Novel WkNN-based technique to improve instantaneous rainfall estimation over the north of Algeria using the multispectral MSG SEVIRI imagery

For the estimation of rainfall in northern Algeria, a new method is proposed in this study. It based on the k nearest weighted neighbours (WkNN) classification algorithm using the dataset from the radiometer-imager SEVIRI (Spinning Enhanced Visible and Infra-Red Imager) boarded on the satellite MSG (Meteosat Second Generation) to determine the pixel rainfall rate among the 16 predefined intensity levels observed in the Setif meteorological radar. The implementation of the classification consists in using the spectral characteristics of a new sample (pixel) as input variables of the WkNN classifier to predict his class of membership based on the weighted distances separating him from the samples of the set of learning. The results thus obtained are validated with respect to the rainfall intensity classes observed and co-located by ground radar. The results showed a significant improvement for instantaneous estimate of precipitation during the day, with a correlation coefficient r = 0.87 and statistical parameters RMSE = 2.29 mm/h, Bias = 0.45 mm/h, MAE = 0.35 mm/h, and we have obtained the results similar overnight.