Dynamic Forests of Great Duck Island

Great Duck Island is a 237-acre island 15 km south of Mount Desert Island, Maine. It served as a manned Coast Guard lighthouse post from 1890 until 1986, when it was automated. Sheep grazed the island from the late 19th century until 1951, dramatically impacting the landscape and ecology of the island. In 1985, the Nature Conservancy and the State of Maine gained control of most of the island, collaborating with the College of the Atlantic Eno Research Station to monitor the ecology of the land.

In the last few years, the recruitment rate of spruces on the island has dropped dramatically--one hypothesis points to the population of hares introduced in the latter half of the 20th century. The concern is the impact of hare browsing on new growth. Increasingly, the older trees are succumbing to blowdowns, and the stabilization of the forest is becoming an increasing concern. The peak of forest recovery was in the early to mid-90s and has been in decline since. This project was made to track the change over time and to inform future understanding of the forests.

Before we get to the data analysis, let's take a look at the digitization process, from image to final polygon. Two polygon layers, representing the forest in 1944 and 2017, respectively, were hand-digitized from aerial photography, over the course of more than 20 hours. The starting images are on the slider below.

Base images for digitization. 1944 aerial map on left, 2017 map on right.

In order to digitize the forest polygons, the Create Feature tool was used, and each vertice was placed individually. Each vertice (represented by the green boxes) along the line is a click, and the sequence of images below provides an idea of the scale the work was done on.

Feel free to navigate the maps by zooming in and out. Some clumps of trees were not digitized, since the canopy coverage was estimated to be less than 20%.

Before I could truly start to see the change across time, the second layer needed to be digitized. Next to turn the vector layers into polygons!

Using the Feature to Polygon tool, each vector layer was converted, and any interior polygons representing gaps in the canopy were selected and deleted, being careful to delete the polygon fill rather than the outline.

Between the two, the grayscale 1944 image was the simplest to digitize--less color variation allowed for more certainty (and likely more error, honestly) in what qualified as canopy. While I worked, I had plenty of time to think, and my mind strayed to how the process might be accelerated. Even in the simpler greyscale, however, I determined that creating software to assess the forest cover and do my work automatically would be difficult due to the fact that where the sun glances off the canopy, the pixel values exactly match those of the surrounding open fields.

Particularly on the southerly side of the middle cape on the west side of the island, the forest of 2017 has standing deadfall and some blowdowns that created complications in digitizing. You can also clearly see the 'stood' in the northernmost patch of woodland (so named because it is, in point of fact, where trees stood) that has been slowly regenerating since logging in the early '80s.

The new growth in the stood is primarily rowan, aspen, and birch, as I understand it. Those deciduous trees create a stark contrast to the dark coniferous canopy of the spruces. It also happens to be nearly the same color values as the open grasslands. The portions that had regenerated enough to count as forest cover were determined, but with the understanding there would be some error.

Now (finally) data analysis can be performed.

When the two layers are juxtaposed, it is possible to see the regeneration of the forest from the heaviest years of occupation--the path running up the southeast coast of the island has been obscured once more by the canopy, and the forest cover has spread, especially on the north end and the eastern side of the island. The majority of the blowdowns seem to be occurring on the west coast.

2017&1944 overlay

When we compare the 2017 polygon to the 2021 google images map, however, there are some portions of the forest that have been affected by more blow-downs, shrinking the forest further. The change is subtle, but on the timescale of a landscape, significant. With the recruitment rate so far down, the forest is unable to maintain size as the old trees fall. This poses a potential threat to the Storm Petrels that nest in the forested areas of the island--as the forest shrinks, their nesting habitat shrinks too. Tracking the rapidity of the loss of forest cover is important to the efforts to stabilize it, providing data on the urgency of action.

In the swipeable map below, the areas to focus on for evidence of blowdowns are circled in red. Because of the potential for error in georeferencing the google images, it is possible that some of the areas that the lines are offset from are not in fact blow-downs, but in those areas it does appear to be the case.

In total, the acreage of Great Duck Island, according to the calculations of ArcGIS Pro, is 234 acres. The area of the 1944 forest polygons totals to 66.3, while the area of the 2017 polygons totals to 77.4. According to these sums, there has almost exactly been a total gain of 11 acres. However, this data does not take into account the 1996 forest cover, the estimated peak documented in aerial photography (there very well may have been greater forest cover in a time before aerial photography, but that data is unavailable and unconfirmable.)

Ultimately, we know the forest is shrinking, and unlike in the past, it is not due to heavy land use by humans. This project attempts to ascertain the rate of decline in an effort to provide more data on the shifting landscape for those managing the land and conservation of Great Duck Island to have at their disposal.

made with

Products by ESRI:

ArcGIS Pro

ArcGIS Online and Webmap Builder

supported by

ArcGIS Storymap

data sources

ESRI & Maxar (Basemap, 2017)

Google Images (2021)

College of the Atlantic (GPSed coastline)

Acadia National Park (1944 aerial image)

Army Air Corps (1944 aerial image)

special thanks to:

Gordon Longsworth

Lundy Stowe

John Anderson

Truth Muller