Landslide history shapes landscape diversity: Applying hyperspectral data for functional diversity monitoring of tropical mountainous ecosystems

The global biodiversity crisis emphasizes the importance of diversity monitoring to examine ecosystem stability and resilience, including regeneration capacity. As a critical driver of change in tropical mountains, landslides alter the structure, composition, and function of landscapes. One possibility to study the large-scale causes and consequences of landslides on diversity is to use remote sensing to characterize ecosystem traits and functions at several spatial and temporal scales. An increasing availability of satellite-borne hyperspectral offers the possibility to capture morphological and physiological traits of vegetation to characterize functional diversity in areas affected by landslides. Using hyperspectral data to characterize functional diversity often involves the removal of bare soil to eliminate background reflectance. Given that landslides of different ages contain a mixture of vegetated and bare soil pixels, the challenge is to incorporate the latter into image processing, and ultimately into metrics that provide an integrative functional characterization of areas undergoing succession. We define landscape diversity as the structural, functional, and historical characteristics of ecosystems and may be useful to expand functional diversity monitoring beyond purely vegetated areas. Using a historical landslide database and the novel PRISMA (PRecursore IperSpettrale della Missione Applicativa) hyperspectral data we addressed two questions to assess the role of landslide history on landscape diversity. First, do areas with different landslide histories exhibit distinct functional trait and landscape diversity patterns? Second, which landscape traits distinguish landslides of different recovery stage considering landslide size, age, frequency, and reactivation status?

To address these two questions, we derived two sets of variables that represent landslide history and landscape diversity. To represent landslide history, we processed historic and current remotely sensed data from the Sierra de Las Minas (SLM) mountains in eastern Guatemala to create a geodatabase that includes landslide inventories (1973 – 2021) in which each landslide is characterized by age, size, and shape. In ArcGIS Pro we identified degree of overlap among landslides from all inventories to mark landslide reactivation. Specifically, a model identifying landslide overlap distinguished landslides that occur once from landslides that occur repeatedly. To represent landscape diversity, we processed PRISMA data to create functional indices representative of vegetation and soil traits across the SLM. To include areas across all stages of succession, both traits of vegetation and bare soil three separate indices were created. A first masked out bare ground, a second masked out vegetation, and a third version combined the vegetation and bare ground. Finally, we examined the relationship between landslide history and the three versions of functional indices using geographic Random Forest algorithm. The outcomes of this study could reveal lasting structural, compositional, and functional impacts of landslides in tropical mountains, which serve as critical safeholds for biodiversity and ecosystem services during drastic global change.