T HE purpose of this paper is to present a decision model which, while relatively simple in structure, has proven valuable in structuring the elements of a rather wide variety of decision situations involving man's individual or collective adjustment to natural events such as weather phenomena. It is felt that the model may be useful in the formulation and solution of still other types of problems. A case study of adjustment to snow hazard is used to illustrate the direct empirical application of the model to an urban decision problem. Time seems to divide itself conveniently into two spans for decision purposes: the oftmentioned short and long runs. While there may exist a spectrum of responses over time to a perceived stimulus, most economic decisions take place within systems which have design capacities for response outside of which response is more difficult or costly than when system design capacity can be adjusted. The economic theory of the firm with its many elaborations is a perfect illustration. The long run-short run distinction seems to hold well also for decision situations which are more accurately characterized as adjustment to natural conditions or environment. In adjusting to floods, the long run holds the possibilities of constructing dams, levees, floodproof buildings, and changing locations, while short run response is restricted to warnings, evacuation, movement of furnishings, etc. There exist other natural hazard situations which are nicely characterized by a similar dichotomy of responses: tornadoes, hurricanes, hail, frost, and snow. The present model was initially constructed to help understand decision processes relating to weather conditions and to assist in valuing changes in the accuracy of weather forecasts. Where weather is an important part of the environment of a production process (including such activities as recreation), there are usually long and short run forms of adjustment. There are also two corresponding types of information: climatological data collected over long time periods, and weather forecasts which reach from the present to perhaps five days in the future. While climatological parameters may be quite stable, particular weather events occur randomly and the forecasts of those events are subject to error. Weather information can thus be capsulated in the joint probability distribution of weather events and forecasts.' The model is constructed to minimize the expected value of the sum of the costs of adjustment and residual damage. In our opinion, therefore, it deals with repetitious, noncatastrophic natural events. The structure of this model is similar to the basic framework of Nelson and Winter (1964). The major differences between the two formulations are that the present model includes both short run and long run responses to the distribution of weather events by the weather information user, but it does not analyze the problem of the transmission of weather information from the weather forecaster's point of view. Nelson and Winter were able to characterize analytically the value of a forecast for cases involving a small number of discrete actions and forecast values. Perhaps more importantly, Received for publication May 28, 1974. Revision accepted for publication December 30, 1974. * This work was started under NSF Grant GA-31298, User Needs and Dissemination Requirements for Weather Information, made to the University of Colorado in 1970. The application to snow hazard was worked out as part of the Assessment of Research on Natural Hazards project under NSF Grant GI-32942 to the University of Colorado. 1 The marginal distribution of weather events would be constructed from the climatological record.