Estimating the parameters of rainfall models using maximum marginal likelihood.
Models based on stochastic point processes have been used widely to characterise the temporal evolution of rainfall at a single location. Such models have many uses in hydrology. However, making inferences about the parameters of these models is notoriously difficult. We propose a method of inference in which rainfall occurrence and rainfall amounts are separated. The temporal parameters of the model, that is, the parameters which govern the occurrence of rainfall, are estimated based on the marginal likelihood of the binary sequence of rainfall occurrence indicators. The remaining parameters are estimated using moment methods. The procedure is illustrated by fitting a rainfall model to data from south west England.
|Title:||Estimating the parameters of rainfall models using maximum marginal likelihood|
|UCL classification:||UCL > School of BEAMS
UCL > School of BEAMS > Faculty of Maths and Physical Sciences
Archive Staff Only