Glowacki, T.J., Powers, P.E. and Keenan, T.D. 1997: Development of A Real Time AWS Data Mesoscale Analysis System. AIFS Nowcasting Workshop 24-26 June 1997. Bureau of Meteorology, Melbourne, Australia.

Development of A Real Time AWS Data Mesoscale Analysis Systems

TOMASZ J. GLOWACKI, PETER E. POWERS AND THOMAS D. KEENAN

SUMMARY

Fine resolution analyses of high frequency surface observations are an important tool for mesoscale forecasting if available on time. Present operational data assimilation systems covering the Australian Domain are not suitable for nowcasting applications for various reasons. The analyses do not make use of all available surface data and important surface variables are not analysed or analysed in insufficient detail for specific mesoscale applications. In addition manual analyses are only produced three hourly, a time scale over which significant mesoscale changes can occur.

The network of Automatic Weather Stations (AWS) is a major source of surface observations and is currently underutilised. The workshop presentation takes a closer look at the current stage of development of an objective, mesoscale surface variable analysis system and addresses problems which influence its performance. The system produces real time analyses of AWS data on a workstation and employs an analysis technique suitable for application to a mesoscale domain of strongly diversified topography. An approach worked out in the MAPS project (Miller and Benjamin 1992) is followed with some modifications, and tested over a smaller domain (of 400 x 400 km centered on Sydney) with a finer grid (4 km). Twenty seven AWS reporting pressure (11 AWS without pressure sensors), temperature, dewpoint temperature and wind components are used. Basic variables of mslp, potential temperature, dewpoint temperature and wind components are analysed first in sequence, using specifically tailored statistical interpolation scheme, then analyses of a number of derived variables are produced. An independent, interactive graphics system displays observational data and analysis results and complements the analysis system. It is written in IDL (Interactive Data Language) and allows user to control graphics through IDL widgets. The whole system is intended to be used as a tool for broadening the basis for manual diagnosis and enhancing nowcasting of local weather events. Exemplary analyses are given.

Statistical interpolation seem to be particularly suitable for application because of its parameterized flexibility e.g. allowance for background and measurement errors. The simplifying assumptions of isotropy made usually about the shape of horizontal autocorrelations and cross-correlations do not hold well for mesoscale circulation systems, especially, where pronounced topographical features like large altitude differences and land-sea boundaries influence the variances of analysis variables. Surface observations are also subject to greater diurnal variability than upper levels data. Consequently, background error correlation adjustments for altitude and potential temperature differences are applied. However, this approach has limited use because of numerical problems. Excessive adjustments lead to a non-positive definite covariance matrix. The following enforced covariance matrix correction has an adverse effect on analysis accuracy. This problem is more severe for the selected NSW domain than for the original MAPS domain, due to a tow information content of the NSW AWS system in relation to assumed analysis grid resolution.

The AWS network within the domain is congested in Greater Sydney area leaving vast mountain areas sparsely covered and ocean areas void of observational data. Short length scales are used because of the importance of the data dense Sydney area, where we want to resolve small scales and analyse local weather systems. This can result in unrealistic analysis extrapolations and analysis stray when analyses are used in cycle in data sparse areas. The assumed remedy i.e. replacement of analysis subdomains of excessive extrapolation with regional (LAPS) forecasts is riot entirely suitable because the model has different time and space scales what affects homogeneity of the resultant mesoanalysis fields.

Production of accurate, small scales resolving analyses, maintaining time and space coherence over the whole analysis domain given the irregularity of data distribution and topographical diversity is a task of somewhat conflicting goals. Enforcing both some specific engineering elements in the analysis system design and an increased subjective input into the 'objective' analysis scheme is necessary. It comes in a form of additional analysis parameters. Prolonged testing of the system in stable conditions (LAPS first guess field) is needed for proper tuning of all the parameters.

Extracting and using information present in time series of observational data seem to be necessary for further increase of analysis accuracy.

REFERENCES

Andersson, E., N. Gustafsson, L. Mueller, and G. Omsted, 1986: Development of meso-scale analysis schemes for nowcasting and very short-range forecasting. SMHI Promis-Rapporter, No. I (January), 33 pp.

Benjamin, S.G., and P.A. Miller, 1990: An alternative sea level pressure reduction and statistical comparison of geostrophic wind estimates with observed surface winds. Mon. Weal Rev., 117, 1586-1605.

Miller, P. A., and S. G. Benjamin, 1992: A system for the hourly assimilation of surface observations in mountainous and flat terrain. Mon. Weal Rev., 120, 2342-2359.

Seaman, R. S. 1996: A comparison of some methods for reduction of pressure to sea level over Australia. Aust. Meteor. Mag., unpublished.

Uccellini, L.W., S.F. Corfidi, N.W. Junker, P.J. Kocin and D.A. Olson, 1992: Report on the Surface Analysis Workshop. Held at the National Meteorological Center 25-28 March 1991. BAMS., 73, 459-472.