HR: 0800h
AN: H11F-0361    [Abstracts]
TI: Space-Time Daily Rainfall Generation Model in the Subtropical Region
AU: * Kim, T
EM: Taewwong_Kim@partner.nps.gov
AF: South Florida Ecosystem Office - ENP/NPS, 950 N. Krome Ave., Homestead, FL 33030 United States
AU: Ahn, H
EM: Hosung_Ahn@nps.gov
AF: South Florida Ecosystem Office - ENP/NPS, 950 N. Krome Ave., Homestead, FL 33030 United States
AB: Daily rainfall is an essential input to hydrologic models that are invaluable tool for evaluating various Everglades restoration projects. In many cases, we used only a limited length of historical rainfalls (<30 years) which is not enough to define extreme hydrologic events such as 1-in-10-year droughts. Thus, it is desirable to use synthesized long-term rainfalls to quantify an uncertainty in hydrologic simulations. The synthetic rainfalls data can be generated by a simple stochastic process for the sake of an intensive and expensive numerical climate models. In general, the previous daily rainfall models consist with two components: occurrence and amounts processes. The occurrence process has been formulated typically by a lag-one, two-state Markov process. Then, rainfall amounts are generated by a separate process. However, this approach could not incorporate the spatial dependence of both occurrence and amounts. Daily rainfalls in south Florida show that the space-time correlation is somewhat significant (rs ,d 0.5). Thus, it is reasonable to introduce the dependence structure in a model. We use a concept of rain storm clustering process to describe adequately the space-time correlation structure that is appeared on the subtropical rainfalls which are driven primarily by convective activities. In our proposed model, a convective storm cluster is randomly identified from the observed probabilities of occurrence on a day. The storm cluster contains gauges where rainfall activity occurs. The rainfall amounts are generated based on the acyclic spatial dependence on wet days. This is a way of considering the spatial intermittence of daily rainfalls. In specific, the temporal dependence of rainfall occurrence is accounted by the Markov process of the rainfalls that are not included in a storm cluster. This study tests the proposed model with historical rainfalls measured from 78 stations in south Florida. Various sample statistics are used to validate the model.
DE: 3354 Precipitation (1854)
DE: 1833 Hydroclimatology
DE: 1854 Precipitation (3354)
DE: 1869 Stochastic processes
SC: Hydrology [H]
MN: 2004 AGU Fall Meeting