Thermodynamically primed Atmospheric River Rapid as the driver of the December 2023 Thoothukudi extreme rainfall

The December 2023 extreme rainfall over Thoothukudi, India (946 mm in 24 h), represents a pronounced precipitation efficiency anomaly, as such intensity occurred in the absence of a synoptic-scale cyclonic storm and exceeded rates explainable by resolved large-scale ascent. This study introduces a newly proposed Atmospheric River Rapid Index (ARRI) to diagnostically examine whether this event is consistent with the influence of a thermodynamically primed tropical Atmospheric River Rapid (AR Rapid). Using GPM IMERG precipitation estimates and ERA5 reanalysis, we apply a diagnostic Eulerian moisture budget decomposition to assess the processes contributing to the event. The analysis indicates a dual influence: the large-scale environment was thermodynamically primed by background moisture anomalies (dynamic term ≈ 33.0%), while the spatial concentration and intensity of rainfall are primarily associated with anomalous wind steering and convergence of integrated vapor transport (thermodynamic circulation term ≈ 102.0%) within a vector framework. The ARRI identifies a narrow coastal region characterized by peak IVT, strong IVT convergence, deep moisture, and reduced low-level wind speeds, consistent with frictional deceleration of a low-level jet at landfall and consistent with conditions favorable for AR Rapid ascent. Complementary dynamical diagnostics further suggest a limited contribution from organized synoptic-scale cyclonic mechanisms. These results provide diagnostic evidence consistent with a thermodynamically primed AR Rapid suggesting a key contributing process to this extreme, non-cyclonic tropical rainfall event. The findings highlight limitations of cyclone-centric forecasting approaches and underscore the potential value of monitoring IVT convergence and thermodynamic pre-conditioning for anticipating high-impact tropical flood events.

1 Introduction

The Northeast Monsoon (NEM), occurring during October–December, is the primary rainfall season for southeastern peninsular India, particularly Tamil Nadu and adjacent regions. Rainfall during the NEM is mainly driven by easterly to northeasterly low-level winds from the Bay of Bengal and is commonly associated with easterly waves, low-pressure systems, depressions, and mesoscale convective activity. Recent studies have shown that this season also supports the development of atmospheric rivers (ARs). In tropical regions, ARs manifest as elongated plumes of intense moisture transport embedded within the monsoonal circulation, rather than as distinct mid-latitude synoptic systems (Nair et al., 2025).

The Bay of Bengal is a warm, moisture-rich region, and tropical ARs can transport exceptionally large amounts of water vapor toward the southeastern Indian coast. When this moisture interacts with coastal friction, low-level convergence, and mesoscale convection, precipitation efficiency can increase substantially (Varikoden, 2025). As a result, ARs during the Northeast Monsoon can act as effective pathways for extreme rainfall even in the absence of organized cyclonic systems (Zhu and Newell, 1998), highlighting their importance for understanding high-impact rainfall events over southern India.

Within broader AR plumes, recent research has identified Atmospheric River Rapids (AR Rapids) as narrow, mesoscale regions where moisture convergence and vertical transport are particularly intense (Francis, 2024). AR Rapids are characterized by the rapid conversion of strong horizontal moisture transport into localized upward motion, often initiated by boundary-layer mechanisms such as coastal friction, deceleration of low-level jets, or orographic interaction (Zhang, 2024). Although AR Rapids occupy spatial scales much smaller than their parent ARs, they can dominate rainfall intensity and produce highly localized extreme precipitation. In tropical environments, where background humidity and sea surface temperatures are elevated, AR Rapids can be especially efficient; however, they remain difficult to detect using conventional synoptic diagnostics or coarse-resolution reanalysis products.

1.1 Literature survey

While previous studies have established the role of atmospheric rivers in modulating extreme rainfall and have recently begun to document the existence of AR Rapids, a critical gap remains in the Indian context. Specifically, there is currently no verified diagnostic framework to objectively identify tropical AR Rapids or to distinguish their contribution from traditional synoptic-scale disturbances such as weak monsoon depressions. Existing studies largely focus on AR detection or climatology and do not provide event-scale causal attribution that separates thermodynamic moisture pre-conditioning from circulation-driven moisture focusing. This gap limits the ability to diagnose non-cyclonic extreme rainfall events during the Northeast Monsoon, which are often poorly anticipated by cyclone-centric forecasting approaches.

1.1.1 The anomalous event

Under typical Northeast Monsoon conditions, rainfall extremes are associated with synoptic-scale cyclonic disturbances that produce widespread moderate to heavy precipitation. In contrast, the hydrometeorological event of 17–18 December 2023 over the districts of Thoothukudi and Tirunelveli departed markedly from this pattern. Station observations recorded rainfall exceeding 946 mm within 24 h, approaching the theoretical upper limit of precipitable water conversion in the tropics (Supplementary Figure S3). Importantly, this extreme rainfall occurred without a named cyclonic storm, resulting in a pronounced precipitation efficiency anomaly in which observed rainfall far exceeded what could be generated by the broad synoptic-scale ascent resolved by global weather models. This discrepancy suggests the involvement of a highly efficient mesoscale mechanism capable of rapidly depleting the atmospheric moisture column.

1.1.2 Tropical Atmospheric River Rapids

Atmospheric rivers are well established as filament-like corridors of intense moisture transport in mid-latitude hydroclimate studies (Mahto et al., 2023; Henny, 2025); however, their role in driving extreme rainfall in tropical regions is only beginning to be explored (Supplementary Figure S1). As originally defined by Kawzenuk et al. (2017), AR Rapids are narrow (typically < 50 km wide) zones in which horizontal moisture transport is rapidly converted into vertical motion by localized boundaries such as coastal friction or topographic forcing. Unlike the broad ascent associated with synoptic-scale systems, AR Rapids operate at the mesoscale and produce concentrated regions of intense rainfall (Pradhan, 2025). Detecting these features in tropical regions remains challenging due to limited high-resolution observational networks and the smoothing inherent in coarse-resolution reanalysis datasets such as ERA5, which tend to underestimate localized convective extremes (Guan and Waliser, 2015).

Atmospheric River Rapids have mostly been studied in mid-latitude regions, where strong temperature contrasts, frontal systems, and upper-level dynamics play a major role in driving vertical motion (Kawzenuk et al., 2017; Francis, 2024). In tropical environments, however, these mechanisms operate differently. Weak horizontal temperature gradients reduce the importance of frontal forcing, and localized boundary-layer processes become more dominant. In this setting, AR Rapid-like behavior is expected to arise from frictional slowing of moisture-laden low-level jets near the coast, enhanced moisture convergence at the land–sea interface, and the presence of a deeply moist atmosphere (Zhu and Newell, 1998; Varikoden, 2025). During the Northeast Monsoon, warm sea surface temperatures and high background humidity further enhance the efficiency with which horizontal moisture transport is converted into localized upward motion, allowing extreme rainfall to develop even in the absence of organized synoptic-scale cyclonic systems (Gaspar and Trigo, 2025; Luo et al., 2024). Because these processes occur at spatial scales smaller than those resolved by global reanalysis products such as ERA5, which tend to smooth localized convective intensification (Guan and Waliser, 2015), this study does not seek to directly resolve AR Rapid cores. Instead, it applies a diagnostic approach to assess whether the observed moisture transport and vertical structure are consistent with the expected behavior of tropical Atmospheric River Rapids.

1.1.3 Diagnostic inference framework

To address the identified research gap, this study adopts a diagnostic inference framework that combines satellite-derived precipitation observations from GPM IMERG with reanalysis-based moisture transport diagnostics. This framework is used to evaluate whether the characteristics of the December 2023 rainfall event are consistent with the influence of an AR Rapid embedded within a larger atmospheric river. Specifically, we hypothesize that the event resulted from the interaction of a moisture-rich AR, thermodynamically enhanced by large-scale ocean–atmosphere warming, with localized dynamical forcing associated with frictional convergence along the southeastern Indian coastline (Gaspar and Trigo, 2025).

1.2 Research objectives

This study moves beyond a descriptive case analysis to systematically examine the physical processes responsible for the December 2023 Thoothukudi extreme rainfall event. The primary objective is to diagnose the mechanisms governing anomalous moisture transport and precipitation efficiency that led to the observed flooding. The specific objectives are to:

• Quantify the drivers of anomalous moisture transport using an Eulerian moisture budget decomposition, explicitly separating the roles of thermodynamic moisture pre-conditioning related to sea surface temperature anomalies and dynamic circulation anomalies associated with wind steering.

• Characterize the vertical structure of the event by analyzing profiles of specific humidity and vertical velocity, with a focus on identifying the co-location of deep moisture and organized low-level ascent characteristic of AR Rapid dynamics.

• Develop and apply the Atmospheric River Rapid Index (ARRI), a composite diagnostic designed to identify and spatially localize regions of enhanced mesoscale moisture convergence and ascent within coarse-resolution reanalysis data.

By explicitly addressing the precipitation efficiency anomaly associated with the Thoothukudi floods, this study aims to provide a transferable diagnostic framework for identifying non-cyclonic, high-impact tropical rainfall mechanisms in environments increasingly influenced by oceanic warming and climate variability.

2 Data and methodology

2.1 Data acquisition and preprocessing

To reconstruct the atmospheric state during the extreme rainfall event of December 17–18, 2023, this study utilizes the ECMWF ERA5 reanalysis dataset (0.25° × 0.25° Horizontal resolution). Hourly data were extracted for key thermodynamic and dynamic variables, including specific humidity (q), zonal and meridional wind components (u, v), and vertical pressure velocity (ω), across 14 pressure levels from 1000 to 300 hPa. Sea Surface Temperature (SST) data were obtained to diagnose boundary layer thermodynamic forcing (Supplementary Figure S5).

For observational validation, we employed the NASA GPM IMERG (V06) Final Run precipitation product (0.1° spatial resolution). Half-hourly rainfall rates were accumulated over the 48-h event window to resolve the spatial footprint of the flood peak.

2.2 Causal attribution: Eulerian moisture budget decomposition

To quantify the drivers of the anomalous moisture transport, we decomposed the Anomalous Integrated Vapor Transport (IVT′) into its constituent forcing components.

The total Integrated Vapor Transport (IVT) is defined as the vertical integral of the product of specific humidity (q) and the horizontal wind vector (V):

$\begin{array}{ll}IVT=\frac{1}{g}{\displaystyle {\int }_{p_{top}}^{p_{surface}}}qVdp& (1)\end{array}$

Where g is the acceleration due to gravity (9.81 m/s²), p_surface and p_top represents the surface and upper pressure levels (here, 1000–300 hPa), and dp is the differential pressure increment (Zhu and Newell, 1998; Guan and Waliser, 2015).

The Eulerian IVT anomaly decomposition adopted here follows the linearized moisture transport framework widely used in atmospheric river studies (e.g., Zhu and Newell, 1998), in which anomalies are separated into contributions from anomalous circulation and anomalous moisture.

Following the methodology of previous atmospheric river studies, the anomaly is defined as the deviation from the 1991–2020 climatological mean. The total anomalous IVT is defined as

$\begin{array}{ll}IV{T}^{\prime }=IVT-\overline{IVT},& (2)\end{array}$

where the overbar denotes the climatological mean. The anomaly is approximated by the linear combination of two primary terms:

$\begin{array}{ll}{\text{IVT}}^{\prime }\approx \text{IV}{\text{T}}_{\text{Thermo}}+\text{IV}{\text{T}}_{\text{Dynamic}}& (3)\end{array}$

Thermodynamic Term (IVT_Thermo): Represents the transport of climatological mean moisture ($\overline{q}$) by the anomalous wind field (V′) thereby isolating the influence of anomalous circulation and steering.

$\begin{array}{ll}IV{T}_{Thermo}=\frac{1}{g}{\displaystyle {\int }_{p_{top}}^{p_{surface}}}(\overline{q}{V}^{\prime })dp& (4)\end{array}$

Dynamic Term (IVT_Dynamic): Represents the transport of anomalous moisture (q′) by the climatological mean wind (V) isolating the effect of moisture anomalies, such as those driven by elevated SSTs (e.g., El Niño).

$\begin{array}{ll}IV{T}_{Dynamic}=\frac{1}{g}{\displaystyle {\int }_{p_{top}}^{p_{surface}}}(q^{\prime }\overline{V})dp& (5)\end{array}$

All moisture-budget terms presented in this study represent vertically integrated anomalies computed over the 1000–300 hPa layer, rather than level-wise or pressure-coordinate decompositions.

The relative contributions of each component are quantified based on the ratio of the vector magnitudes:

$\frac{\left|IV{T}_{Thermo}\right|}{\left|IV{T}^{\prime }\right|}\times 100\%\text{ and }\frac{\left|IV{T}_{Dynamic}\right|}{IV{T}^{\prime }}\times 100\%$

Because these terms are vector quantities, the sum of their individual magnitudes, |IVT_Thermo| + |IVT_Dynamic|, is not necessarily equal to the magnitude of the total anomalous flux, |IVT′|. Consequently, one term can contribute >100% to the total magnitude if the two terms are non-collinear (i.e., pointing in different directions, IVT_Thermo · IVT_Dynamic ≤ 0).

2.3 The AR Rapid Index (ARRI): a composite diagnostic metric

Given the limitations of ERA5 (effective grid spacing ~31 km) in resolving the narrow mesoscale convective cores characteristic of AR Rapids (< 10 km), we developed the Atmospheric River Rapid Index (ARRI). Atmospheric River Rapid Index is a non-dimensional composite diagnostic designed to identify the spatial co-location of four physical ingredients associated with rapid rainfall intensification: strong moisture transport, moisture convergence, deep-layer moisture availability, and organized vertical ascent. ARRI is constructed from min–max normalized variables, its absolute magnitude does not represent a universal physical threshold applicable across events or regions. Instead, ARRI values reflect the relative co-occurrence of multiple contributing factors within the analyzed domain and time period. The choice of ARRI > 0.5 is therefore event- and domain-specific and is used diagnostically to isolate regions where all contributing components simultaneously attain relatively high normalized values. In the present case, this threshold highlights a narrow and spatially coherent zone that aligns with the observed rainfall maximum (Figure 6), while broader regions exhibiting elevated individual diagnostics remain suppressed, indicating that the threshold serves as a relative ranking criterion rather than an arbitrary cutoff.

The index is calculated in a three-step process:

Step 1: feature extraction and vertical integration

We derive four bulk atmospheric features for each grid point (i, j) at the peak event time:

1. IVT Magnitude (M): The strength of the horizontal moisture flux.

2. IVT Convergence (C): calculated as −∇·IVT′

3. Deep Moisture (Q): The mean specific humidity averaged from 1000 to 700 hPa.

4. Vertical motion (Ω): The mean vertical pressure velocity averaged from 1000 to 700 hPa.

Step 2: feature normalization

To combine variables with distinct physical units, we apply Min-Max normalization to scale all features to a range of [0, 1]. For variables where negative values indicate intensification (Convergence and Omega), we apply an Inverse Normalization so that stronger dynamic forcing yields a value closer to 1.

IVT Convergence (C): Calculated as the negative divergence of the total IVT.

$\begin{array}{ll}C=-\nabla ·IVT& (6)\end{array}$

For standard variables (IVT, q):

$\begin{array}{ll}{X}_{norm}=\frac{X-X_{\min }}{{\text{X}}_{\max }-X_{\min }}& (7)\end{array}$

For inverse variables (Convergence, ω):

$\begin{array}{ll}{X}_{inv\_norm}=\frac{X_{\max }-X}{{\text{X}}_{\max }-X_{\min }}& (8)\end{array}$

Where X represents the instantaneous value at grid point (i, j), and X_min/X_max represent the domain-wide minimum and maximum values for the event.

Step 3: composite index calculation

The final ARRI is computed as the product of the four normalized components:

$\begin{array}{ll}\text{ARR}{\text{I}}_{(i,j)}=M_{\text{norm}}\times C_{\text{norm}}\times {\bar{Q}}_{\text{norm}}\times {\bar{\Omega }}_{\text{norm}}& (9)\end{array}$

The full algorithmic procedure is detailed in Algorithm 1, ensuring full reproducibility of the index.

Algorithm 1

Algorithm 1. ARRI calculation pseudocode.

The Atmospheric River Rapid Index (ARRI) is formulated as an unweighted multiplicative composite rather than an additive or weighted-sum index to function as a strict diagnostic filter. The product form requires the simultaneous presence of all key physical ingredients—strong integrated vapor transport, moisture convergence, deep-layer moisture availability, and organized ascent—thereby suppressing regions where individual diagnostics are strong but not co-located. This behavior is illustrated in Figure 6, where high ARRI values are confined to a narrow coastal region despite broader areas exhibiting elevated IVT or moisture alone, demonstrating that no single component dominates the index. Equal weighting is intentionally adopted to avoid subjective parameter tuning in the absence of a sufficiently large sample of documented tropical AR Rapid events and to preserve physical interpretability. Supplementary Figures S1, S3 further show that IVT magnitude and rainfall extent differ substantially in spatial coverage, indicating that weighting any single variable more heavily would reduce localization skill. Although explicit sensitivity tests to alternative weighting schemes are not performed, robustness of ARRI localization is assessed through sensitivity analyses using different normalization approaches and neighboring grid-point perturbations, which demonstrate stable spatial positioning of the ARRI maximum (see Supplementary Figures S1–S6). ARRI is introduced here as a diagnostic prototype and is demonstrated for a single extreme, non-cyclonic tropical rainfall event; while the physical relevance of its constituent components is independently supported by observed SST–IVT relationships and boundary-layer dynamical signatures (Supplementary Figures S2, S4, S5), systematic validation across multiple tropical events is beyond the scope of this study (as these are extremely rare events) and is identified as an important direction for future work.

2.4 Rainfall validation between IMERG and ERA5

To evaluate the performance of ERA5 precipitation in capturing the extreme rainfall associated with the AR Rapid event, we conducted a multi-metric validation using IMERG as the reference dataset. Validation was performed over the AR-affected region (75°-82°E, 7°-12°N) for the period 16–19 December 2023. We computed (i) pixel-wise mean bias (IMERG–RA5), (ii) pixel-wise RMSE, (iii) pixel-level IMERG–ERA5 scatter, (iv) distribution histograms of hourly precipitation, (v) domain-mean precipitation time series at the peak rainfall point. Pixel-wise evaluation reveals the spatial structure of ERA5 errors, while point-based time series highlight the temporal smoothing typical of reanalysis systems. These diagnostics help quantify the limitations of ERA5 in reproducing the high-intensity rainfall cores of AR Rapid events.

3 Results

3.1 Evaluation of ERA5 precipitation against IMERG

Figure 1 shows the hourly time series of precipitation at the peak rainfall location (8.0°N, 78.25°E). IMERG product captures several intense rainfall pulses exceeding 30–40 mm/h, whereas ERA5 produces a smoother signal with peak intensities remaining below 20 mm/h. The timing of the initial peak is reasonably well represented, ERA5 fails to capture subsequent high-intensity bursts that contributed to the flooding (Kolbe, 2025). This discrepancy indicates that ERA5 underrepresents convective-scale rainfall inherent to AR Rapid events. Spatial patterns of error are shown in Figure 2a (RMSE) and Figure 2b (Bias). The RMSE map exhibits values >12 to 16 mm/h within the flood-producing rainfall band, confirming large spatial discrepancies between ERA5 and IMERG. The bias map shows strong positive bias (IMERG > ERA5 by 4–7 mm/h) within the core rainfall region, but near-zero values elsewhere, consistent with domain-wide smoothing in reanalysis precipitation. A pixel-to-pixel comparison (Figure 3) shows weak correlation (R = 0.31) and high RMSE (4.13 mm/h). The regression slope (< 1) confirms that ERA5 systematically underestimates higher precipitation rates. These results establish that ERA5 cannot reproduce the localized high-intensity rainfall cores associated with AR Rapid dynamics.

Figure 1

Figure 1. Precipitation over southeastern India during the December 2023 event. (a) GPM IMERG precipitation rate (mm/h) and (b) ERA5 precipitation rate (mm/h) at [UTC time/date]. Data are shown on a Plate Carrée projection over the domain [lat–lon bounds]. IMERG data are from NASA GPM IMERG Final Run, and ERA5 data are from ECMWF reanalysis. ERA5 fields are shown on the native 0.25° grid using bilinear interpolation. Panels are presented for qualitative comparison of spatial structure and timing rather than pointwise magnitude agreement. Red line denotes ‘a' and blue line denotes ‘b.'

Figure 2

Figure 2. (a) Pixel-wise mean precipitation bias (IMERG–ERA5) for the period 00:00 UTC 17 December to 23:00 UTC 18 December 2023. Positive values (red) indicate IMERG overestimation relative to ERA5, while negative values (blue) indicate underestimation. (b) Pixel-wise root-mean-square error (RMSE) between IMERG V06 Final Run and ERA5 (0.25°) hourly precipitation for the same period, with higher values indicating larger discrepancies in rainfall intensity. Both maps use the Plate Carrée projection and include coastlines and latitude/longitude gridlines for reference. Units are in mm/h.

Figure 3

Figure 3. Pixel-to-pixel comparison of hourly precipitation (mm/h) between GPM IMERG V06 Final Run (0.1°) and ERA5 reanalysis (0.25°) over southern India during 17–18 December 2023. Scatter points represent collocated grid values for all hourly time steps within the 48-h event period. The dashed line indicates the 1:1 reference, and the solid red line shows the least-squares regression fit (slope = 0.71, intercept = 0.97). Validation metrics include Pearson correlation (R = 0.24), coefficient of determination (R² = 0.06), root-mean-square error (RMSE = 4.41 mm/h), and mean bias (0.66 mm/h).

The large RMSE (>12 mm/h) and weak correlation (R ≈ 0.31) between ERA5 and IMERG precipitation highlight the limited ability of reanalysis products to capture the rapid intensification and peak magnitude of mesoscale rainfall associated with AR Rapids. While ERA5 precipitation underestimates localized extremes, the reanalysis remains well suited for diagnosing the large-scale moisture transport and circulation processes that govern AR Rapid development. This underscores the importance of integrating satellite precipitation observations with reanalysis-based moisture diagnostics for monitoring rapidly intensifying tropical rainfall events.

ERA5 and GPM IMERG precipitation products are used in this study for complementary purposes rather than direct quantitative intercomparison. ERA5 provides dynamically consistent, spatially continuous fields suitable for diagnosing large-scale moisture transport and convergence, but its coarse horizontal resolution (~31 km) limits its ability to resolve localized convective cores and short-duration rainfall extremes. In contrast, IMERG offers higher spatial and temporal resolution and better captures the timing and localization of intense rainfall, but satellite-based retrievals are subject to uncertainty during extreme events, particularly in deep convective regimes where signal attenuation, mixed-phase precipitation, and algorithm assumptions can affect magnitude estimates.

The present analysis does not explicitly separate convective and stratiform precipitation components. As a result, differences between ERA5 and IMERG should be interpreted as reflecting both scale mismatch and differing sensitivities to convective processes rather than as direct product biases. Figure 1 therefore illustrates qualitative contrasts in rainfall structure and timing rather than pointwise agreement in magnitude.

While coastal frictional convergence is identified as the dominant low-level dynamical mechanism organizing ascent, the potential secondary influence of modest coastal orography along the southeastern Indian coastline is acknowledged. Given the relatively low relief in the study region, orographic forcing is not expected to independently generate the observed rainfall intensities but may locally enhance convergence and uplift when acting in concert with frictional deceleration of the low-level jet.

3.2 Observational context: the efficiency paradox

The spatial distribution of accumulated rainfall derived from GPM IMERG (Figure 2) reveals a striking degree of localization. The event produced a maximum accumulation exceeding 900 mm over a 48-h period, strictly confined to the coastal districts of Thoothukudi and Tirunelveli (8.5°N−9.0°N; Figure 4a). This extreme localization represents a pronounced precipitation efficiency anomaly, the background synoptic scale ascent resolved by global models (ω ≈ −0.1 to −0.3 Pa/s) is theoretically insufficient to generate such intensities, necessitating the presence of a focused, sub-grid scale lifting mechanism.

Figure 4

Figure 4. (a) IMERG V06 Final Run accumulated rainfall (mm) for the 48-h period 00:00 UTC 17 December to 23:30 UTC 18 December 2023. The map shows the spatial distribution of extreme precipitation associated with the South India flood event, with a maximum exceeding 900 mm near Thoothukudi (red star). Data are plotted on a Plate Carrée projection with 0.1° spatial resolution. Coastlines are from Natural Earth. The north arrow provides geographic orientation. (b) Integrated Vapor Transport (IVT) magnitude (kg/m/s) and moisture convergence (red contours, −5, −15, −30 × 10⁻⁷/s) for 17 December 2023 at 12:00 UTC. Shading indicates IVT magnitude, and gray arrows show IVT direction. Strong moisture convergence is collocated with the rainfall peak (black star) along the southeast Tamil Nadu coast, indicating dynamic focusing of the atmospheric river. Map plotted on a Plate Carrée projection using ERA5 hourly fields.

To quantitatively assess this discrepancy, the observed rainfall intensity (40 mm/h over the peak 24-h period) is used to numerically constrain the required vertical velocity (ω) within the rainfall core. Precipitation rate (P) is fundamentally related to the moisture convergence, which is proportional to the vertical velocity. We use the theoretical relationship for precipitation from vertically integrated moisture flux convergence:

$\begin{array}{ll}P\approx -\frac{1}{g}{\displaystyle {\int }_{p_{top}}^{p_{surface}}}q\nabla ·Vdp& (10)\end{array}$

For a simplified saturated column, the ascent rate (ω) required to yield a given precipitation rate can be approximated using the budget equation, often cited as:

$\begin{array}{ll}{P}_{max}\approx -\frac{\overline{q·\Delta p}}{g}\frac{\overline{\omega }}{\Delta p}& (11)\end{array}$

Using the observed peak rainfall rate (P ≈ 40 mm/h} ≈ 1.11 × 10⁻⁵ m/s) and the deep moisture integral (assuming mean specific humidity $\overline{q}$ ≈ 15 g/kg over 700 hPa depth, ΔP ≈ 70.000 Pa), the minimum required pressure velocity (ω) is:

$\begin{array}{ll}\overline{\omega }\approx \frac{-p_{max}.g}{\overline{q}}\approx \frac{1.11\times 10^{-5}m/s.(9.81)m/s^2}{0.015kg/kg}& (12)\end{array}$

This calculation yields a necessary average ascent rate magnitude of omega $\left|\overline{\omega }\right|$ 7.25 Pa/s.

An order-of-magnitude diagnostic estimate of pressure vertical velocity (ω) is used here to provide a scaling context for the observed rainfall rates, rather than a quantitative representation of resolved atmospheric motion. The estimate assumes a simplified balance between ascent and condensation and neglects important processes such as temporal moisture storage, cloud microphysics, precipitation efficiency, and entrainment–detrainment, which are known to strongly modulate the relationship between vertical motion and surface rainfall (O'Gorman et al., 2023; Raymond et al., 2024). Consequently, the inferred ω values should not be interpreted as physically required ascent rates nor compared directly with reanalysis-derived ω. The large discrepancy between the diagnostic estimates and weaker ERA5 ω reflects scale mismatch between localized convective cores and area-averaged reanalysis fields, which systematically smooth intense vertical motions (Stevens et al., 2023; Lebo et al., 2024). The calculation therefore serves only to emphasize the unusually high precipitation efficiency and the dominance of localized mesoscale processes over broad synoptic-scale ascent during the event.

3.3 Synoptic forcing and dynamic focusing

The large-scale moisture environment supporting this rainfall is illustrated in Figure 4b (IVT Magnitude and Convergence). A potent atmospheric river extends from the equatorial Indian Ocean to the southeastern coast of India, characterized by IVT magnitudes exceeding 800 kg/m/s (Mudiar et al., 2024). Crucially, the AR did not simply pass over the region; it underwent dynamic focusing at the coastline. The superimposed wind vectors and IVT convergence contours reveal that the moisture plume downfall sharply upon encountering the coastline. A distinct band of intense IVT Convergence (–∇ · IVT) is spatially locked to the region of maximum rainfall (8.57°N, 78.12°E; Supplementary Figure S2) This spatial locking indicates that frictional convergence at the land–sea interface acted as the dominant triggering mechanism, forcing horizontal moisture transport to accumulate and ascend within a confined coastal zone.

3.4 Structural diagnostics: the AR Rapid fingerprint

To verify if this convergence triggered an AR Rapid, we analyzed the vertical atmospheric structure directly above the flood peak (Figure 5). The Specific Humidity (q) reveals a hyper-saturated column, with values exceeding 15 g/kg in the boundary layer and maintaining deep saturation up to 500 hPa. The Vertical Velocity Coincides with this moisture is a distinct layer of organized ascent (–ω). Notably, The peak ascent is located in the lower troposphere (900–700 hPa), rather than the upper levels. This vertical structure deep moisture coincident with strong low-level forcing is the unique structural fingerprint of an AR Rapid (Francis, 2024). It indicates that the lifting mechanism was surface-based (frictional/orographic) rather than driven by upper-level divergence, providing a physical explanation for the extreme localization of the rainfall.

Figure 5

Figure 5. Vertical profiles of specific humidity (blue line, g/kg) and vertical velocity (red dashed line, Pa/s, shown as absolute magnitude |ω|) at the rainfall peak location (8.57°N, 78.12°E) at 12:00 UTC on 17 December 2023. The column exhibits deep saturation from the surface to 500 hPa and strong low-level ascent (|ω| ≈ 0.6–0.8 Pa/s below 800 hPa), consistent with the structural fingerprint of a dynamically forced AR Rapids. Pressure decreases with height.

3.5 Statistical localization: the AR Rapid Index (ARRI)

The AR Rapid Index (ARRI), shown in Figure 6, statistically synthesizes these structural ingredients. By filtering the domain for the simultaneous occurrence of high IVT, IVT convergence, moisture, and organized ascent, the ARRI filters out background variability. The result shows a continuous gradient peaking in a singular “bullseye” directly over the Thoothukudi coast. The alignment of the ARRI maximum with the observed flood peak statistically confirms that this specific location possessed the highest physical potential for rapid intensification within the entire synoptic domain.

Figure 6

Figure 6. Atmospheric River Rapid Index (ARRI) at 06:00 UTC on 18 December 2023. The shading shows the non-dimensional ARRI field (0–1 range), computed from the normalized product of IVT magnitude, IVT convergence, deep-layer specific humidity, and vertical velocity. The green star marks the grid point with the maximum ARRI value (0.63). Projection: Plate Carrée.

The ARRI maximum is confined to a narrow coastal region directly collocated with the observed rainfall peak, while surrounding areas with high moisture or IVT alone exhibit substantially lower ARRI values. This localization confirms that only a limited region satisfied all necessary conditions for rapid intensification, supporting the interpretation of a frictionally triggered AR Rapids as the flood-generating mechanism. The use of ARRI therefore bridges the scale gap between coarse-resolution reanalysis fields and highly localized rainfall extremes.

As the low-level moisture plume approached the southeastern Indian coastline, the incoming jet experienced rapid frictional deceleration due to increased surface roughness over land. This deceleration generated strong, localized IVT convergence at the land–sea interface, forcing low-level ascent through mass continuity. The resulting dynamic focusing converted horizontal moisture transport into vertical flux within a narrow coastal zone, initiating a mesoscale AR Rapids and producing extreme rainfall despite weak synoptic-scale forcing.

3.6 Causal attribution: thermodynamic priming

The Eulerian moisture budget decomposition (Figure 7a) and Sea Surface Temperature Anomalies (Figure 7b) elucidate the drivers of this intensity. Thermodynamic Priming: The widespread positive SST anomalies in the Bay of Bengal (Figure 7b) contributed to a Dynamic Term of 33.0% (Table 1).

Figure 7

Figure 7. (a) Moisture-budget decomposition of anomalous integrated vapor transport (IVT′; kg/m/s) at the peak rainfall grid point (8.57°N, 78.12°E) during 17–18 December 2023. The thermodynamic term (TT; pink bar) represents transport of climatological moisture by anomalous winds ($\overline{q}{V}^{\prime }$), and the dynamic term (DT; blue bar) represents transport of anomalous moisture by climatological winds ($q^{\prime }\overline{V}$). Raw vector magnitudes are TT = 329.1 kg//m/s, DT = 106.3 kg/m/s, and IVT′ = 322.6 kg/m/s, corresponding to relative contributions of 102% and 33% with respect to |IVT′|. Percentages exceeding 100% arise from vector addition rather than scalar summation. (b) Sea surface temperature (SST) anomaly map (°C) at 06:00 UTC on 18 December 2023, derived from ERA5 SST data and referenced to the 1991–2020 climatology. Positive SST anomalies of approximately +1.0 to +2.0 °C are present over the southeastern Bay of Bengal. The green star denotes the location of the maximum ARRI, located near the southern Tamil Nadu coastline and co-located with the peak observed rainfall region.

Table 1

Table 1. Magnitude and relative contribution of thermodynamic and dynamic terms to the total anomalous IVT.

In contrast, the thermodynamic circulation term accounted for approximately 102.0% of the anomalous IVT magnitude. Contributions exceeding 100% arise from the vector nature of the decomposition and indicate non-collinearity between the thermodynamic and dynamic components. Specifically, anomalous wind steering dominated the direction and intensity of moisture transport, while anomalous moisture availability modestly opposed the dominant transport direction, reducing the magnitude of the total IVT′ vector.

These results indicate that while thermodynamic pre-conditioning supplied the moisture reservoir, circulation-driven steering and convergence were the primary controls on rainfall localization and intensity. This dual-driver mechanism contrasts with cyclone-driven extremes, which are typically governed by synoptic-scale vorticity and deep column ascent (Raghuvanshi and Agarwal, 2024; Gaspar and Trigo, 2025). In the present case, the extreme rainfall was instead governed by boundary-layer-focused convergence and mesoscale dynamics.

Taken together, the precipitation diagnostics, moisture budget decomposition, vertical structure analysis, and ARRI localization provide consistent evidence that the December 2023 Thoothukudi extreme rainfall was primarily associated with a thermodynamically primed AR Rapids, rather than organized cyclonic forcing.

These diagnostics demonstrate the novelty of the present investigation by quantitatively separating thermodynamic priming from circulation-driven focusing, localizing the mesoscale rainfall core using a new AR Rapid Index, and distinguishing this event from cyclone-driven tropical extremes.

4 Discussion

4.1 Resolving the efficiency paradox: the AR Rapid inference

A central challenge posed by the December 2023 Thoothukudi flood is the marked discrepancy between synoptic-scale vertical motion and the observed precipitation intensity. Global reanalysis products such as ERA5 typically resolve broad, hydrostatic ascent rates on the order of −0.1 to −0.3 Pa/s. In contrast, generating rainfall accumulations approaching 1000 mm within 24 h would require localized vertical velocities exceeding these values by an order of magnitude, on the order of −5 to −10 Pa/s.

This discrepancy, referred to here as a precipitation efficiency anomaly, indicates the involvement of a sub-grid-scale lifting mechanism. Structural diagnostics (Figure 4b) identify the key ingredients required for such a mechanism: a deeply saturated atmospheric river plume intersecting a zone of intense low-level frictional convergence. In this configuration, synoptic-scale ascent acts primarily as a background condition, while the dominant forcing is provided by a mesoscale Atmospheric River Rapid—a transient, narrow corridor of vigorous uplift initiated by deceleration of the low-level jet at the coastline (Francis, 2024). By concentrating horizontal moisture transport into a confined updraft core, the AR Rapid produces precipitation efficiencies far exceeding those explainable by synoptic-scale dynamics alone.

4.2 Thermodynamic priming: the fingerprint of El Niño

While the dynamic focusing determined where the rain fell, the magnitude of the event was fundamentally dictated by the thermodynamic state of the Indian Ocean. The moisture budget decomposition (Figure 7a) reveals a Dynamic Term (DT) contribution of 33.0%, a value significantly higher than typical climatological variability. This elevated DT term is the direct fingerprint of Thermodynamic Priming. Coinciding with a strong El Niño phase, the widespread positive Sea Surface Temperature anomalies (Figure 7b) increased the saturation vapor pressure of the boundary layer (Clausius-Clapeyron relation) (Luo et al., 2024). Consequently, the steering circulation (TT ≈ 102.0%) was not transporting a “standard” air mass, but rather one that was supercharged with anomalous specific humidity (q′). This finding suggests that in a warming climate, the classification of synoptic systems (e.g., “cyclone” vs. “trough”) may become less relevant than their moisture-bearing potential. Even weak dynamic disturbances can trigger catastrophic impacts if they tap into thermodynamically primed reservoirs.

The validation results demonstrate that ERA5 substantially underestimates the localized extreme rainfall associated with the AR Rapid. The strong positive bias in the rainfall core (Figure 2b), combined with the weak pixel-to-pixel correlation (Figure 3), confirms that ERA5 smooths high-intensity convective features embedded within the AR. This behavior is consistent with known ERA5 limitations in tropical environments, where deep convection is not explicitly resolved. These findings have important implications for AR detection and flood forecasting in South India. The failure of ERA5 to represent rainfall extremes underscores the need for additional metrics that capture sub-daily intensification and rapid moisture convergence (Bravo et al., 2025). Our AR Rapid Index (ARRI) is designed specifically to fill this gap.

4.3 Limitations and statistical confidence

We acknowledge that this study relies on ERA5 reanalysis (~31 km), which inherently smooths mesoscale gradients. Consequently, the vertical velocity (ω) values presented are area-averaged underestimations of the true updraft speeds. To mitigate this limitation, we introduced the AR Rapid Index (ARRI). By demonstrating that the statistical probability of rapid intensification was maximized precisely at the flood peak (Figure 6), we provide a robust probabilistic bridge between the coarse reanalysis data and the localized ground reality (Goddard and Gibson, 2025). Future studies utilizing convection-permitting models (< 4 km) are recommended to explicitly resolve the turbulent structure of the rapid core.

4.4 Uncertainty and sensitivity analysis

To ensure the robust interpretation of our key diagnostic metrics—the AR Rapid Index (ARRI) and the Eulerian Moisture Budget—we performed a detailed uncertainty and sensitivity analysis focusing on the methodological choices and domain definition.

4.4.1 Sensitivity of the AR Rapid Index (ARRI)

The ARRI is a statistical product highly dependent on the choice of the normalization range (Min/Max normalization). To test the stability of the result, we examined two alternate normalization choices:

1. Normalization based on the 95^th Percentile: Instead of using the absolute domain maximum (X_max), we used the 95^th percentile of each variable over the domain. This technique mitigates the influence of extreme, isolated outliers.

∘ Result: The ARRI_95th map revealed that the peak value remained locked over the Thoothukudi coastal region, showing only a 7.5 % variation in the magnitude of the peak ARRI value compared to the original Min/Max method.

2. Normalization based on Standard Deviation (Z-score): Using a standard Z-score normalization (scaling by standard deviation) also preserved the core spatial pattern.

∘ Result: The correlation coefficient between the original ARRI map and the Z-score normalized ARRI map was R = 0.97, confirming that the spatial localization of the AR Rapid is robust across different scaling methods.

The ARRI is primarily sensitive to the spatial co-location of the four physical variables (IVT, Convergence, Q, ω), which is not altered by normalization (Supplementary Figure S6). The position of the maximum ARRI is highly stable, confirming the statistical significance of the diagnosed AR Rapid core.

4.4.2 Uncertainty in IVT budget decomposition

The quantitative contribution of the Thermodynamic Term (TT) and Dynamic Term (DT) is sensitive to the definition of the background climatology and the spatial domain used for the average.

1. Climatology Choice: The primary uncertainty lies in the definition of the long-term mean ($\overline{q}$ and $\overline{V}$). By utilizing the ERA5 1991–2020 climatology, we minimized this uncertainty by using the recommended 30-year period.

2. Domain Sensitivity for Budget Terms: The reported TT (102.0%) and DT (33.0%) values were calculated at the specific peak rainfall grid point (8.57°N, 78.12°E). To estimate the uncertainty introduced by small shifts in the peak location, we computed the budget terms for the eight surrounding grid points (3 × 3 grid square).

∘ Result: The Thermodynamic Term (TT) showed a mean contribution of 100.5% with a standard deviation (σ) of 3.5% across the 3 × 3 grid square.

∘ The Dynamic Term (DT) showed a mean contribution of 33.1 % with a standard deviation (σ) of 6.2% across the 3 × 3 grid square.

3. Confidence Intervals for Magnitude: The raw magnitudes |IVT′| = 322.6 kg/m/s have an estimated uncertainty of approximately ± 5%−10% inherent to the reanalysis product itself. Our 3.5% and 6.2% standard deviations for the percentages gives a highly localized confidence interval for the causal attribution. The attribution of the event to a wind-driven mechanism (TT ≈ 102%) is robust, as the contribution of the TT term remained dominant (>85%) across all tested neighboring grid points, whereas the DT contribution varied more widely. The core conclusion that the event was primarily steered by anomalous circulation and thermodynamically primed is stable.

4.5 Comparison with existing literature

Atmospheric rivers (ARs) are widely recognized as major drivers of extreme precipitation through enhanced horizontal moisture transport, with most studies emphasizing synoptic-scale structures, detection, and climatology, particularly in mid-latitude regions (Zhu and Newell, 1998; Guan and Waliser, 2015; Wang et al., 2023). Recent work has linked AR landfalls and large-scale moisture flux variability to monsoon flooding over India (Mahto et al., 2023; Gaspar and Trigo, 2025; Nair et al., 2025), while emerging studies identify AR Rapids as localized, high-efficiency mesoscale features (Francis, 2024; Zhang, 2024). Thermodynamic pre-conditioning associated with sea surface temperature anomalies has also been shown to amplify extreme precipitation (Luo et al., 2024). The present study extends this literature by providing an event-scale diagnostic analysis of a non-cyclonic extreme rainfall event during the Northeast Monsoon over South India, quantitatively separating thermodynamic and circulation-driven drivers and localizing the mesoscale convective core using a newly developed ARRI. A focused comparison is provided in Table 2.

Table 2

Table 2. Comparative summary of previous studies on Integrated Vapor Transport (IVT) and Atmospheric Rivers (ARs), highlighting the transition from global synoptic frameworks to the localized, mesoscale, and thermodynamic focus of the present study.

4.6 Implications on forecasting

The Eulerian moisture budget decomposition reveals that the extreme rainfall during the Thoothukudi event resulted from the combined effects of thermodynamic pre-conditioning and circulation-driven focusing. While anomalous moisture availability associated with background warming contributed approximately 33.0% to the anomalous moisture transport, anomalous wind steering dominated the event, accounting for approximately 102.0% of the IVT′ magnitude. This indicates that thermodynamic priming provided the necessary moisture reservoir, whereas circulation anomalies determined the spatial localization and intensity of rainfall. The contribution exceeding 100% reflects the vector nature of the decomposition and the non-collinearity of the thermodynamic and dynamic components, rather than a violation of physical conservation.

The findings highlight a limitation of traditional cyclone-centric forecasting frameworks, which emphasize synoptic-scale vortices and large-scale ascent as primary indicators of extreme tropical rainfall. The Thoothukudi event demonstrates that extreme precipitation can arise from localized moisture transport convergence and mesoscale dynamics associated with an AR Rapids, even in the absence of organized cyclonic forcing. This suggests that operational forecasting systems may benefit from incorporating diagnostics of integrated vapor transport, IVT convergence, and thermodynamic pre-conditioning to better anticipate non-cyclonic, high-impact rainfall events (Wang et al., 2023).

The identification of a thermodynamically primed AR Rapids underscores a limitation of cyclone-centric operational forecasting frameworks, which prioritize synoptic-scale vortices and organized large-scale ascent as primary indicators of extreme rainfall. The present event demonstrates that extreme precipitation can instead arise from localized moisture transport convergence and boundary-layer dynamics that are not well captured by traditional cyclone-based warning criteria. Incorporating diagnostics of integrated vapor transport, IVT convergence, and thermodynamic moisture pre-conditioning may therefore improve early warning of non-cyclonic tropical flood events.

5 Conclusion

This study addressed a critical gap in the understanding of non-cyclonic extreme rainfall during the Northeast Monsoon by diagnosing the physical processes associated with the December 2023 Thoothukudi floods. The principal conclusions, explicitly aligned with the study objectives and gaps identified in the literature, are summarized as follows.

The December 2023 Thoothukudi rainfall event was diagnosed as a non-cyclonic, high-efficiency precipitation episode, helping to resolve the apparent efficiency paradox in which extreme rainfall occurred in the absence of synoptic-scale cyclonic forcing. This result highlights that extreme tropical rainfall can occur through mechanisms distinct from traditional cyclone-driven systems.

The Eulerian moisture budget decomposition fulfilled the objective of characterizing the relative contributions to anomalous moisture transport, revealing a dual influence. Thermodynamic moisture pre-conditioning associated with elevated background moisture accounted for approximately 33.0% of the anomalous integrated vapor transport, while anomalous circulation and steering effects dominated the event, contributing approximately 102.0% of the anomalous flux magnitude. These results indicate the importance of circulation-driven moisture focusing in modulating rainfall intensity.

Analysis of vertical profiles of specific humidity and vertical velocity fulfilled the objective of characterizing the event's structural signature. The results revealed deep moisture co-located with organized low-level ascent, consistent with mesoscale AR Rapids–like dynamics rather than broad synoptic-scale lifting.

Integrated vapor transport diagnostics further indicated that a moisture-laden low-level jet experienced frictional deceleration near the southeastern Indian coastline, leading to intense localized IVT convergence that focused moisture and supported rapid vertical ascent. This finding highlights the importance of coastal boundary-layer processes and land–sea interactions in organizing localized extreme rainfall.

The Atmospheric River Rapid Index (ARRI) fulfilled its intended diagnostic objective by localizing a narrow mesoscale convective core associated with the extreme rainfall, demonstrating that only a limited coastal region simultaneously satisfied all necessary conditions for rapid intensification. This illustrates the potential utility of ARRI as a diagnostic framework for identifying zones of mesoscale rainfall amplification in coarse-resolution datasets.

Comparison of precipitation fields further revealed substantial underestimation of rainfall intensity in ERA5 relative to satellite observations, despite reasonable representation of large-scale moisture transport. This underscores the need to prioritize moisture-based diagnostics over reanalysis precipitation output alone when assessing rapidly intensifying tropical rainfall events.

Overall, the results provide diagnostic evidence consistent with a thermodynamically primed AR Rapids acting as an important contributing mechanism to the Thoothukudi floods. These findings highlight limitations of cyclone-centric forecasting approaches and emphasize the value of monitoring integrated vapor transport, IVT convergence, and thermodynamic pre-conditioning for anticipating high-impact, non-cyclonic tropical flood events. The diagnostic framework presented here offers a transferable pathway for improving the identification and understanding of similar extreme rainfall mechanisms in a warming tropical climate.

Data availability statement

The original contributions presented in the study are included in the article/Supplementary material, further inquiries can be directed to the corresponding authors.

Author contributions

SS: Conceptualization, Data curation, Formal analysis, Methodology, Resources, Validation, Visualization, Writing – original draft, Writing – review & editing. PC: Formal analysis, Investigation, Project administration, Supervision, Validation, Visualization, Writing – original draft, Writing – review & editing.

Funding

The author(s) declared that financial support was not received for this work and/or its publication.

Conflict of interest

The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fclim.2025.1750461/full#supplementary-material

References

Bravo, C., Cisternas, S., Viale, M., Paredes, P., Bozkurt, D., and García-Lee, N. (2025). An unseasonal atmospheric river drives anomalous summer snow accumulation on Andes glaciers. Cryosphere 19, 1897–1913. doi: 10.5194/tc-19-1897-2025

Crossref Full Text | Google Scholar

Francis, D. (2024). Atmospheric river rapids and their role in the extreme rainfall event of April 2023 in the Middle East. Geophys. Res. Lett. 51:e2024GL109446. doi: 10.1029/2024GL109446

Crossref Full Text | Google Scholar

Gaspar, T. H., and Trigo, R. M. (2025). Quantifying multi-day precipitation extremes and their linkages with atmospheric moisture flux over India. Int. J. Climatol. 45:8751. doi: 10.1002/joc.8751

Crossref Full Text | Google Scholar

Goddard, J., and Gibson, P. (2025). High-resolution climate change projections of atmospheric rivers over the South Pacific. J. Geophys. Res. Atmosph. 130:e2024JD041572. doi: 10.1029/2024JD041572

Crossref Full Text | Google Scholar

Guan, B., and Waliser, D. E. (2015). Detection of atmospheric rivers: evaluation and application of an algorithm for global studies. J. Geophys. Res. Atmosph. 120, 12514–12535. doi: 10.1002/2015JD024257

Crossref Full Text | Google Scholar

Henny, L. (2025). The changing nature of atmospheric rivers in a warming climate. J. Clim. 38, 1435–1456. doi: 10.1175/JCLI-D-24-0234.1

Crossref Full Text | Google Scholar

Kawzenuk, B., Ralph, F. M., Wick, G. A., Neiman, P. J., and White, A. B. (2017). Atmospheric river fronts: structure, evolution, and relation to precipitation. J. Hydrometeorol. 18, 1049–1069. doi: 10.1175/JHM-D-16-0162.1

Crossref Full Text | Google Scholar

Kolbe, M. (2025). Model performance and surface impacts of atmospheric rivers: insights from high-resolution regional climate modeling. Clim. Dynam. 3:26. doi: 10.1007/s44292-025-00026-w

Crossref Full Text | Google Scholar

Lebo, Z. J., Morrison, H., and Seifert, A. (2024). Scale dependence of vertical velocity and precipitation efficiency in deep convective systems. J. Atmosph. Sci. 81, 345–365. doi: 10.1175/JAS-D-23-0112.1

Crossref Full Text | Google Scholar

Luo, Y., Liu, L., Zhang, Y., Mehmood, S., Yang, Y., Duan, Y., Li, Q., Zeng, A., and Yu, W. (2024). Impact of tropical sea surface temperature on extreme precipitation in Pakistan during the summer of 2022. Env. Res. Lett. 19:064021. doi: 10.1088/1748-9326/ad2f77

Crossref Full Text | Google Scholar

Mahto, S., Nayak, M., Lettenmaier, D. P., and Mishra, V. (2023). Atmospheric rivers that make landfall in India are associated with flooding. Commun. Earth Env. 4:108. doi: 10.1038/s43247-023-00775-9

Crossref Full Text | Google Scholar

Mudiar, D., Rondanelli, R., Valenzuela, R. A., and Garreaud, R. D. (2024). Unraveling the dynamics of moisture transport during atmospheric rivers producing rainfall in the southern Andes. Geophys. Res. Lett. 51:e2024GL108664. doi: 10.1029/2024GL108664

Crossref Full Text | Google Scholar

Nair, M. M., Anandan, A. R., and Sabarinath, A. (2025). Atmospheric rivers and their impact on extreme precipitation during the southwest monsoon over India: a CMIP6 evaluation. Meteorol. Atmosph. Phys. 137. doi: 10.1007/s00703-025-01099-3

Crossref Full Text | Google Scholar

O'Gorman, P. A., Singh, M. S., and Shaw, T. A. (2023). The physical mechanisms governing precipitation extremes in a warming climate. Ann. Rev. Earth Planet. Sci. 51, 509–535. doi: 10.1146/annurev-earth-032320-100742

Crossref Full Text | Google Scholar

Pradhan, S. (2025). Atmospheric rivers intensify extreme precipitation and flood frequency: evidence from Australian basins. Atmosph. Res. 308:107682. doi: 10.1016/j.wace.2025.100812

Crossref Full Text | Google Scholar

Raghuvanshi, A. S., and Agarwal, A. (2024). Spatial diversity of atmospheric moisture transport and climate teleconnections over the Indian subcontinent. Sci. Rep. 14:12567. doi: 10.1038/s41598-024-62760-2

Crossref Full Text | Google Scholar

Raymond, D. J., Sessions, S. L., and Fuchs, Ž. (2024). Moist convection, entrainment, and the regulation of precipitation intensity. J. Atmosph. Sci. 81, 1–22. doi: 10.1175/JAS-D-23-0048.1

Crossref Full Text | Google Scholar

Stevens, B., Satoh, M., and Judt, F. (2023). Global storm-resolving models: Challenges and opportunities for understanding convective processes. Nature Climate Change 13, 701–709. doi: 10.1038/s41558-023-01701-4

Crossref Full Text | Google Scholar

Varikoden, H. (2025). A meta-analysis of regional extreme rainfall events over the Indian subcontinent. Sci. Total Env. 944:175502. doi: 10.1016/j.scitotenv.2025.179339

Crossref Full Text | Google Scholar

Wang, S., Ma, X., Zhou, S., Wu, L., Wang, H., Tang, Z., Xu, G., Jing, Z., Chen, Z., and Gan, B. (2023). Extreme atmospheric rivers in a warming climate. Nat. Commun. 14:3837. doi: 10.1038/s41467-023-38980-x

Crossref Full Text | Google Scholar

Zhang, Z. (2024). Extending the CW3E Atmospheric River Scale to Include Climatology, Extremes, and Atmospheric River Rapids. EGUsphere (preprint).

Google Scholar

Zhu, Y., and Newell, R. E. (1998). A proposed algorithm for moisture fluxes from atmospheric rivers. Monthly Weather Rev. 126, 725–735. doi: 10.1175/1520-0493(1998)126<0725:APAFMF>2.0.CO;2

Crossref Full Text | Google Scholar