Comparative verification of GDAS and NCUM analyses using In-situ Radiosonde observations

Abstract

This study quantitatively evaluated the performance of high-resolution global analyses generated by the National Centre for Medium Range Weather Forecasting (NCMRWF) Global Data Assimilation System (GDAS) and the Unified Model (NCUM). The growing density of observational networks has contributed to improvements in numerical model performance, but there is a need for rigorous verification of analysis and forecast products to validate and enhance the competency of Numerical Weather Prediction (NWP) systems. A comparative study between GDAS and NCUM analyses was commenced during the 2024 summer monsoon over 12 distinct regions globally based on Commission for Basic Systems (CBS) guidelines. Various verification skill scores were calculated daily, weekly and monthly timescales from two analyses against the radiosonde (RS/RW) observations.  Regional variations in model performance are observed, with higher errors in geopotential height and temperature noted for the South Pole (SP), Tropics (TR), Regional Specialized Meteorological Centre (RSMC), and India (IN) regions. Wind components exhibit more significant errors over the SP region. However, the GDAS exhibits lower errors in Pearson correlation and anomaly correlation coefficients. These findings facilitate the inter-comparison of model performance at operational centers worldwide, informing future improvements in NWP systems.