What the beach data can tell us

One useful baseline, no demonstrated model improvement

We tested predictions for 36 later City-reported widths at 12 surveyed transects. The average of earlier widths from the same season missed by 10.01 ft on average. Adding a single, pre-specified wave adjustment missed by 11.98 ft. Lower is better. These are checks against later observations, not estimates of how much sand stayed on the beach.

This work began at the request of a Capo Beach resident. It is an independent exploratory analysis, not a City or USACE finding, engineering design, or proof of a nourishment effect.

What we found and what is still missing

The City appendix also lists selected 2014–18 widths. Only 39 of 84 site/date cells contain numbers, and the September 2014, May 2016, and August 2018 entries are LiDAR-derived rather than the same profile-survey method used in the 2022–25 test. They were not pooled into the model setup period.

Coverage, format, access checks, and licensing cautions are in the source index.

Reported placement timeline

Volumes are reported at different times and cannot be added into a measured total. The county later listed approximately 196,000 cy for 2024 federal central beaches and 40,000 cy for North Beach; these likely overlap the contemporaneous federal and City numbers, not extra placements. The phase split, as-built dates and quantities remain unverified. See the placement caveats above.

Wave data: a usable height record, with a seam caveat

CDIP's OC043 and OC131 are nearshore model-output points, not measured buoys. We checked the hourly hindcast-to-nowcast handoff at 31 March / 1 April 2025 against quality flags and a measured offshore buoy. There was no missing hour, duplicate, or obvious step in monthly wave height. The model/buoy median height ratios shifted by roughly 2%, comparable to another within-product buoy-period difference. But the buoy may itself feed the model, so this does not prove nearshore accuracy. Wave direction remains unsuitable for transport claims without an independent check. The wave candidate below used height and period only.

Satellite pilot: useful imagery, unstable agreement

We extracted Sentinel-2 waterlines with CoastSat, using a proxy tide correction and an exploratory wave-runup correction. Visible waterline overlays at the approximate transects generally follow a wet/dry or swash edge; foam and days-to-weeks separation from the field surveys remain problems. The plotted lines are not surveyed MHW contours.

On the fixed 12 previously examined change comparisons, the wave/runup version matched direction in 10/12 and missed by 16.68 ft versus a 32.67-ft no-change comparator. That cleared its stated numerical rule, but the same cases had influenced earlier geometry/window decisions. On six newly evaluable spring-2023 comparisons, direction was only 3/6 and error 46.52 ft versus 30.67 ft for no change. This is a provisional, in-sample diagnostic—not predictive validation or permission to substitute satellite waterlines for City widths. [Fixed diagnostic rows]

Phase 2: predicting later City beach widths

The target was the City's MHW beach width at each transect. We fixed the first four citywide surveys (fall 2022–spring 2024) as the setup period and scored the last three (fall 2024, spring 2025, fall 2025): 12 × 3 = 36 comparisons. Reported zeros remain real zeros. Mean absolute error is the average absolute miss in feet; lower is better. Later survey values had already been seen when the one wave rule was specified, so this is not a blind independent test.

Sources: 36 × 3 baseline predictions, 36 wave predictions, and the Phase 2 method summary.

The single pre-specified wave candidate

The rule added one pooled coefficient for the preceding 90-day, two-point CDIP Hs²Tp anomaly to each transect's same-season average. Its expected sign was negative: rougher waves, narrower beach. We documented and approved that rule before fitting once. Wave-window QC passed 14/14 windows; the coefficient was −4.79. The fall-2024 window was calm (anomaly about −2.0), so the rule predicted widths about 9.7 ft wider than the seasonal mean. Actual fall-2024 widths were close to the plain average. A declared product-seam sensitivity scored 12.14 ft and did not alter the decision. No further candidates will be fitted to these same outcomes.

All 36 later widths

Observed City width and four predictions, in feet. Values rounded to one decimal here; downloads preserve full precision. This table is the check record, not 36 independent weather experiments.

Limits and next steps