NMOC Quarterly Summary April-June 2002

Operational Atmospheric Motion Vectors in Regional Assimilation
2002

Analyses & Numerical Prediction | About Products | Map/Image/Chart Archives

J. Le Marshall, R. Seecamp, P.K. Stewart, K. Puri and R. Bowen

Introduction

The impact of the operational use of locally generated Atmospheric Motion Vectors (AMVs) on Australian Region forecasts has been measured. The motivation for this study was to gauge the impact the winds had on the operational system in the current GMS-5 viewing regime, namely, no Southern Hemisphere hourly images and limited southern extent for nearly all images. It has been found that, despite the reduction in spatial and temporal coverage of the GMS?5 imagery over the Southern Hemisphere, there is still beneficial impact from the derived AMVs.

Background

The Australian Bureau of Meteorology, has been using locally generated AMVs in the operational regional forecast system since 1992. The beneficial impact of these winds on operational forecasts is well documented. In September and October 2000, the impact of hourly infrared, low resolution visible and high resolution visible image based atmospheric motion vectors, in addition to low resolution visible, high resolution visible and water vapour image based vectors, available at 05, 11, 17 and 23 UTC, on the operational regional forecast system was measured. It was found that the addition of these winds to the operational Limited Area Prediction System (LAPS) which already contained locally generated IR channel 1 winds provided, on average, at all levels measured, a modest but consistent improvement (Aust. Meteor. Mag. Sept. 2002). In this study, we have briefly looked at the impact of these additional winds in a GMS-5 viewing regime, which now precludes the generation of hourly winds.

The Winds

a. Generation

The winds used in this study have been generated by tracking features in 3 images, taken with a separation of 30 minutes. The features were tracked automatically, using a first guess field to provide an initial indication of the search area and also to provide a temperature profile to be used in height assignment of the vectors. In the case of visible winds, the IR1 (11 µm) and IR2 (12 µm) channels were used to determine the cloud top for the upper level vectors and the cloud base for the low level vectors, with the wind assignment being to those levels for upper and lower level vectors respectively. For middle level vectors, an algorithm has been developed so that the assignment moves from the cloud top to the cloud base with reduction in altitude. Upper level vectors in the water vapour imagery have been assigned to the cloud top and in the case of middle level, clear air vectors the tracer temperature has been used in height assignment. Table 1 provides a list of the vectors generated routinely in the Bureau. Hourly winds are only produced over the Northern Hemisphere. Figure 1 and Figure 2 give examples of winds generated routinely in the Bureau. The first provides an indication of the coverage available from cloud and water vapour motion vectors over the Tasman Sea near 23 UTC on 24 July 2002. Figure 2 provides an indication of the atmospheric motion vectors around tropical cyclones Feng Shen and Fung Wong in the NW Pacific near 0500 UTC 25 July 2002.

Table 1 - Cloud Drift Wind types generated routinely in the Bureau. The table indicates type, image resolution, frequency of wind extraction and the separation of the image triplets used for wind generation (ΔT). Note: Hourly AMVs are now only generated for the Northern Hemisphere.

	Fig. 1 GMS-5 cloud and water vapour motion vectors over the tasman Sea near 23 UTC 24 July 2002
	Fig. 2 AMVs around tropical cyclones Feng Shen and Fung Wong in the NW Pacific around 0500 UTC 25 July 2002

b. Accuracy and Quality Control

The winds undergo rigorous quality control and are flagged with the local error indicator as well as the international Quality Indicator. Table 2 contains a comparison of radiosondes and atmospheric motion vectors within 150 km of each other over the Australian Region for the period March to June 2002 inclusive. The Mean Magnitudes of Vector Differences indicate error levels consistent with those assigned to the vectors in the operational Regional Analysis Scheme. Currently, the QI appended to the vectors is not used in the operational system for data selection but rather the local system, which is more effective for selecting vectors at a given error level is employed. In the longer term, it is planned to employ the Quality Indicator, although attention has been paid to generating a direct estimate of error with each vector. This appears to be the most effective approach for data selection and use of the AMVs in the operational analysis.

Table 2 - Comparison of radiosonde and atmospheric motion vectors within 150 km radius over the Australian Region, March to June 2002 inclusive. [IR1 = 11 µm imagery based winds, VIS = Low resolution (5 km) visible winds, HR VIS = High resolution (1.25 km) visible winds, WV = Water Vapour based winds and MMVD = mean magnitude of vector difference (ms^-1)]

The Real Time Assimilation Study

The data assimilation study employed the operational LAPS system and the operational data base as the control forecast. In parallel, using the same assimilation system, the local atmospheric motion vectors were added to the operational data base for real time assimilation runs. S1 skill scores of the 24-hour forecasts, tabulated on the NMOC verification grid, from the local AMVs (LAPS + IR) and the matching control forecasts are shown in Table 3. The impact of the winds on S1 skill scores during the same period are also indicated Table 3. Overall, it can be seen that the local AMVs have a small but consistent positive impact on the forecasts and are still contributing to the operational data base. It also should be noted that the operational data base already contains locally generated IR1 vectors which have also been shown to have positive impact on 24-hour regional forecasts.

Table 3 Skill scores for 24-hour forecasts for Operations (Ops. S1) and Operations plus AMVs (Ops + AMVs S1), 3 May to 14 June 2002

Data Access

All winds are currently available through the ADDE server on Gale.

Visualisation tools such as those to be available in VISAD through AIFS will allow 3D depiction, as in Figure 3 where AMVs over the Tasman Sea are shown over an image at 05 UTC on 21 August 2002.

Fig. 3 VISAD image of low level AMVs over the Tasman Sea 05 UTC 21 Aug. 2002

Summary and Conclusions

Local estimation of real time operational AMVs and their impact on local NWP has been further documented. Experiments, using these data, individually and together in an operational NWP trial when GMS provided more complete coverage of the Southern Hemisphere in both space and time have been previously published. The results presented in this article show the benefit of these data to the operational forecast system, even in a restricted GMS-5 viewing regime. In essence, they are still contributing to the operational data base used by the Bureau.

Looking ahead, the continuing trend towards space-based observations with higher spatial, spectral and temporal resolution should allow more accurate estimation of atmospheric motion vectors and result in quantitative benefits to NWP. In particular, the prospects of benefits from the use of sequential observations for MTSat 1R and Feng Yung 2 and new generation ultraspectral measurements such as the Geostationary Imaging Fourier Transform Spectrometer (GIFTS) appear to be very good.

Acknowledgements