Turbidity currents triggered at river mouths form an important highway for sediment, organic carbon, and nutrients to the deep sea. Consequently, it has been proposed that the deposits of these flood-triggered turbidity currents provide important long-term records of past river floods, continental erosion, and climate. Various depositional models have been suggested to identify river-flood-triggered turbidite deposits, which are largely based on the assumption that a characteristic velocity structure of the flood-triggered turbidity current is preserved as a recognizable vertical grain size trend in their deposits. Four criteria have been proposed for the velocity structure of flood-triggered turbidity currents: prolonged flow duration; a gradual increase in velocity; cyclicity of velocity magnitude; and a low peak velocity. However, very few direct observations of flood-triggered turbidity currents exist to test these proposed velocity structures. Here we present direct measurements from the Var Canyon, offshore Nice in the Mediterranean Sea. An acoustic Doppler current profiler was located 6 km offshore from the river mouth, and provided detailed velocity measurements that can be directly linked to the state of the river. Another mooring, positioned 16 km offshore, showed how this velocity structure evolved down-canyon. Three turbidity currents were measured at these moorings, two of which are associated with river floods. The third event was not linked to a river flood and was most likely triggered by a seabed slope failure. The multi-pulsed and prolonged velocity structure of all three (flood- and landslide-triggered) events is similar at the first mooring, suggesting that it may not be diagnostic of flood triggering. Indeed, the event that was most likely triggered by a slope failure matched the four flood-triggered criteria best, as it had prolonged duration, cyclicity, low velocity, and a gradual onset. Hence, previously assumed velocity-structure criteria used to identify flood-triggered turbidity currents may be produced by other triggers. Next, this study shows how the proximal multi-pulsed velocity structure reorganizes down-canyon to produce a single velocity pulse. Such rapid-onset, single-pulse velocity structure has previously been linked to landslide-triggered events. Flows recorded in this study show amalgamation of multiple velocity pulses leading to shredding of the flood signal, so that the original initiation mechanism is no longer discernible at just 16 km from the river mouth. Recognizing flood-triggered turbidity currents and their deposits may thus be challenging, as similar velocity structures can be formed by different triggers, and this proximal velocity structure can rapidly be lost due to self-organization of the turbidity current.