382454 - STOCHASTIC SIMULATION OF DAILY PRECIPITATION AT MULTIPLE SITES: II PERFORMANCE EVALUATION OF THE MODEL
Thursday, June 7
8:30 AM - 10:00 AM
Location: Lake Superior B
Umed Panu, Thunder Bay, ON, Canada – Lakehead University
Weather generation models based on multivariate censored distribution (WG-MCD) and multivariate autoregressive censored process (WG-MACP) have been developed and presented in the paper entitled “Stochastic Simulation of Daily Precipitation at Multiple Sites: I Model Development”. In this paper, the performance of these models is evaluated by comparing discrepancies in attributes (e.g., mean, variance, correlation) obtained from the historical and simulated precipitation. Precipitation records from years 1961-1990 at ten climatic stations located in Manitoba, Canada are adopted for the performance evaluation of models. Three performance measures (i.e., the coefficient of determination, the coefficient efficiency, and the root mean square error) identify a fairly strong relationship between the historical and simulated precipitation.
Proposed models has been found to be suitable for reproducing the statistical characteristics of daily historical precipitations at multiple sites. The spatial and temporal dependencies appear to have been reasonably captured by the covariance and lag-1 covariance of the WG-MACP model. Other descriptors, such as probabilities of wet/dry-day, mean values, and variances, show good statistical agreement with similar statistical characteristics of the historical and simulated data sets. It was found that a better performance of the model could be obtained with use of a smaller number of stations, and with less number of statistical attributes to be preserved.