1. Introduction
Quality and biases of radiosounding observations strongly vary with sensor type, altitude level, and over time. Many previous studies described the adjustment of historical radiosonde temperature measurements to construct Climate Data Records (CDRs) by applying data homogenization techniques. However, none of these CDRs is fully homogeneous or provides estimation of the measurement uncertainties.
To meet the need for homogeneous and fully traceable upper-air measurements with quantified uncertainties, the GCOS Reference Upper-Air Network (GRUAN) was established to provide reference-quality profile measurements of Essential Climate Variables in the upper atmosphere. GRUAN is providing long-term, high-quality radiosounding data at several sites around the world. GRUAN data processing starts from the raw data and applies corrections, ensuring traceability, for all known measurement errors (e.g. due to solar radiation, sensor time-lag, etc.), each with a quantified uncertainty contributing to the final uncertainty budget. Recent progress made by GRUAN includes new reference products for the most recent radiosonde types (RS41, iMS-100, RS-11G).
As a reference network, GRUAN also plays the role to provide observations and methods to allow an enhanced interpretation of the results and the quantification of uncertainties for the baseline observations. Fully exploiting the latter concept, in the frame of the Copernicus Climate Change Service (C3S), a novel approach, named RHARM (Radiosounding HARMonization), has been developed to provide a homogenized dataset of temperature, humidity, and wind profiles along with an estimation of the related uncertainties for a substantive subset of radiosounding stations globally distributed (Madonna et al. 2022) among those available from the Integrated Global Radiosonde Archive (IGRA).
In this paper, a short overview of the RHARM approach is provided.
2. Bias Adjustment of upper-air profiles
RHARM is applied to daily (00:00 and 12:00 UTC) radiosonde data on the 16 standard pressure levels for the IGRA data from 1978 onward. The applied adjustments are interpolated to all the other significant levels. The RHARM method is a hybrid method combining an algorithm mimicking GRUAN for the adjustment of the IGRA radiosondes, applicable to data since 2004 (for all radiosondes included in the GRUAN data products, and for all the radiosonde involved in the 2010 WMO radiosonde intercomparison), and a statistical method for the homogenization of the historical data holdings, before 2004. RHARM is the first data set to provide homogenized time series not affected by cross-contamination of biases across stations and provided with an estimation of the observational uncertainty.
This work focuses on showing the quality of the radiosonde profiles adjusted using the information provided by GRUAN and available from radiosonde intercomparisons.
3. Results
In the top panel of Figure 1, the RHARM-GRUAN and IGRA-GRUAN mean differences are shown. The comparison refers to 00 UTC and 12 UTC ascents from about January 2010 (2008 for two stations, Lindenberg and Ny-Alesund) to July 2023, and refers to mandatory levels from 850 hPa to 10 hPa. Being the GRUAN data at high resolution, with a vertical step of 5-10 m, the IGRA and RHARM data are matched to the GRUAN data by applying a threshold of 40 Pa for levels between 850 hPa and 300 hPa and of 5 hPa below 300 hPa pressure. Then, a linear interpolation is applied between the minimum and the maximum pressures selected according to these criteria.
The comparison shows the capability of RHARM to adjust most of the difference between the IGRA and GRUAN profiles, with a residual difference smaller than ±0.05 K over the entire vertical range. The corresponding value of the IQR (not shown) is smaller than 0.5 K up 200 hPa and smaller than 1.0 K above.
Figure 1: to panel, mean difference of the RHARM and IGRA datasets with respect to GRUAN. Only mandatory levels from 850 hPa to 10 hPa are considered and the period is 2010-2023; bottom panel, Comparison of monthly time series of the water vapor mixing ratio zonal average from IGRA, RHARM, and MLS AURA in the northern tropics (0°N-25°N), in the period 2006-2019.
In the bottom panel of Figure 1, a comparison of the monthly time series of the water vapor mixing ratio zonal average obtained from RHARM in the northern tropics (0°N-25°N) from 2006 to 2019 is compared to the corresponding time series retrieved from the AURA MLS data version 4 (Yan et al., 2016). AURA data have been sub-sampled at the IGRA stations within the considered domain.
The comparison shows the enhanced agreement between RHARM and IGRA due to the adjustment of the dry-bias affecting several radiosonde types in the upper-troposphere and generating smaller values of the water vapor content. The agreement is also improved for the highest values of the water vapor mixing ratio.
4. Conclusions
Ideally, as GRUAN does, it would be desirable to directly reprocess the raw data which are often not preserved at the measurement stations. An alternative could be represented by the adjustment of all upper-air data using the data from available intercomparisons and disclosing corrections used by manufacturers for each sonde type used in the past. At present, it is undoubted that this process is not viable because the lack of metadata does not allow to rebuild the history of each station. The RHARM algorithm is a novel option to improve the adjustment of global upper-air data. This dataset aims to show the importance of the availability of reference data from GRUAN and from WMO/CIMO intercomparison data to quantify the uncertainties in the characterization of present and historical radiosounding datasets.
5. References
Madonna and Co-authors (2022). The new Radiosounding HARMonization (RHARM) data set of homogenized radiosounding temperature, humidity, and wind profiles with uncertainties. Journal of Geophysical Research: Atmospheres, 127, e2021JD035220. https://doi.org/10.1029/2021JD035220
Yan, X and Co-authors: Validation of Aura MLS retrievals of temperature, water vapour and ozone in the upper troposphere and lower–middle stratosphere over the Tibetan Plateau during boreal summer, Atmos. Meas. Tech., 9, 3547–3566, https://doi.org/10.5194/amt-9-3547-2016, 2016.

