This document explores spatial hotspots and seasonality of Flickr images in the Northern Forest. All maps are generated from images taken between 2012 and 2017 within the Northern Forest region (seen below).

 

Prior to analyzing the image data and generating maps, duplicate images (i.e., pictures taken by the same photographer on the same day and at the same coordinates) were omitted to reduce redundancies.

Ultimately, 116,360 images were used to identify regional hotspots. Of these, 89,231 (77%) images were taken in rural areas (i.e., outside of US census designated urban areas). We focus on the spatial and temporal patterns of these rural images in this document.

Images

Below are all of the original image locations, clustered by their density in the landscape. You can zoom in to see individual image locations, each with a url link to the original picture on Flickr.

Hotspots

To visualize areas of high photographic activity, we used kernel density estimation to generate a series of “hotspot” maps. These hotspots were delineated with individual boundaries for further analyses that assessed visitor movements throughout the region.

Regional hotspots

The above map shows rural image locations in the Northern Forest. Each point is sized proportionally to the number of images recorded per square kilometer.

The above map shows rural image locations in the Northern Forest. Each point is sized proportionally to the number of images recorded per square kilometer.

The above map shows the kernel density of rural images (higher values/brighter colors indicate areas of greater image density). A search bandwidth of 7.5 km was used to generate density values.

The above map shows the kernel density of rural images (higher values/brighter colors indicate areas of greater image density). A search bandwidth of 7.5 km was used to generate density values.

The above map groups kernel density image values into three categories: low (0 - 33% percentile), medium (33 - 66% percentile), and high (66 - 100% percentile) areas of rural photography.

The above map groups kernel density image values into three categories: low (0 - 33% percentile), medium (33 - 66% percentile), and high (66 - 100% percentile) areas of rural photography.

The above map shows hotspots of rural images in the Northern Forest. Hotspot regions were delineated using Huang's thresholding algorithm on the original kernel density raster.

The above map shows hotspots of rural images in the Northern Forest. Hotspot regions were delineated using Huang’s thresholding algorithm on the original kernel density raster.

We followed the same procedure as above to define distinct hotspots for the Northern Forest, except we included both urban and rural images. Some large contiguous hotspots (e.g., the Champlain Valley-Green Mountains region) were manually divided to reflect economically and geographically distinct recreation/cultural areas. In total, 29 individual hotspots were mapped as follows.

We used these discrete hotspots to divide the entire Northern Forest into 29 “hotspot regions” using Thiessen polygons. These hotspots/polygons were used to analyze visitor traces throughout the Northern Forest (see “weblink here”).

State hotspots

These next maps apply the same strategy, but to the individual states within the Northern Forest (New York, Vermont, New Hampshire, and Maine) to see more in depth perspectives of social media activity in each state.

New York

New York social media hotspots in the Northern Forest.

New York social media hotspots in the Northern Forest.

Vermont

Vermont social media hotspots in the Northern Forest.

Vermont social media hotspots in the Northern Forest.

New Hampshire

New Hampshire social media hotspots in the Northern Forest.

New Hampshire social media hotspots in the Northern Forest.

Maine

Maine social media hotspots in the Northern Forest.

Maine social media hotspots in the Northern Forest.

Tag patterns

Here we compare what visitors are actually taking pictures of in heavily trafficked areas (i.e., hotspots) versus low-use areas in the Northern Forest. This figure shows the top ten tags from images within hotspots versus images outside of hotspots, ranked by their frequency.

Fairly similar imaging targets within high and low use areas. The only major difference is the greater appearance of mountain-related images in hotspots, which don’t appear as frequently in other parts of the Northern Forest, suggesting that topography is a major natural feature triggering engagement by visitors.

Seasonality and diurnality

Seasonal photography

We look for seasonal hotspots and clustering in the Northern Forest. Here are seasonal kernel density maps for rural photography in the entire region.

A few things to take notice of here. Except for the Bigelow Preserve area, there is relatively little Flickr activity in Maine in the winter compared to the other Northern Forest regions. Flickr activity seems to be most widespread in the spring, with many local and regional hotspots appearing in each state.

Monthly photography

We can also compare the monthly patterns of Flickr photography between states in the Northern Forest.

We see that photography is greatest from June to October for all states. People take fewer photos during the winter, and this is especially true in Maine, where the number of images drops precipitously in November and doesn’t pick up until March.

Hourly photography

Finally, how does photography vary over the course of a single day? Here we use the timestamps associated with each Flickr image in the dataset to show these hourly photography rates for each state.

In the above figures, nighttime periods are shaded blue and daytime periods are shaded yellow. We see a similar pattern for each state, with Flickr activity greatest during daylight to early evening.

©