Analysis and Prediction Operations Bulletin No. 56
Operational Upgrade to LAPS_PT375
24 September 2002

Other Charts | About Products

Introduction

The operational Australian region Limited Area Prediction System known as LAPS_PT375 was upgraded to a new version in NMOC Melbourne on 3 September 2002. This upgraded version (hereafter referred to as 1DVAR LAPS_PT375) was developed by the Model Development Group in BMRC, led by Dr. Kamal Puri, and the Data Assimilation Group in BMRC, led by Dr Bill Bourke. It incorporates a number of model and assimilation scheme changes including:

• Use of 1DVAR ATOVS radiance assimilation
• Reduced atmospheric diffusion to produce deeper systems
• New surface wind extrapolation algorithm to address low wind-speed bias
• Use of soil moisture nudging scheme
• Hourly radiation calculations
• Forecast range extended to 72 hours (from 48 hours).

The upgraded version underwent a parallel operational trial in NMOC during the month of August 2002 after a long period of initial testing in BMRC since late 2001. Both synoptic and objective assessments were carried out. Skill scores indicate that the new 1DVAR LAPS_PT375 generally performed slightly better than the previous model throughout the 48-hour forecasting period at different vertical levels.

 

Overall Scheme

The original configuration of LAPS_PT375 was described in Analysis and Prediction Operation Bulletin (APOB) No 46 of 23 July 1999. Much of the underlying structure of the new system remains as described in that report.  The 1DVAR LAPS_PT375 system continues to be based on the 3-cycle analysis/forecast scheme originally described in that bulletin, in which observational data are inserted every 6 hours (1DVAR LAPS_PT375 time). At each data insertion, an analysis is performed followed by a soil-moisture adjustment and then an initialisation and prediction. The prediction model is integrated forward to either the next analysis time or the end of a longer forecast. A schematic representation of the basic structure of the scheme is shown in Fig. 1 below.

The new system now incorporates 1DVAR (one-dimensional variational) ATOVS radiance assimilation and includes a radiance bias correction. This scheme was developed by members of the BMRC Data Assimilation Group and is very similar to the scheme used in the GASP system (Harris et al, 2000), the operational implementation of which was described in APOB No 52. In brief, instead of using TOVS satellite retrievals produced by NESDIS, the 1DVAR scheme analyses the cloud-cleared radiances from the ATOVS (NOAA-15 & NOAA-16) instruments. The radiances estimated from the first guess profile of moisture and temperature are compared with the observed values and the profile adjusted using a variational scheme to give a best fit to all the observed radiances. This adjusted profile is then used as the 'retrieval' in the analysis scheme with a weight adjusted for the fact that the first guess has already been used in deriving it.

In the new system, an additional preliminary analysis run is required for an ATOVS bias correction at cycle 1. A second bias correction calculation is done for the 06 or 18 UTC data after the analysis step in cycle 2. The results from each bias correction calculation are then used as input to the next assimilation cycle.

A "cold start" is still being used at the commencement of cycle 1, meaning that the GASP analysis becomes the first guess for the 1DVAR LAPS_PT375 analysis in the first cycle. After this, the first guess fields for the 1DVAR LAPS_PT375 analysis in the second and third cycle are provided by 6-hour forecasts from the prediction model in 1DVAR LAPS_PT375 itself. GASP also provides the boundary conditions for 1DVAR LAPS_PT375

The modelled soil moisture/wetness is however now being set in warm running mode, i.e. the initial soil moisture input is obtained from the previous 1DVAR LAPS_PT375 analysis. After each analysis step the soil moisture field is adjusted using a scheme developed at the ECMWF (Viterbo, 1996) whereby the scaled analysis increments in the lower level mixing ratio field are used to adjust the soil moisture field. The idea is that a too dry (or wet) soil will lead to a too dry (or moist) boundary layer in the first-guess forecast compared to the atmospheric analysis (which includes information from the humidity observations). The previously used Pescod soil moisture scheme results (based on rainfall observations and climatological surface evaporation rate estimates) are only used if a cold start is needed due to two missed runs.

 

FIGURE 1. A schematic representation of the basic structure of 1DVAR LAPS_PT375. Modified or added components are shown highlighted.

 

Data Analysis Changes

Analysis program changes include:

• Use of the 1DVAR ATOVS radiance assimilation as described above
• GASP background errors are now used by 1DVAR LAPS_PT375
• Improved quality control and validity checking of GMS moisture data
• The "boxfiles" used to store observational data for the system have been changed to version 6.4 with this upgrade.

 

Model Changes

Forecast model changes include:

• Use of the "unified" bam physics code (version 3.1) with changes such as a new input parameter file format, tuning of parameterisations, and various new options.
• The amount of atmospheric diffusion used in the model has been reduced significantly in order to produce deeper systems. On the domain boundaries the diffusion is doubled relative to the interior to suppress noise developing on the boundaries and infiltrating the inner domain.
• A new surface wind extrapolation technique has been introduced to address the low wind-speed bias found in previous versions of LAPS. New model input parameters now allow the sigma levels for surface wind extrapolation to be chosen, rather than always using the lowest sigma level as was done before.  The winds in the lowest sigma levels are strongly influenced by grid-scale surface roughness effects and these were often unrepresentative of the open conditions found at typical observing stations. Now, over land points, the diagnostic 10m surface wind output to the model history file is being extrapolated down from the third model sigma level using values of roughness appropriate for grass cover. Verification studies by BMRC have shown that this has increased the average 10m winds over land by approximately 0.5 m/s and significantly reduced the average bias with respect to observations.
• A higher resolution 1°x1°solar radiation surface albedo climatology has replaced the coarse-resolution 10°x10°climatology that has previously been in use.  The newer climatology has a significantly higher albedo (up to 10%) over parts of coastal western and northern Australia and lower values over central Australia. Higher albedos could be expected to decrease daily maximum temperatures slightly and vice-versa.
• The frequency of radiation calculations in the 1DVAR LAPS_PT375 has been increased to 1-hourly (previously was 3-hourly).

 

Operational Configuration

The general operational configuration of the 1DVAR LAPS_PT375 system remains essentially as described in APOB No 46, e.g. the model domain, resolution, vertical level structure etc have not been changed and readers should refer to that report for details, noting the following points:

 

• Due to the presence of the extra preliminary analysis and the bias correction suite of programs (balanced by some speedups due to code optimisations etc) there is a several minute increase in overall run time for the analysis steps on the NEC SX-5 supercomputer. The average overall elapsed time has thus increased from approx. 19.2 minutes to 21.3 minutes (4 CPUs, 2.5Gb memory)
• The increased +72 hr forecast range has resulted in the elapsed time for the prognosis step on the NEC SX-5 increasing from approx. 11.2 minutes, for the old system, to 13.2 minutes (8 CPUs, 2.3Gb memory).
• It should be noted that the analysis basetimes for all NWP model systems within NMOC were changed from 05/11/17/23 UTC to the conventional standardised hours of 00/06/12/18 UTC on 18 March 2002.

 

Performance

Synoptic Assessment

Terry Skinner of NMOC undertook a synoptic assessment of the 1DVAR LAPS_PT375 during the parallel trial period of August 2002 comparing mainly the mean sea level pressure analysis and prognosis products with the corresponding NMOC HORACE operational on-screen analysis charts. In addition, the surface temperature, rainfall and 250 hPa wind field forecasts were assessed. The following features emerged from the assessment:

Mean Sea Level Pressure Analyses

• Intensity of lows in westerlies south of Australia in 1DVAR LAPS_PT375 is often underestimated in comparison to the NMOC HORACE analyses and to the previous operational LAPS_PT375.

Mean Sea Level Pressure Prognoses

• Depth of lows in westerlies and cut off lows appear to be more realistic, compared against NMOC HORACE analyses.
• Frontal systems in westerlies in forecasts out to 72 hours appear to be well captured in location and amplitude, but not uniformly so.
• Pre-frontal systems over the continent appear to be more clearly defined.
• There is an increase in "numerical noise" in the unsmoothed full resolution prognoses compared to previous operational LAPS_PT375.  This is attributable to the reduced model diffusion.

Surface Temperature Forecasts

• Compared to the previous operational version, there is a negative bias for temperature predictions at 06 UTC, close to time of T_max. A negative bias appears over the tropics and Pilbara region of WA. MSLP forecasts over the tropics so far appear to be realistic.

Rainfall Predictions

• As in previous operational model, the performance in rainfall prediction is variable.
• Significant rainfall on tropical Queensland coast is often missed.
• Rainfall on Tasmania west coast is often underestimated.
• Occasional excessive rainfall prediction occurs over NSW and Qld interior on frontal changes

250 hPa flow

• Jet stream marginally stronger with cores slightly more extensive in the new model.

 

Objective Assessment

BMRC has been conducting verification studies of the 1DVAR LAPS_PT375 system since January 2002. Their detailed verification results are available.

NMOC also conducted an assessment of 1DVAR LAPS_PT375 in a parallel trial during August 2002 prior to operational implementation. This section presents results from the NMOC verification.

In addition to the conventional verification targets, such as the mean sea level pressure (MSLP) and the geopotential heights (HGHT), surface weather parameters such as the screen temperature dew point and 10m winds were also verified. The results from these two different targeted groups will be discussed separately below.

1. S1 Skill Score, Bias and Root Mean Square error for MSLP and HGHT

Table 1 below presents the mean S1 skill score, bias and root mean square (RMS) error of the previous LAPS_PT375 and the new 1DVAR LAPS_PT375 for August 2002 over the Australian verification region. The 1DVAR LAPS_PT375 generally performed better than the old model throughout the 48-hour forecasting period across the vertical levels. For MSLP, there was a S1 Skill Score gain of 1 at +12 hrs and 0.5 at +24 hrs.

2. Surface Weather Parameter Verification against the Aviation METAR Observations

A new verification display system that highlights the statistical characteristics of surface weather elements has been implemented in NMOC. The system, developed by Kevin Tory of the BMRC Model Development Group, verifies weather elements such as surface (2m, or screen level) temperature and dewpoint and surface (10m) wind direction and speed against METAR observational data. RMS errors and biases are calculated at regular intervals out to a specified forecast period making use of all available METARS (approximately 297 for the LAPS_PT375 domain).

Figures 2, 3 and 4 show the time series plots of the mean bias and RMS errors averaged over all METAR sites and over all hourly model forecast periods out to +48 hrs. Results are shown for 8 days only, since this verification procedure has only recently been implemented. It can be seen that, overall, there is a clear and consistent improvement of 1DVAR LAPS_PT375 over the previous LAPS_PT375 model in predicting surface temperature and wind.  The results for surface dewpoint are more ambivalent.

Further examples of verification against METAR data are available.

Table 1. Mean S1 Skill Score, Bias and RMS Error for MSLP and HGHT comparing the previous LAPS_PT375 model and the new 1DVAR LAPS_PT375.

Region: Standard Australian verification grid.
Total Period: 1 Aug - 31 Aug 2002 using both 00 and 12 UTC runs
Verifying Analyses: Self
Improvement = absolute value of LAPS_PT375 - absolute value of the 1DVAR LAPS_PT375.
A positive improvement implies that the 1DVAR LAPS skilled better than LAPS.
Units for bias and RMS are hPa for MSLP and m for HGHT. S1 skill score is non-dimensional.

Forecast period	Verification statistic	Field	LAPS_PT375	1DVAR LAPS_PT375	Improvement
+12HRS	S1 Skill Score	MSLP	15.0	14.0	1.0
		850 HGHT	13.0	12.0	1.0
		500 HGHT	8.5	8.5	0
		250 HGHT	6.5	6.0	0.5
	BIAS	MSLP	0.5	0.5	0
		850 HGHT	4.0	3.3	0.7
		500 HGHT	-2.1	-2.1	0
		250 HGHT	-3.7	-3.5	0.2
	RMS	MSLP	1.3	1.1	0.2
		850 HGHT	9.7	8.6	0.9
		500 HGHT	11.1	11.0	0.1
		250 HGHT	15.2	14.3	0.9
+24HRS	S1 Skill Score	MSLP	21.0	20.5	0.5
		850 HGHT	19.0	18.0	1.0
		500 HGHT	13.5	13.5	0
		250 HGHT	10.0	10.0	0
	BIAS	MSLP	0.7	0.5	0.2
		850 HGHT	5.5	3.9	1.6
		500 HGHT	-4.1	-3.7	0.4
		250 HGHT	-5.1	-5.0	0.1
	RMS	MSLP	1.9	1.8	0.1
		850 HGHT	14.5	13.5	1.0
		500 HGHT	17.8	18.0	-0.2
		250 HGHT	23.7	23.4	0.3
+36HRS	S1 Skill Score	MSLP	26.5	26.5	0
		850 HGHT	24.0	24.0	0
		500 HGHT	18.0	19.0	-1.0
		250 HGHT	13.0	13.5	-0.5
	BIAS	MSLP	0.7	0.5	0.2
		850 HGHT	5.1	3.1	2.0
		500 HGHT	-6.5	-5.9	0.6
		250 HGHT	-5.5	-8.0	-2.5
	RMS	MSLP	2.5	2.3	0.2
		850 HGHT	19.3	18.1	1.2
		500 HGHT	25.8	26.0	-0.2
		250 HGHT	33.3	33.3	0
+48HRS	S1 Skill Score	MSLP	32.0	32.0	0
		850 HGHT	29.5	29.5	0
		500 HGHT	22.5	23.0	-0.5
		250 HGHT	17.0	17.0	0
	BIAS	MSLP	1.0	0.6	0.4
		850 HGHT	6.9	4.1	2.8
		500 HGHT	-6.4	-5.4	1.0
		250 HGHT	-8.7	-8.5	0.2
	RMS	MSLP	3.1	2.9	0.2
		850 HGHT	24.5	23.1	1.4
		500 HGHT	33.0	33.0	0
		250 HGHT	43.3	43.3	0

 

a)

b)

Figure 2. Screen (2m) dry bulb temperature comparison between the previous LAPS_PT375 and the 1DVAR LAPS_PT375 errors averaged over all (approx. 297) available METAR sites and over all hourly model forecast periods out to +48 hrs for a) Bias and b) Root Mean Square(RMS) Error.

 

a)

 

b)

 

Figure 3. As for Figure 2 but for screen level (2m) dew point.

 

a)

 

b)

 

Figure 4. As for Figure 2 but for 10m wind speed.

 

Product Availability

Apart from the extended 72 hour forecast range, this upgrade should be essentially transparent to most users of LAPS_PT375 products. The only significant product modification is that the cloud amount fields produced from the model are now expressed as percentages (0-100%) rather than the fractional units (0.0-1.0) used previously. This change was made to be consistent with the cloud fields produced by the global GASP system.

DIFACS:

A limited number of extra charts are available on DIFACS to display the new +60 hr and +72 hr forecast periods available from the system. (The displayed fields were chosen on the basis of what is currently available for the +48 hr forecast). The relevant DIFACS slots are:

• 0147: Mean Sea Level Pressure/Thickness (Australian Region) +60 hr
• 0148: Mean Sea Level Pressure/Thickness (Australian Region) +72 hr
• 0159: Precipitation (Australian Region) +60 hr
• 0160: Precipitation (Australian Region) +72 hr
• 0161: Mean Sea Level Pressure (Full Domain) +72 hr
• 0408: Smoothed Mean Sea Level Pressure/Thickness (Australian Region) +60 hr
• 0409: Smoothed Mean Sea Level Pressure/Thickness (Australian Region) +72 hr

It is noted that DIFACS charts are also available through the bureau's internal web.

Any requests concerning DIFACS should be sent by email to: difacs@bom.gov.au

 

REAL-TIME DATABASE (rtdb2):

Data available in the real-time database (rtdb2) is essentially the same as was described in APOB No 46 except for the addition of the extra forecast periods out to +72 hours (+51, +54, +57, +60, +63, +66, +69, +72 hours for the low-resolution 0.75 degree 110x160 grids, and 1-hourly from +49 to +72 hours for the full-resolution 0.375 degree 220x360 single level field grids).

 

MCIDAS & KENNY:

1DVAR LAPS_PT375 fields can be accessed from the real-time database (rtdb2) by using MCIDAS ADDE. The KENNY visualization program is also being modified to allow ADDE access and this method will give access to significantly more data, e.g. 11 levels below .85 as opposed to the 5 levels that are currently accessible. However, to facilitate access to the extra forecast periods out to +72 hrs in a consistent manner for those still relying on existing methods, some additional grids have been added to extend the existing LAPS_PT375 "13**" grid series.  Thus, GRID1311-1314 (pressure) and GRID1331-1334 (sigma) contain the extra forecast fields out to 72 hours in 6 hourly increments.

 

WEB:

A limited number of colour gif charts have been added to the internal LAPS_PT375 web page. Currently only MSLP, rainfall, surface wind and temperature charts are available, but this may be expanded in the next few months.

 

ARCHIVES:

Data from LAPS_PT375 will continue to be archived on SAM-FS with the same name format as before:

Observational boxfiles:	/samnmc/regbox/yyyy/regbox.laps_pt375.mmdd.hh.n2.ieeeseq
Boundary files:	/samnmc/regprep/yyyy/regprep.laps_pt375.mmdd.hh.ncjj.nc
Climate files:	/samnmc/regprep/yyyy/climdata.laps_pt375.mmdd.hh.nc
Output boxfiles:	/samnmc/regbox/yyyy/regbox.laps_pt375.box1_out_n_mmdd.hh.nec
Analysis output:	/samnmc/reganal/yyyy/reganal.laps_pt375.mmdd.hh.nn.nc
Model output:	/samnmc/regprog/yyyy/regprog.laps_pt375.mmdd.hh.0ii_000. nn.nc
A new archive addition is the extended output boxfiles from the 1DVAR assimilation steps:
	/samnmc/regbox/yyyy/regbox.laps_pt375.extbox_out_ n_mmdd.hh.nec

Variables in the names above are:

yyyy	4 digit year, i.e. 2002
mmdd	4 digit month and day, i.e. 0805 is 5th August
hh	2 digit hour of model/analysis run, either 06, 12, 18 or 00
ii	2 digit hour of model output, 00 to 72
jj	1 or 2 digit integer counter (1 to 15) for six-hourly boundary files
N	N=1,2 or 3 for the assimilation cycle, or 1bc forthe extra bias-correction preliminary analysis step

 

Future Developments

It is planned that the other operational LAPS systems run in NMOC (MESO_LAPS_PT125, MESO_LAPS_PT050 & TLAPS_PT375) will be upgraded in the near future.

There are also plans to soon commence ingesting LAPS_PT375 model output into the MARS archive system.

 

References

Harris B. A., Steinle P. and Paevere J., 2000, 'Use of ATOVS Radiances in the 1DVAR/GASP Data Assimilation System', Tech. Proc. ITSC-XI Budapest, Hungary 20-26 September 2000.

Viterbo, P., 1996: 'The representation of surface processes in general circulation models', PhD Thesis, University of Lisbon, 201 pp