Webpark Location Based Services for Species Search in Recration Area

The paper describes research work in the area of mobile cartography as part of IST-project WebPark. It presents the concept and development of a location based service for searching for species information in a recration area. With help of this concrete use case, questions about dynamic cartographic symbolisation of point objects can be investigated, which will be hopefully prove generic enough to draw conclusions for all types of thematic information and points of interest. After a short introduction to the WebPark project, the conclusions of a user questionnaire and results of first trial with ArcPad software are presented. The goal is that research work and software development regularly evaluated against the user requirements. For the species search application, a model of question types is created which considers three things; the spatial nature of the questions, how wildlife semantics are used and the intention of the questions. The answer model includes the manifold research projects, which were done in Swiss national park. Adopting this project view ensures that visitors are provided with answers correct to the best of the park’s knowledge. The suggested user-interface allows the question and answer control, based on the developed models. Finally, scale and generalisation issues in the portrayal of wildlife information are described. Based on the model-view-controller paradigm, three components will be used to generate adapted maps model generalisation, an organisation and interpretation component and view generalisation component. 1. INTRODUCTION AND WEBPARK PROJECT The WebPark project is a European research and development project (running between 10/2001-09/2004), which aims to create a platform to deliver Location Based Services in protected and recreation areas. Location Based Services denotes position dependent, personalised services, which are tailored for the individual requirements of tourists and visitors. As a result, the users are enabled to request information from several databases on the internet, whereby the data are filtered with reference to a personal user profile, time relevance and current position. In most cases the selected information has a spatial reference, therefore presentation with maps is emminently suitable, but also additional explanation in form of text, pictures and video are available. The WebPark consotium consists of partners from industry; European Aeronautic Defence and Space Company, Geodan Mobile Solutions, sciences; City University London, University Zurich, Laboratório Nacional de Engenharia Civil Lissabon, and national parks; the Swiss National Park. The industry partners are mainly responsible for developing architecture and webmapping technology. Swiss National Park provides content, such as animal and plant observations, route descriptions, POI etc. and enables the WebPark services to be tested. The research institutes investigate questions about knowledge discovery, the use of intelligent agents and dynamic visualisation on small displays with help of cartographic generalisation (1). Knowledge discovery in the context of the WebPark project is used to create spatio-temporal metadata for services and information [2]. Results could be visualized as several surfaces such as densitiy, visibility and accessibility. Density surfaces of user tracks indicate where time has been spent. Visibility surfaces give information on whether a user can see a location, or allow queries that include whether a user is within a POI’s viewshed. Surfaces storing mean velocity derived from all users’ mobile trajectories provide a measure of accessibility. Geographic knowledge could be extracted from databases of users’ spatio-temporal behaviour to generate summaries and patterns that agents can refer. The agents can then rank available information dependent on user profile, previous behaviour as well as spatio-temporal context. The geographic knowledge discovery component facilitates the creation of map layers which can classify characteristics of specific users and aggregate behaviour. The agent component can then convert this information into a measure of the accessibility of a specific location. This measure can be used to rank the accessibility of the user to different information sources in a more sophisticated form than simple distance functions. For development of WebPark services some assumption are made [3]: • Each visitor has a defined user profile. • The visitor does not want to perform exhaustive searches for information. • The visitor will not know the structure of information available. • The cartographic visualisation will be on a small screen. 2. USER REQUIREMENTS AND FIRST TRIAL WITH ARCPAD Consultations with users as well as the periodical tests are necessary to consider the end user requirements. The following points are a summary of WebPark questionnaires, which showed the preferences of the users referring to required information: ! Safety information has key importance during recreational activities, e.g. actual information on the state of the trials, weather changes, danger of falling rocks ! Visitors are generally interested in information for orientation. In the context of LBS, not only do digital maps have to be provided but also tools for the profile-visualisation of the planned trip, such as 3D-presentations of topography ! Wildlife information is highly demanded so thematic maps for vegetation and animal occurrences are necessary (this should be enhanced by a third information sources on these topics). ! The information provision has to be quick and to be provided on request. Audio mobile alerts and spoken content were not favoured. ! The general results showed that the user wishes to maintain control over information content, delivery (pull/push, text/picture/video) and personal privacy and security. Table 1. User requirements on possible WebPark services after questionnaires at Waddensee Region (left columns) and the Swiss National Park (right columns), [4] nWaddensee = 77 / nSNP = 1000 Important Nice to have Less important Not necessary i) Security information like weather warnings, shelters *62.5% 51.2% 26.4% 26.7% 2.8% 8.9% 8.3% 4.0% ii) Information on plants and animals 16.7% 28.1% 43.6% 41.3% 24.4% 10.2% 15.4% 8.6% iii) Maps and other information for orientation purposes 38.0% 20.5% 33.8% 37.4% 14.1% 12.8% 14.1% 17.2% iv) Information on local research activities 7.0% 8.7% 23.9% 40.0% 39.4% 26.5% 29.6% 11.9% v) Thematic maps, e.g. geology, tides 16.7% 15.4% 41.7% 45.4% 18.1% 16.3% 23.6% 10.4% * the most frequent answer per question and test area is coloured A first extensive test was carried out in the Swiss National Park in July 2002. The trial mainly aimed to test current technology capabilities (GPS positioning and network connection for data transfer) as well as get additional hints from the users. The preliminary prototype for the test was based on a PDA from Compaq (iPAQ+ Navman GPS) and a GPRS phone from Nokia with bluetooth. The ArcPad software from ESRI was used to show position and trial on a map. Further more the user could ask for habitats of different plants and animals, as well as for additional information of several species which were delivered as text, picture or video (see Figure 1). ArcPad-Software [5] was very helpful for a first realistic trial. Besides allowing the possibility of testing the general technical conditions, we obtained several beneficial hints from the users after demonstration of concrete LBS. Nevertheless ArcPad will be replaced during the project with our own platform for presentation of several WebPark services, in order to avoid several apparent weaknesses of ArcPad. These included slow loading of pictures, instability of application (e.g. when zooming) and use of too many menus and buttons, which are partly more disorienting the users then be helpful. Figure 1. Location Based Services with presentation of habitats of different plants and animals 3. SPECIES SEARCH APPLICATION The species application allows a user to access a variety of data sources related to the wildlife and ecology of a national park in a manner that is relevant to their questions. The application also provides a framework into which park administrators can embed their data holdings, so that these can be published and accessed in a comprehensive way. The following section discusses the requirements for a species application from the perspectives of the user and the national park. It then presents a framework for integrating data sources for the purpose of species identification. Finally it discusses issues concerning interfaces and user interaction with the application. 3.1 Model of question types The aim of this analysis is to try to identify a general model of the types of wildlife questions that national park visitors want to pose and the types of things that users want answers about. This model can then be used to design a webpark service against which users can pose such questions. The analysis involves the consideration of a list of visitor questions collected by staff at the Swiss National Park. These questions provide a basis for a general model for the species service, as well as provide real user input against which the service can later be evaluated. The aim of the analysis is to consider three things; the spatial nature of the questions, how wildlife semantics are used and the intention of the questions. 3.1.1 Spatial/Non Spatial Classification The point of classifying questions according to whether they are spatial or not is for three reasons. It helps to consider the best medium to allow users to pose questions and the service to answer questions e.g. with a map or with textual controls. Secondly, it allows a more thorough consideration of how space is used in questions and answers. Thirdly, it identifies the types of spatial features that will need to be represented in the data model to allow users to pose their questions. Because the authors expertise is in the spatial domain, we focussed our analysis on the identified spatial questions. Classifying a question as spatial or non-spatial is not as obvious as it might seem. Part of the problem is that the spatial element may occur in the question, the answer or in the analysis of the question and formulation of an answer. For example, a question, “What flowers are found in Val Trupchun?” contains a spatial location ‘Val Trupchun’, but the answer need not be a map, a list of flowers would suffice. Hence, the question is spatial but not the answer. Likewise, a question, “Up to what altitude to trees grow?” isn’t directly spatial, but because the answer varies spatially it needs be shown using a map. So the answer is spatial but the question isn’t. A question, “Is this tree a mountain pine?” doesn’t contain a spatial element, nor does it need to be answered spatially, however a knowledge of the users location could be used in the analysis of the question and formulation of the answer by limiting the possible answers according to their current surroundings. In order to be consistent in our discrimination the following rules where applied to classify questions as spatial: ! The question contains a spatial predicate, e.g. Where…?, What is the extent of…? ! The question refers to a location or geographic feature, e.g. ‘Here’, ‘This’ (when referring to a location), ‘Val Trupchun’, ‘SNP’ ! It is possible to answer the question with a map, because there is sufficient spatial variation in the answer to allow it and it provides also the most complete answer. For example, “Do Ibex occur here?”, “What flowers are in blossom?” ! The question involves relationships between spatial variables. For example, “Are mountain pines on south exposed slopes a different colour?”, “Is the presence of larches here dependent on soil or altitude?”, “Is it too wet here for ants?” 3.1.2 Wildlife semantics. The wildlife semantics provide the ontology of the service; what can questions be asked about. For the initial classification wildlife questions were simply categorised as flora or fauna, within these only individual species were assumed. However, it was recognised that a more comprehensive consideration of ontology is necessary. Users want to use terms which are less abstract than ‘animal’ (fauna) but more general than an individual species. For example, terms such as “reptiles”, “fish”, “deer”, “woodpecker”, “forest” etc. These are all composite terms by which groups of individual species can be classified, they have levels of abstraction, each of which may have its own specific attributes independent of its sub-parts. In order that the user and the system have a common understanding and agreement of terms, the user should have control over which specific set of terms they wish to use in there queries. 3.1.3 Question types. Five types of question were identified as covering the majority of questions: • Presence: “What [species_type] can be found here?” • Distribution: “Where do [species_type] live?” • Confirmation: “Are [species_type] found at [location]?” • Identification: “What kind of / which [species_type] is this?”; “Is this a [species_type]?” • Association: Spatial “Is [species_type] associated with [geographical_feature]?”; Object “Is [species_type] associated with [species_type]?” Questions about presence, distribution and confirmation are very similar. They aim to complete knowledge about species and their environments by looking to complement knowledge from one domain with knowledge from another. Presence and Distribution are extreme cases where a user only has knowledge in one domain which they hope to match with information in the other (target) domain. So for presence questions, the user has knowledge about the environment and wants to match it with knowledge about species. For example, by asking the question, “What fish live in this river?” or “What animals might I see around where I am?”. For distribution questions they seek to complement knowledge about a species with where they are found in the environment, for example by asking “Where do alpine jackdaws nest?”. However, both types of question can be asked including partial knowledge about the target domain, so the types of question tend to merge according to how much knowledge is provided. Where they merge the questions become confirmation type questions. For example, in the question “Are there mountain pipits in Val Trupchun?”. Because these questions already contain most of the information to form a piece of knowledge the generally only expect a yes/no answer. Each of these two types of question can be asked repeatedly constraining the target domain. For example, the set of distribution questions; “Where are eagles be found in the national park”, “Where can eagles be found here in Val Trupchun?” and “Where are eagles found around me?”, represent an exploration of the eagle distribution by changing the spatial context, e.g. by zooming or panning a map. Likewise, the set of presence questions; “What animals might be found around me?”, “What predators might be found around me?” and “What eagles might be found around me?”, represent a successive constraining of the semantic context, e.g. by filtering a list of search results. The spatial context and the semantic context relate directly to the discussions above. Figure 2 describes these observations. Figure 2. Model of question types 3.2 Model for answers Answering question involves the application of knowledge that has been previously captured from experts. In the case of WebPark this expertise is provided by the national parks. Consultation was made with the Swiss National Park to consider how WebPark users’ questions could best be answered. The issues under consideration were: ! How is the expert knowledge at the national park (knowledge of individuals, research and data holdings) best used to answer users’ questions ! How is the knowledge structured and organised ! How can uncertainty/quality be dealt with ! How can scale and levels of organisation be dealt with ! Which area should be focussed on first 3.2.1 Research project approach A mayor concern of the national park was that answers should be accurate and reflect the research that has been undertaken at there. This means that, as far as possible, answers should be provided by primary source, surveyed, data as opposed to predicted or modelled data derived from other datasets. To support this concern it was decided that answers to users questions should be based on: ! Projects (e.g., an insect study or an ungulate tracking) o The distribution of projects both spatially and in terms of their hierarchical levels of organisation o The distribution of data elements (Features) within a project o Customisable views of project data according to different attributes o Background (Abiotic) data associated with a specific project (e.g., salt stones and ibex observations) ! Features individual spatial data objects (for example an animal observation, a census count) contained within projects ! Expert defined (multi-scale) datasets (e.g., know plant locations) ! Probability surfaces (e.g., probability of finding ungulates at different points throughout the national park based on observation data) ! Ecological classification units (e.g., vegetation units) ! Habitat preference models, where the modelling has been made as part of a research projects with a know degree of accuracy (e.g., bird habitat models) ! Multimedia documentation related to species produced by the SNP Adopting a project view ensures that visitors are provided with answers to the best of the park’s knowledge. This means the answers are scoped to concern facts that have been established in the park through research or experience. The projects are associated with scale ranges relevant to their data capture and the processes they show. They are arranged topologically according to hierarchical levels of organisation that have been established. Other data associated with a project is attached to it as background data (also with a relevant scale range). Figure 3 describes how projects can be used to organise national park data into a common framework for answering user questions. Figure 3. Example of how projects (whole park, census areas, study plots) can be used 3.2.2 Fitness for purpose A guiding principal for cartography is that maps should be fit for the purpose which they were designed for.The issue of activity which users will use wildlife maps for therefore needs to be considered explicitly. An analysis of different activities and their spatial characteristics was made. Table 2 provides an overview of this. Table 2. Activities regarding wildlife information categorized according to the spatial scope over which the are performed. Spatial Scope Map tasks Species identification Identification of individuals by markings e.g. by tags Spatial history of an identified animal Immediate surroundings Searching for individuals to observe e.g. looking for deer Finding good sites for observing species on route General contextual information about current location e.g. ecotype Awareness of species that are typical or unique to the current area Region of activity Guidance to outstanding wildlife exemplars related to a route Finding out about species distribution and diversity Planning a route comparing routes and their relative opportunities and species strengths Whole resort Temporal variation in species distributions (diurnal/seasonal) Table 2 organizes activites into three spatial scopes based on the spatial range over which they are undertaken. Within the scope of the immediate surroundings, tasks involving interaction with individual objects are performed. This might include looking for animals to observe from a viewpoint and trying to identify a flower. In the region of activity tasks are more related to the surroundings of a user for example knowing what the major land uses is where they are, knowing about rare species they should look out for whilst hiking or knowing how far it is until the next place to stop. Planning activities needs to involve the scope of the whole resort. This might be to compare different areas, for instance hiking trails, based on opportunities to observe different types of wildlife or to learn broad background information about a wildlife area. These different scopes are used ultimately to offer project data portrayed in different ways according to the different types of task.