Predicting habitat suitability for lotic fish: linking statistical hydraulic models with multivariate habitat use models

Quantitative estimates of habitat suitability in a stream reach generally result from coupling a hydraulic habitat model with a biological model of habitat use. The choice of each of these models has led to much controversy and discussion. Nevertheless, most habitat studies of lotic fish use a deterministic hydraulic model and univariate suitability curves. The objective of this contribution is to present a new, alternative method, which relates statistical hydraulic models to multivariate habitat use models. Our statistical hydraulic models predict the frequency distributions of hydraulic variables such as velocity or water depth within stream reaches. Their main advantage is the simplicity of their input variables (mainly discharge and average characteristics of the reach). Our multivariate formulation of habitat use models takes into account the local variability of fish habitat, predicting habitat suitability as a function of the frequency distribution of hydraulic variables within the local fish habitat. We demonstrate how these two model types can be linked to estimate habitat suitability in a stream reach as a function of discharge, focusing on two fish species (barbel, chub) in a regulated reach of the French Rhône River. The main limitations of this new method are a result of mathematical constraints associated with the linkage of the two modelling approaches and to uncertainties in transferring biological models from one stream to another because of insufficient data. Despite these limitations, the method provides solutions to several critical problems facing existing approaches and the simplicity of its input variables can accelerate the validation process of habitat models. Therefore, our first simulations strongly encourage: (i) the use of statistical approaches to describe hydraulic variables; and (ii) the study of multivariate habitat use models that apply to a large variety of streams.