Crowdsourced Indicators of Flora and Fauna Species: Comparisons Between iNaturalist Records and Field Observations

Abstract

Cultural ecosystem services provide intangible benefits such as recreation and aesthetic enjoyment but are difficult to quantify compared to provisioning or regulating ecosystem services. Recent technologies offer alternative indicators, such as social media data, to identify popular locations and their features. This study demonstrates how large volumes of citizen science and social media data can be analyzed to reveal patterns of human interactions with nature through unconventional, scalable methods. By applying spatial statistical methods, data from the citizen science platform iNaturalist are analyzed and compared with ground-truth visitation data. To minimize data bias, records are grouped by taxonomic information and applied to the metropolitan area of Seoul, South Korea (2005–2022). The taxonomic information included in the iNaturalist data were investigated using a standard global biodiversity database. The results show citizen science data effectively quantify public preferences for scenic locations, offering a novel approach to mapping cultural ecosystem services when traditional data are unavailable. This method highlights the potential of large-scale citizen-generated data for conservation, urban planning, and policy development. However, challenges like bias in user-generated content, uneven ecosystem coverage, and the over- or under-representation of locations remain. Addressing these issues and integrating additional metadata—such as time of visit, demographics, and seasonal trends—could provide deeper insights into human–nature interactions. Overall, the proposed method opens up new possibilities for using non-traditional data sources to assess and map ecosystem services, providing valuable information for conservation efforts, urban planning, and environmental policy development.

1. Introduction

We obtain numerous benefits from nature through various forms of relationships with it [1,2]. These diverse values of nature include instrumental (e.g., health benefits of recreation), intrinsic (e.g., the non-use value of species), and relational values (e.g., sense of place, cultural meaning of activities) [2]. Collectively, these non-material contributions of nature to people are often referred to as cultural ecosystem services (CESs), encompassing physical and mental health benefits, landscape aesthetics, and social relations [3,4,5]. CESs play a significant role in enhancing human well-being, fostering enjoyment, and creating a sense of place imparted by nature [6].

Species, as integral components of ecosystems, are pivotal in sustaining the cultural benefits derived from nature [7]. The growing recognition of the role of species and biodiversity in human–nature relationships highlights their importance [8,9]. For instance, certain species enhance the aesthetic and recreational value of landscapes through activities such as wildlife watching [10,11], hiking [12,13], and photography [14,15]. In many regions, specific species hold cultural significance, forming the basis for traditional cultural practices, rituals, and festivals, thereby symbolizing cultural heritage and collective memory [16].

Despite these benefits, increasing urbanization has led to a growing disconnect between people and nature [17,18]. Urban areas, which host dense human populations, present unique challenges and opportunities for biodiversity conservation and the CESs derived from it [19,20]. Urban environments often support diverse species, making them critical for understanding human–nature interactions [21,22]. Activities such as species observation in urban areas not only raise public awareness but also provide valuable data on wildlife distribution and behavior in rapidly changing environments [23,24]. Additionally, the use of digital tools like citizen science platforms and mobile apps amplifies these efforts by enabling broader participation in biodiversity monitoring [25]. These tools have significant social impacts, fostering inclusivity and community engagement, bridging gaps between urban residents and nature, and promoting environmental stewardship among diverse populations [26]. As urbanization is predicted to expand to 68% of the global population by 2050 [27], understanding how urban residents engage with nature and species is essential for fostering a stronger human–nature connection. By enabling public participation in biodiversity monitoring, these tools can provide valuable data to inform conservation strategies, identify at-risk species, and track ecological changes in urban environments [28,29].

However, compared to other ecosystem services, CESs and biodiversity remain challenging to quantify and map, which hinders effective natural resource management [30,31]. To address this, alternative approaches, such as crowdsourced social media data, have been employed to explore human–species relationships and patterns in nature appreciation [32,33]. In recent years, crowdsourced data have proven invaluable for environmental monitoring and species conservation [34,35,36]. These data not only help analyze how people enjoy nature, e.g., [37], but also aid in data acquisition for biodiversity monitoring [38]. Engaging the public in data collection provides information on species distribution, behavior, and population trends that might otherwise be difficult to gather through traditional scientific methods [36,39]. For example, the use of crowdsourced data in species distribution models has more than doubled since the 2010s [36]. By engaging the public in data collection, this collaborative approach enhances our understanding of human–species connections while empowering communities to actively participate in the preservation of biodiversity and the health of our natural environments [23,40,41]. This is particularly significant in urban areas, where crowdsourced species data contribute to a deeper understanding of urban diversity [42].

South Korea is one of the countries experiencing strong industrialization and urbanization [43]. As a result, it has become one of the most urbanized countries globally, with 18.3% of its population residing in Seoul, the capital of the country. This figure rises to 45% when including the greater Seoul Capital Area [44]. Addressing the well-being of this rapidly growing urban population has driven efforts to better understand species dynamics and ecosystem services in the region [45]. For example, Park and Han [45] analyzed the ecological network functions in Seoul, and Jo et al. [46] highlighted the importance of biodiversity enhancement in urban forests for Seoul citizens. Additionally, Serret et al. [47] investigated pollinator families observed in Seoul from 2016 to 2018 using a citizen science program. Despite these efforts, comprehensive information about the distribution of various species and the specific locations people visit remains limited.

To address this gap, we investigate crowdsourced data to better understand the relationship between species and humans. The objectives of this study are to:

• Identify species distribution patterns in the Seoul Capital Area using crowdsourced data;
• Evaluate the consistency of these data with field observations;
• Determine visitors’ preferences for different types of species.

In addition, we explore methodological advancements to better analyze and interpret crowdsourced data.

2. Methods and Materials

2.1. Study Area

This study focuses on the Seoul Capital Area in South Korea (Figure 1). The Seoul Capital Area, located in northwest South Korea, includes the city of Seoul (the capital), the Incheon metropolis, and Gyeonggi Province (Figure 1). With an area of 12,685 km², the Seoul Capital Area is home to approximately 26 million people, of which nearly 10 million reside in Seoul. Tourism in this area is substantial as the area serves many domestic and international visitors [44].

Figure 1. Spatial distribution of visitors (2005–2022) across different tourism site types, reported by the national visitation statistics. Circle colors refer to the different types of visiting sites, while sizes indicate the number of visitors (2005–2022). Korean administrative district boundaries are overlaid for location reference.

2.2. Visitor Statistics

The Korean government publishes the visitor statistics regularly [48], which provide the number of visitors to major tourist destinations across the country. The dataset aims to generate national statistics to estimate tourist demand. Additionally, it evaluates the capacity of tourism facilities and supports planning for resource development. The raw data are collected quarterly by local governments and individual tourist sites, and subsequently reported to the Ministry of Culture, Sports, and Tourism (MCST). The dataset allows for comparisons across different regions and time periods [48].

From 2005 to 2022, the visitor statistics data in the study area covered the number of visitors and locations for the total 694 tourist sites, including natural and cultural attractions (Figure 1). We converted the tourism sites to a set of spatial points using the given coordinates and categorized sites by types of area, namely, ‘sightseeing’, ‘nature’, and ‘culture’, for further analysis following the definition provided by the Korea Culture and Tourism Institute [48]. Generally, indoor facilities such as museums are categorized into the ‘sightseeing’ category, outdoor sightseeing areas such as palaces and temples are categorized into the ‘culture’ category, and natural parks and mountains are categorized into the ‘nature’ category. The information of visitors was categorized into international visitors and domestic visitors depending on their home location. The proportion of domestic visitors was much larger than that of foreigners. Note that the distinction between domestic and foreign visitors is not utilized in this study.

2.3. iNaturalist Database

The iNaturalist platform was utilized to estimate the species distribution patterns by analyzing citizen-science records of biodiversity in the Seoul Capital Area. iNaturalist is a widely used cloud-based platform that enables users to upload photographs of flora, fauna, and insects, which can then be taxonomically identified within the platform. The data from the iNaturalist were acquired via the iNaturalist API and the ‘rinat’ package v0.1.9 [49] for the study period. In the Seoul Capital Area, the acquired data consist of a total of 75,373 records, with 4398 common names and 9049 scientific names. The dataset covers 13 taxonomic groups (kingdom/phylum/class). The breakdown by taxonomic level shows 6 kingdoms, 1 division, and 6 classes: Animalia, Plantae, Insecta, Fungi, Protozoa, and Chromista; Mollusca (phylum), Aves (class), Mammalia (class), Reptilia (class), Amphibia (class), Arachnida (class), and Actinopterygii (class). Most of the downloaded iNaturalist data (99.4%) include scientific names, while 71.6% include common names. We excluded records without scientific names and those labeled as ‘Homo Sapiens’ leaving us with 74,889 records.

2.4. GBIF Database

It is known that the iNaturalist database introduces some uncertainties in its taxonomic classification [28]. To address these uncertainties, we compared the iNaturalist data with the Global Biodiversity Information Facility (GBIF) [50], a global repository that aims to provide a more structured and standardized biodiversity database, incorporating data obscuration where necessary [28,51]. To retrieve standardized taxonomic information for iNaturalist species, we used the scientific names recorded in iNaturalist to query the GBIF database. The data retrieval process was carried out in GNU R, using the ‘rgbif’ package [52,53]. In addition to taxonomic data, we also retrieved the Red List categories from the International Union for Conservation of Nature (IUCN) for species found in iNaturalist. The IUCN Red List provides the conservation status of species, classifying them into nine categories: Extinct (EX), Extinct in the Wild (EW), Critically Endangered (CR), Endangered (EN), Vulnerable (VU), Near Threatened (NT), Least Concern (LC), Data Deficient (DD), and Not Evaluated (NE) [54].

2.5. Point Pattern Analysis

We performed a point pattern analysis using the Cross-K function to assess the spatial relationship between two distinct point patterns: the locations of MCST visitors and the locations of iNaturalist records. Cross-K, an extension of the K function, is utilized to investigate spatial clustering or dispersion between two datasets, primarily for understanding spatial dependencies in tourism and biodiversity data. The Cross-K function computes the expected number of points from one dataset within a specified distance of points from the other, allowing spatial relationships between the two point patterns to be evaluated based on the distribution of one set of points relative to the other [55,56].

Compared to traditional grid-based analyses (e.g., [31,37]), the Cross-K function directly accounts for the spatial configuration of individual points, rather than relying on aggregated gridded data. We adopted this approach to identify underlying spatial patterns that may not be captured by raster-based methods, thereby enabling a more nuanced examination of spatial dependencies.

In this study, the Cross-K function measures how the presence of one set of points (e.g., visitor statistics) is related to the presence of another set (e.g., iNaturalist records). The Cross-K statistic typically measures the difference in spatial patterns between two datasets relative to a Complete Spatial Random (CSR) process, which serves as a null model of spatial randomness [56,57]. Under CSR, points are distributed randomly and independently across the given area. In this study, the CSR model acts as a baseline against which the observed data are compared. The difference between the spatial distribution of the iNaturalist data and CSR represents the deviation of the observed patterns from those expected under CSR. This difference indicates whether the iNaturalist records are clustered, dispersed, or randomly distributed around the locations of visitors compared to a random pattern.

As our study is irregularly shaped, a translation correction was applied in the Cross-K analysis [56]. This correction addresses boundary effects in spatial point pattern analysis by adjusting or shifting the spatial data, thereby mitigating edge bias and ensuring a more accurate representation of spatial relationships [56].

The Cross-K function was evaluated for every 500 m from 0 to 10 km. The 500 m interval was chosen to capture fine-scale spatial relationships while encompassing a 10 km radius to account for broader spatial interactions. To summarize the metrics, we normalized the differences between the estimates of the Visitor–iNaturalist process and the CSR process. These differences were aggregated and normalized on a scale from −1 (representing the largest repulsion) to 1 (representing the largest attraction), across all taxonomic groups. The visitation counts from the MCST visitation statistics and the population size of the iNaturalist records were used as weights in the calculations. In addition to the Cross-K analysis, we visualized the patterns using kernel density estimation, employing a Gaussian kernel with a parameter of $\sigma =5$, which controls the width of the kernel [58]. All computations and visualizations were performed in GNU R, primarily using the ‘spatstat’ v3.3-0 [59] and ‘stats’ packages in GNU R v4.4.2. [60]

3. Results

3.1. Descriptive Statistics

The spatial distribution of the visitor statistics is shown in Figure 1. For the study period (2005–2022), the number of visitors in each category varies significantly. General sightseeing, such as museums and attractions, received the highest number of visitors (1175 million) compared to other types ‘culture’ (205 million) and ‘nature’(105 million). Visitors at these sightseeing spots are concentrated in central Seoul and the northern Gyeonggi-do area near the DMZ (Demilitarized Zone) (Figure 1). Areas classified as ‘nature’, which include mountains and parks, are distributed across the whole region, while ‘culture’ sites are concentrated in densely populated areas. Figure 2 shows the temporal variations in the number of visitors over the study period. The number of visitors significantly decreased starting in 2020, presumably due to the lockdown in South Korea, which lasted until 2022. The visitor statistics suggest that tourism activities were restricted due to the pandemic, leading to fewer movements.

Figure 2. Annual trend in the number of visitors during the study period (2005–2022).

Figure 3 shows the distribution of a total of 74,889 records in the study area. The majority of the iNaturalist data are categorized under the taxa Insecta (40.58%), Plantae (24.49%), and Aves (20.65%), followed by Arachnida (4.00%), Amphibia (3.15%), Fungi (2.35%), Animalia (1.4%), Mammalia (1.23%), Mollusca (0.81%), Reptilia (0.75%), Actinopterygii (0.60%), Protozoa (0.02%), and Chromista (0.01%) (Appendix A Table A1). Given the high prevalence of taxonomic information in most records, iNaturalist data in the area primarily contain information on insects, plants, and birds (Figure 1). The iNaturalist taxonomic group information is not perfectly standardized as the taxonomic levels vary across groups. In particular, the Animalia category contains both freshwater fishes and unidentified Arthropods, leading to heterogeneity within the group. Note that we exclude Chromista in the following analysis due to the extremely low number of records (n = 6). Figure 4 shows the temporal variation in the number of iNaturalist records over the study period. In contrast to the visitor statistics, iNaturalist records actually increased over time, primarily driven by a higher number of bird-related data points in 2020. This suggests that more people went outdoors to photograph birds and uploaded the data to the platform.

Figure 3. iNaturalist data points used for the analysis (2005–2022; n = 74,889). Circle colors represent iNaturalist taxonomic group information included in the metadata. Korean administrative district boundaries are overlaid for location reference.

Figure 4. Annual trend in the number of iNaturalist records during the study period (2005–2022).

3.2. Spatial Point Patterns

To visualize spatial point patterns, kernel density estimation was performed for the visitor statistics. The density estimates were visualized separately for four categories: (a) all visitors together, (b) visitors to cultural heritage sites, (c) visitors to natural sites, and (d) visitors to general tourist facilities (Figure 5). Kernel density estimation was also performed for iNaturalist records. Among all taxonomic groups, the densities of the groups clustered with natural sites are shown (Figure 6). A visual inspection indicates a potential correlation between the iNaturalist record density and the visitor statistics, which revealed iNaturalist records are somewhat concentrated around natural tourism sites.

Figure 5. Estimated kernel density of the visitor statistics for (a) all categories together, (b) culture sites, (c) nature sites, and (d) general sightseeing sites.

Figure 6. Estimated kernel density of the iNaturalist records for the taxonomic groups clustered around the nature-based tourism sites: (a) Animalia, (b) Aves, (c) Mollusca, and (d) Plantae.

3.3. Comparison of iNaturalist and Visitor Statistics

To quantify the relationships more solidly, a Cross-K analysis was conducted to provide a more rigorous statistical comparison between the two datasets. In Figure 7, Cross-K analysis is summarized to illustrate the relationship between the visitor statistics and the iNaturalist records, both in the form of spatial point patterns. When considering all iNaturalist records, no clear association between visitor statistics and the crowdsourced data was observed. However, when analyzed by type of tourism site, certain taxonomic groups showed meaningful associations with visitor statistics. As shown in Appendix A Figure A1, for the selected taxonomic groups, the observed point patterns (solid black line) show higher K values compared to the random process (dotted red line). At nature-based tourism sites, Animalia, Plantae, Amphibia, Mollusca, and Aves exhibited strong clustering. It should be noted that in the dataset, Animalia is somewhat inconsistently categorized, with species including worms and freshwater fishes as well as unidentified species (e.g., scientific name recorded simply ’Animalia’). For Mollusca, the observed associations suggest that snails and slugs are commonly found in nature protection areas that attract visitors, allowing for further interpretation (see species list in Appendix A Table A1).

Figure 7. Summary of the point pattern analysis using Cross-K between MCST visitor statistics and iNaturalist records. Values represent the normalized differences between iNaturalist observations and a Complete Spatial Random (CSR) process.

In contrast, cultural and sightseeing sites exhibited a negative relationship with all taxonomic groups in relation to visitor statistics. The results of the Cross-K analysis showed that most taxonomic groups exhibited strong negative relationships with the number of visitors to cultural tourism sites. This was especially evident after the application of translation correction. Note that in the preliminary analysis, when the baseline Poisson model was applied, a positive relationship was observed across all site types.

3.4. Supplementing iNaturalist Data with Information from the GBIF Database

The comparisons between the taxonomic group information from iNaturalist and the GBIF database are presented in Figure 8 and Figure 9, illustrating how the iNaturalist taxonomic group data differs from the GBIF database at various taxonomic levels. Note that we queried GBIF data for each of the Top 30 species within each iNaturalist taxa group, leaving us with a total of 342 species. In Figure 8, the diagram on the left represents a kingdom-level comparison, showing that a large portion of iNaturalist records fall into the Animalia kingdom. Some records marked as Animalia and Mammalia in iNaturalist data belong to the Plantae kingdom, suggesting possible misclassification in the iNaturalist data. The Phylum-level comparison (Figure 8 on the right) breaks the iNaturalist taxonomic groups into more specific phyla, such as Arthropoda, Chordata, and Basidiomycota. This reveals that the Animalia group in iNaturalist data is substantially mixed up, for example, Arthropoda should not have been included. Except for species belonging to Arthropoda, most of the iNaturalist Animalia group species are worms and algae.

Figure 8. Sankey diagrams comparing iNaturalist taxonomic group data with the GBIF database: the kingdom-level comparison (left) and the Phylum (division)-level comparison (right).

Figure 9. Sankey diagrams comparing iNaturalist taxonomic group data with the GBIF database: the class-level comparison (left) and the order-level comparison (right).

The class-level comparison (Figure 9 on the left) indicates that iNaturalist taxonomic groups are primarily classified at this level, with many groups closely matching their respective classes, except for Plantae and Chromista, aside from Animalia. The order-level comparison (Figure 9 on the right) provides evidence that iNaturalist groups are beyond this level.

We calculated the frequency and distribution of IUCN Red List categories for species found in the iNaturalist data (Figure 10 and Appendix A Table A2). Among the total 342 species from the top 30 species within each iNaturalist taxa group, we retrieved an IUCN class code from the GBIF database for 332 species. Of these, 216 species are categorized as Not Evaluated (NE), followed by Least Concern (LC) (n = 94). The Extinct (EX) and Extinct in the Wild (EW) categories were not identified (Figure 10). Smaller proportions are seen in the Near Threatened (NT) category (four species, 1.20%), Vulnerable (VU) category (six species, 1.81%), and Endangered (EN) category (seven species, 1.79%). The IUCN classification suggests that the study area may not host a large number of threatened species; however, this is insufficient evidence since 65.8% of the species remain unevaluated (NE), and their conservation status is therefore unknown.

Figure 10. Species frequency in the iNaturalist data for each of the Top 30 species per taxa group across IUCN Red List categories. Note that the x-axis is log scaled.

4. Discussions

This study clustered visitor statistics in the Seoul Capital Area using data from a crowdsourced social media database, where users upload species observations. The iNaturalist data were found to be heavily concentrated in tourist hotspots within Seoul’s urban areas and northern Gyeonggi Province. However, categorizing the data by taxonomic groups revealed significantly different patterns. Notably, while the overall correlation between visitor statistics and iNaturalist data is weak, specific taxonomic groups showed strong clustering with ecotourism sites compared to cultural sites, whereas others exhibited negative patterns. These findings suggest that motivations for visiting cultural sites are less related to species observation, whereas natural sites show higher correlations with certain taxonomic groups. For instance, specific groups demonstrated strong positive cross-K relationships with natural sites, though preferences varied among species [61].

Animalia, Aves, Mollusca, and Plantae, in particular, demonstrated a positive cross-K relationship with visitor numbers in natural areas (Figure 7). Charismatic species appear to attract more visitors and foster pro-conservation behaviors [62]. Among the iNaturalist observations, Insecta, Plantae, and Aves were the most frequently recorded groups. Birdwatching, which accounted for the highest number of animal-related records, is especially popular in urban areas due to its psychological benefits and positive effects on human well-being [63]. However, iNaturalist user preferences often exhibit biases toward visually striking or culturally significant species. This leads to the over-representation of charismatic fauna and flora, such as large mammals and colorful birds, while less conspicuous organisms, like fungi and invertebrates, are under-represented [64,65]. Geographical factors, such as road density and accessibility, also influence observation patterns [66].

Given these biases, interpreting the causality of observed relationships requires caution. For example, while people may prefer freshwater environments where Mollusca and fish species are found [67], it remains unclear whether visitors specifically sought these areas for species observation. Over-reliance on simple correlations or iNaturalist data alone may lead to misleading conclusions, as demonstrated in Potsikas et al. [68], Cao and Hochmair [69]. The platform is more suited to exploring the discovery of plant and animal species in relation to the CES of ‘non-use values’ [61]. Complementing social media data with field observations and interviews could provide a deeper understanding of visitor motivations, particularly for cultural and natural sites. While crowdsourced data are rich, they may over- or under-represent specific activities or species depending on location [31,70]. Moreover, incorporating ground data at the national level [71] and enhancing global datasets could align local and global biodiversity records [72]. Such integrated approaches can prioritize conservation planning for endangered species and mitigate biodiversity loss.

Social media data have been widely used to analyze CES, often through aggregated methods like photo-user days (PUDs) using platforms such as Flickr [73]. While these methods offer insights into CES distribution, they risk obscuring finer-scale patterns. This study addresses these limitations by employing point pattern analysis, which preserves the granularity of data. This approach improves the accuracy of spatial analyses and helps identify localized patterns essential for biodiversity conservation and ecosystem management [74].

While the use of social media data for species distribution and CES analysis is cost-effective and captures spatiotemporal patterns of ecosystems [70,71,75], these findings must be contextualized within broader social–environmental frameworks. For instance, how can such patterns inform global discussions on sustainable tourism, biodiversity conservation, and equitable access to natural spaces? This study demonstrates the potential of social media platforms to engage broader audiences in biodiversity awareness but highlights the need for critical evaluations of biases and uneven ecosystem representations [71].

These comparisons reveal how social media-based observations may over-represent certain taxonomic groups, particularly those that are more visible or commonly observed, while GBIF’s more standardized records might show a more even distribution across different groups. The differences between these two datasets underscore the strengths and limitations of each, suggesting the complementary nature of combining both sources for a more holistic understanding of global biodiversity.

Despite its value, the iNaturalist database has limitations, including inconsistent taxonomic classifications and biases in species representation. Many species remain Not Evaluated (NE) under the IUCN Red List within the Seoul Capital Area, limiting biodiversity assessments. Moreover, geographical factors like road connectivity and accessibility may further influence data collection and observation patterns [61]. Future studies should prioritize integrating multiple databases to enhance reliability and reduce biases in conservation planning [31,76].

This research not only advances understanding within the Seoul Capital Area but also contributes to global knowledge on the intersections of visitor behavior, species observation, and ecosystem services. By integrating diverse data sources and considering local socio-ecological contexts, future studies can offer more comprehensive assessments. Ultimately, this study underscores the dual potential of social media data: as a research tool and as a means to inspire inclusive global dialogues on human–nature relationships.

5. Conclusions

This study highlights the potential of crowdsourced data, such as iNaturalist, to provide valuable insights into biodiversity and CESs while addressing its limitations. Patterns of species observations strongly align with ‘nature’ areas, reflecting ecotourism motivations, whereas cultural sites show weaker correlations. Charismatic species attract more visitors and promote conservation awareness, but biases in user preferences and geographic accessibility lead to under-representation of less conspicuous taxa. While social media data are cost-effective and capture spatial and temporal trends, their biases necessitate integration with field observations and alternative datasets for comprehensive assessments. Beyond the Seoul Capital Area, this research underscores the role of citizen science in advancing biodiversity conservation and sustainable tourism, advocating for inclusive approaches that bridge public engagement and equitable access to natural spaces. Moreover, the findings can inform governmental decision-making processes, supporting the development of evidence-based policies for biodiversity conservation and sustainable land use planning.