Half-Life Rainfall Models Improve Landslide Alerts in Brazil

Santos and Guarujá, two densely populated municipalities along the Atlantic coast of São Paulo state in Brazil, have a long and tragic history of shallow landslides. These slope failures, which typically occur in immature residual soils overlying gneissic and migmatitic bedrocks, generally reach depths of up to three meters. The region’s susceptibility to such disasters was starkly highlighted in March 2020, when an extreme rainfall event triggered a major catastrophe that resulted in 41 fatalities. In response to these recurring hazards, researchers have investigated more precise methods for predicting landslide triggers by analyzing the relationship between rainfall patterns and slope stability over a five-year period from 2018 to 2022.

The study, published in the journal Natural Hazards, focuses on the application of effective half-life precipitation indices to improve the accuracy of landslide early warning systems. Traditional methods often rely on simple accumulated rainfall measurements, which may not fully account for the antecedent moisture conditions in the soil. By contrast, the half-life concept, derived from hydrological tank models, defines the time required for the water content in the ground to be reduced by half. This approach allows for a more nuanced understanding of how past rainfall events contribute to current soil saturation and, consequently, the likelihood of slope failure.

Researchers utilized data from the Automatic Rainfall Half-Life Cluster System, or ARHCS, to compute rainfall accumulated over 1, 12, and 96 hours. The dataset included 136 landslide occurrences in Santos and 178 in Guarujá, alongside 665 non-occurrence rainfall events. By applying K-means clustering, the team aimed to identify distinct rainfall reference levels associated with different landslide scenarios, ranging from induced and occasional failures to multiple and generalized events. This statistical approach helps to distinguish between rainfall conditions that are likely to trigger landslides and those that are not, thereby refining the thresholds used in civil defense planning.

The results indicate that effective half-life rainfall indices significantly reduce data dispersion compared to simple rainfall indices. When analyzed using cluster analysis, the half-life method produced more compact and cohesive groups of data points, suggesting a better representation of the hydrological processes controlling slope stability. The established threshold levels were found to be consistent with the empirical curves adopted by the São Paulo Preventive Civil Defense Plan, providing a robust framework for local disaster risk management. Furthermore, soil moisture data from CEMADEN stations offered additional support for interpreting the complex relationship between rainfall and landslide occurrence.

Validation of the models demonstrated a clear advantage for the half-life method over the simple accumulation approach. The predictive performance for the half-life method reached 98%, compared to 88% for the simple method. This improvement is particularly significant for the induced landslide level, where the true positive rate was higher for the half-life indices in both municipalities. For instance, in Guarujá, the true positive rate for the induced level was 1 for the half-life method versus 0.893 for the simple method, while in Santos, it was 0.973 versus 0.868. These figures suggest that the half-life method is more effective at correctly identifying rainfall conditions that lead to shallow landslides.

The study also highlights the importance of antecedent rainfall in triggering shallow landslides. The extreme event of March 2020, which caused widespread destruction, was preceded by months of above-average precipitation. In February 2020, monthly rainfall in Guarujá exceeded 566 mm, and in Santos, it reached 700 mm, both significantly higher than the monthly averages. This prolonged antecedent rainfall led to high soil moisture levels, which, when combined with the intense triggering rainfall of almost 400 mm in 3.5 hours, resulted in numerous slope failures. The data show that volumetric soil moisture peaks often precede rainfall peaks, indicating that the ground was already saturated before the final triggering event.

Despite the promising results, the researchers acknowledge several limitations. The exclusion of low-magnitude rainfall indices associated with anthropogenic factors, such as slope cutting and infrastructure leaks, was necessary to isolate rainfall-triggered landslides but may overlook certain types of failures. Additionally, the clustering approach depends on methodological assumptions that could affect reproducibility. The study emphasizes that rainfall alone does not completely explain landslide occurrences, particularly in urbanized areas where human-induced alterations play a crucial role in shaping slope stability. Therefore, integrating land-use and infrastructure considerations into early warning systems remains essential.

From an operational perspective, the findings have direct implications for the landslide early warning systems in the Baixada Santista region. The improved definition of critical rainfall thresholds through the use of effective half-life indices can enhance the accuracy and reliability of alerts. The combination of effective half-life accumulated rainfall and volumetric soil moisture data offers a more comprehensive approach, capturing both triggering and preconditioning factors. Although the preliminary agreement between soil moisture data and rainfall indices is promising, full quantitative integration requires local-specific soil water retention curves and continued empirical validation.

The study concludes that the integration of effective half-life accumulated rainfall indices, cluster analysis, and soil moisture data provides a flexible framework adaptable to different environments. Future research should focus on incorporating advanced modeling techniques, such as multiple effective antecedent rainfall indices and 3D critical rainfall thresholds, to further improve the spatial and temporal resolution of landslide predictions. The methodological tools developed in this study are intended to be applied to the GeoRisk automated landslide prediction and early warning system of CEMADEN/MCTI, marking a meaningful step toward more reliable and operational landslide early warning systems in Brazil and potentially other regions with similar climatic and geological conditions.

Subject of Research: Rainfall thresholds for shallow landslides using effective half-life indices in Santos and Guarujá, Brazil

Article Title: Exploring rainfall thresholds for shallow landslides using effective half-life precipitation indices and cluster analysis: a case study in Santos and Guarujá, Brazil

Article References: Metodiev, D., Mendes, R. M., de Andrade, M. R. M., Segoni, S., Nocentini, N., Bortolozo, C. A., Canavesi, V., Camarinha, P. I. M., da Silva do Valle, V. A., Kuhlmann, G. G., & Simões, J. V. O. (2026). Exploring rainfall thresholds for shallow landslides using effective half-life precipitation indices and cluster analysis: a case study in Santos and Guarujá, Brazil. Natural Hazards, 122(22), Article 697. https://doi.org/10.1007/s11069-026-08475-1

Image Credits: AI Generated

DOI: 10.1007/s11069-026-08475-1

Keywords: landslides, rainfall thresholds, early warning systems, Brazil, hydrology, cluster analysis, soil moisture, natural hazards, civil defense, slope stability, Exploring, rainfall