Course:EOSC311/2025/Slippery Slopes: A Study of Landslide Patterns and Predictive Risks

Landslides are more than just a force of nature, they are an intersection of geology, climate, human activity and risk. In this project, I wanted to explore how and why landslides occur, and how scientists use data and models to predict their occurrence. My interest in this topic began on a road trip to Spiti Valley where a sudden landslide blocked the route just 10 minutes after my car passed through. That moment made me realise how real and immediate these hazards are, especially in mountainous regions like the Himalayas.

In this project, I look at the science behind landslides, where they commonly occur, and the predictive technologies that are being developed to save lives. By weaving together stories, research, and simple explanations of how models work, I hope to show that even topics grounded in earth science can be approachable, relatable, and deeply important.

Statement of connection and why you chose it

I chose to focus on landslides because of a personal experience that left a deep impression on me. A few years ago, I found myself trapped on a mountain road in northern India which was blocked by one landslide ahead and another rockfall behind. That night, stuck in a car on a narrow muddy road between a mountain and a river, I started thinking about how little control we have over natural disasters like this. It was eerie, and honestly a bit terrifying, to know that had we been slightly delayed, we might have been caught in it. That experience stayed with me.

As a math student who’s also interested in data science, I’ve always looked for real-world problems where numbers and patterns can help us make better decisions. Predicting natural disasters like landslides is one of those areas. Even though I don’t specialise in geology or meteorology, I’ve found that I can understand and appreciate the work that goes into landslide modelling, and that I can communicate it in a way that others might find interesting too. This project gave me a chance to combine a personal connection, an analytical mindset, and a growing curiosity about how we use technology to keep people safe.

Main text

a) What Triggers a Landslide?

Landslides might look like sudden, chaotic events but they’re usually the result of multiple factors that build slowly over time coming together. At the most basic level, a landslide occurs when the pull of gravity overcomes the strength holding soil, rocks, or debris in place on a slope. Once that balance tips, everything starts to slide down.

There are many things that can trigger this imbalance:

Heavy rainfall: Water is a major player. When it rains a lot over a short period, the soil becomes saturated, loses its grip, and begins to slip. In fact, more than 90% of landslides worldwide are triggered by rainfall.
Earthquakes: The shaking from even a moderate quake can loosen slopes, especially if they’re already unstable or saturated.
Human activities: Deforestation, road construction, and mining can remove vegetation and cut into slopes, reducing natural support systems and making landslides more likely.
Geological conditions: The type of rock and soil also matters. Loose, unconsolidated material is more likely to move than solid bedrock. Clay rich soils, for example, are notorious for becoming slippery when wet.
Slope angle: The steeper the slope, the higher the risk especially when the other factors (like rain or weak soil) are in play.

One thing that surprised me was how small changes like a blocked drainage system or a slight shift in the soil’s water content can set off something massive. It really shows how interconnected natural systems are.

b) Where Do Landslides Happen and Why?

When I first started reading about landslides, I assumed they mostly happened in remote mountain areas. But I quickly realized that landslides are a global issue and they're not just limited to rugged terrain. Still, there are certain regions that are far more vulnerable.

Mountainous regions, like the Himalayas, the Andes, and parts of Southeast Asia, are particularly landslide-prone due to steep slopes, tectonic activity, and intense seasonal rainfall. My own experience in the Spiti Valley was part of a broader pattern: according to global landslide inventories, this region faces frequent slope failures, especially during monsoon months.

Interestingly, landslides are also a big concern in urbanising areas, especially where development is happening faster than geological assessments. For example, parts of Central America and Southeast Asia are seeing increased landslide frequency due to rapid deforestation and construction on unstable slopes.

I also came across a concept called “hotspot mapping", where scientists use historical landslide records, slope gradients, rainfall patterns, and land use data to predict future risk zones.

c) How We Predict Them: Models and Math

One of the most fascinating parts of this project was learning how researchers actually predict landslides. It turns out that this is a complex and rapidly evolving field, combining geology, meteorology, math, and machine learning.

There are generally two main types of models used in prediction: empirical models and physics based models. Empirical models are data-driven. They use patterns from past landslide events to estimate future risk similar to how machine learning works. For example, if a certain type of slope in a particular climate has failed before under heavy rainfall, it’s likely to fail again under similar conditions. Physics based models, on the other hand, are more technical. They simulate real world physics by calculating forces acting on the slope like gravity, water pressure, soil strength, and friction to see how likely it is to collapse.

But what I found especially exciting is the use of machine learning and near-real-time monitoring systems. A standout example is NASA’s Landslide Hazard Assessment for Situational Awareness (LHASA) model. LHASA combines satellite data, global rainfall measurements, and a simple decision-tree algorithm to issue near real time landslide warnings across the globe. It uses thresholds based on known rainfall-triggered landslide events, allowing researchers to predict whether current rainfall levels could lead to slope failure in a specific region (Stanley et al., 2021). In its newest version, LHASA 2.0, it even incorporates susceptibility maps that assess underlying terrain conditions like slope angle and land use to better refine its alerts.

What’s impressive is that this model updates every 30 minutes and covers regions globally especially useful in areas where ground based monitoring is limited or too expensive. It’s still far from perfect, but it shows how satellite data and algorithms can provide early warnings and save lives.

d) What Makes Prediction So Difficult?

Despite the progress in data collection and modeling, accurately predicting landslides remains incredibly challenging. One major limitation is the lack of ground-level data, especially in remote or mountainous regions. Models like LHASA depend heavily on satellite-derived rainfall estimates and terrain mapping, but they can't always capture local soil moisture conditions, which are crucial for predicting slope failure (Kirschbaum et al., 2019).

Another issue is the rapidly changing weather patterns due to climate change. Sudden and extreme precipitation events can trigger unexpected failures in areas previously considered low-risk. Even the best models can’t always keep up with how quickly local conditions change, and there's always uncertainty about thresholds, for example, how much rain over how much time actually triggers a slide?

Additionally, terrain accessibility can make installing ground sensors and conducting geological surveys extremely difficult. In places with ongoing conflict, political instability, or difficult access routes, maintaining up-to-date information becomes nearly impossible. These uncertainties mean that landslide warnings, while helpful, must always be interpreted with caution and local context in mind.

e) Applications and Why It Matters

Despite the challenges, landslide prediction and hazard mapping are incredibly important. Knowing which areas are vulnerable helps communities plan safer infrastructure like building roads away from unstable slopes or reinforcing bridges and tunnels (Fell et al., 2008). In some countries, like Japan, slope monitoring systems combined with rainfall thresholds have been used for decades. They even issue automated landslide warnings when conditions are met, which can help save lives by giving residents time to evacuate.

In India, especially in Himalayan regions like Himachal Pradesh and Uttarakhand, there have been major efforts to install retaining walls, steel nets, and drainage systems to stabilize slopes and reduce runoff (Sharma et al., 2020). These kinds of engineering solutions work alongside prediction models to create comprehensive mitigation strategies.

Ultimately, this topic matters because it combines geology, mathematics, and human impact. By understanding how and why landslides happen and how we can better predict and prevent them we’re not just analysing numbers. We’re contributing to public safety, infrastructure resilience, and smarter planning in a world that’s becoming more vulnerable to extreme events.

Conclusion

Working on this project gave me a deeper appreciation for how something as abstract as mathematics can directly impact real world safety and decision-making. Landslides are more than just geological events they’re deeply human in their consequences, and improving how we understand and predict them can literally save lives. As someone who’s personally experienced the uncertainty and fear of being caught in one, learning how experts use logistic regression, probability modelling, and satellite data to map risk felt meaningful.

I’ve also come to realise that while we may never eliminate the uncertainty in natural disasters, we can definitely make better decisions with the right tools. It’s empowering to see how mathematics, combined with technology and earth science, can support early warning systems, guide infrastructure development, and reduce risk in vulnerable areas.

This project reminded me that no field of study exists in a vacuum. The connection between geology and math isn’t just theoretical it’s critical, timely, and very real. I now feel more motivated to pursue opportunities that lie at the intersection of data science, environmental risk, and public impact.

References

Landslide Basics. (2008, November 15). USGS. https://www.usgs.gov/natural-hazards/landslide-hazards/science/landslides-101
Li, X., Ma, P., Xu, S., Zhang, H., Wang, C., Fan, Y., & Tang, Y. (2024). Slow-moving landslide hazard assessment using LS-Unilab deep learning model with highlighted InSAR deformation signal. Remote Sensing, 16(24), 4641. https://doi.org/10.3390/rs16244641
Chen, T.-H. K., Prishchepov, A. V., Fensholt, R., & Sabel, C. E. (2020). Detecting and monitoring long-term landslides in urbanized areas with nighttime light data and multi-seasonal Landsat imagery across Taiwan from 1998 to 2017. https://arxiv.org/abs/2009.07954
Dai, X., Chen, J., Zhang, T., & Xue, C. (2025). Integrated landslide risk assessment via a landslide susceptibility model based on intelligent optimization algorithms. Remote Sensing, 17(3), 545. https://doi.org/10.3390/rs17030545
Guzzetti, F., Reichenbach, P., Cardinali, M., Galli, M., & Ardizzone, F. (2005). Landslide hazard assessment in the Collazzone area, Umbria, Central Italy. Natural Hazards and Earth System Sciences, 5(1), 71–84. https://doi.org/10.5194/nhess-5-71-2005
Kirschbaum, D. B., Stanley, T., & Simmons, J. (2020). A global landslide hazard model for near real-time assessment of rainfall-triggered landslides. *Geophysical Research Letters*, 47(5). https://doi.org/10.1029/2019GL085378