| dc.description.abstract | In this study simulated microwave-link information is used to compute maps of areal precipitation (rainfall fields) for the Netherlands. Simulated information from 2032 microwave links active in January 17th - 2011, are constructed from radar data. Corrected radar data from the KNMI weather radar are used to validate the estimated results. Interpolations are carried out by the geostatistical technique Ordinary Kriging. For this technique, applied semivariograms are obtained on: 1) spherical models fitted to the experimental semivariograms computed for the corresponding data sets; 2) seasonal variogram parameterizations proposed and developed by van de Beek et al. (2011a,b). The estimated rainfall fields show good accuracy at regional and local scales in areas where microwave link density is high. Local cases are presented for the cities of Utrecht and Rotterdam (the Netherlands). Semivariogram temporal downscaling is carried out for time-aggregation scales of 15-minute, 1-, 3-, 6- and 12-hour. For these scales, estimated rainfall fields are obtained and filtered to account for the percentage of microwave links not registering rainfall attenuation in their signals. The Mean Error (ME), Root Mean Squared Error (RMSE) and Variance Ratio (VR) are computed in order to investigate the bias, accuracy and variability of rainfall field estimations for different time scales. The highest correlation is found for the 6-h aggregated time scale. Overall, good results are obtained for the other time scales, suggesting that the developed methodology is suitable for automatic rainfall estimation at small aggregated time scales. | |
