An RGIK rock glacier inventory compiled independently by two novel (trained) operators in Val Grisenche, Italy: Degree of consistency and training needs

Abstract

The Rock Glacier Inventories and Kinematics (RGIK) guidelines provide a standardized methodological framework with technical definitions and criteria for compiling consistent rock glacier inventories. However, the level of consistency attainable under these guidelines remains unclear, particularly regarding which components of the RGIK protocol pose the greatest challenges to novice operators (e.g., students or stakeholders) and how the associated learning curve unfolds. In this contribution, we evaluate inventories compiled in Val Grisenche (Western European Alps) by two graduate fellows (Op2 and Op3) against a reference benchmark produced by an expert scientist (Op1). The two novice operators differed in academic background but shared two months of identical RGIK training. Methods included traditional rock glacier detection and mapping on optical imagery and LiDAR-derived hillshade, complemented by kinematic characterization of rock glacier activity using Sentinel-1 InSAR-derived information. Results indicate that the detection of rock glacier units on optical imagery, as well as the identification of slow-moving areas (1–3 cm yr −1 and 3–10 cm yr −1 ) on long-duration (typically noisier) interferograms, represent the most critical methodological components requiring additional training. Manual detection of rock glacier units achieved a success rate of 60%, implying that 40% went undetected. When considering rock glacier systems, the success rate dropped to 42%, partly due to overall system missed detection (20%), but primarily because of unit misrecognition within systems (38%). Additionally, novice operators inadvertently identified a non-negligible proportion (17–26%) of spurious rock glaciers, most of which proved to be relict glacial depositional features. The dominant types of missed detections, in decreasing order of importance, included: (i) relict, densely vegetated rock glaciers; (ii) composite (mostly active) rock glacier systems inadequately subdivided into units; (iii) textbook-like cases (active, transitional, and relict) that, unexpectedly, were overlooked; and (iv) active and transitional rock glaciers with somewhat ill-defined fronts and lateral margins. Unsurprisingly, landform complexity was found to deteriorate performance scores, with multi-unit systems being more challenging for novice operators than mono-unit ones—particularly multi-lobe, active and transitional rock glaciers, which require the ability to integrate geomorphological and kinematic information across multiple iterative rounds. We argue that this competency develops only with growing experience. Overall, inter-operator heterogeneity emerged in both geomorphological detection and kinematic (InSAR-based) characterization; however, these differences did not appear to depend on the prior academic preparation of the trainee. By identifying likely pitfalls, this work assists novice operators in addressing specific learning objectives while building proficiency in rock glacier inventorying. Importantly, it also provides an empirical basis for developing more targeted RGIK training materials.

Bibliographic Information