Even today the understanding of these processes is still lacking. Our goal is to use observations and models to advance this understanding.
Hurricane forecast models have improved significantly over the last couple of years, due to the new understanding we have obtained using satellite observations and airborne data from dedicated field campaigns. However, deterministic forecasts will always have limitations due to the uncertainty in the representation of the physical processes and the uncertain initial conditions. One of the most crucial improvements that have enabled the more realistic modeling of tropical cyclones is the use of high-resolution ensembles.
Indeed, in our study we adopt the following approach: i) generate an ensemble of high resolution model forecasts using the state-of-the-art HWRF model; ii) use instrument simulators to produce synthetic data (e.g. brightness temperature and radar reflectivity) from the model fields for direct comparison to satellite observations; iii) develop metrics to allow us to sub-select the realistic members of ensemble forecasts, based on objective measures of the similarity between the observed and the forecasted structure of the storms and their environment; iv) for the ensemble members that are most consistent with the satellite observations: determine the skill in forecasting storm genesis/evolution/intensity/tracks to provide“guidance on guidance”; v) use the ensemble members with the best predictive skill to untangle the complex multi-scale interactions and to understand the interplay between the environmental controls and the convective processes.
In this paper we will report on the first three goals of our research, using the forecasts and the observations of hurricane Edouard (2014). Our Edouard ensemble forecast includes members of three different groups: The first group represents uncertainty in the initial conditions. It is driven by GEFS members and it produces the largest spread in the intensity change, ranging from non-intensifying to rapidly-intensifying, but the ensemble members generally follow similar track. The second group focuses on uncertainty in the model physics regarding boundary layer processes and turbulent mixing. Members are generated by adding perturbations to the PBL height while all other conditions are held the same. The third group addresses uncertainty in the model physics regarding the parameterization of cumulus convection. Ensemble members are generated by adding perturbations to the trigger function in SAS scheme. Neither the PBL nor SAS ensembles runs showed a big spread in the intensity and intensity change. Our next efforts are directed toward addressing the uncertainty related to the modeling of the microphysical processes, and, in particular, the assumptions about the properties of the hydrometeors. We will address that by perturbing to the hydrometeor density, such as snow and graupel density, to study the impact on the structure and intensity of the forecasted storm.
After describing the generation of the ensemble forecast, we will briefly describe the challenges in producing synthetic data (satellite observables) and the choices we have made. By looking at the multi-channel joint distributions of microwave brightness distributions, we will demonstrate that simulated brightness temperatures compare reasonably well in a statistical sense to the observed.
We will then describe our approach to developing metrics for the selection of the most appropriate ensemble members. To identify good members, we need to make sure both the environment and the vortex itself are well reproduced in the ensemble members. Regarding the environment, most previous research has relied on SHIPS index. However, SHIPS index is an oversimplified parameter which doesn't take into account of the inhomogeneity of the environment and asymmetric structure of the vortex. In a recently accepted manuscript (Chen et al., 2017), we proposed to refine the SHIPS index by including environmental factors in a shear-relative and storm-motion-relative sense. We can also stratify the environment by radius, such as near-core environment and far filed environment. For the inner core structure evaluation, we focus on convection azimuthal and radial distribution, as well as on analysis of the prevalence of shallow versus deep convection or that of azimuthally symmetric versus asymmetric.
The metrics we are developing are based on applying low-wave number analysis to the observed and forecasted 2D fields to develop objective criteria for consistency (e.g. Hristova-Veleva et al., 2016). We investigate the low-wavenumber cartoons of relative humidity, total precipitable water, precipitation structure, surface convergence and atmospheric temperature as determined from: AIRS, microwave sounders, scatterometer-derived surface winds (ASCAT/OceanScat) and passive microwave imagers (TMI, AMSR-E, AMSR-2, GMI, SSMIS).
We will end by presenting the results of the preliminary selection of most skillful members and will outline the goals of the future research, targeted toward analyzing the multi-scale interactions in these most realistic members.
The research described in this paper was performed at the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration (NASA).