Pith. sign in

REVIEW 3 major objections 6 minor 27 references

Paragliding flight logs can act as dense, quantitative probes of thermal convection, with ceiling heights matching the theoretical lifting-condensation-level slope to within 1% in the reference regime.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review

2026-08-04 00:54 UTC pith:XF62EBL6

load-bearing objection A genuinely new crowdsourced thermal dataset with a surprisingly close LCL slope match, but the cell-hour maximum ceiling is an order statistic that needs a sampling-intensity control before the validation is fully trustworthy. the 3 major comments →

arxiv 2608.00241 v1 pith:XF62EBL6 submitted 2026-07-31 physics.ao-ph cond-mat.stat-mechphysics.flu-dyn

From Flight Logs to Atmospheric Science: Paragliders as Convection Sensors for Identifying Thermal Predictors

C\'esar Hern\'andez-Aguayo , Matthieu Cristelli , Michael Benzaquen This is my paper
classification physics.ao-ph cond-mat.stat-mechphysics.flu-dyn
keywords atmospheric convectionthermalsparaglidingflight logslifting condensation levelboundary-layer heightERA5crowdsourced observations
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper tries to establish that crowdsourced paragliding flight logs can serve as a quantitative observational network for atmospheric thermal convection. From 1.47 million climbing segments, the authors extract three observables: thermal strength, vertical-velocity variability, and the maximum ceiling height reached. They show that the ceiling height's dependence on temperature–dewpoint depression matches the theoretical lifting condensation level slope (123.6 m/°C vs 125 m/°C in Low mountains, warm season, midday), which acts as a physics-based calibration of the method. Using a decorrelation procedure against 118 ERA5 predictors, they further claim that boundary-layer height is the leading independent predictor of both ceiling height and thermal strength across all four terrain classes and both seasons. A sympathetic reader would care because this offers a dense, low-cost complement to radiosondes and models for studying how thermals respond to terrain, soil moisture, and boundary-layer depth.

Core claim

The central claim is that a climbing segment of a paraglider is a valid probe of one thermal column, and that cell-hour aggregates of such segments give physically interpretable measures of convection. The empirical slope of the thermal ceiling with respect to temperature–dewpoint depression is 123.6 m/°C in the Low mountains reference case, matching the theoretical lifting condensation level slope of 125 m/°C to within 1%. After orthogonalizing all 118 predictors against that same dewpoint-depression signal, boundary-layer height emerges as the top independent driver of both ceiling height and mean thermal strength in all terrains and seasons. The authors present the LCL match as a first-pr

What carries the argument

The key object is the paragliding climbing segment, treated as a probe of a thermal column: pilots circle inside an updraft while carrying GPS variometers, and the recorded track samples vertical velocity and maximum altitude. Three observables—thermal strength, vertical-velocity variability, and thermal ceiling—are extracted and aggregated to 0.25° per-hour cells. The argument then runs through two steps: a benchmark regression of ceiling height against dewpoint depression, whose theoretical slope of ~125 m/°C is the lifting condensation level (the height at which a rising parcel saturates), and an orthogonalized rank-feature relevance procedure that removes the LCL signal from all predicto

Load-bearing premise

The ceiling observable assumes that at least one pilot in each cell-hour reaches the true thermal top; if weak or slow thermals prevent pilots from topping out—most likely in the cold season and in plains—the measured ceiling becomes an under-estimate that biases the slope and predictor rankings.

What would settle it

Compare paragliding-derived cell-hour ceiling heights against independent estimates of thermal top from ceilometer cloud-base measurements, radiosonde-derived convective condensation level or boundary-layer top, or large-eddy simulations for the same cells and hours; the central claim fails if H_AGL systematically falls short of the independent top in low-activity cells or cold-season plains, or if the 1% slope agreement disappears when restricting to cells with high segment counts.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

Share X Bluesky LinkedIn Reddit HN

If this is right

  • If the method holds, thermal convection can be observed at a density and resolution unreachable by radiosonde networks, across both mountainous and flat terrain.
  • The 1% agreement of ceiling height with the theoretical lifting condensation level slope elevates the ceiling observable from anecdotal to calibrated, allowing cell-hour ceiling data to test boundary-layer schemes.
  • Boundary-layer height being the leading independent driver suggests thermal ceilings and climb strength are primarily set by how deep the mixed layer grows, not just by surface temperature.
  • Soil moisture behaving as the dominant conditional axis at fixed terrain and hour implies dry-soil days should produce systematically higher and stronger thermals, especially over low mountains.
  • The same pipeline applies wherever flight logs and a reanalysis product overlap, potentially opening global convective observations.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The paper leaves implicit that the ceiling estimator is sensitive to sampling intensity: because it is the maximum over all segments in a cell-hour, its expected value rises with the number of segments, so cell-hours with more flights will tend to report higher ceilings unless normalized.
  • The paper's own conditioning on flyable weather means the reported correlations describe convection under soaring-friendly conditions; extending to the full climatology would require correcting for the selection window or supplementing with non-pilot observations.
  • A testable extension is to use transition phases between climbs to measure inter-thermal spacing, turning the same logs into a probe of thermal organization rather than only single-column statistics.
  • Because vertical-velocity variability is controlled by surface fluxes and wind more than by LCL variables, higher-order moments of climb velocities may encode sub-grid turbulence information useful for model parameterization.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

3 major / 6 minor

Summary. The manuscript introduces a crowdsourced observational dataset of atmospheric thermal convection constructed from 1.47 million climbing segments extracted from 110,730 paraglider flight logs over metropolitan France (2017–2024). Three observables are defined: thermal strength E[V̄z], vertical-velocity variability E[σVz], and thermal ceiling H_AGL, the latter being the maximum altitude over climbing segments in a 0.25° × 0.25° cell and one-hour bin. The paper validates the ceiling observable by regressing H_AGL against the dewpoint depression T−T_d and reports a slope of 123.6 m/°C in Low mountains in the Warm season at midday, matching the theoretical LCL slope of 125 m/°C to within 1%. An orthogonalized rank-feature relevance (ORFR) framework applied to 118 ERA5 predictors is then used to identify boundary-layer height as the leading independent predictor of both H_AGL and E[V̄z] across all terrains and seasons. The authors also present a descriptive climatology of the terrain, seasonal, diurnal, cloud, and soil-moisture dependence of the three observables.

Significance. If the methodology is accepted, this work provides a genuinely novel, high-density observational resource for studying boundary-layer convection, complementing radiosondes and LES with an in situ dataset that is both spatially extensive and temporally resolved. The physical separation of the three observables is sensible, and the descriptive climatology—terrain hierarchy, seasonal and diurnal cycles, and soil-moisture effects—is plausible and internally consistent. The explicit use of a thermodynamic benchmark (LCL scaling) is a commendable attempt at first-principles validation, and the ORFR framework is clearly described. However, the central validation and the subsequent predictor ranking rest on a single observable, H_AGL, whose definition as a cell-hour maximum introduces a sampling-intensity bias that is not addressed. The paper also reports slopes and correlations without uncertainty estimates despite strong spatial and temporal dependence in the cell-hour data. These issues are load-bearing for the paper's main claims, so the manuscript requires substantial additional analysis before the conclusions can be accepted.

major comments (3)
  1. [Sec. III A, Eq. (4), Sec. IV C, Table I] H_AGL is defined as the maximum over all climbing segments in a cell-hour bin. Because the number of segments N_{c,d,h} varies strongly across cell-hours (the top-40 cells capture 64% of all segments), E[max] increases with N even if the underlying thermal-top distribution is fixed. If N is positively correlated with convective activity (and hence with T−T_d or blh), the origin-constrained OLS slope in Eq. (12) and the ORFR rankings in Sec. IV D are inflated by sampling intensity rather than physics alone. The paper acknowledges cold-season under-reach and free-flight conditioning, but never controls for N. A minimal fix is to stratify by N or include log N as a covariate in the LCL regression and in the ORFR ranking; a more direct test is to compare the LCL slope on cell-hours with N above a high threshold, where the maximum is a good approximation to the true ceiling.
  2. [Table I, Figs. 5–7, Sec. IV D] All slopes and Spearman correlations are reported as point estimates without confidence intervals or standard errors. The cell-hour observations are not independent: the same ERA5 grid cell appears across many days and neighboring cells share the same reanalysis forcing. The effective sample size is therefore far smaller than the reported n values (e.g., n=29,023 for Low mountains Warm midday). The paper should report cluster-robust or block-bootstrap uncertainties, or at least identify the effective number of independent cells and days. Without this, the comparisons across terrains and seasons (and the claim that blh 'appears as the top independent driver') are not statistically underpinned.
  3. [Sec. IV C, Eq. (12)] The origin-constrained regression H_AGL = β1·(T−T_d) is an appropriate form for an idealized LCL relation, but the paragliding observable has a positive detection threshold: climbing segments require a gain ≥10 m and mean climb rate ≥0.1 m/s, so H_AGL is bounded below by a positive value even when T−T_d is small. Forcing the intercept to zero can bias β1 upward. The paper should also report the unconstrained fit (with intercept) and discuss whether the nonzero intercept is physical or an artifact of the segment filters. This is particularly relevant for the Cold season, where the fitted slope drops to 108.8 m/°C and the authors attribute the discrepancy to under-reach.
minor comments (6)
  1. [Sec. II C and Sec. III C] There are unresolved reference placeholders: 'ref. [?]' for the terrain thresholds (Eq. 1) and 'Spencer's formula [?]' for solar time. These must be completed before publication.
  2. [Sec. IV B and Appendix B, Table IV] The text says the final predictor set has 118 variables, but Table IV is titled 'The 120 ERA5 predictors used in the orthogonal regression analysis.' The table also lists 'T Td' among the predictors, although T−T_d is the reference variable and should not be a candidate in the ORFR ranking. Please reconcile the count and the table contents.
  3. [Fig. 5] The figure caption says 'Each brown dot is one cell-hour observation,' but with n=29,023 points the scatter is heavily overplotted. Consider using a density-based scatter or hexbin plot, and add a marginal histogram or a confidence band on the OLS line.
  4. [Sec. IV C] The statement that 'Pearson linear correlations r(Y,T−Td) agree with ρS to within ±0.03 across all combinations reported below' is not shown anywhere in the manuscript. Either include a table or remove the claim.
  5. [Sec. V] The limitations paragraph is honest about free-flight conditioning, but it does not mention the order-statistic/sampling-intensity issue with H_AGL. Please add an explicit discussion of this limitation and how it might affect the LCL validation and the ORFR rankings.
  6. [Abstract] The abstract says 'striking precision' for the LCL match; given the order-statistic concern, this phrasing is too strong until the sampling-intensity bias is controlled.

Circularity Check

0 steps flagged

No significant circularity: the LCL benchmark and ORFR analysis are externally grounded; only a minor methodological self-citation is present.

full rationale

Walking the derivation chain: the observables (Eqs. 2-4) are defined from paragliding climb segments; the climb segments are extracted by a phase-detection algorithm cited to Ref. [12], an in-press paper by overlapping authors. That is the only self-citation, and it is a data-processing dependency rather than a result derived from the paper's own target conclusions. The central validation is the LCL benchmark (Sec. IV C): Eq. (12) estimates beta1 by origin-constrained OLS from the observed H_AGL and ERA5 T-Td, then compares to the independent parcel-theory slope of ~125 m/°C (Refs. 22-24). The theoretical value is not used in the fit, so the reported 123.6 m/°C is a genuine external benchmark, not a fitted input called a prediction. The ORFR analysis (Eqs. 14-16) is an empirical residual-correlation ranking of ERA5 predictors after removing linear covariance with T-Td; the claim that blh is the 'top independent driver' is a description of those correlations, not a quantity defined to include itself. No uniqueness theorem, no ansatz-via-citation, and no renaming of a known result is load-bearing. The order-statistic caveat on H_AGL (max over a varying number of segments) and the acknowledged cold-season under-reach are sampling-bias/correctness limitations, not definitional circularity. Score 2 reflects the minor self-citation only; the physical derivation is self-contained and externally checkable.

Axiom & Free-Parameter Ledger

5 free parameters · 6 axioms · 0 invented entities

The paper's quantitative claims rest mainly on hand-chosen thresholds and domain assumptions, not on fitted physical constants. The largest assumption is that a cell-hour max of pilot altitudes equals the thermal ceiling despite variable sampling intensity; the next is that ERA5 fields and the in-press phase-detection algorithm are accurate. No new particles, forces, dimensions, or conserved quantities are introduced; the three observables are statistical summaries, not new physical entities.

free parameters (5)
  • Terrain elevation thresholds = 300, 800, 1500 m AGL
    Hand-chosen boundaries in Eq. (1) (footnote [15] 'ref. [?]'); every terrain-split result and the ORFR rankings depend on these cutoffs.
  • Climbing-segment quality filter bounds = 20–1800 s; ≥10 m gain; ≥0.1 m/s; 50–5500 m; |Vz| ≤ 10 m/s
    Hand-set filters in Section II.A select the 1.47M segments that define all observables; small changes alter the dataset.
  • Cloud-cover threshold τ = 0.1
    Eq. (9) clear/cloudy split; standard operational definition [18], but a hand-set cutoff that affects conditional statistics.
  • SMI dryness threshold = 0.5
    Eq. (11) dry/wet split; standard plant-available-water cutoff, but hand-set.
  • Hour-block boundaries = 9, 11, 12, 14, 15, 17 solar hours
    Eq. (8) Morning/Midday/Afternoon bins; the reference regimes 'Warm Midday' and 'Cold Midday' are defined by these boundaries.
axioms (6)
  • domain assumption ERA5 reanalysis fields faithfully represent the atmospheric state at 0.25°/hourly resolution in the cells used
    All predictor variables and the LCL benchmark use ERA5; if reanalysis is biased (e.g., boundary-layer height), all rankings inherit the bias. Section II.B.
  • domain assumption Vertical velocity measured by GPS is the aircraft velocity; true updraft is recovered as Vz + V_sink with V_sink ≈ 1.0–1.2 m/s
    Acknowledged in Section III.A; sink rate is neither measured per flight nor corrected, so thermal-strength magnitudes and correlations are approximate.
  • domain assumption At least one pilot reaches the true thermal ceiling in each cell-hour, so max altitude is a valid ceiling proxy
    Eq. (4) defines H_AGL as the cell-hour maximum on this expectation; violated in weak thermals or low traffic, especially cold season.
  • domain assumption The phase-detection algorithm of Vilpellet et al. [12] correctly distinguishes climb/search/transition phases
    The entire segment extraction rests on this algorithm, which is in press and not specified in the paper.
  • domain assumption Cell-hour observations are sufficiently independent for OLS/Spearman inference
    Correlations and slopes treat cell-hours as independent; no blocking or autocorrelation analysis is provided.
  • domain assumption LCL scaling H_AGL = β1(T-T_d) with β1 ≈ 125 m/°C is the correct thermodynamic benchmark for wet thermals
    Standard parcel theory [22-24] imported as a benchmark; not derived here, and only applies to cloud-topped thermals.

reviewed 2026-08-04 · how reviews work

0 comments
Cite this review

Pith. "Pith review of From Flight Logs to Atmospheric Science: Paragliders as Convection Sensors for Identifying Thermal Predictors." pith.science (2026). https://pith.science/paper/XF62EBL6

@misc{pith2026260800241,
  author       = {Pith},
  title        = {Pith review of: From Flight Logs to Atmospheric Science: Paragliders as Convection Sensors for Identifying Thermal Predictors},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XF62EBL6}},
  note         = {Machine review of arXiv:2608.00241}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Atmospheric thermal convection drives boundary-layer dynamics and vertical exchange of heat, moisture, and momentum, yet fundamental questions about thermal structure remain open due to limited in situ observational coverage. We introduce a novel high-resolution observational dataset for atmospheric convection based on paragliding flight logs collected over metropolitan France during 2017-2024. Paragliders probe thermal updrafts by circling within them while carrying GPS variometers that record position and altitude; each climbing segment samples the vertical velocity field within a thermal column. Aggregated across 1.47 million climbing segments from 110,730 flights, this dataset provides unprecedented spatial coverage and temporal resolution. To demonstrate the value of this observational resource, we extract three physically distinct observables from climbing segments and characterize their dependence on terrain, season, time of day, cloud state, and soil moisture. Coupling the paragliding observations with global atmospheric reanalysis data (ERA5, 0.25 degrees hourly) through an orthogonalized regression framework against 118 physically interpretable predictors, we identify leading atmospheric predictors of these observables. The empirical relationship between ceiling height and temperature-dewpoint depression matches the theoretical lifting condensation level scaling with striking precision supporting our methodology. Boundary-layer height emerges as the leading independent predictor of both ceiling height and thermal strength across all terrains and seasons, while vertical-velocity variability is controlled by surface heat-flux and wind variables. Taken together, our results highlight the utility of crowdsourced flight-log data for investigating atmospheric convection.

Figures

Figures reproduced from arXiv: 2608.00241 by C\'esar Hern\'andez-Aguayo, Matthieu Cristelli, Michael Benzaquen.

Figure 1
Figure 1. Figure 1: FIG. 1. Schematic of a paragliding climbing segment during which [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. Spatial distribution of climbing activity over France. Active [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. Probability density functions of the three climbing observ [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4. Temporal evolution of the average climbing observables [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5. LCL benchmark scatter for the Low mountains terrain at [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: FIG. 6. ORFR bar plots at Warm midday (12–14h solar) for the four terrains, in reading order: High mountains (top-left, [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: FIG. 7. ORFR bar plots at Cold midday (12–14h solar) for the four terrains, in the same reading order: High mountains (top-left, [PITH_FULL_IMAGE:figures/full_fig_p011_7.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

27 extracted references

  1. [1]

    The two mountain- ous terrains bracket the textbook value of∼125m/ ○C, while the flatter terrains fall∼15–20%below it

    In the Warm season, the empirical slope forH AGL de- creases monotonically from133m/ ○C in High moun- tains through124m/○C in Low mountains and106m/○C in Hills down to101m/○C in Plains. The two mountain- ous terrains bracket the textbook value of∼125m/ ○C, while the flatter terrains fall∼15–20%below it

  2. [2]

    ForH AGL in the Warm season,ρ S lies in the range0.42–0.53, indi- cating a moderate but genuine link

    Within each (terrain, season) configuration, the correlation strength followsρ S(HAGL, T−Td)> ρS(E[ ¯Vz], T−Td)>ρ S(E[σVz], T−Td)without ex- ception across all eight configurations. ForH AGL in the Warm season,ρ S lies in the range0.42–0.53, indi- cating a moderate but genuine link. For the mean climb rateE[ ¯Vz],ρ S drops to0.11–0.23, and for turbulence ...

  3. [3]

    The empir- ical slopeβ 1 decreases in the mountains (133→117, 124→109m/ ○C) but drops much less in Plains (101→ 91m/○C)

    Moving from Warm to Cold at fixed terrain, ρS(HAGL, T−Td)drops from0.47to0.27in High mountains and from0.53to0.40in Low mountains, while it remains essentially flat in Hills (0.46→0.44) andincreasesfrom0.42to0.54in Plains. The empir- ical slopeβ 1 decreases in the mountains (133→117, 124→109m/ ○C) but drops much less in Plains (101→ 91m/○C). Interpreting ...

    2017

  4. [4]

    J. S. Turner, Buoyant plumes and thermals, Annual Review of Fluid Mechanics1, 29 (1969)

    1969

  5. [5]

    R. B. Stull,An Introduction to Boundary Layer Meteorology (Kluwer Academic Publishers, Dordrecht, 1988)

    1988

  6. [6]

    E. M. Sparrow, R. B. Husar, and R. J. Goldstein, Observations and other characteristics of thermals, Journal of Fluid Mechan- ics41, 793 (1970)

    1970

  7. [7]

    D. J. Kirshbaum and C.-C. Wang, Boundary layer updrafts driven by airflow over heated terrain, Journal of the Atmo- spheric Sciences71, 1425 (2014)

    2014

  8. [8]

    D. H. Lenschow and P. L. Stephens, The role of thermals in the convective boundary layer, Boundary-Layer Meteorology19, 509 (1980)

    1980

  9. [9]

    A. M. Blyth, S. G. Lasher-Trapp, and W. A. Cooper, A study of thermals in cumulus clouds, Quarterly Journal of the Royal Meteorological Society131, 1171 (2005)

    2005

  10. [10]

    Hernandez-Deckers and S

    D. Hernandez-Deckers and S. C. Sherwood, A numerical in- vestigation of cumulus thermals, Journal of the Atmospheric Sciences73, 4117 (2016)

    2016

  11. [11]

    J. M. Peters, W. Hannah, and H. Morrison, The influence of ver- tical wind shear on moist thermals, Journal of the Atmospheric Sciences76, 1645 (2019)

    2019

  12. [12]

    Pergaud, V

    J. Pergaud, V . Masson, S. Malardel, and F. Couvreux, A param- eterization of dry thermals and shallow cumuli for mesoscale numerical weather prediction, Boundary-Layer Meteorology 132, 83 (2009)

    2009

  13. [13]

    S. C. Sherwood, D. Hern ´andez-Deckers, M. Colin, and F. Robinson, Slippery thermals and the cumulus entrainment paradox, Journal of the Atmospheric Sciences70, 2426 (2013)

    2013

  14. [14]

    FFVL stands for F ´ed´eration Franc ¸aise de V ol Libre, and CFD for Comp´etition F´ed´erale de Distance

  15. [15]

    Vilpellet, A

    J. Vilpellet, A. Darmon, and M. Benzaquen, From random walks to thermal rides: Universal anomalous transport in soar- ing flights, J. R. Soc. Interface (in press) , arXiv:2601.01293 (2026), arXiv:2601.01293

    arXiv 2026

  16. [16]

    Above Mean Sea Level

  17. [17]

    Hersbach, B

    H. Hersbach, B. Bell, P. Berrisford, S. Hirahara, A. Hor ´anyi, J. Mu ˜noz-Sabater, J. Nicolas, C. Peubey, R. Radu, D. Schep- ers, A. Simmons, C. Soci, S. Abdalla, X. Abellan, G. Balsamo, P. Bechtold, G. Biavati, J. Bidlot, M. Bonavita, G. De Chiara, P. Dahlgren, D. Dee, M. Diamantakis, R. Dragani, J. Flem- ming, R. Forbes, M. Fuentes, A. Geer, L. Haimber...

    1999

  18. [18]

    The thresholds (300m, 800m, 1500m) were chosen to match the global mountain-area classification of ref. [?]

  19. [19]

    For a site nearλ=6 ○E in the French Alps, solar noon lies within±16min of11∶36UTC over the year, i.e

    Solar time is defined by the local sun position and computed from the UTC timestamp of each IGC record using Spencer’s formula [?], which combines the longitude correction4λ(in minutes) with the equation of time. For a site nearλ=6 ○E in the French Alps, solar noon lies within±16min of11∶36UTC over the year, i.e. approximately12∶36CET in winter or13∶36 CE...

  20. [20]

    We also considered the fourth central moment (kurtosis) of the vertical velocity within each climb as a potential additional ob- servable of turbulence structure. However, preliminary analy- sis showed that this quantity has very weak Spearman correla- tions with all ERA5 predictors (∣ρ S∣<0.05across all terrains) and does not provide statistically meanin...

  21. [21]

    A. K. Betts, Understanding hydrometeorology using global models, Bulletin of the American Meteorological Society85, 1673 (2004)

    2004

  22. [22]

    Hirschi, S

    M. Hirschi, S. I. Seneviratne, V . Alexandrov, F. Boberg, C. Boroneant, O. B. Christensen, H. Formayer, B. Orlowsky, and P. Stepanek, Observational evidence for soil-moisture im- pact on hot extremes in southeastern europe, Nature Geoscience 4, 17 (2011)

    2011

  23. [23]

    Balsamo, A

    G. Balsamo, A. Beljaars, K. Scipal, P. Viterbo, B. van den Hurk, M. Hirschi, and A. K. Betts, A revised hydrology for the ecmwf model: Verification from field site to terrestrial water storage and impact in the integrated forecast system, Journal of Hy- drometeorology10, 623 (2009)

    2009

  24. [24]

    Bolton, The computation of equivalent potential tempera- ture, Monthly Weather Review108, 1046 (1980)

    D. Bolton, The computation of equivalent potential tempera- ture, Monthly Weather Review108, 1046 (1980)

    1980

  25. [25]

    M. G. Lawrence, The relationship between relative humidity and the dewpoint temperature in moist air: A simple conver- sion and applications, Bulletin of the American Meteorological Society86, 225 (2005)

    2005

  26. [26]

    D. M. Romps, Exact expression for the lifting condensation level, Journal of the Atmospheric Sciences74, 3891 (2017). 13 Appendix A: More summary statistics of conditional splits TABLE II. Summary statistics of thermal observables by terrain, season, and hour block. Values are means over all cell-hour observations. Terrain Season Hour blockn ¯HAGL [m] ¯Vz...

    2017

  27. [27]

    The 120 ERA5 predictors used in the orthogonal regressionanalysis of Section IV D, grouped by physical family

    m s −1 uv100100 m wind speed (derived) m s −1 fg1010 m wind gust since previous PP m s −1 i10fgInstantaneous 10 m wind gust m s −1 Vertically integrated scalar quantities vimaVI mass of the atmosphere kg m −2 vimatVI mass tendency kg m −2 s−1 vikeVI kinetic energy J m −2 vitheVI thermal energy J m −2 vitoeVI total energy J m −2 vitVI temperature K kg m −2...

This paper was first reviewed by deepseek-v4-flash on August 4, 2026.