Hydro-meteorological Implications for Shallow Landslide Optimization in Utuado and Toro Negro, PR

Abstract

Puerto Rico is extremely susceptible to rainfall induced landslides due to its tropical setting along with the continuous exposure to atmospheric phenomena such as hurricanes, that promote slope instability, especially in sites modified or disrupted by human activity. It is imperative to mitigate the recurrent risk that these hazards pose on society to avoid tragedies like the Mameyes, Ponce rainfall induced landslide in 1985, that claimed approximately 129 lives and is known as the worst landslide disaster in this U.S. jurisdiction to date (Jibson, 1987). In an effort to further understand various parameters that initiate landslides and provide a foundation for a landslide early warning system, this research quantifies and determines hydro-meteorological bilinear thresholds taking into consideration soil moisture values. This was done by analyzing and assessing the data collected from two monitoring stations with different soil types, located in Utuado (sandy, with low storage potential) and Toro Negro (clay and silt, with high storage potential), in the Cordillera Central (about 17 km apart). Both stations have an array of dielectric sensors and a piezometer that record the soil moisture (volumetric water content), suction, and pressure of the subsurface. The bilinear thresholds are optimized by a set of predetermined statistics with hypothesized “landslide times”. All data is processed by a program called ‘Hydromet‘, developed using the Python programming language (Conrad et al., 2021). Potential landslide times were selected by thinking in terms of positive pore-pressure and high volumetric water content in abrupt increments that suggest prime conditions for land mobility. Results indicate that hydro-meteorological bilinear thresholds for soils with low storage potential (i.e. drain out water readily, such as soils in Utuado) performed better with this statistical approach in comparison to the soil types that remain wet throughout the year (i.e. Toro Negro). This is evidenced by the area under the curve (AUC) values and statistical scores used to evaluate the performance for each station. Utuado had a consistent AUC of 0.89 and scores that ranged from 0.36-1.0 (closer to one equals a better score). In contrast, Toro Negro had a consistent AUC value of 0.72 with scores that ranged from 0.0-0.44. However, it was apparent that for the Toro Negro (high storage potential) dataset, some statistics from the Hydromet program showed some inconsistencies when calculating thresholds. This could be attributed to its constant saturated conditions that imply a negligible effect in soil moisture. Overall, this study proved the monitoring system to be functional and beneficial as a future/preliminary forecasting and nowcasting early warning system. It can also serve as an impressive mitigating tool as it takes a different approach to former linear thresholds that only take into consideration rainfall versus intensity duration, disregarding soil moisture, which is a crucial parameter when speaking about low storage potential soils.