PS 34-255 - Can cottonwood establishment locations on a dynamic, regulated river be accurately predicted?

Abstract: Riparian vegetation management is a vital component to river restoration projects, stressing the need to better identify where vegetation is most likely to establish in dynamic river settings. Determination of the most limiting factors influencing vegetation recruitment remains a barrier to river restoration success. The goal of this study was to map the spatial pattern of young cottonwoods along the lower Yuba River (LYR) in Northern California and to investigate the principal topographic and hydro-physical variables driving young cottonwood presence and absence. The Riparian Seedling Recruitment module, within an open-source software called River Architect, was used to predict theoretically suitable areas of Populus fremontii (Fremont cottonwood) seedling establishment along 34 km of the LYR for the years 2017 to 2021 considering 6 levels of suitability. Predictions were compared against 2,957 Fremont cottonwood field observations collected by using an equal effort sampling of randomly selected sites that had been stratified by five geomorphic reaches of the LYR and model results. The 4 hydro-physical outputs of the Riparian Seedling Recruitment module (scour, inundation, recession rate, and bed preparation) and 16 topographic variable rasters (such as channel proximity or bed slope) were then used as predictors in a Random Forest machine learning model to identify the main drivers of young cottonwood presence. The results of the machine learning model indicated that topographic variables were the main drivers in cottonwood presence, while the hydro-physical variables were not as significant. This information can be used to advise suitable locations or topographic characteristics needed for riparian vegetation restoration efforts along the LYR, while the methods could be applied to other rivers given the availability of suitable data inputs.