Shoreline Change Analysis with DSAS 6.1: Step-by-Step

A shoreline is a moving boundary. Waves, tides, longshore drift, sea-level change and storm surge move it; groynes, seawalls, reclamation and sand mining move it faster. Quantifying how far and how fast it moves is the basis of erosion hazard mapping, setback lines, and every coastal EIA that gets past a competent reviewer.

The tool almost everyone uses for that computation is the USGS Digital Shoreline Analysis System (DSAS), cited in well over 800 peer-reviewed papers. This guide walks through the current release, DSAS 6.1, from data preparation in ArcGIS Pro or QGIS to a classified rate-of-change map, with every statistic written out so you can defend it in a methods section.

What changed in DSAS 6.1

DSAS grew up inside Esri software: version 1.0 ran on ArcInfo in 1992, versions 4 and 5 were ArcMap add-ins. ArcMap reached end of life on 1 February 2024, and the USGS no longer supports the Esri-dependent versions (5.1 and earlier). DSAS 6.0 (2024) and 6.1 (February 2026) are standalone desktop applications that sit alongside whichever GIS you use.

The practical consequences:

• You prepare shorelines and a baseline in ArcGIS Pro or QGIS, then export them as shapefile or GeoJSON and upload them to DSAS.
• DSAS stores baselines, shorelines, transects and results in a single GeoPackage (.gpkg).
• Baselines may be onshore, offshore, midshore, or a combination.
• Results export as shapefile, GeoJSON, CSV or GeoPackage, with a plain-text summary report.
• Basic editing tools exist inside DSAS; for anything substantial, export, edit in your GIS, and re-upload.

Four concepts you must get right

Shoreline indicator

The feature you digitise as "the shoreline": the high water line, a datum-based mean high water line, the wet/dry line on imagery, or the vegetation line. Pick one and keep it consistent across every date. Mixing indicators introduces a proxy-datum bias that DSAS can correct only if you record the indicator type per shoreline.

Reference baseline

A line you draw, from which DSAS casts transects. It must follow the general trend of the coast and sit clear of the shoreline cluster. DSAS needs to know which side is land: that is what gives rates their sign, positive for accretion, negative for erosion.

Transects

Straight lines cast perpendicular to the baseline at a fixed alongshore spacing. Three parameters control them: search distance (how far a transect reaches for shorelines), transect spacing (alongshore interval in metres) and smoothing distance (how much the baseline is generalised so local kinks do not swing transect orientation).

Intersection points

Where each transect crosses each shoreline, DSAS records the distance from the baseline, the shoreline date and its uncertainty. On embayed coasts a transect can cross the same shoreline twice; you choose whether DSAS keeps the closest, farthest, seaward or landward crossing.

The statistics DSAS computes

Everything below is per transect. Distances are in metres, rates in metres per year.

Shoreline Change Envelope (SCE)

The greatest distance between any two shorelines crossing the transect, regardless of date or direction. Always positive.

Net Shoreline Movement (NSM)

The distance between the oldest and the most recent shoreline. Positive is net accretion, negative is net erosion.

End Point Rate (EPR)

NSM divided by the years elapsed between the two end shorelines. It needs only two dates, and it ignores every date in between.

Its uncertainty comes from the positional uncertainty of the two shorelines, \(u_A\) (youngest) and \(u_B\) (oldest):

Linear Regression Rate (LRR)

With three or more dates, fit a least-squares line through all intersection points on the transect. The slope is the rate. Here \(x_i\) is the year of shoreline \(i\), \(y_i\) its distance from the baseline, and the bars denote means.

Weighted Linear Regression Rate (WLR)

The same regression, but each shoreline is weighted by the inverse of its uncertainty variance, so a precise LiDAR shoreline counts for more than a 1970s topographic sheet.

DSAS also reports, for both regressions, the coefficient of determination (LR2, WR2), the standard error of the estimate (LSE, WSE) and a confidence interval (LCI, WCI) from the Student's t-distribution at a confidence level you set, 99.7% by default. Report the interval alongside the rate; a rate whose interval spans zero is not evidence of change.

Where shorelines come from

The uncertainty you assign to each shoreline drives EPR uncertainty and WLR weights, so it has to be honest. Typical values from the literature:

For satellite-derived shorelines, download imagery from USGS EarthExplorer or the Copernicus Data Space. Aim for 20 to 50 years of coverage for long-term rates, with at least three dates if you want LRR or WLR.

The workflow

1. Digitise the shorelines

Load each image and build a composite with strong land–water contrast: Landsat 8/9 Bands 5, 6, 4 (NIR, SWIR1, Red) or an NDWI raster; Sentinel-2 Bands 8, 11, 4. Create a polyline feature class in a projected coordinate system in metres and trace your chosen indicator for every date. Assign each shoreline a positional uncertainty that accounts for source resolution, georeferencing error, digitising precision and tidal stage.

2. Add the required attribute fields

DSAS will not run without these. Field names follow the DSAS 5.1/6.1 conventions.

The baseline needs one field: a unique integer ID per segment. Baseline and shorelines must share the same projected CRS. For the southern coast of India that is WGS 1984 UTM Zone 44N (EPSG:32644); set it under Map Properties → Coordinate Systems in ArcGIS Pro, or Project → Properties → CRS in QGIS.

3. Draw the baseline

Three methods, in order of how often I use them:

• Buffer a shoreline. Buffer a representative shoreline 50 to 100 m landward, convert the buffer polygon to a polyline, and delete the seaward side. The baseline inherits the coastal orientation automatically.
• Smooth a shoreline. Run Smooth Line (Cartography toolbox) on an existing shoreline, then offset it landward or seaward.
• Draw it by hand. A smooth polyline following the general trend, placed clear of every shoreline. Fine for short, simple coasts.

Onshore (landward of everything) is the standard placement. Midshore baselines work in 6.1 but only if you set the land-orientation flag correctly, which is the single most common way to get every sign in the study reversed.

4. Export and upload

Export the baseline and every shoreline as shapefile or GeoJSON, then upload them to DSAS 6.1. DSAS writes them into a GeoPackage that becomes your project file. (If you are still on DSAS 5.1 in ArcMap, all feature classes must sit in the same personal or file geodatabase instead.)

5. Cast transects

Then look at every transect. Delete or edit any that cross land features, cross each other, or miss some shorelines. Transects that cross each other almost always mean the baseline has a sharp bend: raise the smoothing distance or simplify the baseline.

6. Calculate statistics

Run the rate calculation. DSAS computes SCE, NSM, EPR, LRR and WLR (with their uncertainty and regression diagnostics) and stores a rates layer and an intersects table in the GeoPackage, plus the text summary. Read the sign convention before you interpret anything:

7. Classify and map

Symbolise the transects by EPR or LRR with a diverging scheme, red for erosion and green or blue for accretion. A classification that reads well and matches most published Indian coastal studies:

The layout needs a title naming the study area, period and metric; north arrow and metric scale bar; a legend with units; an inset location map; a graticule; data credits and the projection; and a 300 DPI export. Reviewers notice when the "stable" class is not tied to the uncertainty.

Errors that get shoreline papers sent back

• Erosion and accretion swapped. Baseline on the wrong side, or the wrong land-orientation flag on a midshore baseline. Check one transect you know by eye before you classify anything.
• Transects missing shorelines. Search distance too short or baseline too far away. Increase the search distance or move the baseline closer to the cluster.
• Missing statistics on some transects. They do not cross the minimum number of shorelines, or a date is mis-typed. Check the DATE_ field is text in MM/DD/YYYY.
• Coordinate system mismatch. Every input in one projected CRS in metres. Geographic degrees will produce nonsense rates.
• GeoPackage will not load. Re-export from your GIS and make sure only polyline features are included.
• Mixed shoreline indicators without a type field. The rates absorb a proxy-datum bias that a reviewer will ask about.

Frequently asked questions

Written by Dr. Aran Castro A J, PhD in Applied Geology on the coastal geomorphology of Kanyakumari, GIS Manager, and author of Coastal Geomorphology: Processes, Landforms, and Human Interactions.

← All blog posts