A Study on the Adjustment of Global Navigation Satellite System Radio Occultation Observation Errors to Improve KIM Forecast Performance

Abstract

This study investigates observation error adjustment strategies for three recent Global Navigation Satellite System Radio Occultation (GNSS RO) datasets (GRACE-D, Sentinel-6 A, and SPIRE) within the Korean Integrated Model (KIM), aiming to mitigate forecast degradation of mid-tropospheric geopotential height. Experiments were conducted using a low-resolution KIM (July-August 2022), with verification against European Centre for Medium-Range Weather Forecasts (ECMWF) Integrated Forecasting System (IFS) analyses. The standard GNSS RO error model (20% observation error at the surface and 1% at 10 km) was used as the control (EXP_ctl). Two adjustment strategies were tested: EXP_zfrac, which increased the lowest-level error to 40%, thereby reducing the influence below 10 km; and EXP_obserr, which applied a uniform inflation factor (√2, from 2.2 to 3.1) across all levels to account for doubled observation counts. Relative to the Baseline, EXP_ctl degraded the global 500 hPa geopotential height analysis by 4.01%, while EXP_zfrac and EXP_obserr improved performance by 2.01% and 2.28%, respectively. For 5-day forecasts of Northern Hemisphere 500 hPa geopotential height, EXP_ctl degraded performance by 0.11%, while EXP_zfrac and EXP_obserr demonstrated improvements of 0.74% and 0.99%, respectively. These findings demonstrate that uniform observation error inflation, which reflects increased data density, is more effective for KIM than selective layer-specific error adjustments.