A Multivariate, Stochastic Approach to Generating Daily Precipitation Series at Ungauged Locations in Mountainous Region

Information on the variability of precipitation in time and space is critical for a large number of water resources projects. However, in most practical applications, precipitation records at the location of interest are often either limited or unavailable due to the lack of adequate network of rainfall measurements. To address this need, regionalization methods have been frequently employed to understand the spatial behavior of precipitation and to transfer precipitation information from one location to the other where records are scarce. However, most conventional regionalization approaches are based on precipitation amounts or extreme rainfalls but do not include information on spatial variation in precipitation occurrence (wet/dry-day). This study hence proposes a stochastic weather generator to estimation of missing daily precipitation series in mountainous regions based on a novel, multivariate regionalization method based on both precipitation amount and occurrence. The proposed approach includes two steps: (i) the combination of Principal Component Analysis (PCA) and Ordinal Factor Analysis (OFA) is implemented for identifying regions of homogeneous precipitation, and (ii) the implementation of a stochastic model for constructing daily precipitation events at ungauged locations. The method calibrated and validated using rain-gauge stations in Catskill Mountains region, from which 90% of the drinking water for the New York City Water Supply System (NYCWSS).