| A new multi-dimensional stochastic precipitation model is proposed with major emphasis on its spectral structure. As a hyperbolic type of stochastic partial differential equation, this model is characterized by having a small set of parameters, which could be easily estimated. These characteristics are similar to those of the noise forced diffusive precipitation model, but representation of the physics and statistical features of the precipitation field is better as in the WGR precipitation model. The model derivation was based on the AR (Auto Regressive) process considering advection and diffusion, the dominant statistical and physical characteristics of the precipitation field propagation. The model spectrum showed a good match for the GATE spectrum developed by Nakamoto et al. (1990). This model was also compared with the WGR model and the noise forced diffusive precipitation model analytically and through applications such as the sampling error estimation from space-borne sensors and raingages, and the ground-truth problem. The sampling error from space-borne sensors based on the proposed model was similar to that of the noise forced diffusive precipitation model but much smaller than that of the WGR model. Similar result was also obtained in the estimation of the sampling error from raingages. The dimensionless root mean square error of the proposed model in the ground-truth problem was in between those of the WGR model and the noise forced diffusive precipitation model, even though the difference was very small. Simulation study of the realistic precipitation field showed the effect of the variance of the noise forcing term on the life time of a storm event. |
