Comparación de metodologías para la imputación de la lluvia diaria en una pequeña cuenca de latitudes medias
Carlos López Vázquez, Celina Gutiérrez, Hugo de los Santos
Facultad de Ingeniería, Centro de Cálculo
CC 30, Montevideo, Uruguay
e-mail: carlos.lopez@ieee.org
Abstract:
This work attempts to compare different methodologies for the missing value problem of daily rain data. A Monte Carlo simulation was designed, randomly choosing both date and place for the missing values and afterwards different imputation procedures were successively applied. We build some statistics which characterize the distribution of the absolute error, namely its expected value, variance and 75, 85 and 95 percentile to compare the results.
Among others, we tested the inverse of distance Cressman’s method, Optimum Interpolation, Ordinary Least Squares and Artificial Neural Networks as a nonlinear example. The test region was the Santa Lucía river catchment area of 13000 km2, at 35°S near the Atlantic; its yearly accumulated precipitation values are around 1000 mm. The dataset has 20 years long and ten stations.
The present results show that it is possible to imputate with a mean error of 2 mm/day and an RMS of 7 mm/day using both linear and nonlinear procedures, while there is still room for improvement in the latter.
Publicado en:
IX Congresso Brasileiro de Meteorologia (Nov/96)
Para los aun interesados, aqui pueden obtener la (161KB)
carlos.lopez@ieee.org