Claim at v1.2.0

further-tidal-flooding-trend

4931e77e06e959dce4c18c318620a551e56f8890586600a42a11236a6f7c0773

Data

{
  "has_outcome": true,
  "kind": "confirmation",
  "lineage": {
    "audit": "docs/lineage_audit.md; data/derived/lineage_audit.json",
    "class": "a",
    "class_meaning": "exposure or treatment only, not the outcome",
    "classes": [
      "a"
    ],
    "early_sources": [
      {
        "class": "a",
        "dataset": "8518750 The Battery",
        "evidence": "Battery water levels are the exposure. Referee: \"The Battery water levels and the rain data were read at 23:39 ... before registration, and the registration discloses this\". 311 counts for 2010-2019 fetched after.",
        "fetched": "2026-11-12T23:39:18-05:00"
      }
    ]
  },
  "outcome": "further-tidal-flooding-trend",
  "prereg_branch": {
    "name": "prereg/further-tidal-flooding-trend",
    "sha": "9ff568d395bba4664fd53c9696b3c4a5731ddcb8"
  },
  "prior": 0.35,
  "registered_at": "2026-11-12T23:44:06-0500",
  "registrant": "Claude (agent), for Jeff Vernon",
  "registration": {
    "alternatives": [
      "Reports follow rain that falls locally but not at the two airport gauges, and sewer surcharge after earlier rain, so a positive slope reflects weather and not tide.",
      "Perigean spring tides cluster in autumn and complaint rates vary by season; month terms address this only partly.",
      "The Battery is a poor proxy for water in Jamaica Bay and the Rockaways (wind set-up, a different range), so a flat slope reflects exposure misclassification.",
      "311 reporting behaviour differs by tract and changed between 2010 and 2019 (the 311 app and web channels grew), so a flat slope may reflect low reporting of tidal nuisance flooding and a year-to-year trend in volume enters through the year terms only.",
      "In 2010-2019 the tidal ZIPs had Hurricane Sandy repairs and Superstorm aftermath, and Sandy-driven changes in sewer use or reporting after 2012; the Sandy window is dropped but later effects are not.",
      "The annual tide-day count varies with the 18.6-year nodal cycle, so a slope over 16 years need not equal the sea-level trend times the density of the distribution."
    ],
    "edge": {
      "cause": "daily maximum water level at the Battery above 0.35 m over MHHW (dry days)",
      "effect": "311 street-flooding and catch-basin complaints per day in the coastal ZIP set, dry days only; and the annual count of such tide days"
    },
    "graph_evidence": [
      {
        "concepts": [
          "flooding",
          "flood",
          "sea level",
          "sea level rise"
        ],
        "direction": "none",
        "n_findings": 0,
        "note": "Concept search on 12 Nov 2026 (cached and new requests, see docs/round2.md). Concepts 'flooding' (AByRAdx2GF), 'flood' (LQWjUNDY2O), 'flood disaster' (NfqwQkrVlq), 'sea level' (4v2R8cNkVy), 'sea level rise' (rSNUatmq8T), 'inundation area' (7kxnfxWr92). 'sea level rise' has no relationships; 'sea level' has 11 neighbours (groundwater, sea surface temperature, atmospheric pressure, El Nino), none of them flooding, complaints or reporting. No graph finding links a water level, a tide or sea level to flood reports. The prediction below therefore rests on no graph finding."
      }
    ],
    "id": "further-tidal-flooding-trend",
    "kind": "confirmation",
    "literature": [
      {
        "citation": "NOAA CO-OPS, Sea Level Trends, The Battery 8518750, monthly means 1856 to 2025, single linear trend.",
        "data_access": "open",
        "data_used": "CO-OPS monthly mean sea level",
        "finding": "Trend 1.16 +/- 0.02 inches per decade, equal to 2.95 +/- 0.05 mm a year (read from the API on 12 Nov 2026; this is the long-term trend, not a 2010-2025 trend).",
        "url": "https://api.tidesandcurrents.noaa.gov/dpapi/prod/webapi/product/sealvltrends.json?station=8518750&affiliation=US&trendType=SINGLE"
      },
      {
        "citation": "Earlier registration further-tidal-flooding-311 and its outcome (no_test, 12 Nov 2026) in this repository.",
        "data_access": "open",
        "finding": "Rain-free days from 2020 to October 2026: 20 days at or above the NOS minor level; tidal-set rate ratio 15.7 (4.3 to 33.4) for street-flooding complaints, a continuous rate ratio of 1.19 per 10 cm (1.06 to 1.32), placebo shifts not near 1. These numbers were seen before this registration and are not a test of this claim."
      }
    ],
    "plan": "WHAT HAS BEEN SEEN. The first run (further-tidal-flooding-311) read 311 erm2-nwe9 for 1 Jan 2020 to 31 Oct 2026 for ALL ZIP codes (aggregated by day and ZIP; the tidal set, a FloodNet-based broad set and every other ZIP as placebo), with complaint sets C1 (street flooding) and C2 (C1 plus catch basin and sewer backup patterns), the descriptor counts for 2020 onward, daily Battery maxima 2020-2026 in MLLW, GHCN precipitation, and FloodNet events and sensors. Its results (listed in the literature field) were seen. No ZIP code is therefore unseen for 2020 onward, so the HELD-OUT unit is the PERIOD: 1 Jan 2010 to 31 Dec 2019 in the separate file 311 Service Requests from 2010 to 2019 (76ig-c548). The project's cache holds one query of that file (noise complaints, 'Noise - Street/Sidewalk' and 'Drinking', daily counts 2010-2019, from other claims); no flooding, sewer or catch-basin count of 2010-2019 has been queried, and the descriptor strings of that file have not been listed. The ZIP sets were fixed in the first registration from neighbourhood names and FloodNet sensor metadata, not from complaint counts, and are reused unchanged. Before registering I read the Battery daily maxima (hourly_height, datum MHHW, verified) and GHCN-Daily precipitation for 2010-2025, which are exposure data, and looked only at aggregate counts of dry days above water levels (by period 2010-2019 and 2020-2025, and for 2010-2025 in total, plus the number in a band of +/- 2.5 cm around 0.35 m). I did not look at counts by year, and I read no 311 flooding count for 2010-2019. For part 2 the 2020-2025 complaint counts are partly seen (daily C1 and C2 panel and yearly means from the first run; C3 below was not computed). The 2010-2025 tide-day counts by year were not seen.\nTHE KNOT. In the 2010-2019 dry sample (1,882 days after exclusions) the number of days with H at or above 0.563 m (the NOS minor flood level, 2.104 m above MLLW with MHHW 1.541 m above MLLW per CO-OPS datums, epoch 1983-2001) is 7, too few for the first claim's threshold. The knot is the highest 5 cm level at which the held-out dry sample holds at least 150 days: 0.35 m above MHHW (1.892 m above MLLW) gives 173 days (0.40 m gives 102, 0.45 m gives 55). The knot was chosen from the tide record alone and is not an estimate of where complaints begin; sensitivity checks move it.\nSAMPLE. Calendar days 1 Jan 2010 to 31 Dec 2019. DRY DAY: GHCN-Daily PRCP under 1.0 mm at both Central Park (USW00094728) and JFK (USW00094789) on the day and the previous day, all four values non-missing. Days from 29 Oct 2012 to 30 Nov 2012 are dropped (Hurricane Sandy and its aftermath, when complaints reflect storm damage); the model with them kept is a sensitivity check. Days with fewer than 20 hourly Battery values are dropped (none in 2010-2025).\nEXPOSURE. NOAA CO-OPS 8518750 The Battery, hourly_height, verified where available, datum MHHW, metric, local time (lst_ldt); calendar day is the local date; H is the daily maximum. x = max(0, (H - 0.35)/0.10).\nZIP SETS. TIDAL (primary): 11414; 11693, 11691, 11692, 11694, 11697; 11231; 11224. TIDAL-BROAD (secondary): TIDAL plus 10464, 11413, 11422, 10305, 10306, 10301, 10310, 10312, 10034. PLACEBO: every other ZIP with at least one 311 record in the sample, excluding TIDAL-BROAD.\nCOMPLAINTS. C3 (primary, 'street-flooding and catch-basin'): created_date in the sample and (upper(descriptor) like 'STREET FLOODING%' or upper(descriptor) like 'CATCH BASIN CLOGGED%' or upper(complaint_type) like '%STREET FLOOD%'). This matches, in the 2020 file, 'Street Flooding (SJ)', 'Catch Basin Clogged/Flooding (Use Comments) (SC)' and 'Catch Basin Clogged' and excludes sunken, grating, search, industrial-waste and sewer-backup descriptors. Patterns are fixed now; the 2010-2019 descriptor list is read only after this registration and any surprise is reported, not changed. C1 (secondary): street flooding only (descriptor like 'STREET FLOODING%' or complaint_type like '%STREET FLOOD%'). C4 (secondary): C3 plus descriptor like 'SEWER BACKUP%'.\nMINIMUM-DATA RULE. The test is run only if the sample has at least 150 dry days with H at or above 0.35 m (exposure count already 173) and at least 1,500 C3 complaints from the TIDAL set on dry days (a mean of about 0.8 a day; below that the standard error exceeds 0.11 and the refute condition cannot be reached). Otherwise the verdict is no_test and C4 is not substituted.\nCAN IT FAIL. Simulations on the actual 2010-2019 dry-day tide values (knot 0.35 m, NB2, dispersion 0.5 from the 2020-2026 C2 tidal series, year, month and weekday terms, 150 replicates per cell) give a model-based standard error of the log rate ratio per 10 cm of 0.11 at a baseline of 0.5 complaints a day, 0.084 at 1.0 and 0.07 at 2.0. Allowing 1.4 times that for serial correlation (block bootstrap), the standard error is 0.12 to 0.15 at a baseline of 0.5 to 1.0 and the minimum detectable effect at 80 percent power (2.8 standard errors) is a rate ratio of about 1.4 per 10 cm. If the true rate ratio is 1.00, the upper 95 percent limit is about 1.26 at a baseline of 1.0, so the refute condition (below 1.30) is reached in a little over half of such samples; if it is 1.2 (the first run's continuous figure), the lower limit is above 1.00 in about 45 percent of samples. The test can therefore fail in both directions but will often be unsettled; this is stated before the data are seen. Part 2a: with 329 dry days at or above 0.35 m in 16 years (mean 20.6 a year) the Poisson standard error of an OLS slope of annual counts on year is 0.25 days per year per year, against an implied slope of 0.55 (0.47 to 0.64 from the band count of 150 days within 2.5 cm of the knot, times 2.95 mm a year); a 95 percent t-interval (14 degrees of freedom) has half-width at least 0.53. Part 2b: the effect expected from part 1 at the registrant's centre (a rate ratio of 1.3 per 10 cm) is about 0.2 to 0.4 extra complaints per tide day on a baseline of about 190 dry-day complaints a year, a change in the annual total of 1 to 3 percent over the period, against year-to-year changes in 311 volume of tens of percent; at 16 years only a correlation of about 0.66 or more is detectable, so 2b is registered as descriptive and cannot support or refute anything.\nMODEL (PART 1, primary). Negative binomial (NB2, log link) of the daily TIDAL count on x with year, month-of-year and day-of-week fixed effects; no linear term in H below the knot (the registered shape is flat below the knot and log-linear above it). The same model on the PLACEBO series. A stacked model with a group indicator and group interactions with all terms gives the tidal-over-placebo ratio of rate ratios. Intervals: moving-block bootstrap over dates (blocks of 14 consecutive days, 1,000 replicates, TIDAL and PLACEBO resampled together, percentile 95 percent interval). The point estimate is exp(coefficient).\nPRE-SPECIFIED SENSITIVITY CHECKS (none decides): (1) knots at 0.30, 0.40 and 0.45 m; (2) log-linear in H over all dry days (per 10 cm), the shape of the first run; (3) binned dose-response, bands of H <0.30 (reference), 0.30-0.35, 0.35-0.40, 0.40-0.45 and 0.45 m or more, with a trend test across the bands; (4) C1 and C4 in place of C3; (5) TIDAL-BROAD in place of TIDAL; (6) Sandy window kept; (7) Poisson with the same terms and bootstrap; (8) leave-one-year-out; (9) the indicator for H at or above 0.563 m (NOS minor) reported with its day count; (10) dry defined as under 0.1 mm; (11) the same model on rainy days (PRCP at least 10 mm at either station) as a reference slope; (12) the 2020-2025 period, which is not held out, with the same model, reported only as a comparison of the held-out estimate with an in-sample estimate.\nPLACEBOS. (a) PLACEBO ZIP set. (b) Tide shifted: x assigned from the calendar day 7 days later, and separately 100 days later, within the same dry sample.\nPART 2a (TREND IN TIDE DAYS, exposure only). N_y = number of dry days in calendar year y with H at or above 0.35 m, y = 2010 to 2025 (dry days as defined above, all days of the year, Sandy window not dropped). OLS of N_y on y with a 95 percent t-interval on the slope (14 df). Checks: Poisson regression with log link and quasi-Poisson interval, Theil-Sen slope with a bootstrap over years, all days (not only dry) at or above 0.35 m, and the same count with H detrended by removing 2.95 mm a year (a null in which the trend has been taken out, to show what the slope would be without sea-level rise). The implied value 0.55 is fixed above; the registrant makes no other claim about the observed slope. Known alternatives: the 18.6-year nodal cycle and decadal wind and storm variability change annual counts by more than Poisson noise, and the local relative sea-level trend in 2010-2025 may differ from the 1856-2025 trend.\nPART 2b (DESCRIPTIVE). For 2010-2025 (2010-2019 from 76ig-c548, 2020-2025 from erm2-nwe9), the annual TIDAL-set C3 complaints on dry days divided by the annual PLACEBO-set C3 complaints on dry days (a volume-normalised series), reported against N_y with a Spearman correlation and a Poisson regression with offset. The seen portion (2020-2025 C1 and C2) is disclosed above. It decides nothing.\nOUTCOME. Estimates and intervals for part 1, all checks, the day and complaint counts, the part 2a slope and interval against 0.55, the 2b series, a figure in figures/round2/, a section in docs/round2.md. Verdict by concord.claims.PRECEDENCE with part 1 deciding.",
    "prediction": {
      "direction": "+",
      "magnitude": "PART 1 (decides the verdict). With x = max(0, (H - 0.35 m)/0.10 m) and H the daily maximum above MHHW, the registrant expects a tidal-set rate ratio per unit of x (per 10 cm above the knot) of 1.10 to 2.0 (centre about 1.3), and 0.95 to 1.10 in the placebo ZIP set; the ratio of the two between 1.05 and 2.0. The standard error of the log rate ratio is expected to be 0.09 to 0.15. PART 2a (scored separately). The slope of the annual count of dry days with H >= 0.35 m against year, 2010 to 2025, is expected at 0.55 days per year per year (Poisson interval of the implied value 0.47 to 0.64), that is a rise from about 16 days in 2010 to about 25 in 2025 around a 16-year mean of 20.6; the Poisson standard error of the fitted slope is 0.25. PART 2b (descriptive). Tidal-set dry-day complaints per year do not detectably follow the annual tide-day count; the predicted effect is under 1 percent of the annual dry-day total per tide day.",
      "refute_if": "PART 1 only. The upper 95 percent limit of the tidal-set rate ratio per 10 cm above the knot is below 1.30 (the data exclude a rise of 30 percent per 10 cm), or the upper limit of the tidal-over-placebo ratio is below 1.00; either holds only if the minimum-data rule is met, otherwise the verdict is no_test.",
      "scoring_notes": "The verdict is decided by part 1 alone, because part 2a has a standardised effect of about 2.2 (implied slope 0.55 over a standard error of 0.25) and cannot separate the implied rate from zero with confidence (power to exclude zero about 0.6 at Poisson variance and lower with overdispersion), and part 2b cannot be tested (see PLAN). Part 2a is scored in prediction_scored as hit (95 percent t-interval of the OLS slope excludes 0 and contains 0.55), partial (contains both 0 and 0.55), or miss (excludes 0.55). Placebos (tide shifted 7 and 100 days; placebo ZIP set) are reported as checks; if the shifted-tide placebo gives a tidal rate ratio with a 95 percent interval above 1.00 at both shifts the verdict is unsettled with that reason. All other results are secondary and never decide.",
      "support_if": "PART 1 only. In the primary model the 95 percent block-bootstrap interval of the tidal-set rate ratio per 10 cm above the knot has a lower limit above 1.00, and the interval of the tidal-over-placebo ratio of rate ratios has a lower limit above 1.00, in the held-out sample 2010-2019 (Sandy window excluded) and with the minimum-data rule met."
    },
    "prior": 0.35,
    "registered_at": "2026-11-12T23:44:06-0500",
    "registrant": "Claude (agent), for Jeff Vernon",
    "sha256": "517759937d19637def545614ad12b6602797658a915567e7ab6799e9575efe78",
    "statement": "In a held-out period (2010 to 2019), on dry days 311 street-flooding and catch-basin complaints from the low-lying coastal ZIP codes rise with the daily maximum water level at the Battery above a knot of 0.35 m above MHHW, and the rise is larger there than in other ZIP codes; and the annual count of dry days with a Battery maximum above that knot rose over 2010 to 2025 at about the rate that the Battery sea-level trend implies."
  },
  "registration_commit": {
    "commit": "5fd01ba768e88801a5f5ee5d2ae9d796b2f9c356",
    "committed_at": "2026-11-12T23:44:14-05:00"
  },
  "registration_file_sha256": "6ee6dea3a3753fda142131545b637e23ca9c3da45e4671b44d018e8ba9db30ca",
  "registration_path": "claims/registered/further-tidal-flooding-trend.json",
  "registration_sha256": "517759937d19637def545614ad12b6602797658a915567e7ab6799e9575efe78",
  "registration_text": "{\n \"id\": \"further-tidal-flooding-trend\",\n \"kind\": \"confirmation\",\n \"statement\": \"In a held-out period (2010 to 2019), on dry days 311 street-flooding and catch-basin complaints from the low-lying coastal ZIP codes rise with the daily maximum water level at the Battery above a knot of 0.35 m above MHHW, and the rise is larger there than in other ZIP codes; and the annual count of dry days with a Battery maximum above that knot rose over 2010 to 2025 at about the rate that the Battery sea-level trend implies.\",\n \"edge\": {\n  \"cause\": \"daily maximum water level at the Battery above 0.35 m over MHHW (dry days)\",\n  \"effect\": \"311 street-flooding and catch-basin complaints per day in the coastal ZIP set, dry days only; and the annual count of such tide days\"\n },\n \"graph_evidence\": [\n  {\n   \"note\": \"Concept search on 12 Nov 2026 (cached and new requests, see docs/round2.md). Concepts 'flooding' (AByRAdx2GF), 'flood' (LQWjUNDY2O), 'flood disaster' (NfqwQkrVlq), 'sea level' (4v2R8cNkVy), 'sea level rise' (rSNUatmq8T), 'inundation area' (7kxnfxWr92). 'sea level rise' has no relationships; 'sea level' has 11 neighbours (groundwater, sea surface temperature, atmospheric pressure, El Nino), none of them flooding, complaints or reporting. No graph finding links a water level, a tide or sea level to flood reports. The prediction below therefore rests on no graph finding.\",\n   \"concepts\": [\n    \"flooding\",\n    \"flood\",\n    \"sea level\",\n    \"sea level rise\"\n   ],\n   \"n_findings\": 0,\n   \"direction\": \"none\"\n  }\n ],\n \"literature\": [\n  {\n   \"citation\": \"NOAA CO-OPS, Sea Level Trends, The Battery 8518750, monthly means 1856 to 2025, single linear trend.\",\n   \"url\": \"https://api.tidesandcurrents.noaa.gov/dpapi/prod/webapi/product/sealvltrends.json?station=8518750&affiliation=US&trendType=SINGLE\",\n   \"finding\": \"Trend 1.16 +/- 0.02 inches per decade, equal to 2.95 +/- 0.05 mm a year (read from the API on 12 Nov 2026; this is the long-term trend, not a 2010-2025 trend).\",\n   \"data_used\": \"CO-OPS monthly mean sea level\",\n   \"data_access\": \"open\"\n  },\n  {\n   \"citation\": \"Earlier registration further-tidal-flooding-311 and its outcome (no_test, 12 Nov 2026) in this repository.\",\n   \"finding\": \"Rain-free days from 2020 to October 2026: 20 days at or above the NOS minor level; tidal-set rate ratio 15.7 (4.3 to 33.4) for street-flooding complaints, a continuous rate ratio of 1.19 per 10 cm (1.06 to 1.32), placebo shifts not near 1. These numbers were seen before this registration and are not a test of this claim.\",\n   \"data_access\": \"open\"\n  }\n ],\n \"prediction\": {\n  \"direction\": \"+\",\n  \"magnitude\": \"PART 1 (decides the verdict). With x = max(0, (H - 0.35 m)/0.10 m) and H the daily maximum above MHHW, the registrant expects a tidal-set rate ratio per unit of x (per 10 cm above the knot) of 1.10 to 2.0 (centre about 1.3), and 0.95 to 1.10 in the placebo ZIP set; the ratio of the two between 1.05 and 2.0. The standard error of the log rate ratio is expected to be 0.09 to 0.15. PART 2a (scored separately). The slope of the annual count of dry days with H >= 0.35 m against year, 2010 to 2025, is expected at 0.55 days per year per year (Poisson interval of the implied value 0.47 to 0.64), that is a rise from about 16 days in 2010 to about 25 in 2025 around a 16-year mean of 20.6; the Poisson standard error of the fitted slope is 0.25. PART 2b (descriptive). Tidal-set dry-day complaints per year do not detectably follow the annual tide-day count; the predicted effect is under 1 percent of the annual dry-day total per tide day.\",\n  \"support_if\": \"PART 1 only. In the primary model the 95 percent block-bootstrap interval of the tidal-set rate ratio per 10 cm above the knot has a lower limit above 1.00, and the interval of the tidal-over-placebo ratio of rate ratios has a lower limit above 1.00, in the held-out sample 2010-2019 (Sandy window excluded) and with the minimum-data rule met.\",\n  \"refute_if\": \"PART 1 only. The upper 95 percent limit of the tidal-set rate ratio per 10 cm above the knot is below 1.30 (the data exclude a rise of 30 percent per 10 cm), or the upper limit of the tidal-over-placebo ratio is below 1.00; either holds only if the minimum-data rule is met, otherwise the verdict is no_test.\",\n  \"scoring_notes\": \"The verdict is decided by part 1 alone, because part 2a has a standardised effect of about 2.2 (implied slope 0.55 over a standard error of 0.25) and cannot separate the implied rate from zero with confidence (power to exclude zero about 0.6 at Poisson variance and lower with overdispersion), and part 2b cannot be tested (see PLAN). Part 2a is scored in prediction_scored as hit (95 percent t-interval of the OLS slope excludes 0 and contains 0.55), partial (contains both 0 and 0.55), or miss (excludes 0.55). Placebos (tide shifted 7 and 100 days; placebo ZIP set) are reported as checks; if the shifted-tide placebo gives a tidal rate ratio with a 95 percent interval above 1.00 at both shifts the verdict is unsettled with that reason. All other results are secondary and never decide.\"\n },\n \"plan\": \"WHAT HAS BEEN SEEN. The first run (further-tidal-flooding-311) read 311 erm2-nwe9 for 1 Jan 2020 to 31 Oct 2026 for ALL ZIP codes (aggregated by day and ZIP; the tidal set, a FloodNet-based broad set and every other ZIP as placebo), with complaint sets C1 (street flooding) and C2 (C1 plus catch basin and sewer backup patterns), the descriptor counts for 2020 onward, daily Battery maxima 2020-2026 in MLLW, GHCN precipitation, and FloodNet events and sensors. Its results (listed in the literature field) were seen. No ZIP code is therefore unseen for 2020 onward, so the HELD-OUT unit is the PERIOD: 1 Jan 2010 to 31 Dec 2019 in the separate file 311 Service Requests from 2010 to 2019 (76ig-c548). The project's cache holds one query of that file (noise complaints, 'Noise - Street/Sidewalk' and 'Drinking', daily counts 2010-2019, from other claims); no flooding, sewer or catch-basin count of 2010-2019 has been queried, and the descriptor strings of that file have not been listed. The ZIP sets were fixed in the first registration from neighbourhood names and FloodNet sensor metadata, not from complaint counts, and are reused unchanged. Before registering I read the Battery daily maxima (hourly_height, datum MHHW, verified) and GHCN-Daily precipitation for 2010-2025, which are exposure data, and looked only at aggregate counts of dry days above water levels (by period 2010-2019 and 2020-2025, and for 2010-2025 in total, plus the number in a band of +/- 2.5 cm around 0.35 m). I did not look at counts by year, and I read no 311 flooding count for 2010-2019. For part 2 the 2020-2025 complaint counts are partly seen (daily C1 and C2 panel and yearly means from the first run; C3 below was not computed). The 2010-2025 tide-day counts by year were not seen.\\nTHE KNOT. In the 2010-2019 dry sample (1,882 days after exclusions) the number of days with H at or above 0.563 m (the NOS minor flood level, 2.104 m above MLLW with MHHW 1.541 m above MLLW per CO-OPS datums, epoch 1983-2001) is 7, too few for the first claim's threshold. The knot is the highest 5 cm level at which the held-out dry sample holds at least 150 days: 0.35 m above MHHW (1.892 m above MLLW) gives 173 days (0.40 m gives 102, 0.45 m gives 55). The knot was chosen from the tide record alone and is not an estimate of where complaints begin; sensitivity checks move it.\\nSAMPLE. Calendar days 1 Jan 2010 to 31 Dec 2019. DRY DAY: GHCN-Daily PRCP under 1.0 mm at both Central Park (USW00094728) and JFK (USW00094789) on the day and the previous day, all four values non-missing. Days from 29 Oct 2012 to 30 Nov 2012 are dropped (Hurricane Sandy and its aftermath, when complaints reflect storm damage); the model with them kept is a sensitivity check. Days with fewer than 20 hourly Battery values are dropped (none in 2010-2025).\\nEXPOSURE. NOAA CO-OPS 8518750 The Battery, hourly_height, verified where available, datum MHHW, metric, local time (lst_ldt); calendar day is the local date; H is the daily maximum. x = max(0, (H - 0.35)/0.10).\\nZIP SETS. TIDAL (primary): 11414; 11693, 11691, 11692, 11694, 11697; 11231; 11224. TIDAL-BROAD (secondary): TIDAL plus 10464, 11413, 11422, 10305, 10306, 10301, 10310, 10312, 10034. PLACEBO: every other ZIP with at least one 311 record in the sample, excluding TIDAL-BROAD.\\nCOMPLAINTS. C3 (primary, 'street-flooding and catch-basin'): created_date in the sample and (upper(descriptor) like 'STREET FLOODING%' or upper(descriptor) like 'CATCH BASIN CLOGGED%' or upper(complaint_type) like '%STREET FLOOD%'). This matches, in the 2020 file, 'Street Flooding (SJ)', 'Catch Basin Clogged/Flooding (Use Comments) (SC)' and 'Catch Basin Clogged' and excludes sunken, grating, search, industrial-waste and sewer-backup descriptors. Patterns are fixed now; the 2010-2019 descriptor list is read only after this registration and any surprise is reported, not changed. C1 (secondary): street flooding only (descriptor like 'STREET FLOODING%' or complaint_type like '%STREET FLOOD%'). C4 (secondary): C3 plus descriptor like 'SEWER BACKUP%'.\\nMINIMUM-DATA RULE. The test is run only if the sample has at least 150 dry days with H at or above 0.35 m (exposure count already 173) and at least 1,500 C3 complaints from the TIDAL set on dry days (a mean of about 0.8 a day; below that the standard error exceeds 0.11 and the refute condition cannot be reached). Otherwise the verdict is no_test and C4 is not substituted.\\nCAN IT FAIL. Simulations on the actual 2010-2019 dry-day tide values (knot 0.35 m, NB2, dispersion 0.5 from the 2020-2026 C2 tidal series, year, month and weekday terms, 150 replicates per cell) give a model-based standard error of the log rate ratio per 10 cm of 0.11 at a baseline of 0.5 complaints a day, 0.084 at 1.0 and 0.07 at 2.0. Allowing 1.4 times that for serial correlation (block bootstrap), the standard error is 0.12 to 0.15 at a baseline of 0.5 to 1.0 and the minimum detectable effect at 80 percent power (2.8 standard errors) is a rate ratio of about 1.4 per 10 cm. If the true rate ratio is 1.00, the upper 95 percent limit is about 1.26 at a baseline of 1.0, so the refute condition (below 1.30) is reached in a little over half of such samples; if it is 1.2 (the first run's continuous figure), the lower limit is above 1.00 in about 45 percent of samples. The test can therefore fail in both directions but will often be unsettled; this is stated before the data are seen. Part 2a: with 329 dry days at or above 0.35 m in 16 years (mean 20.6 a year) the Poisson standard error of an OLS slope of annual counts on year is 0.25 days per year per year, against an implied slope of 0.55 (0.47 to 0.64 from the band count of 150 days within 2.5 cm of the knot, times 2.95 mm a year); a 95 percent t-interval (14 degrees of freedom) has half-width at least 0.53. Part 2b: the effect expected from part 1 at the registrant's centre (a rate ratio of 1.3 per 10 cm) is about 0.2 to 0.4 extra complaints per tide day on a baseline of about 190 dry-day complaints a year, a change in the annual total of 1 to 3 percent over the period, against year-to-year changes in 311 volume of tens of percent; at 16 years only a correlation of about 0.66 or more is detectable, so 2b is registered as descriptive and cannot support or refute anything.\\nMODEL (PART 1, primary). Negative binomial (NB2, log link) of the daily TIDAL count on x with year, month-of-year and day-of-week fixed effects; no linear term in H below the knot (the registered shape is flat below the knot and log-linear above it). The same model on the PLACEBO series. A stacked model with a group indicator and group interactions with all terms gives the tidal-over-placebo ratio of rate ratios. Intervals: moving-block bootstrap over dates (blocks of 14 consecutive days, 1,000 replicates, TIDAL and PLACEBO resampled together, percentile 95 percent interval). The point estimate is exp(coefficient).\\nPRE-SPECIFIED SENSITIVITY CHECKS (none decides): (1) knots at 0.30, 0.40 and 0.45 m; (2) log-linear in H over all dry days (per 10 cm), the shape of the first run; (3) binned dose-response, bands of H <0.30 (reference), 0.30-0.35, 0.35-0.40, 0.40-0.45 and 0.45 m or more, with a trend test across the bands; (4) C1 and C4 in place of C3; (5) TIDAL-BROAD in place of TIDAL; (6) Sandy window kept; (7) Poisson with the same terms and bootstrap; (8) leave-one-year-out; (9) the indicator for H at or above 0.563 m (NOS minor) reported with its day count; (10) dry defined as under 0.1 mm; (11) the same model on rainy days (PRCP at least 10 mm at either station) as a reference slope; (12) the 2020-2025 period, which is not held out, with the same model, reported only as a comparison of the held-out estimate with an in-sample estimate.\\nPLACEBOS. (a) PLACEBO ZIP set. (b) Tide shifted: x assigned from the calendar day 7 days later, and separately 100 days later, within the same dry sample.\\nPART 2a (TREND IN TIDE DAYS, exposure only). N_y = number of dry days in calendar year y with H at or above 0.35 m, y = 2010 to 2025 (dry days as defined above, all days of the year, Sandy window not dropped). OLS of N_y on y with a 95 percent t-interval on the slope (14 df). Checks: Poisson regression with log link and quasi-Poisson interval, Theil-Sen slope with a bootstrap over years, all days (not only dry) at or above 0.35 m, and the same count with H detrended by removing 2.95 mm a year (a null in which the trend has been taken out, to show what the slope would be without sea-level rise). The implied value 0.55 is fixed above; the registrant makes no other claim about the observed slope. Known alternatives: the 18.6-year nodal cycle and decadal wind and storm variability change annual counts by more than Poisson noise, and the local relative sea-level trend in 2010-2025 may differ from the 1856-2025 trend.\\nPART 2b (DESCRIPTIVE). For 2010-2025 (2010-2019 from 76ig-c548, 2020-2025 from erm2-nwe9), the annual TIDAL-set C3 complaints on dry days divided by the annual PLACEBO-set C3 complaints on dry days (a volume-normalised series), reported against N_y with a Spearman correlation and a Poisson regression with offset. The seen portion (2020-2025 C1 and C2) is disclosed above. It decides nothing.\\nOUTCOME. Estimates and intervals for part 1, all checks, the day and complaint counts, the part 2a slope and interval against 0.55, the 2b series, a figure in figures/round2/, a section in docs/round2.md. Verdict by concord.claims.PRECEDENCE with part 1 deciding.\",\n \"alternatives\": [\n  \"Reports follow rain that falls locally but not at the two airport gauges, and sewer surcharge after earlier rain, so a positive slope reflects weather and not tide.\",\n  \"Perigean spring tides cluster in autumn and complaint rates vary by season; month terms address this only partly.\",\n  \"The Battery is a poor proxy for water in Jamaica Bay and the Rockaways (wind set-up, a different range), so a flat slope reflects exposure misclassification.\",\n  \"311 reporting behaviour differs by tract and changed between 2010 and 2019 (the 311 app and web channels grew), so a flat slope may reflect low reporting of tidal nuisance flooding and a year-to-year trend in volume enters through the year terms only.\",\n  \"In 2010-2019 the tidal ZIPs had Hurricane Sandy repairs and Superstorm aftermath, and Sandy-driven changes in sewer use or reporting after 2012; the Sandy window is dropped but later effects are not.\",\n  \"The annual tide-day count varies with the 18.6-year nodal cycle, so a slope over 16 years need not equal the sea-level trend times the density of the distribution.\"\n ],\n \"prior\": 0.35,\n \"registrant\": \"Claude (agent), for Jeff Vernon\",\n \"registered_at\": \"2026-11-12T23:44:06-0500\",\n \"sha256\": \"517759937d19637def545614ad12b6602797658a915567e7ab6799e9575efe78\"\n}",
  "replication": {
    "built_at": "2026-11-13T23:01:49-0500",
    "checked_at": "2026-11-13T23:01:49-0500",
    "decides": "support_if: tidal-set interval lower limit above 1.00 and tidal-over-placebo ratio interval lower limit above 1.00; shifted-tide placebos not above 1",
    "deciding_number": "city_effect: rate ratio of dry-day street-flooding and catch-basin complaints (C3) in the tidal ZIP set per 10 cm of Battery daily maximum above 0.35 m over MHHW, 2010-2019, negative binomial, 14-day moving-block bootstrap 95% interval (seed 20261113, 1,000 replicates); and the tidal-over-placebo-ZIP ratio of rate ratios",
    "inputs_copied": 4,
    "manifest": "replication/further-tidal-flooding-trend/manifest.json",
    "manifest_sha256": "6741e16edd9e35255b54ae7bd00935edb2df6ac43003afa097cb43cd8244d523",
    "package": "replication/further-tidal-flooding-trend/",
    "recorded": {
      "estimate": 2.000694591935365,
      "hi": 2.3843478152869113,
      "lo": 1.6436462504789693,
      "n": 1882,
      "placebo_zip_rr": 1.1530935101354036,
      "ratio_lo": 1.4348997719663479,
      "shift100_tidal_hi": 1.0450819569777863,
      "shift100_tidal_rr": 0.9713548847095617,
      "shift7_tidal_hi": 1.0704564224491058,
      "shift7_tidal_rr": 0.9926041049820505
    },
    "status": "MATCH",
    "status_from": "replication/further-tidal-flooding-trend/verify_result.json",
    "tables_not_rebuilt": 1,
    "tolerance": {
      "estimate": 0.0001,
      "hi": 0.0001,
      "lo": 0.0001,
      "n": 0,
      "placebo_zip_rr": 0.0001,
      "ratio_lo": 0.0001,
      "shift100_tidal_hi": 0.0001,
      "shift100_tidal_rr": 0.0001,
      "shift7_tidal_hi": 0.0001,
      "shift7_tidal_rr": 0.0001
    }
  },
  "statement": "In a held-out period (2010 to 2019), on dry days 311 street-flooding and catch-basin complaints from the low-lying coastal ZIP codes rise with the daily maximum water level at the Battery above a knot of 0.35 m above MHHW, and the rise is larger there than in other ZIP codes; and the annual count of dry days with a Battery maximum above that knot rose over 2010 to 2025 at about the rate that the Battery sea-level trend implies."
}

History in this collection

  1. addedv1.0.0Nov 14, 2026this version