A New Method to Correct Radiosonde Temperature Biases Using Radio Occultation Data

Abstract

A new method to estimate radiosonde temperature biases using radio occultation measurements as a reference has been developed. The bias is estimated as the difference between mean radio occultation and mean radiosonde departures from collocated profiles extracted from the Met Office global numerical weather prediction (NWP) system. Using NWP background profiles reduces the impact of spatial and temporal collocation errors. The use of NWP output also permits determination of the lowest level at which the atmosphere is sufficiently dry to analyze radio occultation dry temperature retrievals. The authors demonstrate the advantages of using a new tangent linear version of the dry temperature retrieval algorithm to propagate bending angle departures to dry temperature departures. This reduces the influence of a priori assumptions compared to a nonlinear retrieval. Radiosonde temperature biases, which depend on altitude and the solar elevation angle, are presented for five carefully chosen upper-air sites and show strong intersite differences, with biases exceeding 2 K at one of the sites. If implemented in NWP models to correct radiosonde temperature biases prior to assimilation, this method could aid the need for consistent anchor measurements in the assimilation system. The method presented here is therefore relevant to NWP centers, and the results will be of interest to the radiosonde community by providing site-specific temperature bias profiles. The new tangent linear version of the linear Abel transform and the hydrostatic integration are described in the interests of the radio occultation community.

Reducing representativeness and sampling errors in radio occultation–radiosonde comparisons

“…Wong et al (2015) used ECMWF forecasts for double-differencing to reduce the sampling differences between Atmospheric Infrared Sounder (AIRS) and RS co-located pairs. Tradowsky et al (2017) calculated an observed-background (O-B) double difference to estimate the mean RS temperature bias using co-located RO profiles and Met Office model background fields. These studies use a double-difference correction, but do not verify or discuss how the correction reduces sampling errors.…”

“…However, as the equivalent Eq. (6a) shows, the corrected difference is also the difference of the departures of the two data sets from a common, reference data set, or the "double-differencing" method (Chander et al, 2013;Tradowsky et al, 2017). In this interpretation, the two data sets are not necessarily required to be close in space or time; the method is valid as long as the biases of the reference data set do not vary over the spatial and temporal scales of the comparison.…”

“…Therefore, we hypothesize that the spatial variability of refractivity within ellipses of semi-major axis a oriented along the wind direction will be reduced compared to the variability within circles of radius a. The second method consists of applying a spatial-temporal sampling correction to the RO-RS co-location pairs using the "doubledifference" method (Chander et al, 2013;Tradowsky et al, 2017). In the double-difference method, each data set is compared to an intermediate reference data set; in our study we use the European Centre for Medium-Range Weather Forecasts (ECMWF) Interim Reanalysis (ERA-Interim) model data (Dee et al, 2011, and https://www.ecmwf.int/ en/forecasts/datasets/reanalysis-datasets/era-interim, last access: 23 March 2018).…”

Reducing representativeness and sampling errors in radio occultation–radiosonde comparisons

“…Wong et al (2015) used ECMWF forecasts to estimate the sampling differences between Atmospheric Infrared Sounder (AIRS) and RS co-located pairs, using a double-difference error estimate to correct AIRS and RS differences. Tradowsky et al (2017) Office NWP background fields, in which the double-difference is applied to reduce the effects of temporal and spatial sampling errors. These studies use a double-difference sampling correction, but do not verify or discuss how the correction reduces sampling errors.…”

“…By subtracting the model data from both the RO and RS the background and representativeness errors are removed, leaving only the RO and RS observational errors. The approach has been called the "double-difference" method (Haimberger et al, 2012;Wong et al, 2015;Tradowsky et al, 2017). In addition 2 Atmos.…”