Leveraging the potential of satellite time series, cloud computing and artificial intelligence to quantify fluvial biogeomorphology across multiple scales

Riverine landscapes are shaped by the feedbacks between hydrological, geomorphological and ecological processes. These feedbacks occur across multiple scales, from the scale of single plants modifying the hydraulic forces around it to the formation of landforms like islands which in turn lead to the emergence of specific river types such as braided or anastomising. Over the past years, the field of biogeomorphology has significantly improved the understanding of the interaction of vegetation and hydro-morphological processes. Despite recent scientific progress, research gaps remain. In particular, it is still poorly understood, how processes happening on small scales, such as sedimentation in the lee of an individual plant or a piece of large wood, lead to the emergence of landforms and reach scale river types and how – vice versa – the specific landform pattern within river types foster small scale processes. The concept of Panarchy considering a number of adaptive cycles linking the different scales of the fluvial biogeomorphic system is a promising candidate for analyzing cross-scale vegetation-hydromorphology feedbacks. However, developing methods for quantitative studies is still an ongoing challenge in biogeomorphological research.

We introduce an empirical approach for filling this research gap driven by a combination of field mapping and state-of-the-art remote sensing taking the Naryn River, a large free flowing river in Kyrgyzstan, as a case study. In the field, we map vegetation traits and geomorphic characteristics and link them to the stages of the biogeomorphic succession concept. Then, we utilize the computational potential of the “Terrabyte” cloud computing platform of the German Aerospace Center (DLR) to analyze temporally dense time series from the Sentinel-1 and -2 archives. To map vegetation and hydro-geomorphic characteristics (vegetation height, density, biomass, share of bare sediment, grainsize, duration of inundation) and to assess how these biogeomorphic traits change over time, we make use of the capabilities of the recently available foundational deep learning model “Clay” as state-of-the-art artificial intelligence method in earth observation. This enables river corridor scale analysis of the spatial-temporal dynamics of hydro-geomorphic disturbance, rejuvenation potential (windows of opportunity), vegetation growth as well as the emergence of biogeomorphic feedback windows and therefore tracking the biogeomorphic succession. This gives us the possibility to study adaptive cycles on different scales and construct Panarchies for different river types occurring along the Naryn River. Our approach is a significant step towards the quantification of biogeomorphic feedbacks across multiple scales and advances the empirical understanding of the role of scale dependence of biogeomorphic feedbacks which lead to the emergence of riverine landscape pattern.