Abstract:
|
Measurements of wind speed, precipitation, and other climate-related data often show spatial, temporal, or spatio-temporal clusters or repeated behaviors, some of which may be evident only in specific representations of the data. To obtain realistic simulations and forecasts, models for these processes should reflect similar patterns of behavior. We demonstrate that some such clustering behavior can be recreated simply from the local dependence structure in a model. Next, we discuss modifications to standard models that ensure the preservation of clustering behavior, either by matching the overall level of "clusteredness" in the data or by recreating specific cluster patterns found in the original data.
|