The feasibility of using the proposed relationship between model-generated upper-tropospheric potential vorticity (PV) fields and Meteosat water vapour images in forecast initialisation has been investigated by a number of case studies involving model forecasts of North Atlantic cyclones. The objective of each case study was to reduce the errors in the forecast track and depth of a cyclone by re-initialising the forecast via upper-level PV adjustment and PV inversion. The methodology has employed the PV-water vapour relationship to detect mismatches between PV fields and water vapour images, and has attempted to correct these mismatches by manual adjustment of the PV fields. These PV adjustments have been converted to wind and temperature increments by means of PV inversion and used to initialise the model fields in a way that preserves dynamical balance. Five case studies with small initialisation errors were carried out. The results of the case studies show some partial positive impacts on forecasts of cyclonic development, but overall the results do not demonstrate that the use of the PV-water vapour relation for forecast initialisation is a practical proposition for operational use.