Wet weather

Daily flow with precipitation

The raw read behind everything below: daily plant flow with rainfall bars. Wet-weather inflow shows up as flow spikes following rain — the visible signature of inflow & infiltration (I&I) into the collection system, and a sizing input for any Phase II expansion. The charts further down quantify how strong that rain–flow coupling is and how it differs between the two plants.

Gaps in the lines mean no DMR row was filed for that plant on that date (not a zero). Precipitation bars here are the operator’s own DMR gauge readings (larger of the two plants’ filings when both report); the analytical charts below switch to NOAA observations — see the credibility check next for why.

Operator-reported precip vs NOAA — record-keeping credibility

Every analytical chart on this page that involves rainfall uses NOAA observations from station US1NYDT0024(RED HOOK 1.2 NNE, 1.4 mi from the Village). The operator's gauge readings on the DEC monthly forms are kept around for this credibility check only — they are not the basis of any of the wet-weather analyses below. This chart is the comparison.

Each point is one day where both sources reported. Perfect agreement would put every point on the dashed y = x line. Off-diagonal points (red, > 0.5″ disagreement) are days where one source recorded a major storm and the other missed it entirely — almost always because the operator did not read the gauge on that day. Days where only one source reported are tallied at the bottom; the operator-side shortfall is the larger one.

Stormwater memory (antecedent rain → flow)

For each candidate window, total rainfall over the last N days vs same-day plant flow, expressed as Pearson r. The highest-correlated window per plant is that plant's effective stormwater memory — how far back rainfall is still visibly affecting today's flow.

A peak at 1–2 days means rainfall mostly enters through open inflow points (cross-connected catch basins, leaky manhole covers) and clears quickly. A peak at 5–14 days means stormwater enters through subsurface infiltration and the soil-saturated pipes keep delivering elevated flow for a week-plus after the storm. Single-day cross-correlation is too noisy to use here because most days have zero rain — the cumulative-window approach is the standard I&I framing.

Rainfall-window response surface (by operator era)

Each cell shows the Pearson correlation between "cumulative rain over days [d − start, d − end]" and today's plant flow. The 1D antecedent-window chart above is just the bottom-left column (end = 0); this view also shows what happens when you exclude the most recent days from the window. Each plant gets two panels — before and after the April 1, 2025 contract-operator handoff — using identical color scales so the eye can compare directly where the "hot zone" of rainfall response sits in each era.

All-time pattern (using NOAA observations): Red Hook Commons' hot cells cluster along the bottom row at wide cumulative windows — peak r ≈ 0.16 at the full 14-day cumulative window, consistent with groundwater-driven infiltration. Village WWTP shows essentially no rainfall–flow correlation at any window (peak |r| < 0.05). If the hot zone moves between eras, that's evidence either the collection-system leakage profile changed, or the way flow is being measured / reported shifted across the boundary.

Caveat: peak |r| even on the all-time data tops out at ~0.16 (Commons). Rainfall explains a small share of daily flow variance — most variance is operational and below the resolution of monthly DMR data. A sub-daily flow study would be needed to characterise the response further. Per-era cells with fewer than n=30 paired observations are omitted (gray). The earlier version of this chart used operator-reported precip and showed a weak Village fast-inflow signal that turned out to be a record-keeping artifact — see the validation chart above for why we switched to NOAA.

Wet-weather sensitivity at each plant's preferred window (by operator era)

Daily flow versus cumulative rainfall over the window the heatmap above identified as each plant's strongest rainfall response — same window for both eras, so the comparison is apples-to-apples on the x dimension. Three regressions overlaid per plant: all-time (neutral), pre- April 2025 (gray dashed), and post-April 2025 (plant color, solid).

Reading the comparison: a materially flatter post slope on the same window means either the system's wet-weather sensitivity actually decreased (sewer rehabilitation, different storm frequency), or the way flow is being measured / reported changed across the operator handoff. A steeper post slope would suggest the opposite. The slope value (MGD/in) and R² for each fit appear in the legend.

Pump-out days vs typical days: was it rainier than usual?

For each QuickBooks "sludge removal" pump-out date, we compute cumulative rainfall over the days preceding it. We compare that to the same statistic for every non-pump-out day in the same date range. If pump-outs are functionally rainwater removal, pump-out days should sit in the wetter tail of the rainfall distribution; if not, the two distributions overlap.

Important caveat: these are QuickBooks invoicedates, not the dates the pump actually ran. Where a memo records the real event date it often differs by days to a month, and some invoices bundle multiple pumps. So this comparison is closer to “rainfall vs. billing date” than “rainfall vs. pump event,” and should be read as suggestive only. The capacity-censoring view (which needs no event dates) is the more reliable line of evidence.

The Mann–Whitney U test reported below the chart is a rank-based comparison of medians — robust to the small event sample and the skewed shape of the rainfall distribution. Two windows are shown to check whether the answer is window-sensitive: 7 days (matches the fast-inflow signal in the wet-weather sensitivity charts above) and 14 days (matches the slow-infiltration signal). Same-date pump-outs are aggregated. We can't reliably tell which plant is being pumped from the QuickBooks memos, so we treat all events together — village-level rainfall is the same input either way.

7-day window

14-day window

Caveat: events whose lookback window crossed a missing precip day are excluded. The biggest gap is November 2025, where the operator left the precip column blank on the source DMRs (verified across 4 independent extraction sources). A NOAA backfill is the obvious next data improvement.