HR: 11:35h
AN: H22A-06    [Abstracts]
TI: A nonparametric weather-state approach for downscaling of multi-site precipitation occurrences
AU: Mehrotra, R
EM: raj@civeng.unsw.edu.au
AF: School of Civil and Environmental Engineering, The University of New South Wales, Sydney, NSW 2052 Australia
AU: * Sharma, A
EM: a.sharma@unsw.edu.au
AF: School of Civil and Environmental Engineering, The University of New South Wales, Sydney, NSW 2052 Australia
AB: The physical linkages between climate on the large scale and weather on the local scale form the basis of downscaling approaches for assessing the impact of climate variability at point locations. The common approach frequently used for downscaling of precipitation, considers discrete weather classes of the atmospheric patterns and simulates precipitation conditioned on these weather-states. This paper presents the development of a nonparametric weather-state downscaling approach (KNN-W) and its comparison with traditional KNN resampling approach (KNN) and a parametric Non-homogenous Hidden Markov Model (NHMM). The KNN-W defines local scale weather as a function of a weather state that is continuous and auto-regressive in nature and depends on predictor variables representing synoptic atmospheric patterns. Such a formulation offers a simpler alternative to the weather-state based parametric approaches like NHMM. The KNN resampling approach considers a direct probabilistic relationship between the larger scale climatic variables and the local scale weather. On the other hand, the weather-state KNN downscaling approach being structured on continuous weather-state formulation is more opt at representing temporal persistence. A weather-state of KNN-W is defined based on spatial rainfall distribution over the study region. The paper also considers the relative influence of atmospheric circulation variables on the conditional density formulation in the form of influence weights. In the comparison presented here, we applied these downscaling approaches conditional on four atmospheric circulation variables, to estimate precipitation occurrences at a network of 30 raingauge locations around Sydney, Australia. Our results suggest that all the models perform well at representing spatial variations while they lack at representing temporal dependence at scales longer than a few days as exhibited through wet spell length characteristics. The weather-state based KNN approach is more successful in capturing the longer duration as well as extreme rainfall characteristics in comparison to direct KNN approach and NHMM. Local scale features that are difficult to represent through the large scale climate predictors are expectedly not reproduced by any approach.
DE: 1833 Hydroclimatology
DE: 1854 Precipitation (3354)
DE: 1869 Stochastic processes
SC: Hydrology [H]
MN: 2004 AGU Fall Meeting