Evaluating the Impacts of COSMIC-2 GNSS RO Bending Angle Assimilation on Atlantic Hurricane Forecasts Using the HWRF Model

William J. Miller aCooperative Institute for Satellite Earth System Studies, Earth System Science Interdisciplinary Center, University of Maryland, College Park, College Park, Maryland

Search for other papers by William J. Miller in
Current site
Google Scholar
PubMed
Close
,
Yong Chen bNOAA/National Environmental Satellite, Data, and Information Service, Center for Satellite Applications and Research, College Park, Maryland

Search for other papers by Yong Chen in
Current site
Google Scholar
PubMed
Close
,
Shu-Peng Ho bNOAA/National Environmental Satellite, Data, and Information Service, Center for Satellite Applications and Research, College Park, Maryland

Search for other papers by Shu-Peng Ho in
Current site
Google Scholar
PubMed
Close
, and
Xi Shao aCooperative Institute for Satellite Earth System Studies, Earth System Science Interdisciplinary Center, University of Maryland, College Park, College Park, Maryland

Search for other papers by Xi Shao in
Current site
Google Scholar
PubMed
Close
Free access

Abstract

This study evaluates the impact of assimilating Global Navigation Satellite System (GNSS) radio occultation (RO) bending angles from Formosa Satellite Mission-7/Constellation Observing System for Meteorology, Ionosphere, and Climate-2 (COSMIC-2) receiver satellites on Hurricane Weather Research and Forecasting (HWRF) Model tropical cyclone (TC) forecasts. Launched in June 2019, the COSMIC-2 mission provides significantly higher tropics data coverage compared to its predecessor COSMIC constellation. GNSS RO measurements yield information about atmospheric pressure, temperature, and water vapor profiles. HWRF is cycled with and without COSMIC-2 bending angle data assimilation for six 2020 Atlantic hurricane cases. COSMIC-2 assimilation has little impact on HWRF track forecasts, consistent with HWRF’s design limiting cycled data assimilation impacts on surrounding large-scale flows; however, COSMIC-2 assimilation results in a statistically significant ∼8%–12% mean absolute forecast error reduction in minimum central sea level pressure for t = 36-, 54-, 60-, and 108–120-h lead times. Forecasts initialized from analyses assimilating COSMIC-2 observations also have a 1%–4% smaller 600–700-hPa specific humidity (SPFH) root-mean-squared deviation compared to radiosondes and dropwindsondes for most lead times. While not all HWRF intensity forecasts benefit from COSMIC-2 assimilation, a few show notable improvement. For example, assimilating two COSMIC-2 profiles within the inner core of developing Hurricane Hanna (2020) increases 800-hPa SPFH by up to 1 g kg−1 locally, helping to correct a dry bias. The forecast initialized from this analysis better captures Hanna’s observed intensification rate, likely because its moister inner core facilitates development of persistent deep convection near the TC center, where diabatic heating is more efficiently converted to cyclonic wind kinetic energy.

Significance Statement

Tropical cyclone (TC) intensification can be strongly sensitive to the lower-to-midtropospheric water vapor distribution near the storm. The COSMIC-2 GNSS radio occultation (RO) receiver satellite mission provides denser spatial coverage of atmospheric water vapor and temperature profiles over the tropics compared to other GNSS RO observation platforms. Herein, using six 2020 Atlantic TC cases, we evaluate the impacts of assimilating COSMIC-2 RO bending angles into a regional forecast model that already assimilates clear-sky satellite radiances. It is shown that COSMIC-2 assimilation yields a modest ∼10% intensity forecast skill improvement for several lead times, although more substantial intensity forecast improvement is found for a few forecasts where the COSMIC-2 observation assimilation helps correct a lower-to-midtropospheric water vapor bias.

© 2023 American Meteorological Society. This published article is licensed under the terms of the default AMS reuse license. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).

Corresponding author: Dr. William J. Miller, wmiller1@umd.edu

Abstract

This study evaluates the impact of assimilating Global Navigation Satellite System (GNSS) radio occultation (RO) bending angles from Formosa Satellite Mission-7/Constellation Observing System for Meteorology, Ionosphere, and Climate-2 (COSMIC-2) receiver satellites on Hurricane Weather Research and Forecasting (HWRF) Model tropical cyclone (TC) forecasts. Launched in June 2019, the COSMIC-2 mission provides significantly higher tropics data coverage compared to its predecessor COSMIC constellation. GNSS RO measurements yield information about atmospheric pressure, temperature, and water vapor profiles. HWRF is cycled with and without COSMIC-2 bending angle data assimilation for six 2020 Atlantic hurricane cases. COSMIC-2 assimilation has little impact on HWRF track forecasts, consistent with HWRF’s design limiting cycled data assimilation impacts on surrounding large-scale flows; however, COSMIC-2 assimilation results in a statistically significant ∼8%–12% mean absolute forecast error reduction in minimum central sea level pressure for t = 36-, 54-, 60-, and 108–120-h lead times. Forecasts initialized from analyses assimilating COSMIC-2 observations also have a 1%–4% smaller 600–700-hPa specific humidity (SPFH) root-mean-squared deviation compared to radiosondes and dropwindsondes for most lead times. While not all HWRF intensity forecasts benefit from COSMIC-2 assimilation, a few show notable improvement. For example, assimilating two COSMIC-2 profiles within the inner core of developing Hurricane Hanna (2020) increases 800-hPa SPFH by up to 1 g kg−1 locally, helping to correct a dry bias. The forecast initialized from this analysis better captures Hanna’s observed intensification rate, likely because its moister inner core facilitates development of persistent deep convection near the TC center, where diabatic heating is more efficiently converted to cyclonic wind kinetic energy.

Significance Statement

Tropical cyclone (TC) intensification can be strongly sensitive to the lower-to-midtropospheric water vapor distribution near the storm. The COSMIC-2 GNSS radio occultation (RO) receiver satellite mission provides denser spatial coverage of atmospheric water vapor and temperature profiles over the tropics compared to other GNSS RO observation platforms. Herein, using six 2020 Atlantic TC cases, we evaluate the impacts of assimilating COSMIC-2 RO bending angles into a regional forecast model that already assimilates clear-sky satellite radiances. It is shown that COSMIC-2 assimilation yields a modest ∼10% intensity forecast skill improvement for several lead times, although more substantial intensity forecast improvement is found for a few forecasts where the COSMIC-2 observation assimilation helps correct a lower-to-midtropospheric water vapor bias.

© 2023 American Meteorological Society. This published article is licensed under the terms of the default AMS reuse license. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).

Corresponding author: Dr. William J. Miller, wmiller1@umd.edu

1. Introduction

After decades of limited improvement, National Hurricane Center (NHC) Atlantic basin tropical cyclone (TC) intensity forecast skill has steadily improved since 2010, owing in part to improved operational dynamical forecast model guidance (Cangialosi et al. 2020). TC intensification is controlled by complex multiscale interactions between large-scale conditions (Kaplan and DeMaria 2003; DeMaria et al. 2005) and inner-core processes (e.g., Rogers et al. 2013; Montgomery and Smith 2014). The National Oceanic and Atmospheric Administration (NOAA)’s operational Hurricane Weather Research and Forecasting (HWRF) regional model intensity guidance has become particularly skillful in recent years (Cangialosi et al. 2020), likely benefiting from a series of upgrades, which include improved physics schemes, increasingly convection-permitting resolution, and ability to assimilate in situ inner-core observations (Biswas et al. 2018b). However, HWRF still struggles to capture significant Atlantic TC intensity changes, such as Hurricane Hanna (2020)’s last-minute rapid strengthening as it approached the southern Texas coast (Brown et al. 2021). Developing regional numerical weather prediction (NWP) model data assimilation (DA) algorithms, which optimally utilize in situ and satellite observations for sensing the structures important to TC intensification, is a high NOAA research-to-operations priority moving forward (Zawislak et al. 2022).

The joint U.S.–Taiwan Formosa Satellite Mission-7/Constellation Observing System for Meteorology, Ionosphere, and Climate-2 (COSMIC-2) Global Navigation Satellite System (GNSS) radio occultation (RO) receiver satellite mission was launched in June 2019 to replace the FORMOSAT-3/COSMIC (COSMIC) constellation which no longer transmits data. Compared to COSMIC, the six COSMIC-2 satellites provide substantially higher data coverage over the tropics—about five daily profiles per 500 × 500 km2 box; COSMIC-2 receivers also have a higher signal-to-noise ratio, resulting in more than 85% of all soundings penetrating below the 1-km height, compared to 80% of all COSMIC soundings based on an initial assessment (Ho et al. 2020b). Raw radio signals intercepted by GNSS RO receivers can be processed into various retrieval products, including refractivity and bending angle, which yield information about the atmospheric pressure, temperature, and water vapor profile intercepted by the RO ray (Kuo et al. 2004). COSMIC-2 RO measurements have high (<100 m) vertical resolution, all-sky observability, deep tropospheric signal penetration, and unlike satellite radiances, they do not require bias correction (Ho et al. 2020a). Given that NWP model forecasts of TC genesis and intensification have shown strong sensitivity to environmental lower-to-midtropospheric water vapor (Sippel and Zhang 2008; Doyle et al. 2012; Teng et al. 2021), assimilated COSMIC-2 RO profiles could potentially provide a valuable sampling of TC environments and improve the model forecasts.

Previous studies have shown beneficial impacts of RO refractivity assimilation on regional model TC forecasts. Examining ten developing typhoon cases, Chen et al. (2020) showed how assimilating COSMIC RO data into the Advanced Research Weather Research and Forecasting (WRF-ARW) Model over multiple DA cycles increased TC cyclogenesis probability of detection from 30% to 70%. They showed how the increased lower-to-midtropospheric moisture in Typhoon Nuri (2008) analyses assimilating COSMIC refractivity yielded stronger inner-core updrafts later in the cycling period compared to their “GTS” experiment without COSMIC observations; the updrafts coincided with a developing midlevel circulation and cyclogenesis, both of which GTS failed to capture. Other studies have shown how RO refractivity assimilation in regional models could improve forecasts of TC genesis, intensity changes, and rainfall (Huang et al. 2005; Kueh et al. 2009; H. Liu et al. 2012; Teng et al. 2021, 2023) as well as TC motion (Chen et al. 2015; Huang et al. 2010). Recently, Ruston and Healy (2021) and Lien et al. (2021) reported positive impacts of COSMIC-2 DA on global NWP model forecasts–particularly for midtropospheric water vapor in the tropics—using gridpoint-based metrics. Lien et al. (2021) did not find statistically significant impacts of COSMIC-2 assimilation on the Taiwan Central Weather Bureau (CWB)’s global NWP model typhoon track forecasts, although their sample size was small. Several other recent NWP modeling case studies evaluating typhoons Mitag (2019), Haishen (2020), and Hagupit (2020) have shown generally neutral to positive impacts of COSMIC-2 DA on TC track, intensity and structure prediction (S.-Y. Chen et al. 2021; Y.-C. Chen et al. 2021; Chien et al. 2022).

Despite the beneficial RO observation assimilation impacts on TC forecasts reported in the earlier studies, it is possible that the operational HWRF may not substantially benefit from COSMIC-2 DA. Unlike the regional models used in most of the studies mentioned above, the operational HWRF assimilates clear-sky satellite radiances, which provide rich sampling over data-sparse oceans. Also, while several RO DA impact studies have focused on TC cyclogenesis and early development (H. Liu et al. 2012; Chen et al. 2020; Teng et al. 2021), relatively little attention has been given to how RO DA impacts intensity forecasts of mature TCs. Despite their increased tropics coverage compared to other RO missions, COSMIC-2 observations provide limited sampling of TC inner cores where important intensification processes occur. Finally, it is possible that HWRF’s default RO DA algorithm may not optimally utilize COSMIC-2 observations. The objectives of this study are twofold. First, simulating a set of 2020 Atlantic hurricane cases using an offline HWRF configuration, we ask: can COSMIC-2 RO observation assimilation improve TC track and intensity forecasts? And if so, through what physical mechanisms? Second, we evaluate the performance of HWRF’s default RO DA algorithm when assimilating COSMIC-2 observations and explore whether any adjustments may be needed.

The remainder of the paper is organized as follows. Section 2 describes the selected hurricane cases, experiment design, and the HWRF model configuration, including its COSMIC-2 DA algorithm. A preliminary evaluation of the HWRF COSMIC-2 DA algorithm’s performance follows in section 3. Section 4 describes COSMIC-2 DA impacts on HWRF forecast track, intensity, and gridpoint-based error statistics. An analysis of COSMIC-2 DA impacts on TC structure and potential intensification pathways, using selected HWRF forecasts, is presented in section 5. Section 6 provides a summary and the conclusions.

2. Experiment design, model configuration, and COSMIC-2 RO data assimilation algorithm

a. Summary of cases

We have selected six TC cases from the highly active 2020 Atlantic hurricane season for running HWRF COSMIC-2 assimilation impact experiments: Hurricane Hanna, Hurricane Isaias, Major Hurricane Laura, Hurricane Sally, Major Hurricane Delta, and Major Hurricane Zeta. Criteria for selection include a U.S. landfall of at least hurricane intensity and the availability of verification dropsonde observations from aircraft reconnaissance missions. Figure 1 shows the six storms’ observed tracks and Saffir–Simpson scale intensities. For brevity, we refer the reader to the NHC Tropical Cyclone Reports covering each case for details on the meteorological characteristics, operational forecast critique, and storm impacts: Brown et al. (2021; Hanna), Latto et al. (2021; Isaias), Pasch et al. (2021; Laura), Berg and Reinhart (2021; Sally), Cangialosi and Berg (2021; Delta), and Blake et al. (2021; Zeta). We summarize the Hanna and Zeta cases below, given our extra focus on them in section 5.

Fig. 1.
Fig. 1.

Observed tracks, colored by the current stage of TC development, of the six 2020 Atlantic hurricanes selected for the HWRF retrospective forecast experiments. Hurricane intensity is stratified by VMAX according to the Saffir–Simpson scale: category 1, 64–82 kt; category 2, 83–95 kt; category 3, 96–112 kt; category 4, 113–136 kt; and category 5, ≥137 kt. Observed position and intensity data are taken from the NHC best track database. Closed circles and × symbols denote each case’s beginning and end of the HWRF cycling period. The HWRF ghost d02 and ghost d03 domain sizes are also shown as gray rectangles.

Citation: Monthly Weather Review 151, 7; 10.1175/MWR-D-22-0198.1

Hurricane Hanna (2020) developed into a tropical depression around 0000 UTC 23 July from a westward-moving tropical wave in the central Gulf of Mexico. The storm’s precursor disturbance had struggled to organize while crossing the central tropical Atlantic and later the Caribbean, initially due to midlevel dry air and unfavorable upper-level winds, despite increased deep convection accompanying the wave. Vertical wind shear (VWS) began to relax once the wave entered the southeastern Gulf of Mexico, coinciding with the disturbance gaining increased convective organization and developing a surface low pressure center. Deep-layer winds south of a subtropical high steered Hanna westward, then west-southwestward toward the southern Texas coast. Although Hanna initially struggled to intensify due to midlevel dry air to its west, the TC was otherwise embedded in an environment favorable for intensification, with high ocean heat content and low-to-moderate VWS. Around 0000 UTC 24 July Hanna began a sustained period of moderate intensification. Hanna’s intensification rate increased markedly after 0600 UTC 25 July, when maximum surface wind speed (VMAX) increased from 60 to 80 kt (1 kt ≈ 0.51 m s−1) over the next 12 h. Notably, this rate exceeded the 30 kt (24 h)−1 rapid intensification (RI) threshold defined in Kaplan and DeMaria (2003), which corresponds to the 95th percentile of overwater 24-h Atlantic TC intensity changes from the 1989–2000 period.1

Hanna made landfall around 2200 UTC 25 July on South Padre Island, Texas, as a category-1 hurricane with an 80-kt VMAX. Hanna’s most destructive impacts came from inland flooding rains. Over 6 in. (1 in. = 2.54 cm) of rain fell across the Rio Grande Valley with isolated totals exceeding 12 in., resulting in extensive damage to agriculture and infrastructure and four flooding-related fatalities in northeast Mexico. Despite Hanna “only” reaching category-1 intensity, it was one of the more problematic 2020 Atlantic TC cases from an operational forecasting standpoint. As Hanna approached Texas, the operational HWRF and other global NWP model guidance continually under-forecast the storm’s intensification rate. NHC forecasts did not predict the storm to reach hurricane intensity until 24 h before landfall (Brown et al. 2021).

Hurricane Zeta (2020) developed into a tropical depression at 1200 UTC 24 October from a northward drifting surface trough about 110 km southwest of Grand Cayman. Over the next two days, Zeta intensified steadily in a favorable environment of low VWS and high ocean heat content. Meanwhile, the storm began accelerating northwestward, steered by a subtropical high centered over the Gulf of Mexico. Zeta made landfall as a 75-kt category-1 hurricane on the east coast of Mexico’s Yucatán Peninsula around 0400 UTC 27 October. Over the next ∼12 h, Zeta weakened while crossing the northeast Yucatán. Nevertheless, by 1800 UTC later that day, Zeta remained a 55-kt strong tropical storm, its inner core having moved back over water in the southern Gulf of Mexico. Then, 6 h later, Zeta began to undergo RI as it accelerated northward around the western periphery of the subtropical high, with its VMAX increasing from 55 to 95 kt over the next 18 h. Zeta made landfall in southeastern Louisiana as a 100-kt category-3 hurricane around 2100 UTC 28 October. The combination of Zeta’s fast forward motion and its large wind field resulted in widespread wind damage to trees and buildings over southeastern Louisiana and nearby parts of southern Mississippi and Alabama. Zeta’s winds and storm surge caused five fatalities and about $4.4 billion (U.S. dollars) of damage in the United States.

NHC official intensity forecast errors for Zeta exceeded the previous 5-yr Atlantic TC seasonal mean for most lead times, resulting partly from poor NWP model intensity guidance (Blake et al. 2021). Common themes among the NWP model suite included (i) insufficient strengthening before the Yucatán landfall, (ii) missing the subsequent land-induced weakening in early cycle forecasts due to a right-of-track bias that minimized land interaction, and (iii) failure to predict the storm’s RI in the northern Gulf of Mexico. Favorable environmental conditions likely facilitated Zeta’s RI; they included low VWS, warm sea surface temperatures, and upper-level divergence downstream of a deep-layer cutoff low over Texas. Given Zeta’s fast forward motion in the Gulf of Mexico, an additional factor likely contributing to Zeta’s attaining Major Hurricane intensity before its Louisiana landfall was the relatively early RI onset at 0000 UTC 28 October, only about 12 h after the center re-emerged over water. The lack of substantial weakening while Zeta crossed the northeast Yucatán could have resulted from a combination of (i) the low-lying, flat terrain and (ii) Zeta occurring during the local climatological rainy season when wet antecedent conditions could have partially mitigated the loss of ocean surface latent heat fluxes.

b. HWRF Model configuration

Used operationally at the National Centers for Environmental Prediction (NCEP) for generating real-time Atlantic TC forecasts prior to its retirement after the 2022 season, HWRF is a cloud-resolving regional forecasting model featuring 75 vertical levels and a triply nested storm-following grid with a 1.5-km-resolution innermost nest. This study uses an offline HWRF v4.0a configuration (Biswas et al. 2018a) that is similar to the 2020 operational HWRF, unless otherwise noted. Whenever an Atlantic basin TC underwent cyclogenesis while HWRF was operational, NCEP activated an HWRF cold start in real time using background fields and lateral boundary conditions provided by its operational Global Forecast System (GFS) model.

HWRF is cycled every 6 h beginning with the cold start. Prior to assimilating any observations, the background TC vortex is relocated to the forecaster-estimated TCVitals (Trahan and Sparling 2012) position and corrected to better match the TCVitals vortex size and intensity (Biswas et al. 2018b; Liu et al. 2020). If an HWRF 6-h forecast is available (i.e., this is not a cold start) and the TCVitals VMAX ≥ 14 m s−1, the HWRF 6-h background vortex kinematic and thermodynamic fields are extracted from the forecast and then relocated/corrected. Otherwise, the vortex fields are extracted from the currently valid GFS 6-h forecast. For both cases, the relocated background vortex is added to the GFS 6-h forecast “environmental” fields interpolated to a large regional domain, where the “environment” is defined as the residual following extraction of the GFS 6-h forecast TC vortex. The size, surface pressure, and three-dimensional wind, temperature, and water vapor fields of the relocated background forecast vortex are corrected to match better the TCVitals size, surface pressure, and VMAX intensity under hydrostatic conditions while keeping background relative humidity (RH) unchanged. Next, the background state consisting of the relocated/corrected TC vortex superimposed on the GFS environment is interpolated to HWRF’s 11° × 11° ghost d03 and 20° × 20° ghost d02 domains. The Gridpoint Statistical Interpolation (GSI) DA algorithm (Wang et al. 2013) then assimilates observations into ghost d02 and ghost d03. Next, the ghost d02 and ghost d03 analyses are interpolated to the smaller intermediate and innermost model integration nests, which are the 17.7° × 17.7° d02 and 5.9° × 5.9° d03, respectively. The HWRF outermost 77° × 77° parent d01 domain does not assimilate observations, but rather it uses fields interpolated from the currently valid GFS analysis. To summarize: (i) only the vortex region background fields are cycled (provided that the TCVitals VMAX ≥ 14 m s−1), (ii) the remaining portions of ghost d02 and ghost d03 assimilate observations on top of the GFS 6-h background, and (iii) the HWRF outermost d01 domain is initialized from the current GFS analysis.

By default, HWRF v4.0a runs GSI in hybrid-ensemble 3DVAR mode, using background ensemble covariances downscaled from NCEP’s 80-member operational Global Data Assimilation System (GDAS) ensemble. One difference between our HWRF configuration and the operational HWRF is the latter’s activation of a high-resolution cycled HWRF ensemble used for supplying mesoscale background error covariances for high-priority storms; our HWRF experiments forgo the use of this option due to computational constraints. Like the operational HWRF, our configuration assimilates a rich set of observations, which include: conventional radiosonde, buoy, surface ship, and commercial aircraft measurements; GNSS RO bending angles; in situ inner-core observations from manned aircraft reconnaissance missions such as dropwindsonde wind, temperature, water vapor and surface pressure, Tail Doppler Radar radial velocities, and flight-level wind, temperature, and dewpoint temperature; as well as clear-sky microwave and infrared satellite radiances (Biswas et al. 2018b; Wu et al. 2019). The satellite radiance assimilation in our HWRF-GSI configuration is a particularly notable difference between our study and some others evaluating RO observation impacts on regional model TC forecasts that did not assimilate radiances (e.g., H. Liu et al. 2012; Chen et al. 2020; Teng et al. 2021). See Biswas et al. (2018b) for a detailed description of HWRF v4.0’s critical components, including the physics parameterizations, the vortex correction procedure, and the GSI DA system.

c. HWRF observing system experiment design

Each HWRF cycling experiment covers a multiday period, beginning when the system develops into a tropical depression and generally ending 12–24 h before final landfall (Fig. 1); 126-h HWRF free forecasts are initialized from each 6-hourly DA cycle. Therefore, these multiday cycling experiments include weaker early developing and more mature TC analyses. Note that the weaker (i.e., tropical depression and tropical storm intensity) TC analyses tend to occur early in the cycling period (Fig. 1) when fewer COSMIC-2 profiles have been assimilated over prior cycles. Two cycled HWRF experiments are run for each hurricane case: (i) Control, using all available observations (section 2b), but with COSMIC-2 data withheld, and (ii) C2, identical to Control except that COSMIC-2 bending angles are assimilated. The COSMIC-2 data, which have a median latency of less than 30 min (Weiss et al. 2022), had been processed in real time by the University Corporation for Atmospheric Research (UCAR) and provided to NCEP in Binary Universal Form for the Representation of meteorological data (BUFR) for assimilation in the operational GFS for all 2020 case study periods. Both the C2 and Control configurations assimilate GNSS RO bending angles from other receiver platforms, which are MetOp-A/B/C, TerraSAR-X, TanDEM-X, and KOMPSAT-5. These non-COSMIC-2 platforms provide only a small fraction of the number of COSMIC-2 observations available for assimilation. Although we have not run additional sensitivity tests evaluating the non-COSMIC-2 RO observation assimilation impacts on HWRF, we expect their impacts to be relatively small. On average, about three COSMIC-2 profiles are assimilated in the HWRF ghost d03 nest per cycle, although this number varies between 0 and 8. Figures 2a and 2b show the number of COSMIC-2 profiles with bending angles below 800 hPa passing HWRF-GSI’s quality control checks (section 2f) that are assimilated into HWRF ghost d03 and ghost d02, respectively, during each DA cycle for three selected cases. Control and C2 run the same number of DA cycles for each case; they are 6, 25, 24, 18, 20, and 15 for Hanna, Isaias, Laura, Sally, Delta, and Zeta, respectively. Note that both the Control and C2 experiments initialize HWRF’s outermost d01 domain with the same operational GFS analysis fields; both experiments also use the same lateral boundary conditions supplied by the operational GFS forecast.

Fig. 2.
Fig. 2.

(a) Number of COSMIC-2 profiles with portions below 800 hPa passing quality control checks that are assimilated into the HWRF ghost d03 during each DA cycle for the Hanna, Zeta, and Isaias (2020) cases. DA cycles are labeled on the x axis by their cycle number n in the ordered sequence i = 1, …, n, …i_last where i = 1 and i = i_last are the cold start and final DA cycles, respectively. (b) As in (a), but for ghost d02.

Citation: Monthly Weather Review 151, 7; 10.1175/MWR-D-22-0198.1

d. HWRF-GSI RO bending angle forward operator

The UCAR COSMIC Data Analysis and Archive Center (CDAAC) processes the raw COSMIC-2 radio signals into various retrieval products. The two processed RO products most suitable for assimilation into atmospheric models are (i) the bending angle (α) between the occultation ray asymptotes emanating from the transmitter and receiver satellites; and (ii) refractivity N = (n − 1) × 106 at the ray perigee, where the index of refraction n–the ratio of the speed of light in a vacuum to that in a medium–can be approximated as a scalar field dependent on pressure, temperature, and water vapor. Refractivity can be assimilated using a simple forward operator that computes N as a function of background fields interpolated to the observation location. However, compared to bending angles, a disadvantage of assimilating refractivity is the extra processing needed for the N retrieval—an additional source of retrieval error. For example, CDAAC uses climatological data to smooth and adjust upper portions of α profiles before converting them to N profiles via an Abel transform (Kuo et al. 2004).

HWRF-GSI computes RO bending angles from the background state using the local forward operator NCEP Bending Angle Model (NBAM), which is currently used in the operational GFS (Cucurull et al. 2013).2 The term “local” means that NBAM integrates the background n field over a single column above each bending angle observation’s tangent point location rather than along the two-dimensional limb sounding traced by the RO ray path. A nonlocal RO bending angle forward operator, such as the one used in the European Center for Medium-Range Weather Forecasting (ECMWF) global model (Healy et al. 2007), also depends on background refractivity away from the tangent point column along at least a portion of the approximated RO ray. NBAM computes bending angle α as a function of its associated impact parameter a using an inverse Abel transform:
α(a)=2aadlnn/dxx2a2dx,
where a, supplied by the processing center, is the distance from the RO ray tangent point to Earth’s center of curvature (ra) multiplied by n at the tangent point; and refractional radius x = nr, where n and the distance r to Earth’s center of curvature are evaluated at a point in the vertical column above the ray tangent point. Equation (1) assumes that the atmospheric refractivity along the RO ray path is spherically symmetric, so horizontal gradients are neglected. Whether this assumption is reasonable in the tropical lower troposphere, where horizontal water vapor fields tend to be more spatially inhomogeneous compared to higher latitudes, could be worth exploring in a future study. For example, Chen et al. (2020) showed that replacing WRF-ARW’s local RO refractivity forward operator with a nonlocal excess phase forward operator resulted in a substantial improvement in tropical cyclogenesis detection.

Global NWP models run at most operational centers, including NCEP (Cucurull et al. 2013), ECMWF (Ruston and Healy 2021), and the Met Office (UKMET; Bowler 2020), directly assimilate bending angles rather than refractivity. However, regional models such as the WRF-ARW typically assimilate local refractivity or nonlocal excess phase (e.g., H. Liu et al. 2012; Chen et al. 2020) because they usually have lower model tops, rendering them less suitable for a bending angle forward operator that requires background refractivity data extending to a great height. However, HWRF v4.0a has implemented a global-regional blended vertical coordinate in the GSI analysis that raises the model top to 0.26 hPa by blending HWRF and GFS vertical levels in the lower stratosphere, smoothly transitioning to the GFS levels with increasing height (Biswas et al. 2018b).

e. HWRF-GSI RO observation error specification

HWRF-GSI assigns the final observation error standard deviation (σo) to each RO bending angle as the representativeness error (σrep), which approximates the expected forward operator and retrieval uncertainty, multiplied by a “superobs factor” proportional to the square root of the number of profile observations in the nearest model layer that compensates for possible σo correlations among nearby observations. GSI computes σrep via a series of piecewise height-dependent quadratic functions formulated as
σrep=103×eγ,γ=c0+c1(arc)+c2(arc)2,
where rc is Earth’s local radius of curvature, arc is the impact height, and coefficients c0, c1, and c2 (provided in Table 1) depend on the latitudinal zone, impact height, and processing protocol for different RO receiver platforms. There are two RO observation processing protocols: the CDAAC protocol (COSMIC-2 and KOMPSAT-5 platforms) and the UKMET protocol (MetOp-A/B/C, TerraSAR-X, TanDEM-X platforms). The Desroziers et al. (2005) method, which estimates observation error variance using vectors of observation-minus-background (OB) and observation-minus-analysis (OA) departures, was previously used to determine coefficients c0, c1, and c2 from a global sample of RO bending angle observations assimilated into the GFS model (Cucurull 2010; Cucurull et al. 2013). Our HWRF C2 configuration assigns the same representativeness errors previously set for COSMIC bending angles to the COSMIC-2 bending angles. Diagnostics from our C2 experiments—shown in section 3a—suggest that these settings may not be optimal for certain heights in HWRF.
Table 1

Coefficients c0, c1, and c2 used in the HWRF-GSI v4.0a RO bending angle representativeness error specification [Eq. (2)], which depend on the data processing protocol, latitude band, and impact height arc.

Table 1

f. HWRF-GSI RO observation quality control algorithm

Observations with substantial departures from the background state can adversely affect the cost function minimization. Therefore, a critical component of any DA system is a well-tuned quality control (QC) algorithm, which identifies and rejects poor quality observations and those that the forward operator cannot sufficiently represent. HWRF-GSI v4.0a rejects RO bending angle observations that meet any of several criteria which include:

  1. Statistical check: when the absolute OB bending angle fractional innovation |OB|/O exceeds a cutoff value that has been empirically tuned for the GFS using global RO data before the COSMIC-2 era. The cutoff value is a piecewise function of impact height, and for some layers, it also depends smoothly on latitude or temperature (Table 2).

  2. Superrefraction (SR) check. SR layers, most common near the top of the maritime boundary layer in the subtropics or trade wind regions off the west coasts of major continents, are characterized by large vertical refractivity gradients that cause RO rays to become trapped within the atmosphere (Xie et al. 2010). NBAM could yield unrealistic values for observations below a SR layer because an infinite number of atmospheric states could reproduce the same observed RO profile (Xie et al. 2006; Cucurull 2015). HWRF-GSI v4.0a rejects RO bending angles located near or below a layer where SR likely occurred, either in the model background or observation space. Two criteria trigger a SR rejection for observations below a 6-km impact height: (i) if the background vertical N gradient exceeds (0.75 × 157) N-units km−1 near the observation; and (ii) if the observed bending angle exceeds 0.03 radians and the background vertical N gradient exceeds (0.5 × 157) N-units km−1 near the observation. For (ii), the profile observation with the largest bending angle is first identified, and then the portion of the profile below it is rejected (Cucurull 2015).

Table 2

HWRF-GSI v4.0a statistical QC check cutoff values for RO bending angle |OB|/O, which depend on temperature (T; K), latitude (λ; radians), and impact height arc (km). Dependent variables in the expressions below are highlighted in boldface font for emphasis.

Table 2

See Lien et al. (2021) section 3d for a description of the GSI RO bending angle DA algorithm’s other QC checks, namely, the out-of-model boundary check, “n-time observation error” gross error check, and height threshold.

3. COSMIC-2 bending angle assimilation statistics in HWRF-GSI

a. Innovation and observation error statistics

Before describing COSMIC-2 DA impacts on HWRF forecasts, we first examine HWRF-GSI diagnostic output from the C2 experiments to verify that the COSMIC-2 bending angle assimilation algorithm is functioning as expected. Figure 3a shows COSMIC-2 bending angle observation-minus-background fractional innovation [i.e., (OB)/O] root-mean-squared deviation (RMSD) and mean bias profiles generated using all ghost d03 observations3 collected from the six hurricane cases, where “observation” herein refers to a bending angle at a single height within an RO profile. Profiles are truncated at the 18-km impact height to emphasize the atmospheric layer most important to TC prediction, namely, the troposphere and lower stratosphere. These results are generally consistent with the COSMIC-2 (OB)/O RMSD and mean bias profiles obtained from the CWB-adapted GFS model shown in Lien et al. (2021; see their Fig. 6a). The pre-QC (OB)/O RMSD profile peaks in the lower troposphere, likely resulting in part from both (i) a larger forward operator uncertainty in the moist lower troposphere (section 2d) and (ii) model water vapor forecast errors. The (OB)/O RMSD reaches a minimum in the upper troposphere, where RO data are expected to be of the highest quality (Kuo et al. 2004). Although COSMIC-2 observations are nearly unbiased to HWRF above the 7.5-km impact height, they show a negative OB bias in the lower troposphere, a feature also noted by Schreiner et al. (2020) and Lien et al. (2021), both of whom used background datasets other than HWRF. Thus, the negative COSMIC-2 OB bias in the lower troposphere likely results partially from biased RO observations. Comparing the post-QC COSMIC-2 bending angle (OA)/O and (OB)/O RMSD profiles (cf. Figs. 3a,b; Table 3), we find the former to be ∼40% smaller, indicating that the COSMIC-2 observations are having a positive impact on the HWRF analyses by reducing the RMSD.

Fig. 3.
Fig. 3.

(a) Profiles of COSMIC-2 bending angle OB fractional innovation RMSD (red lines) and mean bias (blue lines) generated using all observations available for assimilation into HWRF ghost d03 from the six TC case experiments. Dashed and solid colored lines show statistics computed before and after observation QC, respectively. Dashed and solid black lines show the number of COSMIC-2 bending angles available for assimilation and assimilated, respectively. Data are binned by impact height using 1-km bin widths. (b) As in (a), but for the COSMIC-2 bending angle OA fractional innovation. (c) As in (a), but for profiles of post-QC COSMIC-2 bending angle observation-minus-background normalized innovation [(OB)/σo]. Also shown are normalized σo and normalized σrep, where data in each height bin are averaged over all post-QC COSMIC-2 observations. For normalized σo and normalized σrep, the mean post-QC COSMIC-2 bending angle observation value for each height bin is used as the denominator.

Citation: Monthly Weather Review 151, 7; 10.1175/MWR-D-22-0198.1

Table 3

Assimilated COSMIC-2 bending angle OB and OA fractional innovation RMSD values (%) using HWRF ghost d03 data from all six C2 case experiments as for Fig. 3, listed by impact height.

Table 3

According to Cucurull et al. (2013), RO bending angle observation errors can be reasonably well characterized if the (OB)/σo RMSD ranges between one and two. For our assigned COSMIC-2 errors (section 2e), we find this to be the case below the 5-km impact height and in the 8–15-km impact height layer (Fig. 3c). However, the COSMIC-2 bending angle (OB)/σo RMSD exceeds two over portions of the middle troposphere and lower stratosphere (Fig. 3c), suggesting the assigned σo values may be smaller than optimal there. Figure 3c also shows the mean COSMIC-2 σrep and σo profiles normalized by the mean observation value in each height bin (note the separate x-axis scale at the panel top). The spread in σrep among these observations is nearly zero (not shown), which is expected given that nearly all come from within the 40°S–40°N latitude band [Eq. (2) and Table 1].

b. Quality control statistics

The histogram shown in Fig. 4a plots the percentage of available COSMIC-2 bending angle observations rejected by HWRF-GSI’s QC algorithm binned by pressure level height. Although the rejection percentage remains relatively small—under 10%—for most heights, it becomes larger in the lower troposphere. For example, nearly 30% of COSMIC-2 observations are rejected from the 850–950-hPa layer. When considering only COSMIC-2 rejections below 800 hPa, the SR check generates the largest fraction, followed by the Statistical check (Fig. 4b), and the other QC criteria account for very few (not shown). Figure 4b also shows that the bulk of COSMIC-2 observations assimilated into ghost d03 come from the near-storm environment, in a 200–700 km radial band surrounding the TC center. The lower-tropospheric COSMIC-2 observation rejection percentage increases with increasing background SPFH above 12 g kg−1 (Fig. 4c). This results in part from Statistical check rejections becoming more frequent with increasing lower-tropospheric water vapor, consistent with our expectation that the forward operator error (and therefore the OB spread) could become larger in moister regions (section 2d); however, interestingly, the SR check rejection percentage increases even more dramatically with increasing background lower tropospheric moisture (cf. the relative heights of black, red, and blue bars in Fig. 4d).

Fig. 4.
Fig. 4.

(a) Histogram of the percentage of COSMIC-2 bending angle observations available for assimilation in HWRF ghost d03 from all six TC case experiments that are rejected by the HWRF-GSI QC algorithm, binned by pressure height. (b) As in (a), but showing a radius-binned histogram of the number of COSMIC-2 bending angles below 800 hPa taken from (i) the full dataset prior to HWRF-GSI QC screening (black bars), (ii) the subset rejected by the SR QC check (blue bars), and (iii) the subset rejected by the Statistical QC check (red bars). (c) As in (a), but for the COSMIC-2 observations below 800 hPa, binned by background forecast specific humidity at the observation location. (d) As in (b), but for COSMIC-2 observations binned by background forecast specific humidity at the observation location.

Citation: Monthly Weather Review 151, 7; 10.1175/MWR-D-22-0198.1

Further work is needed to better understand the characteristics of the lower-troposphere COSMIC-2 bending angles rejected by HWRF-GSI’s SR and Statistical checks. While it is important to reject poor-quality observations, an optimal data assimilation algorithm should maximize the use of good-quality observations, provided the forward operator can accurately simulate them. This better understanding could help guide future work to optimize and test the QC algorithm for COSMIC-2 observations in our offline HWRF-GSI model.

4. COSMIC-2 observation assimilation impacts on HWRF forecast error statistics

a. Mean track and intensity errors

The GFDL Vortex Tracker (Biswas et al. 2018b) provides the HWRF TC center position and minimum central sea level pressure (PMIN) intensity every 3 h, while the 6-hourly NHC best track TC position and PMIN intensity data (Jarvinen et al. 1984) are used for verification.4 Figure 5a compares the time series of Control and C2 TC absolute position errors averaged at each 6-hourly verification time over all 108 free forecast initializations from the six case experiments. Following Z. Liu et al. (2012), we generate 90% confidence intervals using a 10 000-iteration bootstrap technique (Wilks 2011) that resamples the set of C2-minus-Control forecast absolute position error differences at each time and recomputes the mean difference for every iteration; results are considered statistically significant improvements (degradations) for C2 if the confidence interval remains below (above) zero. Figure 5c shows the track forecast relative skill gained from the COSMIC-2 DA, defined as the Control-minus-C2 mean absolute error difference normalized by the Control mean absolute error. Although the Control and C2 mean absolute position errors are similar before t = 60 h, C2 track forecasts show a statistically significant ∼10% relative skill improvement very late in the forecast period over the t = 102–114-h window (Figs. 5a,c). The absence of significant track forecast differences between Control and C2 throughout most of the forecast period is not surprising, given that HWRF replaces the 6-h background fields outside of the TC vortex with the GDAS background (section 2b). Prior research has shown that despite some exceptions, TC motion is often more strongly controlled by the larger-scale environment than by vortex-scale processes (e.g., Chan and Gray 1982; Chan et al. 2002; Galarneau and Davis 2013). Therefore, we expect the cumulative impact of COSMIC-2 DA over multiple cycles to have a stronger impact on the cycled TC vortex region—typically bounded by the outermost closed isobar’s approximate radius–compared to the surrounding environment.

Fig. 5.
Fig. 5.

(a) Absolute TC position error (km), averaged over all HWRF forecasts from the six cases valid at each 6-hourly lead time, where the dashed green line denotes the number of forecasts used in the average, for the Control (blue) and C2 (orange) configurations. Black solid and dashed lines plot the NHC official forecast TC position errors averaged over the six TC cases and all 2015–19 Atlantic TC cases, respectively. Green shading denotes the 90% confidence interval for the mean C2-minus-Control absolute position error difference generated by a bootstrap resampling of individual forecast differences; the mean C2-minus-Control absolute position error difference is considered statistically significant at lead times when the interval does not include zero, as denoted by the triangles. (b) As in (a), but for mean PMIN absolute error. (c),(d) Relative skill, or the normalized percent absolute error reduction relative to Control, for TC position and PMIN, respectively. For (c) and (d), positive (negative) relative skill indicates that C2 forecasts are, on average, improved (degraded) relative to Control.

Citation: Monthly Weather Review 151, 7; 10.1175/MWR-D-22-0198.1

C2 PMIN intensity forecast improvement over Control is statistically significant for t = 36, 54, 60, and 108–120 h (Fig. 5b), when absolute errors are reduced by ∼8%–12% (Fig. 5d). Unlike for track forecasts, COSMIC-2 DA improves PMIN forecast relative skill at all lead times (Fig. 5d). When evaluating a research-configured HWRF’s Atlantic TC forecasts run from high-frequency cycling experiment final analyses, Christophersen et al. (2018) found a roughly similar PMIN forecast improvement after adding unmanned Global Hawk (GH) dropwindsonde assimilation (see their Fig. 8d). However, they found more notable improvements for analyses that did not already assimilate in situ observations from manned aircraft reconnaissance missions.

Next, we consider the subset of 70 Control and C2 forecast pairs initialized from intensifying (IN) TC analyses. Following Christophersen et al. (2018) and Rogers et al. (2013), we label an HWRF analysis as “IN” if the NHC best track VMAX increase over the subsequent 12 h exceeds 20 kt (24 h)−1. Unlike those studies, we relax the IN criterion to include analyses where the observed 20 kt (24 h)−1 VMAX intensification threshold is exceeded over any 12-h time window that begins within 12 h following the analysis time. We include these “pre-intensification” analyses in the IN group based on our expectation that there may be a time delay between COSMIC-2 DA and the TC intensification episode in the free forecast, given that most COSMIC-2 observations are assimilated outside of the inner core (Fig. 4b). Changes to TC environmental conditions, which have longer distance scales, typically influence intensity changes on longer time scales compared to inner-core processes (Fritz and Wang 2013; Rogers et al. 2013). Compared to the full forecast set, the IN sample shows a more substantial 10%–20% C2 PMIN forecast relative skill improvement over Control for the t = 48–84-h period, when the differences are statistically significant for most lead times (cf. Figs. 5b,d and cf. Figs. 6b,d). Despite COSMIC-2 RO profiles’ limited sampling of TC inner-core regions, it is encouraging to find that their assimilation on average modestly improves HWRF PMIN forecasts—particularly at medium-range lead times—for the cases evaluated here.

Fig. 6.
Fig. 6.

As in Fig. 5, but for the subset of HWRF analyses valid within 12 h of the beginning of an observed intensification episode, defined by a NHC best track VMAX rate of change exceeding 20 kt (24 h)−1.

Citation: Monthly Weather Review 151, 7; 10.1175/MWR-D-22-0198.1

b. Gridpoint-based verification against radiosondes and dropwindsondes

Control and C2 forecast temperature, specific humidity (SPFH), and horizontal wind RMSD profiles are computed against synoptic radiosondes and dropwindsondes released from NOAA P-3, NOAA G-IV, and U.S. Air Force C-130 aircraft reconnaissance missions. The RMSD profile is computed at each 6-hourly verification time using HWRF free forecast output aggregated from all six cases and observations collected within a 500-km radius and within ±1.5 h. Dropwindsonde observation locations account for their Global Positioning System (GPS)-computed horizontal drift from their release points. To mitigate HWRF TC position error contributions to the RMSD, dropwindsonde coordinates are translated on the HWRF grid to match their observed storm-relative locations.

Figure 7a shows the time series of the fractional change in the C2 forecast temperature RMSD profile relative to Control [i.e., (RMSDC2 − RMSDControl)/RMSDControl] superimposed over the Control temperature RMSD profile. Although COSMIC-2 DA degrades HWRF forecast temperature RMSD for some heights and lead times, its overall impact is neutral to slightly beneficial, especially for the 500–800- and 150–350-hPa layers where the fractional RMSD change ranges between 0% and −4% for most lead times. A more substantial—and statistically significant—C2 forecast temperature RMSD reduction relative to Control is evident for the 500–700-hPa layer around t = 54 and 72 h.

Fig. 7.
Fig. 7.

(a) Time series of the fractional change in the HWRF C2 forecast temperature RMSD profile relative to Control (shaded; %), using dropwindsonde and radiosonde data from the six hurricane cases. Black contours show the HWRF Control forecast temperature RMSD (K) profile time series. Green dashed contours show the number of observation–gridpoint pairs used. (b) As in (a), but for specific humidity (g kg−1). (c),(d) As in (a), but for u- and υ-wind components (m s−1), respectively. Black (brown) triangles in (a)–(d) show times and heights where the improvement (degradation) in C2 forecast RMSD is statistically significant (computed using a two-sample bootstrap method) at the 90% level; black (brown) closed circles denote times and heights where C2 RMSD improvement (degradation) over Control is statistically significant at the 95% level.

Citation: Monthly Weather Review 151, 7; 10.1175/MWR-D-22-0198.1

Compared to temperature, the SPFH C2 forecasts show less overall RMSD improvement over Control (Figs. 7b and 8). COSMIC-2 DA has the most beneficial impact on SPFH in the 600–700-hPa layer, where a negative 1%–4% fractional RMSD change persists through the t = 12–54-h verification period; however, this improvement is generally not statistically significant. Nevertheless, the modestly improved C2 midtropospheric water vapor field representation around t = 24 h shown in Figs. 7b and 8 is consistent with the systematic improvement in mean C2 PMIN intensity forecasts first appearing around t = 30 (Fig. 5b), given our expectation that convection-modulated TC vortex structural and intensity changes should lag near-storm water vapor field modifications. When evaluating COSMIC-2 DA impacts on the Naval Research Laboratory (NRL) and ECMWF global NWP models’ short-range water vapor forecasts in the tropics, Ruston and Healy (2021) notably did find a statistically significant improvement against radiosondes in the middle-to-upper troposphere, especially around 500–700 hPa (see their Figs. 1 and 2). Several factors could explain their stronger positive COSMIC-2 DA impact on middle-to-upper-troposphere water vapor forecasts, including their substantially larger observation sample size, different DA algorithms, different background states, and evaluation of the global tropics rather than exclusively the TC environment. In agreement with our results, Ruston and Healy (2021) did not find statistically significant NRL model water vapor forecast improvement below 800 hPa with COSMIC-2 DA. COSMIC-2 assimilation reduces u- and υ-wind RMSD by ∼1%–15% over most times and heights (Figs. 7c,d); the most notable improvements, some statistically significant, are found in the t = 48–60-h window when C2 absolute PMIN error improvements are also statistically significant (Fig. 5b).

Fig. 8.
Fig. 8.

Profile of the t = 24-h C2 forecast SPFH RMSD normalized by the t = 24-h Control forecast SPFH RMSD, with error bars denoting the 95% confidence interval (black lines). The RMSD is computed as in Fig. 7. The green line shows the number of observation–gridpoint pairs used for each 100-hPa-deep bin.

Citation: Monthly Weather Review 151, 7; 10.1175/MWR-D-22-0198.1

5. COSMIC-2 observation assimilation impacts on HWRF analysis and forecast TC structure

Among our six Atlantic hurricane cases, Hanna (2020) and Zeta (2020) showed greater medium-range PMIN forecast relative skill improvement with COSMIC-2 assimilation (not shown), compared to the six-case composite (Figs. 5b,d). These two storms also proved particularly challenging for NWP model intensity forecasting (section 2a). This section compares these storms’ C2- and Control-generated structures and explores possible mechanisms through which COSMIC-2 DA can improve TC intensity forecasts.

a. Hurricane Hanna (2020) case study

1) HWRF cycled analyses

Kaplan and DeMaria (2003) identified high relative humidity (RH) in the TC environment 700–850-hPa layer as an important RI predictor. The NHC’s Statistical Hurricane Intensity Prediction Scheme (SHIPS) statistical-dynamical intensity forecasting model also uses 700–850-hPa layer-mean environmental RH as an input parameter (DeMaria et al. 2005). Dry lower-to-midtropospheric air intrusions into a TC’s inner core can disrupt the organization of inner-core deep convection into an axisymmetric eyewall through several proposed mechanisms, which include: (i) weakening updraft buoyancy through entrainment (Cram et al. 2007; Fritz and Wang 2013; Onderlinde and Nolan 2016); (ii) enhancing evaporation-driven downdrafts, thereby stabilizing the boundary layer (Emanuel 1989; Molinari et al. 2013); and (iii) generating asymmetries in diabatic heating leading to less efficient conversion to the cyclonic wind’s kinetic energy (Braun et al. 2012).

The Hovmöller diagram shown in Fig. 9a plots the time evolution and radial distribution of azimuthally averaged RH vertically averaged over the 700–850-hPa layer in the C2 cycled analyses and its difference from Control. Differences are negligible out to the r = 300-km radius for the first two cycles, which is consistent with HWRF C2 not assimilating any COSMIC-2 profiles near Hanna’s inner core. However, note the two COSMIC-2 profiles assimilated close to the C2 TC center near r = 80 km and r = 110 km in the fourth analysis cycle at 0600 UTC 23 July. Beginning at this time and continuing for the rest of the cycling period, the C2 analysis 700–850-hPa layer-averaged azimuthal-mean RH exceeds that of Control by 1%–2% inside of r = 150 km. Unlike the 0600 UTC 23 July initialized Control free forecast, which has a low-intensity bias resulting from a delayed intensification onset, the C2 free forecast initialized at this time generates PMIN and VMAX tendencies that track closely with the NHC best track observations (Figs. 9b,d). However, the C2 storm’s intensification is halted around t = 48 h when it makes a premature landfall near Galveston, Texas (not shown). Both C2 and Control severely under-forecast Hanna’s intensification from prior analysis cycles (some forecasts not shown). In the 850–950-hPa layer (Fig. 9c), a moister C2 analysis eventually develops inside of r = 150 km, albeit one cycle later at 1200 UTC 23 July when the C2- and Control-initialized free forecasts both capture the observed RI well (Figs. 9b,d).

Fig. 9.
Fig. 9.

(a) Hovmöller plot showing the HWRF C2 cycled analysis 700–850-hPa layer-averaged and azimuthally averaged RH (shaded; %) and its difference from the cycled Control analysis (C2 − Control; contours; %) as a function of radius and UTC time, for the Hurricane Hanna (2020) cycling period. The × symbols mark the approximate 800-hPa radial locations of assimilated COSMIC-2 profiles; the total number of COSMIC-2 profiles with observations below 800 hPa assimilated in ghost d03 per cycle, including outer portions not shown here, is labeled on the right-hand y axis. (b) PMIN time series for selected Hurricane Hanna (2020) HWRF free forecasts initialized from cycled Control (blue lines) and C2 (orange lines) analyses, where different initialization times are shown in different line styles (see inset key). The black line shows the NHC best track PMIN. (c) As in (a), but for the 850–950-hPa layer. Dashed black line in (c) denotes the approximate 925-hPa C2 analysis radius of maximum wind. (d) As in (b), but for VMAX. Red squares in (b) and (d) show the C2 analysis intensity; if different from C2, the Control analysis intensity is shown as a blue square.

Citation: Monthly Weather Review 151, 7; 10.1175/MWR-D-22-0198.1

Focusing on the 0600 UTC 23 July analysis cycle, Fig. 10 compares horizontal slices of the ECMWF Reanalysis v5 (ERA5) SPFH field against Control and C2 analysis SPFH for a few selected levels. Hanna, then a tropical depression, has not yet developed a closed lower tropospheric circulation (Figs. 10a–f), although a closed 600-hPa circulation is centered about 2°S latitude of the best track center (Figs. 10g–i). At 950 and 800 hPa, the Control analysis has a mostly <1.5 g kg−1 dry bias with respect to ERA5 within 3° longitude of the TC center except for the northeast quadrant (Figs. 10a,d). The Control 600-hPa circulation generally has a moist bias (Fig. 10g). The Control analysis-minus-background SPFH difference fields show that the HWRF-assimilated observations other than COSMIC-2 have not alleviated the lower-to-midtropospheric dry bias in Hanna’s western quadrants (Figs. 10b,e), which was present in the background (cf. Figs. 10a,d,b,e). Because Hanna’s current observed VMAX is <14 m s−1, both Control and C2 assimilate observations into the same background fields, which are interpolated to the HWRF grid from the 6-h GDAS forecast (section 2b). Therefore, differences between the Control and C2 0600 UTC 23 July analyses should only result from the four COSMIC-2 profiles assimilated in that cycle. Figure 10c shows that the two COSMIC-2 profiles assimilated close to Hanna’s center (i.e., profiles 844 and 12) have a small <0.4 g kg−1 drying effect on the 950-hPa analysis. However, at 800 hPa, their combined impact is much stronger; here, they increase SPFH by 0.2–1 g kg−1 throughout Hanna’s western inner-core region. Their DA impact on 600-hPa SPFH is smaller (Fig. 10i), although profile 844 reduces it by 0.2–1 g kg−1 over a localized area where Control has a moist bias (Fig. 10g).

Fig. 10.
Fig. 10.

(a) HWRF Control analysis minus ERA5 950-hPa qυ difference (shaded; g kg−1) with ERA5 950-hPa qυ (contours; g kg−1) and horizontal wind vectors (m s−1), valid at 0600 UTC 23 Jul 2020. (b) Control analysis minus background 950-hPa qυ difference (shaded; g kg−1) with Control analysis horizontal wind vectors (m s−1), valid at 0600 UTC 23 Jul 2020. (c) HWRF C2 minus Control analysis 950-hPa qυ difference (shaded; g kg−1) with C2 analysis horizontal wind vectors (m s−1), valid at 0600 UTC 23 Jul 2020. COSMIC-2 profiles assimilated in the 0600 UTC 23 Jul C2 analysis cycle are also shown, labeled by the identification number; magenta (green) colors show portions of each profile below (above) 800 hPa. (d)–(f) As in (a)–(c), but for the 800-hPa level. (g)–(i) As in (a)–(c), but for the 600-hPa level.

Citation: Monthly Weather Review 151, 7; 10.1175/MWR-D-22-0198.1

2) 0600 UTC 23 July initialized free forecast

Figures 11a and 11b compare the time evolution of azimuthally averaged 700–850-hPa layer-mean RH (shaded) and 800-hPa SPFH (contoured) from the Control and C2 0600 UTC 23 July initialized free forecasts, respectively. The pattern of a moister inner-core C2 analysis relative to Control over this layer (Figs. 9a and 10f) persists through the first 18 h, when C2 has ∼3%–6% higher RH and ∼0.5 g kg−1 higher SPFH inside of r ∼ 200 km (Box 1). Closer to the boundary layer, the C2 850–950-hPa layer inside of r = 150 km becomes moister compared to Control by t = 12 h (Figs. 11c,d), coincident with C2 developing stronger lower tropospheric azimuthally averaged radial inflows and higher composite reflectivity in the r = 60–150-km radial band (Box 2; Figs. 11c–f). We speculate that C2’s moister 700–850-hPa layer environment between r = 60 and 150 km could have facilitated more widespread convective development there over the first 12 forecast hours, which in turn could have helped increase low-level convergence and inward advection of moist air near the boundary layer, assuming a limited low-level drying impact from convective downdrafts.

Fig. 11.
Fig. 11.

(a) Hovmöller plot showing the radius–time dependence of the Hurricane Hanna 0600 UTC 23 Jul initialized Control forecast RH, azimuthally averaged and vertically averaged over the 700–850-hPa layer (shaded; %), 800-hPa azimuthal mean SPFH (contours; g kg−1), and the 700–850-hPa layer-averaged azimuthal mean radial wind component (VR; vectors; m s−1). (c) As in (a), but for 850–950-hPa mean RH and VR, with 925-hPa SPFH. (e) As in (a), but for composite reflectivity (shaded; dBZ) and the 925-hP atangential wind component (VT; contours; m s−1). (g) As in (a), but for 700-hPa relative vorticity (×104 s−1; shaded) and 400–700-hPa layer-averaged w (contours; m s−1; thin dotted for −0.1, thin solid for 0.1, thick solid for 0.3, 0.5, and 1.0). (b),(d),(f),(h) As in (a), (c), (e), and (g), respectively, but for the Hurricane Hanna 0600 UTC 23 Jul initialized C2 forecast. Note the smaller radial range shown for (g) and (h) compared to the other panels. Numbered rectangles highlight periods and regions discussed in the text.

Citation: Monthly Weather Review 151, 7; 10.1175/MWR-D-22-0198.1

Notable differences between the Control and C2 0600 UTC 23 July initialized Hanna free forecast convective structures first appear around t = 24 h (i.e., 0600 UTC 24 July), when an outbreak of deep convection develops very close to the C2 storm’s center inside of r = 30 km (as shown by azimuthal mean composite reflectivity and layer-averaged updraft speeds in Figs. 11f and 10h). Then, 6 hours later, the C2 storm develops a reasonably axisymmetric eyewall near r = 30 km, which persists through the forecasted landfall around t = 48 h, as outlined by the 35-dBZ reflectivity contour in Fig. 11f and w = 0.3 m s−1 contour in Fig. 11h. The C2 storm’s inner-core deep convection outbreak at t = 24 h coincides with an abrupt increase in cyclonic vorticity inside of the developing eyewall, indicating a strengthening midlevel circulation (Fig. 11h). Although the Control storm also developed deep convection near its center coincident with a strengthening midlevel circulation, these features appear 12 h later compared to C2, and they are less intense in the azimuthal mean (cf. Figs. 11e–h). Control’s azimuthally averaged inner-core structure remains poorly defined through the forecast period (Figs. 11e,g).

Comparing the Control and C2 three-dimensional convective structures at t = 18 h, we find that C2’s greater deep convection coverage near Hanna’s center is more in agreement with Special Sensor Microwave Imager and Sounder (SSMIS) microwave observations. However, both Control and C2 differ from the SSMIS observations by focusing the outer rainband convection north and east, rather than south, of the TC center (Figs. 12a–c). Then, 12 hours later, the Control vortex’s inner-core coverage of deep convection has increased (Fig. 12d). The C2 vortex’s low-level wind circulation is stronger and more tightly contracted (Fig. 12e). Unlike Control, C2 has also developed convection very close to the TC center by this time, in better agreement with the SSMIS observations (Fig. 12f); the potential significance of this feature will be discussed in section 5c.

Fig. 12.
Fig. 12.

(a) Composite reflectivity (shaded; dBZ), 400–700-hPa layer-averaged vertical velocity (contours at 0.5 and 3 m s−1), and 850-hPa horizontal flow vectors (m s−1) from the 18-h HWRF Control Hurricane Hanna (2020) forecast initialized at 0600 UTC 23 Jul. (b) As in (a), but for the 18-h HWRF C2 forecast initialized at 0600 UTC 23 Jul. (c) SSMIS passive 91-GHz microwave radiance-derived color composite imagery of developing TC Hanna (2020) and its environment, obtained from a Defense Meteorological Satellite Program overpass valid at 2218 UTC 23 Jul (image provided courtesy of the Naval Research Laboratory Monterey, available for public access at https://www.nrlmry.navy.mil/tc-bin/tc_home2.cgi). (d),(e) As in (a) and (b), but for the 30-h HWRF Control and C2 forecasts, respectively, initialized at 0600 UTC 23 Jul. (f) As in (c), but valid at 1047 UTC 24 Jul. Black (white) triangles denote the approximate HWRF-forecast (observed) TC Hanna center position.

Citation: Monthly Weather Review 151, 7; 10.1175/MWR-D-22-0198.1

b. Hurricane Zeta (2020) 1200 UTC 26 October initialized free forecast

Turning now to the Hurricane Zeta (2020) case, we compare the time evolution of C2 and Control analysis azimuthal mean RH vertically averaged over the 700–850- and 850–950-hPa layers in Figs. 13a and 13c, respectively. The Zeta (2020) C2 analysis RH is generally similar or 1%–3% drier compared to Control in the TC environment outside of the radius of maximum wind (RMW). Some local maxima in the C2-minus-Control RH absolute difference fields first appear radially collocated with COSMIC-2 profile locations and then persist for several subsequent cycles, suggesting that the “memory” of these assimilated observations’ impact is preserved through the model integration. One example is the negative C2 RH anomaly first appearing around a COSMIC-2 profile assimilated at r ∼ 235 km in the 0000 UTC 25 October C2 analysis (Figs. 13a,c).

Fig. 13.
Fig. 13.

As in Fig. 9, but for the Hurricane Zeta (2020) cycled HWRF experiments. Black rectangles in (a) and (c) highlight the period of interest, i.e., the 0600 and 1200 UTC 26 Oct analyses. In (a) and (c), C2 − Control layer-averaged and azimuthally averaged RH differences (%) are dotted-line contoured for values −6%, −3%, and −1%; thin solid line contoured for values 1% and 3%; and thick solid line contoured for values 6%, 9%, and 12%.

Citation: Monthly Weather Review 151, 7; 10.1175/MWR-D-22-0198.1

Figures 13b and 13d show Zeta’s PMIN and VMAX intensities, respectively, generated from C2 and Control free forecasts initialized at 0600 and 1200 UTC 26 October when the intensifying TC is approaching the Yucatán Peninsula. Although these are not the only Zeta (2020) initializations with significant free forecast intensity differences between C2 and Control (not shown), we selected these forecasts to show examples of COSMIC-2 assimilation reducing an HWRF over-intensification bias. The 1200 UTC 26 October initialized Control free forecast begins to deepen Zeta’s PMIN after t = 12 h rapidly, while the storm is still crossing the Yucatán (Figs. 16b,c), and this intensification continues uninterrupted over the Gulf of Mexico until shortly before landfall on the U.S. Gulf Coast when a peak PMIN and VMAX of 949 hPa and 55 m s−1, respectively, are achieved. Although the 1200 UTC 26 October initialized C2 free forecast eventually intensifies Zeta at a similar rate, its intensity changes track more closely with the best track data, particularly for PMIN, where intensification onset is delayed until after t = 24 h, leading to a peak intensity of 963 hPa. Since both the Control and C2 1200 UTC 26 October initialized forecasts bring Zeta over the northeast Yucatán in nearly identical tracks (not shown), their intensity forecast differences likely do not result from differences in land interaction.

Figures 14a, 14d, and 14g plot the 1200 UTC 26 October Control analysis SPFH bias against ERA5, shown as horizontal cross sections through the 900-, 750-, and 600-hPa levels, respectively. The lower-to-midtroposphere ERA5 SPFH field near Zeta features a deep moist pocket surrounding the TC center that extends northwestward over the Yucatán and a dry tongue wrapping northeastward from the Honduras region into Zeta’s circulation about 200 km east of the center (Figs. 14a,d). Control poorly captures this dry tongue, particularly at 900 hPa, and keeps its source region over Honduras and Belize too moist (Figs. 14a,d). Figures 14b, 14e, and 14h show this cycle’s C2-minus-Control background SPFH differences at the same three levels. The shaded region extending out to r ∼ 300 km approximately demarcates the area covered by the HWRF 6-h forecast vortex, which has been relocated and modified to better match the real-time position, intensity, and size estimates (section 2b). Despite the vortex improvement applied to both, differences between Control and C2 background SPFH are evident, likely a result of COSMIC-2 DA and model advances over prior cycles. Figures 14c, 14f, and 14i show the 1200 UTC 26 October C2-minus-Control analysis SPFH differences at the three levels overlaid by locations of the four COSMIC-2 profiles (profile index of 361, 928, 933, and 1218) assimilated into HWRF C2’s ghost d03 in this cycle. Changes to the SPFH difference fields shown in the right-hand and middle columns of Fig. 14 should reflect DA impacts of the four COSMIC-2 profiles assimilated in this cycle (although the discrepancies could also reflect some impacts of assimilated non-COSMIC-2 observations, given the different C2 and Control background states). Among them, profiles 361 and 1218 are located in Zeta’s southwestern outer circulation near the Central American coast (boxed region in Figs. 14d–f) where the Control analysis generally has a 1–3 g kg−1 moist bias at 750 hPa (Fig. 14d). Here, the 750-hPa C2 background SPFH is up to 1 g kg−1 drier compared to the Control background (Fig. 14e). However, the 750-hPa C2 analysis SPFH is ∼1–4 g kg−1 lower than the Control analysis SPFH (Fig. 14f), suggesting that profiles 361 and 1218 are having a local drying impact on the midtroposphere, helping to correct the moist Control bias with respect to ERA5.

Fig. 14.
Fig. 14.

As in Fig. 10, but for the Hurricane Zeta (2020) HWRF Control and C2 1200 UTC 26 Oct cycled analysis times at the (a)–(c) 900-, (d)–(f) 750-, and (g)–(i) 600-hPa levels. Unlike Fig. 10, (b), (e), and (h) show the C2 6-h background forecast horizontal wind vectors (m s−1) and the C2-minus-Control 6-h background forecast qV difference (g kg−1; shaded); the −3 and −1 g kg−1 and 1 and 3 g kg−1 contours of these difference fields are also shown in purple and black, respectively, in (c), (f), and (i). The rectangular box shown in (d)–(f) highlights a region discussed in the text.

Citation: Monthly Weather Review 151, 7; 10.1175/MWR-D-22-0198.1

Figure 15 compares the Hovmöller time series of selected azimuthally averaged fields from the 1200 UTC 26 October initialized C2 and Control free forecasts. Compared to Control, portions of the C2 vortex are slightly drier over the first 24 h, namely, in the 700–850-hPa layer r = 130–300-km radial band (Box 1 in Figs. 15a,b) and the 850–950-hPa layer r = 110–210-km radial band (Box 2 in Figs. 15c,d). These azimuthally averaged moisture differences reflect C2’s more vigorous lower-to-midtroposphere dry air intrusion into the eastern circulation ∼200 km from Zeta’s center over the first 18 h that appears to originate from the Belize and Honduras coast region (Fig. 16) where COSMIC-2 profiles 361 and 1218 had a drying impact on the 750-hPa C2 analysis (Figs. 14e,f).

Fig. 15.
Fig. 15.

As in Fig. 11, but for the (a),(c),(e),(g) Control and (b),(d),(f),(h) C2 HWRF forecasts for Hurricane Zeta (2020) initialized at 1200 UTC 26 Oct. For (g) and (h), w is thin-dotted contoured for –0.5 and −0.1 m s−1, thin-solid contoured for 0.3 m s−1, and thick-solid contoured for 0.6 and 1.0 m s−1. Also note the smaller radial ranges shown for (e) and (f) and for (g) and (h) compared to the other panels.

Citation: Monthly Weather Review 151, 7; 10.1175/MWR-D-22-0198.1

Fig. 16.
Fig. 16.

(a) HWRF Control Hurricane Zeta (2020) 700–850-hPa layer-averaged RH (shaded; %) and 750-hPa horizontal wind vectors (m s−1) from the 1200 UTC 26 Oct cycled analysis. (b),(c) As in (a), but for the 12- and 18-h verification times, respectively, from the Control forecast initialized from the 1200 UTC 26 Oct cycled analysis. (d)–(f) As in (a)–(c), but for the HWRF C2 forecast initialized from the 1200 UTC 26 Oct cycled analysis. Rectangular boxes highlight the dry tongue discussed in the text.

Citation: Monthly Weather Review 151, 7; 10.1175/MWR-D-22-0198.1

Although C2 forecasts a stronger storm compared to Control through t = 24 h (Fig. 13b; cf. Figs. 15e,f), the Control storm becomes the stronger of the two afterward (Figs. 13b,d), coincident with a more rapid contraction of the Control storm’s RMW and associated eyewall deep convection and updraft core (see Box 3 in Figs. 15e–h). The more tightly contracted eyewall in Control versus C2 beginning around t = 24 h is also evident when comparing their composite reflectivity, 925-hPa wind vectors, and eyewall updraft vertical velocity horizontal distributions in Fig. 17. A comparison of their forecast convective structures to SSMIS microwave observations reveals that (i) both Control and C2 capture Zeta’s observed inner core and outer rainband structure reasonably well prior to the Yucatán landfall (Figs. 17a,d,g); (ii) both forecasts move the storm too quickly off the northern Yucatán coast (Figs. 17b,e,h), which may contribute to them both under-forecasting the storm’s weakening over land (Figs. 13b,d); and (iii) the C2 storm’s less sharply contracted ∼1° wide eyewall convective ring following the storm’s re-emergence in the Gulf of Mexico, compared to Control, is more consistent with the SSMIS observations (Figs. 17c,f,i).

Fig. 17.
Fig. 17.

(a)–(c) As in Fig. 12a, but for the Control 12-, 24-, and 30-h Hurricane Zeta (2020) forecasts initialized at 1200 UTC 26 Oct, respectively, with 925-hPa Control forecast horizontal wind vectors (m s−1). (d)–(f) As in (a)–(c), but for the C2 Hurricane Zeta (2020) 12-, 24-, and 30-h forecasts initialized at 1200 UTC 26 Oct, respectively. (g)–(i) As in Fig. 12c, but for SSMIS 91-GHz color composite imagery valid at 2225 UTC 26 Oct, 1058 UTC 27 Oct, and 0030 UTC 28 Oct, respectively (images provided courtesy of the Naval Research Laboratory Monterey, available for public access at https://www.nrlmry.navy.mil/tc-bin/tc_home2.cgi). The white triangle in (h) marks the approximate NHC best track center position at 1200 UTC 27 Oct (21.3°N, 89.0°W).

Citation: Monthly Weather Review 151, 7; 10.1175/MWR-D-22-0198.1

c. Discussion

The Hurricane Hanna (2020) 0600 UTC 23 July analysis cycle provides the most direct evidence of beneficial COSMIC-2 DA impacts because the Control and C2 experiments use the same background forecast provided by the operational GDAS. Two COSMIC-2 profiles assimilated ∼100 km northwest and southwest of the early developing TC’s center increase SPFH by up to 1 g kg−1 locally at 800 hPa, helping to mitigate an HWRF dry bias with respect to ERA5 there (Figs. 10d,f). This is consistent with the 0600 UTC 23 July initialized C2 free forecast generating greater coverage of near-saturated air (i.e., azimuthal mean RH exceeding 90%) below 700 hPa within r = 150 km compared to Control through t = 18 h (Figs. 11a–d). Given that subsaturated lower-to-midtropospheric air can hinder the organization and maintenance of TC inner-core convection through entrainment and/or downdraft enhancement (section 5a), it is plausible that the C2-forecast storm’s moister inner core may have helped facilitate its earlier and more persistent outbreak of deep convection near Hanna’s center compared to Control (cf. Figs. 11e–h; and cf. Figs. 12a,b). Might the low-to-midtropospheric moistening caused by assimilating two COSMIC-2 profiles near Hanna’s inner core at 0600 UTC 23 July be the primary reason why the C2-forecast storm initialized from this cycle better captures Hanna’s intensification compared to Control? A complete answer to this question would necessitate a more comprehensive analysis—including investigating possible COSMIC-2 DA impacts on analysis temperature and horizontal winds (through multivariate covariances)—beyond this study’s scope.

Notable inner-core structural differences between C2 and Control emerge after deep convection develops near the C2 Hanna’s center—including C2’s more rapid and significant 700-hPa cyclonic vorticity intensification (cf. Figs. 11g,h). Chen et al. (2020) similarly showed how, compared to their “GTS” control experiment, cycled WRF-ARW forecasts of developing Typhoon Nuri (2008) that assimilated COSMIC refractivity developed a moister midtroposphere in Nuri’s inner core over a ∼30 h period; this preceded a rapid strengthening of Nuri’s inner-core updrafts coinciding with a midlevel circulation intensification, both of which were poorly captured by GTS (see their Fig. 12). The important role of inner-core convective updrafts in supporting TC genesis and intensification has been well established (Montgomery and Smith 2014). However, questions remain in developing a universal physical explanation of the link between inner-core updraft organization and the cyclonic wind field intensification. They include the relative importance of the axisymmetric balanced dynamical response to eyewall diabatic heating (e.g., Charney and Eliassen 1964; Shapiro and Willoughby 1982) versus three-dimensional processes, such as (i) the updraft-facilitated aggregation, stretching, and tilting of low-level cyclonic vorticity anomalies (e.g., Hendricks et al. 2004; Van Sang et al. 2008; Montgomery and Smith 2014), (ii) inward-directed compensating subsidence contributing to eye warming, which relates to the VT field spinup through thermal wind balance (Heymsfield et al. 2001; Chen and Zhang 2013), or (iii) updraft-assisted vortex vertical alignment (Molinari et al. 2006; Rogers et al. 2015). The relative importance of inner-core deep convective updrafts (Nolan 2007; McFarquhar et al. 2012; Rogers et al. 2013, 2015) versus shallow convective rings (Zagrodnik and Jiang 2014) as TC intensification precursors also remains an open question.

In a balanced vortex theoretical framework, diabatic heating in updrafts inside the RMW can be more efficiently converted to cyclonic wind kinetic energy due to inertial stability being maximized there (Hack and Schubert 1986). Rogers et al. (2013) used observational composites to show that deep convection focused inside the RMW was a characteristic of intensifying TCs. Similarly, we find that for the 0600 UTC 23 July initialized C2 forecast, the previously mentioned deep convection outbreak persists for 12 h inside of the developing RMW, which contrasts with Control’s generating weaker and more transient convection inside its RMW around t = 36 h (cf. their 40-dBZ reflectivity contours in Figs. 11e,f). Nolan (2007)’s idealized simulations of early-developing TCs similarly showed that both deep tropospheric moistening of the inner core as well as a sufficiently strong and inertially stable midlevel vortex were necessary precursors for tropical cyclogenesis, which coincided with the sudden development of a tightly contracted low-level circulation embedded within the broader parent vortex.

The Hurricane Zeta (2020) 1200 UTC 26 October analysis cycle presents an additional perspective because it differs from the 0600 UTC 23 July Hanna cycle in several ways, including (i) a more mature initial TC, (ii) different C2 and Control background states due to prior cycling, and (iii) the C2 free forecast reducing a Control over-intensification bias. Here, in addition to prior background cycling, two COSMIC-2 profiles assimilated in the southwestern outer circulation likely help reduce nearby 750-hPa SPFH in the C2 analysis, moving it in closer agreement with ERA5 (Figs. 15d–f). This region, near the Central American landmass, appears to be the source of low-to-midtroposphere dry air intrusion that wraps into the east side of Zeta’s circulation over the first 18 free forecast hours; the dry air intrusion is more pronounced in C2, consistent with the analysis differences (Fig. 16). Although both the Control and C2 inner-core regions within r ∼ 150 km remain relatively “protected” by a moist envelope throughout this period, it is possible that the stronger C2 dry air intrusion helped to put a brake on the storm’s intensification as it quickly crossed the northeast corner of the Yucatán and emerged into the Gulf of Mexico. Interestingly, a more tightly contracted eyewall compared to C2 appears in the Control storm after it re-emerges over water (cf. Figs. 17b,e and cf. Figs. 17c,f). Assuming that inward advection of absolute angular momentum overcomes frictional losses, eyewall contraction implies boundary layer VT intensification (Smith et al. 2017). Braun et al. (2012)’s idealized modeling study showed that dry air intrusions close to a TC center could delay intensification by introducing inner rainband asymmetries that temporarily moved convective heating outward to a lower inertial stability region. Further investigation of potential causal relationships between the more robust dry air intrusion, reduced eyewall contraction, and reduced over-intensification bias in the 1200 UTC 26 October initialized C2 Zeta free forecast will be deferred to a future study.

6. Summary and conclusions

We have analyzed COSMIC-2 bending angle assimilation impacts on an offline HWRF-GSI v4.0 system using cycled HWRF forecasting experiments run for six 2020 Atlantic hurricane cases. The pre and post-QC COSMIC-2 (OB)/O mean and RMSD profiles broadly resembled those reported by Lien et al. (2021) for the global CWB-GFS, with a negative bias in the lower troposphere and a maximum RMSD in the lower to midtroposphere. The COSMIC-2 σrep and QC parameters used here were set to values previously tuned for COSMIC bending angle DA in GSI. Diagnostics revealed that the prescribed COSMIC-2 σrep may be too small for the 5–8-km impact height layer. The SR and Statistical QC checks accounted for nearly all lower-troposphere COSMIC-2 observation rejections, which were most likely to occur in regions of high background SPFH exceeding 16 g kg−1.

Comparing Control and C2 mean absolute track errors for all 108 free forecasts drawn from the six cases, differences were small until very late in the forecast period after t = 102 h, when the C2 forecast improvement was statistically significant. However, since it is well known that larger-scale flows typically control TC motion, the relatively muted impact of COSMIC-2 DA on the HWRF track forecasts could in part result from the fact that HWRF v4.0 replaces background fields outside of the vortex region with GDAS fields, thereby reducing the cycled impact of any observations assimilated outside of the TC vortex circulation. On the other hand, Control mean absolute PMIN intensity forecast errors were reduced in C2 by a statistically significant (90% confidence interval) ∼8%–12% for t = 36, 54, 60, and 108–120 h. COSMIC-2 DA also reduces the Control SPFH RMSD against radiosondes and dropwindsondes by 1%–4% in the 600–700-hPa layer over the t = 12–54-h period; this improvement is not statistically significant, and the sample size is relatively small.

Focusing on the Hanna (2020) and Zeta (2020) case experiments, we found that COSMIC-2 bending angle assimilation and subsequent model advances resulted in changes to their azimuthally averaged analysis RH fields below 700 hPa that persisted for multiple cycles. For Hanna, the C2 analysis 700–850-hPa layer azimuthal mean RH inside of r = 150 km exceeded that of Control by >1% beginning with the fourth cycle (0600 UTC 23 July), when two COSMIC-2 profiles were assimilated near (then Tropical Depression) Hanna’s center. The C2 intensity forecast initialized from this analysis significantly outperformed its Control counterpart by reducing a severe low-intensity bias. It is possible that the improved C2 intensity forecast could have resulted in part from its moister inner core facilitating an earlier and more organized/persistent outbreak of deep convection near the developing TC’s center where updraft latent heating is more efficiently converted to the VT field’s kinetic energy. This C2-forecast event immediately preceded the development of an eyewall convective ring, which remained more poorly organized in the Control forecast. Also focusing on the 1200 UTC 26 October Zeta (2020) analysis cycle, we found that assimilation of two COSMIC-2 profiles in the TC’s southwestern outer circulation likely helped to reduce a 750-hPa moist Control bias with respect to ERA5, which could have contributed to C2 having a more robust dry air intrusion and a reduced over-intensification bias compared to Control.

Our results suggest that although COSMIC-2 bending angle assimilation, on average, has a neutral to modestly positive impact on HWRF TC intensity forecasts, more substantial intensity forecast improvement is possible if the HWRF background (i) assimilates COSMIC-2 profiles near the TC inner core and (ii) has a biased midtroposphere moisture field. Although these results are encouraging, our offline HWRF-GSI system only assimilates clear-sky satellite microwave and infrared radiances; thus, it is unclear how much benefit COSMIC-2 DA could provide if HWRF could assimilate radiances in cloudy precipitating regions. Several studies have shown that assimilating microwave or infrared radiances from cloudy regions into regional NWP models can improve forecast TC convective structure and intensity, compared to assimilating clear-sky radiances only (Minamide and Zhang 2018; Wu et al. 2019). However, cloudy radiance assimilation in regional cloud-resolving NWP models presents several unique challenges, such as sensitivity to the microphysics scheme, possible imbalances between updated cloud microphysical and dynamical fields, and bias correction (Wu et al. 2019). We speculate that further intensity forecast benefits could be realized if HWRF-GSI’s RO observation QC algorithm could be tuned to allow greater utilization of good-quality lower-troposphere COSMIC-2 bending angles; a future study could test implementation of a local spectral width (LSW)-based QC criterion that screens out observations with higher wave optics retrieval uncertainty resulting from greater horizontal moisture inhomogeneities along their observed ray paths (Liu et al. 2018). Further work is also needed to better understand HWRF-GSI’s NBAM forward operator uncertainty in the moist tropical troposphere, and to test whether modifying HWRF’s COSMIC-2 RO bending angle observation error specification to account for additional factors such as retrieval uncertainty (Zhang et al. 2023) and representativeness error dependence on background meteorological fields (e.g., as in Bowler 2020) can further improve HWRF analyses and forecasts. Future work exploring the impact of RO observations from COSMIC-2 and other platforms on TC precipitation forecasts would also be beneficial, given that inland flooding is a significant TC-related hazard. Our results suggest that increased utilization of good-quality RO data from additional receiver satellite platforms, including commercial sources, may provide further value to regional NWP model TC intensity forecasts.

1

Despite the 24-hr time period used in Kaplan and DeMaria (2003)’s RI criterion, numerous TC research studies have defined RI (or “ordinary” intensification, typically using a 20 kt day−1 threshold) over a 12-h time window (e.g., McFarquhar et al. 2012; Rogers et al. 2013; Christophersen et al. 2018; Wadler et al. 2018), arguing that this period length better captures inner-core convective-scale and vortex-scale processes as opposed to environmental influences.

2

The default HWRF-GSI v4.0a RO observation forward operator code used in this study predates the updated NBAM code used for assimilating RO bending angles, including those from COSMIC-2, in the current operational GFS. These modifications include slight improvements to the forward operator and quality controls, as well as COSMIC-2 specific observation error and quality control settings (L. Cucurull 2022, personal communication).

3

In this section we show DA statistics from ghost d03 COSMIC-2 observations rather than from the larger set of ghost d02 COSMIC-2 observations (cf. Figs. 2a,b) because the former are expected to have a larger impact on the cycled TC vortex evolution given HWRF’s design (section 2b). We have repeated the analysis described herein for the set of COSMIC-2 observations assimilated in ghost d02 (not shown) and the results do not differ substantially from those shown in Figs. 3 and 4.

4

The GFDL vortex tracker also outputs TC intensity in terms of VMAX, which it computes as the gridpoint-maximum z = 10-m HWRF output wind speed. However, we do not use it for our statistical intensity forecast verification because we found it to be quite noisy, probably due to its sensitivity to instantaneous gusts generated on the high-resolution grid. The NHC best track VMAX is estimated as a 1-min-average sustained wind. Since the temporal resolution of our HWRF output is 3 h, it would be difficult for us to generate a temporally smoothed VMAX postprocessed product suitable for comparison against the best track VMAX.

Acknowledgments.

This study was supported by NOAA Grant NA19NES4320002 [Cooperative Institute for Satellite Earth System Studies (CISESS)] at the University of Maryland/Earth System Science Interdisciplinary Center (ESSIC). The scientific results and conclusions, as well as any views or opinions expressed herein, are those of the author(s) and do not necessarily reflect those of the NOAA or the Department of Commerce. We thank Dr. James Yoe for his very helpful feedback when internally reviewing this manuscript.

Data availability statement.

The HWRF simulation datasets and GSI data assimilation diagnostic files used for generating the figures shown herein are stored on the NOAA Research and Development HPC System’s High Performance Storage System (HPSS) drive and can be made available upon request.

REFERENCES

  • Berg, R., and B. J. Reinhart, 2021: Tropical cyclone report: Hurricane Sally (11–17 September 2020). NHC Tech. Rep. AL192020, 69 pp., https://www.nhc.noaa.gov/data/tcr/AL192020_Sally.pdf.

  • Biswas, M. K., L. Carson, K. Newman, D. Stark, E. Kalina, E. Grell, and J. Frimel, 2018a: Community HWRF users’ guide V4.0a. Developmental Testbed Center, 163 pp., https://dtcenter.org/sites/default/files/community-code/hwrf/docs/users_guide/HWRF-UG-2018.pdf.

  • Biswas, M. K., and Coauthors, 2018b: Hurricane Weather Research and Forecasting (HWRF) model: 2018 Scientific Documentation. Developmental Testbed Center, 112 pp., https://dtcenter.org/sites/default/files/community-code/hwrf/docs/scientific_documents/HWRFv4.0a_ScientificDoc.pdf.

  • Blake, E., R. Berg, and A. Hagen, 2021: Tropical cyclone report: Hurricane Zeta (24–29 October 2020). NHC Tech. Rep. AL282020, 56 pp., https://www.nhc.noaa.gov/data/tcr/AL282020_Zeta.pdf.

  • Bowler, N. E., 2020: Revised GNSS-RO observation uncertainties in the Met Office NWP system. Quart. J. Roy. Meteor. Soc., 146, 22742296, https://doi.org/10.1002/qj.3791.

    • Search Google Scholar
    • Export Citation
  • Braun, S. A., J. A. Sippel, and D. S. Nolan, 2012: The impact of dry midlevel air on hurricane intensity in idealized simulations with no mean flow. J. Atmos. Sci., 69, 236257, https://doi.org/10.1175/JAS-D-10-05007.1.

    • Search Google Scholar
    • Export Citation
  • Brown, D. P., R. Berg, and B. Reinhart, 2021: Tropical cyclone report: Hurricane Hanna (23–26 July 2020). NHC Tech. Rep. AL082020, 49 pp., https://www.nhc.noaa.gov/data/tcr/AL082020_Hanna.pdf.

  • Cangialosi, J. P., and R. Berg, 2021: Tropical cyclone report: Hurricane Delta (4–10 October 2020). NHC Tech. Rep. AL262020, 46 pp., https://www.nhc.noaa.gov/data/tcr/AL262020_Delta.pdf.

  • Cangialosi, J. P., E. Blake, M. DeMaria, A. Penny, A. Latto, E. Rappaport, and V. Tallapragada, 2020: Recent progress in tropical cyclone intensity forecasting at the National Hurricane Center. Wea. Forecasting, 35, 19131922, https://doi.org/10.1175/WAF-D-20-0059.1.

    • Search Google Scholar
    • Export Citation
  • Chan, J. C. L., and W. M. Gray, 1982: Tropical cyclone movement and surrounding flow relationships. Mon. Wea. Rev., 110, 13541374, https://doi.org/10.1175/1520-0493(1982)110<1354:TCMASF>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Chan, J. C. L., F. M. F. Ko, and Y. M. Lei, 2002: Relationship between potential vorticity tendency and tropical cyclone motion. J. Atmos. Sci., 59, 13171336, https://doi.org/10.1175/1520-0469(2002)059<1317:RBPVTA>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Charney, J. G., and A. Eliassen, 1964: On the growth of the hurricane depression. J. Atmos. Sci., 21, 6875, https://doi.org/10.1175/1520-0469(1964)021<0068:OTGOTH>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Chen, H., and D.-L. Zhang, 2013: On the rapid intensification of Hurricane Wilma (2005). Part II: Convective bursts and the upper-level warm core. J. Atmos. Sci., 70, 146162, https://doi.org/10.1175/JAS-D-12-062.1.

    • Search Google Scholar
    • Export Citation
  • Chen, S.-Y., Y.-H. Kuo, and C.-Y. Huang, 2020: The impact of GPS RO data on predicting tropical cyclogenesis using a nonlocal observation operator: An initial assessment. Mon. Wea. Rev., 148, 27012717, https://doi.org/10.1175/MWR-D-19-0286.1.

    • Search Google Scholar
    • Export Citation
  • Chen, S.-Y., T.-C. Nguyen, and C.-Y. Huang, 2021: Impact of radio occultation data on the prediction of Typhoon Haishen (2020) with WRFDA hybrid assimilation. Atmosphere, 12, 1397, https://doi.org/10.3390/atmos12111397.

    • Search Google Scholar
    • Export Citation
  • Chen, Y.-C., M.-E. Hsieh, L.-F. Hsiao, Y.-H. Kuo, M.-J. Yang, C.-Y. Huang, and C.-S. Lee, 2015: Systematic evaluation of the impacts of GPSRO data on the prediction of typhoons over the northwestern Pacific in 2008–2010. Atmos. Meas. Tech., 8, 25312542, https://doi.org/10.5194/amt-8-2531-2015.

    • Search Google Scholar
    • Export Citation
  • Chen, Y.-C., C.-C. Tsai, Y.-C. Wu, A.-H. Wang, C.-J. Wang, H.-H. Lin, D.-R. Chen, and Y.-C. Yu, 2021: Evaluation of operational monsoon moisture surveillance and severe weather prediction utilizing COSMIC-2/FORMOSAT-7 radio occultation observations. Remote Sens., 13, 2979, https://doi.org/10.3390/rs13152979.

    • Search Google Scholar
    • Export Citation
  • Chien, T.-Y., S.-Y. Chen, C.-Y. Huang, C.-P. Shih, C. S. Schwartz, Z. Liu, J. Bresch, and J.-Y. Lin, 2022: Impacts of radio occultation data on typhoon forecasts as explored by the global MPAS-GSI System. Atmosphere, 13, 1353, https://doi.org/10.3390/atmos13091353.

    • Search Google Scholar
    • Export Citation
  • Christophersen, H., A. Aksoy, J. Dunion, and S. Aberson, 2018: Composite impact of Global Hawk unmanned aircraft dropwindsondes on tropical cyclone analyses and forecasts. Mon. Wea. Rev., 146, 22972314, https://doi.org/10.1175/MWR-D-17-0304.1.

    • Search Google Scholar
    • Export Citation
  • Cram, T. A., J. Persing, M. T. Montgomery, and S. A. Braun, 2007: A Lagrangian trajectory view on transport and mixing processes between the eye, eyewall, and environment using a high-resolution simulation of Hurricane Bonnie (1998). J. Atmos. Sci., 64, 18351856, https://doi.org/10.1175/JAS3921.1.

    • Search Google Scholar
    • Export Citation
  • Cucurull, L., 2010: Improvement in using an operational constellation of GPS radio occultation receivers in weather forecasting. Wea. Forecasting, 25, 749767, https://doi.org/10.1175/2009WAF2222302.1.

    • Search Google Scholar
    • Export Citation
  • Cucurull, L., 2015: Implementation of quality control for radio occultation observations in the presence of large gradients of atmospheric refractivity. Atmos. Meas. Tech., 8, 12751285, https://doi.org/10.5194/amt-8-1275-2015.

    • Search Google Scholar
    • Export Citation
  • Cucurull, L., J. C. Derber, and R. J. Purser, 2013: A bending angle forward operator for global positioning system radio occultation measurements. J. Geophys. Res. Atmos., 118, 1428, https://doi.org/10.1029/2012JD017782.

    • Search Google Scholar
    • Export Citation
  • DeMaria, M., M. Mainelli, L. K. Shay, J. A. Knaff, and J. Kaplan, 2005: Further improvements to the Statistical Hurricane Intensity Prediction Scheme (SHIPS). Wea. Forecasting, 20, 531543, https://doi.org/10.1175/WAF862.1.

    • Search Google Scholar
    • Export Citation
  • Desroziers, G., L. Berre, B. Chapnik, and P. Poli, 2005: Diagnosis of observation, background and analysis-error statistics in observation space. Quart. J. Roy. Meteor. Soc., 131, 33853396, https://doi.org/10.1256/qj.05.108.

    • Search Google Scholar
    • Export Citation
  • Doyle, J. D., C. A. Reynolds, C. Amerault, and J. Moskaitis, 2012: Adjoint sensitivity and predictability of tropical cyclogenesis. J. Atmos. Sci., 69, 35353557, https://doi.org/10.1175/JAS-D-12-0110.1.

    • Search Google Scholar
    • Export Citation
  • Emanuel, K. A., 1989: The finite-amplitude nature of tropical cyclogenesis. J. Atmos. Sci., 46, 34313456, https://doi.org/10.1175/1520-0469(1989)046<3431:TFANOT>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Fritz, C., and Z. Wang, 2013: A numerical study of the impacts of dry air on tropical cyclone formation: A development case and a nondevelopment case. J. Atmos. Sci., 70, 91111, https://doi.org/10.1175/JAS-D-12-018.1.

    • Search Google Scholar
    • Export Citation
  • Galarneau, T. J., Jr., and C. A. Davis, 2013: Diagnosing forecast errors in tropical cyclone motion. Mon. Wea. Rev., 141, 405430, https://doi.org/10.1175/MWR-D-12-00071.1.

    • Search Google Scholar
    • Export Citation
  • Hack, J. J., and W. H. Schubert, 1986: Nonlinear response of atmospheric vortices to heating by organized cumulus convection. J. Atmos. Sci., 43, 15591573, https://doi.org/10.1175/1520-0469(1986)043<1559:NROAVT>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Healy, S. B., J. R. Eyre, M. Hamrud, and J.-N. Thépaut, 2007: Assimilating GPS radio occultation measurements with two-dimensional bending angle observation operators. Quart. J. Roy. Meteor. Soc., 133, 12131227, https://doi.org/10.1002/qj.63.

    • Search Google Scholar
    • Export Citation
  • Hendricks, E. A., M. T. Montgomery, and C. A. Davis, 2004: The role of “vortical” hot towers in the formation of Tropical Cyclone Diana (1984). J. Atmos. Sci., 61, 12091232, https://doi.org/10.1175/1520-0469(2004)061<1209:TROVHT>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Heymsfield, G. M., J. B. Halverson, J. Simpson, L. Tian, and T. P. Bui, 2001: ER-2 Doppler radar investigations of the eyewall of Hurricane Bonnie during the Convection and Moisture Experiment-3. J. Appl. Meteor., 40, 13101330, https://doi.org/10.1175/1520-0450(2001)040<1310:EDRIOT>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Ho, S.-P., and Coauthors, 2020a: The COSMIC/FORMOSAT-3 radio occultation mission after 12 years: Accomplishments, remaining challenges, and potential impacts of COSMIC-2. Bull. Amer. Meteor. Soc., 101, E1107E1136, https://doi.org/10.1175/BAMS-D-18-0290.1.

    • Search Google Scholar
    • Export Citation
  • Ho, S.-P., and Coauthors, 2020b: Initial assessment of the COSMIC-2/FORMOSAT-7 neutral atmosphere data quality in NESDIS/STAR using in-situ and satellite data. Remote Sens., 12, 4099, https://doi.org/10.3390/rs12244099.

    • Search Google Scholar
    • Export Citation
  • Huang, C.-Y., Y.-H. Kuo, S.-H. Chen, and F. Vandenberghe, 2005: Improvements in typhoon forecasts with assimilated GPS occultation refractivity. Wea. Forecasting, 20, 931953, https://doi.org/10.1175/WAF874.1.

    • Search Google Scholar
    • Export Citation
  • Huang, C.-Y., and Coauthors, 2010: Impact of GPS radio occultation data assimilation on regional weather predictions. GPS Solutions, 14, 3549, https://doi.org/10.1007/s10291-009-0144-1.

    • Search Google Scholar
    • Export Citation
  • Jarvinen, B. R., C. J. Neumann, and M. A. S. Davis, 1984: A tropical cyclone data tape for the North Atlantic basin, 1886–1983: Contents, limitations, and uses. NOAA Tech. Memo. NWS NHC 22, 24 pp., https://www.nhc.noaa.gov/pdf/NWS-NHC-1988-22.pdf.

  • Kaplan, J., and M. DeMaria, 2003: Large-scale characteristics of rapidly intensifying tropical cyclones in the North Atlantic basin. Wea. Forecasting, 18, 10931108, https://doi.org/10.1175/1520-0434(2003)018<1093:LCORIT>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Kueh, M.-T., C.-Y. Huang, S.-Y. Chen, S.-H. Chen, and C.-J. Wang, 2009: Impact of GPS RO refractivity soundings on a simulation of Typhoon Bilis (2006) upon landfall. Terr. Atmos. Oceanic Sci., 20, 115131, https://doi.org/10.3319/TAO.2008.01.21.03(F3C).

    • Search Google Scholar
    • Export Citation
  • Kuo, Y.-H., T.-K. Wee, S. Sokolovskiy, C. Rocken, W. Schreiner, D. Hunt, and R. A. Anthes, 2004: Inversion and error estimation of GPS radio occultation data. J. Meteor. Soc. Japan, 82, 507531, https://doi.org/10.2151/jmsj.2004.507.

    • Search Google Scholar
    • Export Citation
  • Latto, A., A. Hagen, and R. Berg, 2021: Tropical cyclone report: Hurricane Isaias (30 July–4 August 2020). NHC Tech. Rep. AL092020, 84 pp., https://www.nhc.noaa.gov/data/tcr/AL092020_Isaias.pdf.

  • Lien, G.-Y., and Coauthors, 2021: Assimilation impact of early FORMOSAT-7/COSMIC-2 GNSS radio occultation data with Taiwan’s CWB Global Forecast System. Mon. Wea. Rev., 149, 21712191, https://doi.org/10.1175/MWR-D-20-0267.1.

    • Search Google Scholar
    • Export Citation
  • Liu, H., J. Anderson, and Y.-H. Kuo, 2012: Improved analyses and forecasts of Hurricane Ernesto’s genesis using radio occultation data in an ensemble filter assimilation system. Mon. Wea. Rev., 140, 151166, https://doi.org/10.1175/MWR-D-11-00024.1.

    • Search Google Scholar
    • Export Citation
  • Liu, H., Y.-H. Kuo, S. Sokolovskiy, X. Zou, Z. Zeng, L.-F. Hsiao, and B. Ruston, 2018: A quality control procedure based on bending angle measurement uncertainty for radio occultation data assimilation in the tropical lower troposphere. J. Atmos. Oceanic Technol., 35, 21172131, https://doi.org/10.1175/JTECH-D-17-0224.1.

    • Search Google Scholar
    • Export Citation
  • Liu, Q., and Coauthors, 2020: Vortex initialization in the NCEP operational hurricane models. Atmosphere, 11, 968, https://doi.org/10.3390/atmos11090968.

    • Search Google Scholar
    • Export Citation
  • Liu, Z., C. S. Schwartz, C. Snyder, and S.-Y. Ha, 2012: Impact of assimilating AMSU-A radiances on forecasts of 2008 Atlantic tropical cyclones initialized with a limited-area ensemble Kalman filter. Mon. Wea. Rev., 140, 40174034, https://doi.org/10.1175/MWR-D-12-00083.1.

    • Search Google Scholar
    • Export Citation
  • McFarquhar, G. M., B. F. Jewett, M. S. Gilmore, S. W. Nesbitt, and T.-L. Hsieh, 2012: Vertical velocity and microphysical distributions related to rapid intensification in a simulation of Hurricane Dennis (2005). J. Atmos. Sci., 69, 35153534, https://doi.org/10.1175/JAS-D-12-016.1.

    • Search Google Scholar
    • Export Citation
  • Minamide, M., and F. Zhang, 2018: Assimilation of all-sky infrared radiances from Himawari-8 and impacts of moisture and hydrometeor initialization on convection-permitting tropical cyclone prediction. Mon. Wea. Rev., 146, 32413258, https://doi.org/10.1175/MWR-D-17-0367.1.

    • Search Google Scholar
    • Export Citation
  • Molinari, J., P. Dodge, D. Vollaro, K. L. Corbosiero, and F. Marks Jr., 2006: Mesoscale aspects of the downshear reformation of a tropical cyclone. J. Atmos. Sci., 63, 341354, https://doi.org/10.1175/JAS3591.1.

    • Search Google Scholar
    • Export Citation
  • Molinari, J., J. Frank, and D. Vollaro, 2013: Convective bursts, downdraft cooling, and boundary layer recovery in a sheared tropical storm. Mon. Wea. Rev., 141, 10481060, https://doi.org/10.1175/MWR-D-12-00135.1.

    • Search Google Scholar
    • Export Citation
  • Montgomery, M. T., and R. K. Smith, 2014: Paradigms for tropical cyclone intensification. Aust. Meteor. Oceanogr. J., 64, 3766, https://doi.org/10.22499/2.6401.005.

    • Search Google Scholar
    • Export Citation
  • Nolan, D. S., 2007: What is the trigger for tropical cyclogenesis? Aust. Meteor. Mag., 56, 241266.

  • Onderlinde, M. J., and D. S. Nolan, 2016: Tropical cyclone–relative environmental helicity and the pathways to intensification in shear. J. Atmos. Sci., 73, 869890, https://doi.org/10.1175/JAS-D-15-0261.1.

    • Search Google Scholar
    • Export Citation
  • Pasch, R. J., R. Berg, D. P. Roberts, and P. P. Papin, 2021: Tropical cyclone report: Hurricane Laura (20–29 August 2020). NHC Tech. Rep. AL132020, 75 pp., https://www.nhc.noaa.gov/data/tcr/AL132020_Laura.pdf.

  • Rogers, R., P. Reasor, and S. Lorsolo, 2013: Airborne Doppler observations of the inner-core structural differences between intensifying and steady-state tropical cyclones. Mon. Wea. Rev., 141, 29702991, https://doi.org/10.1175/MWR-D-12-00357.1.

    • Search Google Scholar
    • Export Citation
  • Rogers, R., P. Reasor, and J. A. Zhang, 2015: Multiscale structure and evolution of Hurricane Earl (2010) during rapid intensification. Mon. Wea. Rev., 143, 536562, https://doi.org/10.1175/MWR-D-14-00175.1.

    • Search Google Scholar
    • Export Citation
  • Ruston, B., and S. Healy, 2021: Forecast impact of FORMOSAT-7/COSMIC-2 GNSS radio occultation measurements. Atmos. Sci. Lett., 22, e1019, https://doi.org/10.1002/asl.1019.

    • Search Google Scholar
    • Export Citation
  • Schreiner, W. S., and Coauthors, 2020: COSMIC-2 radio occultation constellation: First results. Geophys. Res. Lett., 47, e2019GL086841, https://doi.org/10.1029/2019GL086841.

    • Search Google Scholar
    • Export Citation
  • Shapiro, L. J., and H. E. Willoughby, 1982: The response of balanced hurricanes to local sources of heat and momentum. J. Atmos. Sci., 39, 378394, https://doi.org/10.1175/1520-0469(1982)039<0378:TROBHT>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Sippel, J. A., and F. Zhang, 2008: A probabilistic analysis of the dynamics and predictability of tropical cyclogenesis. J. Atmos. Sci., 65, 34403459, https://doi.org/10.1175/2008JAS2597.1.

    • Search Google Scholar
    • Export Citation
  • Smith, R. K., J. A. Zhang, and M. T. Montgomery, 2017: The dynamics of intensification in a Hurricane Weather Research and Forecasting simulation of Hurricane Earl (2010). Quart. J. Roy. Meteor. Soc., 143, 293308, https://doi.org/10.1002/qj.2922.

    • Search Google Scholar
    • Export Citation
  • Teng, H.-F., Y.-H. Kuo, and J. M. Done, 2021: Importance of midlevel moisture for tropical cyclone formation in easterly and monsoon environments over the western North Pacific. Mon. Wea. Rev., 149, 24492469, https://doi.org/10.1175/MWR-D-20-0313.1.

    • Search Google Scholar
    • Export Citation
  • Teng, H.-F., Y.-H. Kuo, and J. M. Done, 2023: Potential impacts of radio occultation data assimilation on forecast skill of tropical cyclone formation in the western North Pacific. Geophys. Res. Lett., 50, e2021GL096750, https://doi.org/10.1029/2021GL096750.

    • Search Google Scholar
    • Export Citation
  • Trahan, S., and L. Sparling, 2012: An analysis of NCEP tropical cyclone vitals and potential effects on forecasting models. Wea. Forecasting, 27, 744756, https://doi.org/10.1175/WAF-D-11-00063.1.

    • Search Google Scholar
    • Export Citation
  • Van Sang, N., R. K. Smith, and M. T. Montgomery, 2008: Tropical-cyclone intensification and predictability in three dimensions. Quart. J. Roy. Meteor. Soc., 134, 563582, https://doi.org/10.1002/qj.235.

    • Search Google Scholar
    • Export Citation
  • Wadler, J. B., R. F. Rogers, and P. D. Reasor, 2018: The relationship between spatial variations in the structure of convective bursts and tropical cyclone intensification as determined by airborne Doppler radar. Mon. Wea. Rev., 146, 761780, https://doi.org/10.1175/MWR-D-17-0213.1.

    • Search Google Scholar
    • Export Citation
  • Wang, X., D. Parrish, D. Kleist, and J. Whitaker, 2013: GSI 3DVar-based ensemble-variational hybrid data assimilation for NCEP Global Forecast System: Single-resolution experiments. Mon. Wea. Rev., 141, 40984117, https://doi.org/10.1175/MWR-D-12-00141.1.

    • Search Google Scholar
    • Export Citation
  • Weiss, J.-P., W. S. Schreiner, J. J. Braun, W. Xia-Serafino, and C.-Y. Huang, 2022: COSMIC-2 mission summary at three years in orbit. Atmosphere, 13, 1409, https://doi.org/10.3390/atmos13091409.

    • Search Google Scholar
    • Export Citation
  • Wilks, D. S., 2011: Statistical Methods in the Atmospheric Sciences. 3rd ed. International Geophysics Series, Vol. 100, Academic Press, 704 pp.

  • Wu, T.-C., M. Zupanski, L. D. Grasso, C. D. Kummerow, and S.-A. Boukabara, 2019: All-sky radiance assimilation of ATMS in HWRF: A demonstration study. Mon. Wea. Rev., 147, 85106, https://doi.org/10.1175/MWR-D-17-0337.1.

    • Search Google Scholar
    • Export Citation
  • Xie, F., S. Syndergaard, E. R. Kursinski, and B. M. Herman, 2006: An approach for retrieving marine boundary layer refractivity from GPS occultation data in the presence of super-refraction. J. Atmos. Oceanic Technol., 23, 16291644, https://doi.org/10.1175/JTECH1996.1.

    • Search Google Scholar
    • Export Citation
  • Xie, F., D. L. Wu, C. O. Ao, E. R. Kursinski, A. J. Mannucci, and S. Syndergaard, 2010: Super-refraction effects on GPS radio occultation refractivity in marine boundary layers. Geophys. Res. Lett., 37, L11805, https://doi.org/10.1029/2010GL043299.

    • Search Google Scholar
    • Export Citation
  • Zagrodnik, J. P., and H. Jiang, 2014: Rainfall, convection, and latent heating distributions in rapidly intensifying tropical cyclones. J. Atmos. Sci., 71, 27892809, https://doi.org/10.1175/JAS-D-13-0314.1.

    • Search Google Scholar
    • Export Citation
  • Zawislak, J., and Coauthors, 2022: Accomplishments of NOAA’s airborne hurricane field program and a broader future approach to forecast improvement. Bull. Amer. Meteor. Soc., 103, E311E338, https://doi.org/10.1175/BAMS-D-20-0174.1.

    • Search Google Scholar
    • Export Citation
  • Zhang, H., Y.-H. Kuo, and S. Sokolovskiy, 2023: Assimilation of radio occultation data using measurement-based observation error specification: Preliminary results. Mon. Wea. Rev., 151, 589601, https://doi.org/10.1175/MWR-D-22-0122.1.

    • Search Google Scholar
    • Export Citation
Save
  • Berg, R., and B. J. Reinhart, 2021: Tropical cyclone report: Hurricane Sally (11–17 September 2020). NHC Tech. Rep. AL192020, 69 pp., https://www.nhc.noaa.gov/data/tcr/AL192020_Sally.pdf.

  • Biswas, M. K., L. Carson, K. Newman, D. Stark, E. Kalina, E. Grell, and J. Frimel, 2018a: Community HWRF users’ guide V4.0a. Developmental Testbed Center, 163 pp., https://dtcenter.org/sites/default/files/community-code/hwrf/docs/users_guide/HWRF-UG-2018.pdf.

  • Biswas, M. K., and Coauthors, 2018b: Hurricane Weather Research and Forecasting (HWRF) model: 2018 Scientific Documentation. Developmental Testbed Center, 112 pp., https://dtcenter.org/sites/default/files/community-code/hwrf/docs/scientific_documents/HWRFv4.0a_ScientificDoc.pdf.

  • Blake, E., R. Berg, and A. Hagen, 2021: Tropical cyclone report: Hurricane Zeta (24–29 October 2020). NHC Tech. Rep. AL282020, 56 pp., https://www.nhc.noaa.gov/data/tcr/AL282020_Zeta.pdf.

  • Bowler, N. E., 2020: Revised GNSS-RO observation uncertainties in the Met Office NWP system. Quart. J. Roy. Meteor. Soc., 146, 22742296, https://doi.org/10.1002/qj.3791.

    • Search Google Scholar
    • Export Citation
  • Braun, S. A., J. A. Sippel, and D. S. Nolan, 2012: The impact of dry midlevel air on hurricane intensity in idealized simulations with no mean flow. J. Atmos. Sci., 69, 236257, https://doi.org/10.1175/JAS-D-10-05007.1.

    • Search Google Scholar
    • Export Citation
  • Brown, D. P., R. Berg, and B. Reinhart, 2021: Tropical cyclone report: Hurricane Hanna (23–26 July 2020). NHC Tech. Rep. AL082020, 49 pp., https://www.nhc.noaa.gov/data/tcr/AL082020_Hanna.pdf.

  • Cangialosi, J. P., and R. Berg, 2021: Tropical cyclone report: Hurricane Delta (4–10 October 2020). NHC Tech. Rep. AL262020, 46 pp., https://www.nhc.noaa.gov/data/tcr/AL262020_Delta.pdf.

  • Cangialosi, J. P., E. Blake, M. DeMaria, A. Penny, A. Latto, E. Rappaport, and V. Tallapragada, 2020: Recent progress in tropical cyclone intensity forecasting at the National Hurricane Center. Wea. Forecasting, 35, 19131922, https://doi.org/10.1175/WAF-D-20-0059.1.

    • Search Google Scholar
    • Export Citation
  • Chan, J. C. L., and W. M. Gray, 1982: Tropical cyclone movement and surrounding flow relationships. Mon. Wea. Rev., 110, 13541374, https://doi.org/10.1175/1520-0493(1982)110<1354:TCMASF>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Chan, J. C. L., F. M. F. Ko, and Y. M. Lei, 2002: Relationship between potential vorticity tendency and tropical cyclone motion. J. Atmos. Sci., 59, 13171336, https://doi.org/10.1175/1520-0469(2002)059<1317:RBPVTA>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Charney, J. G., and A. Eliassen, 1964: On the growth of the hurricane depression. J. Atmos. Sci., 21, 6875, https://doi.org/10.1175/1520-0469(1964)021<0068:OTGOTH>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Chen, H., and D.-L. Zhang, 2013: On the rapid intensification of Hurricane Wilma (2005). Part II: Convective bursts and the upper-level warm core. J. Atmos. Sci., 70, 146162, https://doi.org/10.1175/JAS-D-12-062.1.

    • Search Google Scholar
    • Export Citation
  • Chen, S.-Y., Y.-H. Kuo, and C.-Y. Huang, 2020: The impact of GPS RO data on predicting tropical cyclogenesis using a nonlocal observation operator: An initial assessment. Mon. Wea. Rev., 148, 27012717, https://doi.org/10.1175/MWR-D-19-0286.1.

    • Search Google Scholar
    • Export Citation
  • Chen, S.-Y., T.-C. Nguyen, and C.-Y. Huang, 2021: Impact of radio occultation data on the prediction of Typhoon Haishen (2020) with WRFDA hybrid assimilation. Atmosphere, 12, 1397, https://doi.org/10.3390/atmos12111397.

    • Search Google Scholar
    • Export Citation
  • Chen, Y.-C., M.-E. Hsieh, L.-F. Hsiao, Y.-H. Kuo, M.-J. Yang, C.-Y. Huang, and C.-S. Lee, 2015: Systematic evaluation of the impacts of GPSRO data on the prediction of typhoons over the northwestern Pacific in 2008–2010. Atmos. Meas. Tech., 8, 25312542, https://doi.org/10.5194/amt-8-2531-2015.

    • Search Google Scholar
    • Export Citation
  • Chen, Y.-C., C.-C. Tsai, Y.-C. Wu, A.-H. Wang, C.-J. Wang, H.-H. Lin, D.-R. Chen, and Y.-C. Yu, 2021: Evaluation of operational monsoon moisture surveillance and severe weather prediction utilizing COSMIC-2/FORMOSAT-7 radio occultation observations. Remote Sens., 13, 2979, https://doi.org/10.3390/rs13152979.

    • Search Google Scholar
    • Export Citation
  • Chien, T.-Y., S.-Y. Chen, C.-Y. Huang, C.-P. Shih, C. S. Schwartz, Z. Liu, J. Bresch, and J.-Y. Lin, 2022: Impacts of radio occultation data on typhoon forecasts as explored by the global MPAS-GSI System. Atmosphere, 13, 1353, https://doi.org/10.3390/atmos13091353.

    • Search Google Scholar
    • Export Citation
  • Christophersen, H., A. Aksoy, J. Dunion, and S. Aberson, 2018: Composite impact of Global Hawk unmanned aircraft dropwindsondes on tropical cyclone analyses and forecasts. Mon. Wea. Rev., 146, 22972314, https://doi.org/10.1175/MWR-D-17-0304.1.

    • Search Google Scholar
    • Export Citation
  • Cram, T. A., J. Persing, M. T. Montgomery, and S. A. Braun, 2007: A Lagrangian trajectory view on transport and mixing processes between the eye, eyewall, and environment using a high-resolution simulation of Hurricane Bonnie (1998). J. Atmos. Sci., 64, 18351856, https://doi.org/10.1175/JAS3921.1.

    • Search Google Scholar
    • Export Citation
  • Cucurull, L., 2010: Improvement in using an operational constellation of GPS radio occultation receivers in weather forecasting. Wea. Forecasting, 25, 749767, https://doi.org/10.1175/2009WAF2222302.1.

    • Search Google Scholar
    • Export Citation
  • Cucurull, L., 2015: Implementation of quality control for radio occultation observations in the presence of large gradients of atmospheric refractivity. Atmos. Meas. Tech., 8, 12751285, https://doi.org/10.5194/amt-8-1275-2015.

    • Search Google Scholar
    • Export Citation
  • Cucurull, L., J. C. Derber, and R. J. Purser, 2013: A bending angle forward operator for global positioning system radio occultation measurements. J. Geophys. Res. Atmos., 118, 1428, https://doi.org/10.1029/2012JD017782.

    • Search Google Scholar
    • Export Citation
  • DeMaria, M., M. Mainelli, L. K. Shay, J. A. Knaff, and J. Kaplan, 2005: Further improvements to the Statistical Hurricane Intensity Prediction Scheme (SHIPS). Wea. Forecasting, 20, 531543, https://doi.org/10.1175/WAF862.1.

    • Search Google Scholar
    • Export Citation
  • Desroziers, G., L. Berre, B. Chapnik, and P. Poli, 2005: Diagnosis of observation, background and analysis-error statistics in observation space. Quart. J. Roy. Meteor. Soc., 131, 33853396, https://doi.org/10.1256/qj.05.108.

    • Search Google Scholar
    • Export Citation
  • Doyle, J. D., C. A. Reynolds, C. Amerault, and J. Moskaitis, 2012: Adjoint sensitivity and predictability of tropical cyclogenesis. J. Atmos. Sci., 69, 35353557, https://doi.org/10.1175/JAS-D-12-0110.1.

    • Search Google Scholar
    • Export Citation
  • Emanuel, K. A., 1989: The finite-amplitude nature of tropical cyclogenesis. J. Atmos. Sci., 46, 34313456, https://doi.org/10.1175/1520-0469(1989)046<3431:TFANOT>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Fritz, C., and Z. Wang, 2013: A numerical study of the impacts of dry air on tropical cyclone formation: A development case and a nondevelopment case. J. Atmos. Sci., 70, 91111, https://doi.org/10.1175/JAS-D-12-018.1.

    • Search Google Scholar
    • Export Citation
  • Galarneau, T. J., Jr., and C. A. Davis, 2013: Diagnosing forecast errors in tropical cyclone motion. Mon. Wea. Rev., 141, 405430, https://doi.org/10.1175/MWR-D-12-00071.1.

    • Search Google Scholar
    • Export Citation
  • Hack, J. J., and W. H. Schubert, 1986: Nonlinear response of atmospheric vortices to heating by organized cumulus convection. J. Atmos. Sci., 43, 15591573, https://doi.org/10.1175/1520-0469(1986)043<1559:NROAVT>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Healy, S. B., J. R. Eyre, M. Hamrud, and J.-N. Thépaut, 2007: Assimilating GPS radio occultation measurements with two-dimensional bending angle observation operators. Quart. J. Roy. Meteor. Soc., 133, 12131227, https://doi.org/10.1002/qj.63.

    • Search Google Scholar
    • Export Citation
  • Hendricks, E. A., M. T. Montgomery, and C. A. Davis, 2004: The role of “vortical” hot towers in the formation of Tropical Cyclone Diana (1984). J. Atmos. Sci., 61, 12091232, https://doi.org/10.1175/1520-0469(2004)061<1209:TROVHT>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Heymsfield, G. M., J. B. Halverson, J. Simpson, L. Tian, and T. P. Bui, 2001: ER-2 Doppler radar investigations of the eyewall of Hurricane Bonnie during the Convection and Moisture Experiment-3. J. Appl. Meteor., 40, 13101330, https://doi.org/10.1175/1520-0450(2001)040<1310:EDRIOT>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Ho, S.-P., and Coauthors, 2020a: The COSMIC/FORMOSAT-3 radio occultation mission after 12 years: Accomplishments, remaining challenges, and potential impacts of COSMIC-2. Bull. Amer. Meteor. Soc., 101, E1107E1136, https://doi.org/10.1175/BAMS-D-18-0290.1.

    • Search Google Scholar
    • Export Citation
  • Ho, S.-P., and Coauthors, 2020b: Initial assessment of the COSMIC-2/FORMOSAT-7 neutral atmosphere data quality in NESDIS/STAR using in-situ and satellite data. Remote Sens., 12, 4099, https://doi.org/10.3390/rs12244099.

    • Search Google Scholar
    • Export Citation
  • Huang, C.-Y., Y.-H. Kuo, S.-H. Chen, and F. Vandenberghe, 2005: Improvements in typhoon forecasts with assimilated GPS occultation refractivity. Wea. Forecasting, 20, 931953, https://doi.org/10.1175/WAF874.1.

    • Search Google Scholar
    • Export Citation
  • Huang, C.-Y., and Coauthors, 2010: Impact of GPS radio occultation data assimilation on regional weather predictions. GPS Solutions, 14, 3549, https://doi.org/10.1007/s10291-009-0144-1.

    • Search Google Scholar
    • Export Citation
  • Jarvinen, B. R., C. J. Neumann, and M. A. S. Davis, 1984: A tropical cyclone data tape for the North Atlantic basin, 1886–1983: Contents, limitations, and uses. NOAA Tech. Memo. NWS NHC 22, 24 pp., https://www.nhc.noaa.gov/pdf/NWS-NHC-1988-22.pdf.

  • Kaplan, J., and M. DeMaria, 2003: Large-scale characteristics of rapidly intensifying tropical cyclones in the North Atlantic basin. Wea. Forecasting, 18, 10931108, https://doi.org/10.1175/1520-0434(2003)018<1093:LCORIT>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Kueh, M.-T., C.-Y. Huang, S.-Y. Chen, S.-H. Chen, and C.-J. Wang, 2009: Impact of GPS RO refractivity soundings on a simulation of Typhoon Bilis (2006) upon landfall. Terr. Atmos. Oceanic Sci., 20, 115131, https://doi.org/10.3319/TAO.2008.01.21.03(F3C).

    • Search Google Scholar
    • Export Citation
  • Kuo, Y.-H., T.-K. Wee, S. Sokolovskiy, C. Rocken, W. Schreiner, D. Hunt, and R. A. Anthes, 2004: Inversion and error estimation of GPS radio occultation data. J. Meteor. Soc. Japan, 82, 507531, https://doi.org/10.2151/jmsj.2004.507.

    • Search Google Scholar
    • Export Citation
  • Latto, A., A. Hagen, and R. Berg, 2021: Tropical cyclone report: Hurricane Isaias (30 July–4 August 2020). NHC Tech. Rep. AL092020, 84 pp., https://www.nhc.noaa.gov/data/tcr/AL092020_Isaias.pdf.

  • Lien, G.-Y., and Coauthors, 2021: Assimilation impact of early FORMOSAT-7/COSMIC-2 GNSS radio occultation data with Taiwan’s CWB Global Forecast System. Mon. Wea. Rev., 149, 21712191, https://doi.org/10.1175/MWR-D-20-0267.1.

    • Search Google Scholar
    • Export Citation
  • Liu, H., J. Anderson, and Y.-H. Kuo, 2012: Improved analyses and forecasts of Hurricane Ernesto’s genesis using radio occultation data in an ensemble filter assimilation system. Mon. Wea. Rev., 140, 151166, https://doi.org/10.1175/MWR-D-11-00024.1.

    • Search Google Scholar
    • Export Citation
  • Liu, H., Y.-H. Kuo, S. Sokolovskiy, X. Zou, Z. Zeng, L.-F. Hsiao, and B. Ruston, 2018: A quality control procedure based on bending angle measurement uncertainty for radio occultation data assimilation in the tropical lower troposphere. J. Atmos. Oceanic Technol., 35, 21172131, https://doi.org/10.1175/JTECH-D-17-0224.1.

    • Search Google Scholar
    • Export Citation
  • Liu, Q., and Coauthors, 2020: Vortex initialization in the NCEP operational hurricane models. Atmosphere, 11, 968, https://doi.org/10.3390/atmos11090968.

    • Search Google Scholar
    • Export Citation
  • Liu, Z., C. S. Schwartz, C. Snyder, and S.-Y. Ha, 2012: Impact of assimilating AMSU-A radiances on forecasts of 2008 Atlantic tropical cyclones initialized with a limited-area ensemble Kalman filter. Mon. Wea. Rev., 140, 40174034, https://doi.org/10.1175/MWR-D-12-00083.1.

    • Search Google Scholar
    • Export Citation
  • McFarquhar, G. M., B. F. Jewett, M. S. Gilmore, S. W. Nesbitt, and T.-L. Hsieh, 2012: Vertical velocity and microphysical distributions related to rapid intensification in a simulation of Hurricane Dennis (2005). J. Atmos. Sci., 69, 35153534, https://doi.org/10.1175/JAS-D-12-016.1.

    • Search Google Scholar
    • Export Citation
  • Minamide, M., and F. Zhang, 2018: Assimilation of all-sky infrared radiances from Himawari-8 and impacts of moisture and hydrometeor initialization on convection-permitting tropical cyclone prediction. Mon. Wea. Rev., 146, 32413258, https://doi.org/10.1175/MWR-D-17-0367.1.

    • Search Google Scholar
    • Export Citation
  • Molinari, J., P. Dodge, D. Vollaro, K. L. Corbosiero, and F. Marks Jr., 2006: Mesoscale aspects of the downshear reformation of a tropical cyclone. J. Atmos. Sci., 63, 341354, https://doi.org/10.1175/JAS3591.1.

    • Search Google Scholar
    • Export Citation
  • Molinari, J., J. Frank, and D. Vollaro, 2013: Convective bursts, downdraft cooling, and boundary layer recovery in a sheared tropical storm. Mon. Wea. Rev., 141, 10481060, https://doi.org/10.1175/MWR-D-12-00135.1.

    • Search Google Scholar
    • Export Citation
  • Montgomery, M. T., and R. K. Smith, 2014: Paradigms for tropical cyclone intensification. Aust. Meteor. Oceanogr. J., 64, 3766, https://doi.org/10.22499/2.6401.005.

    • Search Google Scholar
    • Export Citation
  • Nolan, D. S., 2007: What is the trigger for tropical cyclogenesis? Aust. Meteor. Mag., 56, 241266.

  • Onderlinde, M. J., and D. S. Nolan, 2016: Tropical cyclone–relative environmental helicity and the pathways to intensification in shear. J. Atmos. Sci., 73, 869890, https://doi.org/10.1175/JAS-D-15-0261.1.

    • Search Google Scholar
    • Export Citation
  • Pasch, R. J., R. Berg, D. P. Roberts, and P. P. Papin, 2021: Tropical cyclone report: Hurricane Laura (20–29 August 2020). NHC Tech. Rep. AL132020, 75 pp., https://www.nhc.noaa.gov/data/tcr/AL132020_Laura.pdf.

  • Rogers, R., P. Reasor, and S. Lorsolo, 2013: Airborne Doppler observations of the inner-core structural differences between intensifying and steady-state tropical cyclones. Mon. Wea. Rev., 141, 29702991, https://doi.org/10.1175/MWR-D-12-00357.1.

    • Search Google Scholar
    • Export Citation
  • Rogers, R., P. Reasor, and J. A. Zhang, 2015: Multiscale structure and evolution of Hurricane Earl (2010) during rapid intensification. Mon. Wea. Rev., 143, 536562, https://doi.org/10.1175/MWR-D-14-00175.1.

    • Search Google Scholar
    • Export Citation
  • Ruston, B., and S. Healy, 2021: Forecast impact of FORMOSAT-7/COSMIC-2 GNSS radio occultation measurements. Atmos. Sci. Lett., 22, e1019, https://doi.org/10.1002/asl.1019.

    • Search Google Scholar
    • Export Citation
  • Schreiner, W. S., and Coauthors, 2020: COSMIC-2 radio occultation constellation: First results. Geophys. Res. Lett., 47, e2019GL086841, https://doi.org/10.1029/2019GL086841.

    • Search Google Scholar
    • Export Citation
  • Shapiro, L. J., and H. E. Willoughby, 1982: The response of balanced hurricanes to local sources of heat and momentum. J. Atmos. Sci., 39, 378394, https://doi.org/10.1175/1520-0469(1982)039<0378:TROBHT>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Sippel, J. A., and F. Zhang, 2008: A probabilistic analysis of the dynamics and predictability of tropical cyclogenesis. J. Atmos. Sci., 65, 34403459, https://doi.org/10.1175/2008JAS2597.1.

    • Search Google Scholar
    • Export Citation
  • Smith, R. K., J. A. Zhang, and M. T. Montgomery, 2017: The dynamics of intensification in a Hurricane Weather Research and Forecasting simulation of Hurricane Earl (2010). Quart. J. Roy. Meteor. Soc., 143, 293308, https://doi.org/10.1002/qj.2922.

    • Search Google Scholar
    • Export Citation
  • Teng, H.-F., Y.-H. Kuo, and J. M. Done, 2021: Importance of midlevel moisture for tropical cyclone formation in easterly and monsoon environments over the western North Pacific. Mon. Wea. Rev., 149, 24492469, https://doi.org/10.1175/MWR-D-20-0313.1.

    • Search Google Scholar
    • Export Citation
  • Teng, H.-F., Y.-H. Kuo, and J. M. Done, 2023: Potential impacts of radio occultation data assimilation on forecast skill of tropical cyclone formation in the western North Pacific. Geophys. Res. Lett., 50, e2021GL096750, https://doi.org/10.1029/2021GL096750.

    • Search Google Scholar
    • Export Citation
  • Trahan, S., and L. Sparling, 2012: An analysis of NCEP tropical cyclone vitals and potential effects on forecasting models. Wea. Forecasting, 27, 744756, https://doi.org/10.1175/WAF-D-11-00063.1.

    • Search Google Scholar
    • Export Citation
  • Van Sang, N., R. K. Smith, and M. T. Montgomery, 2008: Tropical-cyclone intensification and predictability in three dimensions. Quart. J. Roy. Meteor. Soc., 134, 563582, https://doi.org/10.1002/qj.235.

    • Search Google Scholar
    • Export Citation
  • Wadler, J. B., R. F. Rogers, and P. D. Reasor, 2018: The relationship between spatial variations in the structure of convective bursts and tropical cyclone intensification as determined by airborne Doppler radar. Mon. Wea. Rev., 146, 761780, https://doi.org/10.1175/MWR-D-17-0213.1.

    • Search Google Scholar
    • Export Citation
  • Wang, X., D. Parrish, D. Kleist, and J. Whitaker, 2013: GSI 3DVar-based ensemble-variational hybrid data assimilation for NCEP Global Forecast System: Single-resolution experiments. Mon. Wea. Rev., 141, 40984117, https://doi.org/10.1175/MWR-D-12-00141.1.

    • Search Google Scholar
    • Export Citation
  • Weiss, J.-P., W. S. Schreiner, J. J. Braun, W. Xia-Serafino, and C.-Y. Huang, 2022: COSMIC-2 mission summary at three years in orbit. Atmosphere, 13, 1409, https://doi.org/10.3390/atmos13091409.

    • Search Google Scholar
    • Export Citation
  • Wilks, D. S., 2011: Statistical Methods in the Atmospheric Sciences. 3rd ed. International Geophysics Series, Vol. 100, Academic Press, 704 pp.

  • Wu, T.-C., M. Zupanski, L. D. Grasso, C. D. Kummerow, and S.-A. Boukabara, 2019: All-sky radiance assimilation of ATMS in HWRF: A demonstration study. Mon. Wea. Rev., 147, 85106, https://doi.org/10.1175/MWR-D-17-0337.1.

    • Search Google Scholar
    • Export Citation
  • Xie, F., S. Syndergaard, E. R. Kursinski, and B. M. Herman, 2006: An approach for retrieving marine boundary layer refractivity from GPS occultation data in the presence of super-refraction. J. Atmos. Oceanic Technol., 23, 16291644, https://doi.org/10.1175/JTECH1996.1.

    • Search Google Scholar
    • Export Citation
  • Xie, F., D. L. Wu, C. O. Ao, E. R. Kursinski, A. J. Mannucci, and S. Syndergaard, 2010: Super-refraction effects on GPS radio occultation refractivity in marine boundary layers. Geophys. Res. Lett., 37, L11805, https://doi.org/10.1029/2010GL043299.

    • Search Google Scholar
    • Export Citation
  • Zagrodnik, J. P., and H. Jiang, 2014: Rainfall, convection, and latent heating distributions in rapidly intensifying tropical cyclones. J. Atmos. Sci., 71, 27892809, https://doi.org/10.1175/JAS-D-13-0314.1.

    • Search Google Scholar
    • Export Citation
  • Zawislak, J., and Coauthors, 2022: Accomplishments of NOAA’s airborne hurricane field program and a broader future approach to forecast improvement. Bull. Amer. Meteor. Soc., 103, E311E338, https://doi.org/10.1175/BAMS-D-20-0174.1.

    • Search Google Scholar
    • Export Citation
  • Zhang, H., Y.-H. Kuo, and S. Sokolovskiy, 2023: Assimilation of radio occultation data using measurement-based observation error specification: Preliminary results. Mon. Wea. Rev., 151, 589601, https://doi.org/10.1175/MWR-D-22-0122.1.

    • Search Google Scholar
    • Export Citation
  • Fig. 1.

    Observed tracks, colored by the current stage of TC development, of the six 2020 Atlantic hurricanes selected for the HWRF retrospective forecast experiments. Hurricane intensity is stratified by VMAX according to the Saffir–Simpson scale: category 1, 64–82 kt; category 2, 83–95 kt; category 3, 96–112 kt; category 4, 113–136 kt; and category 5, ≥137 kt. Observed position and intensity data are taken from the NHC best track database. Closed circles and × symbols denote each case’s beginning and end of the HWRF cycling period. The HWRF ghost d02 and ghost d03 domain sizes are also shown as gray rectangles.

  • Fig. 2.

    (a) Number of COSMIC-2 profiles with portions below 800 hPa passing quality control checks that are assimilated into the HWRF ghost d03 during each DA cycle for the Hanna, Zeta, and Isaias (2020) cases. DA cycles are labeled on the x axis by their cycle number n in the ordered sequence i = 1, …, n, …i_last where i = 1 and i = i_last are the cold start and final DA cycles, respectively. (b) As in (a), but for ghost d02.

  • Fig. 3.

    (a) Profiles of COSMIC-2 bending angle OB fractional innovation RMSD (red lines) and mean bias (blue lines) generated using all observations available for assimilation into HWRF ghost d03 from the six TC case experiments. Dashed and solid colored lines show statistics computed before and after observation QC, respectively. Dashed and solid black lines show the number of COSMIC-2 bending angles available for assimilation and assimilated, respectively. Data are binned by impact height using 1-km bin widths. (b) As in (a), but for the COSMIC-2 bending angle OA fractional innovation. (c) As in (a), but for profiles of post-QC COSMIC-2 bending angle observation-minus-background normalized innovation [(OB)/σo]. Also shown are normalized σo and normalized σrep, where data in each height bin are averaged over all post-QC COSMIC-2 observations. For normalized σo and normalized σrep, the mean post-QC COSMIC-2 bending angle observation value for each height bin is used as the denominator.

  • Fig. 4.

    (a) Histogram of the percentage of COSMIC-2 bending angle observations available for assimilation in HWRF ghost d03 from all six TC case experiments that are rejected by the HWRF-GSI QC algorithm, binned by pressure height. (b) As in (a), but showing a radius-binned histogram of the number of COSMIC-2 bending angles below 800 hPa taken from (i) the full dataset prior to HWRF-GSI QC screening (black bars), (ii) the subset rejected by the SR QC check (blue bars), and (iii) the subset rejected by the Statistical QC check (red bars). (c) As in (a), but for the COSMIC-2 observations below 800 hPa, binned by background forecast specific humidity at the observation location. (d) As in (b), but for COSMIC-2 observations binned by background forecast specific humidity at the observation location.

  • Fig. 5.

    (a) Absolute TC position error (km), averaged over all HWRF forecasts from the six cases valid at each 6-hourly lead time, where the dashed green line denotes the number of forecasts used in the average, for the Control (blue) and C2 (orange) configurations. Black solid and dashed lines plot the NHC official forecast TC position errors averaged over the six TC cases and all 2015–19 Atlantic TC cases, respectively. Green shading denotes the 90% confidence interval for the mean C2-minus-Control absolute position error difference generated by a bootstrap resampling of individual forecast differences; the mean C2-minus-Control absolute position error difference is considered statistically significant at lead times when the interval does not include zero, as denoted by the triangles. (b) As in (a), but for mean PMIN absolute error. (c),(d) Relative skill, or the normalized percent absolute error reduction relative to Control, for TC position and PMIN, respectively. For (c) and (d), positive (negative) relative skill indicates that C2 forecasts are, on average, improved (degraded) relative to Control.

  • Fig. 6.

    As in Fig. 5, but for the subset of HWRF analyses valid within 12 h of the beginning of an observed intensification episode, defined by a NHC best track VMAX rate of change exceeding 20 kt (24 h)−1.

  • Fig. 7.

    (a) Time series of the fractional change in the HWRF C2 forecast temperature RMSD profile relative to Control (shaded; %), using dropwindsonde and radiosonde data from the six hurricane cases. Black contours show the HWRF Control forecast temperature RMSD (K) profile time series. Green dashed contours show the number of observation–gridpoint pairs used. (b) As in (a), but for specific humidity (g kg−1). (c),(d) As in (a), but for u- and υ-wind components (m s−1), respectively. Black (brown) triangles in (a)–(d) show times and heights where the improvement (degradation) in C2 forecast RMSD is statistically significant (computed using a two-sample bootstrap method) at the 90% level; black (brown) closed circles denote times and heights where C2 RMSD improvement (degradation) over Control is statistically significant at the 95% level.

  • Fig. 8.

    Profile of the t = 24-h C2 forecast SPFH RMSD normalized by the t = 24-h Control forecast SPFH RMSD, with error bars denoting the 95% confidence interval (black lines). The RMSD is computed as in Fig. 7. The green line shows the number of observation–gridpoint pairs used for each 100-hPa-deep bin.

  • Fig. 9.

    (a) Hovmöller plot showing the HWRF C2 cycled analysis 700–850-hPa layer-averaged and azimuthally averaged RH (shaded; %) and its difference from the cycled Control analysis (C2 − Control; contours; %) as a function of radius and UTC time, for the Hurricane Hanna (2020) cycling period. The × symbols mark the approximate 800-hPa radial locations of assimilated COSMIC-2 profiles; the total number of COSMIC-2 profiles with observations below 800 hPa assimilated in ghost d03 per cycle, including outer portions not shown here, is labeled on the right-hand y axis. (b) PMIN time series for selected Hurricane Hanna (2020) HWRF free forecasts initialized from cycled Control (blue lines) and C2 (orange lines) analyses, where different initialization times are shown in different line styles (see inset key). The black line shows the NHC best track PMIN. (c) As in (a), but for the 850–950-hPa layer. Dashed black line in (c) denotes the approximate 925-hPa C2 analysis radius of maximum wind. (d) As in (b), but for VMAX. Red squares in (b) and (d) show the C2 analysis intensity; if different from C2, the Control analysis intensity is shown as a blue square.

  • Fig. 10.

    (a) HWRF Control analysis minus ERA5 950-hPa qυ difference (shaded; g kg−1) with ERA5 950-hPa qυ (contours; g kg−1) and horizontal wind vectors (m s−1), valid at 0600 UTC 23 Jul 2020. (b) Control analysis minus background 950-hPa qυ difference (shaded; g kg−1) with Control analysis horizontal wind vectors (m s−1), valid at 0600 UTC 23 Jul 2020. (c) HWRF C2 minus Control analysis 950-hPa qυ difference (shaded; g kg−1) with C2 analysis horizontal wind vectors (m s−1), valid at 0600 UTC 23 Jul 2020. COSMIC-2 profiles assimilated in the 0600 UTC 23 Jul C2 analysis cycle are also shown, labeled by the identification number; magenta (green) colors show portions of each profile below (above) 800 hPa. (d)–(f) As in (a)–(c), but for the 800-hPa level. (g)–(i) As in (a)–(c), but for the 600-hPa level.

  • Fig. 11.

    (a) Hovmöller plot showing the radius–time dependence of the Hurricane Hanna 0600 UTC 23 Jul initialized Control forecast RH, azimuthally averaged and vertically averaged over the 700–850-hPa layer (shaded; %), 800-hPa azimuthal mean SPFH (contours; g kg−1), and the 700–850-hPa layer-averaged azimuthal mean radial wind component (VR; vectors; m s−1). (c) As in (a), but for 850–950-hPa mean RH and VR, with 925-hPa SPFH. (e) As in (a), but for composite reflectivity (shaded; dBZ) and the 925-hP atangential wind component (VT; contours; m s−1). (g) As in (a), but for 700-hPa relative vorticity (×104 s−1; shaded) and 400–700-hPa layer-averaged w (contours; m s−1; thin dotted for −0.1, thin solid for 0.1, thick solid for 0.3, 0.5, and 1.0). (b),(d),(f),(h) As in (a), (c), (e), and (g), respectively, but for the Hurricane Hanna 0600 UTC 23 Jul initialized C2 forecast. Note the smaller radial range shown for (g) and (h) compared to the other panels. Numbered rectangles highlight periods and regions discussed in the text.

  • Fig. 12.

    (a) Composite reflectivity (shaded; dBZ), 400–700-hPa layer-averaged vertical velocity (contours at 0.5 and 3 m s−1), and 850-hPa horizontal flow vectors (m s−1) from the 18-h HWRF Control Hurricane Hanna (2020) forecast initialized at 0600 UTC 23 Jul. (b) As in (a), but for the 18-h HWRF C2 forecast initialized at 0600 UTC 23 Jul. (c) SSMIS passive 91-GHz microwave radiance-derived color composite imagery of developing TC Hanna (2020) and its environment, obtained from a Defense Meteorological Satellite Program overpass valid at 2218 UTC 23 Jul (image provided courtesy of the Naval Research Laboratory Monterey, available for public access at https://www.nrlmry.navy.mil/tc-bin/tc_home2.cgi). (d),(e) As in (a) and (b), but for the 30-h HWRF Control and C2 forecasts, respectively, initialized at 0600 UTC 23 Jul. (f) As in (c), but valid at 1047 UTC 24 Jul. Black (white) triangles denote the approximate HWRF-forecast (observed) TC Hanna center position.

  • Fig. 13.

    As in Fig. 9, but for the Hurricane Zeta (2020) cycled HWRF experiments. Black rectangles in (a) and (c) highlight the period of interest, i.e., the 0600 and 1200 UTC 26 Oct analyses. In (a) and (c), C2 − Control layer-averaged and azimuthally averaged RH differences (%) are dotted-line contoured for values −6%, −3%, and −1%; thin solid line contoured for values 1% and 3%; and thick solid line contoured for values 6%, 9%, and 12%.

  • Fig. 14.

    As in Fig. 10, but for the Hurricane Zeta (2020) HWRF Control and C2 1200 UTC 26 Oct cycled analysis times at the (a)–(c) 900-, (d)–(f) 750-, and (g)–(i) 600-hPa levels. Unlike Fig. 10, (b), (e), and (h) show the C2 6-h background forecast horizontal wind vectors (m s−1) and the C2-minus-Control 6-h background forecast qV difference (g kg−1; shaded); the −3 and −1 g kg−1 and 1 and 3 g kg−1 contours of these difference fields are also shown in purple and black, respectively, in (c), (f), and (i). The rectangular box shown in (d)–(f) highlights a region discussed in the text.

  • Fig. 15.

    As in Fig. 11, but for the (a),(c),(e),(g) Control and (b),(d),(f),(h) C2 HWRF forecasts for Hurricane Zeta (2020) initialized at 1200 UTC 26 Oct. For (g) and (h), w is thin-dotted contoured for –0.5 and −0.1 m s−1, thin-solid contoured for 0.3 m s−1, and thick-solid contoured for 0.6 and 1.0 m s−1. Also note the smaller radial ranges shown for (e) and (f) and for (g) and (h) compared to the other panels.

  • Fig. 16.

    (a) HWRF Control Hurricane Zeta (2020) 700–850-hPa layer-averaged RH (shaded; %) and 750-hPa horizontal wind vectors (m s−1) from the 1200 UTC 26 Oct cycled analysis. (b),(c) As in (a), but for the 12- and 18-h verification times, respectively, from the Control forecast initialized from the 1200 UTC 26 Oct cycled analysis. (d)–(f) As in (a)–(c), but for the HWRF C2 forecast initialized from the 1200 UTC 26 Oct cycled analysis. Rectangular boxes highlight the dry tongue discussed in the text.

  • Fig. 17.

    (a)–(c) As in Fig. 12a, but for the Control 12-, 24-, and 30-h Hurricane Zeta (2020) forecasts initialized at 1200 UTC 26 Oct, respectively, with 925-hPa Control forecast horizontal wind vectors (m s−1). (d)–(f) As in (a)–(c), but for the C2 Hurricane Zeta (2020) 12-, 24-, and 30-h forecasts initialized at 1200 UTC 26 Oct, respectively. (g)–(i) As in Fig. 12c, but for SSMIS 91-GHz color composite imagery valid at 2225 UTC 26 Oct, 1058 UTC 27 Oct, and 0030 UTC 28 Oct, respectively (images provided courtesy of the Naval Research Laboratory Monterey, available for public access at https://www.nrlmry.navy.mil/tc-bin/tc_home2.cgi). The white triangle in (h) marks the approximate NHC best track center position at 1200 UTC 27 Oct (21.3°N, 89.0°W).

All Time Past Year Past 30 Days
Abstract Views 654 21 0
Full Text Views 3625 3364 176
PDF Downloads 572 273 32