The latest Met Office Global model improvements
November 2016 – The way we use satellite data in our Global Numerical Weather Prediction (NWP) model has changed, leading to significant improvements to our global forecast skill.
Satellite data makes up a large proportion of the observational data used to initialise our forecasts. Having accurate observations of the state of the atmosphere is an essential starting point for NWP.
Satellites have the advantage of being able to observe large portions of the globe, including areas where ground-based observations are sparse. The Met Office is constantly working to improve our use of satellite data and to bring in observations from more satellites to give better forecasts.
The Global model has benefited from a package of satellite changes as part of the Parallel Suite 38 upgrade, which became operational earlier this month. These changes are described in this article.
Improved use of infrared and microwave satellite data over land
The Global model has benefited from improved use of infrared and microwave satellite data over land, providing improved temperature and humidity information at near-surface levels in the analysis.
Assimilating satellite radiances with strong surface sensitivity has long proved a challenge due to the difficulty in estimating the spectral emissivity of the land surface accurately. This has been solved using a dynamic, cycle-by-cycle, variational retrieval of surface emissivity and skin temperature. This is carried out in a pre-processing step prior to the main assimilation for the additional IR channels and by use of an emissivity atlas for microwave channels. These retrieved estimates are then used to model the observed radiances in our main 4-dimensional variational analysis system (4D-Var).
Improved global humidity analyses and forecasts
Global humidity analyses and forecasts have been improved by giving increased weight to satellite humidity information from microwave radiances in our analysis. Observations errors, which determine the weight given to radiances in the analysis, have been diagnosed and tuned to give optimal fits to independent humidity sensitive observations in short range (T+6 hour) forecasts.
Chinese polar orbiting satellite data
More Chinese polar orbiting satellite data has been incorporated, from the FY-3B microwave humidity sounder (MWHS-1), following our initial introduction in March 2016 (FY-3C MWHS-2). The first incorporation of polar orbiting data from Chinese satellites represents a landmark in the expansion of the satellite constellation assimilated in our Global model. Beyond the benefits of improved coverage, redundancy and resilience, data from China’s satellites are expected to enable us to continue to drive down analysis errors over the coming decade.
Japanese Geostationary satellite data
Data from a new Japanese Geostationary satellite, Himawari-8, has been introduced. Clear sky radiances from the Advanced Himawari Imager (AHI) have improved the analysis of temperature and humidity. AHI is a 16-channel imager (spanning the visible to thermal IR) offering a temporal resolution of 10 minutes at a resolution of 1-2 km.
Improvements to cloud-tracked winds
Wind information from satellites is obtained by tracking clouds or water vapour features in consecutive satellite images. These are known as Atmospheric Motion Vector (AMV) observations. The latest upgrade has seen an improvement in our use of AMVs through the inclusion of (Suomi-NPP) VIIRS winds, height correction for low level AMVs in the presence of a temperature inversion and quality control of AMVs assigned in dry layers of the model. The inversion correction and quality control changes aim to address known issues with AMV height assignment and result in an improved fit of short-range forecasts to the AMVs and humidity-sensitive sounding channels.
The new global forecasting system configuration has been run as Parallel Suite 38 during the period 13th July – 8th November 2016 (providing 174 forecasts at 00Z and 12Z), permitting a reliable verification of the upgrade. The changes result in a significant improvement in our global forecast skill (see Figure) amounting to +1.0 / +1.6 NWP index points against observations/analysis respectively. This represents around half of the normal annual improvements in skill (typically 2 points per year). For example, extra-tropical sea level pressure errors are reduced by 0.5-2.0 % for forecast days 3-5, and errors in extra-tropical geopotential 500hPa heights are reduced by 1-2%.
The figure above shows the improvements in forecast errors for sea level pressure, 500 hPa geopotential height and winds at 250 hPa in the Northern and Southern extra-tropics, and additionally 250 and 850 hPa winds in the Tropics. Improvements (indicated in green) in RMS errors are shown for forecast ranges from T+12 hours to T+120 hours. Verification is relative to observations (left) and own analysis (right).