Mesoscale Improved Data Assimilations of Scatterometer winds (MIDAS)

MIDAS wass a collaborative project between The Portuguese Institute for Sea and Atmosphere (IPMA) and The Royal Netherlands Meteorological Institute (KNMI). Its aim was to provide EUMETSAT users and the convection-permitting Numerical Weather Prediction (NWP) community valuable insights on the use of scatterometers in Data Assimilation (DA).

Objectives

The main objectives of the study were as follows:

(1) Assess the impact of scatterometer data assimilation (DA) on mesoscale NWP.

(2) Investigate strategies to maximise the benefits of denser space-time scatterometer wind observations in mesoscale DA.

(3) Determine the optimal spatial-temporal coverage of scatterometer DA in mesoscale NWP.

Through the benefits in regional mesoscale NWP, optimal use of scatterometer winds are expected to yield several benefits for marine alert advisories and the offshore economy, including transport, energy and tourism sectors. In addition, it will provide improved high-resolution winds for driving ocean, wave and surge models. Moreover, insights gained from this study might contribute to the future transition to Metop-SG.

Overview

IPMA and KNMI conducted this study using the HARMONIE-AROME 4D-Var Data Assimilation (DA) system within a domain covering south-west Europe. This investigation focused on integrating scatterometer wind data into the non-hydrostatic convective-scale-resolving model HARMONIE-AROME. Evaluation of the 4D-Var DA configuration was performed in comparison to the 3D-Var formulation still in use within Numerical Weather Prediction (NWP) centres employing HARMONIE-AROME for operational purposes. The study reveals the superior performance of the 4D-Var system, particularly in predicting 10-meter winds over both ocean and land surfaces. The benefit of including scatterometer winds in the assimilation system was also demonstrated. Several observing system experiments (OSE) were conducted to optimise the utilisation of scatterometer winds within HARMONIE-AROME 4D-Var (Fig. 1). These experiments involved testing thinning procedures, observation error prescription and superobbing of observations.

While the benefits of using the 4D-Var configuration and incorporating scatterometer observations into the assimilation system were evident, the various approaches tested for utilising scatterometer winds showed only marginal differences. However, it was observed that HARMONIE-AROME exhibits systematic biases, both in speed and direction, over ocean when compared to ScatSat and HSCAT derived winds (Fig. 2). Model winds consistently display a clockwise rotation compared to observations, along with systematic overestimation of speed. Although the biases are partially corrected by 4D-Var assimilation and the inclusion of ASCAT winds, the model quickly reverts to its biased atmospheric state. This study demonstrates the necessity of further investigating the underlying causes of these biases in HARMONIE-AROME to effectively address them.

Fig. 2 - 7 February to 15 March 2020 +0h forecast departures (observations - forecast) of 10-m model wind over oceans for the zonal (upper row) and meridional (central row), and wind speed (bottom row) with respect to ScatSat observations. Left panels for 3DVar twin experiment, and right panels 4DVar_ctrl. In the bottom panel observation (black) and model (red) median wind direction taken over the experimentation period are overlaid. In the statistics computation, all analysis cycles are considered and observations and forecasts are binned on a 0.5 by 0.5 degrees grid. Number of matches between observations and model forecasts is ~80 for each grid cell over the domain.

MIDAS conducted an extensive and detailed investigation into the use of ASCAT data in HARMONIE-AROME, leading to the identification of several errors in the utilisation of this data. A subsequent study at Météo-France revealed a bug in the code for surface observations assimilation in the 4D coupling with SURFEX, which affected the global model. This bug was associated with two significant forecast errors in the ARPEGE model over Europe. As a result, this led to a correction in the common code by Météo-France.

Furthemore, MIDAS contributed to the operationalisation of existing software for verifying HARMONIE-AROME forecasts against independent scatterometer winds, initially developed within the framework of NWP SAF (Visiting-Scientists/NWP SAF). This contribution enabled the operational use of the software in the ACCORD community (scr4verification).

MIDAS also introduced new research topics and provided improved insights into HARMONIE-AROME. One significant finding was the identification of systematic biases in both speed and direction of HARMONIE-AROME over the ocean, compared with ScatSat and HSCAT derived winds. This study documented, for the first time, a clockwise rotation in surface wind direction in the model relative to observations, which is one of the main outcomes of MIDAS.

Full details of the study can be found in the Final Report linked below.