📊 What the data say

Swim days: daily mean at the Camp Ellis upper mooring sensor ≥ 18.3 °C (65 °F); that sensor is below the surface, so the true count is likely a bit higher. Lobster stress: hours above 20 °C, a commonly used threshold for American lobster. Growing degree-days: sum of max(0, (Tmax + Tmin)/2 − 10 °C) since April 1. Low oxygen: share of hours below 2 mg/L (hypoxic) in the worst month.

Great Pond's oxygen sensor has never read above 3.0 mg/L; if that seems low for the pond, the sensor's placement is worth checking.

Yearly averages of NOAA's monthly mean sea level (years with at least 11 months), with a least-squares trend. Because one year's sea level is related to the next, the error bar is widened to account for that.

Assumes the land under the gauge is stable. Maine's land is close to stable, so this is close to true sea-level rise.

Daily highest water from NOAA's verified hourly heights at Portland (station 8418150), 1912 to today. A day counts if any hourly reading reached 12.0 ft (3.66 m) above MLLW: the National Weather Service minor flood stage for this gauge, at which wharves and the most vulnerable side streets near Portland Pier flood 4 to 6 inches deep. Years with fewer than 330 days of data are left out. Hourly readings can miss the very top of a tide by a few centimetres, so counts are slightly conservative.

Each year's highest monthly reading, shifted to today's sea level using the trend above. A generalized extreme value (GEV) distribution is fitted by maximum likelihood; the likely range comes from 300 bootstrap resamples.

"1-in-10-year" means a 10% chance in any year, not once every 10 years like clockwork. Assumes years are independent and storms aren't changing apart from sea level. Portland is about 20 km from Saco Bay, so local water can differ by a few centimetres.

Record values from NOAA via the National Weather Service (14.17 ft MLLW in Feb 1978; 14.57 ft MLLW on Jan 13, 2024). Sea-level rise between them uses the long-term trend (1.97 mm/yr). Year-to-year swings in average sea level (a few centimetres) mean the split is approximate.

Hourly measured water level minus NOAA's hourly prediction, with the year's average offset removed, then a 25-hour running average. Storm periods are those on the Camp Ellis wave buoy page.

Daily means of NOAA buoy 44007 water temperature (about 1 m deep). Normal: 1991–2020 values within ±5 days of each date, smoothed over 31 days; band = 10th to 90th percentile. Marine-heatwave test follows Hobday et al. (2016). Trend: least squares on yearly averages for the 28 years with at least 330 days of data (p = < 0.001).

Camp Ellis is shown for comparison: its water is shallow and close to shore, so it heats up more in summer (about 0.5–1.3 °C warmer than the NOAA buoy) and cools more in winter (about 1–1.5 °C colder).

Daily mean = (max + min) / 2 from GHCN-Daily station USW00014764. Normal = the 1991–2020 average within ±7 days of each date. Chart shows 7-day averages.

Daily means of the three temperature sensors on the Camp Ellis smart mooring (positions 1, 2 and 3, shallowest to deepest). Their exact depths aren't recorded in the data. Temperature sets most of the density here, but salinity also matters near the river plume.

Compiled from: UMaine Sea Grant: A brief history of Camp Ellis; Maine Monitor: Unstoppable ocean (Saco); Kelley et al., GSA Northeastern Section 2006.

Wave energy by direction: sum of H_s²·T_p (proportional to wave energy flux) in each 22.5° sector of the buoy's mean wave direction (where waves come from; energy-weighted mean 81°). The beach from the jetty to Ferry Beach runs north-northwest, so a wave hitting it straight on comes from about 65° (estimated from map coordinates; a surveyed shoreline would sharpen it).

The answer depends on that angle: with these waves the drift flips direction if the beach faced 82° or more (see the next card). Farther north, where the shore turns to face east-southeast at Old Orchard, the same waves would push sand the other way, so this estimate applies to the Camp Ellis – Ferry Beach stretch only.

Each hour, waves are refracted and shoaled from the buoy (12 m) to breaking (H = 0.78·depth) with Snell's law over straight, parallel depth contours, then

Assumptions: beach slope 0.02, grain size 0.2 mm, porosity 0.4. Wood Island, Ram Island and the jetty bend waves in ways straight contours miss, and the shoreline curves through the bay, so a wave model (such as SWAN) run along a surveyed shoreline would sharpen both the amounts and where the drift changes direction. Storm periods are those on the Camp Ellis wave buoy page.

Bed orbital velocity U = πH_s/(T·sinh kh) from linear wave theory (depth 12.7 m including mean tide); skin-friction factor f_w = 1.39(A/z₀)^−0.52 with z₀ = 2.5·D50/30 (Soulsby 1997); θ = ½f_wU² / [(s−1)g·D50]; threshold θ_cr from Soulsby & Whitehouse. D50 = 0.2 mm (fine sand).

Waves only: adding the measured bottom current and local grain-size samples would sharpen it.

Hourly speed and direction (rotated from magnetic to true north, declination −14.6°) are converted to east/north components and summed: x(t) = Σ u·Δt, y(t) = Σ v·Δt. This assumes the current at the buoy represents a wider area, a standard caveat for these diagrams. Direction is the direction the current flows toward, as labelled in the NERACOOS data catalogue for this buoy (CF standard name direction_of_sea_water_velocity).

USGS daily mean discharge, site 01066000, 1916 to one year ago as the baseline. For each date, the median and percentiles of all flows within ±3 days. Recent values are provisional.

Descriptive: daily totals and means over the last 45 days. River flow and tidal range overlap in time, so their separate effects are estimated in the salinity-model card.

Hourly mean salinity, last 31 days, no smoothing. Purely descriptive.

Daily averages of the 13-hour maximum-minus-minimum of salinity and of water level. Neighbouring days overlap slightly, so the correlation is a little optimistic, but the relationship is strong and expected from physics: ΔS ≈ (∂S/∂x) × tidal excursion, and the excursion grows with tidal range.

Fitted by non-linear least squares to tide-filtered salinity (Godin 24-24-25 filter, sampled every 6 hours, up to the last 120 days). Q = USGS Cornish daily flow; TR = 25-hour tidal range at Camp Ellis.

The model's errors are strongly related from one step to the next (lag-1 correlation 0.98), so the error bars are widened by √((1 + r)/(1 − r)) ≈ 10.4. River flow and tidal range are correlated over this period (r = -0.37), so their effects are only partly separable.

• Tidal excursion L = 2U/ω (how far water travels on one flood), ω = M2 frequency
• Salinity gradient along the estuary ∂S/∂x ≈ ΔS_tide / L
• Steady salt balance: K_x = Q·S / (A·∂S/∂x), A = width × depth
• Estuarine Richardson number Ri_E = (g·Δρ/ρ)·Q / (W·U³), Δρ/ρ ≈ 7.7×10⁻⁴ × ΔS. Below 0.08 well mixed; 0.08–0.8 partially mixed; above 0.8 layered.

Default river flow is the Cornish gauge (last 30 days) × 1.3, a rough allowance for the watershed below Cornish. Vertical mixing can't be measured with one sensor depth.

Daily means over the 32 most recent days when all four buoys reported (Sep 13 – Oct 18, 2026). For each pair, the daily difference is tested with a t-test whose sample size is cut to the effective number of independent days, n_eff = n(1 − r₁)/(1 + r₁), where r₁ is the day-to-day correlation of the difference. Because six pairs are tested for each measurement, p-values are adjusted for multiple comparisons (Holm method).

For reference, the classic tests on salinity, which treat days as independent: one-way ANOVA F = 255, Kruskal–Wallis H = 108, Friedman χ² = 88; effect size η² = 0.86. Their p-values are too small for the reason above, so the table uses the adjusted tests.

Nightly medians (10 pm – 4 am), last 60 days. NOAA ASSETS / National Estuarine Eutrophication Assessment levels: low ≤ 5, medium 5–20, high 20–60, very high > 60 µg/L. NOAA's ratings use the 90th percentile over a whole season, so this is an early look. Chlorophyll–salinity correlation (night) r = 0.45.

Median and middle half (interquartile range) of hourly chlorophyll by hour of day, Sep 12 – Oct 21. Night = 10 pm – 4 am; midday = 10 am – 2 pm. Values use the sonde's factory calibration, so comparisons within one sensor are more reliable than between sensors.

Least-squares fit of seven tidal constituents (M2, S2, N2, K1, O1, M4, M6) to the last 30 days of hourly water level. Thirty days separates M2 from N2, but not K1 from P1 (that needs 6 months), so K1 includes some P1.

Rise and fall times come from 5-minute data, between successive highs and lows. The gauge sits on the creek bed and can't see the very lowest water, which slightly distorts the shallow-water beats (M4, M6).

The heat equation ∂T/∂t = κ ∂²T/∂z², driven by a daily sine wave at the surface, predicts amplitude A(z) = A₀e^(−z/d) and delay Δt = z/(ωd), with damping depth d = √(2κ/ω). Daily cycles were fitted at each depth for June 1 – September 15, 2026, after removing the 24-hour running mean.

Assumes uniform soil and steady moisture, with no rain carrying heat downward. The deepest sensors' swings are below the 0.1 °C reporting step, so 40 and 45 cm are shown but not used in the fit.

All hourly Ram Island readings since Dec 2025, in 16 direction sectors and 4 speed classes, as a share of all hours.

972 scans matched to hourly wind. Correlations are descriptive; no p-values are given because hourly values aren't independent. The relationship is curved (a threshold), so the rank correlation (0.69) is also reported.