Comparison of two approaches for downscaling synoptic atmospheric patterns to multisite precipitation occurrence

摘要: [1] The physical linkages between climate on the large scale and weather local allow formulation of downscaling approaches for assessing impact variability at point locations. This paper presents a comparison two such applied synoptic atmospheric patterns to rainfall occurrences rain gauge network. The evaluated are parametric nonhomogenous hidden Markov model (NHMM) nonparametric k-nearest neighbor approach. NHMM defines local-scale as function discrete state that is Markovian depends predictor variables representing patterns. As defined parametrically, number parameters need specification increases one considers more states. Consequently, parameter identification generalization ungauged sites becomes difficult. On other hand, resampling attractive because its efficiency simplicity, being structured direct probabilistic relationship larger-scale climatic weather. Such offers simpler alternative approach using intermediate but less capable persistence introduced through assumptions in NHMM. In presented here, we weather-state-based nonhomogeneous bootstrap estimate precipitation network 30 locations around Sydney, Australia. Our results suggest both models perform well spatial variations while they show lack temporal dependence scales longer than few days exhibited wet spell length characteristics. Local-scale features difficult represent large-scale predictors are, expected, not reproduced by either