Abstract
Over recent years, the forecast sensitivity observation impact tool has been used to quantify the impact of the assimilated observations on 24 h short‐range forecasts. The verification field used to compute the forecast error is the analysis and the forecast error objective function is the total energy norm. However, the analysis includes contributions not only from observations but also from the model employed; it can thus be influenced by biases in the model. In this article, an observation‐only based verification is proposed using a large variety of both space‐ and surface‐based meteorological observations, opportunely normalized by their variances. A comprehensive comparison between the standard forecast sensitivity observation impact based on an analysis‐energy and the observation‐only forecast error objective function is presented. The comparison shows a general agreement on the observation impact ranking derived from both measures with some differences due to the different normalization used in the two objective functions and due to the effect of model biases on the analysis‐energy objective function. It is also found that the number of assimilated observations that reduce the forecast error is on average 1% higher when the observation‐based forecast verification is used.

This publication has 11 references indexed in Scilit: