such as Ru, shear strain, initial static shear stress ratio, and material type, along with targeted single-element verification using advanced constitutive models like P2PSand. While developing such comprehensive criteria extends beyond this technical exchange, our work demonstrates how advanced numerical modeling can distinguish subtle but critical differences in soil behavior.​​ 

I’ll be presenting related research on advanced constitutive modeling for tailings at upcoming geotechnical conferences (Geo-Congress, U.S. Rock Mechanics / Geomechanics Symposium, and Tailings and Mine Waste). To learn more about how ITASCA’s P2PSand model can improve your liquefaction assessments for tailings dams and other critical infrastructure, contact our geotechnical engineering team or explore our resources on cyclic loading and liquefaction analysis.

References

Kim, J., Athanasopoulos-Zekkos, A., & Zekkos, D. (2024). The effect of initial static shear stress on liquefaction triggering of coarse-grained materials. J. Geotech. Geoenviron. Eng. 150(10), 04024099. https://doi.org/10.1061/JGGEFK.GTENG-12282 

Reid, D., Fanni, R., & Fourie, A. (2024). Drained static bias effects on very loose silt tailings. Jpn. Geotech. Soc. Spec. Publ. 10(27), 995–1000. https://doi.org/10.3208/jgssp.v10.OS-16-03 

Author