Author(s): Beatrice Pinna; Alex Laini; Gemma Burgazzi; Simone Guareschi; Paolo Vezza

Linked Author(s): Paolo Vezza

Keywords: Macroinvertebrate communities; Random Forest; Mesoscale habitat modeling; Ecological flows

Abstract: The Flow-T model, the first mesoscale habitat model developed for river macroinvertebrate communities, was refined and tested across multiple hydromorphological contexts in Northern Italy. The study aimed to evaluate its transferability and predictive accuracy in different river typologies, including braided, meandering, cascade, step-pool, and straight channel morphologies. New datasets collected in 2023-2024 from nineteen river and stream reaches were integrated with previous dataset (2019–2020), resulting in an expanded model calibrated and validated using a Random Forest (RF) approach. The refined model selected the 12 descriptors among classes of water surface gradient, flow velocity, substrate composition, water depth, and longitudinal connectivity, and achieved higher accuracy in both cross-validation (R² = 0.79) and pure validation (R² = 0.68). The demonstrated transferability and predictive power of the re-trained model confirm the extension of the MesoHABSIM methodology to a new ecological target.

DOI: https://doi.org/10.64697/iahr.proc.ise2026.abs.199

Year: 2026