{"id":"74d11d0b-8c18-49a5-9bad-735fbcbf97ea","arxiv_id":"1909.00132","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Heterogeneous sand beds need much stronger wind to sustain saltation than well-sorted beds with equal median grain size, and they can show more than one distinct transport threshold.","lead":"Wind tunnel measurements show that the wind speed needed to keep sand saltating is 60 to 250 percent higher for poorly sorted, mixed-size sand beds than for well-sorted beds with the same median grain size. The result matters because dust emission and dune models typically assume a single median grain size sets the transport threshold.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 60–250% quantitative claim and the second-threshold inference hinge on fitting two straight regimes to a u*–U∞ relation that the paper concedes has a transitional region; a reanalysis of the archived raw data with a smooth-transition model is needed.","rationale":"The central claim has two parts: (i) dynamic thresholds are much larger for heterogeneous beds, and (ii) there is a distinct second threshold due to size selectivity. Part (i) is supported by two independent methods and the internal arithmetic checks out; visual ratios of 59% and 202% do not depend on the roughness method. However, the specific 60–250% range and its upper end come from the roughness method, which is the only method yielding 249%. That method is a fitting construction rather than a direct measurement, and the paper itself concedes the relevant caveats: Eq. 5b is extrapolated into the intermittent-transport regime, and an 'obvious transitional region' is acknowledged for Sample 5. The reported confidence intervals capture run-to-run scatter for a fixed fitting model, not the uncertainty caused by the choice of two straight lines. Since size-resolved sampling of the saltating grains is absent, interpreting uzo_t as the threshold of the whole ensemble rests on this two-line intersection; if a smooth-transition model fits the data as well or better, the second-threshold evidence for Sample 6 is weakened. Therefore the reader's conditional verdict remains appropriate: the qualitative finding that heterogeneity strongly affects the threshold is credible, but the quantitative upper bound and the two-threshold conclusion should be re-analyzed from the archived data before being treated as established.","tokens_in":15083,"tokens_out":6788,"duration_ms":70065,"concrete_test":"Use the publicly archived wind-tunnel data (Zenodo record 2550975) and, for each run of Samples 4 and 6, fit the u*–U∞ pairs with a single model that reduces to Eq. 5a below a free onset and to Eq. 5b above a free offset, with a smooth bridging function in between (or a piecewise linear fit with free breakpoints compared by AIC). Report the fitted transition u* and its uncertainty. If the best-fit transition for Sample 6 is below the reported 65.2 cm/s, or the Sample 4/6 transition ratio falls below about 2, then the 249% effect and the second-threshold inference are artifacts of the two-line extrapolation; if the transition remains near 65 cm/s and the ratio stays above 2, the headline upper bound survives.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The ceiling of the headline effect is the roughness-method ratio for Samples 4 and 6: uzo_t = 65.2 ± 2.3 cm/s versus 18.7 ± 3.3 cm/s, i.e., 249%, and this same method is the only evidence that Sample 6 has a second threshold. The method defines uzo_t as the intersection of Eq. 5a (u* = α1 U∞, constant zo) and Eq. 5b (u* = α2(U∞ − u_f), exponential zo). But Eq. 5b is a Bagnold-focus approximation valid for saturated transport; the paper explicitly says in Section 3.2.2 that using it to find the threshold 'neglects the transitional region that occurs near the dynamic threshold because transport is intermittent,' and in Section 4/Figure 4c it flags an 'obvious transitional region' for Sample 5. If the real u*–U∞ relationship bends smoothly from one branch to the other, the fitted intersection is not a physically defined threshold; its location depends on how many saturated-regime points are included and on the chosen line fit. This is not a minor calibration detail because the reported confidence intervals (±2.3, ±3.3 cm/s) are fit-based and do not include the model-form uncertainty. A smooth transition could move the Sample 6 intersection below 65.2 cm/s and shrink the Sample 4 versus Sample 6 ratio, weakening the 249% upper bound and the claim of a distinct ensemble threshold. The visual thresholds still show a 202% difference, so the qualitative 'heterogeneity matters' conclusion is more robust than the specific number in the abstract.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper presents wind tunnel measurements of the dynamic saltation threshold for four well-sorted and two poorly sorted sand beds. The threshold is determined by two methods: a visual method and a roughness-based method that locates the intersection of constant-roughness and Bagnold-focus branches in the u* versus U∞ plane (Eqs. 5a, 5b, 6). The authors report that both methods yield thresholds 60–250% larger for poorly sorted beds than for well-sorted beds with similar median diameter, that the visual and roughness thresholds differ for the two poorly sorted samples (which they interpret as evidence for more than one dynamic threshold), and that rescaling by the 90th percentile diameter d90 does not eliminate the difference. They also revisit the extrapolation method for transport-rate data and argue that threshold estimates depend strongly on the fitting procedure.","tokens_in":15392,"tokens_out":6728,"duration_ms":77104,"significance":"The central claim—that sand size heterogeneity can substantially raise the dynamic saltation threshold beyond what d50-based scaling predicts—is physically important and, if well supported, has consequences for dust-emission modeling and planetary geomorphology. The paper has notable strengths: the paired comparisons (Samples 1–5 and 4–6) show threshold differences well beyond the reported 95% confidence intervals; both measurement methods lean in the same direction; and the underlying velocity-profile data are publicly archived on Zenodo (record 2550975). The paper is also candid about the transitional-region limitation of the roughness method. However, the quantitative range (60–250%) and the 'more than one threshold' interpretation rest on assumptions (two-line intersection; fine-subset identification by visual inspection) that require additional support, so the present version needs revision before the claims can be accepted at face value.","major_comments":[{"comment":"The roughness-method threshold is defined as the intersection of the two fitted branches (Eq. 6), but the paper itself acknowledges in Section 3.2.2 that this neglects the transitional region near the dynamic threshold and flags an 'obvious transitional region' for Sample 5 in Figure 4c. Because the headline 60–250% range and the upper-bound 249% ratio for Samples 4 and 6 depend on the roughness threshold for Sample 6 (65.2 ± 2.3 cm/s), and because the reported confidence intervals include only line-fit uncertainty and not model-form uncertainty from the two-line approximation, the quantitative claim is not yet robust. I request a sensitivity analysis on the archived u*–U∞ data—for example, varying the range of points included in each branch, excluding points nearest the intersection, or fitting a smooth transition function—to show how uzo_t and the Sample 4-versus-6 ratio change. Without such an analysis, the specific 60–250% values in the abstract should be treated as provisional.","section":"Section 3.2.2, Eq. (6), Fig. 4c"},{"comment":"The conclusion that the visual threshold represents saltation of a fine-particle subset whereas the roughness threshold represents the whole ensemble is inferred from the visual observation that 'only relatively fine particles were saltating' in the intermediate range for Samples 5 and 6. No quantitative size-resolved data on the saltating grains are presented (e.g., particle tracking, image analysis, or size-selective traps), so the 'more than one dynamic threshold' interpretation is currently under-supported. Either provide direct evidence of size-selective transport or explicitly present the fine-subset interpretation as a hypothesis that is consistent with, but not uniquely determined by, the observations.","section":"Section 5.1"},{"comment":"The heterogeneity effect is anchored on only two poorly sorted samples (Samples 5 and 6), and the 95% confidence interval for the Sample 4 visual threshold (16.3 ± 6.0 cm/s) overlaps with the well-sorted samples. The qualitative conclusion that heterogeneity raises the threshold is supported by the two paired comparisons, but the paper should avoid presenting '60–250%' as a precise empirical bound until the model-form uncertainty in the roughness method is addressed and more heterogeneity levels are tested. The abstract and conclusions should be reworded to reflect this uncertainty.","section":"Section 4, Table 1"}],"minor_comments":[{"comment":"The color-coded symbols will not be distinguishable in greyscale printing; please add distinct symbol shapes or markers in addition to color.","section":"Figure 5"},{"comment":"The header on each page reads 'Confidential manuscript submitted to JGR-Earth Surface'; if this is posted as a preprint, the header should be updated or removed.","section":"Header"},{"comment":"The supporting information Figure S1 referenced in Section 2 is not included; please make it available with the submitted version or remove the reference.","section":"Section 2"},{"comment":"The statement that 'for all our tested sand beds, the visually estimated dynamic threshold is smaller than the one estimated from the roughness method' is based on point estimates; for Samples 1–4 the differences are not statistically significant, so the sentence should be qualified to emphasize that the significant difference occurs only for the poorly sorted samples.","section":"Section 5.1"},{"comment":"The notation M(Θ) and Mc is introduced; consider defining Mc explicitly on first use (it is currently defined parenthetically in the text) to improve readability.","section":"Equation (2)"},{"comment":"In the caption of Figure 5b, specify which measurements lack known d90 values and are therefore excluded from the comparison.","section":"Figure 5b caption"}],"recommendation":"major_revision","confidential_remarks":"The paper is closely connected to ongoing work by Pähtz and co-authors, and one of the authors is a proponent of the rebound hypothesis used in the interpretation; this is not improper, but the referee report focuses on the evidence. The recommendation of major revision is driven by the model-form uncertainty in the roughness method and by the unquantified size-selectivity claim; the qualitative heterogeneity effect appears defensible from the visual thresholds alone."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Worth reading, and worth sending out, but the abstract's 60-250% range should be handled with care. The paper's core message — that poorly sorted sand can have dynamic thresholds far above those of well-sorted sand with the same median diameter — is genuinely new and almost certainly right in direction. First controlled wind tunnel study to report full size distributions, and they've put the velocity profiles on Zenodo, which makes the work reproducible rather than just credible. I also found the Section 2 comparison of extrapolation procedures (least-squares vs. weighted, linear vs. nonlinear transport law) useful on its own: it shows threshold estimates from existing datasets can vary by a factor of 1.7, which is a healthy warning to anyone using flux-matching to infer thresholds. The authors are appropriately hedged and openly list limitations. Credit where due: the internal arithmetic is consistent, the two methods agree on the well-sorted beds, and the visual thresholds alone show a 202% difference for Samples 4 and 6, so the finding does not rest solely on the fancier method.\n\nThe soft spots are real but not fatal. The roughness-method threshold is the intersection of two straight-line fits, and the authors themselves concede that the true relation has a transitional region near the threshold (they flag an 'obvious transitional region' for Sample 5). The 95% intervals on u_zo are fit-based and don't include the model-form uncertainty, so the 249% upper end could indeed shift downward if a smooth-transition model were applied. The stress-test note is right about that; it is also right that the 202% visual difference survives, so the paper's principal claim is robust. The second, smaller issue is that only two poorly sorted distributions anchor the heterogeneity comparison, and the fine-participating-subset interpretation rests on visual observation rather than size-resolved sampling. Both caveats are acknowledged in the text, but they should be stated more prominently in a revision.\n\nWho is this for? Aeolian geomorphologists, dust-emission modellers, and anyone pushing planetary-threshold predictions from median diameter alone. It deserves a serious referee, not a desk reject. I'd send it out with explicit help: ask referees to re-fit the roughness-method thresholds with a smooth-transition model on the archived data and to assess how much the Sample 4 vs. Sample 6 ratio changes. If the 249% shrinks to, say, 150%, the paper still makes its point; if it collapses, the authors need a somewhat more modest headline. Either way, the qualitative heterogeneity effect is going to stand, and the field should know about it.","headline":"A real and important result about particle-size heterogeneity inflating the dynamic saltation threshold, but the headline 249% number is softer than it looks; the qualitative claim holds, the quantitative ceiling needs a re-analysis of the roughness method.","tokens_in":16002,"tokens_out":3121,"would_cite":true,"duration_ms":31886,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Sand beds with mixed grain sizes need 60 to 250 percent stronger wind to stop saltation than uniform sand of the same median size.","keywords":["dynamic saltation threshold","particle size heterogeneity","poorly sorted sand","roughness method","visual threshold","wind tunnel","aeolian saltation","dust aerosol emission"],"falsifier":"Collect and size the grains that are saltating in the wind-speed window between the visual and roughness thresholds for a poorly sorted bed (e.g., with a trap or high-speed imaging): the multi-threshold claim predicts they should be markedly finer than the bed's coarse tail; finding coarse grains in that cloud would refute the size-selective interpretation.","tokens_in":14758,"feed_emoji":"🌬️","tokens_out":8019,"duration_ms":72400,"temperature":0.7,"pith_summary":"This paper tries to establish that the wind speed at which saltation stops depends strongly on how mixed the sand bed is: for beds with similar median grain diameter $d_{50}$, poorly sorted sands have dynamic saltation thresholds 60–250 percent higher than well-sorted sands, measured both visually and by a surface-roughness regime-shift method. It also argues that sufficiently heterogeneous beds have more than one dynamic threshold: a lower, visually observed threshold at which only relatively fine grains saltate, and a higher roughness-based threshold at which the whole grain mixture participates. Even rescaling the threshold by the 90th percentile diameter $d_{90}$ instead of $d_{50}$ does not erase the difference, indicating that a bed's very coarse tail exerts a strong control. If correct, models of dust emission, dune dynamics, and planetary surface evolution that use a median-diameter threshold would predict saltation at winds where real heterogeneous beds have already stopped.","feed_headline":"Mixed-grain sand needs 60–250% stronger winds to stop saltation","feed_subtitle":"Real desert sands with mixed grain sizes quit saltating at winds far above what median-diameter models predict.","key_machinery":"The paper's central measuring device is the roughness-regime-shift method, which exploits that absent transport the bed surface roughness $z_o$ stays roughly constant, so the free-stream wind velocity and shear velocity obey $u_* = \\alpha_1 U_{\\infty}$, while saturated transport suppresses near-bed wind and raises roughness exponentially (the Bagnold-focus approximation), giving $u_* = \\alpha_2 (U_{\\infty} - u_f)$. The dynamic threshold is the intersection $u_t^{zo} = \\alpha_1 \\alpha_2 u_f /(\\alpha_2 - \\alpha_1)$. Complemented by the classic visual method, this yields the two thresholds compared across samples; the results are nondimensionalized as $A_{50} = u_t / \\sqrt{(\\rho_p/\\rho_a - 1) g d_{50}}$ and $A_{90}$ with $d_{90}$ to separate median-size scaling from the effect of the coarse tail.","core_discovery":"On its own terms, the paper claims that the dynamic saltation threshold (the minimal shear velocity that keeps saltation going once started) is not a property of median grain size alone. In wind-tunnel experiments comparing four well-sorted and two poorly sorted sand beds with overlapping $d_{50}$, both the visual method and the roughness method give thresholds for the poorly sorted beds that are 60–250 percent larger. The paper attributes part of this to hiding effects: fine grains sheltered by coarse neighbors are harder to keep rebounding, and coarse grains require stronger flow to stay airborne, so a heterogeneous bed resists sustained transport until the wind is much stronger. A second finding is that poorly sorted beds exhibit two distinct cessation thresholds—a lower visual threshold marking the stop of fine-particle saltation and a higher roughness threshold marking the stop of the full ensemble—and the paper interprets the gap as size-selective, erosion-limited, undersaturated transport sustained by aerodynamic entrainment rather than the splash feedback that saturates uniform beds.","pith_inferences":["The coarse-tail control suggests a testable generalization: for mixed beds, the threshold should scale with a high percentile such as $d_{84}$ or $d_{90}$ plus a sorting parameter; reanalyzing existing threshold data sets with sorting metadata would reveal whether a universal two-parameter curve collapses aeolian and fluvial thresholds.","The 'more than one threshold' picture implies hysteresis in heterogeneous soils: once full-ensemble saltation is running, transport can continue down to the lower visual threshold, so onset and cessation differ not only by turbulence but by grain-size selectivity; dust-emission schemes may need a memory or state variable for bed armoring.","If the visual threshold is indeed a fine-particle subset threshold, then ripple-crest armoring and the saltation threshold are coupled: surface sorting changes the threshold as transport proceeds, so threshold is not a fixed bed property but co-evolves with bed texture."],"forward_implications":["Threshold predictions for dust emission and dune mobility that use only $d_{50}$ will systematically underpredict the wind needed to stop transport on heterogeneous desert soils and will overpredict transport just above the median-based threshold.","A heterogeneous bed can have a wind window in which fine grains saltate while coarse grains stay put, so single-threshold transport laws cannot describe the cessation branch there.","Because $d_{90}$ does not collapse the data, the relevant characteristic size for thresholds lies above $d_{90}$ or is a fuller function of the size distribution, not a single percentile.","The extrapolation method for inferring thresholds from transport-rate measurements should use weighted least squares that preserve near-threshold data; unweighted fits can shift inferred thresholds by up to a factor of 1.7.","Field observations that find a single, size-independent threshold may simply reflect well-sorted beds; the size-selective behavior is expected to appear only when bed sorting is poor (roughly $d_{90}/d_{50} > 2$ in this study)."],"supporting_citations":[{"why":"Provides the classic visual-method dynamic threshold data set that this study extends and challenges.","marker":"[Bagnold, 1937]"},{"why":"Provides the other classic visual-threshold data set; both classic sets lacked particle size distributions.","marker":"[Chepil, 1945]"},{"why":"Supplies the wind-tunnel transport-rate measurements and linear transport law used in the extrapolation-method analysis.","marker":"[Creyssels et al., 2009]"},{"why":"Supplies the second wind-tunnel transport-rate dataset for the extrapolation comparison; its modified data (Ho, 2012) are used in Figure 1.","marker":"[Ho et al., 2011]"},{"why":"Provides the rebound-based cessation-threshold theory and the nonlinear transport law (Eq. 2) used to interpret size selectivity.","marker":"[Pähtz and Durán, 2018a]"},{"why":"Reports the field finding of size-independent participation in saltation against which the multiple-threshold claim is contrasted.","marker":"[Martin and Kok, 2019]"},{"why":"Shows that particle protrusion controls resisting forces on entrainment, supporting the hiding-effect explanation.","marker":"[Yager et al., 2018]"},{"why":"Reports the fluvial analogue of very coarse particles dominating bed mobility, cited as the water-transport parallel.","marker":"[MacKenzie and Eaton, 2017]"}],"fun_headline_variants":["Mixed sand needs up to 250% stronger winds to stop saltation","Grain size mix raises saltation stop threshold by up to 250%","Heterogeneous sand quits saltating at winds far above median-model","Poorly sorted sand: double threshold, up to 250% higher wind needed"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The claim that heterogeneous beds have distinct thresholds for fine grains and the whole ensemble stands on the visual identification that only relatively fine particles saltate between the two thresholds; no size-resolved sampling of the saltating grains was made to confirm that identification.","fun_headline_variants_meta":{"raw":{"variants":["Mixed sand needs up to 250% stronger winds to stop saltation","Grain size mix raises saltation stop threshold by up to 250%","Heterogeneous sand quits saltating at winds far above median-model","Poorly sorted sand: double threshold, up to 250% higher wind needed"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000414,"raw_usage":{"total_tokens":2142,"prompt_tokens":953,"completion_tokens":1189,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":569,"completion_tokens_details":{"reasoning_tokens":1107}},"tokens_in":569,"tokens_out":1189,"duration_ms":16069,"temperature":1.0,"reasoning_tokens":1107,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T06:01:41.448923+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Collect and size the grains that are saltating in the wind-speed window between the visual and roughness thresholds for a poorly sorted bed (e.g., with a trap or high-speed imaging): the multi-threshold claim predicts they should be markedly finer than the bed's coarse tail; finding coarse grains in that cloud would refute the size-selective interpretation.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the classic visual-method dynamic threshold data set that this study extends and challenges."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the other classic visual-threshold data set; both classic sets lacked particle size distributions."},{"cited_title":"Dupont, A","cited_arxiv_id":null,"evidence_quote":"Supplies the wind-tunnel transport-rate measurements and linear transport law used in the extrapolation-method analysis."},{"cited_title":"L., and J","cited_arxiv_id":null,"evidence_quote":"Reports the field finding of size-independent participation in saltation against which the multiple-threshold claim is contrasted."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Shows that particle protrusion controls resisting forces on entrainment, supporting the hiding-effect explanation."},{"cited_title":"G., and B","cited_arxiv_id":null,"evidence_quote":"Reports the fluvial analogue of very coarse particles dominating bed mobility, cited as the water-transport parallel."}],"review_version":1}