{"id":"6e8ad93b-071b-45f9-9fa7-8ae0fbf9520f","arxiv_id":"1908.04579","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":8,"one_line_summary":"Across lab data and Common Land Model simulations, the Balland and Arp [2005] thermal conductivity scheme ranked first, while scheme choice shifted deep soil temperatures by up to 2.5 K in dry regions.","lead":"This paper compares seven formulas for soil thermal conductivity, first against laboratory measurements and then inside a land surface model, and finds that the Balland and Arp (2005) formula performs best. The choice of formula matters most for simulated soil temperature in dry, cold regions, which is where the paper says land models should switch from the current default.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The CoLM-based ranking that supports the 'superior scheme' recommendation may be an artifact of CoLM2014's soil hydraulic parameter estimation, a limitation the authors explicitly concede in Section 4.","rationale":"The paper is a well-structured and honest evaluation; it does not hide the main caveat, and the reader's conditional verdict is appropriate. I considered whether the small RMSE differences alone should be the primary objection, or whether the lack of released code is decisive; neither is as load-bearing as the hydraulic coupling. The lab comparison is informative but does not by itself establish superiority for land modelling, since the ranking by saturation level is split and the overall RMSE differences are small (0.28 vs 0.30 vs 0.32 W/m/K). The added value of the paper is the CoLM comparison, and that comparison is exactly where the confounding can enter: soil moisture is a state variable produced by a model with hydraulic parameters that the authors themselves single out in Section 4. Without a perturbation of those parameters, the conclusion that Balland and Arp will outperform Farouki in other land models or with other soil datasets is an extrapolation. The proposed test is feasible: CoLM2014 is a community model, and the five sites are described in Table 1, so the experiment can be reproduced. If the ranking is preserved, the recommendation would be substantially strengthened; if not, the paper should be revised to a conditional recommendation or a sensitivity statement. I therefore leave the reader's CONDITIONAL verdict unchanged.","tokens_in":16896,"tokens_out":6518,"duration_ms":69490,"concrete_test":"Repeat the five-site CoLM2014 evaluation with the seven thermal conductivity schemes unchanged but with soil hydraulic parameters drawn from an independent, observationally based source (e.g., site-calibrated Clapp-Hornberger parameters or measured retention curves at the Nagqu site), and recompute the RMSE rankings in Table 2. If Balland and Arp no longer yields the lowest ground heat flux and soil temperature RMSE at all or most sites, the claim that it is a superior scheme for land modelling use is not robust to the acknowledged hydraulic-parameter coupling.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim - that Balland and Arp [2005] can be chosen as a superior scheme for land modelling use - depends on the CoLM2014 evaluation, not only on the laboratory comparison. In CoLM, thermal conductivity is computed from simulated liquid and ice water content, and soil moisture is produced by the Richards equation using CoLM2014 pedotransfer functions. All seven schemes are therefore scored against the same, potentially biased, soil moisture states. Because the schemes have different saturation-dependent errors (Fig. 3: Cote and Konrad best at dry conditions, Balland and Arp best at S_r=0.5-1.0, Tarnawski and Leong best at low saturation), a systematic wet or dry bias in simulated soil moisture will systematically favor one scheme. The paper's Section 4 states that simulated performances 'can be significantly affected by the estimations of soil hydraulic parameters,' and no test is presented to show the ranking survives alternative hydraulic parameter estimates. The Table 2 margins are also small for surface temperature (0.1-0.3 K RMSE), so the most visible separation is in ground heat flux RMSE, a quantity directly tied to conductivity and sensitive to the same soil moisture state. The laboratory evaluation alone does not settle the issue because Balland and Arp is not best at all saturation levels. Thus the 'consistently best' claim is conditional on CoLM2014's hydrology being representative.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"Seven soil thermal conductivity schemes—Farouki 1981, Johansen 1975, Côté-Konrad 2005, Balland-Arp 2005, Lu et al. 2007, Tarnawski-Leong 2012, and De Vries 1963—are evaluated for land surface modeling. Each scheme is first modified to include soil organic matter and gravel effects, then assessed in two ways: direct comparison with laboratory thermal conductivity measurements from 61 soils across multiple saturation levels, and incorporation into the Common Land Model (CoLM2014) with simulations evaluated against observations at five sites spanning dry to wet climates (one Tibetan Plateau site with soil temperature profiles and four FLUXNET sites). The Balland-Arp scheme shows the lowest overall lab RMSE and the smallest CoLM RMSE for ground surface temperature and ground heat flux at most sites, leading the authors to recommend it as a superior scheme. The paper additionally quantifies the uncertainty introduced by thermal conductivity formulation, showing that in dry regions inter-scheme differences propagate to deep soil temperature and freeze-thaw timing, while wet regions are buffered by high heat capacity. A global comparison between the Balland-Arp scheme and the mean of the other six schemes shows large differences in soil temperature, heat fluxes, snow cover, and soil moisture over arid and semi-arid regions.","tokens_in":17264,"tokens_out":6825,"duration_ms":62694,"significance":"If the ranking holds, the paper provides actionable guidance for land surface modelers: replacing the default Farouki scheme with Balland-Arp would alter simulated ground heat flux and deep soil temperature in dry and cold regions, with implications for permafrost and freeze-thaw representation. The study is valuable in combining laboratory evaluation with land model evaluation and in explicitly quantifying the uncertainty in simulated temperatures from thermal conductivity formulation. The paper's strengths include a broad lab dataset, a consistent modification of all schemes, the use of observed site data, and the clear statement of the sensitivity of results to soil hydraulic parameters in Section 4. However, the strength of the recommendation is conditional on CoLM2014's hydrology and on the modification procedure, so the significance depends on the extent to which these caveats are addressed.","major_comments":[{"comment":"The recommendation of Balland and Arp [2005] as a 'superior scheme for land modelling use' is contingent on CoLM2014's soil hydraulic parameter estimation, as the authors themselves note in Section 4 ('the performances of soil thermal conductivity schemes can be significantly affected by the estimations of soil hydraulic parameters'). All seven thermal schemes are evaluated against the same simulated soil moisture and ice states; because the schemes have saturation-dependent errors (Figure 3), a systematic wet or dry bias in the simulated moisture could systematically favor one thermal scheme. No experiment is presented that varies the pedotransfer functions or hydraulic parameter scheme, so the transferability of the ranking to other LSMs or to CoLM with different hydraulic parameters is not established. Please either add sensitivity experiments with alternative hydraulic parameter estimation, or restrict the recommendation to CoLM2014.","section":"Section 4, Table 2, Section 3.1"},{"comment":"The margins in favor of Balland and Arp at the four FLUXNET sites are small for ground surface temperature (e.g., 3.40 vs 3.51 K at US-FPe; 2.60 vs 2.68 K at IT-Ro1; 2.18 vs 2.27 K at US-Bo1; 1.72 vs 1.79 K at AU-How), and the most visible separation is in ground heat flux RMSE. At three of these sites, soil properties are taken from the coarse global datasets GSDE and SoilGrids rather than from site measurements, and the paper does not quantify how uncertainty in these soil property estimates (porosity, organic matter, gravel fraction) propagates into the scheme ranking. If the soil property uncertainty produces changes in simulated thermal conductivity comparable to the inter-scheme differences, the ranking could change. Please add an uncertainty analysis of the soil input data, or discuss why the ranking is robust to it.","section":"Section 2.2.2, Table 2"},{"comment":"The laboratory evaluation shows that Balland and Arp is not uniformly best across all saturation levels: Côté and Konrad is best at S_r=0, Tarnawski and Leong is best at low saturations (S_r=0.1, 0.25), and Balland and Arp is best only at S_r=0.5–1.0. The claim in the abstract and Section 3.1 that Balland and Arp 'consistently perform best among all the schemes' overstates the evidence; the relative-error analysis supports a more nuanced conclusion that it performs well across the full saturation range and has the best overall RMSE. Please qualify the statement accordingly.","section":"Section 3.1, Figure 3"},{"comment":"The seven schemes are all modified into a volumetric weighted combination of mineral soils, SOM, and gravels, and the evaluation (both laboratory and CoLM) is performed on these modified versions. The paper does not validate that the modified formulations preserve the behavioral characteristics of the original published schemes, nor does it state whether the original fitted parameters were retained unchanged. Since the recommendation is stated in terms of the original scheme names, and since a reader may wish to implement the original Balland and Arp [2005] scheme, please clarify whether the conclusion applies to the modified versions only, and provide a validation or a sensitivity test of the modification procedure.","section":"Section 2.1"}],"minor_comments":[{"comment":"The site is referred to as 'Naqu' in the table header but 'Nagqu' elsewhere in the text; please standardize the spelling.","section":"Table 2"},{"comment":"The phrase 'global comparison between CoLM performances with the identified superior scheme and the mean of the other schemes' is ambiguous; it should specify that the 'mean' is the ensemble mean of the simulated variables from the other six schemes, not a mean of the schemes themselves.","section":"Section 2.2.3"},{"comment":"The caption lists panels (a), (b), (c), (e), and (f) as vertical profiles and then says 'with (d) precipitation variations given as well'; this is confusing. Please clarify that (d) is a separate precipitation panel.","section":"Figure 5 caption"},{"comment":"Equation (14), describing the Balland and Arp Kersten number, is garbled in the manuscript rendering and is not readable as written; please fix the typesetting so that the equation can be verified.","section":"Section 2.1.4, Equation (14)"},{"comment":"The sentence 'These patterns provide rationality to use these measurements as a reference' is awkward; consider 'These patterns provide a rationale for using these measurements as a reference.'","section":"Section 2.2.1"}],"recommendation":"major_revision","confidential_remarks":"The paper is within scope and addresses a useful applied question, but the central recommendation is stronger than the evidence supports given the acknowledged dependence on CoLM2014's hydraulic parameter estimation. A revision that either adds sensitivity experiments on hydraulic parameters or explicitly scopes the recommendation to CoLM2014, plus a tempering of the laboratory-based claims, would make the paper publishable. The global analysis comparing one scheme against the mean of six others is a notable weakness; comparing against the default Farouki scheme would strengthen the practical message."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First thing to know: this paper is worth a serious referee, and I think the Balland-Arp recommendation will hold up, but it is not as airtight as the abstract implies. The genuinely new bit is integrating a lab evaluation with a land surface model evaluation: seven published schemes, all modified to include SOM and gravel, tested against thermal conductivity measurements and then against observed soil temperature and heat flux in CoLM2014. That dual test is new, and it is a real improvement over prior lab-only comparisons. The lab comparison is clean: independent data, no parameters fitted here, and Balland-Arp comes out with the lowest RMSE overall. The CoLM comparison is consistent across five sites, with Balland-Arp best for ground heat flux and surface temperature almost everywhere.\n\nThe soft spots are where the stress-test concern lands. The CoLM ranking could be contingent on CoLM2014's soil hydraulic parameter estimation, because conductivity is computed from simulated soil moisture, and all seven schemes are scored against the same possibly biased moisture states. The authors acknowledge this in Section 4, but they do not test whether the ranking holds with alternative pedotransfer functions. Given the lab data alone show scheme performance varies by saturation level—Côté-Konrad is best dry, Tarnawski-Leong best at low saturation, Balland-Arp best from 0.5 to 1.0—a systematic wet or dry bias in simulated moisture could shift the ranking. This does not sink the paper, because the lab result independently favors Balland-Arp, but it weakens the 'superior scheme for land modelling use' headline. Also, three of the five FLUXNET sites use global soil datasets rather than site measurements, the surface-temperature margins are small (0.1–0.3 K RMSE), and the global experiment compares one scheme against the mean of six others, which is not a clean comparison. No code or data is released, so reproducing the exact implementations is harder than it should be.\n\nWho gets value: LSM developers choosing thermal conductivity schemes, and soil physicists interested in how lab rankings transfer to models. The central argument holds—Balland-Arp is a reasonable recommendation—but the paper would be stronger with a sensitivity test around soil hydraulic parameters and a more cautious claim in the abstract. I would send it to peer review, and I would ask for those sensitivity runs before accepting, but I would not desk reject it.","headline":"A solid, honest evaluation with a new lab-plus-model comparison; the Balland-Arp recommendation is plausible but the model-side evidence is conditional on CoLM2014's hydrology.","tokens_in":17756,"tokens_out":2641,"would_cite":true,"duration_ms":25534,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The Balland and Arp soil thermal conductivity scheme ranks first among seven formulas in both laboratory and land-model tests.","keywords":["soil thermal conductivity","land surface modeling","Common Land Model","Kersten number","Balland and Arp 2005","Farouki 1981","freeze-thaw cycles","soil temperature uncertainty"],"falsifier":"Run the same seven schemes in an independent land surface model at the Nagqu site with identical forcing and soil data; if the Farouki [1981] scheme matches observed 40-cm soil temperature as well as or better than Balland and Arp [2005], the transferable superiority claim is falsified. A cheaper test: prescribe observed soil moisture at the Nagqu profile; if Balland-Arp's edge disappears, the result is an artifact of simulated moisture rather than of thermal conductivity physics.","tokens_in":16718,"feed_emoji":"🌡️","tokens_out":6841,"duration_ms":63296,"temperature":0.7,"pith_summary":"This paper asks which of seven widely recommended formulas for soil thermal conductivity land surface models should actually use. It answers that the Balland and Arp [2005] scheme is the most accurate both against laboratory measurements across saturation levels and when embedded in the Common Land Model at five sites spanning dry to wet climates. A sympathetic reader would care because most land models currently default to the Farouki [1981] scheme, which the paper finds systematically overestimates conductivity at middle and high saturations. If the recommendation holds, replacing the default would change simulated deep soil temperature, freeze-thaw timing, ground heat flux, and snow cover, mainly in arid and semi-arid mid- and high-latitude regions, while wet regions remain largely insensitive.","feed_headline":"Balland-Arp soil heat formula beats six rivals in land models","feed_subtitle":"In dry, cold regions the choice shifts deep soil temperature by over a degree and alters freeze-thaw timing.","key_machinery":"The load-bearing object is the Johansen-type interpolation formula, in which bulk soil thermal conductivity is the dry-soil value plus a Kersten number $K_e$ times the gap between saturated and dry values, with $K_e$ a function of the degree of saturation and soil texture. Balland and Arp's variant replaces the piecewise $K_e$ relations of earlier schemes with a single continuous function that varies smoothly from dry to saturated and from fine- to coarse-textured soils, and the paper adds soil organic matter and gravel as volumetric components. Inside CoLM, $k$ enters the soil heat diffusion equation $C \\frac{\\partial T}{\\partial t} = \\frac{\\partial}{\\partial z} \\left(k \\frac{\\partial T}{\\partial z}\\right)$, so the scheme controls downward heat conduction and freeze-thaw energy release; that is the mechanism by which the ranking translates into model performance.","core_discovery":"On the paper's own terms, the discovery is a ranking with a winner: among the seven schemes, Balland and Arp [2005] gives the best thermal conductivity estimates (overall bias 0.01 W/m/K, RMSE 0.28 W/m/K) and, when incorporated into CoLM2014, produces the lowest RMSE for surface temperature and ground heat flux at all five sites, with the clearest gains at the dry Nagqu site (soil-temperature RMSE 1.93 K at 5 cm and 1.03 K at 40 cm, and roughly 3 W/$m^{2}$ lower ground-heat-flux RMSE). The authors therefore recommend Balland and Arp [2005] as a superior scheme for land modelling use. They also show that the scheme choice is not neutral: over dry regions inter-scheme conductivity differences of about 0.5 W/m/K propagate into deep-soil temperature differences up to 2.5 K, while over wet regions high soil heat capacity buffers the soil so that simulations barely change.","pith_inferences":["Beyond the paper, the ranking's portability is untested: because the paper notes soil hydraulic parameter estimates can significantly affect results, a different set of pedotransfer functions in another land model could reorder the seven schemes.","The deep-soil temperature spread implies that adopting Balland and Arp could alter simulated permafrost extent and soil carbon decomposition timing in Earth system models, since both depend on deep thermal regimes.","A sharper test would prescribe observed soil moisture profiles instead of simulating them, isolating the thermal-conductivity effect from the hydraulic feedbacks the paper's coupled simulations mix in.","Because the Balland and Arp parameters were fitted to unfrozen experimental data, its behavior in very dry frozen soils with ice-dominated conductivity is an open target for a dedicated validation."],"forward_implications":["The default Farouki [1981] scheme is not the best choice: at middle and high saturation it overestimates conductivity, so switching to Balland and Arp [2005] would improve simulated deep soil temperatures.","In dry, cold regions, inter-scheme conductivity differences of about 0.5 W/m/K produce deep-soil temperature differences up to 2.5 K at the Nagqu site and more than 1 K over northern frozen areas in global runs.","Wet regions such as AU-How show almost no soil temperature sensitivity, because high water content raises heat capacity and damps the response.","Global simulations with Balland and Arp versus the mean of the other schemes put about 2 W/m^2 more ground heat flux into soil, reducing snow cover and shallow soil moisture while increasing deep moisture via infiltration.","A global dataset of Balland-Arp parameters is needed before it can fully replace Farouki as the default scheme in global land models."],"supporting_citations":[{"why":"Supplies the winning scheme's continuity-improved Kersten-number relation that the paper recommends.","marker":"Balland and Arp [2005]"},{"why":"Provides the dry/saturated interpolation structure that all five derivative schemes build on.","marker":"Johansen [1975]"},{"why":"Is the default scheme in many land models, the main baseline the paper hopes to replace.","marker":"Farouki [1981]"},{"why":"Supplies a competing derivative scheme and the dry-soil and Kersten-number parameterization used in comparison.","marker":"Côté and Konrad [2005]"},{"why":"Supplies the series-parallel mechanistic scheme used as another comparator.","marker":"Tarnawski and Leong [2012]"},{"why":"Provides the Canadian and other field-soil laboratory measurements against which all schemes are scored.","marker":"Tarnawski et al. [2015; 2018]"},{"why":"Defines CoLM2014, the land model into which the seven schemes are plugged for simulation evaluation.","marker":"Dai et al. [2014]"},{"why":"Provides the Nagqu site observations of soil temperature and moisture profiles used for the dry-site evaluation.","marker":"Pan et al. [2017]"},{"why":"Supplies the GSDE global soil property data used for both site and global simulations.","marker":"Shangguan et al. [2014]"},{"why":"Supplies the SoilGrids data used for gravel fractions in the global and site simulations.","marker":"Hengl et al. [2017]"}],"fun_headline_variants":["Balland-Arp scheme tops 7 soil heat formulas for land models","Best soil heat formula for land models: Balland-Arp wins","In dry soils, heat scheme choice shifts deep temp by 2.5 K","Wet soils buffer heat formula differences; dry soils don't"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The conclusion assumes that the ranking seen in one model (CoLM2014) at five sites holds for land models generally, and that the model's co-estimated soil moisture does not bias the comparison; the authors explicitly caution that soil hydraulic parameter estimates can significantly affect simulated performance.","fun_headline_variants_meta":{"raw":{"variants":["Balland-Arp scheme tops 7 soil heat formulas for land models","Best soil heat formula for land models: Balland-Arp wins","In dry soils, heat scheme choice shifts deep temp by 2.5 K","Wet soils buffer heat formula differences; dry soils don't"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000456,"raw_usage":{"total_tokens":2329,"prompt_tokens":1023,"completion_tokens":1306,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":639,"completion_tokens_details":{"reasoning_tokens":1228}},"tokens_in":639,"tokens_out":1306,"duration_ms":9744,"temperature":1.0,"reasoning_tokens":1228,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:38:16.978237+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same seven schemes in an independent land surface model at the Nagqu site with identical forcing and soil data; if the Farouki [1981] scheme matches observed 40-cm soil temperature as well as or better than Balland and Arp [2005], the transferable superiority claim is falsified. A cheaper test: prescribe observed soil moisture at the Nagqu profile; if Balland-Arp's edge disappears, the result is an artifact of simulated moisture rather than of thermal conductivity physics.","supporting_citations":[],"review_version":1}