Exploring the impact of urban development on mountain view visualization using a GIS-based landscape assessment model: a case study in Lishui, China

Abstract

Amid China’s high-speed urbanization, the lack of urban landscape and homogenization of urban image are becoming increasingly serious, prompting discussion on how to preserve the natural urban landscape. This study extends the existing landscape assessment model to evaluate the urban hillscape of the Lishui urban area of Nanjing over the past 50 years. First, a landscape quality analysis is conducted based on raster images, and a landscape quality map of the study area is drawn. Then, a visibility analysis of the site is conducted, key observation points (KOPs) are screened using a digital surface model, and the visual magnitude of each KOP is quantitatively calculated. Finally, the visual impact scores of the mountains in the site are synthesized. The results show that the model can accurately quantify the visual impact of historical mountain views. The following conclusions are drawn (1) The visual quality of the landscape in the area has improved as the urban area has spreads out. (2) The viewshed and magnitude of the area are trending downward over time. (3) The Chengnan area’s visual impact score increased by 25.7%, but that of the Tuanshan and Chengzhong areas decreased by 48% and 26.9%, respectively. This study presents a spatial decision support system that can obtain better landscape assessment results with more accessible open-source data, and can be used for historical landscape evaluation. By comparing the characteristics and trends of urban hillscapes in different historical periods, this study’s landscape assessment model can help managers better understand the impact of development on urban intentions and make better decisions.

Keywords:

• GIS
• landscape assessment model
• city mountain view
• visual impact assessments

1. Introduction

The combination of mountains and cities produces a distinctive urban landscape, and numerous countries engage in urban planning for mountainous regions. A notable example is Montreal, Canada, where city skyline regulations prioritize a harmonious relationship between buildings and natural ridgelines (Zacharias 1999). This results in the building skyline of the central business district either conforming to the contours of the ridgeline or adopting an alternate profile depending on the viewing perspective. In Kyoto City, Japan, the municipal government introduced the Kyoto City Overlook Landscape Creation Ordinance in 2007, aimed at enhancing the city’s scenic vistas (Xiang et al. 2000). Similarly, Vancouver, Canada, manages the development of buildings within the central business district through the allocation of landscape cones, ensuring unobstructed views of the two prominent mountains in North Vancouver (Liu et al. 2019). These initiatives are not unique: such cities as Athens, Rome, Edinburgh, Prague, San Francisco and Rio de Janeiro, among many others, are renowned for their exceptional urban landscapes set against mountainous backdrop (Geng et al. 2021).

In 1960, Kevin Lynch, in his book The Image of the City, introduced the notion that urban identity emerges from the dynamic interplay between the urban environment and the observer (Gao et al. 2023). The environment offers distinctions and relationships, while the observer brings adaptability. Consequently, it is important to examine the formation of mountain-inclusive urban cityscapes and underscore the seamless integration of artificial and natural landscapes to craft cityscapes with distinct local characteristics.

The emergence of visual evaluation research has garnered significant attention. The roots of early landscape visual evaluation research can be traced to the environmental movement of the 1960s, which sought to address the impacts of land and resource development activities on environmental quality. In response, the United States took the pioneering step of enacting the National Environmental Policy Act (NEPA) in 1969 (Hanks and Hanks 1969). This landmark legislation aimed to enhance people’s access to a better environment. It marked a turning point by granting legal status to landscape visual resources, which, although long enjoyed, had not previously received the same legal recognition as other economically valuable landscape resources.

In the same year, the U.S. Forest Service established the Visual Management System and Scenic Management System (Bacon Warren 1979). Subsequently, the U.S. Bureau of Land Management introduced a Visual Resource Management System (Smardon 1982). These systems set clear objectives for landscape resource management and provide invaluable guidance to landscape professionals. Furthermore, various related initiatives have been implemented, such as the London Landscape Management Framework in the United Kingdom (He and Hu 2020), and the Landscape Act in Japan (Heng and Liang 2006). In Hong Kong, China, the Hong Kong Urban Design Guidelines launched in 2003 mandate the protection of three major landscape elements: ridgelines, heritage buildings and view corridors (Yu 2003).

Since the 1960s, scholars have conducted research on various aspects of visual landscape control, including the influence of buildings on landscape shading, the impact of resource development on landscape erosion, the evaluation of viewshed landscapes of high-rise buildings (Steffen et al. 2018) and the examination of preferences for city skylines (Xu et al. 2013). The aim of such research is to imbue cities and neighborhoods with distinctive visual characteristics against the backdrop of urbanization. Natural urban landscapes typically rely on relatively stable elements, such as mountain ranges and water systems. However, most current research concentrates on assessing the visual landscape quality and extent of these natural urban landscapes. For instance, Sepideh et al. (2019) explored the visibility of mountain views with different developmental patterns using an urban development prediction model. However, specific analyses on the extent to which mountain views affect the size of urban areas are lacking. One of the factors contributing to this gap is the difficulty in acquiring relevant data, including urban building base maps, building floor heights and mountain data. With the advent of big data, there is potential to solve this issue using open-source databases.

The initial step in conducting this study involves establishing a visual impact assessment (VIA) model. Although the current VIA model still adheres to the fundamental structure of its predecessors, there is room for enhancement and refinement to align it more effectively with the research objectives of this study. The landscape assessment model in this study encompasses four primary components: visual landscape assessment, visual domain analysis, visual magnitude analysis and VIA. This model is designed to comprehensively meet the research objectives of this study.

Visual quality assessment (VQA) originally relied on expert evaluations combined with the principles of landscape garden design. Early expert evaluation systems emerged during the landscape planning movement from the 1950s to the 1960s. This movement encompasses research on the aesthetics of countryside landscapes, as seen in the UK (Crowe 1964) and the United States, such as Litton’s (1968) research on national forest parks. Over time, the field of visual landscape quality assessment has evolved into four distinct schools of thought: expert, cognitive, psychophysical and empirical schools. Within this framework, some scholars have explored the integration of objective data to determine the visual landscape quality over large areas. For instance, Roth assessed the overall aesthetic appeal of Germany by creating a map reflecting Germany’s aesthetic quality, which largely follows the methods developed in earlier works (Michael and Dietwale 2018). Contemporary VQA still largely follows the methodology developed in earlier work. It involves quantitative measurements of visual patches using such tools as the Landscape Index (LM). These measurements provide insight into the spatial complexity and fragmentation of the visual landscape, shedding light on the connections between urban characteristics, development processes, aesthetics and visibility (Alberti and Waddell 2000; Jafarnejad et al. 2015).

The entire geographic area that can be seen from a given point in space is known as the viewshed, a concept first proposed in 1967 by the landscape architect Clifford Tandy (Tandy 1967). The field of view in which an observer can see influences all aspects of their experience of the environment and quantitative analysis of visibility using viewsheds is the most common approach in research in many different fields, including urban planning (Anderson and Rex 2019; Inglis and Vukomanovic 2020), architecture (Rød and van der Meer 2009; Weitkamp 2011), archaeology (Garcia-Moreno 2013; Van et al. 2016) and natural resource management (Chamberlain and Meitner 2013; Depellegrin 2016; Aben et al. 2018). Visual domain analysis has led to various approaches, including the fuzzy visual domain, probabilistic visual domain and cumulative visual domain. Among these, the cumulative visual domain is prominent, because it succinctly summarizes the visual attributes of a set of points within a spatial structure. This approach holds significant promise in the field of landscape visual analysis. As technology continues to advance, more precise raster images can be leveraged to accurately calculate visual domain results within geographic information system (GIS) platforms. However, this process is time consuming. There are two potential solutions: (1) utilize a high-precision digital surface model (DSM) in the primary target area and employ lower-precision maps in the surrounding areas with less obstruction; and (2) enhance computational efficiency by transforming surface structures into vectors and modeling the triangulated irregular network within tools, such as Arc Scene, as suggested by Sepideh et al. (2019).

Visual magnitude is a function that depends on both the visibility of the target and the viewer’s distance. It represents the proportion of the target within the viewer’s field of vision and is closely associated with the level of visual impact. Complex GIS models and software have enhanced the precision of visual magnitude calculations. These include the cone of vision method and spatial openness index models (Dafna and Israel 2006). However, because of the challenges associated with data acquisition and analysis, these methods have not been widely adopted, and some scholars have proposed more concise and effective approaches for calculating the visual magnitude. For example, Integral classified visual saliency into three levels, in which a target is considered ‘obvious’ if it occupies an angle greater than 30° within the viewer’s field of view. Targets with an angle of 5° to 30° are categorized as ‘more obvious’, while those with a visual field of less than 5° fall into the ‘less obvious’ category (Integra 2010).

VIA and VQA share a close relationship, focusing on the repercussions of changes in land use on the visual and aesthetic aspects of landscapes (Palmer 2019). VIA studies encompass a broad spectrum of issues related to landscape transformation, including deforestation, reforestation, urban landmark construction and siting, land-use changes and energy development. In recent years, a considerable amount of scholarly attention has been devoted to examining the visual impacts of renewable energy utilization, particularly wind power, on landscapes. More recently, scholars have focused on the consequences of power facility construction on visual landscapes in the context of renewable energy utilization, with wind power often at the forefront (e.g. de Vries et al. 2012). In such scenarios, mountain landscapes are often viewed as stakeholders affected by these changes, prompting the implementation of VIA. Moreover, advancements in landscape assessment models have enabled the quantification of visual impacts on mountain landscapes. This study aims to extend the existing landscape assessment model beyond quantifying only the visual impacts of the current urban landscape, so that it can be applied to restore the visual impacts of an historical urban landscape and provide a reference for urban planning.

2. Materials and methods

2.1. Study area

The selected area for this case study is situated in the Lishui District in the southern part of Nanjing, China. The study area is determined based on the most recent publicly available national land spatial planning program. It encompasses the urban area, comprising the Tuanshan, Chengzhong and Chengnan areas, in conjunction with the mountainous terrain of the Wuxiangshan area. Together, these areas constitute the study zone, located between latitudes 31°33′ and 31°42′N and longitudes 118°59′E and 119°5′E, covering an approximate total land area of 9740 hectares (Figure 1). Within this study area, Wuxiangshan Mountain, known as ‘the first scenic spot of Lishui’, is situated to the south. However, owing to urbanization, the visibility of the mountain landscape from the old city has significantly diminished. Since 2013, Lishui District has actively promoted new urbanization, including the relocation of the Wuxiangshan Mountain landscape area into the urban fabric as part of broader southern city planning. This move aimed to reintegrate the mountain into the urban environment and make it accessible to the public. Urban development has evolved from a north–south axis to a recent focus on the southern core area, where the mountain plays a central role.

Figure 1. Positioning of the study area.

These urban development dynamics have led to a gradual reduction in visibility of the mountain for urban residents. Conducting a visual landscape analysis of the area provides valuable insights that are applicable not only to this specific area but also to many cities with sprawling development patterns. Understanding the extent of the erosion of mountain views due to urban development is essential for envisioning the future of the study area. The objective of this study is to address these issues and provide valuable insights for urban planning and development.

The building data used in this study were sourced from the Baidu Map open-source database. This database provides information on building locations and building height data, which serve as the basis for determining building heights used in the modeling. Regarding historical building data, while direct access to such data was unavailable, this study employed historical image maps in conjunction with the determined building heights to create a projection for building maps, thus, enabling historical building restoration. The historical image maps for the study area were procured from the U.S. Geological Survey. Based on these maps, this study categorized land use in the study area. All digital elevation models (DEMs) for the study area had an accuracy of 5 m and were obtained from Tuxin GIS (www.tuxingis.com).

2.2. Landscape evaluation models and validation methods

This study emphasized the use of raster data to model the historical restoration of the study area and evaluate the urban hillscape over the past six decades. Although numerous landscape assessment models have been developed in academia, none of the existing models are suitable for historical restoration because of their complexity or the challenges associated with acquiring the necessary raw data. Therefore, this study enhanced existing landscape assessment models to create an improved model that comprises four key analytical components: visual quality, visibility, visual magnitude and visual impact (as depicted in Figure 2).

Figure 2. Landscape evaluation model for assessing visual landscape impacts.

The effectiveness of this landscape evaluation model, particularly in terms of visual quality, visual magnitude and ultimate visual impact results, was assessed by comparing the 2020 GIS evaluation outcomes with real-world conditions observed at selected observation points within the site. To validate the results of the visual magnitude analysis, this study employed KaPPA coefficients, as indicated in Eq. (1)(1) $$\mathrm{\text{Kappa}}=\frac{p_0-p_e}{1-p_e}$$(1) . Then, 100 sampling points within the study area were randomly selected to determine the consistency between the GIS-calculated results and the actual on-site conditions. (1) $$\mathrm{\text{Kappa}}=\frac{p_0-p_e}{1-p_e}$$(1) where $p_0$ is the total number of samples in which the prediction is consistent with the actual value, and $p_e$ is the total number of samples in which the prediction is inconsistent with the actual value.

This study used the scenic beauty estimation (SBE) method to validate the evaluation results for 2020 (Wang and Chen 1999). The VQA and the final results were further evaluated using questionnaires at 30 points chosen from the observation sites. These points were captured by two landscapers using a camera over an 8-month period at various sampling locations within the study area with different landscape characteristics. All photographs captured during this process were geo-tagged in the field. Upon completion, a human perspective photo that best represents the point was selected by an expert as the basis for judging, and landscape professionals and general residents were invited to participate in scoring the photo using Richter’s five-point scale. The results of the questionnaire consisted mainly of a visual quality assessment and a VQA of the mountain. The questionnaire results were validated using Cronbach’s alpha (as specified in Eq. (2)(2) $$a=\frac{n}{n-1}(1-\frac{\sum S_i^2}{S_t^2})$$(2) ). This statistical measure helps assess the internal consistency and reliability of the questionnaire responses. (2) $$a=\frac{n}{n-1}(1-\frac{\sum S_i^2}{S_t^2})$$(2) where $n$ is the number of quiz questions, $S_i^2$ the variance of each subject’s score for each question and $S_t^2$ the variance of the total score obtained by all subjects.

To confirm the validity of the model, a Pearson regression analysis was conducted, as described in Eq. (3)(3) $$r=\frac{{\sum }_{i=1}^n(x_i-\overline{X})(y_i-\overline{Y})}{\sqrt{{\sum }_{i=1}^n(x_i-\overline{X})}\sqrt{{\sum }_{i=1}^n(y_i-Y)}}$$(3) : This analysis combined the outcomes of questionnaire evaluations with the results of objective GIS assessments. Pearson regression analysis is a statistical method used to evaluate the relationship between two sets of data, in which the subjective evaluations from the questionnaire and the objective evaluations are derived from GIS data. (3) $$r=\frac{{\sum }_{i=1}^n(x_i-\overline{X})(y_i-\overline{Y})}{\sqrt{{\sum }_{i=1}^n(x_i-\overline{X})}\sqrt{{\sum }_{i=1}^n(y_i-Y)}}$$(3) where $n$ is the number of key observation points (KOPs) in the current year, and $\overline{X}$ and $\overline{Y}$ denote their sample means.

2.2.1. Visual quality evaluation based on raster datasets

Traditional visual quality evaluation systems often involve an extensive set of indicators, making the process complex and data-intensive. To simplify this approach, some scholars have developed a method for mapping the visual quality of a study area based on raster data, as seen in Marks et al. (1989). This method primarily relies on two indicators: edge density and topographic relief, both of which can be calculated based on the structural parameters available within the raster cells.

In comparison to landscape indexes, this method offers advantages such as easier data acquisition and simplified calculations. Herbst et al. (2009) found that edge density, in particular, is a suitable evaluation index for objectively assessing visual quality.

Adapting this research method, this study divided the study area into 200-m by 200-m test areas, each covering 400 m². This approach enabled the use of the edge effect of habitat structure as a diversity indicator (as outlined in Table 1). Additionally, to differentiate artificial water bodies and human-made landscapes, such as street trees, they were weighted at 0.5 times and corrected by incorporating the terrain undulation index into the final result. This modification enhanced the accuracy and efficiency of obtaining a historical visual quality map of the study area.

Table 1. Edge effect evaluation grading based on raster map.

Edge of vegetation
<100	1
100–300	2
300–600	3
600–1000	4
>1000	5
Edge of water bodies
<50	1
50–150	2
150–300	3
300–600	4
>600	5

2.2.2. Viewshed calculating

Determining the extent of mountain visibility in the study area forms the foundation for modeling visual landscape impact assessments. The U.S. Forest Service recommends exclusively using topographic elevation data to conduct long-term visual impact analyses. To ensure more accurate results, this study utilized a DEM with a 5-m resolution within the study area, in conjunction with urban building data, to create an urban DSM for the study area. After the DSM was established, the visual field of the urban area was delineated using mountain extraction points. The results of all the extraction points were aggregated to determine the cumulative visual field range.

For street trees, this study determined a height of 5 m based on Open Street Map data and field research. In the case of forest-covered areas, such as mountains, this study added 12 m to the digital elevation base map, aligning with the findings of relevant scholars (Palmer 2016).

Based on the results of field research, this study found that the slopes of the mountains with a height of less than 100 m are smaller, there are more buildings, and the ornamental properties are not as strong, thus parts of the mountains with a height of more than 100 m were chosen as the ornamental objects for the study. The target mountain sections Were divided into 50-m by 50-m control grids to establish mountain control points and facilitate subsequent calculations (Zhang and Wang 2022). Following visual domain analysis using the model, this study identified the key viewing points. These were strategically selected from main roads, boulevards, plazas, road intersections, parks, green spaces, shopping malls and other locations where people typically congregate and have a visual range (Yusoff et al. 2014).

2.2.3. Visual magnitude analysis

In recent years, a plethora of methods for evaluating visibility magnitude using GIS and statistical modeling have emerged (Dean and Lizarraga-Blackard 2007). This study opted for a simplified approach that involved calculating the distance from the viewpoints to the number of visible structures (as detailed in Table 2). The maximum visual distance of the human eye is closely related to atmospheric visibility. Based on relevant data, the average visibility in Nanjing is approximately 9.79 km (Wu et al. 2016). By reviewing existing literature, this study identified five distance zones: immediate (<0.02 km), foreground (0.02–0.4 km), middle ground (0.4–2.5 km), near background (2.5–5 km) and background (5–9.79 km). These distance zones help categorize and effectively analyze the visibility magnitude.

Table 2. Classification of visual magnitude based on the number of visible structures per distance zone.

Distance zone	Visual magnitude
	Very high	High	Medium	Low	Very low
Immediate	>1	–	–	–	–
Foreground	>4	3–4	2	1	–
Middle ground	>25	21–25	11–20	5–10	≤4
Near background	>52	41–52	21–40	9–20	≤8
Background	>102	84–102	42–83	18–41	≤17

The total number of visible structures on the mountain was determined based on the findings of the visual magnitude studies. The number of mountain structures that a viewer can observe from ground level serves as an indicator of the proportion of mountains within the field of view. As the viewer’s position changes from a distant to a closer location, the visual magnitude rating of the mountain fluctuates, ranging from very low (1) to very high (5).

2.2.4. Visual impact assessment analysis

Visual magnitude serves as a representation of the direct impact of landscape changes over extended periods, with the inherent change in visual sensitivity captured by visual quality, as suggested by Palmer (2019). By combining these two metrics, this study determined the extent of the visual impact of urban development, as shown in Table 3. Scoring considers factors such as the scale of the visual field and characteristics of the visual landscape, allowing for a more accurate reflection of the visual impact of mountain scenery on viewers in each area.

Table 3. Combining visual quality and visual magnitude for visual impact grading.

Visual quality	Visual magnitude
	Very high (5)	High (4)	Medium (3)	Low (2)	Very low (1)
Very high (5)	5	5	4	3	2
High (4)	5	4	3	2	1
Medium (3)	4	3	2	1	1
Low (2)	3	2	1	1	1
Very low (1)	2	1	1	1	1

Although some scholars argue that using the KOP method for calculation may overlook areas with lower scores in the study area, it is important to note that a sufficient sample size can effectively capture the general visual perception of pedestrians. Additionally, the VIA has the effect of slightly reducing other ratings and expanding the range of ‘very low’ (1) ratings, broadening the scope of assessment for a more comprehensive understanding of the visual impact.

3. Results

3.1. Visual quality analysis

Validation of visual quality calculations using the SBE method and the regression analysis demonstrated the model’s capability to effectively assess the visual impact of the mountains in the study area (α = 0.969, p < .001).

Next, this study sequentially computed and mapped the visual quality of Lishui’s urban area for the years between 1970 and 2020. Figure 3 shows the changes in the visual quality within the study area during this period. In the figure, lighter colors indicate areas with lower visual quality, whereas darker colors indicate areas with higher visual quality. The analysis reveals that the visual quality of the urban area has improved; during the 1970–1990 period, the built-up area of Lishui’s urban area was relatively small and surrounded by villages, farmland, quarries and ponds. This area features minimal topographic undulation and a relatively uniform landscape. Areas with higher visual quality scores were concentrated near Tuanshan Mountain and Wuxiangshan Mountain.

Figure 3. Visual quality assessment map for the study area, 1970–2020.

The continuous development of urban area post-1990 has brought in significant improvements in visual quality. New urban greening efforts have notably enhanced the visual landscape quality of water systems within urban areas. This improvement is particularly pronounced with the establishment of the new Tuanshan Park and the greening of the industrial park in the northern part of Tuanshan. Furthermore, the focus on landscape creation, coupled with the development centered around the southern area of the city, has led to enhancements in the overall visual quality. The construction of new amenities, such as the Xinzhuang Wetland Park, Wuxiang Water Town and Lishui Sports Park, has further contributed to raising the visual quality in the southern area of the city, surpassing that of the central city and Tuanshan area. After 2010 especially, Chengnan area further integrated and developed the existing landscape resources and created a scenic belt around Wuxiangshan Mountain, which substantially increased the landscape score of the Chengnan area relative the other two areas. It should also be noted that the Tuanshan area in the north of the city showed a trend of landscape decline after 2010, which may be due to the negative impact of the development of industrial areas on the landscape of Tuanshan Mountain.

3.2. Visibility analysis

The results of the 2020 visibility analysis were validated through point checks, and the outcomes confirm the high reliability of the visual field analysis model (KaPPa = 0.797, p < .001).

The evolution of the visual area within an urban area over time is presented in Table 4. With the acceleration of urban construction after 2000, the visible area in the urban area steadily decreased. By 2020, the proportion of the visible area in the urban area in relation to the total area had shrunk to 44.32%. This reduction includes 17.35% in the Tuanshan area and 28.25% in the city center area. Compared to these two areas, a lot of new construction has taken place in Chengnan, but it has effectively ensured the viewable area is 84.94% even in 2020. Following the determination of the visible area, this study selected 989 viewing points for all years within this visual area, as presented in Table 5.

Table 4. Results of visibility analysis.

	City area	Tuanshan area	Chengzhong area	Chengnan area
Total area of visibility (km^2)
1970	59.70	15.19	23.20	21.31
1980	58.46	15.02	22.17	21.26
1990	53.44	14.50	17.94	21.00
2000	47.76	12.80	14.29	20.67
2010	34.88	5.79	9.40	19.69
2020	29.43	3.23	7.47	18.74
Incremental increase in visibility (km^2)
1970–1980	−1.25	−0.16	−1.03	−0.05
1980–1990	−5.02	−0.53	−4.23	−0.26
1990–2000	−5.68	−1.70	−3.65	−0.33
2000–2010	−12.88	−7.01	−4.89	−0.98
2010–2020	−5.45	−2.57	−1.93	−0.95
Relative area of visibility (%)
1970	89.72	81.66	89.63	96.61
1980	87.85	80.78	85.65	96.38
1990	80.30	77.96	69.30	95.20
2000	71.77	68.84	55.20	93.69
2010	52.42	31.16	36.30	89.24
2020	44.23	17.35	28.85	84.94
Incremental increase in relative area of visibility (%)
1970–1980	−1.87	−0.88	−3.98	−0.24
1980–1990	−7.54	−2.83	−16.35	−1.18
1990–2000	−8.53	−9.11	−14.10	−1.50
2000–2010	−19.36	−37.68	−18.90	−4.45
2010–2020	−8.19	−13.81	−7.46	−4.30

Table 5. Distribution of viewpoints over the years.

	City area	Tuanshan area	Chengzhong area	Chengnan area
Number of KOPs
1970	76	7	57	12
1980	118	10	61	47
1990	132	22	61	49
2000	167	27	70	70
2010	228	27	84	117
2020	268	26	92	150

Despite the rapid decrease in the overall visible area, most of the obscured areas are situated in undeveloped areas, like cultivated land, with very few obscured areas within the urban zone, except for the central location. When the observation points were initially established, both the Tuanshan and Chengzhong areas experienced an increase in KOPs over the years. However, these viewpoints gradually shifted toward the city’s periphery, and by 1990, the 57 observation points in the city center area had vanished. New observation points emerged in the south and east of the city, coinciding with the development of the urban fringe areas.

In 1970, both the Tuanshan area and the southern part of the city were predominantly undeveloped farmland, and the number of observation points was limited. As the city expanded outward, the number of new viewpoints generally exceeded the number of obscured viewpoints. The Chengnan area witnessed a significant increase in viewpoints after 2000, in response to the southward shift of the urban development center of gravity. By contrast, Tuanshan was the only one of the three areas to exhibit a decrease in the number of viewpoints owing to a considerable reduction in the visual field area.

3.3. Visual magnitude analysis results

The results of the visual magnitude analysis are detailed in Table 6, which illustrates the overall trend of decreasing visual magnitude over time. There was a slight increase in the 1980s and the 2000s, attributed to the repositioning of viewpoints from within the city to the city’s periphery and the substitution of these internal viewpoints with those offering poorer visibility. This trend persists until approximately 2000.

Table 6. Visual magnitude analysis result.

	City area	Tuanshan area	Chengzhong area	Chengnan area
Mean visual magnitude
1970	4.47	3.29	4.51	4.99
1980	4.55	3.40	4.43	4.96
1990	3.70	2.14	3.44	4.71
2000	3.90	2.59	3.49	4.83
2010	3.87	2.00	3.36	4.68
2020	3.75	1.58	3.08	4.53
Incremental increase in mean visual magnitude
1970–1980	0.08	0.11	−0.08	−0.03
1980–1990	−0.85	−1.26	−0.98	−0.24
1990–2000	0.21	0.46	0.04	0.11
2000–2010	−0.03	−0.59	−0.13	−0.15
2010–2020	−0.13	−0.42	−0.28	−0.14

As external urban development continued, especially in the Tuanshan area, which was influenced by construction in the city center and southern area, the average visual magnitude score for its viewpoints plummeted to just 1.58 by 2020, representing the most substantial decrease among the three areas. This suggests that within this area, even if mountain views are visible, their percentage is so small that it is barely noticeable. The higher visual magnitude score for mountain views in the city center area may be because of the broader roads and greater spacing between buildings in newly developed urban areas, which have preserved mountain views. The southern part of the city is almost entirely composed of new urban zones, with development focusing on showcasing mountain views. Consequently, the overall visibility in this area consistently ranks very high.

3.4. Visual impact assessments

Following the calculation of the results using this assessment model, this study validates the evaluation results for 2020. The outcome of this validation process affirmed that the model effectively and accurately reflects the degree of visual impact of urban mountain views (α = 0.969, p < .001).

Table 7 provides an overview of the VIA results for various urban areas over different years. Figures 4 and 5 depict the visual impact scores for all viewpoints in the study area. Notably, the visual impact in the core area to the south of the city has consistently exceeded that of the Tuan Shan area and the old city center since 1980. This suggests that changes in a city’s visual quality exert less influence on visual impact, with the overall VIA results being more heavily influenced by visual magnitude. Only the southern part of the city has seen a steady improvement, owing to an increase in the number of viewpoints. After the 1990s, the number of viewpoints in the southern core area increased from 49 to 150, with the highest increase. The visual magnitude of the new viewpoints and the visual quality ratings are also higher. It is worth noting that while 1970 recorded the highest average visual impact, it was affected by the lower visual quality of the viewpoints, with a maximum score of 4, which was the lowest among all the years assessed.

Figure 4. Results of visual imaging evaluation of KOPs in previous years.

Figure 5. Visual impact assessment rating of KOPs over the years.

Table 7. Visual impact assessment results.

	City area	Tuanshan area	Chengzhong area	Chengnan area
Mean visual impact
1970	2.47	2.29	2.53	2.33
1980	2.42	2.20	2.38	2.53
1990	2.01	1.45	1.85	2.45
2000	2.23	1.63	2.01	2.67
2010	2.30	1.33	1.95	2.77
2020	2.38	1.15	1.85	2.93
Standard deviation of visual impact
1970	1.03	1.38	1.07	0.49
1980	0.99	1.23	1.02	0.91
1990	1.02	1.01	0.96	0.91
2000	1.05	1.15	1.00	0.86
2010	1.13	0.88	1.04	1.00
2020	1.24	0.61	1.09	1.10
Incremental increase in mean visual impact
1970–1980	−0.05	−0.09	−0.15	0.20
1980–1990	−0.42	−0.75	−0.53	−0.08
1990–2000	0.22	0.17	0.16	0.22
2000–2010	0.07	−0.30	−0.06	0.10
2010–2020	0.09	−0.18	−0.10	0.16

The visual impact standard deviation index reflects the degree of dispersion in the visual impact scores across different areas. It can be observed that the standard deviation of the visual impact in the Tuanshan area is decreasing, whereas in the southern part of the city, it is gradually rising. This discrepancy reflects differing trends in the development of visual impact concerning mountain views in these two areas. Despite the addition of new viewpoints during the development process, the visual impact scores remain mostly very low owing to the influence of visual field and distance. However, the new viewpoints in the southern part of the city exhibited more fluctuations, typically ranging between two and five points. This variation is influenced by such factors as visual quality and extent of the visual field.

4. Discussion

The empirical evidence presented by this improved landscape assessment model supports this study’s assessment of historical visual landscape resources. The landscape assessment model proposed in this study not only requires simpler and easier access to experimental materials than those of previous studies but also does not sacrifice the accuracy of the results obtained, which is why this model can be used as a historical landscape resource assessment. Nonetheless, this study considers questions regarding VIAs that warrant further consideration and exploration.

4.1. Extended application of GIS-based visual impact assessment models

At the current stage of research, this study remains grounded in mainstream 2.5D analysis, the advantage of using this method for the study is the possibility of fast calculations from DSM with raster images only. However, during this phase of technological evolution, scholars have already begun to explore the use of drones to create point-cloud datasets on a smaller scale, enabling detailed descriptions of the site’s three-dimensional attributes of the site. This approach differs from traditional 2D or 2.5D analyses and further enhances the accuracy of visual analysis. Additionally, unmanned aerial vehicles (UAVs) provide clear geographic information and enable the scanning of areas with insufficient raw data during fieldwork. Expanding the study area would make it less efficient to use this method, and in addition, local government policies on banning drones would have to be taken into account as well.

When only open-source datasets are used, visual analysis methods continue to undergo updates and refinements. For example, Wang et al. (2023) introduced an innovative tree structure to efficiently extract line-of-sight (LOS) information from Digital Elevation Model (DEM) data. This approach involves quantifying the visual landscape value of a site and subsequently filtering the optimal tour routes by combining them with ecological resistance (Xu and Matsushima 2023). These methods can be improved so that they can be embedded in the models of this study to obtain more targeted results.

4.2. Impacts of different urban development patterns on urban hillscapes

Chinese cities have adopted two typical development patterns: sprawl and compact development. Sprawl development has been the preferred mode for most Chinese cities because it allows rapid urban expansion. The time span of our study enables analysis of changes in the urban hillscape in the context of this urban development pattern. An analysis of the historical evolution of mountain views in urban areas reveals that under this development pattern, the visual impact of mountain views across the city as a whole increase when the city’s boundaries expand even the areas furthest away from the mountain have raised their visual impact scores over the period 1990–2000 (Table 7). This increase is likely because newly constructed roads, plazas and green spaces provide larger viewing areas than those blocked by newly erected buildings (Table 4).

Nevertheless, there appears to be a threshold for positive effects. This threshold signifies the point at which the expansion of the urban boundary results in the visual impact of mountain views within the newly expanded area, which is less significant than the obscured area within the city. Based on the findings, it appears that preserving specific visual corridors and allowing for the free development of the city’s mountain-facing side is the most appropriate way to preserve mountain views throughout the urbanization process.

4.3. Directions for future research and development

This study introduced a model for assessing the visual impact of historical mountains in urban areas and successfully achieved the intended results. However, several issues warrant further investigation. The absence of historical data limited the ability to account for the impact of streetscapes and historical cultural landscapes on visual landscape evaluation. Furthermore, numerous methods for urban visual landscape evaluation exist, such as the entropy power method (Han and Dong 2013) and remote sensing observation (Xi et al. 2022); however, their implementation is often complex and may require experimental materials that are not readily available. Consequently, these methods are not universally applicable and may not be suitable for modeling the visual impacts of mountain landscapes in historical urban areas.

This study’s approach, which compares the urban construction process over the past six decades, provides valuable insights, even without collecting detailed urban zoning or land cover data for the study area. Although the exact impact factors were not determined owing to the methodology used, buildings clearly have the most substantial impact on the visual field. According to the results of the visibility analysis in this study (Table 4), new taller buildings outside the city block views of the inner city more significantly. Urban greenery also significantly influences the visibility of mountain views. There are various approaches for modeling urban greening, including raster image recognition (Xi et al. 2022), 3D modeling of LiDAR point-cloud datasets (Zhang et al. 2021) and software simulation (Zhang and Wang 2022). These methods provide valuable insights into the impact of urban greening on mountain views in the context of urban development and change.

Indeed, the size of the mountain view in different scenarios plays a significant role in the visual impact; however, there are other crucial factors to consider when assessing the visual impact of mountain views in urban areas. For example, some scholars have studied the impact of skyline composition within different numbers of urban view corridors looking at mountains. The results suggest that the highest level of satisfaction among residents occurs when there are three view corridors (Xu et al. 2013).

This study considers only the objective visual impact scenarios of the urban hillscape and does not address the human attributes within the site. Historical and cultural factors play a crucial role in urban development. Areas with strong historical and cultural significance such as temples or heritage sites can have a more profound impact on people, even if only a limited part of the mountain view is visible. These cultural and historical factors can significantly enhance the perceived value of mountain views within urban landscapes.

5. Conclusion

This study improves the existing VIA model that it can be applied to the VIA of urban historical hillscapes, and quantitatively analyzing the visual impact of urban hillscapes in the study area over 50 years.

The results show that VIA, as a common method of visual landscape analysis, can also achieve corresponding results when applied to urban hillscapes, and that the integration of such factors as viewable area, visual magnitude and urban visual landscape, can restore the visual impacts of historic mountain views in the study area, makeing the evaluation more comprehensive and the results more representative. After analyzing and studying the Lishui urban area of Nanjing through this model, it is found that the visual quality score of the area increased as the city continued to develop outward, but the overall viewshed area and visual magnitude of the area decreased by 45.49% and 16.1%, respectively, due to the shading of the new buildings. The final visual impact scores of the mountains in the area increased by 25.7% from 2.33 to 2.93, except for Chengnan area, where the visual impact scores of Tuanshan area and Chengzhong area decreased by 48% from 2.29 to 1.15 and 26.9% from 2.53 to 1.85. By quantitatively analyzing the urban hillscape through this model, it can be more intuitively showen that the areas with high visual impacts of the hills in the city can help to provide effective references in the process of urban planning.

Finally, this study can be expanded in several ways, such as by adding the impact of different urban functional zones, which not only represent the value of the mountain view in the area but also the number of viewers. The relationship between the degree of urban boundary expansion and the addition of new mountain view visual impact zones can also be discussed to facilitate the development of better strategies for urban expansion to preserve the natural urban landscape.

Authors’ contributions

Conceptualization and methodology, Y.X. and H.X.; software, Y.X. and W.L.; data analysis, Y.X., Y.Z and X.B.; writing-original draft preparation, Y.X.; writing-review and editing, Y.X. and H.X.; supervision and project administration, H.X. All authors have read and agreed to the published version of the manuscript.

Data availability statement

The data presented in this study are available on request from the corresponding author.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This research was funded by the Priority Academic Program Development of Jiangsu Higher Education Institutions (PAPD) and a project funded by the Jiangsu Postgraduate Research and Practice Innovation Program project, grant number (KYCX23_1203).