(en) Soil hydraulic properties are needed for modelling below-ground water flow and solute movements. They are very variable in space and time and across scales and their characterisation is tedious. Pedotransfer functions (PTF) are tools developed to predict hydraulic properties from more readily available information. This thesis provides PTF predicting the parameters of a closed-form model of the soil hydraulic conductivity and moisture retention curves based on the soil texture, bulk density and organic carbon content. These PTF show their limits very close to saturation because the soil structure is not well apprehended by these indicators. With the aim of developing PTF considering the soil macroporosity, soils of homogeneous texture but contrasted structure (silt loam) are studied with a multistep outflow system and with fine resolution moisture retention measurements. It is first shown that the choice of the model has more impact than the spatial variability on extrapolation results and that, even with very good fitting performances, the prediction uncertainty can be very large. Assuming that land cover and horizon depth control the root architecture, the soil biological activity and land management practices, their impact on close to saturation hydraulic properties of silt loam soils is investigated. Results suggest that land cover should be included into the next generation of PTF. It is also shown that the presence of macropore is generalised in these soils. Finally, hydraulic properties of tropical soils derived from volcanic ashes are successfully linked with pedo-chemical and morphological indicators determined on exactly the same sample as the hydraulic measurements. Future use of this combination of information for developing PTF for tropical soils is therefore advised.