Improvement of forecast skill for severe radar-based extrapolation and storm-sca
Gaili Wang ⁎, Wai-Kin Wong c,1 a State Key Laboratory of Severe Weather, Chinese Academ b Jiangsu Institute of Meteorological Science, 16 Kunlun Lo c Hong Kong Observatory, 134A, Nathan Road, Kowloon 9 d School of Civil Engineering and Environmental, Universit e Department of Hydraulic Engineering, Tsinghua Universi f d Predi
Article history: hyperbolic tangent weight scheme. The comparison of forecast skill betweenMTaRE and ARPS in spatial and temporal
Atmospheric Research 154 (2015) 14–24
Contents lists available at ScienceDirect
Atmospheric j ourna l homepage: www.e lsresolutions for the next few hours (nowcasting) plays an important role in severe rainfall prediction, meteorological disaster warnings, and the meteorological services of major
E-mail addresses: firstname.lastname@example.org (G. Wang), email@example.com (W.-K. Wong), firstname.lastname@example.org (Y. Hong), email@example.com (L. Liu), Jili.Dong@noaa.gov (J. Dong), firstname.lastname@example.org. Introduction
Deterministic forecasting with high ⁎ Corresponding author at: State Key Laboratory of SevereWeather, Chinese
Academy of Meteorological Science, Beijing, 46 Zhongguancun South Street,
Haidian District, Beijing 100081, China. Tel.: +86 10 5899 3540.high spatial resolution of 0.01° × 0.01° and high temporal resolution of 5min showed thatMTaRE outperformed ARPS in terms of index of agreement andmean absolute error (MAE).MTaRE had a better Critical Success Index (CSI) for less than 20-min lead times andwas comparable to ARPS for 20- to 50-min lead times, while ARPS had a better CSI for more than 50-min lead times. Bias correction significantly improved ARPS forecasts in terms of MAE and index of agreement, although the CSI of corrected ARPS forecasts was similar to that of the uncorrected ARPS forecasts.
Moreover, optimally merging results using hyperbolic tangent weight scheme further improved the forecast accuracy and became more stable. © 2014 Elsevier B.V. All rights reserved.
NWP forecasts(M. Xue). 1 Tel.: +852 2926 8642. 2 Tel.: +1 405 325 3644. 3 Tel.: +86 10 5899 3540. 4 Tel.: +1 405 325 3502. http://dx.doi.org/10.1016/j.atmosres.2014.10.021 0169-8095/© 2014 Elsevier B.V. All rights reserved.the bias corrections were performed to improve the forecast accuracy of ARPS forecasts. Finally, the corrected ARPS forecast and radar-based extrapolation were optimally merged by using aReceived 18 June 2014
Received in revised form 16 October 2014
Accepted 28 October 2014
Available online 6 November 2014, Yang Hongd,e,2, Liping Liu a,3, Jili Dong f,4, Ming Xue f,4 y of Meteorological Science, 46 Zhongguancun South Street, Haidian District, Beijing 100081, China ad, Nanjing, Jiangsu 210009, China 99077, Hong Kong, China y of Oklahoma, 120 David L. Boren Blvd., National Weather Center Rm. 4610, Norman, OK 73072, USA ty, Beijing, 100084, China ction of Storms, University of Oklahoma, 120 David L. Boren Blvd., National Weather Center Rm. 4610, Norman, a b s t r a c t
The primary objective of this study is to improve the performance of deterministic high resolution rainfall forecasts caused by severe storms bymerging an extrapolation radar-based schemewith a storm-scale Numerical Weather Prediction (NWP) model. Effectiveness of Multi-scale Tracking and Forecasting Radar Echoes (MTaRE) model was compared with that of a storm-scale NWP model named Advanced Regional Prediction System (ARPS) for forecasting a violent tornado event that developed over parts of western andmuch of central Oklahoma onMay 24, 2011. ThenSchool of Meteorology, and Center for Analysis an
OK 73072, USA a r t i c l e i n f oa,b,corrected forecastweather by merging le NWP
Research ev ie r .com/ locate /atmossports events. Three primary nowcasting methods are used operationally.
The first group includes a number of techniques that rely on extrapolation of radar images, and is widely applied in operational nowcasting systems such as the Auto-Nowcast
System (ANC; Mueller et al., 2003) developed by the National
Center Atmosphere Research and the McGill Algorithm for
PrecipitationNowcasting by Lagrangian Extrapolation (MAPLE;
Turner et al., 2004) used atMcGill. Extrapolation techniques are divided into pixel-based and object-based approaches (Zahraei et al., 2012). The pixel-based technique extrapolates radar reflectivity observations using motion estimation from two consecutive radar images (Rinehart and Garvey, 1978; Li et al., 1995; Grecu and Krajewski, 2000; Germann and Zawadzki, 2002; Zahraei et al., 2012;Wang et al., 2013; Sokol et al., 2013).
The object-based technique identifies 3D convective cells, tracks, and forecasts storm-related parameters assuming linear trends (Dixon and Wiener, 1993; Johnson et al., 1998; Hong et al., 2004; Vila et al., 2008; Zahraei et al., 2013).
The second group consists of storm-scale NWP models.
Recently, the “spin-up” problem of NWP models was reduced significantly using the rapid-update-cycle (RUC) approach. 15G. Wang et al. / Atmospheric Research 154 (2015) 14–24The High-Resolution Rapid Refresh (HRRR; Zahraei et al., 2012) developed by the National Oceanic and Atmospheric
Administration, and ARPS used at the Center for Analysis and
Prediction of Storms (CAPS) are the outstanding representatives of the second group. The forecast accuracy at the first several hours has been improved significantly by assimilating various types of observation data (Macpherson, 2001;
Weygandt et al., 2002; Benjamin et al., 2004; Caya et al., 2005; Tong and Xue, 2005; Sokol, 2007; Sokol and Pesice, 2012; Wong et al., 2009; Zahraei et al., 2012).
Predictive accuracy of radar-based extrapolation rapidly decreases within the first several hours of severe weather development, because the growth and decay of storms are not taken into account. Extrapolation of observations is most likely more accurate in the shorter terms (Austin et al., 1987; Golding, 1998; Lin et al., 2005;Wong et al., 2009; Zahraei et al., 2012), as shown in Fig. 1. On the other hand, some comparisons show that the NWP models outperform radar-based extrapolation methods over longer time scales as they dynamically resolve large-scale flow. However, they may not produce optimal predictions at the first short-term, because they are sensitive to the initial condition, spatial resolution, and assimilation data (Golding, 1998; Ganguly and Bras, 2003; Lin et al., 2005;