the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 18 May 2026
Revisiting Discharge Envelope Curves: Hydrometeorological Analysis and Lessons from the 4 July 2025 Kerr County Flash Flood
Abstract. Flash floods are a primary driver of flood mortality in semi-arid regions where compound risk factors amplify convective extremes. The 4 July 2025 Kerr County flash flood in the Texas Hill Country-a region colloquially designated "Flash Flood Alley"-represents a benchmark hydrometeorological event. Triggered by a quasi-stationary mesoscale convective system derived from Tropical Storm Barry remnants, the event delivered more than 508 mm of rainfall over 72 hours, with peak 3-hour intensities exceeding the local 1,000-year recurrence threshold (National Weather Service, 2025). Using NOAA Stage IV rainfall estimates, USGS streamflow records, and mass-balance modeling, we reconstruct the hydrometeorological response of the South Fork and mainstem Guadalupe River. Estimated peak discharges at the South Fork ranged from 7,221 to 7,505 m3 s−1 (255,000–265,000 ft3 s−1), approaching or exceeding the regional discharge envelope curve and surpassing the historical record set in July 1932. Antecedent drought may have contributed to runoff through hydrophobic soil behavior, while overnight timing and critical communication gaps compounded human impacts (137 fatalities). Numerous affected structures lay outside FEMA-designated Special Flood Hazard Areas, exposing the limitations of historically calibrated static hazard maps. The findings support a transition toward impact-based warning protocols, expanded monitoring networks, and climate-aware hazard mapping for topographically complex, data-sparse regions.
- Preprint
(6565 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2026-1750', Sheeba Thomas, 30 Jun 2026
-
AC1: 'Reply on RC1', Ayantika Rinti Bose, 30 Aug 2026
We thank the reviewer for their careful reading and constructive comments. The revised manuscript addresses each point below. Our responses follow each comment, with the location of the new or modified text indicated by section and, where available, by line number in the revised version.
Reviewer 1
Comment 1. Please include and label USGS gauges in Figure 3 or, if it is better, create another figure with the gauges clearly labeled, and be consistent in the narrative when referencing them.
Response.
We agree and have revised Figure 3 accordingly. The four USGS streamflow gauging stations used in the analysis are now plotted on the basin map as colored triangles labeled 1 through 4, each connected to its label with a leader line. The expanded legend below the map lists the full station names and numbers:
(1) North Fork Guadalupe River near Hunt (08165300),
(2) Guadalupe River at Hunt (08165500),
(3) Guadalupe River above Bear Creek at Kerrville (08166140),
(4) Guadalupe River at Kerrville (08166200).
The triangle symbols and their fill colors are distinct from the square symbols used for communities and for Camp Mystic, so the legend now separates gauging stations, communities, and Camp Mystic into three symbol classes. We retained a single figure rather than adding a new one because the gauge locations are most informative when viewed together with the basin outline and the North Fork–South Fork confluence.
We have also standardized the narrative. Each station is now identified by name and station number, and the same naming convention is used in Figures 3, 5, and 6 and in Sections 3.2 and 3.3.
Comment 2. Figure 6 looks like a very important figure; the current explanation is not enough. Consider adding more details on how the envelope curve was created and more details on how the other data points were included into this figure.
Response. We agree that Figure 6 carries much of the argument and that its construction required fuller explanation. We have added a paragraph to Section 3.3 (lines 159-171) describing the provenance of the curve and of the plotted points:
- The envelope curve is the potential extreme peak discharge relation of Asquith and Slade (1995), which extends the conterminous United States maximum flood flow envelope of Crippen and Bue (1977) to Texas. It is an empirical upper bound fitted to the largest peak discharges as a function of contributing drainage area, not a probabilistic quantile, and it is plotted here on log-log axes over the drainage area range spanned by the Guadalupe system.
- Solid symbols are documented annual and historical peak discharges at USGS gauging stations in the Guadalupe River basin and adjacent Hill Country basins, taken from the National Water Information System peak-flow files.
- Hollow squares are peak discharges at sites without permanent gauges, compiled from USGS indirect measurements (slope-area, slope-conveyance, and critical-depth methods) and from the documented extreme peak discharge inventory of Asquith and Slade (1995). These points carry larger uncertainty than the gauged values, and the distinction is now stated in the caption as well as the text.
- The four 2025 points are: the North Fork peak from the rating at station 08165300; the Guadalupe River at Kerrville peak from station 08166200; the Guadalupe River at Hunt peak from the USGS high-water-mark indirect measurement at station 08165500; and the South Fork peak from the mass-balance reconstruction described in Section 3.3, plotted at the midpoint of the estimated range with the range itself now shown as a vertical error bar.
The caption has been expanded correspondingly, and the unit-discharge comparison with the 1935 Seco Creek event is now stated in both customary and SI units.
Citation: https://doi.org/10.5194/egusphere-2026-1750-AC1
-
AC1: 'Reply on RC1', Ayantika Rinti Bose, 30 Aug 2026
-
RC2: 'Comment on egusphere-2026-1750', Anonymous Referee #2, 05 Aug 2026
This manuscript reconstructs and assesses the catastrophic 4 July 2025 flash flood in Kerr County, Texas. The extreme event offers valuable insights into climate-driven flood extremes in semi-arid zones and carries great scientific and practical value. However, enhancements are needed in the description of technical details and discussions. Specific recommendations are as follows:
1. The core conclusion that the flood exceeded the regional discharge envelope curve relies only on a first-order mass-balance model. Cross-verification with standard hydrological models would strengthen this result.
2. As revisiting discharge envelope curves is the main theme, supporting evidence needs enrichment. Comparisons with other extreme flash floods are missing, and the inference that warming shifts envelope curves upward lacks quantitative climate or return-period data.
3. Discussions are restricted to Texas watersheds. Global comparisons of similar semi-arid flash flood risks and detailed solutions for non-stationary flood models and impact-based warnings will broaden the study’s applicability.
4. Fatality data are only displayed in simple classified charts without quantitative risk analysis. Quantify exposed populations via inundation data, and build correlations between water depth, warning delay and mortality to distinguish vulnerabilities of camps, residences and RV parks.Citation: https://doi.org/10.5194/egusphere-2026-1750-RC2 -
AC2: 'Reply on RC2', Ayantika Rinti Bose, 30 Aug 2026
We thank the reviewer for their careful reading and constructive comments. The revised manuscript addresses each point raised below. Our responses follow each comment, with the location of the corresponding new or revised text identified by section and, where applicable, by line number in the revised manuscript.
Comment 1. The core conclusion that the flood exceeded the regional discharge envelope curve relies only on a first-order mass-balance model. Cross-verification with standard hydrological models would strengthen this result.
Response. We appreciate the concern and have added an independent cross-check. We respectfully suggest, however, that a calibrated rainfall-runoff model would not provide the verification the reviewer seeks in this particular basin, for the following reasons.
The envelope-curve conclusion does not rest on the mass-balance estimate alone. The manuscript’s central claim, that the 4 July 2025 event approached or exceeded the regional discharge envelope, is anchored by the USGS indirect measurement of 8,892 m3 s-1 (314,000 ft3 s-1) at the Hunt gauge (08165500), derived from surveyed high-water marks by the U.S. Geological Survey and independent of any modeling performed here. That value alone exceeds the previous basin record of July 1932 and plots at or above the Asquith and Slade (1995) curve. The mass-balance reconstruction serves the narrower purpose of partitioning that flow between the two forks. We have revised the abstract and Section 3.3 to make this hierarchy of evidence explicit, so that the modeled South Fork range is presented as a corroborating rather than a load-bearing estimate. (Line 10-16, Line 199-205)
A second, fully observational cross-check is now included. At the reviewer’s prompting, we added a confluence mass balance that uses only measured quantities. Subtracting the gauged North Fork peak of 1,659 m3 s-1 (58,600 ft3 s-1) from the Hunt indirect measurement of 8,892 m3 s-1 yields a residual South Fork contribution of approximately 7,233 m3 s-1 (255,400 ft3 s-1). This falls just within the lower bound of the independently derived rainfall-based range of 7,221 to 7,505 m3 s-1 (255,000 to 265,000 ft3 s-1). Because the North Fork peaked at 05:30 CDT, roughly 20 minutes after the last stage measurement at Hunt, the North Fork contribution at the time of the Hunt peak was smaller than its own peak, so the residual is a conservative lower estimate of the South Fork peak. Two methods that share no common assumptions, one radar-rainfall based and one gauge based, therefore converge to within better than 1% at the lower bound. This exchange has materially improved the paper and we thank the reviewer for it. The analysis is now presented in Section 3.3 (lines 172-182). better than (Line172-182)
A calibrated hydrologic model could not be validated for this event. The conditions required to constrain a distributed or lumped rainfall-runoff model are absent here. The Hunt gauge failed at 05:10 CDT on the rising limb, before the peak; there is no streamflow gauge on the South Fork; and discharge data are unavailable for the Guadalupe River above Bear Creek. No observed hydrograph therefore exists against which infiltration parameters, channel roughness, or routing parameters could be calibrated or validated for the 4 July event. A model run under these conditions would reproduce whatever peak its assumed parameter set produced, and reporting that peak alongside the observational estimates would convey a precision the data do not support. The authors have published distributed, physically based modeling of Texas flash floods, including the May 2015 Blanco River event (Furl et al., 2017), and the contrast is instructive: that study had a surviving peak-flow record to constrain the model, which the present event does not. (Line 183-192)
First-order mass balance can be a reasonable method in this situation. Post-event reconstruction of peak discharge at ungauged sites, or at sites where the gauge was destroyed, is conventionally performed by indirect and mass-balance methods rather than by simulation, and the regional envelope-curve dataset itself (Crippen and Bue, 1977; Asquith and Slade, 1995) is populated largely by such indirect estimates. Using a mass-balance estimate to compare against that curve is therefore methodologically consistent with the reference dataset. (Line 193-198)
Full hydrodynamic reconstruction of the event, incorporating post-flood lidar and surveyed high-water marks along the South Fork, is a substantial undertaking and is identified in the Conclusions as a priority for follow-on work (line 462).
Comment 2. As revisiting discharge envelope curves is the main theme, supporting evidence needs enrichment. Comparisons with other extreme flash floods are missing, and the inference that warming shifts envelope curves upward lacks quantitative climate or return-period data.
Response. Adopted. We have added a table to Section 3.3 comparing the 2025 South Fork unit discharge with other documented extreme unit discharges in Texas and the south-central United States, including the 1935 Seco Creek event, the 1978 Medina River flood near Pipe Creek, the 1998 and 2002 Blanco River floods at Wimberley, and the May 2015 Blanco River flood (Furl et al., 2017), each reported with contributing drainage area, peak discharge, and unit discharge so that the 2025 values can be read against comparable basins. (Line 206-209,Table 1)
On the climate inference, we agree that the manuscript stated more than it demonstrated. We have revised the text to present the upward-shift argument as a hypothesis supported by thermodynamic reasoning and by existing attribution work rather than as a result of this study. The revised passage cites the Clausius-Clapeyron scaling of atmospheric moisture holding capacity at approximately 6-7% per degree Celsius of warming, the observed increase in extreme precipitation intensity in semi-arid regions (Tabari, 2020; IPCC, 2021), and the event-specific attribution assessments already referenced (ClimaMeter, 2025; Faranda et al., 2022). We have added the return-period context available for this event, namely the approximately 250-year 3-hour basin-average estimate and the exceedance of the local 1,000-year 3-hour threshold at individual radar pixels from NOAA Atlas 14. We now state explicitly that a formal detection-and-attribution analysis of envelope-curve non-stationarity is beyond the scope of this event reconstruction and requires regional analysis across many basins. (Line 210-226)
Comment 3. Discussions are restricted to Texas watersheds. Global comparisons of similar semi-arid flash flood risks and detailed solutions for non-stationary flood models and impact-based warnings will broaden the study’s applicability.
Response. Adopted. We have added a paragraph to Section 6 placing the event in an international context, drawing on documented semi-arid and Mediterranean flash flood disasters with comparable rainfall intensities, short response times, and high fatality tolls, and on flash flood work in arid catchments of the Arabian Peninsula, including work by members of the author team (Hussein et al., 2025). The comparison highlights the recurring combination of thin soils, limited infiltration capacity, steep confined channels, and populations concentrated in valley bottoms. (Line 405-418)
We have also expanded the treatment of the two operational recommendations. For non-stationary flood frequency, the revised text identifies the specific components required, namely time-varying or covariate-dependent distribution parameters, incorporation of large historical and paleoflood peaks to constrain the upper tail, and regional pooling to compensate for short at-site records. For impact-based warnings, the revised text describes the elements of an operational implementation for riverside camps and low-water crossings: site-specific upstream rainfall and stage triggers, pre-defined protective actions tied to each trigger, and redundant delivery pathways, consistent with WMO impact-based forecast and warning guidance.paleo flood(Line 427-432) (Line 388-393)
Comment 4. Fatality data are only displayed in simple classified charts without quantitative risk analysis. Quantify exposed populations via inundation data, and build correlations between water depth, warning delay and mortality to distinguish vulnerabilities of camps, residences and RV parks.
Response. We have strengthened the mortality analysis as far as the data permit, and we appreciate the reviewer directing us toward a more quantitative treatment. Two of the specific requests cannot be met with defensible data for this event, and we describe the reasons and our alternative below.
What we have added.
Section 5.2 now reports site-specific fatality proportions where compatible published occupancy denominators are available and presents structured qualitative comparisons where such denominators are incomplete or unavailable. This allows a direct comparison of outcomes at Camp Mystic and Heart O’ the Hills with those at Waldemar, Mo-Ranch, and Camp La Junta without implying unsupported exposure rates. (Table 2, Line 323-328)
We have added the timing sequence explicitly, from the 01:14 CDT Flash Flood Warning through the 05:01 CDT countywide CodeRED alert against the 03:00 to 05:30 CDT peak inundation window, so that the interval between warning issuance and life-threatening water levels is quantified for each affected reach (lines 338-344). We have also added age-specific and sex-specific fatality proportions with comparison to the national flood-fatality distribution reported by Han and Sharif (2021), which places the strong 0-9 and 60-79 age concentrations in this event against the national pattern (lines 301-309). (Line 301-309)
Exposed population cannot be quantified reliably for this event. The population at risk at 03:00 CDT on 4 July consisted largely of transient occupants: campers and staff at seasonal youth camps, occupants of RV parks and short-term rentals, and holiday visitors. Census and residential parcel data do not represent this population, and site-level occupancy rosters for the affected camps and RV parks are not in the public record. Any exposure denominator derived from residential data would substantially misstate the population actually at risk, and normalized fatality rates computed from it would be misleading. We have stated this limitation directly in Section 5.2 rather than presenting an exposure estimate we cannot support. (Line 329-337)
A depth-warning-mortality relation is not statistically identifiable from a single event. Establishing a correlation among water depth, warning delay, and mortality requires depth and warning-receipt information at the level of the individual fatality. The NCEI Storm Events records provide age, sex, and location-of-death category, but neither depth at the location of death nor the time or channel through which any warning was received. Depths would have to be assigned from a hydraulic model that, as discussed in our response to Comment 1, cannot be validated for this event, and warning delay is effectively constant across the affected corridor because a single countywide alert sequence applied to all locations. A regression fitted across 106 records under these conditions would be an artifact of the assumptions rather than a measured relationship. (Line 345-352)
We have instead presented the comparison across setting types as a structured qualitative contrast, which the data do support, and we identify individual-level exposure and warning-receipt reconstruction as a priority for future survey-based follow-up studies (lines 353-361).
We believe these revisions address the reviewers’ concerns and have improved the manuscript, particularly the independent confluence-based verification of the South Fork discharge and the expanded documentation of Figure 6. We thank both reviewers for their time.
Citation: https://doi.org/10.5194/egusphere-2026-1750-AC2
-
AC2: 'Reply on RC2', Ayantika Rinti Bose, 30 Aug 2026
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 218 | 222 | 25 | 465 | 28 | 24 |
- HTML: 218
- PDF: 222
- XML: 25
- Total: 465
- BibTeX: 28
- EndNote: 24
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
1)Please include and label USGS gauges in Figure 3 or if it is better, create another figure with the gauges clearly labeled and be consistent in the narrative when referencing them.
2) Figure 6 looks like a very important figure; the current explanation is not enough. Consider adding more details on how the envelope curve was created and more details on how the other data points were included into this figure, Figure 6.