Various characteristics of the hydrological regime (frequency and intensity of low and high stage level, seasonality and predictability) influence the structure of aquatic communities. Amongst tools available to assess and simulate the impacts of flow management, the use of physical habitat models is well spread. These models link a hydraulic model of stream reaches, that predicts how microhabitat hydraulics change as a function of discharge, with biological models reflecting organism preferences for their microhabitat conditions. In the last decades, developments in Europe and others continents have led to the generation of new habitat models based on statistical approaches. We describe here the development of statistical hydraulic models in reaches of steep tropical streams. The method consists in modelling (by a parametric probability function) the frequency distributions of point hydraulic variables (e.g., velocity and depth) within the reach. Previous works in temperate systems showed that the parameters of the distribution function could be predicted from reach-characteristics, as mean reach width and depth. In turn, mean reach width-discharge and depth-discharge of reaches (i.e. at-a-reach hydraulic geometry relationships) can be modelled from simple measurements at two or more discharge rates. Therefore, the use of statistical hydraulic model is less time-consuming and contributes to simplify the technical application of habitat models for ecologists. Statistical hydraulic models haven't been tested in tropical climate and in steep streams. Therefore, we tested the transferability of European models (predicting water depth and velocity distributions) to tropical rivers in French islands (Caribbean and Indo-Pacific region), where strong climatic events occur. We collected extensive hydraulic data in 40 sites situated in a variety of geomorphologic contexts. These sites had complex flow structure, had slopes generally >5%, high relative roughness, torrential flows and were characterized by step-pool sequences.