Sensitivity of simulating Typhoon Haiyan (2013) using WRF: the role of cumulus convection, surface flux parameterizations, spectral nudging, and initial and boundary conditions

Abstract

Abstract. Typhoon (TY) Haiyan was one of the most intense and highly destructive tropical cyclones (TCs) to affect the Philippines. As such, it is regarded as a baseline for extreme TC hazards. Improving the simulation of such TCs will not only improve the forecasting of intense TCs but will also be essential in understanding the potential sensitivity of future intense TCs with climate change. In this study, we investigate the effects of model configuration in simulating TY Haiyan using the Weather Research Forecasting (WRF) Model. Sensitivity experiments were conducted by systematically altering the choice of cumulus schemes, surface flux options, and spectral nudging. In addition to using the European Centre for Medium-Range Weather Forecasts Reanalysis fifth-generation (ERA5) single high-resolution realization as initial and boundary conditions, we also used 4 of the 10 lower-resolution ERA5 data assimilation system (EDA) ensemble members as initial and boundary conditions. Results indicate a high level of sensitivity to cumulus schemes, with a trade-off between using Kain–Fritsch and Tiedtke schemes that have not been mentioned in past studies of TCs in the Philippines. The Tiedtke scheme simulates the track better (with a lower mean direct positional error, DPE, of 33 km), while the Kain–Fritsch scheme produces stronger intensities (by 15 hPa minimum sea level pressure). Spectral nudging also resulted in a reduction in the mean DPE by 20 km, and varying the surface flux options resulted in the improvement of the simulated maximum sustained winds by up to 10 m s−1. Simulations using the EDA members initial and boundary conditions revealed low sensitivity to the initial and boundary conditions, having less spread than the simulations using different parameterization schemes. We highlight the advantage of using an ensemble of cumulus parameterizations to take into account the uncertainty in the track and intensity of simulating intense tropical cyclones.

Future projections of hurricane intensity in the southeastern U.S.: sensitivity to different Pseudo-Global Warming methods

“…However, the focus of this study is to specifically extract the uncertainties in the simulated hurricane intensity resulting solely from the PGW set up, which may add to the general uncertainty level when simulating tropical cyclones. Therefore, we adopted an established WRF set up that has proven to deliver acceptable results in the context of hurricane simulations, in detail described by Delfino et al (2022Delfino et al ( , 2023.…”

Impact of Aerosols on NPP in Basins: Case Study of WRF−Solar in the Jinghe River Basin

“…WRF is a mesoscale weather forecast model developed by the National Center for Atmospheric Research of the United States. This model has been successfully applied to data assimilation research [29,30], air quality modeling [31,32], and regional climate simulations, such as short−term rainfall forecasts [33] and typhoon simulation [34]. The WRF was dynamically scaled down using region nesting to improve the simulation resolution.…”

The Effect of the Cordillera Mountain Range on Tropical Cyclone Rainfall in the Northern Philippines

“…A summary of the options used is listed in Table 1. Choices of physical parameterisation schemes and model configuration were adapted from the operational forecasting configurations of the Philippine weather bureau, Philippine Atmospheric, Geophysical and Astronomical Services Administration [24], optimisation studies by Tolentino and Bagtasa (2021) [25], historical comparisons by Bagtasa (2021) [26], and sensitivity tests for TC simulations conducted by Delfino et al (2022) [27]. Additionally, we adapted the Planetary Boundary Layer scheme based on the study of Cruz and Narisma [28] due to its performance in simulating heavy precipitation over Luzon.…”

“…For the model runs, the "adaptive time-step" option of WRF was enabled for model stability and efficiency [23], and the model output was set at 3-hourly intervals. To achieve a balance between longer simulation periods to allow for spin up, TC development and maturation [27] and shorter simulation periods for higher precipitation accuracy [30], simulations were initialised at 96 h before TC landfall and ended at 24 h after landfall. This allowed for a simulation period of 5 days: 48 h spin up time, 48 h pre-landfall TC development, and 24 h post-landfall analysis.…”