EPS-Aeolus-ASpot

Contribution to the NOAA OSSE to estimate EPS-Aeolus impact on NWP

 

EPS-Aeolus-ASpot
EPS-Aeolus-ASpot

This study aims to aid in assessing the individual effects of current Aeolus observations and potential future EPS-Aeolus-like observations on NWP.

Last Updated

10 June 2024

Published on

19 April 2024

About

The purpose of this study is to help estimate the individual impacts of current Aeolus observations and future EPS-Aeolus-like observations on NWP. This goal has been achieved by incorporating NOAA OSSE simulations of both Aeolus and EPS-Aeolus, which have been generated by KNMI using the Lidar In-space Performance Analysis System (LIPAS; Marseille et al., 2003, 2011) tool.

aeolus-rayleigh-clear-HLOS-wind
Figure 1 - Courtesy KNMI (from mid-term review presentation): LIPAS simulated EPS-Aeolus HLOS wind along 1 orbit for Rayleigh-clear clear (Top) and Mie-cloudy (Bottom)

Objectives

The main tasks considered during this study are as follows:

  • Configuring the Aeolus (-2) simulation through LIPAS.
  • Providing advice on, discussing, and testing data assimilation aspects.
  • Supporting OSSE calibration.
  • Assisting in OSSE analysis and drawing conclusions.

Overview

Aeolus was the first and remains the only satellite capable of performing global 3D wind profile observations. It has demonstrated significant positive impact on global NWP models, as evidenced by ECMWF, Météo-France, Met Office, DWD, NOAA, JMA, NCMRWF and ECCC, surpassing initial expectations prior to its launch. Building on these successful outcomes, the phase 0 of preparation activities for a potential follow-on mission, EPS-Aeolus, commenced at ESA and EUMETSAT in 2020. Consolidated requirements have been compiled in a EURD document, and assessing their potential impact on NWP has emerged as a crucial focus. Typically, such assessment is conducted by OSSE exercises, which are complex due to the need to simulate all contributing components accurately, including realistic error variances and possible error correlations in both space and time.

Already deeply engaged in the preparation of the Aeolus mission, KNMI developed the Lidar In-space Performance Analysis System (LIPAS; Marseille et al., 2003, 2011) tool, capable of simulating Aeolus observations. The initial task of this study was to adapt LIPAS to simulate EPS-Aeolus lidar measurements, aligning with the new requirements outlined in the DWL EURD.

Table 1 presents a summary of the statistics regarding LIPAS-simulated Aeolus and EPS-Aeolus HLOS winds. The rejection rate is determined based on the estimated random error standard deviation, with threshold values set at 8 m/s for Rayleigh winds and 5 m/s for Mie winds. The table highlights two significant findings (highlighted in bold): a substantial reduction in overall random error, decreasing from 5.74 m/s for Aeolus to 1.66 m/s for EPS-Aeolus (at comparable resolution), and a notable increased in the number of Mie-cloudy winds (factor of ~3).

 Rayleigh-clearMie-cloudy
 Nstd (m/s)rej (%)Nstd (m/s)rej (%)
Aeolus36,6745.741.62,1362.5751.0
EPS-Aeolus36,0731.660.16,5461.704.7

Table 1: Courtesy KNMI (From final report): Summary of statistics of LIPAS simulated Aeolus and EPS-Aeolus HLOS winds based on 6-hours (4 orbits) of data. N is the number of data, ‘std’ the random error standard deviation and ‘rej’ the rejection rate.

NOAA performed the OSSE calibration for the Aeolus experiment. Subsequently, they conducted an additional OSSE with EPS-Aeolus, during which observation errors were tuned to their system. The experiment highlighted the significantly enhanced impact of EPS-Aeolus compared to Aeolus. In particular, the root mean square difference (RMSD) reduction was approximately doubled with respect to the baseline, as depicted in Figure 2 this reduction amounted to about 8% for EPS-Aeolus and only 4% for Aeolus at the initial time.

EPS-Aeolus
Figure 2. Courtesy NOAA (from final report): EPS-Aeolus (green), Aeolus (red) and baseline (without Aeolus) RMSD with respect to the nature run truth of the 200 hPa southern hemisphere extratropics wind vector field at OOUTC for a period from mid-June to end-July. The vertical bars denote the mean spread. The bottom plot shows the RMSD reduction with respect to the baseline, which is 8% for EPS-Aeolus and 4% for Aeolus at the initial time, while this percentage is reducing with forecast time as the forecast error grows.