Exploring Biodiversity Through Citizen Science: A Case Study of Green Roofs at the Calouste Gulbenkian Foundation Garden in Lisbon

Highlights

-

Comparison of biodiversity data in green roofs and garden areas

-

Citizen Science proves effective to collect large-scale biodiversity data

-

Spatial observations show that combining gardens and green roofs provides continuous ecosystems

-

Structured activities increase taxonomic diversity and monitoring in less-visited areas.

-

Birds and insects observed in gardens and green roofs show a positive correlation

Abstract

Citizen science is rising and expanding as an approach to data collection, enabling the acquisition of data through the voluntary involvement of citizens in scientific activities. This study explores the effectiveness of citizen science in collecting biodiversity data in urban green infrastructure, focusing on a case study at the Calouste Gulbenkian Foundation in Lisbon, which includes both green roofs and traditional garden areas. Data collected via the iNaturalist platform through structured and unstructured citizen science activities were analyzed to compare patterns of biodiversity observation. Results show that unstructured activities attract more participants but produce fewer observations per person, mainly focusing on more familiar taxa, such as birds. In contrast, structured events concede a higher number of observations per observer, including less commonly recorded taxa like insects, and provide greater coverage of green roofs, since routes are predefined. Seasonal and temporal trends were also noted, with a higher concentration of observations in spring and summer and a significant increase on weekends, indicating the influence of participants’ availability. Spatial observations show that combining gardens and green roofs provides continuous and rich ecosystems that are crucial for city urban planners.

1. Introduction

According to the UN, in 2022, the global population surpassed eight billion people, highlighting the significant growth observed worldwide in recent decades. Moreover, it is estimated that approximately 55% of the current population resides in urban areas, with this percentage projected to increase to 68% by the year 2050 [1]. Thus, the evolution and development of urban areas are highly influenced by this significant population growth. Projections indicate that, by 2030, urban areas will expand by 1.5 million km², which is three times the urban area that existed worldwide in the year 2000 [2]. The increasing demand for urban infrastructure has been resulting in the disorderly expansion of built and paved areas [3], globally causing a gradual disappearance of green spaces in cities [4], which has negative consequences for communities, since the benefits they provide are vast and widely recognized. These spaces have a beneficial effect on the physical and mental well-being of residents, playing a crucial role in the social, cultural, and psychological needs of people, enabling the practice of physical activities and improving the aesthetic value of cities [2,5]. Various studies also highlight the positive role of green areas in the urban environment, including the improvement of air quality, the regulation of temperature, and a reduction in noise pollution levels [2,5].

The awareness of society is growing, and there is a raising tendency to preserve and, whenever possible, promote urban green areas. One of the most common approaches is the introduction of nature-based solutions (NBSs) in planning measures [6,7]. These solutions use natural ecosystems as a support and aim to assist communities in addressing environmental sustainability, providing environmental, social, and economic benefits [6,8,9]. NBSs come in various forms, such as parks, rain gardens, street trees, urban drainage systems, green walls, and green roofs [2,7,10]. Green roofs emerge as one of the most important forms of NBSs [4,11,12], since buildings occupy a significant area in cities [2], and green roofs can help adapt and mitigate the adverse effects of excessive urbanization [13,14]. Since the use of green roofs has increased in urban areas, it is crucial to promote research on their impact to better understand their benefits and limitations.

Green roofs are designed to allow the development of vegetation over built spaces, with maintenance requirements comparable to those found in traditional soil [15,16]. In the literature, this construction technique is referred to by various names, such as eco-roofs, living roofs, vegetated roofs, and roof gardens [4,12,13]. Green roofs are defined whenever vegetation is installed on top of the roof. The roof can be located at street level, e.g., when on top of a parking lot, and in that case, it can be confused with green areas, but it is still a green roof. Green roofs can be classified as extensive, intensive, or semi-intensive [4,13], depending on the properties, type of vegetation, and purpose of the roofs [16].

Green roofs offer various economic and socio-environmental benefits that can be analyzed from different perspectives. At the building scale, the green roof can contribute to reduced energy consumption, improved sound and thermal insulation, reduced fire risk, increased efficiency of photovoltaic panels, financial valorization of the structure, and increased lifespan of its elements, which are mentioned in the literature [13,17,18]. At the urban scale, one can expect reductions in the urban heat island effect, improvements in stormwater management and water quality, reductions in urban noise pollution, improvements in air quality, the enhancement of physical and mental health of inhabitants, and the conservation of ecosystems and biodiversity preservation [4,12,19].

The green roofs are adaptive solutions [10], also serving as a mean to restore part of nature to cities, so they can act as an ecological refuge, providing food and habitats and creating conditions for biodiversity to increase [16,20]. However, it is essential to understand the requirements of local fauna and flora species to develop effective green roof projects that actively contribute to ecosystem recovery [21] and to biodiversity management and preservation [22]. Different authors have called for more studies to fill gaps in understanding the true impact of green roofs on biodiversity recovery in urban areas [23]. One possible explanation for this gap is that the costs associated with data collection by professionals might be high and time-consuming for green roofs, mainly because they are typically distributed over the cities. Citizen science (CS) can help overcome this problem [23]. Data from surveys conducted by CS represent one of the most significant sources of information on biodiversity distribution [24].

The present study seeks to contribute to research related to green roofs in urban environments and their impact on the preservation and development of biodiversity and ecosystems in urban areas. Although it is expected that green roofs can help in enhancing urban biodiversity, there are few studies that directly compared green roofs and urban garden biodiversity, as well as the positive input of citizen science activities to this problem [25].

The chosen method of information gathering was CS, using the worldwide iNaturalist platform as a tool. This platform allows for monitoring biodiversity and recording observations through the participation of users. The data are made available to all. The platform iNaturalist can be used either in structured or non-structured monitoring. Structured activities follow a protocol attempting to record all general observations or those of a given taxonomic group. Activity structures do not correspond to case-by-case records made spontaneously by people, so it is more difficult to ensure that all species are recorded. Structured CS events include BioBlitzes that are activities that join citizens, researchers, and naturalists for a certain period of time in a limited space area to register the biodiversity presented there.

Given the growing importance of urban green areas and the increasing implementation of nature-based solutions, such as green roofs, it becomes crucial to understand their actual ecological impact and the role of CS in supporting this understanding. Despite the acknowledged benefits of green roofs, their contribution to urban biodiversity remains insufficiently explored. This study addresses the following research questions: To what extent do green roofs support urban biodiversity compared to traditional ground-level gardens? How effective is CS, particularly through structured and unstructured iNaturalist activities, in collecting biodiversity data on green roofs? What are the limitations and potentials of using CS data to inform biodiversity-related planning in urban green infrastructure?

To address the research questions, this study sets out with the following objectives: (1) to evaluate and compare biodiversity levels observed in green roofs and traditional garden areas within the Calouste Gulbenkian Foundation in Lisbon, which is committed to environmental sustainability and focused on the implementation of green infrastructure [26]; (2) to assess the effectiveness of citizen science as a tool for biodiversity monitoring in urban green roofs; and (3) to analyze the potential and limitations of using data collected through the iNaturalist platform—both from structured and unstructured activities—as a support for urban ecological planning.

The paper is organized as follows: Section 2 outlines the general methodology used, explaining how the case study was characterized and how the biodiversity data were collected and analyzed. Following this, Section 3 presents the case study, explaining the different characteristics of the Calouste Gulbenkian Foundation and specifying which areas of this space are green roofs and which areas are traditional gardens. Section 4 details the methodology used to collect and treat the biodiversity data, namely how the iNaturalist application works, what is a structured CS event, and how to optimize data treatment. Section 5 presents and discusses the results, comparing the data obtained with structured and unstructured CS and in green roof areas and traditional garden areas. Finally, Section 6 discusses the conclusions reached by this study.

2. Research Approach

The Calouste Gulbenkian Foundation was selected as the case study, since it presents unique characteristics both in terms of the built spaces (existence of gardens and green roofs with and without continuity) and available information, namely the existence of a multi-year series of biodiversity observation records using CS. The flowchart in Figure 1 details, schematically, the methodology adopted in the study.

The detailed description of the case study was done by consulting various documents provided by the Calouste Gulbenkian Foundation (e.g., building drawings) complemented with visits to the site. This provided a characterization of the space in terms of the services offered, uses by the public, and features of its green spaces. Information regarding the area, location, typology, accessibility, and type of vegetation was retrieved to categorize both garden areas and green roofs.

The biodiversity data were acquired exclusively through CS, combining unstructured (spontaneous participation of citizens) and structured citizen participation in both formats of CS. Data were gathered through Biodiversity4All, the Portuguese node of the iNaturalist platform, described in more detail in Section 4.

Finally, the data treatment entails cleaning the database and conducting a statistical analysis in SPSS 29 to discuss the number of biodiversity observations in the green roofs and the other garden areas.

3. Calouste Gulbenkian Foundation

3.1. Description and History

The Calouste Gulbenkian Foundation is located in Lisbon (Figure 2) and is comprised of several buildings, namely (i) the Foundation’s Headquarters, (ii) the Gulbenkian Museum, and (iii) the Centre for Modern Art. The buildings are integrated in a green space of approximately 7.5 hectares of gardens belonging to the Foundation and with controlled access, implying that not even the green spaces are accessible 24/7. The construction of the complex began in the 1960s and was awarded the Valmor Prize in 1975. It was declared a National Monument in 2010, making it the first contemporary work to be recognized as heritage in Portugal [27,28].

Due to its characteristics and central location in the city of Lisbon, it is one of the most well-known places in the city, renowned for its cultural, artistic, and educational activities. Its green spaces are also unique and widely used by nearby residents and tourists, reflected in the 835,000 visitors registered in 2022 [29].

Figure 2. Geographic location of the Calouste Gulbenkian Foundation in Lisbon, Portugal, green roofs and gardens areas under study (source: maps are from [30], and the satellite image is from Google Earth).

Figure 2. Geographic location of the Calouste Gulbenkian Foundation in Lisbon, Portugal, green roofs and gardens areas under study (source: maps are from [30], and the satellite image is from Google Earth).

3.2. Green Spaces Characterization

The green spaces present a very diversified planted flora, comprising over 230 species, with areas with different characteristics in terms of density and type of vegetation. This creates a favorable environment for rich biodiversity, reflected in the presence of more than 270 taxa [29]. These green spaces include garden and green roofs, both elevated and at ground level. The latter contribute to making this construction one of the most emblematic architectural symbols of Portugal. The green roofs of the Calouste Gulbenkian Foundation are located at various points within the complex, as shown in Figure 2. Roofs A, B, and D are located at the ground level, and Roof C is elevated.

The green roof to the north of the Headquarters building (Roof A) covers approximately 8855 square meters. It features a wide variety of flora, including ground cover vegetation, shrubs of various sizes, and some large trees. The roof connecting the main building to the Centre for Modern Art (Roof B) has an area of approximately 2050 m². Similarly, it boasts a diverse range of vegetation, with a great variety of species and sizes. The green roof adjacent to the Headquarters building and the Grand Auditorium (Roof C) covers around 3917 m², and its vegetation consists of smaller plants, including ground cover and shrubs. The green roofs of the cloisters of the Gulbenkian Museum (Roof D) are smaller in size, covering about 490 m², and the vegetation is diverse. All green roofs under analysis are of intensive typology.

Due to the vegetation used in the green roofs at street level, one can say that they are intensive green roofs, while the green roofs on top of the building are extensive ones, since they use mostly grass.

The areas of the Foundation’s garden that do not constitute green roofs have been divided into four sections to assist in the upcoming data analysis, as shown in Figure 2. Area 1, located in the northern part of the Foundation’s garden in front of the main building, covers approximately 14,327 m². This area is adjacent to the green roof (Roof A) and features a diverse range of vegetation, including grass, small shrubs, and large trees. Area 2 is situated in the western part of the Foundation and covers 11,265 m²; it is characterized by dense vegetation, featuring large trees and shrubs. This area is adjacent to the green roofs (Roof B and Roof C). The third zone of the common garden that has been defined is Area 3 and is located in the southeast area of the Foundation covering of 16,190 m². This area includes the Foundation’s lake, and it is the area most prone to having a higher concentration of visitors, in part due to the access to the bar. It also presents a wide variety of vegetation, ranging from small plants to shrubs of various sizes and several types of trees and borders (Roof B and Roof C). Lastly, Area 4, covering an area of 5197 m², is situated in the western part of the Foundation’s garden. This area also features dense and varied vegetation, and it includes a terrace area for the use of garden visitors.

The Calouste Gulbenkian Foundation is served by a dense network of paved pathways and trails (Figure 3) that extend to virtually all areas of the garden. It is predominantly on these pathways that people move around, and at various points along these trails, there are benches where people can stop to sit and rest. These pathways are very important for the dynamics of garden use, being sought after by users primarily for leisure activities, such as walking, reading, or photography.

4. Materials and Methods for Biodiversity Assessment

4.1. Data Acquisition

4.1.1. Platform

The data were obtained through the citizen science (CS) iNaturalist platform, a multi-taxa CS tool developed by the California Academy of Sciences and the National Geographic Society, launched in 2008. It is one of the most successful CS initiatives in terms of the quantity of data collected and participation, recording over 198 million observations and more than 7.5 million users in 252 countries. The objectives of this platform include generating scientific data about biodiversity and building a network for sharing information about nature [31,32]. iNaturalist features a mobile application with automatic species identification software and relies on a global community of experts for the taxonomic confirmation of species [33,34,35]. BioDiversity4All is a Portuguese CS platform launched in May 2010 and, nowadays, is the Portuguese node of iNaturalist contributing to increasing the national community by organizing activities including BioBlitz, creating projects, and enhancing the connection between different groups of citizens and stakeholders and the scientific community.

The operating mode of the application (Figure 4) consists of:

1st step: When a participant wishes to make an observation of an organism, they should ideally use the camera or audio recorder on their mobile phone (or both) to record it in the application.

2nd step: Once these data are uploaded to the application, information about the date and location of the observation is also recorded. Then, the application’s trained algorithm for species recognition helps the user determine the taxonomy of the recorded organism and share the observation with the community.

3rd step: If the observation is properly documented and identified by two or more users, and if more than two-thirds of the identifications agree, then the observation is recognized with research grade status [31,33,36]. This classification relies on the evaluation and consensus among the identification communities of the iNaturalist and BioDiversity4All platforms. This serves as an indicator of the community’s confidence in the accuracy of the observations made.

The data compiled on iNaturalist can be accessed and exported to allow for different analyses. For example, the information can be accessed by taxonomic groups, type of observation, its classification, and quality. It is also possible to view observations from specific areas, locations, and projects, including the use of geographic coordinates. Additionally, users can define time intervals during which the observations were made and uploaded and can search observations from specific observers.

Figure 4. Operating mode of the iNaturalist app (adapted from [37]).

Figure 4. Operating mode of the iNaturalist app (adapted from [37]).

4.1.2. BioBlitz and Others Structured CS Events

A BioBlitz is a CS initiative in which, during a specific period, citizens, scientists, and naturalists collaborate to record as many species as possible in a particular area. This methodology allows for a rapid assessment of the biodiversity of a location, contributing to valuable scientific purposes, while also providing enjoyable moments for participants [38]. The objectives of a BioBlitz event are the collection and sharing of information about biodiversity, increasing scientific literacy among both experts and the general public [39].

During the period under review in this paper, three BioBlitz were carried out to collect biodiversity observations. In addition to these BioBlitzes, two training sessions on the use of the application and species identification were conducted, bringing together Foundation employees, people who work and live nearby, and specialists. The dates of these events were 18 September 2021, 23 April 2022, 22 April 2023 (BioBlitzes), 29 May 2023, and 19 June 2023 (training activities). The first three events lasted approximately three and a half hours each and the last two only one hour. All the events began with a brief introductory session, followed by sessions of observation gathering focusing observations of insects and birds, and concluded with a closing session. The durations of the various phases of the event differed significantly depending on whether they were BioBlitzes or training sessions. Since both the BioBlitzes and the training sessions were accompanied by specialists in specific groups (birds and insects), they will be referred to from now on as structured events. The path of one of the BioBlitz is depicted in Figure 4, along with photographic documentation of some of the locations along the route.

4.2. Data Treatment

The biodiversity data collected for this study spans the period from 2009 to 2023. The data retrieval was carried out in July 2023. The data cleaning procedure aimed at excluding incomplete records, e.g., data with unavailable taxonomic group or location. The number of observations until 2016 was residual; therefore, all statistical tests will only include the observations from 2016 onwards.

The statistical analysis included descriptive statistics and univariate analysis. The univariate analysis includes correlation and mean comparisons. For the correlation, the parametric Pearson correlation and non-parametric Spearman correlation tests were used depending on the nature of the relation between the variables, the presence of outliers, and the normality of the data. The comparison of means was also carried out using the parametric t-test (for two categories) and analysis of covariance (ANOVA) (for more than two categories) or the non-parametric Mann–Whitney (for two categories) and Kruskal–Wallis (for more than two categories) tests depending on the normality of the data. The Shapiro–Wilk test was used to test the normality of the data.

Since the number of observations obtained through CS varies with the time and method used (structured or unstructured), the statistical analysis was carried out also using several indicators. For unstructured CS observations, annual data were used. For structured events, only data collected on the days of the events were considered.

• Insects to Birds ratio $={\displaystyle \frac{\mathrm{Number} \mathrm{of} \mathit{Insects} \mathrm{observations}}{\mathrm{Number} \mathrm{of} \mathit{Birds} \mathrm{observations}}}$
• Proportion of observations $={\displaystyle \frac{\mathrm{Number} \mathrm{of} \mathit{Birds}\left(\mathrm{or} \mathit{Insects}\right) \mathrm{observations} \mathrm{in} \mathrm{the} \mathrm{category}}{\mathrm{Total} \mathit{Birds}\left(\mathrm{or} \mathit{Insects}\right) \mathrm{observations}}}\times 100$
• Proportion of observations/day $={\displaystyle \frac{\frac{\mathrm{Number} \mathrm{of} \mathrm{Bird}s \left(\mathrm{or} \mathit{Insects}\right) \mathrm{observations} \mathrm{in} \mathrm{the} \mathrm{category}}{\mathrm{Total} \mathit{Birds} \left(\mathrm{or} \mathit{Insects}\right) \mathrm{observations}}}{\mathrm{Number} \mathrm{of} \mathrm{days} (\mathrm{weekdays} \mathrm{or} \mathrm{weekends})}}$
• Proportion observations/area $={\displaystyle \frac{\frac{\mathrm{Number} \mathrm{of} \mathit{Birds} \left(\mathrm{or} \mathit{Insects}\right) \mathrm{observations} \mathrm{in} \mathrm{the} \mathrm{category}}{\mathrm{Total} \mathit{Birds} \left(\mathrm{or} \mathit{Insects}\right) \mathrm{observations}}}{\mathrm{Area} \mathrm{of} \mathrm{the} \mathrm{green} \mathrm{space}}}$

5. Results and Discussion

5.1. Dataset

A total of 1280 observations have been recorded since 2009, with the number of observations increasing steadily over time, until July 2023. Only 8 have no identified taxonomic group, and 6 are missing the date, so the sample has 1266 complete observations. In total, 961 of these observations were obtained through unstructured CS, and the remaining 305 observations were recorded in the context of structured CS. This number of observations was achieved thanks to the contribution of 238 citizen scientists using the iNaturalist application. Of these 238 users, 6 contributed both in the context of structured and unstructured CS.

The data include observations of individuals from the taxonomic groups Actinopterygii, Amphibia, Animalia, Arachnida, Aves, Fungi, Insecta, Mammalia, Mollusca, Plantae, and Reptilia. The groups referring to birds and insects will have greater emphasis in the analysis because (i) they are the groups with the largest number of observations (568 and 193, respectively, corresponding to 45% and 15% of the total observations); and (ii) the main objective of the structured events was to collect observations for these groups. As such, most analysis will be exclusively dedicated to the observations on these groups. Plants are also a group with a substantial number of observations (377, 30%), but it should be taken into consideration that the green spaces of the Calouste Gulbenkian Foundation were planned and are regularly maintained.

Most of the observations recorded (750) are classified as research grade, with the remaining splitting between casual (250), without media source, and need identification (266) grades.

In Figure 5, it is possible to observe the geographical distribution of biodiversity observations in the context of unstructured and structured CS.

5.2. Unstructured vs. Structured Citizen-Science

In Figure 6, it is possible to observe the distribution of the number of observations per user in the context of data collected through unstructured and structured CS.

In unstructured CS, most citizens make between one and three observations, with a very sharp decrease in observers for the subsequent intervals of the number of observations. Still, it is worth noting that there is a small niche of observers who, in this context, contribute with a high number of observations.

In structured CS, the trend observed previously of a high concentration of observers performing a small number of observations is true. However, it can be observed that, in this context, there is a very significant number of observers making large quantities of observations. In fact, the number of observers conducting more than fifteen observations surpasses the number of observers conducting between one and three observations. This information provides valuable insight into the contribution of citizen scientists when collecting data in each of the different contexts. It is evident that the commitment of participants to the observation collection is higher when they are in the structured CS scenario, with more observations per person, whereas in unstructured CS, the vast majority of observers contribute with a low quantity of data.

Figure 7 illustrates the proportion between the number of structured or unstructured observations obtained in each year and the total number of observations made that year. The same relationship is also present for the groups of birds and insects. This ratio is calculated by dividing the number of observations obtained in each year by the total number of observations, whether in the context of structured or unstructured CS.

It is observed that, since the beginning of the observation survey in 2016, there has been a gradual and continuous growth in the amount of data collected by unstructured CS. It is important to consider that the year 2023 only has data collected in the first six months of the year. On the other hand, in structured CS, the growth has been sustained, with a significant increase in the amount of data collected from 2022 to 2023. Part of the growth recorded in 2023 can be attributed to the fact that more structured events were carried out in that year than in previous years.

One possible reason for this increase in the number of observations is a growing interest of citizens in biodiversity registration. Specifically in Portugal, awareness of this type of tool has also increased specially from 2018 onwards when Biodiversity4All became the connecting node to iNaturalist in Portugal, which helped raise awareness on the topic and consequently increased the number of observations.

Figure 8 presents the proportion of observations made in each month, for all taxonomic groups (Total) and specifically for birds and insects. This proportion consists of the number of observations in each month divided by the total number of observations.

This graph highlights that, in the context of unstructured CS, there is a greater concentration of observations in the months between March and August. The explanation for these values might be the influence of seasons that impacts the quantity of observations, the life cycles of animals and plants, and weather conditions. In Portugal, spring begins at the end of March and lasts until the end of June, when summer begins and lasts until the end of September. This favorable weather conditions increase people’s predisposition to spend time outdoors and engage in activities in nature, therefore probably having an impact on the number of registered observations. In the case of structured CS, the influence of the season is not as noticeable, since these are planned activities, so there is control over the date of execution. However, it can be observed that there is a greater tendency to schedule these types of events during months when higher biodiversity is expected, and the weather conditions are more adequate.

Figure 9 shows the percentage between the total number of recorded observations and the day of the week when the observations were made. This proportion is calculated for all taxonomic groups (Total) and for the taxonomic groups of birds and insects.

Regarding unstructured CS observations, it is possible to observe a higher concentration of observations on Friday, Saturday, and Sunday. This trend is somewhat expected, as there is generally greater availability for recreational activities on the weekends. In fact, during weekends, the gardens of the Calouste Gulbenkian Foundation are one of the most sought-after spaces in the city by its inhabitants for leisure activities. The Calouste Gulbenkian Foundation also promotes various cultural and educational activities with scheduling that often focuses on weekends. All this results in a higher flow of people on these days, possibly contributing to the increase in observations made during these periods.

Regarding only the unstructured citizen data, a Mann–Whitney U test was performed to compare the proportion of the annual observations of birds and insects during the weekdays and weekends. Including Fridays in the weekend group, the proportion per day of bird observations during weekends is statistically significantly higher than during the weekdays (U = 54.50, p = 0.015).

Table 1. Number of days required to unstructured CS reach the same number of observations as structured CS, in the period after two of the BioBlitzes.

Number of Observations Obtained During the Structured CS (BioBlitz)		Number of Days Required by Unstructured CS to Provide the Same Number of Observations as the BioBlitz
18 September 2021	Birds: 17	106
	Insects: 17	356
23 April 2022	Birds: 22	21
	Insects: 18	324

compares the collection of observations in the context of unstructured and structured CS, showing the number of observations obtained in two BioBlitz and the number of days required for unstructured CS to reach the same number of observations of insects and birds.

Regarding the BioBlitz on the 18 September 2021, unstructured CS takes 106 days to reach the same 17 bird observations and 356 days to achieve the same 17 insect observations. In the BioBlitz of 23 April 2022, unstructured CS took 21 days to obtain the same 22 bird observations and 324 days to achieve the same 18 insect observations. These data highlight differences in observer profiles in the two different contexts. This is particularly evident in the case of insects, where unstructured CS takes nearly a year to record the same volume of observations as a BioBlitz event that lasts a few hours. Observing individuals from the taxonomic group of insects presents additional difficulties and may sometimes generate less interest to the average observer. This is not the case for structured CS observers, who receive assistance from experts and are more predisposed and efficient in observing organisms with the aim of gathering as much information as possible.

Finally, these data help to confirm the influence of the seasons. The first BioBlitz takes place on September 18, practically at the beginning of autumn, while the second takes place on April 23, during spring. This factor may influence the subsequent period and is reflected in the time required for CS to obtain the same number of observations as those obtained in the BioBlitz. For both taxonomic groups, the required time was longer in the case of the first Bioblitz, supporting the argument of the influence of seasons on observations.

To enhance the comprehension of the data gathered in the context of unstructured and structured CS, statistical analyses were conducted to provide a more objective perspective on the obtained information about birds and insects.

Table 2. Spearman’s rho correlation between the number of bird and insect observations for all green areas.

		INSECTS
		Unstructured	Structured
BIRDS	Correlation Coefficient	0.647	0.564
	Sig. (2-tailed)	0.083	0.322
	N	8	5

presents the correlation between the number of observations of birds and insects obtained from the Spearman’s coefficient. No statistically significant correlation was found between the number of birds and insects observed in both structured and unstructured CS data. However, the correlation is positive in both cases, and in the case of the unstructured data, the result is statistically significant for a significance level of 10%.

This means that the number of birds is positively related to the number of insects observed, i.e., when more insects are observed, more birds are also observed (and vice versa). Although the result is not statistically significant, we can say that both survey methods are capturing the biodiversity of birds and insects equivalently.

5.3. Green Roofs vs. Garden Areas

Figure 10 shows the density of collected observations, i.e., the number of observations per 1000 square meters of area, in different locations of the Calouste Gulbenkian Foundation identified in Figure 3 and Figure 4.

This graph allows us to understand how the collection of biodiversity data differs between areas located on green roofs and garden areas. What can be observed from this analysis is that, in most of these defined areas, there is a density of observations lower than 1 observation per 1000 square meters. However, there are two areas where the density is higher than 1, one in a garden area (Area 3) and another in a green roof area (Roof B), with the density being higher in the garden area, reaching over two observations per square meter. These two zones are contiguous and are located in an area where, typically, from the user’s point of view, there is a greater immersion in the garden vegetation and, consequently, a higher concentration of people.

In Figure 11, the graph illustrates the difference in density between bird and insect observations collected through unstructured CS and those from structured CS. These data refer to both garden areas and green roofs.

It is evident that the density of bird observations obtained through unstructured CS is higher than those obtained through structured CS in all garden areas. Additionally, there is a clear superiority in observation density in Area 3. A significant portion of the Foundation’s lake is contained within this area. This lake is used by water birds, such as ducks, geese, and moorhens, which attract considerable attention from visitors to the Foundation’s gardens. Moreover, these birds are accustomed to human presence, making them easy to photograph and generating a great deal of interest among citizen scientists in recording their observations on the platforms.

In the case of green roofs, the density of observations in structured CS exceeds that of unstructured CS in two of the green roofs, which does not occur in the garden areas. However, it is observed that the highest density of observations occurred in the green roof (Roof B) in the context of unstructured CS. These values may be explained by the fact that the green roof (Roof B) is located between the garden areas (Area 2 and Area 3), which, as explained earlier, are central areas of the garden, easily accessible, and with a high concentration of people, benefiting the number of observations in unstructured CS. Unstructured CS records a decrease in observation density on green roofs, and a possible explanation for this is that these areas of the Foundation are areas less frequented by visitors, either because they are inaccessible, or because they do not appear to be preferred leisure areas.

Compared to bird data, it seems to be a lesser tendency among observers to survey data on insects. This could be symptomatic of the characteristics of this taxonomic group. Insects are typically small-sized and possess physical characteristics that aid in their camouflage with the surrounding environment, making detection by citizen scientists more difficult. Moreover, many of these individuals do not produce sounds, further complicating the task of detecting them. It is also not uncommon for citizens to have aversions or even phobias of insects, which could also contribute to the lower volume of observations in this group.

In structured CS, the growth in the volume of observations may have been influenced by the presence of specialists at these events. Their technical knowledge of species, combined with methodologies acquired through practical fieldwork, enables them to optimize data collection and find individuals that would go unnoticed by most citizen scientists without formal scientific training. This, coupled with the fact that, in structured events, several people are recording simultaneously, promotes the same individual being observed multiple times, leading to a more pronounced effect, especially in the case of insects.

Once again, it is in Area 3 where the highest density of observations is recorded, as was also observed in the case of birds.

In all green roof areas (Figure 11), it is the structured CS that gather a higher density of observations, by a larger magnitude compared to any of the previous situations. Additionally, the density of insect observations by unstructured CS is notably lower than in the other scenarios analyzed earlier.

Furthermore, it is also observed that the more isolated green roof areas, such as Roof A and Roof C, have a lower observation density compared to Roof B, which is surrounded by garden areas. The elevated nature of Roof C and Roof A being a circulation zone may help explain this difference.

As previously stated, some statistical analyses were conducted to better understand the relationship between bird and insect observations and their distribution across garden and green roof areas.

Table 3. Spearman’s rho correlation between the number of bird and insect observations separately analyzing gardens and green roofs.

			INSECTS
			Unstructured	Structured
BIRDS	Garden	Correlation Coefficient	0.855	0.300
		Sig. (2-tailed)	0.007	0.624
		N	8	5
	Green roof	Correlation Coefficient	0.852	0.821
		Sig. (2-tailed)	0.007	0.089
		N	8	5

illustrates the correlation between bird and insect observations, determined through Spearman’s coefficient. This test aims to determine if there is a difference in the correlation between birds and insects in gardens and green roofs. A statistically significant positive correlation between the number of bird and insect observations was found for the unstructured CS data, for both green roofs and gardens. For the structured CS data, it reveals a positive correlation, but not statistically significant, only for the green roofs. This is possibly explained by the limited number of Bioblitz. Still, the fact that both correlations are positive indicates that an increase in the number of observations positively affects the bird and insect records.

Regarding only the unstructured CS observations, the Mann–Whitney U test indicates a statistically significant lower proportion of yearly observations of birds (U = 2.00, p = 0.001) and insects (U = 1.00, p = 0.003) in the green roof areas compared to the garden areas. This holds also with the proportion normalized by the total area of green roofs and gardens, but in this case, the significance level of the insects is of 10% (birds: U = 9.00, p = 0.015; insects: U = 11.00, p = 0.097). In fact, the proportion of observations is always higher in the garden areas, except in 2020, as can be observed in Figure 12 for the bird observations.

6. Conclusions

The aim of this study was twofold: (i) to understand the role of citizen science for the collection of biodiversity data in green roofs, including the influence of different data collection methodologies, and (ii) to allow the comparison of observations in a case study that combines an urban garden with green roofs.

This information is missing in the literature, and the Calouste Gulbenkian Foundation possesses unique characteristics that proved to be very important for a study of this nature, since it has a great variety and complexity of vegetation of different sizes, as well as the peculiarity of having a lake, expanding the spectrum of ecosystems present. Moreover, it is a location with a high flow of people, as the Foundation is part of the daily life of many citizens. Besides, the gardens have a well-defined network of paths, so it allows a better understanding of the routes used by people and garden areas with highest concentration of users. This diversity of characteristics makes this case study a very comprehensive sample, gathering conditions both in terms of green roofs and green areas, and also in terms of users and citizen scientists.

This work showed that, through CS, it is possible to collect a large amount of biological observation data. Structured citizen science observations allow for the collection of a large amount of information in a short period of time and focusing on groups that are not always the most observed by citizen scientists when they do unstructured records. On the other hand, by following a route defined by scientists, participants can visit areas of the garden, including green roofs, which are not preferred during unstructured observation recording.

Regarding the data collected, it is generally observed that unstructured CS has more observers but produces fewer observations, usually between 1 and 3 observations. In structured CS, the opposite occurs, with fewer observers but producing more observations, usually above 20 observations per observer. It is also noted that unstructured CS produces more observations of taxonomic groups with which observers are more familiar, such as birds, while in structured CS, observations focus equally on groups that usually do not capture as much interest from observers, such as insects.

It is also noticed that, for unstructured CS, the proportion per day of birds’ observations during weekends is statistically significantly higher than during the weekdays, considering Friday a weekend day. This suggests that citizen scientists do more observations when they have more free time. The data also seem to indicate that seasons have influence on data collection through CS, with more observations during the months of spring and summer.

Besides, this study helps to explain one of the main advantages of CS, namely the accessibility to real data on biodiversity of a certain location in a free and simple way. The information used was not obtained specifically for this study but rather the opposite. This study was possible, because these data were already available on the iNaturalist/BioDiversity4All platform, allowing for the development of studies in a more expedited manner.

Regarding the data obtained by unstructured CS, it was found that the correlation between the number of birds and insects observed is statistically significant and identical in both gardens and green roofs. In fact, it is possible that the proximity of green roofs to common garden might contribute to an increase in biodiversity in the area. Future studies regarding this possibility should be carried out to confirm this suggestion.

The use of data obtained through CS brings the great advantage of understanding people’s preferences regarding the observation and register of biodiversity, but some limitations may exist in these analyses.

The volume of data analyzed, although covering a period of about nine years, is still insufficient for carrying out more in-depth statistical analyses. In particular, most observations under analysis were recorded from 2018 onwards, due to the growing interest registered in recent years. In addition, there is also a significant discrepancy between the quantity of observations made in the context of unstructured and structured CS that may influence the comparative analysis.

The analysis was performed assuming no relevant limitations associated to the use of the iNaturalist platform. The iNaturalist application automatically associates GPS coordinates with observations made. Plus, the accuracy of this location may not always entirely correct, depending on the mobile device used. Also, since the green area under study has a continuous design between garden areas and green roofs, the accuracy of the GPS coordinates that are used by iNaturalis might, in some cases, play a role. In most studies, this accuracy is not important, but since the green areas are continuous and jump from green roofs to urban gardens, in this case specifically, it is not always easy to be sure if the monitoring is on a green roof or not. However, this is crucial for the work of this paper, since the goal is to show that the urban green area work as a continuous green space to increase biodiversity.

Future studies should also focus on comparing biodiversity records from CS with data collected systematically by scientists and on evaluating different types of green roofs. For example, it will be important to also evaluate biodiversity on green roofs that, unlike the ones from Gulbenkian Foundation Garden, are not interconnected with other garden areas, to understand how this aspect may interfere with their biodiversity and, thus, help urban planners maximize the design of new green roofs areas in the city.