In this thesis a methodological framework for assessing non-catastrophic weather risk is presented and applied on the winter tourism industry in Austria by examining the impacts of snow conditions on tourism demand in 185 ski areas in the period 1972/1973 to 2006/2007. A three-step approach is proposed: (1) modelling the distribution of several weather indices, e. g. days with snow depth >1 cm, (2) estimating the dependency of overnight stays on snow conditions by means of an Autoregressive Distributed Lag (ADL) model, (3) measuring Value at Risk(weather), in short VaR(weather), corresponding to the maximum loss from adverse weather conditions which is not exceeded with a given probability level over a given period of time. Results emphasize the importance of considering both the probability of an event and its potential impact for estimating weather risks. Trend analyses provide evidence that the probability of seasons with adverse natural snow conditions substantially increases. At the same time, analyses show a predominantly positive dependence of overnight stays on snow conditions, but also suggest that impacts have decreased in recent years, probably owing to the major increase in snowmaking. Overall, estimates of the 95%-VaR(weather), corresponding to a 1 in 20 year event, range from a 1.7 % to 50.5 % loss in overnight stays in ski areas (median ski area: 7.2 %). This is equivalent to a loss in sales of up to 19 million Euro (median ski area: 500 000 Euro) and yields an aggregate 95%-VaR for the accommodation industry of 157 million Euro. Potential sources of biases in these estimates are e. g. uncertainties in the meteorological data and changes in the level of adaptation. Finally, a linking of weather risk estimates to financial ratios for hotels reveals that hotels in lower lying and smaller areas do not only face higher weather risk, but also tend to be less profitable and exhibit higher debt ratios.