Characterizing root architecture of riparian vegetation for assessing bank erosion potential in Queensland Rivers: A stochastic framework to integrate field and LIDAR data

Abstract

Key Points 堉n this study we outline a method to upscale site specific species assemblage root characteristics to the catchment scale 堅valuation of model results for 49 sites, across a broad geographic range along the east coast of Queensland, produced unbiased predictions. 堒esults show that is possible to focus on assemblages instead of species, making it easier than other models to apply to the diverse species compositions found in Australia. 堏ur analysis showed that there was as much variation in root strength characteristics within species (from different sites) as between species. Abstracts Sediment budget models predict riverbank erosion using the presence of riparian vegetation as a main factor controlling bank erosion across river systems. The way these catchment scale models relate bank erosion to riparian vegetation is; however, extremely crude. Site scale models rely on the parameters of root tensile strength and root architecture to quantify the reinforcement provided by roots in riverbanks. Root diameter is used to estimate the tensile strength of individual roots, whereas root architecture describes the abundance, root diameter and spatial distribution of roots across the bank face. Limitations of detailed site scale models are that they have been developed a low diversity of tree species in temperate climes, this model structure does not provide the necessary information for application at a catchment scale in tropical environments. In this work, we propose a stochastic approach to upscale root architecture data, collected during extensive fieldwork, to the catchment scale using 1 m 2 resolution vegetation information (canopy high and projected foliage cover) derived from LIDAR. We focused our data collection and analysis at species and assemblage levels in order to better characterize forest structures and targeting key dominant species in the reach. The non-parametric Spearman Rank Correlation Coefficient was calculated between field site root architecture data and LIDAR imagery. A probability density function was then fitted to the field data. Several analytical functions were tested using a Kolmogorov-Smirnov and ranked according to their respective p-values. Monte Carlo simulations were performed, constrained by the Spearman index, using the parameters of the chosen analytical functions found for each site and each distinct riparian forest structure. This method provided a way to upscale root characteristics, essential for sediment budget models. The stochastic nature of the process allowed the quantification and reporting of model uncertainties. Finally, this framework is capable of characterizing the strength provided to a riverbank by the roots of a vegetation community with a highly diverse species composition, typical of those found in Queensland rivers.