Correlations among AHP-based scenic beauty estimation and water quality indicators of typical urban constructed WQT wetland park landscaping

With the increasing demand for advanced sewage treatment in cities, urban constructed water quality treatment (WQT) wetlands are widely used to treat lightly polluted waterbodies as one component of nature-based solutions (NBS), which are often effective in improving water quality indicators (WQIs). The pollutant removal rates of constructed WQT wetlands are generally higher than other treatment measures when treating sewage with high-level (80–99%) pollutant concentrations (Girts et al. 2012). Once more attention is given to scenic beauty in the WQT wetlands, it will help to make more efficient use of limited spaces in the city and re-design more spaces for biophilic aesthetic experiences. Constructed WQT wetlands have been extensively utilised in urban built environments due to their impressive geo-ecological and landscape characteristics in eco-engineered landscaping (EEL) projects (Xie et al. 2009; Addo-Bankas et al. 2022; Li et al. 2022a). Therefore, the growth, use, and changes of urban constructed wetlands should be necessarily considered for wetland protection and management in the urban built environment (Geng et al. 2023).

Scenic beauty often relates to cultural, biological, and individual dimensions. Scenic beauty estimation (SBE) has been studied widely for landscape perceptions, which have been conducted on biological, cultural, or social behaviour (Bourassa 1991; Wang et al. 2020; Zhou et al. 2023). Since semi-natural elements have been regarded as one of the significant effectors for optimistic prediction, SBE has often been judged by vegetation and physical and human geographical conditions (Junge et al. 2009). Related studies on SBE have primarily occurred in geography and forestry. SBE judging has been commonly utilised to quantify the scenic beauty index (SBI) voting, while effectors of the SBI vary in existing studies.

Estimations of SBIs can help to achieve a refined landscape design and resource management for the aquatic environments. The SBI is one of the common indices in evaluating geo-sites and villages for conservation (Long et al. 2023), heritage protection (Tessema et al. 2021), and ecological tourism (Jia et al. 2022). The SBE has been used to assess landscape resources in agroecosystems, in which landscape values were found to be affected by conditions of vegetation (van Zanten et al. 2016), land cover structure, vegetation, livestock, and historic architecture (van Zanten et al. 2014). Forestry-related studies indicated that the proportion of green space, canopy coverage, colours, morphology, and the combination of vegetation and the wilderness of the landscape can affect the qualities of forest landscapes (Yang 2014). SBIs have dynamic effects during the process, particularly in the spatial and temporal patterns of change perceived by human beings (Pierskalla et al. 2016). Such understanding enhances the ability to comprehend the real-time experience of the landscape. These over-time dynamic effects of SBIs were also empirically found in a study on forest landscape in a national forest park (Deng et al. 2014).

SBIs can also be integrated with Geographic Information System (GIS) techniques and environmental psychological studies. An SBE-based study indicated that the beauty of urban forests could impact tourists' behaviour (Pierskalla et al. 2016). Based on the expert-based SBI voting and remote sensing (RS) technology, landscape beauty patterns of the surroundings of Alps Mountain have been studied, demonstrating that the complexity of landscape morphology and diversity is positively correlated with visual qualities (Schirpke et al. 2013b). A simple SBI scoring method was proposed to analyse the landscape beauty of semblable forests (Hull & Buhyoff 1983). As for urban landscape open space units, the ‘sequence of penetration’, ‘scale’, and ‘naturalness’ have positive impacts on citizens' scenic beauty preference, whereas ‘complexity’ and ‘rhythm’ are negative (Wang et al. 2020).

Nevertheless, the SBI scoring method has rarely been adapted to evaluate design schemes of urban constructed WQT wetlands. Landscape preferences and physical geographical conditions can affect the aesthetic appreciation of landscapes. Other landscape-oriented factors, i.e., on-site scenery, cultural and educational value, and geoscience knowledge, are often not concerned. Relationships between wetlands' scenic beauty, diverse landscape perceptions, and WQT effects have not been empirically investigated thoroughly. SBIs of constructed WQT wetlands should be further studied. WQIs–SBIs correlations should also be analysed to assist the landscape design of urban constructed WQT wetlands, because the related field is still much more to learn for geographers, landscape architects, and environmental engineers.

The mathematical modelling of SBIs should be applied to landscape dynamics evaluation, which is beneficial to guide decision-making in landscape design (Grêt-Regamey et al. 2007). To evaluate the beauty quantitatively by the SBE method, it is also necessary to integrate the methodology of expert ratings and volunteer SBI voting for practical experiments. The analytic hierarchy process (AHP) method has been widely used as an evaluation algorithm in the multi-criteria decision-making (MCDM) method in the recent decade. However, there have been few studies on a direct comprehensive evaluation of WQT wetland landscape resources based on the SBE.

The details of two wetland parks in this study are introduced as follows:

WQI sampling at each sampling site (mentioned in Tables 1 and 2) was conducted in situ on sunny days in June 2021, during which the temperature of all sites varied between 24.5 and 29.5 °C. All samples were taken at 10–20 cm depth underneath water surfaces. DO was determined in situ by the electrochemical probe of 86031 AZ IP67 Combo Water Meter (manufactured by AZ Instrument Corp., P. R. China). The LOHAND water test field kit (manufactured by LOHAND Corp., P. R. China) measured levels of COD_cr, NH₃-N, and TP. An existing study (Li et al. 2022a) indicated that there was no obvious accuracy difference between measurement methods of in situ monitoring and laboratory methods; therefore, it was appropriate to utilise the water test field kit. Experiments in situ were performed in triplicate; collected data are presented as the mean ± standard deviation, which will be demonstrated in the later section. Simultaneously, the buoy method assessed water flow velocities at each sampling site. The average water velocity was determined by measuring the time per floating buoy cost to pass through a fixed distance of two points above the waterbodies.

Numerous studies (Hull & Stewart 1992; Schirpke et al. 2013a; Deng et al. 2014) have indicated photographs' reliability as an optimal medium for SBE. The photographs were taken with a Canon SLR digital camera to capture the scenic beauty features of sites and landscape resources. The following principles were applied during selecting photographs:

• (1)

  Photographs were taken from proper perspectives at the eye level (at 1.7 m above ground).

• (2)

  No non-landscape elements exist in photographs, i.e., people, vehicles, and animals.

• (3)

  The same shooting conditions were always maintained, while weather conditions were acceptable during the field research. No addictive image processing has been conducted.

As a common decision analysis technique, the AHP was put forward and developed by operation research expert T. L. Saaty. The AHP method has been commonly used to tackle complex, unstructured decision-making problems involving numerous levels, elements, objectives, and criteria (Saaty 1980). The AHP has emerged as one of the most frequently used mathematical tools in studies related to quantitative geography since the 1980s and has been applied as one of the standard methods for evaluating the scenic values of the landscape (Schirpke et al. 2013a; Li & Li 2021; Jia et al. 2022; Li et al. 2022b). The AHP-based decision process consists of four steps, i.e., defining aims, structuring decision hierarchies, constructing pairwise comparison matrices, and adapting the priorities, from comparisons to weighing priorities below the immediate level (Saaty 1980). The AHP-based decision-making process can be determined in terms of the number of attributes, criteria, and alternatives (Nowak 2010). The AHP model is comprised of three-level hierarchical structures, i.e., goal (objective), criteria, sub-criteria, and alternatives.

Considering the wide range of indicators available in the relevant literature and the design expertise of WQT wetlands and urban waterscapes in Southern China (Xie et al. 2009; Li & Li 2021; Zhang et al. 2021; Li et al. 2022a), the SBI model in the research considers three criteria in the criterion layer (Conditions of vegetation (B1), Physical geographical conditions (B2), and Human geographical conditions (B3)) and their eight sub-criteria in the first-level indicator layer (Density (C1), Combination (C2), Form of vegetation (C3), Micro-topology (C4), Form of waterbodies (C5), Openness of waterbodies (C6), Human geographical uniqueness (C7), Value for natural education (C8)), which are primary factors being believed to influence SBIs for WQT wetland landscaping. The ‘openness of water bodies’ refers to the degree of visual openness that would be perceived by a human being when viewing waterbodies in 180° viewsheds (Zhang et al. 2021). The detailed meanings of each criterion are listed in Table 3.

Five voluntary experts with experiences in landscape architecture, geography, and ecology are invited for the face-to-face ranking process. Experts are invited to vote for the weights of each factor and these weights will be applied in calculating the SBI voting by judges, as mentioned in Section 3.3. Ranking of the criteria and sub-criteria based on Saaty's 1- to 9-point scale (Saaty 1980), which contains nine intensities, includes five primary levels of intensity (1 = equal importance, 3 = moderate importance, 5 = vital importance, 7 = very vital importance, 9 = extreme importance) and four intermediate levels (2 = weak, 4 = moderate plus, 6 = strong plus, 8 = extreme plus). The class with the most suited and potential in each factor map was given the most weight and vice versa.

A statistic Python programme developed with ‘Scikit-learn’ and ‘Matplotlib’ packages was used for ranking criteria and sub-criteria, during which the most potent factors are given the most significant weight at each level. In contrast, the least ones receive the least weight. The priority value of the sub-criterion was calculated by the arithmetic average of the response of experts, and their answers were ranked in matrices. After the priority value in the criterion layer is calculated, an inverse linear matrix is created. To assess whether the pairwise comparisons were consistent, each weight of factors in the ‘criterion layer’ (W_i) and the CR of all data (Table 4) were also calculated by the statistical programme, according to Saaty (1990). A CR of 0.08 for four compared elements and a CR of 0.05 for three compared features were satisfied, ensuring the reasonable level of consistency of the AHP model (Saaty 1995).

The CR of the first layer (CR = 0.048) suggested acceptable consistency for weight analysis and was appropriate for identification. CRs of the pairwise comparison matrix are calculated. The weights of each criterion are sufficiently consistent; thus, other re-evaluation processes are not required (CR < 0.05). The normalised weight matrix of each criterion includes ones for the criterion layer (W_Bi) and first-level indicator (W_Ci), which have been further set (Table 5). It will be adapted in the SBI and calculated in later sections.

SBIs of each site are listed in Table 6.

At Yangbei Lake Wetland, Site No. 1 achieved the highest mean SBI (6.2252) due to its dense and abundant vegetation along the lake, highly naturalised waterbodies, and a wooden bridge. In contrast, Site No. 2 reached the lowest mean SBI (4.9352) because of its highly modified landscapes, linear embankments, fewer types and densities of vegetation, and a stone bridge. For Shuijing Park, Site No. 3 reached the highest level of well-managed vegetation along buffers, wooden boardwalks, and pavilions. However, a subsurface flow wetland at Site No. 1 has the lowest SBI.

The statistical paired t-test proceeded to conduct differences between SBI voting by a group of ‘Non-professionals’ and ‘Professionals’ (Table 7). The accuracy and stability of the group of ‘Professionals’ and the group of ‘Non-professionals’ were adjacent to each other (paired mean difference = −0.12; p = 0.052 > 0.05), and no deviation was observed. In general, SBI voted by the group of ‘Professionals’ was only slightly higher than that of ‘Non-professionals’.

The aesthetic values of ecological landscape should be considered by landscape architects and engineers when designing constructed WQT wetlands. To determine the contributions of multi-effectors to SBIs, the mean SBIs of each criterion layer are shown in Table 8. Distinctive properties of different criteria have been indicated at other sites, which will be analysed in detail later.

The constructed WQT wetland is one of the most effective media to purify domestic and industrial wastewater. Following the design process of eco-landscaping engineering as ‘design–management–monitoring–evaluation–adjustment’ (Xiang et al. 2022), first-hand in situ WQT monitoring is essential for comprehending WQI–SBI correlation to summarise design strategies. In situ monitoring of each WQI per site was conducted in June 2021 (Table 9). Detailed methods have been introduced in Section 2.3. It is indicated that WQIs were improved after successive purification processes of WQT wetlands. However, slightly adverse purification effects for COD_Cr have existed.

To assess the WQT effects of each purification stage in constructed WQT wetland systems, the removal (or increase) contributions (%) of WQIs of each water quality purification stage have also been computed, respectively (Table 10).

To figure out relations among WQIs for indicating pollutant removal contributions (i.e., DO, COD_Cr, NH₃-N, and TP) and mean SBIs, correlation analysis has been processed by Python 3.0 with the ‘NumPy’ package. Firstly, Pearson correlation test analysis (Table 11) proceeded.

DO increase, COD_Cr removal, and NH₃-N removal can explain 68.8% of the changes in the SBI with R² = 0.688 (F = 2.936, p = 0.163). These phenomena were listed as follows.

A significant positive correlation between the DO increase rate and the SBI is found in both Yangbei Lake and Shuijing Park. DO levels related to variations in the intensity of nitrification and denitrification in substrates. FWIs at site Shuijing No. 3 were planted with more amounts and types of aquatic vegetation than other sites, while high different water levels are led by weirs, which contribute to the highest DO increase rate and the highest SBI. Other sites presented less DO increase contributions.

In addition, a relatively high positive correlation between COD_Cr and NH₃-N removal was indicated. COD_Cr in the constructed WQT wetland was believed to be removed by the absorption and bio-metabolic degradation effects of plants' roots. At the same time, substrates and microorganisms grown on surfaces played a vital role in NH₃-N removal (Cao et al. 2020). Noticeable results for COD_Cr and NH₃-N have been shown at specific sites. The most apparent NH₃-N reduction was observed at site Yangbei No. 3 (51.11%), at which SBI was also the highest. Meanwhile, a VF wetland at Shuijing No.1 attained the lowest SBI (4.3967). At the same time, improvements in WQIs (24.75% of COD_Cr and 24.32% of NH₃-N removal) have occurred due to the lack of diverse vegetation, highly homogenised physical geographical conditions, as well as insufficiencies of human geographical elements. In addition, Site Yangbei No. 1 (SP) and Shuijing No. 3 (FWI) present negative removal contributions of NH₃-N, which might be influenced by uncertain external pollution, although they obtained a higher SBI. TP removal was believed to be accomplished by adsorption, complexion, and precipitation by plants and the biochemical action of microorganisms (Ebrahimi et al. 2022). However, no obvious TP–SBI correlations have been indicated (R= − 0.0278, p < 0.01) on account of in situ conditions during the field research.

Two aspects are commonly admitted for integrations of scenic beauty and ecological effects. On the one hand, knowledge of environmental geography plays a role in moderating beauty cognitions to be ecologically aware. On the other hand, it was believed that landscape aesthetic experience was a psychological process with perceptual, affective, and cognitive characteristics, which were often obtained from patterns of waterbodies, vegetation, and micro-topologies (Gobster et al. 2007). Phenomena and rules can be indicated from SBI voting as follows.

Plants were often beautifully shaped for sites with higher scores, while dense vegetation promoted higher green-looking ratios. The high density of arbours contributed more to SBIs. Scenes with multi-layered vegetation (e.g., Yangbei No. 1, SBI = 6.2252) were more favoured by participants than those with a medium density of shrubs (e.g., Yangbei No. 2, SBI = 5.8146) or floating-leaf plants (e.g., Yangbei No. 3, SBI = 5.7278). Based on these phenomena, arbours, shrubs, herbaceous plants, and floating vegetations should be arranged to maintain proper vegetation structures. Sites planted with homotypic deciduous vegetations were often attributed to lower SBIs, while sites vegetated with greater amounts of evergreen perennial aquatic plants reached higher SBIs. Neat Cinnamomum camphora groves planted further up the bank as the backview woods can lead to emotions of fatigue or dullness (e.g., Shuijing No. 2, SBI = 5.0473). Also, substrate beds planted with high-density and single-typed Cyperus involucratus Rottboll were the most unattractive scene to participants (SBI = 4.3967). Thus, the landscape design of SSF wetlands should focus on improving varieties and integrations of vegetation.

Sites with semi-natural geo-features were often rated higher SBIs, which is influenced by micro-topology and the form of the waterbody. Sites with slope revetments and well-vegetated buffers were often more ornamentally valuable than those with linear embankments (e.g., Yangbei No. 3 and 4). Sites with vegetated wavy micro-topologies were often more attractive than the others (e.g., Shuijing No. 2 and 4), more multi-layered spaces could be created by setting appropriate micro-topologies and wetland EEL measures can be formed delicately. Sites with high degrees of waterbody openness were rated higher for SBIs. Waterscapes with a higher degree of openness demonstrated a higher level of potential than those of lower openness. This phenomenon can be explained by psychological perceptions of human beings. Empirical research (Völker & Kistemann 2013) demonstrated that open waterscapes can improve citizens' mental health, probably because waterscapes were often regarded as clean, peaceful, and worth preserving. Well-designed waterscapes are a primary factor contributing to the improved restorative potential of the urban environment. Also, shape, flow velocity, and transparency of water can affect people's perceptions of aesthetics in semi-natural environments (Yamashita 2002). Semi-natural waterscapes were supported for rendering visitors' attention. Waterscapes at those sites of higher waterbody openness are beneficial to improve participators' attention restoration into senses of relaxation, immersion, and calmness (Kaplan 1995). However, those with vast and empty water were often non-attractive (e.g., the sediment pond at Yangbei No. 4), which should be avoided in practical design as much as possible.

Sites with human geographical elements have more potential for natural education, which is often attributed to higher SBIs. As for Yangbei Lake Wetlands, the bridge with ornamental boats (Site No. 1) and waterwheels (Site No. 2) attributed to higher human geographical characteristics, which highlighted features of hydraulic landscapes. As for Shuijing Park, boardwalks and pavilions (Site No. 3) and a wooden hut at the OP (Site No. 4) supported cultural aesthetics. All sites in both wetlands owned specific potentials for activities of natural education, especially for presentations of water purification processes. However, no related appliances or signals have been set in situ.

Humans both perceive and experience landscapes, which demonstrate an intertwined relationship. In urban environments, people's perceptions of landscapes are often correlated to a sense of ‘place attachment’ and their ‘perceived restoration’ (Chen et al. 2014; Goetcheus et al. 2016). The perception realm theory (PRT) by Gobster et al. (2007) suggests that humans often cannot perceive large-scale ecological phenomena that occur in natural ecosystems. Humans are limited to perceiving small-scale cognitive landscape characters and patterns. Therefore, EEL projects should focus on human's landscape aesthetic experience at a specific scale, which would encompass broader ecological processes. EEL measures can be integrated into landscaping, i.e., re-designing buffer zones, constructing green infrastructures, and re-arranging vegetation structures (Li et al. 2022b). Also, waterscapes in different sections of WQT wetlands are recommended to be designed for higher cognitions. Wetland cells should be instructed to design in shapes with length/width ratios of less than 4:1 for treatment purposes. Curvilinear shapes that follow existing topographical contours should be used. Also, WQT wetlands should be designed to balance the friendly biodiversity and the WQT benefits (Addo-Bankas et al. 2022; Liu et al. 2023). Overall, when planning and designing WQT wetlands, it is crucial to incorporate both aesthetic design factors and technical engineering considerations of EEL (Huang et al. 2022). Embankments, shoals, ponds, and pools can be innovated in practical design for purifying waterbodies and reducing flow rates.

Constructed WQT wetland EEL integrates a highly complex WQT system, which normally consists of vegetation, microorganisms, and substrates. Among them, microorganisms played a vital role in improving water qualities, whereas substrates' WQT capabilities often relate to vegetation and physical geographical conditions. Suggestions for further improvements of WQT techniques are advocated as follows. For the two wetlands mentioned in the research, EEL measures were indicated to be vital for arousing visitors' biophilic attention to the physical geographical phenomena and EEL principles of WQT processes. Both physical and human geographical conditions should be optimised, especially for SPs and VFs. Node design of the SP and VF wetlands was found to be omitted by landscape architects. By adapting waterbodies with higher openness, greater amounts of natural-form micro-topologies at buffers can contribute to more significant WQI improvements and SBIs. Vegetation can remove multiple pollutants by assimilation and oxidation, e.g., Phragmites australis and Canna indica L. are effective for COD_Cr and NH₃-N removal. Increasing vegetation density and planting more floating-leaf vegetation at APs and FSF wetlands are recommended. Otherwise, the constructions of siltation ponds are not appropriate in urban environments, as they are inclined to be polluted significantly.

Exposure to semi-natural environments may enhance attention performance through two distinct mechanisms: the modulation of alertness and the connection between humans and nature (Johnson et al. 2022). According to the attention restoration theory (ART) suggested by Kaplan (1995), four types of restorative natural environments, including ‘Being away’, ‘Fascination’, ‘Compatibility’, and ‘Extent’, can be attributed to semi-natural wetland waterscapes. Waterscapes with higher naturalness levels usually demonstrate a better effect of participating in restoration than those of lower naturalness (Yamashita 2002; Nassauer & Faust 2013). Among them, ‘being away’ means to be distant from highly modified artificial landscape elements. Sites with characteristics of ‘fascination’ are often arranged with abundant landscape elements. Sites with spaces for developing diverse educational and recreational activities are regarded with high ‘compatibility’. High and medium naturalness can create restorative semi-natural environments of ‘being away’ and ‘extent’ (Kaplan 1995). In practical wetland design, waterscapes of higher naturalness and openness with semi-natural micro-topologies can significantly improve WQTs and scenic beauty (Li et al. 2022a). Therefore, wetland landscapes should be designed for better attention to restorative achievements by designing waterscapes with naturalised characters, proper vegetation, and abundant landscape elements, which was mentioned by Sevenant & Antrop (2009). Also, the urban constructed wetlands are important for the management of the urban water environment and the promotion of the water-sensitive urban design (WSUD) because they are resilient to the challenges of climate change and were often the surrounding places where citizens potentially want to live. Urban constructed WQT wetlands designed and constructed with integrated urban water management solutions were often developed in the context of social and aesthetic values.

So far, two main methods have been widely adopted in the SBE, i.e., expert-based evaluations (Sowińska-Świerkosz & Chmielewski 2016) and the perception-based method (Peng & Han 2018). In terms of cost and time, expert-based methods outperformed perception-based methods. In the study, we found that it was easier to verify the reliability and validity of perception-based methods with AHP voting than expert-based methods. Further improvements in the following can be adapted in the experiment.

• (1)

  The SBE process can be disturbed by weather conditions and participants' viewpoints of the scenery. If states were permitted, panoramic photographs or virtual reality (VR) presentations could be adapted for the SBI judging.

• (2)

  Apart from the visual sense, other senses, i.e., hearing, smell, and touch, can also affect the perception of the voluntary judges. Soundscapes have intensive relationships with human perception. Fountains with waterfalls or jects can produce acoustic aversions (Patón et al. 2020). Also, integrating the soundscape with the visual landscape in urban outdoor spaces can increase the aesthetic attractiveness for citizens and tourists. Therefore, multi-sensory perception and the related SBE scoring should further proceed to improve the credibility of experiments.

• (3)

  Some changes of hydrological and vegetation conditions in urban WQT wetland parks due to temperature may affect visitors' visual perceptions and the results of the SBI voting for landscapes (Ma et al. 2023). The experiment in this study measured water temperature between 24.5 and 29.5 °C. It remains uncertain whether the SBI–WQI relationships found in the study remain valid for colder water temperatures (0–10 °C), and further studies are required.

This research focuses on the scenic beauty values of the EEL measures for the design and management of WQT wetlands. During processes of the quantitative AHP-based SBE, with investigations by experts and SBI voting by participators, SBIs at each site have been obtained. WQI–SBI correlations have been summarised. DO increase, COD_Cr, and NH₃-N removal contributions are significantly related to SBIs, while no obvious TP–SBI correlations have been indicated. Adapting SBIs to constructed WQT wetlands can offer proper guidance for both landscape architects and engineers. EEL strategies for wetlands have been concluded, i.e., following strategies of EEL, improving WQT techniques, and applying ART in landscape design. Adjusted vegetation and physical and human geographical conditions will enable the creation of sustainable landscapes that promote eco-tourism and natural education. A deeper understanding of the SBIs will enhance the practical design, management, and maintenance of urban WQT wetlands.

Y.H. and T.L. wrote the main manuscript text, Y.J. and T.L. prepared some data, and W.W. reviewed the paper and offered some advice. All authors reviewed the manuscript.

The research was supported by the National Natural Science Foundation of China (Project #51208467) and the Scientific Research Foundation for the High-level Talent Introduction, Zhejiang University of Technology (Project #118001929).

This study recruited some voluntary participants from the campus of the Zhejiang University of Technology. Free and informed consent of the participants or their legal representatives was obtained. All the research protocol of the study was approved by the Ethics Committee of Zhejiang University of Technology (Zhejiang Province, P. R. China). All participants gave their verbal informed consent before participating in the experiment. The authors would like to express their sincere gratitude to them.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.