Operational Numerical Analysis & Prediction System

Introduction | GASP | LAPS | TXLAPS | WAM | Storm Surge Model
Sea Surface Temperatures | Sub Surface Temperatures
Coupled Ocean-Atmosphere Models | OCF | Other Numerical Model Output

Introduction

NMOC runs numerical models over several domains to provide analyses and predictions for the atmosphere and ocean on a range of time-scales. The models have been developed in the Bureau of Meteorology Research Centre. The most detailed predictions are for targeted areas over the Australian region and extend to 36 or 48 hours. Medium range predictions to 7 days are made with the global model. NMOC also runs numerical models to predict the state of the ocean. These range from products relevant to marine meteorological forecasts such as surface wind and wave forecasts to coupled atmosphere- ocean systems used in the preparation of seasonal outlooks. Many diagnostic fields are available from the models to assist in forecasting such as stability indices and deep layer mean winds. The models are used with other computer systems to produce guidance on sea state, atmospheric dispersion of contaminants, and UV-B prediction.

GASP (Global AnalysiS and Prediction)

The global model (GASP) uses spectral numerical techniques and is currently run with triangular wave number T_L239 truncation (approximately 85 km horizontal resolution). The model's domain is global, and has 29 levels from the surface to 10 hPa. Analyses of observations are made at 6-hourly intervals and twice per day forecasts are generated out to 7 days. The GASP system provides boundary conditions for the limited area systems and also provides the surface winds used in the global sea state prediction scheme.

• Operations Bulletin No. 52
  Operational Implementation of 1DVAR GASP in NMOC 18 August 2000
• GASP Charts

LAPS (Limited Area Prediction System)

The prediction component of LAPS is a grid point primitive equation model with a grid spacing of 0.375^o (about 37 km) in the horizontal, and 29 levels in the vertical, from the surface to 50 hPa. The lateral boundary conditions are obtained from the GASP system. The analysis component involves a 6 hourly data assimilation cycle. The system operates with a short data cut-off time to ensure timely delivery of forecast guidance to forecast offices. Consequently, the data analysis stage places special emphasis on the use of locally derived satellite observations.

LAPS operates over several telescoping domains. The largest extends from about 17^oN to 65^oS and 65^oE to 184^oE and is used for broad-scale predictions over the Australian region to 48 hours.

Several Meso-LAPS systems operate to provide finer scale predictions for sub-areas of the larger Australian region domain. The resolution of this next level is 0.125^o (about 12 km). Several test systems at an even finer scale of 0.05^o (5 km) are in test mode for areas round Sydney and Melbourne. These fine-scale predictions are being used by BMRC and CSIRO in a test of an air quality prediction system.

• Analysis and Prediction Operations Bulletin No. 58
  Operational Upgrade to LAPS_PT375 11 April 2003
• Analysis and Prediction Operations Bulletin No. 64
  Operational Upgrade LAPS_PT375 to 51 levels 3 May 2005
• LAPS Charts

TXLAPS (Tropical eXtended Area Prediction System)

(A joint effort between Northern Territory Regional Office and NMOC) TXLAPS is a version of LAPS customised to describe the tropical atmosphere. In particular an attempt is made to capture areas of convection that are observable as cloud formations on satellite imagery and which can lead to
the development of tropical depressions. In addition to available conventional data, input to the analysis includes bogus moisture values derived from satellite cloud imagery. Tropical Cyclones are incorporated into the analysis using a bogus system operated by Darwin RSMC staff. The first guess field for this analysis is a forecast from GASP. TXLAPS currently operates over the domain 48^oN to 45^oS and 60^oE to 143^oW with a grid spacing of 0.375^o.
TXLAPS was upgraded from 29 levels to 51 levels in April 2005.

• Analysis and Prediction Operations Bulletin No. 59
  Operational implementational of TXLAPS_PT375 1 December 2004
• Analysis and Prediction Operations Bulletin No. 63
  Operational Upgrade TXLAPS_PT375 to 51 levels 19 April 2005
• TXLAPS Charts

WAM (WAve Model)

The WAM model solves equations for the directional wave spectrum on a regular grid based on local wind input, wave dissipation, nonlinear wave-wave interactions and propagation of waves from nonlocal sources (swell).

The Wave Model is run over three domains each run twice a day and including a 12 hour assimilation period using satellite derived wave height data. A global model at 3 degree resolution forced by surface winds from GASP produces forecasts generated out to 6 days and boundary conditions for the regional system. Regional (1 degree resolution) and Meso- scale (0.25 degree resolution) versions of the model using surface wind forecasting from the respective atmospheric models are used for local forecasting out to 48 hours.

• Analysis and Prediction Operations Bulletin No. 55
  Changes to the Operational Sea State Forecast System 27 August 2002
• WAM charts

Storm Surge Model

Strong onshore winds associated with tropical cyclones and other intense storms produce a rise in the sea level which can be the cause of a lot of major flooding. A system has been set up based on a simple ocean model to forecast the storm surge associated with tropical cyclones. The wind forcing is based on an ideal distribution calculated from the central pressure of the cyclone at any time and the radius of maximum winds.

Sea Surface Temperature (SST) Analysis

A global SST analysis is performed weekly on a 1 degree resolution grid. In the analysis measurements of the sea temperature from voluntary observing ships and drifting buoys are used to remove biases in remotely sensed data from satellites before being blended into the final product. A finer resolution product for the Australian region is produced daily based on locally received satellite data. The global analysis is used by operational atmospheric models and in the production of climate forecasts and seasonal outlooks.

Sub-surface Temperature Analysis

An analysis of monthly temperatures at a number of levels in the ocean down to 500 metres is updated weekly. This analysis uses data from moored buoys, bathythermograph reports from the ship-of-opportunity program and profiling floats.

Coupled Ocean-Atmosphere Models

NMOC runs an operational coupled ENSO forecast model. This is an intermediate class model which is initialised using sub-surface ocean data and surface wind data. It produces forecasts of the SST anomaly in the central Pacific Ocean out to 2 years which are used operationally in climate assessments and outlooks.

Operational Consensus Forecasts (OCF)

The OCF system is a new type of automatic weather forecasting scheme developed by the Bureau of Meteorology. It uses forecasts derived from the Australian Regional numerical model (LAPS) and from global models from Australia (GASP), the European Centre for Medium Range Weather Forecasts, Japanese Meteorological Agency, United Kingdom and United States of America.

There are two ways that objective forecasts are derived from models. The simplest is Direct Model Output (DMO) where the weather at a particular location is interpolated from model grid values. In OCF, DMO forecasts are derived from all the available models and include LAPS runs at 5, 12.5 and 37.5 km resolution. DMO forecasts are currently available from the Bureau as meteograms. While DMO forecasts respond quickly to model upgrades, interpolations and parameterisations in the models can result in forecasts with large biases.

A more complex forecast is available using relationships built on a history of model archives and their corresponding weather. The most successful of these methods is known as Model Output Statistics. These forecasts, known locally as MOF, are also available from the Bureau of Meteorology. A significant problem with MOF is that about three years of data are required before the relationships between the weather and the models stabilise by which time the models have been upgraded. The resultant MOS forecasts then tend to become biased and occasionally they become erratic.

OCF has been designed to overcome the problems of MOF and DMO forecasts. Everyday the bias in MOS and DMOs for each site is computed by comparing the forecasts with the verifying observations over the last 30 days. Then the last 30 days of forecasts are corrected for bias and the resulting mean absolute error (MAE) of the corrected forecasts are stored.

In order to make new OCF forecasts, the latest MOS and DMOs are first corrected by their 30-day bias and then combined using weighted average of their MAEs. This means that the OCF forecast is weighted towards the historically best performing, bias-corrected, contributing forecasts and it can adapt to model upgrades quickly.

In verification results over the last year under parallel, real-time conditions OCF has produced absolute errors in day 1 to day 3 temperature forecasts that are 20% to 60% lower than corresponding MOS and DMO forecasts.

Rainfall forecasts are derived by averaging model DMO rains at the forecast site. Similar averaging is used for probability of rain with one being assigned to model rain if it exceeds 0.2mm at the site and zero otherwise.

Other numerical model output

An atmospheric transport system can be run on demand to produce forecast trajectories and concentrations for volcanic ash, smoke, radioactive debris from nuclear accidents or atmospheric contaminants arising from other environmental emergencies.

Upper wind and temperature forecasts for the aviation industry.

Analysis and forecast charts derived from the global analysis and prediction models of several international centres, including the European Centre for Medium Range Forecasts (ECMWF), United States, Japan and the United Kingdom. Forecasts out to 36 hours of the UV-B index by coupling an atmospheric UV radiation scheme to analyses and predictions of vertical profiles of ozone concentration and temperature.

Statistical guidance (Model Output Forecasts), derived from the numerical model predictions, of weather elements such as maximum and minimum temperatures, probability of rainfall and cloud cover.

Samples:
Samples of some of these models may be found on the Bureau's ftp site: Samples