{"id":"d99159e8-9748-4932-83dd-ec448ea225d8","arxiv_id":"2505.05310","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"GNSS-derived precipitable water vapour measurements and model-based transmission estimates indicate that Gamsberg Mountain is drier than the H.E.S.S. site and is the preferred AMT site, with 345 GHz possible only in winter.","lead":"Two candidate sites for a new African millimetre radio telescope were compared using GNSS satellite measurements and weather reanalysis: the Gamsberg Mountain and the H.E.S.S. site. The paper reports that both are usable at 86 and 230 GHz and that Gamsberg is drier, with 345 GHz possible there in winter.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Transfer relation from a 6-week autumn overlap (Eq. 8) is extrapolated to winter; the winter 345 GHz claim depends on unverified seasonal stationarity of that relation.","rationale":"The reader identified the transfer relation in Section 3.2.2 as the weakest assumption, and I agree: it is load-bearing for nearly every quantitative Gamsberg result, especially the winter 345 GHz claim. The paper is otherwise competent: GNSS-derived PWV using NGL products is standard, the MERRA-2 validation at H.E.S.S. (92% correlation) is reasonable, and the am-based opacity/transmission conversion is well established. The authors also honestly acknowledge the short Gamsberg record and the need for a radiometer. The concern is not that the conclusion is impossible—Sarazin (1995) and altitude arguments make Gamsberg plausibly drier—but that the paper presents specific winter transmission numbers as if they were measured, when they are extrapolations from an untested seasonal invariance assumption. The proposed MERRA-2 test is cheap and uses data already in hand, so the reader's CONDITIONAL verdict is appropriate; no change to that verdict is needed from this stress-test pass.","tokens_in":14858,"tokens_out":2278,"duration_ms":24489,"concrete_test":"Use the paper's own 24-year MERRA-2 dataset (2000-2024, interpolated to both sites) to test seasonal stationarity of Eq. 8. Fit the H.E.S.S.-to-Gamsberg PWV relation separately for winter (JJA) and for the April-May calibration window, then compare slopes, intercepts, and predicted winter PWV distributions. If the winter residual between MERRA-2 Gamsberg PWV and Eq. 8 transformed H.E.S.S. PWV exceeds about 1 mm, or if the JJA slope differs from 0.83 by more than the calibration uncertainty, the 345 GHz winter claim is not supported and the conclusion should be downgraded to a conditional possibility. A direct winter GNSS campaign at GBGB would provide the definitive check.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that Gamsberg is the best AMT site and that 345 GHz is possible in winter rests on PWV_Gam = 0.83 * PWV_HESS - 2.61 mm (Eq. 8, Section 3.2.2). This relation is fitted to only six weeks of concurrent GNSS data, 2 April to 16 May 2024, under the explicit assumption that 'the two sites experience the same atmospheric conditions'. Nothing in the paper tests whether this linear mapping is seasonally invariant. The sites differ by 518 m in altitude and 30 km in horizontal separation; orographic lifting, valley fog, and seasonal moisture advection could easily change the slope or intercept during winter. All Gamsberg winter statistics (median PWV 2.62 mm, 345 GHz median transmission 62%, 77% at the 25th percentile) are synthesized by applying Eq. 8 to H.E.S.S. data from June-August, a period with no direct Gamsberg GNSS data. If the winter slope were 0.6 instead of 0.83, the winter median PWV would shift by roughly 1 mm and the 345 GHz transmission percentiles would move materially. The paper also gives no uncertainty for Eq. 8 and does not propagate any error through the opacity and transmission calculations. Sarazin (1995) independently shows Gamsberg is dry in winter, but that historical photometric-night sample does not quantitatively support the 2.62 mm median or the specific 345 GHz winter transmission claims. Thus the strongest quantitative statement, that 345 GHz is possible at Gamsberg during winter, is not directly supported by the data presented.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript compares GNSS-derived precipitable water vapour (PWV) at two candidate sites for the Africa Millimetre Telescope, the H.E.S.S. site and Gamsberg Mountain, and uses MERRA-2 reanalysis together with the am atmospheric model to derive opacities and atmospheric transmissions at 86, 230, and 345 GHz. The authors report that both sites are viable at 86 and 230 GHz, that Gamsberg has lower PWV overall (median 9.25 mm versus 14.27 mm), and that 345 GHz observations are possible at Gamsberg during winter, with a winter median PWV of 2.62 mm and a median 345 GHz transmission of 62%. The analysis rests on a short six-week overlap of GNSS data at the two sites (2 April to 16 May 2024), from which a linear transfer relation is derived and then applied to the full H.E.S.S. GNSS record to synthesize the Gamsberg seasonal statistics.","tokens_in":15204,"tokens_out":5370,"duration_ms":53222,"significance":"If the central seasonal results can be supported, the paper provides a valuable contribution to AMT site selection and to the broader EHT site-testing literature: it supplies a direct GNSS PWV record at H.E.S.S., demonstrates a high (92%) correlation between GNSS and MERRA-2 PWV, and applies a standard radiative-transfer tool to produce opacity and transmission estimates at three frequencies. The qualitative ranking of Gamsberg as drier is consistent with its 518 m altitude advantage and with Sarazin (1995), and the H.E.S.S. PWV analysis is largely sound. The main unresolved significance-risk is that the winter Gamsberg statistics, including the headline 345 GHz claim, are not directly measured but are synthesized through an extrapolated linear relation.","major_comments":[{"comment":"The entire multi-season Gamsberg PWV series, including the winter median of 2.62 mm and the winter 345 GHz transmission values in Section 3.2.3, Table 6, and Figure 9, is synthesized by applying the relation PWV_Gam = 0.83 * PWV_HESS - 2.61 mm to the H.E.S.S. GNSS record. This relation is fitted to only the six-week concurrent period between 2 April and 16 May 2024 and is applied under the explicit assumption that the two sites experience the same atmospheric conditions. The sites differ by 30 km in horizontal separation and 518 m in altitude, so seasonal changes in moisture advection, orographic lifting, or valley-level processes could alter the slope or intercept during winter. Because the abstract's central claim that 345 GHz is possible at Gamsberg during winter rests on this extrapolation, the paper should (i) test the seasonal stationarity of Eq. (8) using the 24-year MERRA-2 record at both sites, (ii) provide uncertainties on the fitted parameters and propagate them through the opacity and transmission calculations, and (iii) state explicitly that no direct Gamsberg GNSS data exist for June-August.","section":"§3.2.2, Eq. (8)"},{"comment":"The polynomial coefficients A, B, C in Eq. (6) are fitted to MERRA-2 PWV and opacity, but the transmission statistics in Tables 3, 4, 6, and 7 are computed by inserting GNSS PWV into these same fits. Section 3.1.1 reports a mean 7.45% difference (0.84 mm) between MERRA-2 and GNSS PWV at the H.E.S.S. site, yet this calibration offset is not propagated through Eq. (6). The resulting opacities and transmissions therefore carry an unquantified systematic uncertainty that is relevant to the 345 GHz conclusions. Please add an error-propagation or sensitivity analysis that shows how a PWV bias of the reported magnitude changes the transmission percentiles, especially at 345 GHz.","section":"§3.1.2 and §3.2.1, Tables 2 and 5"},{"comment":"The text states that during winter the Gamsberg 345 GHz median transmission is 62% and the 25th-percentile value is 77%, while Table 6 lists an overall 345 GHz median transmission of 18% and a 25th-percentile value of 40%. These numbers are not contradictory only if the winter statements are understood to be conditioned on the June-August weeks, but the paper never states how many winter weeks or seasons contribute to Figure 9b or Table 6. Please make the conditioning explicit and report the sample size behind each winter percentile; as written, the abstract's phrase '345 GHz possible at the Gamsberg Mountain during winter' can easily be misread as being supported by the overall statistics in Table 6.","section":"§3.2.3, Figure 9b and Table 6"}],"minor_comments":[{"comment":"The typesetting of Eq. (5) is difficult to parse: the factor 10^6 and the combination k'_2 + k_3 T_m^{-1} should be displayed with standard mathematical notation and clear parentheses.","section":"§2.1, Eq. (5)"},{"comment":"The text says the GNSS-MERRA-2 comparison covers 'over a year,' but the period from September 2022 to April 2024 is actually about 19 months; please state the exact span.","section":"§3.1.1"},{"comment":"The sentence 'only spans over 2 Months from 2 April 2024 and May 2024' should give the exact end date, 16 May 2024, as shown in Figure 7.","section":"§3.2.2"},{"comment":"The reported '92% correlation' should specify whether this is Pearson's r, Spearman's rank, or R^2, and should be accompanied by the scatter or residual statistics, since the transfer relation in Eq. (8) is also a correlation-based fit.","section":"§3.1.1, Figure 3a"},{"comment":"The paper should briefly reconcile its H.E.S.S. winter median PWV of 6.29 mm with the Backes et al. (2024) winter value of 3.04 mm cited in Section 1, rather than leaving the reader to infer all of the differences from the discussion of instrument biases.","section":"§1 and §4"},{"comment":"Please add a small table or text comparing the six-week concurrent GNSS statistics at both sites with the full synthesized year, so that the proportion of directly measured versus extrapolated data behind the final Gamsberg statistics is transparent.","section":"§3.2.2"}],"recommendation":"major_revision","confidential_remarks":"The H.E.S.S. GNSS analysis and the overall wet-dry ranking are credible, but the winter Gamsberg statistics and the 345 GHz claim are entirely dependent on the six-week transfer relation in Eq. (8). If the authors can demonstrate, e.g. with MERRA-2, that the relation is seasonally stable, the paper could become acceptable; otherwise the winter 345 GHz statements should be substantially weakened. No concerns about citation or novelty disclosure."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know this paper is worth reading if you care about the AMT site decision or about how GNSS PWV can be used for millimeter site testing in data-poor regions. It does something genuinely new: it installs and analyzes GNSS-derived PWV at both the H.E.S.S. site and Gamsberg Mountain, validates the H.E.S.S. record against MERRA-2 (92% correlation), and produces opacity and transmission statistics at 86, 230, and 345 GHz. The direct H.E.S.S. data are solid, the am-modeling is standard, and the authors are straight about the limitations of earlier instrument-based PWV estimates. The conclusion that Gamsberg is drier than H.E.S.S. is plausible and independently supported by Sarazin's 1994/1995 photometric-night data, which also showed winter dryness.\n\nThe soft spot is exactly where the stress-test lands. The Gamsberg GNSS station contributed only six weeks of direct data (2 April to 16 May 2024). To say anything about other seasons, the authors fit a linear relation PWV_Gam = 0.83 * PWV_HESS - 2.61 mm (Eq. 8) during that overlap and then apply it to the full H.E.S.S. record, assuming the two sites experience the same atmospheric conditions. That assumption is unverified, and the period of overlap (autumn) may not represent the winter relation. The 345 GHz winter transmission numbers (median 62%, 25th percentile 77%) are therefore not directly supported. The authors do say the Gamsberg record is short and that they rely on the assumption, but they do not give the transfer relation any uncertainty, nor do they propagate errors through the opacity and transmission calculations. A winter slope of 0.6 instead of 0.83 would shift the winter median PWV materially and change the 345 GHz story. This is not a fatal flaw — the qualitative ranking of Gamsberg as better, especially in winter, is well supported — but the quantitative winter claims should be toned down or flagged as model-dependent until direct winter GNSS or radiometer data exist.\n\nMinor points: the EHT-window analysis uses only two campaigns, which is a small sample, and the paper could compare against other EHT sites to contextualize the numbers. These are minor.\n\nI would send this to peer review. It deserves referee time because it is relevant to a real telescope decision and demonstrates a useful method. I would ask the authors to (1) state explicitly that winter 345 GHz claims are extrapolated, (2) provide uncertainty on Eq. 8 and propagate it, and (3) if possible, get any winter GNSS data from the Gamsberg station before publication. The paper, revised, would be a valuable contribution to site testing literature.","headline":"A useful, honest site assessment that correctly favors Gamsberg, but the headline 345 GHz winter claim rests on a six-week transfer relation extrapolated across all seasons.","tokens_in":15815,"tokens_out":1802,"would_cite":false,"duration_ms":21519,"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":"Gamsberg Mountain is the drier, more suitable site for the Africa Millimetre Telescope, and supports 345 GHz observing in winter.","keywords":["precipitable water vapour","GNSS","site testing","millimetre astronomy","submillimetre astronomy","atmospheric opacity","Event Horizon Telescope","Africa Millimetre Telescope"],"falsifier":"Put a PWV radiometer, or a sustained GNSS receiver, on Gamsberg Mountain through June, July, and August and compare the measured winter median PWV with the predicted 2.62 mm. If the measured median comes out instead near the H.E.S.S. winter value, or if the Gamsberg-to-H.E.S.S. offset changes seasonally, the 345 GHz winter transmission claim fails.","tokens_in":14620,"feed_emoji":"📡","tokens_out":10720,"duration_ms":88138,"temperature":0.7,"pith_summary":"This paper tries to establish which of two Namibian sites should host the Africa Millimetre Telescope by measuring the column of water vapour above each. It finds a median precipitable water vapour of 14.27 mm at the H.E.S.S. site and 9.25 mm at Gamsberg Mountain, and argues that the lower water burden at Gamsberg makes it the better site. If the estimates are right, the telescope would work well at 86 and 230 GHz from either site, and at 345 GHz from Gamsberg during the southern winter, when dry air lets more millimetre-wavelength light through.","feed_headline":"Gamsberg is the drier, better site for the Africa Millimetre Telescope","feed_subtitle":"Gamsberg's median water vapour is 9.25 mm vs 14.27 mm at H.E.S.S., making it the better site.","key_machinery":"The argument runs on a conversion chain from raw GNSS signal delays to PWV and then to opacity. The zenith total delay is split into hydrostatic and wet parts using on-site pressure and the Saastamoinen-Davis hydrostatic models; the wet delay becomes PWV through the Bevis water-vapour constants and the weighted mean temperature. Because the Gamsberg GNSS record covers only six weeks, the paper manufactures a multi-season Gamsberg series by applying the linear relation $\\mathrm{PWV}_{\\mathrm{Gam}} = 0.83\\,\\mathrm{PWV}_{\\mathrm{H.E.S.S.}} - 2.61\\ \\mathrm{mm}$, calibrated on the overlapping April-May 2024 data. PWV is then turned into opacity at 86, 230, and 345 GHz with quadratic fits built from MERRA-2 and the am atmospheric model, and transmission is $t(\\nu)=e^{-\\tau(\\nu)}$.","core_discovery":"The central claim is that the Gamsberg Mountain, which stands 518 m higher, has consistently lower precipitable water vapour than the H.E.S.S. site and is therefore the most suitable location for the AMT. The paper derives PWV from GNSS signal-delay measurements, checks them against MERRA-2 reanalysis data (92% correlation), and then fills in the short Gamsberg record by converting H.E.S.S. PWV with the relation $\\mathrm{PWV}_{\\mathrm{Gam}} = 0.83\\,\\mathrm{PWV}_{\\mathrm{H.E.S.S.}} - 2.61\\ \\mathrm{mm}$. On that basis it reports overall PWV medians of 14.27 mm at H.E.S.S. and 9.25 mm at Gamsberg, EHT-window medians of 16.62 mm and 11.20 mm, and a Gamsberg winter median of 2.62 mm, for which the 345 GHz transmission reaches a 62% median (77% at the 25th percentile). At the H.E.S.S. site, 345 GHz is effectively ruled out in the March-April EHT window, where even the best quartile passes only 10%.","pith_inferences":["Inference: because the Gamsberg calibration record runs from April to May only, the same-atmosphere assumption carries the entire winter and 345 GHz result; an independent winter measurement could overturn it.","Inference: the 518 m altitude advantage is doing most of the work in the PWV gap, so a short dedicated winter campaign at Gamsberg, even a few weeks of radiometer data, would directly test the most consequential claim.","Inference: the exponential sensitivity of 345 GHz transmission to PWV means that a small dry-end error in the linear conversion changes the winter 345 GHz verdict; a modest PWV error near the winter median would move the transmission result by tens of percent.","Inference: a natural next check is to use the same MERRA-2 reanalysis to predict seasonal Gamsberg PWV directly, since the paper uses MERRA-2 for opacity calibration but not as an independent seasonal Gamsberg PWV estimate."],"forward_implications":["Gamsberg would deliver median atmospheric transmission of 91% at 86 GHz and 62% at 230 GHz, versus 87% and 46% at the H.E.S.S. site.","During the current March-April EHT window, a Gamsberg AMT would receive median 55% transmission at 230 GHz (68% in the best quartile), compared with 40% (52%) at H.E.S.S.","345 GHz observing is viable only from Gamsberg and only in winter, with a median transmission of 62%; at H.E.S.S. the EHT-window best is 10%.","If the EHT extends beyond March-April, both sites could host 230 GHz EHT observations in the southern winter, with Gamsberg providing the most transparent sky."],"supporting_citations":[{"why":"It supplies the atmospheric constants $k'_2$ and $k_3$ used to convert zenith wet delay into PWV.","marker":"Bevis et al. 1994"},{"why":"It provides the hydrostatic-delay model used to separate the wet delay from the measured total delay.","marker":"Saastamoinen 1972"},{"why":"It provides the alternative hydrostatic-delay formula used in the same separation.","marker":"Davis et al. 1985"},{"why":"It describes the GNSS processing products that supply the zenith total delay and mean temperature for both stations.","marker":"Blewitt et al. 2018"},{"why":"It documents the GipsyX software used in producing those GNSS tropospheric products.","marker":"Bertiger et al. 2020"},{"why":"It shows that GNSS-derived PWV is reliable for site evaluation and quantifies its averaging uncertainty.","marker":"Sugiyama et al. 2024"},{"why":"It gives the MERRA-2-based EHT site study that included Gamsberg and provided opacity baselines for candidate sites.","marker":"Raymond et al. 2021"},{"why":"It provides the earlier in-situ Gamsberg PWV measurements from 1994-1995 used as a comparison baseline.","marker":"Sarazin 1995"},{"why":"It provides the am atmospheric model used to compute opacity and transmission from PWV and MERRA-2 profiles.","marker":"Paine 2022"},{"why":"It supplies the earlier multi-instrument PWV study at H.E.S.S. whose biases motivate the GNSS approach and comparison values.","marker":"Backes et al. 2024"}],"fun_headline_variants":["Gamsberg's lower water vapour makes it the top AMT site","For the Africa Millimetre Telescope, Gamsberg is the drier pick","Gamsberg beats H.E.S.S. for AMT with 9.25 mm vs 14.27 mm PWV","Drier Gamsberg wins out for Africa Millimetre Telescope site","Gamsberg's winter skies enable 345 GHz, H.E.S.S. can't"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the two sites, 30 km apart, experience the same atmospheric conditions all year, so that a Gamsberg PWV series can be reconstructed from H.E.S.S. measurements using a relation fitted to six weeks of overlap.","fun_headline_variants_meta":{"raw":{"variants":["Gamsberg's lower water vapour makes it the top AMT site","For the Africa Millimetre Telescope, Gamsberg is the drier pick","Gamsberg beats H.E.S.S. for AMT with 9.25 mm vs 14.27 mm PWV","Drier Gamsberg wins out for Africa Millimetre Telescope site","Gamsberg's winter skies enable 345 GHz, H.E.S.S. can't"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000308,"raw_usage":{"total_tokens":1864,"prompt_tokens":1155,"completion_tokens":709,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":771,"completion_tokens_details":{"reasoning_tokens":596}},"tokens_in":771,"tokens_out":709,"duration_ms":6734,"temperature":1.0,"reasoning_tokens":596,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T23:07:18.913468+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Put a PWV radiometer, or a sustained GNSS receiver, on Gamsberg Mountain through June, July, and August and compare the measured winter median PWV with the predicted 2.62 mm. If the measured median comes out instead near the H.E.S.S. winter value, or if the Gamsberg-to-H.E.S.S. offset changes seasonally, the 345 GHz winter transmission claim fails.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It provides the hydrostatic-delay model used to separate the wet delay from the measured total delay."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It provides the alternative hydrostatic-delay formula used in the same separation."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It describes the GNSS processing products that supply the zenith total delay and mean temperature for both stations."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It documents the GipsyX software used in producing those GNSS tropospheric products."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It provides the earlier in-situ Gamsberg PWV measurements from 1994-1995 used as a comparison baseline."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It provides the am atmospheric model used to compute opacity and transmission from PWV and MERRA-2 profiles."},{"cited_title":"F., Frans L","cited_arxiv_id":null,"evidence_quote":"It supplies the earlier multi-instrument PWV study at H.E.S.S. whose biases motivate the GNSS approach and comparison values."}],"review_version":1}