Development and Testing of a New Cloud Analysis Package Using Radar, Satellite, and Surface Cloud Observations Within GSI for Initializing Rapid Refresh
Total Page:16
File Type:pdf, Size:1020Kb
Preprint, 18th Conference on Numerical Weather Prediction American Meteorological Society, Park City, UT P2.5 Development and Testing of a New Cloud Analysis Package using Radar, Satellite, and Surface Cloud Observations within GSI for Initializing Rapid Refresh Ming Hu1, Steve Weygandt3*, Ming Xue1,2, and Stan Benjamin3 1Center for Analysis and Prediction of Storms, University of Oklahoma 2School of Meteorology, University of Oklahoma 3Global System Division, Earth System Research Laboratory, NOAA 1. Introduction documented by Zhang (1999) and Brewster (2002). It was used with the WSR-88D data through frequent *The spin-up problem, which shows as the sig- intermittent assimilation cycles in several studies of nificant delay in the development of cloud and pre- tornadic thunderstorms at horizontal resolutions of 3 cipitation at the early stage of a model forecast, is a km or higher (Xue et al. 2003; Hu et al. 2006; Hu and critical problem faced by the short-range forecasts of Xue 2007a) and been recently applied to initializing aviation sensitive weather parameters and high- WRF also (Hu and Xue 2007b) Those studies clearly impact weather. The spin-up problem is due to the show that the cloud analysis procedure can effec- absence or improper initialization of the cloud and tively build up storms in the initial condition and precipitation systems and related thermodynamical therefore reduce the spin-up problem. and dynamical features in the initial condition and The RUC cloud analysis is used by the opera- therefore can be mitigated through improving the tional RUC run at the National Centers for Environ- analysis of such features, which include in-cloud mental Prediction (NCEP, Benjamin et al. 2004b). It temperature, moisture, and cloud and hydrometeor is formulated to update 5 fully cycled cloud (water fields. However, the cloud and hydrometeor fields are and ice) and precipitation (rain, snow and graupel) typically poorly analyzed by most analysis systems species. Observations used include GOES cloud-top even though surface METAR, satellite, radar, and data and surface cloud, visibility and current weather other observing systems provide a great deal of cloud information. Experimental versions of the RUC cloud and precipitation information. The spin-up problem is analysis run at GSD have also included 2D radar re- also responsible for the inability of numerical flectivity and lighting data (Benjamin et al. 2004b; weather prediction (NWP) models to beat extrapola- Weygandt et al. 2006a,b). The experiments show the tion-based nowcasting systems for very short-range use of the RUC cloud analysis improves the analysis precipitation forecasting (Wilson et al. 1998). and forecast of aviation weather sensitive elements. To initialize cloud and precipitation systems as More recently, a procedure for dynamically initializ- well as associated temperature and moisture fields ing ongoing precipitation systems based on national using surface, satellite, and radar observations, both radar reflectivity mosaic data has been developed for the Center for Analysis and Prediction of Storms the RUC and is in real-time testing (Benjamin et al. (CAPS) at the University of Oklahoma and the 2007; Weygandt and Benjamin 2007). Global Systems Division (GSD) of the NOAA Earth The frequently updated guidance produced by System Research Laboratory have developed semi- RUC (using the latest observations within a mesocale empirical cloud analysis packages within their analysis and prediction system) has been used heavily mesoscale numerical forecast systems, namely the for short-range forecast applications, mainly for avia- Advanced Regional Prediction System (ARPS, Xue tion, storm forecasting, and other transportation areas et al. 2000; Xue et al. 2001; Xue et al. 2003) of (Benjamin et al. 2006). Building upon this success, a CAPS and the Rapid Update Cycle (RUC, Benjamin new system, known as the Rapid Refresh (RR), is et al. 2004a; Benjamin et al. 2004c) of GSD, respec- being developed in GSD to replace RUC with a tively. WRF-based short-range forecast system. The new The ARPS cloud analysis has evolved from that RR is able to cover a larger area including Alaska, of the Local Analysis and Prediction System (LAPS, Canada, and Puerto Rico and use more high- Albers et al. 1996) with significant modifications frequency observations over the wider areas. In RR, NCEP Grid-point Statistical Interpolation (GSI, Wu et al. 2002) is being used to analyze conventional * Corresponding author address: Steve Weygandt, data and initialize one of the WRF-ARW cores. NOAA/GSD, R/E/GSD, 325 Broadway, Boulder, CO To improve the initialization of the cloud and 80305, [email protected] precipitation system in RR, CAPS and GSD have 1 collaborated to develop a generalized cloud analysis procedure within GSI. During this period, a case study of a Central Plains storm cluster on 23 May 2005 (Hu and Xue 2007b) was conducted to investi- gate the impact of the ARPS cloud analysis proce- dure with WSR-88D radar reflectivity data on the forecast of the storm cluster when it is used within GSI framework and with the Advanced Research WRF (WRF-ARW, Skamarock et al. 2005) as fore- cast model. After that, the ARPS and RUC cloud analysis packages were incorporated into a GSI framework and tested individually with a case of squall lines striking the central US on 13 March 2006. More recently, a prototype of the new general- ized cloud analysis procedure that combines the strengths of the both RUC (for stable clouds) and ARPS (for explicit deep convection) cloud analysis Fig. 1 Schematic diagram depicting the various modules packages has been developed to improve the analysis and options within the general cloud analysis solver. of both stable layer and convective cloud and precipi- tation systems over a large domain. observations and facilitate update of the background The components of the new generalized proce- fields (cloud building or clearing) only where the dure are shown in Fig. 1. The ingest of the cloud and observations warrant, the cloud observations are precipitation observations and the 1-h forecast cloud blended together and used to distinguish three classi- and hydrometeor fields (from the previous RR cycle) fications: 1) observed clear, 2) observed cloudy, 3) is followed by the stable cloud analysis solver and the clouds unknown from observations. This composite convective cloud analysis solver. Recognizing the observed cloud information field is then blended with different treatments of convection within numerical the background cloud information to produce an op- forecast models, the convective cloud package in- timal estimate of the 3D cloud and precipitation cludes a choice of modules: one for a model setup fields. In this section, we will introduce the main with parameterized convection and one for a model observations used in the current cloud analysis pro- setup with explicitly resolved convection. Consis- cedure with their characteristics related to the cloud tency between the cloud analysis packages with the analysis. model microphysics is also sought. This paper documents some details of the new 2.1 METAR DATA cloud analysis procedure and some verification ex- periments using the 13 March 2006 squall line case METAR data are typically generated once an under the RR configurations. Section 2 describes the hour and some of them come from an Automated observations used in the cloud analysis. In section 3, Surface Observing System located at airports, mili- the 13 March 2006 squall line case is briefly intro- tary bases and other sites and some are from aug- duced. The new cloud analysis procedure is described mented observations or from trained observers or in detail in section 4 with the analysis of testing ex- forecasters. On 1 June 1996, METAR replaced the periments. The impacts of the cloud analysis used in Surface Aviation Observation (SAO) and becomes assimilation cycles are briefly described in section 5 the primary observation code used in the United and a summary of the results is given in section 6. States to satisfy requirements for reporting surface meteorological data. 2. Cloud and Precipitation Observations A regular METAR contains a report of basic atmospheric elements such as wind, temperature, dew Many meteorological observations include either point, and barometric pressure together with weather direct cloud and precipitation elements or informa- information such as precipitation type and trend, tion related to them, but no one single observation or cloud height and cover, lightning, and visibility. even the complete set of available observations can The METAR data used in current cloud analysis fully describe the state of all ongoing cloud and pre- are from the operational NCEP BUFR data file in- cipitation systems. Furthermore, many of the cloud gested in GSD and include cloud amount and the observations provide a “one-way” look, where cloud height of cloud base for up to 3 layers, horizontal information above or below an observed cloud layer visibility, and current weather. Before the cloud is unknown. To accommodate this aspect of cloud analysis, METAR data are decoded and digitalized. 2 2.2 SATELLITE DATA 2.4 LIGHTNING DATA Satellites can provide a comprehensive view of Lightning ground stroke data from the National cloud systems on a scale not possible by other means, Lightning Detection Network (NLDN) can provide especially in the area that has limited human activi- thunderstorm information in areas without radar cov- ties like ocean and desert. But most of current satel- erage and are used as a proxy for reflectivity in the lite data contains only one level of cloud observation cloud analysis in regions where reflectivity data are and need to be used with other cloud observation to not available. generate a column of cloud. In this study, satellite cloud products from the NOAA National Environ- 3. March 13, 2006 Central US Squall Case mental Satellite, Data, and Information Service (NESDIS), which is included in the same NCEP From 15 UTC March 12 to 09 UTC March 13, BUFR file as METAR data, are used in the analysis.