Scaling of GNSS Radio Occultation Impact with Observation Number Using an Ensemble of Data Assimilations

Abstract

Abstract An ensemble of data assimilations (EDA) approach is used to estimate how the impact of Global Navigation Satellite System (GNSS) radio occultation (RO) measurements scales as a function of observation number in the ECMWF numerical weather prediction system. The EDA provides an estimate of the theoretical analysis and short-range forecast error statistics, based on the ensemble “spread,” which is the standard deviation of the ensemble members about the ensemble mean. This study is based on computing how the ensemble spread of various parameters changes as a function of the number of simulated GNSS RO observations. The impact from 2000 up to 128 000 globally distributed simulated GNSS RO profiles per day is investigated. It is shown that 2000 simulated GNSS RO measurements have an impact similar to real measurements in the EDA and that the EDA-based impact of real data can be related to the impact in observing system experiments. The dependence of the ensemble statistics on observation error statistics assumed when assimilating the data, rather than the actual observation errors, is emphasized. There is no evidence of “saturation” of forecast impact even with 128 000 GNSS RO profiles per day. However, this result is a well-known consequence of always improving the theoretical analysis and short-range forecast error statistics when adding new observations that are assumed to have uncorrelated observation errors. In general, it is found that 16 000 GNSS RO profiles per day have around half the impact of 128 000 profiles, based on the reduction of ensemble spread values where the GNSS RO measurements have the largest impact.