Optical network operators face the challenge of upgrading the WDM network capacity to adapt to estimated traffic growths. Network upgrades are commonly carried out in scheduled intervals (i.e. every six months), using traffic forecasts. The uncertainty in the forecasts is a major issue in the capacity upgrading process. If it is not handled appropriately, the network is exposed to service degradation caused by an unexpected traffic progression. Despite of its relevance, the effects of uncertainty in the forecasts is a factor that has not been well studied in the literature. In this paper, we apply the robust optimization paradigm to incorporate this uncertainty into the network upgrade problem. Under this robust network upgrade model, we can dimension the network by tuning the tradeoff between network cost and robustness level. This proposal is applied to a case study where several experiments are conducted comparing different levels of robustness and different WDM technologies, namely pure 10G (single line rate), pure 40G SLR, pure 100G SLR and 10/40/100G MLR (mixed line rate).