1

Fig. 1. A map of Finland and location of study area of forest attractiveness in Ruunaa area in Lieksa, eastern Finland.

2

Fig. 2. Evaluated Scots pine dominated Vaccinium-type forest views in panorama images including two replicate sites which were unharvested (control, basal area 26 m2 ha–1), a selective cutting site (basal area 18 m2 ha–1), small openings sites (gap cut) with 20 and 5% retained trees, respectively, and one site which was clear cut with 3% retained trees.

3

Fig. 3. Example photo and video views used in our study (selective cut). Standard photo size cropped with a red dotted line from a panoramic image.

Table 1. Evaluation event of measuring forest attractiveness: grouping variables related to the evaluation event and methods (% of respondents).
Grouping variable: Category:
To evaluate this number of photos and videos *) Moderately or very effortless 51.4
The evaluation of attractiveness in general *) Moderately or very easy 52.9
Photos contra videos (the easy of evaluations): *) Easier from the photos 17.1
Easier from the videos 27.1
Equally easy 55.7
Imagining the forest from the photos *) Well or very well 75.7
Imagining the forest from the videos *) Well or very well 84.3
The quality of the images was good Yes 100
The quality of the videos was good Yes 95.8
The time of looking photos was Too long 5.7
About right 91.4
Too short 2.9
The length of the videos was Too long 41.4
About right 55.7
Too short 2.9
The vertical motion in the videos: *) Positive 40.0
Neutral 50.0
Negative 10.0
The poorer quality of images and videos would have affected to the evaluation *) Yes 56.3
Noises (e.g., birds and wind) would have affected to the evaluation *) Yes 53.5
Particular attention to the forest floor *) Yes 45.7
Particular attention to the trees *) Yes 54.2
*) Involved in group analyses according to the grouping presented here.
Table 2. Statistical differences in forest attraction between visualization methods: entire data and treatments. Significant < 0.01 in bold and significance 0.01–0.05 in italic and underscore.
Treatment: F p Contrast: F p
Average of all treatments 19.992 <0.001 Video vs. Photo 25.465 <0.001
Video vs. Panorama 24.336 <0.001
Photo vs. Panorama 0.119 0.731
Untreated 16.722 <0.001 Video vs. Photo 22.863 <0.001
Video vs. Panorama 17.719 <0.001
Photo vs. Panorama 2.490 0.119
Selectively cut 18.601 <0.001 Video vs. Photo 25.909 <0.001
Video vs. Panorama 20.132 <0.001
Photo vs. Panorama 1.508 0.224
Gap cut 20% 3.071 0.054
Gap cut 5% 4.018 0.021 Video vs. Photo 3.660 0.060
Video vs. Panorama 7.456 0.008
Photo vs. Panorama 0.391 0.534
Clear cut 4.059 0.029 Video vs. Photo 2.599 0.111
Video vs. Panorama 6.532 0.013
Photo vs. Panorama 1.880 0.175
4

Fig. 4. Statistical differences in forest attraction between visualization methods: entire data and treatments. Equal line connects methods that did not differ statistically from each other (variance analysis of repetition measurements). Statistical significance: * < 0.05; ** < 0.01; *** < 0.001. Forest treatments presented in the Fig. 2 caption.

Table 3. Statistical differences in forest attraction between visualization methods in group and pair comparisons: entire data and treatments. Presentation methods: V = Video; Ph = Photo; Pa = Panorama. Significant < 0.01 in bold and significance 0.01–0.05 in italic and underscore.
Grouping factor Treatment Group * Method Contrasts of Interaction
F p Contrast F p
Easiness of assessing attractiveness Entire data 5.847 0.005 V vs. Ph 4.294 0.042
V vs. Pa 9.955 0.002
Ph vs. Pa 1.625 0.207
Untreated 3.644 0.033 V vs. Ph 4.576 0.036
V vs. Pa 4.816 0.032
Ph vs. Pa 0.110 0.741
Selectively cut 3.772 0.030 V vs. Ph 1.242 0.269
V vs. Pa 7.106 0.010
Ph vs. Pa 3.406 0.069
Gap cut 5% 3.619 0.031 V vs. Ph 1.237 0.270
V vs. Pa 7.930 0.006
Ph vs. Pa 2.528 0.117
The easiness of methods Entire data 3.123 0.020 V vs. Ph 3.813 0.027
V vs. Pa 3.846 0.026
Ph vs. Pa 0.474 0.624
Untreated 3.206 0.017 V vs. Ph 4.101 0.021
V vs. Pa 3.367 0.040
Ph vs. Pa 1.250 0.293
Selectively cut 4.295 0.003 V vs. Ph 6.570 0.002
V vs. Pa 3.157 0.049
Ph vs. Pa 2.205 0.118
Gap cut 20% 2.549 0.045 V vs. Ph 2.298 0.108
V vs. Pa 3.592 0.033
Ph vs. Pa 1.046 0.357
Particular attention to undergrowth Entire data 5.564 0.006 V vs. Ph 6.028 0.017
V vs. Pa 8.067 0.006
Ph vs. Pa 0.152 0.697
Untreated 4.382 0.017 V vs. Ph 5.093 0.027
V vs. Pa 6.477 0.013
Ph vs. Pa 0.009 0.925
Selectively cut 3.738 0.031 V vs. Ph 0.489 0.487
V vs. Pa 6.664 0.012
Ph vs. Pa 5.626 0.021
Gap cut 20% 3.592 0.033 V vs. Ph 3.275 0.075
V vs. Pa 5.349 0.024
Ph vs. Pa 0.880 0.352
Table 4. Statistical differences in forest attraction between visualization methods in intragroup comparisons: entire data and treatments. Presentation methods: V = Video; Ph = Photo; Pa = Panorama. Significant < 0.01 in bold and significance 0.01–0.05 in italic and underscore.
Treatment Contrast Easiness of assessing attractiveness
Moderately or very easy Less easy
t p t p
Entire data V vs. Ph 2.091 0.044 5.994 <0.001
V vs. Pa 1.707 0.096 5.363 <0.001
Ph vs. Pa 1.168 0.251 0.630 0.533
Untreated V vs. Ph 2.042 0.049 4.784 <0.001
V vs. Pa 1.635 0.111 4.155 <0.001
Ph vs. Pa 0.988 0.330 1.359 0.184
Selectively cut V vs. Ph 2.953 0.006 4.201 <0.001
V vs. Pa 1.467 0.151 5.184 <0.001
Ph vs. Pa 2.577 0.014 0.380 0.707
Gap cut 5% V vs. Ph 0.567 0.574 2.059 0.048
V vs. Pa 0.086 0.932 3.479 0.001
Ph vs. Pa 0.666 0.509 1.507 0.142
Treatment Contrast The easiness of methods
Easier from the photos Easier from the videos Equally easy
t p t p t p
Entire data V vs. Ph 4.720 0.001 3.416 0.003 2.457 0.019
V vs. Pa 3.398 0.006 3.130 0.006 2.410 0.021
Ph vs. Pa 0.621 0.547 0.628 0.538 0.708 0.483
Untreated V vs. Ph 3.680 0.004 3.024 0.007 2.100 0.042
V vs. Pa 2.207 0.050 3.474 0.003 1.652 0.107
Ph vs. Pa 1.287 0.224 0.317 0.755 1.160 0.253
Selectively cut V vs. Ph 4.072 0.002 3.456 0.003 2.476 0.018
V vs. Pa 3.500 0.005 2.891 0.010 1.957 0.058
Ph vs. Pa 2.493 0.030 0.403 0.692 0.796 0.431
Gap cut 20% V vs. Ph 2.930 0.014 1.102 0.285 0.864 0.393
V vs. Pa 2.449 0.032 1.481 0.156 0.157 0.876
Ph vs. Pa 0.484 0.638 1.022 0.320 1.000 0.324
Treatment Contrast Particular attention to undergrowth
Yes No
t p t p
Entire data V vs. Ph 1.744 0.091 5.495 <0.001
V vs. Pa 1.313 0.199 5.747 <0.001
Ph vs. Pa 0.631 0.533 0.100 0.921
Untreated V vs. Ph 1.681 0.103 4.863 <0.001
V vs. Pa 1.169 0.251 4.304 <0.001
Ph vs. Pa 1.000 0.325 1.339 0.189
Selectively cut V vs. Ph 2.483 0.019 4.946 <0.001
V vs. Pa 0.975 0.337 6.533 <0.001
Ph vs. Pa 2.436 0.021 0.666 0.510
Gap cut 20% V vs. Ph 0.395 0.696 2.675 0.011
V vs. Pa 0.431 0.669 2.701 0.010
Ph vs. Pa 1.027 0.313 0.397 0.694
5

Fig. 5. The visualization method averages with groupings that showed a statistical difference in forest attraction (y-axis). At the top side by side are the groups that had statistical significant differences in the entire data. Below them are the statistical significant differences that those groups had with different forest treatments. Statistical significant differences between groups in Table 3 and internal in Table 4. Forest treatments presented in the Fig. 2. caption.

6

Fig. 6. Attractiveness of forest stands (y-axis) by various presentation methods (video, photo, panorama, and field assessment). Field assessment carried out in 2017 on the same forests but with a different person (Silvennoinen et al. 2022). Forest treatments presented in the Fig. 2 caption.