Interpreting estimated Observation Error Statistics of Weather Radar Measurements using the ICON-LAM-KENDA System

Abstract

Abstract. Assimilation of weather radar measurements including radar reflectivity and radial wind data has been operational at the Deutscher Wetterdienst, with a diagonal observation error (OE) covariance matrix. For an implementation of a full OE covariance matrix, the statistics of the OE have to be a priori estimated, for which the Desroziers method has been often used. However, the resulted statistics consists of contributions from different error sources and are difficult to interpret. In this work, we use an approach that is based on samples for truncation error in radar observation space to approximate the representation error due to unresolved scales and processes (RE) and compare its statistics with the OE statistics estimated by the Desroziers method. It is found that the statistics of the RE help the understanding of several important features in the variances and correlation length scales of the OE for both reflectivity and radial wind data and the other error sources from the microphysical scheme, radar observation operator and the superobbing technique may also contribute, for instance, to differences among different elevations and observation types. The statistics presented here can serve as a guideline for selecting which observations to assimilate and for assignment of the OE covariance matrix that can be diagonal or full and correlated.

Including observation error correlation for ensemble radar radial wind assimilation and its impact on heavy rainfall prediction

“…As Zeng et al . (2021) pointed out, the error correlation scale increases with height. Nevertheless, the choice of $$$ {L}_r=25 $$$ km in our study represents the mean behavior of the observation error statistics during the assimilation period.…”

“…With ICON‐LAM (ICOsahedral Nonhydrostatic‐Limited Area Model) at DWD, Zeng et al . (2021) further showed that for the same elevation, the correlation length scales increased with height, and the lower elevations generally exhibited longer correlation length scales for the same height. With a variational data analysis system, Simonin et al .…”

“…A similar estimated error correlation length scale for both sets of superobservations is obtained, possibly because observation error is largely contributed by representative error as suggested by Zeng et al . (2021), which is also related to the model setting and weather conditions. Additionally, the difference between the spacing of two sets of superobservations is not as large as one order.…”

“…The error structure of radar radial wind is estimated by the Desroziers diagnostic approach, which has been commonly adopted in previous studies (e.g., Waller et al ., 2016b; Waller et al ., 2019; Zeng et al ., 2021). The estimation is performed based on the innovation statistics of the assimilation experiments from 1,600 UTC to 1800 UTC on June 1, 2017.…”

“…However, over‐resampling sacrifices the advantages of high‐resolution data since small‐scale features could be related to rapid intensification. Recently, it has been recognized that numerous types of observations, especially those with dense spatial distributions, such as radar and satellite data, have significant error correlations, even after resampling (Keeler and Ellis, 2000; Waller et al ., 2016b; Waller et al ., 2019; Zeng et al ., 2021). Generally, observation errors can arise from instrument errors and representative errors (Janjić et al ., 2018), and the latter may vary spatially and temporally (Waller et al ., 2014).…”