Improved Prediction of Bay of Bengal Tropical Cyclones Through Assimilation of Doppler Weather Radar Observations
Total Page:16
File Type:pdf, Size:1020Kb
NOVEMBER 2015 O S U R I E T A L . 4533 Improved Prediction of Bay of Bengal Tropical Cyclones through Assimilation of Doppler Weather Radar Observations KRISHNA K. OSURI AND U. C. MOHANTY School of Earth, Ocean and Climate Sciences, Indian Institute of Technology, Bhubaneswar, India A. ROUTRAY National Centre for Medium Range Weather Forecasting, Noida, India DEV NIYOGI Purdue University, West Lafayette, Indiana (Manuscript received 27 November 2013, in final form 16 March 2015) ABSTRACT The impact on tropical cyclone (TC) prediction from assimilating Doppler weather radar (DWR) obser- vations obtained from the TC inner core and environment over the Bay of Bengal (BoB) is studied. A set of three operationally relevant numerical experiments were conducted for 24 forecast cases involving 5 unique severe/very severe BoB cyclones: Sidr (2007), Aila (2009), Laila (2010), Jal (2010), and Thane (2011). The first experiment (CNTL) used the NCEP FNL analyses for model initial and boundary conditions. In the second experiment [Global Telecommunication System (GTS)], the GTS observations were assimilated into the model initial condition while the third experiment (DWR) used DWR with GTS observations. Assimilation of the TC environment from DWR improved track prediction by 32%–53% for the 12–72-h forecast over the CNTL run and by 5%–25% over GTS and was consistently skillful. More gains were seen in intensity, track, and structure by assimilating inner-core DWR observations as they provided more realistic initial organization/ asymmetry and strength of the TC vortex. Additional experiments were conducted to assess the role of warm- rain and ice-phase microphysics to assimilate DWR reflectivity observations. Results indicate that the ice- phase microphysics has a dominant impact on inner-core reflectivity assimilation and in modifying the intensity evolution, hydrometeors, and warm core structure, leading to improved rainfall prediction. This study helps provide a baseline for the credibility of an observational network and assist with the transfer of research to operations over the India monsoon region. 1. Introduction represent the realistic position, strength, and structure of TC vortices. Osuri et al. (2013) analyzed forecasts of 100 Tropical cyclones (TCs) are one of the most hazard- TC cases over the NIO, and concluded that the average ous weather events affecting the Indian monsoon region. initial position and initial intensity errors are about The devastation from TCs is particularly notable over 2 57 km and 8–10 m s 1, respectively. These errors prop- heavily populated, long, and low-lying coastal regions of agate and increase with time and limit the predictive skill the Bay of Bengal (BoB). of high-resolution mesoscale models for TC movement, Conventional observations are scarce over the north intensification, and decay (Elsberry et al. 2007; Osuri Indian Ocean (NIO) where the BoB TCs form and et al. 2013). As a result, a number of recent efforts over evolve over the deep ocean/sea. Initial conditions de- the NIO basin have been directed to assess the impact of rived from coarser-resolution global analyses cannot assimilating remotely sensed data to improve the posi- tion and structure of the initial vortex and consequently the forecast of TCs (Mandal and Mohanty 2006; Singh Corresponding author address: U. C. Mohanty, School of Earth, Ocean and Climate Sciences, Indian Institute of Technology, et al. 2008; Xavier et al. 2008; Osuri et al. 2012a). Toshali Bhavan, Satya Nagar, Bhubaneswar, 751007, India. Responding to the need for severe weather monitor- E-mail: [email protected] ing and prediction associated with a landfalling TC, the DOI: 10.1175/MWR-D-13-00381.1 Ó 2015 American Meteorological Society Unauthenticated | Downloaded 09/28/21 03:35 PM UTC 4534 MONTHLY WEATHER REVIEW VOLUME 143 India Meteorological Department (IMD) extended the automated weather stations (AWS), satellite-derived winds, Doppler Weather Radar (DWR) network along the east and aircraft observations} on the intensity and spatial coast. The DWRs can provide a high-resolution, spa- rainfall distribution of landfalling TC. These studies over tiotemporal perspective of landfalling TCs (Marks the Indian monsoon domain have focused primarily on 2003). The DWR-based high-resolution reflectivity and assimilating DWR observations only within the inner- radial velocity fields are important for weather analysis core region, and the impact of assimilating the DWR and forecasting at meso- and microscales. When the TC observations of the TC environment and its interaction is within the DWR’s range, the reflectivity and velocity with TC evolution has not been addressed. fields can provide significant details of the TC eyewall, In this study, the objective is to assess how important inner-core winds, and hydrometeor structures (Marks are radar observations in defining (or correcting) TC en- and Shay 1998; Marks 2003). When the TC is out of the vironment, and the TC inner-core region, and in turn DWR range, it can still provide potentially useful in- impacting the TC forecasts. The DWR observations formation about the TC environment. include both radar reflectivity and radial wind impact, Past studies, principally over the United States and for while the TC environment refers to a region where the landfalling TCs in the Atlantic basin (e.g., Xiao and Sun TC center is away from the radar coverage and the TC 2007; Zhao and Xue 2009), have shown that high- outer bands are detected. resolution radar inner-core observations have an over- In general, S-band weather radar observes hydrome- all positive impact on analysis and prediction. Studies teors including rain, hail, and snow. Hail in particular such as Gao et al. (1999), Xiao et al. (2005), and Zhao can have a very strong reflectivity signal. By assuming and Jin (2008) used the three-dimensional variational warm-rain processes in reflectivity assimilation, the data assimilation (3DVAR) systems in the framework of contribution of hail and snow, which exist in tropical various regional/mesoscale models, such as the ARPS cyclone systems, is ignored but can have a large impact model, the Coupled Ocean–Atmosphere Mesoscale on TC intensity forecast (Houze 2010; Zhu and Zhang Prediction System (COAMPS), and the fifth-generation 2006). Therefore, in addition to the above objective, the Pennsylvania State University–National Center for At- assimilation of reflectivity observations in the TC inner- mospheric Research Mesoscale Model (MM5), and core with different (warm rain and ice phase) micro- showed improvement in the prediction of hurricane physics parameterization schemes is also investigated. track, intensity, and structure. These studies demon- The study investigates the necessity and potential con- strated that radial velocity assimilation can improve tribution of DWR observations for consistent and im- track forecasts, while the assimilation of reflectivity data proved prediction of TC track, intensity, and structure can help improve intensity forecasts. The assimilation of over the Bay of Bengal. radial velocity and reflectivity combined help improve track and intensity predictions as compared to the in- 2. Methodology and numerical experiments dividual data alone. The assimilation of DWR observa- tions (both ground and airborne based) using the ensemble The Advanced Research version of the Weather Re- Kalman filter (EnKF) is a recent effort (although com- search and Forecasting (WRF) Model (ARW) V3.3 putationally expensive) to improve the predictions of with a single domain covering 38–288N and 758–1058E, at hurricanes (Aksoy et al. 2012, 2013) and severe con- 9-km horizontal resolution and 51 vertical levels was vection (Dowell et al. 2011). used for this study. The model followed Arakawa C-grid Over the Indian monsoon region, studies with DWR staggering and the model integration time step was 30 s. data assimilation are still evolving. Kiran Prasad et al. The model physics used include the Kain–Fritsch con- (2014) showed the positive impact of assimilating DWR vection scheme, the Yonsei University (YSU) planetary reflectivity and radial wind observations on the simula- boundary layer (PBL) scheme, the WRF single-moment tion of severe thunderstorms over northeastern parts of 3-class (WSM3) microphysics scheme, the Monin– India. Routray et al. (2010, 2013) noted that the assim- Obukhov similarity scheme, the thermal diffusion land ilation of DWR radial velocity and reflectivity helped surface scheme, and the Rapid Radiative Transfer improve predictions associated with monsoon depres- Model (RRTM) for longwave and Goddard for short- sion forecasts for the Indian region. Govindankutty wave atmospheric radiation schemes. Osuri et al. (2012b) et al. (2010) demonstrated the positive impact of DWR identified the above parameterization choices to be better radial wind and the Global Telecommunication System for TC track and intensity predictions for the Bay of (GTS) data {which includes radiosonde atmospheric Bengal domain. The details of the model physics, dy- profiles [the radiosonde/radiowind (RS/RW) and the pilot namics, and description of the model equations can be balloon (pibal)], surface synoptic observations (SYNOP), found in Skamarock et al. (2005). Unauthenticated | Downloaded 09/28/21 03:35 PM UTC NOVEMBER 2015 O S U R I E T A L . 4535 A 3DVAR is available within the ARW modeling Note that all of the strong cyclones (.48 kt) during system (WRF-Var). This variational assimilation system 2007–11 that were monitored by the DWR are consid- blends in the observations with the global model analysis ered in this study; however, some of the marginal cy- through an iterative solution of prescribed cost function clones (,47 kt) are not included. The synoptic situations [see details in Barker et al. (2004)]. The cost function and detailed characteristics of each cyclone were ob- J(x) is described as tained from the Regional Specialized Meteorological Centre reports (Regional Specialized Meteorological 1 2 Centre 2007, 2009, 2010, 2011) from IMD, New Delhi, J(x) 5 Jb 1 Jo 5 (x 2 xb)TB 1(x 2 xb) 2 India.