Etude de l'ajustement quadratique uniforme des estimations de pluie par radar à l'aide des mesures pluviométriques

Previous studies have shown large discrepancies between radar and gage measurements. Both random and bias errors contribute to alter the radar-estimated rainfalls. This paper examines a new approach to combine radar and gage measurements. In the proposed method, radar calibration by gages is based upon a second-order correction applied uniformly to the whole observed domain. The quadratic expression is derived from a least-squares fit of radar estimates by gage observations at gage locations. The use of such a technique is dictated by the fact that bias can vary as a nonlinear function of the observed rainfall (differential bias). The effectiveness of the quadratic adjustment in improving the radar measurements is appreciated through comparative numerical and experimental studies with classical methods using linear representation: the uniform linear adjustment and the spatial adjustment as proposed by Brandes (1975). It is found, compared to the former, that the proposed method brings a substantial improvement of radar estimates by performing a more effective reduction of differential bias, while it gives similar results with respect to those provided by the latter. Experimental results show that the initial error of 43% in the radar estimates is reduced to 15% after uniform quadratic adjustment, against 34% after uniform linear adjustment. Its performances, in addition to its simple implementation, make the uniform quadratic adjustment very attractive for operational uses. © 1989.