Aims : The quality of the Committee on Space Research (COSPAR) International Reference Atmosphere models NRLMSISE-00, JB2008, and DTM2013 in the 150–300 km altitude range has never been thoroughly evaluated due to a lack of good density data. This study aims at providing the model accuracies thanks to the recent high-resolution high-accuracy Gravity field and steady-state Ocean Circulation Explorer (GOCE) density dataset. The evaluation was performed on yearly, monthly, and daily time scales, which are important for different applications such as mission design, mission operation, or re-entry predictions. Methods : The accuracy of the models was evaluated by comparing to the GOCE density observations of the Science Mission (1 November 2009–20 October 2013) and new density data at the lowest altitudes derived for the last weeks before the re-entry (22 October–8 November 2013) according to a metric, which consists of computing mean, standard deviation and root mean square (RMS) of the observed-to-model ratios, and correlation. Mean statistics are then calculated over the three time scales. Results : The range of model biases, standard deviations, and correlations becomes larger when the time interval decreases, and this study provides COSPAR International Reference Atmosphere (CIRA) model statistics in the altitude range of 275–170 km. DTM2013 is the least biased and most accurate model on all time scales, essentially thanks to the database, notably containing two years of GOCE densities, to which it was fitted. NRLMSISE-00 performs worst, with considerable bias of about 20% in 2009 and 2013, and systematically higher standard deviations (lower correlations) than JB2008 and DTM2013. The performance of JB2008 is presently only slightly behind DTM2013, thanks to the new release 4_2g solar activity proxies. However, it still presents some weakness under the lowest solar activity conditions in 2009 and 2010. Comparison to Challenging Mini-Satellite Payload (CHAMP) density data showed that the results based on GOCE densities, despite limited local solar time coverage of 6–8 am & pm, are representative of model performance.
Read full abstract