From rockfall source areas identification to susceptibility zonation: a proposed workflow tested in El Hierro (Canary Islands, Spain)

Abstract. Accurate rockfall modeling is crucial for evaluating rockfall hazards and requires consideration of several inputs data, including parameters that control boulder trajectories and source areas. Inaccurate definitions of source areas can lead to unrealistic representations of the rockfall process. In this study, we analyze how different approaches used to define source areas can affect the accuracy of rockfall modeling. The island of El Hierro (Canary Islands, Spain) is selected due to its geological and geomorphological characteristics, as well as the socio-economic importance of rockfalls on the island.

To assess rockfall source areas, three different approaches were considered, ranging from situations with limited data availability to scenarios with many topographic, geological and geomorphological information.

A morphometric firstly approach establishes a slope angle threshold above which block detachment zones are considered. For the second approach, we have employed a statistical method to identify rockfall source areas, using Empirical Cumulative Distribution Functions (ECDF) of slope angle values. The third method was a probabilistic modeling framework that applies a combination of multiple multivariate statistical classification models. These models use the mapped source areas as a dependent variable, as well as a set of thematic information as independent variables.

The source area maps obtained from the three methods were used as inputs for a rockfall runout model, to establish a classification of rockfall susceptibility areas.

One of the main outcome of the rockfall modeling simulations on El Hierro is the rockfall trajectory counts maps, showing areas prone to rockfalls. These maps indicate the probability of a given pixel being affected by a rockfall event. Two classification approaches were applied to generate the probabilistic susceptibility maps: unsupervised and supervised statistical methods by using distribution functions. The unsupervised classification only employs as input the raster map of the rockfall trajectory counts. In contrast, the supervised classification requires additional data on the areas already affected by rockfalls. Finally, six susceptibility maps are developed and compared to highlight the influence of source areas definition on the distribution of rockfall trajectories.