The use of analog method for statistical downscaling of 500-mb level height to autumn precipitation in Slovenia
2000
Bergant, K. | Kajfez-Bogataj, L. (Ljubljana Univ. (Slovenia). Biotechnical Fac., Agronomy Dept.)
Analog method was used to estimate the October and November precipitation at 5 locations in Slovenia in time period 1958-1999. The 500-mb level height was chosen as a predictor. Predictor data were taken from area of size 20 degrees P 20 degrees with center at 45 degrees N and 15 degrees E. The resolution of predictor data was 2.5 degrees P 2.5 degrees. Empirical orthogonal function analysis was performed prior the application of analog method. Reasonable analogs for all predictor fields were found with 1 exception. The cross validation approach was used for testing the employability of analog method for statistical downscaling of 500-mb level height to October and November precipitation. The analog method performed well in estimating the real values of averages of October and November precipitation and their variability in treated time period. Significant, but not extremely high correlation between observed and estimated values was detected. Calculations were performed also for other months in year. The statistical models are beside for October and November also employable for estimating the monthly precipitation in winter months, but not successful in estimating the amount of precipitation in spring and summer months. In comparison to location Ljubljana, Ratece and Novo mesto, lower skill of analog method for estimation of monthly precipitation was found for location Maribor and slightly higher for location Slap. Different performance on different locations probably originates in the interaction of Mediterranean cyclones with agitated Slovenian relief, especially with Alpine-Dinaric barrier. Statistical models based on analog method can be used for rough estimation of precipitation in October and November and also for precipitation in winter months in Western and Central Slovenia. For better results a multiple predictor and more complicated downscaling schemes should be used.
显示更多 [+] 显示较少 [-]