FIONA: Multidisciplinary Science & Applications

• FIONA is a multifaceted term describing methodologies in meteorology, audio forensics, gravitational-wave lensing, and particle physics.
• It underpins advanced wind reconstruction in hurricanes and rare-event cyclogenesis studies by integrating sparse satellite data with high-quality in situ observations.
• FIONA also drives innovative fusion frameworks for deepfake detection and optimizes FFT-based lensing computations, while symbolizing an exotic gauge boson in theoretical models.

FIONA is used in recent research literature in several distinct senses. In tropical-cyclone and climate-impact studies, Fiona denotes Hurricane Fiona (2022), which is treated as a real-data case for inner-core wind reconstruction, rare-event tropical cyclogenesis diagnostics, gray-swan scenario generation, and coupled climate–energy cascade modeling (Han et al., 18 May 2026, Schreck et al., 29 Jun 2026, Hakim et al., 1 Apr 2026, Xu et al., 2024). In audio forensics, “FIONA” abbreviates “FusION through Kernel Alignment,” a fusion framework for singing voice deepfake detection (Phukan et al., 2024). In gravitational-wave lensing, “FIONA” abbreviates “Fresnel Integral Optimization with Non-uniform trAnsforms,” an FFT-based code for evaluating the Fresnel diffraction integral (Ephremidze et al., 12 Mar 2026). In a curved-space-time Pati–Salam construction, Fiona is also the name assigned to the gauge boson $W_\mu^0$ (Li, 2014).

1. Hurricane Fiona in inner-core vector-wind reconstruction

In "Global kilometre-scale tropical cyclone inner-core vector winds from sparse scalar CYGNSS observations" (Han et al., 18 May 2026), TC FIONA (2022) is the principal real-data fusion case study for QiFeng, a framework that reconstructs the 10 m vector wind field $\mathbf{x}=(u_{10},v_{10})$ on a $384\times384$ km domain at 1.5 km resolution. The study assimilates CYGNSS scalar wind speeds through the nonlinear observation operator

$$H_{\text{CYG}}(\mathbf{x})_k=\sqrt{u_{i_k,j_k}^2+v_{i_k,j_k}^2},$$

and, for Fiona, augments them with 11 dropsonde vector winds via a linear identity operator. The reconstruction is guided by a score-based diffusion prior and three boundary-layer constraints: low-level divergence, bounded inflow angle, and cyclonic vorticity (Han et al., 18 May 2026).

Fiona was selected because it combined unusually poor CYGNSS sampling with unusually strong independent in situ and remote-sensing constraints. For 2022-09-18 12 UTC, only 95 CYGNSS specular points were available in the analysis domain, mostly in the outer circulation and with almost no inner-core coverage; this is far below the Observation Coverage Sufficiency threshold $n_{\mathrm{obs}}\ge 300$, so the case is explicitly treated as coverage-insufficient. At the same time, 11 aircraft dropsondes partially sampled the eyewall and inner core, and a Tail Doppler Radar composite at nearly the same time provided an independent benchmark after correction of the 0.5 km wind layer to 10 m with a radially dependent reduction factor. IBTrACS gives Fiona a best-track $V_{\max}=33\,\mathrm{m\,s^{-1}}$ at that analysis time (Han et al., 18 May 2026).

The Fiona results quantify the effect of heterogeneous observation fusion. For the west–east cross-eye 10 m wind-speed profile, CYGNSS-only QiFeng has an RMSE of $9.7\,\mathrm{m\,s^{-1}}$ against TDR, while CYGNSS plus 11 dropsondes reduces this to $5.7\,\mathrm{m\,s^{-1}}$, a 42% reduction. At dropsonde locations, the CYGNSS-only reconstruction underestimates eyewall winds by 30–50%, whereas the fused reconstruction reduces the bias to within $3\,\mathrm{m\,s^{-1}}$ at most sites. Spatially, the TDR field shows a clearly defined eye with winds below $5\,\mathrm{m\,s^{-1}}$ and an asymmetric eyewall strongest in the southeast quadrant; CYGNSS-only QiFeng is too smooth and nearly axisymmetric, whereas the fusion run restores a sharper eye, steeper eyewall gradient, and the southeast–northwest asymmetry qualitatively (Han et al., 18 May 2026).

Within the broader study, Fiona is not an OCS-qualified case and is not used for hyperparameter tuning or observing-system simulation experiments. Its role is instead methodological: it demonstrates that when CYGNSS-only reconstruction is not recommended, a small number of high-quality vector observations can materially constrain the posterior and recover a dynamically consistent inner-core flow (Han et al., 18 May 2026).

2. Fiona as a precursor environment in rare-event tropical cyclogenesis

In "Conditional Tropical Cyclogenesis Rates via Rare-Event Sampling in a Neural Weather Emulator" (Schreck et al., 29 Jun 2026), the Fiona case does not denote the mature hurricane itself but the Fiona precursor environment, represented by the 2022-09-09 12Z ERA5 initial condition. The study couples Forward Flux Sampling to SDL-WXFormer, a stochastic neural weather emulator on a $\mathbf{x}=(u_{10},v_{10})$0 grid advanced in 6-hour steps out to 15 days. The order parameter is the minimum mean sea-level pressure over the Atlantic basin,

$\mathbf{x}=(u_{10},v_{10})$1

with state $\mathbf{x}=(u_{10},v_{10})$2 defined by $\mathbf{x}=(u_{10},v_{10})$3 hPa and state $\mathbf{x}=(u_{10},v_{10})$4 defined by $\mathbf{x}=(u_{10},v_{10})$5 hPa together with a warm-core tropical structure by Cyclone Phase Space and an in-domain vortex (Schreck et al., 29 Jun 2026).

The conditional genesis rate is factorized as

$\mathbf{x}=(u_{10},v_{10})$6

with interfaces at $\mathbf{x}=(u_{10},v_{10})$7, $\mathbf{x}=(u_{10},v_{10})$8, $\mathbf{x}=(u_{10},v_{10})$9, $384\times384$0, and $384\times384$1 hPa. For the Fiona precursor environment, the paper reports

$384\times384$2

$384\times384$3

and therefore

$384\times384$4

with a direct-sampling estimate of $384\times384$5 and an FFS-to-direct ratio of 1.21. This is the highest genesis rate in the 98-initial-condition dataset, and the speedup over direct sampling for equal 20% uncertainty is $384\times384$6 (Schreck et al., 29 Jun 2026).

The study characterizes Fiona by “uniformly high crossing probabilities.” That description refers primarily to the first three interfaces: once a disturbance in that environment reaches 1000 hPa, it is very likely to deepen through 987.2, 984.7, and 981.2 hPa before returning to the unorganized state. By contrast, the principal remaining barrier is the final transition to $384\times384$7 hPa while retaining warm-core tropical structure and remaining within the basin. Composite diagnostics of reactive trajectories show a compact 850-hPa vorticity core, high $384\times384$8-hPa relative humidity, and a growing 500-hPa warm-core anomaly, together with westward tracks into the Caribbean that are consistent with Fiona’s observed evolution (Schreck et al., 29 Jun 2026).

A plausible implication is that this precursor environment is best understood not merely as “favorable” in a generic sense, but as a regime in which early organization is unusually non-rate-limiting. The paper contrasts this with Earl, for which the principal bottleneck is initial organization, and Ian, for which the principal bottleneck is the late intensification stages (Schreck et al., 29 Jun 2026).

3. Fiona as seed storm in gray-swan scenario generation

In "Gray Swan Factory: Making Extreme Events from Ordinary Cyclones" (Hakim et al., 1 Apr 2026), Hurricane Fiona is the central case for a differentiable-weather-model optimization that constructs a Sandy-like catastrophe from a nearby initial condition. The forecast model is NeuralGCM v1 at $384\times384$9 horizontal resolution. The optimization is performed over the 48-hour window from 00 UTC 22 September 2022 to 00 UTC 24 September 2022 and minimizes a loss

$$H_{\text{CYG}}(\mathbf{x})_k=\sqrt{u_{i_k,j_k}^2+v_{i_k,j_k}^2},$$0

where $$H_{\text{CYG}}(\mathbf{x})_k=\sqrt{u_{i_k,j_k}^2+v_{i_k,j_k}^2},$$1 drives the 1000 hPa geopotential at a grid point near Sandy’s landfall toward a prescribed low target, $$H_{\text{CYG}}(\mathbf{x})_k=\sqrt{u_{i_k,j_k}^2+v_{i_k,j_k}^2},$$2 penalizes departures from the ERA5 initial temperature field, and $$H_{\text{CYG}}(\mathbf{x})_k=\sqrt{u_{i_k,j_k}^2+v_{i_k,j_k}^2},$$3 penalizes large first-step geopotential tendencies (Hakim et al., 1 Apr 2026).

In the control NeuralGCM forecast initialized from ERA5, Fiona recurves northward and reaches Atlantic Canada, broadly matching the observed track. The gray-swan optimization instead produces a storm that remains similar to control through roughly 23/12, then turns sharply westward toward the U.S. Mid-Atlantic coast in a Sandy-like manner. The optimized storm deepens from 979 hPa at 22/06 to 952 hPa at 23/12 and then undergoes 33 hPa of deepening in 12 hours, reaching 919 hPa at 24/00. The abstract characterizes this as a warm-core extratropical cyclone with a minimum sea-level pressure more than 20 hPa lower than Sandy (Hakim et al., 1 Apr 2026).

The paper attributes the key sensitivity not to direct perturbation of the hurricane core but to perturbation of the extratropical flow. In the optimized solution, the upstream 500-hPa trough is shifted westward and shortened in zonal wavelength; afterward, the trough and hurricane co-rotate and merge into a single vertically aligned vortex. Experiments that restrict perturbations to the tropical-cyclone region do not produce the Sandy-like outcome, whereas environment-only perturbations produce a qualitatively similar, though weaker, coastal-impact scenario. The paper therefore states that perturbations to the extratropical state are more important than to the hurricane (Hakim et al., 1 Apr 2026).

This suggests that, within the model’s coarse-resolution dynamics, Fiona functions as a “near miss” analogue of Sandy: an observed storm whose realized trajectory was ordinary relative to a nearby but much more damaging dynamical possibility. The paper also stresses a major weakness: the hurricane core is not resolved by the model used for optimization, and the impact of that limitation is unknown (Hakim et al., 1 Apr 2026).

4. Fiona and the Puerto Rico system-wide blackout

In "Quantifying cascading power outages during climate extremes considering renewable energy integration" (Xu et al., 2024), Hurricane Fiona is the validation event for CRESCENT, a coupled climate–energy model of cascading outages. Fiona made landfall in Puerto Rico on 18 September 2022 as a Category 1 hurricane. The entire island, about 1.5 million customers, lost power, and outages escalated from below 50% customers out to 100% in a single 10-minute step at 18:00 UTC. The paper describes this as the first-ever system-wide blackout event with complete weather-induced outage records (Xu et al., 2024).

CRESCENT combines a physics-based tropical-cyclone wind model with component fragility or resistance formulations for transmission towers, transmission lines, distribution feeders, and solar infrastructure, and embeds these within a cascade model that includes DC power flow, OPA-style overload tripping, RoCoF-based inertial stability, and a MILP real-time operation layer. The system imbalance in sub-grid $$H_{\text{CYG}}(\mathbf{x})_k=\sqrt{u_{i_k,j_k}^2+v_{i_k,j_k}^2},$$4 is written as

$$H_{\text{CYG}}(\mathbf{x})_k=\sqrt{u_{i_k,j_k}^2+v_{i_k,j_k}^2},$$5

and the maximum Rate of Change of Frequency is constrained by

$$H_{\text{CYG}}(\mathbf{x})_k=\sqrt{u_{i_k,j_k}^2+v_{i_k,j_k}^2},$$6

with $$H_{\text{CYG}}(\mathbf{x})_k=\sqrt{u_{i_k,j_k}^2+v_{i_k,j_k}^2},$$7 used as the stability threshold for Puerto Rico (Xu et al., 2024).

Against Fiona, the model reproduces both progressive degradation and sudden collapse. Among 1000 realizations, 60% produce catastrophic blackouts, and the peak in blackout occurrences is at 18:00 UTC, matching the observed collapse time. The study also reports a novel resilience pattern: early failure of certain critical transmission lines can improve survival odds. For the Costa Sur–Manatí line, forcing it to be among the weakest 1% of components decreases the probability of catastrophic blackout by 17%, and placing it among the weakest 10% decreases that probability by 9% (Xu et al., 2024).

The Fiona event is also used to study behind-the-meter solar integration. Puerto Rico’s current BTM PV penetration in the case is about 16.1% of peak demand. Sensitivity analysis finds that integration levels below about 45%, including the current level, exhibit minimal impact on system resilience in this event; above that level, without additional flexibility resources, blackout probability rises super-linearly because hurricane-induced PV output losses enlarge supply–demand imbalances while synchronous inertia is reduced. Fiona therefore serves simultaneously as a meteorological hazard case and as a systems-resilience benchmark for climate–energy interactions (Xu et al., 2024).

5. FIONA as a singing voice deepfake detection framework

In "Are Music Foundation Models Better at Singing Voice Deepfake Detection? Far-Better Fuse them with Speech Foundation Models" (Phukan et al., 2024), FIONA stands for “FusION through Kernel Alignment.” It is a fusion framework for singing voice deepfake detection that combines a speech foundation model and a music foundation model. The best-performing instantiation synchronizes the speaker-recognition speech model x-vector with the music model MERT-v1-330M and reports the lowest Equal Error Rate of 13.74%, outperforming all individual foundation models and baseline fusion schemes (Phukan et al., 2024).

The framework operates on clip-level embeddings extracted by average pooling the last hidden layer of each foundation model. Each feature stream is passed through a separate CNN branch, then through a sigmoid-based gating mechanism, and then through a linear projection to a common dimensionality. The two projected streams are concatenated and passed to an FCN classifier. Training minimizes a joint objective

$$H_{\text{CYG}}(\mathbf{x})_k=\sqrt{u_{i_k,j_k}^2+v_{i_k,j_k}^2},$$8

where $$H_{\text{CYG}}(\mathbf{x})_k=\sqrt{u_{i_k,j_k}^2+v_{i_k,j_k}^2},$$9 is the standard cross-entropy loss and $n_{\mathrm{obs}}\ge 300$0 is a Centered Kernel Alignment term that encourages the two branches to align in representation space (Phukan et al., 2024).

The empirical motivation for FIONA comes from a systematic comparison of music and speech foundation models on the CtrSVDD dataset. Among single models, x-vector is the strongest, reaching 14.18% EER with a CNN classifier, while the best music foundation model, MERT-v1-330M, achieves 27.50% EER. The fusion of x-vector and MERT-v1-330M via simple concatenation reaches 13.85% EER, while FIONA further improves this to 13.74%. The paper interprets this pattern as evidence that speaker-recognition representations capture pitch, tone, intensity, and related singer-specific cues more effectively than music-only representations, while the music branch still contributes complementary structure when the two are explicitly aligned (Phukan et al., 2024).

Within the supplied literature, this is the clearest case in which FIONA is an acronym rather than a proper name. The term denotes a specific neural fusion architecture rather than a dataset, benchmark, or pretrained model (Phukan et al., 2024).

6. FIONA as a code for gravitational-wave lensing

In "Fast Fourier Transform evaluation of the Fresnel integral for gravitational-wave lensing" (Ephremidze et al., 12 Mar 2026), FIONA stands for “Fresnel Integral Optimization with Non-uniform trAnsforms.” The code evaluates the wave-optics amplification factor

$n_{\mathrm{obs}}\ge 300$1

where $n_{\mathrm{obs}}\ge 300$2 is given by the two-dimensional Fresnel integral

$n_{\mathrm{obs}}\ge 300$3

The central methodological observation is that the source-position dependence can be rewritten as a two-dimensional Fourier transform, enabling simultaneous evaluation over the source plane with non-uniform FFTs (Ephremidze et al., 12 Mar 2026).

For general, non-axisymmetric lenses, FIONA uses Gauss–Legendre quadrature and Type-1 NUFFTs after subtracting the free-space contribution and windowing the lens potential to a finite radius. For axisymmetric lenses, the Fresnel integral reduces to a zeroth-order Hankel transform and is evaluated with a non-uniform fast Hankel transform. The code also vectorizes derivatives with respect to lens parameters by reusing the same transform structure with modified weights, so gradients can be obtained with only incremental additional cost (Ephremidze et al., 12 Mar 2026).

The principal performance claim is comparative. For dense source grids, FIONA is significantly faster than contour-based methods such as GLoW, reaching two to three orders of magnitude speedups for approximately $n_{\mathrm{obs}}\ge 300$4 GW-emitting points. The paper reports that, for a $n_{\mathrm{obs}}\ge 300$5 source grid and 560 frequencies between $n_{\mathrm{obs}}\ge 300$6 and $n_{\mathrm{obs}}\ge 300$7, evaluation for simple multicomponent lenses takes about 6 seconds on 112 CPU cores; in axisymmetric examples, 560 frequencies between $n_{\mathrm{obs}}\ge 300$8 and $n_{\mathrm{obs}}\ge 300$9 and 400 source positions can be evaluated in about 0.7 CPU seconds on a MacBook M2 (Ephremidze et al., 12 Mar 2026).

Scientifically, the code is presented as a tool for wave-optics studies of dark-matter substructure, including galactic subhalos and multi-component lens models, where dense sampling in source position and repeated evaluation across frequency are operationally important. The paper also notes that the vectorized non-uniform fast Hankel transforms developed as part of FIONA may have uses beyond gravitational-wave lensing, such as cosmological two-point correlation calculations (Ephremidze et al., 12 Mar 2026).

7. Fiona as an exotic gauge boson

In "Pati-Salam model in curved space-time from square root Lorentz manifold" (Li, 2014), Fiona is neither a storm nor an acronym but a named particle. The paper states in its particle spectrum discussion that $V_{\max}=33\,\mathrm{m\,s^{-1}}$0 and $V_{\max}=33\,\mathrm{m\,s^{-1}}$1 are gauge-boson fields and identifies $V_{\max}=33\,\mathrm{m\,s^{-1}}$2 as the dark photon and $V_{\max}=33\,\mathrm{m\,s^{-1}}$3 as the Fiona particle. Fiona is therefore the $V_{\max}=33\,\mathrm{m\,s^{-1}}$4 gauge boson associated with the $V_{\max}=33\,\mathrm{m\,s^{-1}}$5 factor in a $V_{\max}=33\,\mathrm{m\,s^{-1}}$6 bundle over a “square root Lorentz manifold” (Li, 2014).

Within the model’s matter Lagrangian,

$V_{\max}=33\,\mathrm{m\,s^{-1}}$7

the Fiona contribution comes from the $V_{\max}=33\,\mathrm{m\,s^{-1}}$8 term, where $V_{\max}=33\,\mathrm{m\,s^{-1}}$9 is proportional to the identity generator. The details block interprets this as a universal coupling to a flavor-number current. The paper also places Fiona among a set of exotic gauge bosons that includes the dark photon, $9.7\,\mathrm{m\,s^{-1}}$0, and $9.7\,\mathrm{m\,s^{-1}}$1 (Li, 2014).

The broader gauge-theoretic role of Fiona is tied to the decomposition from $9.7\,\mathrm{m\,s^{-1}}$2 to a chiral flavor structure and a Pati–Salam-type non-Abelian subgroup. The details supplied with the paper state that Fiona is the flavor-sector $9.7\,\mathrm{m\,s^{-1}}$3 gauge boson, in parallel with the dark photon as the color-sector $9.7\,\mathrm{m\,s^{-1}}$4 gauge boson. The paper does not provide a mass term, a mass matrix, or explicit mixing angles for Fiona, and it does not specify whether Fiona is massless or massive, stable or unstable. The name “Fiona” is also described there as specific to this paper rather than a standard designation in the field (Li, 2014).

Across these literatures, FIONA is therefore not a single concept but a label reused for sharply different entities: a 2022 hurricane employed as a benchmark in meteorology, power-systems, and rare-event studies; a kernel-aligned fusion architecture for audio deepfake detection; a high-performance code for wave-optics gravitational-wave lensing; and an exotic gauge boson in a speculative curved-space-time gauge model.