In‑situ observations, and in particular surface land observations, play an important role in numerical weather prediction (NWP) models. At ECMWF, these land observations are used in the atmospheric data assimilation system (4D‑Var) as well as in the Land Data Assimilation System (LDAS). Over land, SYNOP weather reports from ground stations are exchanged via the World Meteorological Organization (WMO) Global Telecommunication System (GTS). They are the primary source of measurements for near-surface atmospheric parameters, such as surface pressure, 2‑metre temperature, 10‑metre wind, 2‑metre humidity, and snow depth. For example, ECMWF’s atmospheric 4D‑Var currently assimilates surface pressure and daytime relative humidity, and there is work in progress towards assimilating 2‑metre temperature and humidity (both day and night). Good coverage and better use of these near-surface observations contribute to a more accurate estimate of the initial atmospheric state (analysis). This in turn leads to a better forecast, particularly for near-surface parameters, which are key for many forecast users.
In recent years, ECMWF has made an effort to acquire and use additional surface observations not being distributed on the GTS. ECMWF’s support for the WMO-initiated South-East European Multi-Hazard Early Warning Advisory System (SEE-MHEWS‑A) project represented an important milestone for the development of custom acquisition and pre-processing workflows enabling the use of additional surface observations.
Additional surface observations
Some tools were developed within the Scalable Acquisition and PreProcessing (SAPP) system at ECMWF to facilitate the acquisition of SEE-MHEWS-A SYNOP data that could not be provided in the standard BUFR format. This includes handling local and national identifiers for stations not yet registered in the WMO’s repository of metadata for surface-based observing stations, OSCAR/Surface. These developments also made it possible to acquire and use observations from the Forecasting & Monitoring of Weather Related Natural Disasters unit of the National Observatory of Athens (METEO/NOA) and Météo-France’s extended network, including the RADOME network (MFR RADOME/Etendu).
Customised workflows were also implemented to enable the acquisition of open data from the Meteo Italian Supercomputing Portal (MISTRAL) and open surface observations from the German National Meteorological Service (DWD) and the Finnish Meteorological Institute (FMI). The first figure shows the locations of stations from which data were acquired from the sources mentioned.
Major developments made in handling WMO Integrated Global Observing System (WIGOS) Station Identifiers (WSI) were crucial for enabling the use of surface observations from newly registered observing stations. WSI processing was introduced in ECMWF’s Integrated Forecasting System (IFS Cycle 47r1). The entire data handling software stack had to be adapted to account for the increased size of the new alphanumeric identifier compared to the five-digit numeric Traditional Station Identifiers (TSI). Thanks to these developments, it was possible to start processing additional data available on the GTS with only WSI. The second figure shows the locations of stations with WSI whose BUFR SYNOP data are acquired through the GTS. More than 450 of these stations have only WSI available, reporting either hourly (e.g. the ones from Colombia and Israel) or sub-hourly (e.g. the ones from Hungary and Slovenia). The rest of the stations have both WSI and TSI.
When batches of new stations are made available to 4D‑Var by the acquisition system, their observations are not assimilated until their quality has been assessed based on statistics of their departures from a short-range forecast (the ‘background’). This procedure is intended to ensure that new observations are safely introduced into the process of generating the initial conditions for a new forecast (the ‘analysis’) without the risk of introducing spurious features. Despite a huge increase in the number of non‑GTS stations, a large majority do not report surface pressure observations, which is one of the most important atmospheric quantities for global models. Of those that do, a large proportion are providing observations of a quality deemed good enough to be assimilated. The data selection is reviewed periodically, and good-quality stations will be added. On the other hand, most of the non‑GTS stations report 2‑metre temperature observations, which at the moment are only used in LDAS. However, only a small percentage is actually used in LDAS due to poor observation quality, observation redundancy or elevation differences between the station and the corresponding model grid cell.
The way ahead
ECMWF has made an effort to acquire and process a wealth of surface data that had not been exploited for global NWP purposes before. As a result, some issues in the quality of the data have been exposed. However, this effort will eventually pay off if the quality issues that are preventing the use of some of the new observations are fixed.
Enhancing observation capabilities is also a goal of the EU’s Destination Earth initiative (DestinE), in which ECMWF is involved. Observations are not just crucial for data assimilation but also very useful for forecast evaluation. This applies in particular to forecasts of extreme weather events, which is one of DestinE's areas of interest. ECMWF is working together with several European national meteorological services and EUMETNET in the EU-funded RODEO project (https://rodeo-project.eu), which will support the provision of open access to public meteorological data.
In addition, the WMO Global Basic Observing Network (GBON), which went into effect on 1 January 2023, represents a cornerstone of the WMO’s strategy to secure observational data for critical global weather and climate applications. The GBON provisions for surface stations require the exchange of hourly observations. Many ECMWF Member and Co‑operating States already provide hourly data, but where 3- or 6‑hourly observations are currently exchanged, we encourage the exchange of hourly data to enable improved forecast products.