Describing and comparing landscapes using tags, texts, and free lists: an interdisciplinary approach

Figures & data

Table 1. Study sites and their landscape types in the Swiss landscape typology.

Location	Landscape type
Oeschinensee	Limestone mountain landscape of the Alps
Seealpsee	Limestone mountain landscape of the Alps
Thurauen, River Thur	River landscape
Bremgarten, River Reuss	River landscape
Robenhuserriet, Pfäffikon	Moor-influenced landscape
Ägeriried, Rothenthurm	Moor-influenced landscape
Ufschötti, Lake Lucerne	Urban landscape [with lake]
Zürichhorn, Lake Zurich	Urban landscape [with lake]
Hochwacht, Lägern	Landscape of hills of the Central Plateau with a focus on forage production
Hochwacht, Pfannenstiel	Landscape of hills of the Central Plateau with a focus on agricultural production

Figure 1. Overview of the methodological sequence applied to extract and compare landscape descriptions.

Figure 2. Processing steps to create footprints from web documents and extract georeferenced images.

Figure 3. Relative prominence of different landscape aspects depending on the data source.

Table 2. Cosine similarity values for the comparison between hiking blogs and free lists.

Study site	Cosine similarity (all terms)	Cosine similarity (landscape terms, no toponyms)	Difference
Oeschinensee	0.2622	0.3339	0.0716
Seealpsee	0.1991	0.2843	0.0852
River Thur	0.2852	0.2848	−0.0003
River Reuss	0.2579	0.2860	0.0281
Robenhuserriet	0.1555	0.1719	0.0164
Ägeriried	0.2220	0.3227	0.1007
Lake Lucerne	0.1343	0.2027	0.0684
Lake Zurich	0.1946	0.2392	0.0446
Lägern	0.1713	0.2156	0.0443
Pfannenstiel	0.1661	0.2165	0.0504

Table 3. Mann-Whitney-U values for within-document-type vs. across-document-type cosine similarity comparisons (statistically significant results in bold).

	All terms (including non-landscape)	All landscape terms	Landscape terms, no toponyms	Biophysical terms only	Sense of place terms only
Free lists	U = 8970, p< .001	U = 8873, p< .001	U = 8480, p< .001	U = 7429, p< .001	U = 8078, p< .001
Hiking blogs	U = 8898, p< .001	U = 5568, p = .013	U = 5921, p< .001	U = 4273, p = .598	U = 5834, p = .002
Flickr tags	U = 9000, p< .001	U = 8998, p< .001	U = 8374, p< .001	U = 7926, p< .001	U = 7190, p< .001

Table 4. Mann-Whitney-U values for within-landscape-type vs. across-landscape-type cosine similarity comparisons (statistically significant results in bold).

	All terms (including non-landscape)	All landscape terms	Landscape terms, no toponyms	Biophysical terms only	Sense of place terms only
Mountain landscape	U = 1344, p = .120	U = 1402, p = .058	U = 1344, p = .106	U = 1337, p = .131	U = 1444, p = .031
River landscape	U = 1405, p = .056	U = 1512, p = .011	U = 1680, p < .001	U = 1936, p < .001	U = 1200, p = .479
Moor landscape	U = 1161, p = .635	U = 1189, p = .523	U = 1126, p = .789	U = 1096, p = .927	U = 993, p = .606
Lake landscape	U = 1200, p = .481	U = 1267, p = .272	U = 1375, p = .083	U = 1676, p < .001	U = 913.5, p = .300
Hill landscape	U = 1206, p = .460	U = 1263, p = .282	U = 1150, p = .682	U = 1228, p = .385	U = 908, p = .302

Table 5. Mann-Whitney-U values for within-landscape-type vs. across-landscape-type cosine similarity comparisons, within a data source (statistically significant results in bold).

	All terms (including non-landscape)	All landscape terms	Landscape terms, no toponyms	Biophysical terms only	Sense of place terms only
Free lists: same vs. different landscape types	U = 197, p < .001	U = 197, p < .001	U = 199, p < .001	U = 194, p < .001	U = 118, p = .527
Hiking blogs: same vs. different landscape types	U = 117, p = .551	U = 163, p = .024	U = 160, p = .032	U = 167, p = .016	U = 140, p = .154
Flickr tags: same vs. different landscape types	U = 131, p = .271	U = 128, p = .321	U = 156, p = .045	U = 181, p = .004	U = 93, p = .814