Assessment of the Visual Quality of Sediment Control Structures in Mountain Streams

Abstract

Sediment control structures such as check dams, groundsills, and revetments are commonly used to balance sediment transport. In this study, we investigated the visual quality of sediment control structures that have been installed to manage mountain streams by analyzing images from the Soil and Water Conservation Bureau (SWCB) of Taiwan. We used visual preference (P) as an indicator in the evaluation of visual quality and considered two softscape elements and four cognitive factors associated with P. The two softscape elements were the visible body of water and vegetation, which were represented by the percentage of visible water (W_R) and the percentage of visible greenery (G_R). We considered four cognitive factors: naturalness, harmony, vividness, and closeness. Using a questionnaire-based survey, we asked 212 experts and laypeople to indicate their visual preferences (P) for the images. We examined the associations of the P ratings with cognitive factors and softscape elements and then established an empirical relationship between P and the cognitive factors using multiple regression analysis. The results showed that the subjects’ visual preferences were strongly affected by the harmony factor; the subjects preferred the proportion of softscape elements to be 30% W_R and 40% G_R for optimal harmony, naturalness, and visual quality of the sediment control structures. We discuss the visual indicators, visual aesthetic experiences, and applications of the empirical relationship, and offer insights into the study’s implications.

1. Introduction

Sediment-related disasters in natural rivers usually occur as a result of erosion, landslides, debris flow, and other sources, especially during high pulses or floods, as particulates move downstream, altering the river’s geomorphology [1]. The interaction between riverbeds and sediment transport has interested scholars because sediment transport is closely associated with the erosion process, landscape evolution, and river geomorphology [2,3]. Thus, many researchers have conducted studies on sediment transport on the basis of theories, experiments, and numerical simulations [4,5,6,7,8]. In the field, sediment control facilities such as check dams, groundsills, and revetments are commonly constructed and used to reduce erosion by stabilizing torrents or streams and preventing flood and sediment disasters in mountainous areas. The top priorities for these facilities must be their function and safety; however, the intrusion of artificial structures into the natural landscape can have a negative visual impact, and these structures are sometimes criticized by the public as discordant and aesthetically unpleasant. Ideally, such structures would be functional, stabilizing, and visually acceptable. However, to determine how to design sediment control facilities that realize all three goals—to maintain the scenic beauty of the natural landscape—we first need a method to assess their visual impact. To date, there has been a lack of literature on the aesthetics of sediment control works, and there are no standard measures of visual quality, which has led to instances of jarring incompatibility between engineering structures and the natural environment [9]. More studies are needed on landscape aesthetics and public preferences to better integrate artificial but necessary engineered structures with the landscape. This study sought to quantify the public’s preferences for the integration of sediment control facilities into the natural landscape of mountain streams, with the underlying assumption that those considered more attractive would share measurable characteristics.

Many researchers have examined the visual quality of water landscapes using various metrics, such as the naturalness of stream landscapes, riverbank landscapes, or coastal environments [10,11,12,13]; the aesthetic value of the river flow [14]; and the visual preferences for a river corridor [15] and waterscape beauty [16]. However, only a few studies have focused on the aesthetic or visual quality assessment of hardscape structures such as hydraulic and sediment control engineering. In [17], an image power spectrum of the Nan-Shi-Ken Stream of Taiwan was used to study changes in visual preferences before and after stream remediation. The authors of [9] assessed the aesthetic quality of erosion control engineering structures using the scenic beauty estimation (SBE) method [18].

Studies of visual landscape quality can be classified into five approaches: ecological, formal aesthetic, psychophysical, psychological, and phenomenological [19]. A merger of the psychophysical and psychological approaches might provide a basis for a reliable, valid, and useful system of landscape quality assessment [19]. A basis for aesthetic design in soil and water conservation engineering is provided in [9]; however, their findings were based on the SBE results in a psychophysical approach and focused on qualitative analysis. We used the psychophysical and psychological approaches in this study and focused more on quantitative analysis.

Visual preferences for a landscape reflect the degree to which people experience it as beautiful (or ugly). Preference is a cognitive process and a product of perception that should be considered in studies involving people’s aesthetic judgments and selectivity, including perceptions of the environment [20]. Visual preferences for landscapes can be influenced by people’s psychological or cognitive characteristics and by the physical characteristics of the landscape. This study focused on people’s assessments of the visual quality of 16 different sediment control structures in mountain streams in Taiwan according to four cognitive factors (i.e., naturalness, harmony, vividness, and closeness) and the softscape elements of the landscape (i.e., water and vegetation). The objectives of this study were twofold: (1) to investigate people’s visual preferences associated with the four cognitive factors and the visible softscape and (2) to develop an empirical model for measuring visual preference and cognitive factors. The results can be used as a reference to assess the visual quality of sediment control structures in mountain streams and guide future designs.

2. Methods

The landscape that surrounds sediment control structures for mountain streams comprises many physical elements, including natural components (e.g., water flow, vegetation, soil, rock, sediment, and landforms) and artificial components (e.g., artificial structures and manufactured materials). The arrangement of these elements affects the overall scenic aesthetics. The aesthetics of each physical element could have been evaluated separately, but this would have been impractical and time-consuming and would have defeated the study’s purpose of assessing the integration of the parts. Therefore, only the main softscape elements—the water flow and vegetation—were quantified in this study, and the preference and cognitive factors were applied in the analysis of the overall visual impact. The four cognitive factors considered were naturalness, harmony, vividness, and closeness, all terms that have been commonly used in other studies of visual preferences associated with stream- or river-management engineering projects [16,17,21]. The terms are defined as follows: naturalness describes the perception of how close a landscape is to its original wild, undisturbed state [22]; harmony describes the perceived balance of elements to form a pleasing, unified whole; vividness describes the perception of the landscape as memorable, stimulating (positively or negatively), distinctive, or visually striking [23,24]; and closeness describes the perception of the landscape as confined, contained, or limited versus open or expansive.

Figure 1 shows a flowchart of the methodology used in this study. First, we collected images from the Soil and Water Conservation Bureau (SWCB) in Taiwan and processed them to calculate the areas occupied by vegetation and water. Second, we conducted a questionnaire survey by showing the participants the selected images and asking them to rate their preferences vis-à-vis the cognitive factors. Third, we performed a statistical analysis of these preferences, cognitive factors, and the percentages of vegetation and water in the images. Finally, we developed a model of visual preferences. Details of each step in the flowchart are provided in the following sections.

2.1. Image Collection and Processing

The selected study area, Taiwan, covers about 36,179 km²; it is 394 km long and 144 km wide at its widest point. Mountain ranges extend almost the entire length of the island, with 258 peaks that are higher than 3000 m. All of its rivers originate in the mountains in the central part of the island. Because the mountains in Taiwan are relatively high and steep, the flow velocity on hillslopes and in mountain streams is so high that it leads to severe soil erosion and sediment transport. Therefore, many engineering facilities, such as check dams, groundsills, and revetments, have been constructed in many mountain streams to reduce water velocity, trap sediment in gullies and streams, and stabilize the banks and riverbeds.

In this study, 16 engineering projects associated with mountain streams in Taiwan were collected (Figure 2). Some of these engineering projects have won Golden Quality Awards in Public Construction and Outstanding Agricultural Construction Awards from Taiwan’s SWCB. We collected 16 images of these projects in areas of Taiwan, including Yilan County, Miaoli County, Hsinchu County, Taichung City, Nantou County, Yunlin County, and Chiayi County. The collected images were shot from river banks or on a bridge. They can generally be captured as a naked-eye view, which mainly imparts the feeling of being at the scene to a viewer. Aerial, long-range, and panoramic shots were excluded because they cannot evoke the same immersive experience as normal visibility. Our selection criteria for images were that (1) the structure was easily visible (aerial, long-range, and panoramic images were excluded); (2) the image was sharp and clear (shot on a sunny day with high visibility); (3) the engineered elements were large enough to be visible (>30% of the visual field); (4) the image showcased engineering structures, streambeds, and streambanks; and (5) there was a variety of distinctive designs (arrangement, form, texture). We did not consider any seasonal or temporal changes at the sites of the images in this study. The facilities included check dams, grade control structures, and revetments (Figure 2 and Figure 3). The width of the streams ranged from 5 m to 20 m.

All of the images were adjusted to the same size (72.25 × 54.19 cm; 2048 × 1536 pixels) and resolution (72 dpi) in Adobe Photoshop. We calculated the percentages of vegetation ($G_R$) and water area ($W_R$) by identifying the dominant content of each pixel and determining the percentages of greenery and water in each image as follows [25]:

$$G_R=P_G/P_T$$

(1)

$$W_R=P_w/P_T$$

(2)

where P_G and P_W are the numbers of pixels occupied by green vegetation and water, respectively, in the whole image, and P_T is the total number of pixels of an image. Using Photoshop, we determined which pixels were occupied by green vegetation and which ones contained water elements using simple logic and eyesight; for pixels located on the boundary between the two, we zoomed in with reference to the color-palette sliders to decide whether they were more like “water” or “vegetation.” The calculation results of $G_R$ and $W_R$ are shown in Figure 3.

2.2. Questionnaire Survey

The platform for administering the questionnaire surveys in this study was built with the cloud questionnaire software SurveyCake. We distributed the questionnaire to related licensed engineers, academic researchers, consultant companies, and the general public through email and social media channels (LINE, WeChat, Facebook, etc.); these people and groups were then asked to help further disseminate the questionnaire. Although the online questionnaire was anonymous, participants provided their gender, age, and professional background (to distinguish between the general public and the experts) and then rated their visual preferences ($P$) and assessed the cognitive factors of each image: vividness ($V$), harmony ($H$), naturalness ($N$), and closeness ($C$). The participants were asked to quantify the visual preference ($P$) and cognitive factors ($V,H,N$ and $C$) using a five-point Likert scale: 5 = very high, 4 = high, 3 = medium, 2 = low, and 1 = very low. We sent 412 invitations asking Internet users to take part in this survey. There were two types of survey respondents: experts and the general public. The expert group comprised people who had either a master’s degree (or higher) or at least two years of related work experience in the fields of hydraulic engineering, civil engineering, soil and water conservation, or landscape architecture; the rest were classified as the general public. We received 223 responses (101 experts and 122 in the general public), and after excluding incomplete or invalid responses, we had a final sample of 212 participants (95 experts, 117 in the general public), distributed as follows: 20–30 years old (79%), 30–40 years old (14%), 40–50 years old (6%), and older than 50 years (1%).

2.3. Statistical Analysis

We first determined and analyzed the statistical properties of (1) the visual preferences ($P$) and the cognitive factors (harmony H, naturalness N, vividness V, and closeness C) according to the questionnaire samples; (2) the relationship between the visual preferences and the cognitive factors; and (3) the cognitive factors associated with G_R and W_R. We then developed models for $P$ associated with the cognitive factors and $P$ associated with G_R and W_R with multiple regression analysis using SPSS Statistics for Windows. Figure 3 shows the mean values of H, N, V, and C for the two types of survey respondents (experts and the general public).

3. Results

3.1. Relationship between Cognitive Factors and Softscape Elements

Landscape elements include water, vegetation, soil and rock, and landforms. The landscape elements focused on in this paper are water and vegetation, represented by the percentage of visible water (W_R) and the percentage of visual greenery (G_R). This section describes the W_R and G_R values associated with the cognitive factors V, H, N, and C.

In the questionnaire survey, participants ranked their visual preferences and all of the cognitive factors for each image from 1 to 5. We visualized the statistical properties of all of the cognitive factors for each image as boxplots (Figure 4, Figure 5, Figure 6, Figure 7, Figure 8, Figure 9, Figure 10 and Figure 11) and determined their mean values to analyze the cognitive factors associated with W_R and G_R.

3.1.1. Percentage of Visible Greenery (G_R)

The percentage of visual greenery (G_R) calculated by Equation (1) represents the amount of vegetation in a scene within a visual image. Figure 4 shows the percentage of visible green (G_R) associated with naturalness (N) for the two sample groups (experts and the general public). The two groups had similar trends: N increased with an increase in G_R when G_R < 40%, and N increased slowly and remained almost constant when G_R > 40%.

A high harmony ranking means that the participants perceived the composition of the elements as being balanced, unified, and coordinated as a whole. Visual harmony is the satisfying effect of combining similar or related visual elements such as adjacent colors, similar shapes, and related textures. Harmony can be enhanced (diminished) by introducing similar (discordant) elements and pleasing (jarring) spatial arrangements, and the level of harmony affects viewers’ visual preferences. Figure 5 shows that the harmony rankings trended upward as the G_R value increased to 40%, but they declined when the G_R value exceeded 40%. The reason may be that the participants perceived the stream landscape as disorderly or wild when there was “too much” greenery, which affected the overall sense of cohesion (G_R > 40%), resulting in a decrease in the image’s perceived harmony. A high vividness ranking means that the participants found the image memorable. Figure 6 shows that the vividness rankings trended downward as the G_R value increased. A high closeness ranking means that the participants viewed the image as relatively contained. As Figure 7 shows, the closeness rankings did not vary significantly when the G_R value was less than 40%, although it increased slightly when the G_R value was greater than 40%.

3.1.2. Percentage of Visible Water (W_R)

The percentage of visible water (W_R) represents the proportion of water in each image. Figure 8, Figure 9, Figure 10 and Figure 11 show the relationship between W_R and various cognitive factors. Among the four cognitive factors, naturalness (N) had a low association with W_R. The N value trended upward as the W_R value increased as long as W_R < 20%, and it decreased slightly when W_R > 20% (Figure 8). The value of harmony (H) trended upward slightly as W_R increased when W_R < 30%, and it decreased slightly as W_R increased when W_R > 30% (Figure 9). The vividness (V) had a positive association with W_R in both the general public and the expert responses (Figure 10). The V value trended upward as the W_R value increased. This result implies that vividness may be improved when the proportion of water flow in the image increases. Figure 11 shows that closeness (C) trended downward as W_R increased. A high closeness value means that the viewer perceived the space as confined, cramped, or limited. This implies that increasing the visibility of the water may help reduce viewers’ negative perceptions of the closeness of the space.

3.2. Relationship between Visual Preferences and Cognitive Factors

Figure 12a shows the relationship between visual preferences (P) and naturalness (N). P was positively correlated with N in the responses of both the general public and experts, showing the same tendency: P trended upward with an increase in N. Perceptions of naturalness are mainly affected by the amount and type of vegetation: positive N perceptions can be promoted by increasing (or improving the arrangement of) the vegetation or G_R; the N values were significantly higher when the G_R value was high (up to 40%) (Figure 4). These results are consistent with those of previous studies that indicate that vegetation can affect N, which, in turn, affects P [10,17,26,27].

Figure 12b shows a high correlation between P and H (R² = 0.96 for the general public group; R² = 0.80 for the expert group). In both groups, the results revealed that P increased with an increase in H. Visual harmony is related to unity and balance, which reflects a pleasant arrangement of landscape attributes that increases perceptions of beauty and public acceptance [28]. As mentioned previously, while function and safety are the top priorities, it is also important to consider visual harmony during the design phase for sediment control structures such as check dams [29]. Our study shows that harmony had a high weighting factor on visual preferences when the respondents evaluated the images of mountain stream engineering projects in the landscape. Moderate increases in the amount of water and vegetation—for example, G_R < 40% (Figure 5) and W_R < 30% (Figure 9)—tended to enhance perceptions of harmony and positive visual assessments.

Figure 12c shows that the visual preference value (P) was positively correlated with vividness (V) in the responses of both the expert and general public groups, suggesting that P can be effectively improved by increasing vividness. Perceptions of vividness can be affected by the presentation and characteristics of the elements, including the volume and velocity of the visible water and the type, amount, and arrangement of the vegetation. Our results reflect the fact that, in our study images, most of the vegetation was distributed along the banks of the streams and had similar forms and colors. Thus, a general increase in G_R might lead to increases in uniformity or decreases in variety, which would result in a decrease in perceived vividness; in contrast, a professionally designed selection and arrangement of vegetation (e.g., artful groupings, attractive plantings) might increase both G_R and V. In mountain streams with steep slopes and high flow velocities, the visual water flow varies, and increasing W_R can improve vividness.

Closeness (C)—the perception of a lack of openness, an isolated and confined space—had a significant negative association with visual preference (P) in the responses of both the expert and general public groups, as shown in Figure 12d; that is, P decreased as C increased. This suggests that the amount of visible water and vegetation influenced the respondents’ perceptions of closeness. Increasing the water’s visibility reduced C, as shown in Figure 11; that is, C decreased as W_R increased. In our study, excessive vegetation (G_R > 40%), however, led to slight increases in C (Figure 11) and decreases in visual preferences.

3.3. Relationship between Visual Preferences and Softscape Elements

The previous section reports that vividness (V) was significantly affected by the percentage of visible water (W_R); naturalness (N) was significantly affected by the percentage of visible green vegetation (G_R); and closeness (C) and harmony (H) were both significantly affected by G_R and W_R. The H values decreased when G_R was above 40% (Figure 5), and the H values began to decline as W_R rose above 30% (Figure 9). The value of H was highest when the G_R value was approximately 40% and the W_R value was approximately 30%. Because visual preference (P) had a high correlation with H, there was a decrease in the P value when the G_R value was higher than 40% (Figure 13), and the P level decreased slightly when the W_R value was higher than 30% (Figure 14).

3.4. Model of Visual Preference

We randomly divided the 212 valid questionnaires into two sets, and we used one set to develop a visual preference model and the other set to verify the model. The two sets had the same sample size of 106. The association of visual preference P with the cognitive factors H, N, V, and/or C was modeled by an ordinal multiple regression analysis using SPSS software. In the ordinal multiple regression, four models with different variables were tested, as shown in

Table 1. Results from the ordinal multiple regression in the model of visual preference.

Model Type	Variables xi and Constant in Model	Regression Coefficient	Standard Error	T	Significance Value	R2 of Model	Standard Error of Model
I	Constant	0.724	0.041	17.7	<0.001	0.604	0.804
	H	0.799	0.011	72.6	<0.001
II	Constant	0.383	0.043	8.9	<0.001	0.639	0.768
	H	0.659	0.013	50.7	<0.001
	N	0.241	0.014	17.2	<0.001
III	Constant	1.116	0.077	14.5	<0.001	0.652	0.754
	H	0.589	0.014	42.1	<0.001
	N	0.216	0.013	16.6	<0.001
	C	−0.151	0.013	−11.6	<0.001
IV	Constant	0.961	0.079	12.2	<0.001	0.658	0.748
	H	0.560	0.015	37.3	<0.001
	N	0.206	0.013	15.8	<0.001
	C	−0.134	0.013	−10.3	<0.001
	V	0.079	0.011	7.2	<0.001

Note: Model type P = ${\displaystyle \sum }{c}_i{x}_i$ + constant; c_i = coefficient in variable; T = regression coefficient/standard error; R² = coefficient of determination.

. Model IV, which included all four cognitive factors, had a higher determination coefficient (R² = 0.658) than the other three models. Thus, this model was adopted to assess visual preference. The model of visual preference can be expressed as follows:

$$P=0.56H+0.206N+0.079V-0.134C+0.961$$

(3)

Equation (3) shows that visual preference (P) was positively correlated with harmony (H), naturalness (N), and vividness (V) and negatively correlated with closeness (C). Among the four cognitive factors, harmony (H) had the strongest effect on visual preference (P). The respondents’ visual preference for the sediment control structures in the 16 images could be determined by factoring in the cognitive factors H, N, V, and C.

The second set of 106 questionnaires (the ones that were not used to develop the model) were used with the 16 images to verify the model. These 106 questionnaires were divided into an expert group (40) and a general public group (66). The mean values of P, H, N, V, and C of each image were determined from the questionnaires through statistical analysis. We were also able to evaluate the values of P using Equation (3) by factoring in the mean values of H, N, V, and C for an image. Figure 15 shows comparisons between the P values determined by the model (Equation (3)) and those obtained from the questionnaires. We also collected and compared previous study data [30] on P associated with perception factors and found 29 images of mountain stream-related engineering projects in Taiwan.

Figure 15 shows that the correlation coefficient between the statistical value and the model’s predicted value for visual preference (P) was 0.94. The average error was 4.9%, which is within an acceptable range. The visual preference value predicted by the model was in agreement with the results of the questionnaires.

4. Discussion

4.1. Visual Indicators

In this study, we developed an approach for explaining environmental preferences. However, in previous research, nine types of visual indicators were summarized, and a conceptual framework was proposed for understanding these indicators and a landscape’s physical attributes. The nine visual indicators were stewardship, coherence, disturbance, historicity, visual scale, imageability, complexity, naturalness, and ephemera. This terminology may not correspond exactly to that used by all researchers, but their general meanings are similar. For example, the terms coherence, harmony, and unity are used interchangeably in the literature; the term vividness is also described as imageability, and the term openness (the opposite of closedness) is often used as an indicator of visual scale. The four terms chosen for this work were included in the framework proposed by [22].

Different landscape types have different physical characteristics that require a different selection of indicators. For example, three indicators—coherence, complexity, and mystery—were introduced as predictors of preferences in the assessment of building façades in [20]. In our study, we used naturalness, harmony, vividness, and closeness to assess the respondents’ perceptions of the visual impact of the sediment control structures on mountain streams in Taiwan. It is necessary to test the applicability of visual indicators in different contexts and landscape types to provide a comprehensive framework for landscape assessment, including the development of visual indicators, and show that water and vegetation (the softscape elements considered in this study) can be regarded as potential indicators or landscape attributes when analyzing perceptions of visual quality [22]. In [31], vividness was strongly linked to water bodies. Vegetation has often been related to naturalness, and visual preferences can usually be enhanced by increasing this parameter [32,33], but not always [34]. Our findings are generally consistent with those of previous studies.

4.2. Visual Aesthetic Experience

Perceptions are subjective. Therefore, different individuals can have different visual aesthetic experiences of the same landscape. Respondents’ visual aesthetic experiences can be affected by their cultural background [35,36], education level [37,38], gender [37,39], age [40,41], professional knowledge [39,42], familiarity with the environment [43], and living environment [35,44,45]. The authors of [46] concluded that cultural background has an important influence on the perceptions of landscape beauty. However, the results in [47] showed that aesthetic perceptions of natural landscapes are generally similar, even for people from different cultural backgrounds. Nevertheless, such demographic factors should be taken into account in landscape planning and design [39,48]. Our study found no significant effects of education level, gender, age, or professional knowledge on the respondents’ assessments of visual quality. This result may be due to the respondents all sharing a cultural background and living environment (all were Taiwanese Internet users, and the study area was located in Taiwan). Future studies should be conducted with other types of respondents and other study locations to ensure generalizability.

4.3. Model Applications

A viewer’s visual preference P can be affected by many physical, psychological, and sociocultural factors. Among the physical variables related to stream landscapes that contain sediment control structures are the structural design elements (e.g., form, color, materials, or texture); the volume and velocity of the water; the amount and type of vegetation (e.g., naturally occurring and reproducing shrubbery, formal or decorative plantings added to the landscape); and the arrangement of the structures and vegetation in relation to the water (e.g., a structure sited on a purely functional basis with the surrounding vegetation reclaiming its original place, vegetation added or rearranged strategically for protective or aesthetic reasons). Among the psychological variables to consider are the four cognitive factors of naturalness, harmony, vividness, and closeness, although there may be others that could be considered in future studies, such as heightened safety concerns (e.g., fears of flooding) or biases against built structures in the wild or the use of certain materials (e.g., concrete versus stone).

In this study, an empirical model (Equation (3)) of the relationship between P and four specific cognitive factors was developed. Softscape elements (G_R and W_R) were also associated with cognitive factors. Landscape designers and structural engineers could use Equation (3) to factor people’s visual preferences into their plans and identify environmental characteristics that might make the project more aesthetically pleasing. When designing sediment control structures, the physical factors may need to be weighted more heavily than the psychological factors because the former are directly linked to the top design priorities of function and safety. Nevertheless, Equation (3) shows that harmony should be seriously considered. Harmony can be achieved by ensuring that all the hardscape elements (e.g., form, color, texture, or materials) and softscape elements (e.g., water and vegetation) are considered holistically, with an eye toward visually unifying and balancing the completed project. For example, engineering structures that use natural materials and forms, such as stone revetments and stone masonry in general, are more easily integrated with the surrounding environment than shapes and materials not found in nature. Ecological engineering structures with natural materials and vegetation are more readily accepted by the public [9]. Our results could be helpful for those who design and develop sediment control structures. However, more studies are needed to fully grasp the relationship between visual preferences and physical variables.

5. Conclusions

In this study, images of 16 sediment control structures in Taiwan were collected. Through questionnaire surveys and statistical analyses, we performed an analysis of the association between four cognitive factors and the softscape elements of water and vegetation and between visual preferences and the cognitive factors. The results can be summarized as follows:

• Higher values for naturalness (N) are associated with higher proportions of visible green vegetation (G_R), but the effect is less significant when the percentage is higher than 40%. Closeness values increase as G_R increases, but higher closeness values can decrease P values. Increasing the proportion of visible water (W_R) can effectively enhance the vividness of sediment control structures.
• Visual preference values improve significantly when vividness, naturalness, and harmony increase or when closeness decreases. The visual preference (P) value is highly correlated with harmony (H); P increases as H increases. Harmony can be affected by the proportions of water and vegetation (W_R and G_R, respectively). H increases with moderate increases in W_R and G_R; however, H decreases when G_R and W_R increase too much, that is, when G_R > 40% or W_R > 30%. P follows a similar pattern of increasing with increases in G_R or W_R (as long as G_R < 40% and W_R < 30%); however, P decreases when G_R > 40% or W_R > 30%. G_R can be maintained and managed by humans to keep the amount of greenery at about 40%, avoiding the overgrowth of vegetation that affects the visual preference of the public.
• Water and vegetation are softscape elements in landscape design. Engineering structures (one type of hardscape) can be “softened” by the strategic use of water and vegetation in a stream landscape. The respondents in this study preferred a proportion of 30% water and 40% green vegetation to achieve the most desirable balance of closeness, harmony, vividness, and naturalness. Water and vegetation (values of W_R and G_R) may change with seasons, so there could be an optimal season, such as spring or autumn, that has a high visual preference for a viewer. Determining which season is better for visual preference requires future study.
• We developed an empirical model of visual preference based on the relationship between visual preferences and four cognitive factors (Equation (3)). The evaluation results of the model are consistent with the results of the questionnaires and previous studies. When the four cognitive factors are factored in, the model determines the degree of visual preference.
• The images analyzed in this study were limited to sediment control structures in Taiwan’s mountain streams. The visual field of the human eye can enable people to simultaneously perceive both sides of a small- to medium-sized stream and its facilities from what might be called, idiomatically, “nearby.” Much more can be seen from a distant view—a panorama or aerial view—but with a loss of perceptible detail. This study did not use overhead or aerial views (images taken from high above), distant wide-field views, or panoramic views. Future studies should replicate this one but focus on sediment control facilities for larger streams or rivers (with widths of >20 m) and use different perspectives.