An Analysis of Hurricane Edouard SST Sensitivity Runs By HWRF in a Neutral-Shear Environment

The association of ocean fluxes to tropical cyclone (TC) intensity, as well as intensity changes, is well-known. However, the timing of rapid intensity changes and steady-state conditions still presents forecasting challenges. In 2016, NHC validation analyses show the previous cycle operational HWRF, interpolated to the operation’s hour (HWFI), is now more skillful than statistical intensity guidance such as SHIPS and LGEM. Therefore, HWRF datasets may reliably provide insight into the intensity response to ocean forcing. We present results of the 2016 and operational version of HWRF to eight constant sea surface temperature (SST) configurations, and one coupled POM run, in a neutral shear environment (8 m/s<shear<14m/s). This is achieved using a control run for Hurricane Edouard (2014) with the GFS SST, then rerun with SST configurations from NCODA for 2014; RTOFS for 2014; GDEM climatology; and NCODA for 2010-2013. In all runs, the TC develops then approaches a steady-state. Intensity responses over the warmer and wider SST pools result in moderate-to-rapid intensification. These runs do not include coupling, and are a direct response to an unchanging warm pool configuration. Environmental conditions are favorable (and similar) for all runs until the end of the simulation, isolating the intensity response to the track over each different SST field when shear < 14 m/s.

Statistical assessments are performed for the 8 runs concatenated into one dataset. Surface pressure has more scatter than maximum sustained one-minute winds (V_max), hence the analysis is performed on V_max. All variables are available in 3-h segments. SST, sensible heat flux, latent heat flux, Convective Available Potential Energy (CAPE), surface relative humidity, and surface dewpoint temperature are determined at the radius of maximum winds (R_max) in compass 30-deg directions, and averaged using the 12 points. Filtered wind shear and precipitable water are determined at the 100-, 200-, 300-, 400-, and 500-km radii in compass 90-deg directions, and averaged using the 20 points. Wind shear is computed as the vectorial wind difference between 200- and 850-mb pressure levels.

Linear correlation results show V_max is correlated to SST with a variance explained of 59-61%, but higher correlations exist for the fluxes with a variance explained of 72-80%. However, the intensification cycle (defined as ΔV_max over a 24-h period) is mostly linearly correlated to CAPE with r=0.7 and variance explained of 49%. Generally, other correlations are either low for V_max, low for ΔV_max, or highly correlated with CAPE or sensible heat flux (multicollinearity).

Moreover, when a sigmoidal correlation relationship (“s-curve pattern”) is computed for sensible heat flux versus V_max, r=0.94 and the variance explained is 88%. Hence, for a favorable environmental environment in this HWRF environment of 8 cases, the TC slowly develops as it approaches the warm pool with a correspondence to heat flux, intensifies in response to a “CAPE boost” as it begins moving over SST of 27.5-28.5°C with an intensity change linearly proportional to CAPE, then approaches steady-state as CAPE is reduced partially due to the developing warm-core aloft. A simple model that uses an empirically derived heat flux sigmoidal equation and linear CAPE equation explains 75% of the variance. Implications of this concept towards understanding intensity change, facilitating Maximum Potential Intensity theory, and possible utilization in statistical intensity schemes such as SHIPS, LGEM, and the Rapid Intensification Index will be briefly discussed.

Wind shear calculations required removing the vortex influence of Edouard, which was a large storm. Using the traditional 2-3-4-2-5-6-7-2-8-9-2 Kurihara filter failed, but was remedied by expanding to a 2-3-4-2-5-6-7-2-8-9-2-10-11-2-12-13-2-3-2 filter. A response function analysis of both filters shows the new scheme is more suitable for higher-resolution models, as Kurihara originally tuned his scheme for a coarse model to remove ten-degree features. An interesting necessity is the need for an intermittent 2Δx pass or the scheme becomes unstable — a property not discussed in Kurihara’s original paper. As many more passes are required for the new technique, additional modifications to remove 3-5Δx noise may be required. Ultimately, as model resolution continues to improve, the original Kurihara scheme will become ineffective, and we will discuss necessary steps to remedy this problem.