The characterisation of errors and uncertainties in numerical weather prediction (NWP) model fields is a major challenge that is addressed as part of the Horizon 2020 Gap Analysis for Integrated Atmospheric ECV CLImate Monitoring (GAIA-CLIM) project. In that regard, observations from the GCOS (Global Climate Observing System) Reference Upper-Air Network (GRUAN) radiosondes are being used at the Met Office and European Centre for Medium-Range Weather Forecasts (ECMWF) to assess errors and uncertainties associated with model data.
The software introduced in this study and referred to as the GRUAN processor has been developed to collocate GRUAN radiosonde profiles and NWP model fields, simulate top-of-atmosphere brightness temperature at frequencies used by space-borne instruments, and propagate GRUAN uncertainties in that simulation. A mathematical framework used to estimate and assess the uncertainty budget of the comparison of simulated brightness temperature is also proposed.
A total of 1 year of GRUAN radiosondes and matching NWP fields from the Met Office and ECMWF have been processed and analysed for the purposes of demonstration of capability. We present preliminary results confirming the presence of known biases in the temperature and humidity profiles of both NWP centres. The night-time difference between GRUAN and Met Office (ECMWF) simulated brightness temperature at microwave frequencies predominantly sensitive to temperature is on average smaller than 0.1 K (0.4 K). Similarly, this difference is on average smaller than 0.5 K (0.4 K) at microwave frequencies predominantly sensitive to humidity.
The uncertainty estimated for the Met Office–GRUAN difference ranges from 0.08 to 0.13 K for temperature-sensitive frequencies and from 1.6 to 2.5 K for humidity-sensitive frequencies. From the analysed sampling, 90 % of the comparisons are found to be in statistical agreement.
This initial study has the potential to be extended to a larger collection of GRUAN profiles, covering multiple sites and years, with the aim of providing a robust estimation of both errors and uncertainties of NWP model fields in radiance space for a selection of key microwave and infrared frequencies.