Joseph HeinrichSoftware engineer
← All Side Quests
Oct 24, 2026Visualization

The Moving Edge

See where published shoreline measurements moved landward or seaward over decades.

Try the project ↓
How to use this project

Help

How to explore

A shoreline can move toward land or toward water over time. This page compares long-term rates measured at cross-shore lines in 21 coastal states.

  1. Start with Louisiana

    Read its median yearly rate, then choose another coastal state.

  2. Test the direction

    Compare the sign of each estimate with a stricter reading that requires its 90% range to stay on one side of zero.

  3. Read the spread

    Use the rate bands to see how much local measurements differ within the selected state.

  4. Compare counties

    Rank represented counties, choose one, and carry it back to the coordinate sketch.

Measuring line (transect)
A cross-shore line where old and newer shoreline positions are compared.
Shoreline-change rate
The slope of a best-fit line through shoreline positions, reported here in metres per year.
Erosion
Landward shoreline movement, shown as a negative rate.
Accretion
Seaward shoreline movement, shown as a positive rate.
Median
The middle value after the published rates are put in order.
90% confidence interval
The published uncertainty range around a rate; this project calls a direction clear only when the full interval stays above or below zero.
Side Quest / VisualizationStart with Louisiana, choose any of 21 coastal states, compare the reported direction with a stricter 90% interval test, read the state's full rate distribution, and select a county to connect its exact summary with the sampled coordinate marks.

Long-term shoreline change · 21 states

The moving
edge

See where published shoreline measurements moved landward or seaward over decades.

A shoreline does not stay fixed. The U.S. Geological Survey compared old and newer shoreline positions at thousands of cross-shore measuring lines, then estimated an average yearly pace at each line.

landwater

− landward

+ seaward

01 · Choose a state

Follow the coast around the country.

Choose a state to update every view. The large number is the median rate across that state's published measuring lines: negative is landward, positive is seaward.

WestPacific coast
GulfGulf coast
EastAtlantic coast

Louisiana · Gulf coast

−6.43 metres per year

Median long-term rate across 3,451 published measuring lines. The middle half runs from −12.64 to −2.46 metres per year.

Record range
1853–2001
Counties represented
5
Lines moving landward
89%

The lines are not spaced evenly among states. These summaries describe the published records, not a percentage or average of total coastline length.

02 · See measuring lines

Direction is clear in some places and uncertain in others.

Each mark is a sampled measuring line placed by its recorded coordinates. Circles moved landward, squares moved seaward, and diamonds do not have a clear direction under the active reading.

Louisiana coordinate sketch

All represented counties

LandwardSeawardUnclear

This is a coordinate sketch of up to 240 evenly selected source records, not a navigational map or a drawing of the shoreline. County percentages beside it use every source record.

03 · Read the spread

One state can contain very different local rates.

Bar height shows how many Louisiana records fall in each rate band. The zero line separates landward movement on the left from seaward movement on the right.

04 · Compare counties

Nearby records can point in opposite directions.

Rank the represented counties in Louisiana, then choose one. The selection highlights sampled marks above and replaces the direction summary with the county's full set of records.

What to carry forward

A direction is not the whole story.

59.1% of all published records have a negative rate, but only 35.2% stay negative across their 90% interval.

Louisiana has a median of −6.43 metres per year. Washington has a median of +1.97.

45% of intervals cross zero. That does not prove the shoreline stood still; it means this test does not support one clear direction.

Read the data limits

The release combines regional USGS studies with different date ranges, source shorelines, line spacing, and methods. Every rate spans at least 50 years, but the individual start and end years vary by place. The project preserves each published rate and its 90% confidence interval.

Counts and shares describe measuring-line records, not miles of coast, land area lost, property risk, causes, or future change. Adjacent lines are not independent observations. A shoreline proxy can also differ from the legal boundary, the water's edge on a particular day, or a safe place to build.

The downloaded table contains 133,113 active records. Its May 2026 metadata reports 126,831 vector objects. Because those counts differ, this project pins and validates the actual archive and reports the discrepancy instead of silently substituting the metadata count.

A standard-library Python builder verifies the exact 10,063,056-byte USGS source archive, parses 133,113 dBase records, checks the schema and published domains, and writes a 672,278-byte static browser snapshot with no runtime request.

Why this exists

The question

Carrying one state through a coordinate sketch, a rate distribution, an uncertainty test, and county comparisons will make long-term shoreline movement easier to understand than a national average alone.

What to try

Start with Louisiana, choose any of 21 coastal states, compare the reported direction with a stricter 90% interval test, read the state's full rate distribution, and select a county to connect its exact summary with the sampled coordinate marks.

Method and limits

About the data

I selected the U.S. Geological Survey's May 2026 national shoreline-change release through the allowlisted USGS provider, then opened the official release page, ScienceBase records, full attribute metadata, exact download, and USGS licensing guidance. The release is marked CC0 1.0. It compiles long-term rates from regional USGS studies for the contiguous United States. The shoreline dates represented by those studies range from 1830 through 2018, and each long-term rate spans at least 50 years.

The ScienceBase landing-page download contains a nested file named `National_Shoreline_Change_Rates_LT.zip`. That exact source archive is 10,063,056 bytes with SHA-256 `eff2beedaa2727623e6c7456fc01e3f3588de94dc1f00f46bcf0c1ce7899653f`. A standard-library Python builder opens the nested archive and parses its 89,585,627-byte dBase table without a GIS dependency. It verifies the 17-field schema, 133,113 active records, unique national identifiers, 21-state set, three coastal areas, coordinate bounds, nonnegative confidence intervals, documented shoreline-count domain, and published −39.62 to +33.51 metre-per-year rate range. Repeated reductions produce the same 672,278-byte JSON snapshot with SHA-256 `b07869414d49860e53cc882ca390192b68cf7a432fb10304417753bd60150e88`. The page imports the snapshot and makes no runtime request.

The downloaded table declares and contains 133,113 active records, while the release metadata reports 126,831 vector objects. I preserve and report that discrepancy. All counts and summaries use the records in the pinned archive. State and county medians, quartiles, direction shares, and histograms use every applicable source record. The coordinate sketch keeps at most 240 records per state, selected at even intervals through the published national ordering; it is explicitly labeled as a sample and is not used to calculate displayed shares.

The published linear regression rate is the slope of a best-fit line through shoreline positions. Negative values indicate landward movement, or erosion; positive values indicate seaward movement, or accretion. The stricter direction view uses the published 90% confidence interval: erosion requires the rate plus the interval to remain below zero, and accretion requires the rate minus the interval to remain above zero. Intervals that cross zero are labeled unclear, not motionless.

The release combines regional studies with different dates, shoreline sources, line spacing, and processing histories. Measuring lines are not spaced evenly among states or counties, and nearby lines are not independent observations. Counts and percentages therefore describe published transect records, not shares or averages of coastline length. These data do not measure land area lost, property or safety risk, causes, legal boundaries, a shoreline on a particular day, or future movement. A smoothed long-term rate can hide storms, reversals, seasonal movement, management work, and recent change.

Skills used

  • Data analysis
  • Data engineering
  • Statistical reasoning
  • Systems design
  • Visualization
  • Information design
  • Interaction design
  • Visual design
  • Creative coding
  • Product thinking
  • Editorial storytelling
  • Accessibility
  • Performance

What the data shows

  1. Across the 133,113 records in the pinned table, 59.1% have a negative rate. Only 35.2% remain negative across their published 90% confidence interval, while 45.0% of all intervals cross zero.
  2. Louisiana's 3,451 published measuring lines have a median rate of −6.43 metres per year. The middle half runs from −12.64 to −2.46 metres per year, and 71.7% remain negative under the stricter interval test.
  3. Washington's 2,154 published lines have a median rate of +1.97 metres per year, while Delaware's 777 lines have a median of −1.01. These are record summaries, not length-weighted state coastline averages.
  4. The exact source table contains 133,113 active records although the release metadata reports 126,831 vector objects. The project pins the table, validates its contents, and keeps that source discrepancy visible.

Sources

U.S. Geological Survey · CC0 1.0 UniversalUSGS National Shoreline Change — national compilation ↗U.S. Geological Survey · CC0 1.0 UniversalLong-term shoreline-change transect dataset ↗U.S. Geological Survey · Public-domain U.S. government metadata with USGS attributionTransect dataset attribute metadata ↗