Validating SNAP climate models

Climate models as predictive tools Given the unreliability of even relatively shortterm weather forecasting, how useful are climate models that project as far as a century into the future? The answer to this question depends on the needs of the model user. In general, climate model outputs simulate both the predictable and the unpredictable nature of climate. They are intended to predict overall patterns of temperature, precipitation, winds, and ocean currents rather than specific weather events. Climate patterns can be considered as a combination of several concurrent patterns. These include mean annual values (e.g., average annual temperature or precipitation); intraannual variability (e.g., the variability between seasons within a year); interannual variability (e.g. the variation in mean annual values over the course of a decade or longer); and trend (any persistent increase or decrease in mean annual values). In general, statistical analysis is necessary to determine whether changes from year to year are the result of random variation, short-term effects such as La Nina, or a long-term trend. Cold years can occur in the midst of warming trends, and wet years can take place during drying trends. Even if the overall tendency is towards more severe storms, large forest fires, or extreme floods, many years may go by without these events occurring. Precipitation tends to be even more variable over space and time than temperature. A localized cloudburst can drench one valley but leave the neighboring valley dry. Snowfall for the month of January might be ten times greater one year as compared to the following year. In order to provide useful simulations of the real world, climate models must depict the same types of variability and the same degree of variability as the real world, as well as the same mean values. A good climate model can predict the frequency and magnitude of random fluctuations around mean values. However, it is impossible for any climate model to predict the exact timing or location of these fluctuations. Thus, climate model projections are useful for predicting changes in mean annual conditions, such as average summer high temperatures, average winter low temperatures, or average length of growing season. These projections can in turn be used to help inform fire management, agricultural systems, pest control, permafrost engineering, wildlife management, and many other planning tasks. Models can also be used to estimate the frequency and magnitude of events such as storms, flooding, heat waves, winter thaws, rain-on-snow events, or summer frosts. However, they cannot be used to predict the dates on which such events will occur. Therefore, climate models might help inform a longterm decision to enlarge a culvert or relocate a village, but planners would need to make such decisions before the need became pressing.