ASCAT ocean surface wind assessment

The European Centre for Medium-Range Weather Forecasts has been contracted by the European Organization for the Exploitation of Meteorological Satellites (EUMETSAT) to perform an evaluation of ASCAT wind measurements, assess their impact on the Global Observing System (GOS) and optimize the assimilation strategy. This report presents the results of the work done during the two years (February 2013-February 2015) of the contract (Project Ref. EUM/CO/12/4600001149/JF).  

The impact of the ASCAT-A and ASCAT-B winds has been assessed over different GOS scenarios: one is replicating the operational ECMWF system and is using all the available observations; two scenarios use a subset of the GOS (all observations except wind observations and all observations except wind observations and AMSU-A) to assess the interaction between scatterometer observations and other sensors.

The assessment of scatterometer winds has been performed using a range of diagnostics, from the traditional forecast scores to the verification against independent observations such as altimeter winds, wave height and wind speed buoy data. The verification methods show similar results.

The main positive impacts are made when either or both ASCAT datasets are assimilated together with OSCAT data; ASCAT-A and ASCAT-B have the same impact on the system. From all the verification methods, it is shown that, in a Full System configuration, the assimilation of scatterometer observations is globally beneficial on the analysis; however the benefit is not propagated into the forecast. Verifications against buoy and altimeter winds show that when other wind observations are removed from the GOS, the positive impact of assimilating scatterometer observations at analysis time is larger and is propagated to longer forecast range. Regional statistics show that overall the largest benefit is coming from the Tropics.

The assimilation of ASCAT-B winds has a positive impact on the analysis departure of ASCAT-A. It has a neutral to positive impact on the analysis departure of OSCAT in the Northern Hemisphere and in the Tropics but is slightly negative in the Southern Hemisphere. This is most likely due to the OSCAT wind speed bias seen in the Southern Ocean (mostly south of 50° S), which is known and already partially corrected by KNMI.  It is found that Scatterometer observations have impact only up to 600 hPa; scatterometer assimilation appears to make almost no difference above this height. This is also found to be the case for other near surface observations, especially wind, and it is not specific to scatterometer observations. To better understand the ability of 4D-var to propagate the scatterometer increments from the surface to higher model levels, single observation experiments were run assimilating only one scatterometer observation close to the centre of a Tropical Cyclone (TC) and in an area where the scatterometer wind and the model value were close. Results showed that close to the TC, the 4D-Var is able to propagate the scatterometer wind information from the surface to the upper troposphere. To assess the synergies of scatterometer observations with other types of observation, experiments assimilating one scatterometer and a couple of AMSU-A observations were run showing that the analysis increments structure is not modified when AMSU-A observations are also assimilated, either at low or high model levels. This suggests that the large impact of AMSU-A is not limiting the impact of ASCAT.

Forecast Error Contribution statistics show overall a higher impact for OSCAT winds due to the higher number of assimilated observations. However, statistics computed for each single observation show that a single ASCAT observation has higher impact than a single OSCAT observation. ASCAT-A and ASCAT-B have the same impact. Moreover, regional statistics show that the largest scatterometer observations impact is from the Southern Hemisphere.

The impact of ASCAT winds has been evaluated on tropical cyclone events. Global statistics of mean sea level pressure and storm centre position error do not show a clear benefit when assimilating scatterometer winds when based on all the TCs occurring during the test period. However when the analysis is repeated taking into account only TCs where scatterometer observations where available at analysis time root mean square (RMS) forecast error of the minimum sea level pressure (SLP) in the centre of the storm is reduced. It was found that during the test period the position errors for all the configurations were, in general, small in comparison to model resolution, such that differences in performance between the configurations were negligible.

A detailed analysis of the impact of ASCAT-A winds on the analysis and forecast of the Typhoon Haiyan, which hit the Philippines in November 2013 has been performed. Overall the assimilation of Scatterometer observations is beneficial for the storm analysis and forecast. However it was noticed that some strong ASCAT winds in the area of maximum storm intensity were rejected prior to the assimilation partially because of the thinning applied (only one observation out of four is assimilated) and partially due to the quality control. If the wind vector difference between the background and the observation is too high the observations are rejected. In this case, the wind vector difference is likely due to a displacement of the storm location in the background. Tests were performed for four TCs where this problem occurred, which showed that there is a general sensitivity of the data assimilation to changes in thinning and quality control set-up. Preliminary tests on the use of an alternative method to the current quality control, the Huber norm, were also run. This is a robust method which allows observations with large background departure to still give some weight into the analysis. The results showed that there is indeed potential to increase scatterometer impact further through fine tuning of these components. This will be important for the new SCA scatterometers on EPS-SG scatterometer, as it will better observe high winds.

The impact on the surface stress were also evaluated. The assimilation of scatterometer winds increases the surface stress almost globally. Few areas in the Tropics showed lower values when the observations are assimilated. Since surface stress is strongly connected to the surface winds, and in-situ measurements are not available, the verification in few tropical sub-areas was done using Altimeter and buoy winds. Results confirmed that the assimilation of scatterometer winds is beneficial in these areas in terms of winds, and thus also in terms of surface stress.