This paper aims at proposing a way to get round the intrinsic deadlocks of the economic assessment of climate change impacts (absence of consistent baseline scenario and of credible description of adaptation behaviours under uncertainty). First, we use climate scenarios from two models of the PRUDENCE project (HadRM3H and ARPEGE) to search for cities whose present climates can be considered as reasonable analogues of the future climates of 17 European cities. These analogues meet rather strict criteria in terms of monthly mean temperature, total annual precipitations and monthly mean precipitations. Second, we use these analogues as a heuristic tool to understand the main features of the adaptation required by climate change. The availability of two analogues for each city provides a useful estimate of the impact of uncertainty on the required adaptation efforts. Third, we carry out a cost assessment for various adaptation strategies, taking into account the cost of possible ill-adaptations due to wrong anticipations in a context of large uncertainty (from sunk-costs to lock-in in suboptimal adaptation choices). We demonstrate the gap between an enumerative approach under perfect expectation and a calculation accounting for uncertainty and spillover effects on economic growth.