Several methods have been reported in the literature for the measurement of riverbank retreat (Lawler, 1993). During recent years, particular interest has been aroused in the application of terrestrial photogrammetry because of several advantages over other more traditional methods (Barker et alii, 1997; Pyle et alii, 1997). In this study, digital terrestrial photogrammetry has been applied to measure the retreat of a riverbank reach along Sieve River (Tuscany), with the following objectives: a) to verify potentialities and to evaluate advantages and disadvantages in applying such a method to monitor riverbank erosion processes; b) to combine measurements and field observations with modelling and analysis of the main processes responsible for the bank retreat. The riverbank has a height of 4.5-5 m, and it is composite, with a basal portion predominantly composed of gravel and an upper portion of fine-grained material (silt and sand). The length of the reach studied is about 150 m. Five photogrammetric survey points have been arranged along a baseline on the opposite river side, at a constant distance of about 23 m. A Metric Wild P31 Camera has been used for the photogrammetric surveys. Each photograph has a coverage of 22 m, with a stereoscopic overlap of 60% for each pair of photos. A series of targets of aluminium (size 15X15 cm) have been inserted on metallic pins. The pins have been arranged with a configuration including a series of 3 horizontal pins, inserted at different heights of the bank, and a vertical pin on the scarp edge, at a short distance from it. Such an arrangement has been adopted for 11 bank profiles, with a total of 44 targets. A detailed topographic survey has been performed using a total station Nikon DTMA5LG for the determination of plano-altimetric co-ordinates of the targets and of the photogrammetric survey points. Two photogrammetric surveys have been carried out, the first during March 1999 and the second during December 1999, after a flow event (19/11/1999) that caused a significant bank retreat. Image elaboration has been carried out adopting a specific software (StereoView) for digital photogrammetry. Digital Terrain Models (DTM) of the bank surface for the two photogrammetric surveys (before and after the event) have been obtained, allowing detailed reconstruction of the retreat which occurred. For this aim, the bank surface has been divided in 4 sub-reaches (each one corresponding to a stereoscopic pair). The comparison of DTMs has highlighted a significant retreat for most of the entire study reach, with more than 2 m of retreat in some places. Based on the two DTMs, it has been possible accurately to evaluate, with automatic procedure, the difference in volume (corresponding to the eroded material during the flow event), resulting in a total of 1007 m <sup>3</sup> . From field observations conducted during the flow event, the main mechanisms determining the bank retreat have been identified. Most of the processes of retreat have been observed during the drawdown phase of the event. The main mechanisms have been: a) small slides and/or falls in the upper portion of the bank; b) piping/sapping phenomena (Hagerty, 1991; Thorne, 1998), corresponding with a series of centimetre-sized layers of sorted sand in the middle portion of the bank; c) planar-rotational or slab-type failures (Thorne, 1998), involving all or most of the fine-grained portion of the bank. For a better understanding of causes and triggering conditions of the observed instability mechanisms, modelling of saturated/unsaturated flow within the bank has been performed, while hydraulic analyses for studying fluvial erosion have been carried out. Flow modelling has been performed by the software SEEP/W for two representative sections. Results have highlighted: a) a localised increase of seepage velocity corresponding to the sandy layers, due to the differences in conductivity with the other materials, such to explain the triggering of piping processes; b) the interaction between pore water pressures and river confining pressures is particularly unfavourable during the drawdown phase of the flow event, during which most of the mass movements have been observed along the bank. The stability of the basal portion of the bank has been taken into account introducing a dimensionless hydraulic stability number (Casagli et alii, 1999) and analysing its variations during the flow event. The intensity of fluvial erosion has been shown to follow the trend of river level, that is maximum at the peak stage. Based on the results of the analyses, combined with field observations, two schemes of bank evolution have been obtained for the two representative sections. In conclusion, the main advantages and disadvantages in using digital terrestrial photogrammetry have been summarised and discussed. Among the main advantages are: a) the possibility to measure the retreat of all the points of the investigated surface, rather than for a limited number of points; b) high degree of accuracy of the measurements; c) limited disturbance of the investigated surface. Among the disadvantages, the following need to be noted: a) disturbance due to vegetation on the bank surface; b) the method is expensive, in terms of equipment needed, costs, field operations, data processing; c) the method has limited applicability to relatively large rivers.