Analyses & Numerical Prediction

Introduction

The last derivation of Model Output Statistics (MOS) coefficients was in 1986 and there have been extensive changes to operational models since then. Also, many observational sites have been automated while others have either disappeared or changed their reporting pattern.

There has been a growing demand for an updated Model Output Statistics(MOS) system based on more recent Regional NWP output and providing forecasts for a greater number of stations.

These requirements have been addressed with a new system, to be known as Model Output Forecast (MOF), and applied to the Australian region LAPS model. The new MOF runs twice a day, extends the prediction time to 48 hours at 3-hourly intervals and covers an increased range of predictands.

This bulletin briefly describes the new LAPS MOF(Version 1).

The current MOS products will continue to be distributed for the time being.

Further information is available in the references 'Derivation of new Model Output Statistics Coefficients' and 'Model Output Forecast Reference Manual' both of which are available through the Bureau's internal web server (See References)

These documents provided much of the information summarised in this bulletin.

Method:

The equations are derived from a linear regression of a historical set of dependent variables provided by a numerical model (as potential predictors) and near concurrent weather elements (as independent variable predictands).

The Developmental Data Period (DDP) was 1 June 1993 to 31 August 1997. During that period the operational regional models were RASP and LAPS.

The RASP (Regional Assimilation Model) model ran until 03 May 1994 when RASP75 was introduced with 75km horizontal resolution. LAPS (Limited Area Prediction System) replaced RASP75 from 03 July 1996 and has been running ever since. This DPP is barely sufficient for generating stable coefficients.

MOS previously generated forecasts for 185 stations once a day. MOF will issue forecasts twice daily for the 447 stations whose observational record during the DDP was sufficient to generate regression equations.

An important difference between daily and 3-hourly predictands is measuring period. Whereas the 3-hourly values refer to a mearuring period of three or less hours, the daily values are recordings taken at 9am local time and refer to events within the previous 24hr period.

Forecast elements will be a sub-set of the following

Daily Forecast Fields

3-hourly Forecast Fields

NB. Additionally 3-hourly relative humidity is derived from the dry-bulb and dewpoint temperatures and disseminated as part of the MOF output.

Performance:

Verification of 22 sites for the independent data period April, May and June 1998 was performed for the daily elements Maximum and Minimum temperatures. These were compared with the results obtained by MOS for the same period.

MOS and MOF compared over independent data (1998)

(Results supplied by F.Woodcock (Techniques Development))

BIAS
Month	Max Temp		Min Temp
	MOS	MOF	MOS	MOF
April	-0.3	1.2	0.2	0.3
May	-0.7	1.2	0.4	0.4
June	0.2	0.7	0.8	0.0
Average	-0.3	1.0	0.5	0.2
RMSE
April	2.4	2.7	2.4	2.1
May	2.3	2.5	2.4	2.2
June	2.0	2.0	2.5	1.9
Average	2.2	2.4	2.4	2.1

MOF consistently produced better minimum temperature forecasts than MOS but the maximum temperature forecasts were too warm. This was due to a strong warm bias in Canberra, Charleville, Brisbane and Dubbo. It is thought to be due to the unusual weather associated with the El-Nino event and the short data base from which MOF was developed. This deficiency can be rectified by applying a Kalman-filter as a post-processing step to the operational product. Preliminary testing of this process improved the April bias for Canberra from 2.6 to 0.4 deg C and the RMSE from 3.6 to 2.8 deg C. In May, the corresponding improvements were 3.0 to 0.7 deg C and 3.7 to 2.6 deg C. This step requires further development and will be implemented in the next version.

Verification details of MOF forecasts for individual stations for each month of the test period can be seen in the 'Derivation of new Model Output Statistics Coefficients' documentation. MOS verification statistics are already routinely distributed.

Limitations:

It is acknowledged that the length of time covered by the DDP is rather small and that the period of independent data used for verification may be atypical. The verification results are for a small sample of 22 stations and only three months data and this suggests that in the full set of 447 sites there could be a considerable number of stations showing significant warm biases in the maximum temperature predictions during the warmer months. Also, there has been no verification data on elements other than max and min temperature so users should be aware of potential problems.

Future plans:

• Add Kalman-filter post processor
• Add a quality of forecast flag if the forecast deviates more than two standard deviations from a climatology generated from model interpolated fields.
• Regular upgrades to increase both the data development period and incorporate any improved methodology
• Expand to forecasts based on alternative operational models
  eg. GASP, MESO_LAPS
• Integrate with AIFS Forecast Guidance and Verification Systems to routinely provide verification statistics
• Possible extension of number of forecast sites

Communications:

The operational MOF will be produced twice daily in NMOC and distributed via CMSS using ftp. All Regional offices will receive a text message which will be sent to the AIFS Product Store or in the case of non-AIFS regions their local WWW data store. MOF text messages will be viewed using any web based browser and in AIFS regions additional tabular, map or timeseries displays are available through the Forecast Guidance System module.

Help Desk:

The first contact point for problems with receipt of the MOF messages should be to the NMOC Help Desk on (03) 9662 2182. This is a 24hr service.

Documentation:

For fuller details of derivation and implementation, the web sites in the references are recommended. The 'Derivation of new Model Output Statistics Coefficients' includes a description of the coefficient files and hyper-links to these files for on-line viewing.

References:

Model Output Forecast Reference Manual, Version 1

Derivation of new Model Output Statistics Coefficients, Version 1

Home | About Us | Learn about Meteorology | Contacts | Search | Help | Feedback Weather and Warnings | Climate | Hydrology | Numerical Prediction | About Services | Registered Users | SILO

© Copyright Commonwealth of Australia 2008, Bureau of Meteorology (ABN 92 637 533 532) Please note the Copyright Notice and Disclaimer statements relating to the use of the information on this site and our site Privacy and Accessibility statements. Users of these web pages are deemed to have read and accepted the conditions described in the Copyright, Disclaimer, and Privacy statements. Please also note the Acknowledgement notice relating to the use of information on this site. No unsolicited commercial email.