Impact of precipitation assimilation on tropical cyclone structure in the North American Regional Reanalysis

Continued advancement in the realm of tropical cyclone (TC) forecasting requires a more accurate depiction of these storms at model initialization. This study examines the representation of TCs in the North American Regional Reanalysis (NARR) before and after the 2004 introduction of precipitation assimilation over ocean in the vicinity of TCs. Coinciding with this transition, there is a statistically significant break point in TC rainfall forecast skill, as measured versus rainfall rates derived by radars (for locations over land) and satellites (for locations over ocean). The probability distribution function of rainfall rates indicates that light (heavy) precipitation is over-forecast (under-forecast) by NARR in the early time period. Mesoscale features, such as spiral outer rainbands, are also more skillfully forecast in post-2004 storms. Since the precipitation assimilation is applied through an adjustment to the latent heating distribution, the data assimilation system in the later time period initializes a low-level moisture and heating profile that is more conducive to the initiation of deep convection and the generation of precipitation. Consequently, the deep convection and enhanced latent heat release lead to a more robust warm core temperature perturbation and a better developed secondary circulation, which supplies the TC with a larger volume of moisture that is sourced from the large-scale environment. Additionally, the evolution of TC size, which was objectively estimated though the radius of outermost closed isobar (ROCI), is significantly more skillful in post-2004 storms. Based on this study, precipitation assimilation leads to a better analysis of temperature, winds, and moisture in the vicinity of TCs, resulting in an improved representation of the water budget and the storm lifecycle. Therefore, we conclude that efforts towards the development of precipitation assimilation techniques from radar and satellite datasets will be valuable toward the construction of improved TC forecasting tools.