Using a network of temperature lidars to identify temperature biases in the upper stratosphere in ECMWF reanalyses

Abstract

Abstract. To advance our understanding of the stratosphere, high-quality observational datasets of the stratosphere are needed. It is commonplace that reanalysis datasets are used to conduct stratospheric studies. However, the accuracy of these reanalyses at these heights is hard to infer due to a lack of in situ measurements. Satellite measurements provide one source of temperature information. As some satellite information is already assimilated into reanalyses, the direct comparison of satellite temperatures to the reanalysis is not truly independent. Stratospheric lidars use Rayleigh scattering to measure density in the middle and upper atmosphere, allowing temperature profiles to be derived for altitudes from 30 km (where Mie scattering due to stratospheric aerosols becomes negligible) to 80–90 km (where the signal-to-noise ratio begins to drop rapidly). The Network for the Detection of Atmospheric Composition Change (NDACC) contains several lidars at different latitudes that have measured atmospheric temperatures since the 1970s, resulting in a long-running upper-stratospheric temperature dataset. These temperature datasets are useful for validating reanalysis datasets in the stratosphere, as they are not assimilated into reanalyses. Here, stratospheric temperature data from lidars in the Northern Hemisphere between 1990–2017 were compared with the European Centre for Medium-Range Weather Forecasts ERA-Interim and ERA5 reanalyses. To give confidence to any bias found, temperature data from NASA's EOS Microwave Limb Sounder were also compared to ERA-Interim and ERA5 at points over the lidar sites. In ERA-Interim a cold bias of −3 to −4 K between 10 and 1 hPa was found when compared to both measurement systems. Comparisons with ERA5 found a small bias of magnitude 1 K which varies between cold and warm bias with height between 10 and 1 hPa, indicating a good thermal representation of the middle atmosphere up to 1 hPa. A further comparison was undertaken looking at the temperature bias by year to see the effects of the assimilation of the Advanced Microwave Sounding Unit-A (AMSU-A) satellite data and the Constellation Observing System for Meteorology, Ionosphere, and Climate GPS Radio Occultation (COSMIC GPSRO) data on stratospheric temperatures within the aforementioned ERA analyses. It was found that ERA5 was sensitive to the introduction of COSMIC GPSRO in 2007 with the reduction of the cold bias above 1 hPa. In addition to this, the introduction of AMSU-A data caused variations in the temperature bias between 1–10 hPa between 1997–2008.