Introducing the combined atlas framework for large‐scale web‐based data visualization: The GloNAF atlas of plant invasion

Abstract Large‐scale biodiversity data, for example, on species distribution and richness information, are being mobilized and becoming available at an increasing rate. Interactive web applications like atlases have been developed to visualize available datasets and make them accessible to a wider audience. Web mapping tools are changing rapidly, and different underlying concepts have been developed to visualize datasets at a high cartographic standard. Here, we introduce the Combined Atlas Framework for the development of interactive web atlases for ecological data visualization. We combine two existing approaches: the five stages of the user‐centred design approach for web mapping applications and the three U approach for interface success. Subsequently, we illustrate the use of this framework by developing the Atlas of Plant Invasions based on the Global Naturalized Alien Flora (GloNAF) database. This case study illustrates how the newly developed Combined Atlas Framework with a user‐centred design philosophy can generate measurable success through communication with the target user group, iterative prototyping and competitive analysis of other existing web mapping approaches. The framework is useful in creating an atlas that employs user feedback to determine usability and utility features within an interactive atlas system. Finally, this framework will enable a better‐informed development process of future visualization and dissemination of biodiversity data through web mapping applications and interactive atlases.

One established way of visualizing spatial data is through the development of an atlas, that is, a tool historically defined as a collection of maps that is comprehensive in its field, arranged systematically, authoritatively edited and presented in a unified format (Alonso, 1968). Modern definitions of atlases, though similar, have become more flexible regarding the medium, organization, spatial extent and content due to the emergence of technology (e.g. digital media, the world wide web, geographical information system software) within the field (Panchaud et al., 2013). The modernization of this definition for cartography has brought about many changes in the way that atlases are now conceived, produced, disseminated and used (Vozenilek, 2019).
A major dissemination pathway to increase the accessibility of large-scale biodiversity datasets is web-based applications (or online atlases). There has been substantial progress in the development of relevant web mapping and web geographical information systems (GIS) technologies (Farkas, 2017); thus, such tools have become more important in ecological and other biological fields (Janicki et al., 2016). A major advantage of web mapping applications is that they offer an effective way to provide geospatial information without the need for additional software (Machwitz et al., 2019).
Open-source web mapping and data visualization JavaScript libraries, such as Leaflet (Agafonkin, 2010) and data-driven documents (D3; Bostock et al., 2011), combined with a user-centred design approach, can lead to the successful creation of online interactive maps (Roth & Harrower, 2008).
At the same time, new challenges emerge concerning web mapping applications, such as geospatial organization, access, display and the use of maps as dynamic portals to inter-connected, distributed, geospatial data resources (MacEachren & Kraak, 2001).
Interactive web mapping applications further urge the developer to design a user interface that is relevant and intuitive in its usability to the target audience, that is, focus on the utility-usability trade-off of web mapping applications. This interactivity enhances the experience of using the web mapping application and allows the user to adapt the cartographic image of the data to one that matches their own view and needs (Ormeling, 1995). If this interaction process works smoothly, a web mapping application will benefit the target audience and have the desired impact.
To facilitate the future development of web mapping applications of large-scale biodiversity data while ensuring successful and intuitive target audience usability (i.e. interface success; Roth, Ross, et al., 2015), we introduce a reusable framework for interactive atlas creation-the Combined Atlas Framework. Subsequently, we will illustrate the application of the newly developed Combined Atlas Framework by applying it to the Global Naturalized Alien Flora Furthermore, we assess mapping functionalities from other web mapping initiatives commonly used. Finally, we discuss the advantages of applying the Combined Atlas Framework in biodiversity research to facilitate data dissemination and simultaneously ensure the usage by the respective target audiences.

| MATERIAL S AND ME THODS
To create an interactive atlas, we propose to employ a user-centred design (UCD) framework, which is considered essential for many web mapping projects (Tsou, 2011). In a UCD framework, web cartographers design an effective and intuitive cartographic representation by focusing on creating user interfaces, mapping functions and dynamic map content (Roth, Ross, et al., 2015). An effective web mapping application framework should be user-centred but should also consider the utility and usability of the application (Roth, Ross, et al., 2015). The Combined Atlas Framework combines the five-stage (strategy, scope, structure, skeleton and surface stage) UCD framework (Tsou, 2011;Tsou & Curran, 2008) with the three U (Usability, Utility and Users) approach for interface success by . Integration of both theoretical approaches is achieved visualization and dissemination of biodiversity data through web mapping applications and interactive atlases.

K E Y W O R D S
atlas, cartography, D3, framework development, GloNAF, invasive alien species, JavaScript, web mapping, workflow by incorporating an iterative element to the five-stage UCD approach to gather feedback on the utility and usability of the atlas ( Figure 1a). This enables the developer to better measure interface success and increases the flexibility of the process. While the user → utility → usability loop is most advantageous between the skeleton and surface stages, it can in practice be integrated at any of the five stages as new information gathered from the users will result in updated utility and usability purposes (Figure 1a). Additional background information to both theoretical approaches with a description of the content of each of the five stages and three U approach is found in Material S1.

| Strategy stage
This stage involves defining both the target users and their needs, recognizing the differences between the wide range of potential users in the field (Haklay & Zafiri, 2008).
To determine the needs of the GloNAF core team for an Atlas of Plant Invasion, a questionnaire was sent out to the target audience (i.e. the GloNAF core team consisting of 11 people), to determine their visualization needs. The questionnaire included the following four questions: 1. Is there any specific data in the GloNAF dataset that you are interested in seeing visualized? (e.g. Counts of taxa per TDWG region? Families per region? Mapping where a specific plant is invasive? Naturalized vs. alien for a region? Visualizing inventory completeness?) 2. Would a data download feature be useful?
3. If yes, I know there is a main focus on global patterns, but would a feature that lets a user download data for specific regions or taxa be useful? (e.g. If you were only interested in Hawai'i you could download that data only, or if you were only interested in one plant/taxa/family you could download just that information).

Would having multiple map views at the same time be useful?
This would give the possibility for comparisons between different regions.

| Structure stage
This stage involves the formalization of the mapping functionalities of the atlas by creating a list of tools needed in the atlas.
As part of the structure stage in the development of the Atlas of Plant Invasions, we performed a competitive analysis. We assessed a set of available biodiversity web mapping platforms and online atlases to gather information on their specific strengths and weaknesses, and methods. We selected a set of online atlases, though not all of them self-define as atlases, with different scopes (Table 1): the Map of Life (Jetz et al., 2012), Ant Maps (Janicki et al., 2016), the Global Inventory of Floras and Traits-GIFT (Weigelt et al., 2020), the Allen Coral Atlas (Allen Coral Atlas, 2020), the Atlas of Biodiversity Conservation in the Coral Triangle (Asaad et al., 2019), the Fish Atlas of Germany and Austria (Brunken & Vatterrott, 2020) and the Gender Atlas of Austria (Wenk et al., 2015). The assessment aimed to identify map types and interaction methods commonly used in F I G U R E 1 (a) The developed framework for interactive atlas creation. On the right, the triangle shows the three U approach of interface success, on the left the five stages user-centred design is shown. Both approaches can be combined using iterative feedback loops at each and between all five stages. The competitive analysis and user test illustrate overlap between procedures undertaken at different stages and are the most important positions to initiate a feedback look. (b) Development stages and key procedures during the development of the atlas of plant invasions following the combined atlas framework comparable projects and gain insights into patterns or similarities between the functionalities. For comparison purposes, we measured representation methods, interaction methods and the technology stack of each web mapping platform, which were compared subsequently.
Representation is described as the way the information on the map is encoded and here assessed by the presence or absence of Following the results of the competitive analysis, concrete sketches of the atlas designs or graphical user interface (GUI) were created and a wireframe (Roth et al., 2017) was developed to better visualize the overall structure of the atlas on screen ( Figure S1).

| Skeleton stage
The skeleton stage involves the arrangement of data objects into meaningful categories and the design of the overall structure and display of the atlas (e.g. the map display window, the sidebar menu and the pop-up windows).
To better structure the users' interactions with the data, the Atlas of Plant Invasions was divided into three views: world view, continent view and plant view. The views would be accessible through a homepage, functioning like a table of contents. The decision was made to better structure the users' interaction with the data and to address the issue of geographical scale following the visual information seeking mantra 'Overview first, zoom and filter, then details-ondemand'. As the dataset is large, it also gives the user the ability to filter out the data they do not need.
The wireframe prototype created in this stage was sent out as a user test (Figures S2 and S3) to the target user group (including 10 people from the GloNAF core team and nine randomly chosen people from a cartographic background). It tested the utility and usability of the atlas while also collecting input and feedback. The utility questions assessed the ability of the atlas to be a resource for GloNAF-related questions, while the usability questions asked about ease of use and learnability of the views. The final questions collected feedback regarding subjective satisfaction with the prototype.

| Surface stage
The surface stage is arguably the most important stage of the framework. This stage focuses on bringing together the actual design of TA B L E 1 Name, URL and a basic description of the web mapping platforms included in the competitive analysis  patterns (an example for the world view is given in Figure 2b and for the continental view in Figure 2c).

| The GloNAF atlas of plant invasions
Each view uses the Robinson projection, a compromise projection that balances geographical accuracy with aesthetics. Equal area projections are then used for each continent view. Finally, the plant view is used to visualize the spatial distribution of specific plant taxa.
The user can search for particular taxa and is provided with a global map highlighting the regions where the taxon is naturalized (an example for the plant view of Ficus carica is given in Figure 2d).
The results of the three comparisons from the competitive analysis are shown in Figure 3, where we display the cumulative number of available features of the interaction, representation and technology methods for each project. It is, however, important to state that low scores in any of the methods do not necessarily correlate with a low-quality atlas or web mapping platform. One reason for this is that the competitive analysis does not consider the target audience of each project or the specific needs and goals it is trying to reach but aims to identify key features in comparable applications and to serve as a brainstorming exercise. In

| DISCUSS ION
The proposed Combined Atlas Framework provides a conceptual approach to developing an interactive web atlas and offers multiple methods to achieve interface success at each stage of the interactive atlas development process. By combining the five-stage UCD approach with the three U approach of interface success, we integrated two existing theoretical concepts that complement each other and increase the robustness of the development process by adding an iterative feedback element. This iterative feedback process between the user and the developer is vital to ensure that all relevant facets are integrated in the atlas, and the provision of intermediate products to the target groups strongly contributes to refining the final product along the process (Roth, Quinn, et al., 2015).

Each atlas and web mapping process is unique because it caters
to the specific needs of the defined user group. In the Atlas of Plant laboration and interaction between developers and users. This will not only ensure the adequate focus of the final product, but also reduce the necessity for time-consuming updates. This assumption is in contrast to other approaches, where early interaction with stakeholders (i.e. target users) has been avoided to provide a more polished product for the first presentation (Slocum et al., 2004).
However, such practice only supports the belief that interacting with the target audience using an 'imperfect' product has no benefit, whereas on the contrary, these interactions are highly valuable for the development success.
The biodiversity community will greatly benefit from more web mapping applications on the vast amount of mobilized biodiversity data that explicitly consider the relevant target audience.
The Combined Atlas Framework is one tool to increase the utilityusability trade-off of such new web mapping applications by considering the background, expertise and knowledge of the audience it wants to address.  202548816). We appreciate the comments by the handling editor (Aaron Ellison), Piero Zannini, and one anonymous reviewer.

CO N FLI C T O F I NTE R E S T
The authors have no conflict of interest to declare.

PEER R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1111/2041-210X.13820