Pioneering Easy-to-Use Forestry Data withForest ExplorerGuillermo Vega-Gorgojo ab Joseacute M Gimeacutenez-Garciacutea a Cristoacutebal Ordoacutentildeez cd andFelipe Bravo cd

a Group of Intelligent and Cooperative Systems Universidad de Valladolid SpainE-mails guivegteluvaes jmgimenezgarciagsicuvaesb Departamento de Teoriacutea de la Sentildeal y Comunicaciones e Ingenieriacutea Telemaacutetica ETS Ingenieros deTelecomunicacioacuten Universidad de Valladolid Valladolid Spainc Sustainable Forest Management Research Institute iuFOR Universidad de Valladolid amp INIA SpainE-mails a_cristopvsuvaes fbravopvsuvaesd Departamento de Produccioacuten Vegetal y Recursos Forestales ETS Ingenieriacuteas Agrarias Universidadde Valladolid Palencia Spain

Abstract Forest Explorer is a web tool that can be used to easily browse the contents of the Cross-Forest dataset aLinked Open Data resource containing the forestry inventory and land cover map from Spain The tool is purposedfor domain experts and lay users to facilitate the exploration of forestry data Since these two groups are notknowledgable on Semantic Web the user interface is designed to hide the complexity of RDF OWL or SPARQLAn interactive map is provided for this purpose allowing users to navigate to the area of interest and presentingforestry data with different levels of detail according to the zoom level Forest Explorer offers different filter controlsand is localized to English and Spanish All the data is retrieved from the Cross-Forest and DBpedia endpointsthrough the Data manager This component feeds the different Feature managers with the data needed to bedisplayed in the map The Data manager uses a reduced set of SPARQL templates to accommodate any datarequest of the Feature managers Caching and smart geographic querying are employed to limit data exchangeswith the endpoint A live version of the tool is freely available for everybody that wants to try it ndash any devicewith a modern browser should be sufficient to test it Since December 2019 more than 3200 users have employedForest Explorer and it has appeared 12 times in the Spanish media Results from a user study with 28 participants(mainly domain experts) show that Forest Explorer can be used to easily navigate the contents of the Cross-Forestdataset No important limitations were found only feature requests such as the integration of new datasets fromother countries that are part of our future work

Keywords forestry geospatial Linked Data data access map visualizations user interfaces

1 Introduction

Forest science and forestry rely on the use oflarge-scale datasets that cover lengthy periods oftime [21] This is because trees are long-lived or-ganisms that require continuous monitoring to ob-tain sound and accurate information Similarlylarge-scale monitoring systems are used to capture

Corresponding author E-mail guivegteluvaes

the complexity and structure of forests in their fullextend Permanent and extensive data recordingsystems particularly land cover maps and forestinventories [31] are needed to implement soundsustainable forest management [4 24 25] to en-sure a constant flow of ecosystem services such ashabitat conservation and raw materials (timberresin cork )

Due to long-term scale of forestry actions theprivate sector has no incentives to conduct this

type of data collection and curation As a resultthe public sector is the main responsible for mon-itoring forests and providing this information tosociety Such information is consumed for differ-ent purposes by diverse end users including foreststakeholders (governments at different levels en-vironmental NGOs and other lobby groups) oper-ational foresters data and environmental journal-ists interested citizens and start-up promotersHowever exploiting forest inventories and land

cover maps is a non-trivial task that requires bothdomain expertise and technical skills Forestrydatasets are typically isolated described with dis-parate data schemas and using unfamiliar (some-times even proprietary) formats Given these lim-itations Linked Open Data and Semantic Webtechnologies can help to facilitate the integrationand accessibility of forestry data Towards thisgoal the EU project Cross-Forest1 publishes asLinked Open Data a set of national forestry datasources from Spain ndash the Cross-Forest dataset fromnow on This resource has been released throughthe creation of a suite of ontologies to representforestry data the transformation of their forestinventories and land cover datasets and the con-nection among them and with relevant externaldatasetsThe Cross-Forest dataset is available as Linked

Data through the SPARQL endpoint at httpsforestexplorergsicuvaessparql This datasetis a complex resource that integrates spatial fea-tures corresponding to plots trees and land coverpatches Target users from the forestry domainneed to explore and analyze this dataset in or-der to fulfill their goals This is especially chal-lenging as this user group is not fluent in Seman-tic Web languages While there are some LinkedData browsers for lay users [8 9 14] most of themhave limited support for geospatial data More-over visualization tools of geospatial Linked Dataare very scarce and typically require knowledge ofSPARQL eg Sextant [18]In this paper we present Forest Explorer a web

tool for easily accessing the contents of the Cross-Forest dataset Forest Explorer is designed for non-Semantic Web experts so the user interface isbased on an interactive map that completely hidesthe manipulation of RDF data and exchange of

1httpscrossforesteu

SPARQL queries behind the scenes The rest ofthe paper is organized as follows Section 2 reviewsexploration tools of geospatial Linked Data aswell as existing approaches for visualizing forestrydata Section 3 presents the functional require-ments of Forest Explorer Section 4 describes thedesign and implementation of the tool In Section 5we give evidence of the impact of Forest Explorerincluding the results of a user study The paperends with a discussion and future work lines inSection 6

2 Background

Since the earlier years of the Semantic Web thechallenge of exploring RDF data was evident [13]Preliminary projects in this area were targeted toeither Semantic Web experts or technology enthu-siasts willing to put in the time to learn More re-cent works [9] have stressed the importance of sup-porting both lay users and domain experts eg for-est managers Such users may not have any knowl-edge of SPARQL OWL or RDF and require ap-propriate tools to work with Linked Open Data

There are some Linked Data browsers visualquery builders and exploration tools that do notrequire expert knowledge of Semantic Web tech-nologies [8 9 14] A recent example is RDF Sur-veyor [33] a lightweight exploration tool targetedto lay users that is part of our previous work Un-fortunately most of the exploration tools availablehave limited support for geospatial Linked Dataif any eg the visualization of an entity in RDFSurveyor includes a geo widget with a marker if apoint location annotated with the Basic Geo Vo-cabulary2 is found

Some systems have been designed to supportthe visualization of geospatial Linked Data Againmany proposals are targeted to Semantic Web ex-perts GeoYASGUI [1] is a SPARQL editor thatnatively supports GeoSPARQL and provides a re-sult set visualizer Sgvizler [28] is a JavaScript li-brary that can produce different charts ndash includ-ing maps ndash with the results of SPARQL queriesand Sextant [18] is an advanced visualization toolthat can combine spatial data from several end-points and represent the temporal dimension al-

2httpwwww3org200301geo

though it still requires knowledge of SPARQLin order to use it There are seldom visualizersof geospatial Linked Data for non-Semantic Webexperts LinkedGeoData browser [29] is a dedi-cated visualization tool for OpenStreetMap andMap4RDF [10] is a browsing tool of geospatialRDF datasets that uses a faceted interface to con-trol the information to displayIn the forestry domain we can find several ini-

tiatives at national or regional level focused on thedelivery of raw data but not using Linked Dataand Semantic Web technologies A remarkableexample at pan-European level is the EFISCENdatabase portal [27] that offers forestry datasetsfrom 32 countries but using disparate schemataand data formats National data portals are com-mon although the typical case is downloadableraw data in proprietary format (eg Spanish for-est inventory3) or just aggregated results (eg thefirst edition of the Spanish forest inventory4)

Visualization of forestry data is commonlyachieved through a Geographical Information Sys-tem (GIS) For example Global Forest Watch5

is a web application that provides informationabout forest status and land use management atglobal scale GISs do not rely on Linked Data tech-nologies and use instead their own formats typ-ically standardized by the Open Geospatial Con-sortium6 As a result geospatial data is syntacti-cally interoperable but the integration of externaldatasets into a GIS is time-consuming and com-plex [17]Other forestry tools focus on the analysis of

data This is the case of BASIFOR7 [3 22] aflexible and powerful forestry analysis tool thatcan process data from the Spanish forest inven-tory Initially designed as a tool for forestry re-search BASIFOR is also used for management andplanning purposes as it calculates timber stocksdensity forest structure specific composition andother forestry variables in a geographical regiondefined by the user BASIFOR holds strong capa-

3httpwwwmitecogobesesbiodiversidadserviciosbanco-datos-naturalezainformacion-disponibleifn3aspx

4httpwwwmitecogobesesbiodiversidadserviciosbanco-datos-naturalezainformacion-disponibleprimer_inventario_nacional_forestalaspx

5httpwwwglobalforestwatchorg6httpswwwogcorg7httpwwwbasifores

bilities in terms of data manipulation but is notbased on Semantic Web technologies and is thustied to the schemata of the Spanish forest inven-tory

3 Requirements

The Spanish forest inventory [2] is a continuousdataset that is updated every 10 years A plot isan homogeneous and small area of the territorythat constitutes the sampling unit Forest techni-cians survey plots to gather tree data (locationspecies diameter and height) The current versionof this inventory accrues 14M trees 919K plotsand 43M positions The Spanish land cover map8

contains patches of terrain with similar character-istics described using polygons over the territoryPatch data includes soil use and dominant treespecies The current version of this dataset accrues6802K patches Spanish provinces are employedto aggregate data in both cases

The Cross-Forest dataset integrates the latestversions of the Spanish forest inventory and landcover map in Linked Data format thus fulfillingone of the main goals of the Cross-Forest projectProvinces patches plots and trees are modeledas spatial features with a geometry and a set ofmeasures extracted from the corresponding sourceThe details of the main features of each datasetare shown in Table 1 Their positions are rep-resented using a simple ontology that indicatesthe Coordinate Reference System and the coor-dinates of the position This ontology makes safereuse [7] of relevant geographical ontologies in-cluding GeoSPARQL [20] the W3C Basic Geo Vo-cabulary [5] and the ISA Programme LocationCore Vocabulary [19]

The original datasets includes features withabsolute positions using UTM coordinates butsome features namely trees are only annotatedwith relative positions using plot centers as refer-ence When making the conversion into RDF wehave enriched the original data with the inclusionof WGS84 positions for every feature includingtrees UTM coordinates and relative positions arestill available in the Linked Open Data version of

8httpswwwmitecogobesesbiodiversidadserviciosbanco-datos-naturalezainformacion-disponiblemfe50aspx

Table 1Feature types and feature data of the Cross-Forest dataset

Feature type Feature data

Province Number of treesBasal area (m2)Volume with bark (m3)

Patch ProvincePolygonArea (m2)Soil useCanopy coverTree species

Plot ProvinceCoordinatesNumber of trees ha

Basal area (m2ha)Volume with bark (m3ha)

Tree PlotCoordinatesHeight (m)Diameter (mm)Tree species

the datasets so there is no impact for users of theoriginal data sources We used Proj4js9 and Map-shaper10 to transform coordinates from one coor-dinate system to another When developing a mapapplication it is quite convenient to use WGS84coordinates as they are supported in the majorityof GISsThe transformation of the land cover map into

RDF includes the creation of a new layer ofpatches in a lower resolution in order to makeefficient the drawing of polygons on top of aWeb map (see Best Practice 4 in [30]) The tax-onomical structure of trees and bushes (includ-ing class genus family and species) is linked torelevant datasets in forestry and biology fieldsthe NCBI taxonomy [12] and the CrossNaturedataset11 They are also linked to other well-known cross-domain knowledge graphs DBpe-dia [16] and Wikidata [11] Links to DBpedia andCrossNature datasets are made using owlsameAs

9httpproj4jsorg10httpsmapshaperorg11httpscrossnatureeudata

Table 2Statistics of the Cross-Forest dataset

Item Inventory Land cover Total

Distinct classes 89 204 325Distinct properties 49 23 210Distinct individuals 194M 391M 585MDistinct subjects 104M 206M 310MDistinct objects 129M 302M 412MDistinct literals 40M 100M 129MGeometries (polygons) ndash 9223K 9223KSize of TTL files (GB) 27 75 108Triples 551M 1424M 1978M

and schemasameAs since they involve individu-als In the case of Wikidata and NCBI we inter-linked classes by means of rdfssubClassOf Theoriginal sources were transformed into RDF usingSPARQL generate [15] The resulting Cross-Forestdataset is currently published in a SPARQL end-point12 using Openlink Virtuoso v07203230 Ta-ble 2 provides some statistics of the dataset thatwere computed in October 2020

We identified a target group interested in theresulting integrated dataset domain experts suchas forest managers and operational foresters thathave to invest a lot of time with manual inte-grations or using forest analysis tools like BASI-FOR [3] due to strong coupling to schematalegacy formats or limited tool scope For instance[23] reports a significant effort integrating the dataof three provinces of the Spanish forest inventory(this dataset is sliced in 50 files one per province)ndash note that this is a simple integration case asthe database schema is the same for the threeprovinces and no other data sources were requiredMoreover analyzing forestry data from differentcountries is a time-consuming and difficult taskrequiring schema harmonization and data conver-sions Another target group corresponds to layusers such as data journalists or citizens interestedin forestry These two groups are not knowledge-able about Semantic Web technologies so theyneed a tool that shows the data while hiding thecomplexities of its representation As a result ofa collaborative design effort among two forestryexperts and two Semantic Web practitioners (the

12httpsforestexplorergsicuvaessparql

authors of this paper) we identified the followingrequirements

R1 Portable The tool should run in different de-vices ranging from desktop computers to mo-bile phones Moreover the installation pro-cess should be as simple as possible to facili-tate the adoption of the tool

R2 Effective hiding of RDFOWLSPARQL Theend user does not need to know about RDFOWL SPARQL or any other Semantic Weblanguage Instead the user interface has tooffer appropriate visualizations that presentthe information in an appropriate way to theusersrsquo needs

R3 Interactive map Since lay users have em-braced map applications [32] and the tar-get dataset is about spatial data an inter-active map seems a suitable visualization forthis case Typical map navigation controls likepanning or zooming can be added to easilyexplore the zone of interest

R4 Adaptable to different zoom levels The toolshould serve to explore large or small areasIn the former case lower resolution of geome-tries should be served When zooming in toa small area a high level of detail of geome-tries is appropriate and tree markers shouldbe drawn

R5 Filtering capabilities The user needs to con-trol the information at display In particularthey should be able to set filters of tree speciesand land uses as well as controlling which el-ements of the view to show eg choosing be-tween province or landscape visualizations

R6 Multilingual The tool should be localized toEnglish and Spanish

4 Design and implementation

In this section we present the tool devised to eas-ily access the contents of the Cross-Forest datasetThe proposed tool is Forest Explorer and satisfiesthe requirements described in Section 3 The fol-lowing subsections dive deep into the design andimplementation of Forest Explorer

Cross-Forest endpoint

Data cacheDBpediaendpoint

Map server

Data manager

Province manager

Patch manager

Plot manager

Tree manager

Map generator

SPARQL

Figure 1 Logical architecture of Forest Explorer

41 Logical architecture

The logical architecture of Forest Explorer is de-picted in Figure 1 The Map generator is in chargeof displaying the view for the end user This com-ponent employs a base map obtained from the Mapserver and listens to the requests made from thedifferent Feature managers for showing markerspolygons popups or tooltips on top of the mapMore specifically the Province manager gathersprovince geometries and forest inventory data ag-gregated by province and prepares a suitable dis-play request to the Map generator Patch Plotand Tree managers work in a similar way obtain-ing first the information about the correspondingfeatures (see Table 1) located in the geographicalbounds of the current map view and then sendingdisplay requests to the Map generatorIn order to comply with requirement R4 the dif-

ferent Feature managers are activated dependingon the zoom level In case of large areas the usercan choose between province or patch views (R5)In the former case the Province manager takescontrol and requests the presentation of inventorydata aggregated by provinces In the latter casethe Patch manager is activated and requests thelowest resolution layer available of the land covermap With intermediate zoom levels the Patchand Plot managers are both enabled to present theplots on top of a land cover map of the visible areaWith high zoom levels the Patch and Tree man-agers are activated ndash the highest resolution layeravailable of the land cover map is employed in thiscase while the Tree manager requests the presen-tation of tree markers to the Map generator basedon the inventory data for the target area Notethat this design is compliant with Best Practice 4

in [30] recommending to provide multiple resolu-tion versions of features at different zoom levelsThe Data manager handles all the data requests

from the Feature managers This component isable to communicate via SPARQL and is config-ured to use the Cross-Forest and DBpedia end-points As described in Section 3 the Cross-Forestdataset contains a Linked Open Data version ofthe Spanish forest inventory and land cover mapDBpedia is employed as a source of tree species in-formation providing images and multilingual de-scriptions (R6) Upon receipt of a request theData manager first checks if the result is alreadyavailable in the Data cache In case of a miss theData manager sends one or more SPARQL queriesto the endpoints Section 42 gives further detailsof the functioning of the Data manager

42 Data gathering

As depicted in Section 41 the Data manager isin charge of all data gathering operations in ForestExplorer The design of this component is chal-lenging due to a number of reasons (1) requestslook quite varied referring to different types offeatures (provinces patches plots and trees) andall their associated data (2) the size of the Cross-Forest dataset is not small ndash 43GB correspond-ing to 737M triples and (3) requests can be verynumerous since any change in the interactive mapwill trigger data requests by one or more FeaturemanagersFortunately the different request types can be

abstracted to the identification of features local-ized in a specific area and then obtaining their cor-responding feature data The Data manager usesSPARQL template queries for retrieving the fea-tures localized in the map view Since plots andtrees have points as geometries a suitable querybasically has to check which points are includedin a bounding box Listing 1 shows the tem-plate used for plots in which latnorthlatsouth lngwest lngeast areplaceholders for the map view bounds

Listing 1 SPARQL query template for retrievingplots in the map view

SELECT plot lat lngWHERE

plot a finvPlot

poshasPosition pos pos poshasCoordinateReferenceSystem crs4326

axis1 lat axis2 lng

FILTER (lat gt latsouth) FILTER (lat lt latnorth) FILTER (lng gt lngwest) FILTER (lng lt lngeast)

As patches and provinces have polygons as ge-ometries queries have to be adapted to find thepolygons intersecting with the map view The tem-plate used for patches is included in Listing 2Note that we use the bounding box of the poly-gon (included for all polygons in the dataset)to simplify the detection of patches in the mapview In order to comply with requirement R4the template is parametrized to select a specificlayer and to define a minimum threshold forthe area of polygons ie minarea The Datamanager chooses the most appropriate layer basedon the zoom level of the map view while it sets theminimum area to 10 pixels in the user device13

Listing 2 SPARQL query template for retrievingpatches in the map view

SELECT patch poly west east north south areaWHERE

patch a mapPatch poshasPolygon poly

poly epsghasLeftBound west epsghasRightBound east epsghasUpperBound north epsghasLowerBound south poshasAreaInSquareMeters area posisInLayer layer

FILTER (south lt latnorth) FILTER (north gt latsouth) FILTER (west lt lngeast) FILTER (east gt lngwest) FILTER (area gt minarea)

Once the Data manager obtains the set of fea-tures in the area of interest it retrieves the cor-responding data in a next step Listing 3 showsthe query template used for this purpose it is ex-

13The assumption here is that patches of less than 10pixels are barely visible so it is better to not waste timeand computation resources with them The area of a pixelchanges with the zoom level so a patch discarded for beingtoo small can be later displayed in case of zooming in

tremely generic and easily adaptable to each fea-ture type For example obtaining the height of acollection of trees just requires setting their IRIsand the height property IRI defined in the Cross-Forest ontology As a result gathering feature datais just a matter of selecting the set of properties toextract for each feature type (see Table 1) Classmembership is required in some cases ndash for exam-ple to obtain tree species ndash so we have includedanother generic template for retrieving the classesof a set of individuals

Listing 3 SPARQL query template for obtainingthe values of a property for a set of individuals

SELECT DISTINCT iri valueWHERE

iri propiri value VALUES iri iris

The Data manager stores all retrieved data ina Cache so as to reduce exchanges with the end-point This is quite effective with feature datasince the template query in Listing 3 can be eas-ily parametrized to avoid requests of feature dataalready available in the Cache With respect tofeature locations a similar approach consists oncaching the results for every map view bounds re-quest ie queries built with Listing 1 for plotsListing 2 for patches and so on However this so-lution is too naive as the cache will only get a hitin case of a request with exactly the same mapboundaries as a previous one Instead the Datamanager keeps track of the map regions with lo-calized features and exploits this information torestrict queries to unknown areas Figure 2 illus-trates the idea

(a) At the beginning the Data manager has no in-formation of any part of the map

(b) Upon receiving a feature location request forregion R0 it asks the endpoint about thefeatures in such area ndash the Data managerstores the results in the Cache and updatesits tracked area to the bounding box of R0

(c) A request about region R1 is decomposed intosubregions R1prime and R1primeprime ndash R1prime is the inter-section of R1 and the tracked area so theincluded features can be obtained from theCache Since R1primeprime is unknown to the Datamanager it has to query the endpoint to re-

trieve the features in R1primeprime Then the Datamanager updates its tracked area to R0 cup R1

(d) A request about region R2 does not requirefurther queries to the endpoint as R2 is con-tained in the tracked area

43 User interface

In order to comply with requirement R3 theuser interface of Forest Explorer exposes an inter-active map as its main component Figure 3 showssome snapshots that illustrate the design of theuser interface Since the tool has to work with a di-versity of devices (R1) the map is set in fullscreenmode to easily adapt to different screen sizes Sim-ilar to other map applications panning and zoom-ing are naturally supported for both point-and-click and touch interfaces In this regard zoombuttons are included in the lower-right corner theextra button with a person icon is used to navigateto the user location ndash this latter functionality isespecially useful when employing Forest Explorerwith a mobile device in a field trip

As the user navigates with the map a Featuremanager takes control by obtaining feature infor-mation from the Data manager and then sendingdisplay requests to the Map generator (see Sec-tion 41) For example the Province manager con-trols the view of Figure 3(a) the Patch manageris activated in Figure 3(b) the Patch and the Plotmanagers collaborate to build the view in Fig-ure 3(c) and the Tree manager is active in Fig-ure 3(d)

Beyond map navigation the user interface needsto include controls to allow the user to adjust theinformation to display (R5) For this purpose For-est Explorer includes a form in the upper-left partndash see Figure 3(a) This form can take a significantpart of the screen in mobile devices so it can becollapsed by pushing the lsquo-rsquo button ndash the collapsedform is shown in Figure 3(b)(c)(d)

A key characteristic of the dataset is the treespecies of land cover map and forestry inventoryThus the form includes a lsquoFilter speciesrsquo buttonthat can be pushed to easily browse the taxonomyof tree species (obtained from the Cross-Forest on-tology) and then select one or more filters Theform in Figure 3(a) includes two species filters Pi-nus sylvestris in indigo color and Pinus pinasterin brown Each filter includes buttons for removal(lsquoxrsquo icon) color change (lsquotintrsquo icon) and more info

Figure 2 The Data manager limits query exchanges with the endpoint by keeping track of the map regions with localizedfeatures The map is initially unknown to the Data manager (a) so a feature location request for region R0 (b) requiresquerying the endpoint ndash the Data manager stores the results in the Cache and updates its tracked area to R0 In (c) arequest about region R1 is decomposed into subregions R1prime and R1primeprime ndash the features in R1prime are obtained from the Cache andthe endpoint is queried to retrieve the features in R1primeprime The tracked area now includes R0 cup R1 As region R2 is containedin the tracked area the feature location request in (d) is answered with the Cache

(lsquoinforsquo icon) ndash the latter one displays a popup withadditional information about a tree species such asan image and a localized description obtained fromDBpedia Species filters have a significant impactin the map view the color code of species filtersis applied to the different features while speciesdata is also displayed in the different tooltips andpopovers ndash see the snapshots in Figure 3In addition the form includes a textbox for

searching places (obtained from the GeoNamesdataset14) that sets the view of the map to thelocation of the selected place Unsurprisingly thelsquoScientific namesrsquo switch allows the user to choosebetween scientific and vulgar names of tree speciesOther controls are dependent on the zoom levelthe lsquoShow provincesrsquo switch is displayed with lowzoom levels so as to choose between the province ndashFigure 3(a) ndash and patch ndash Figure 3(b) ndash visualiza-tions With intermediate zoom levels the user canhide or show the plots in the map ndash this appliesto Figure 3(c) although the form is collapsed in

14httpwwwgeonamesorg

this snapshot Color saturation for plots can alsobe adjusted with a range input Finally it is pos-sible to highlight patches by land use eg to findplantation forests dehesas thickets and so on

44 Implementation

Forest Explorer is coded in JavaScript to facil-itate its deployment as a web application As aresult the tool can be used in any device with amodern web browser15

The code is organized in several files that re-flect the logical architecture in Figure 1 The im-plementation effort has been considerably reducedby the integration of a number of JavaScript li-braries We use Leaflet16 for the interactive mapthus fulfilling the role of the Map generator com-ponent (see Section 41) This library offers al-most all the mapping functionality needed for For-

15We have tested Forest Explorer with the latest versionsof Mozilla Firefox and Google Chrome in a variety of mobilephones tablets and desktop computers

16httpsleafletjscom

in [30] recommending to provide multiple resolu-tion versions of features at different zoom levelsThe Data manager handles all the data requests

from the Feature managers This component isable to communicate via SPARQL and is config-ured to use the Cross-Forest and DBpedia end-points As described in Section 3 the Cross-Forestdataset contains a Linked Open Data version ofthe Spanish forest inventory and land cover mapDBpedia is employed as a source of tree species in-formation providing images and multilingual de-scriptions (R6) Upon receipt of a request theData manager first checks if the result is alreadyavailable in the Data cache In case of a miss theData manager sends one or more SPARQL queriesto the endpoints Section 42 gives further detailsof the functioning of the Data manager

42 Data gathering

As depicted in Section 41 the Data manager isin charge of all data gathering operations in ForestExplorer The design of this component is chal-lenging due to a number of reasons (1) requestslook quite varied referring to different types offeatures (provinces patches plots and trees) andall their associated data (2) the size of the Cross-Forest dataset is not small ndash 43GB correspond-ing to 737M triples and (3) requests can be verynumerous since any change in the interactive mapwill trigger data requests by one or more FeaturemanagersFortunately the different request types can be

abstracted to the identification of features local-ized in a specific area and then obtaining their cor-responding feature data The Data manager usesSPARQL template queries for retrieving the fea-tures localized in the map view Since plots andtrees have points as geometries a suitable querybasically has to check which points are includedin a bounding box Listing 1 shows the tem-plate used for plots in which latnorthlatsouth lngwest lngeast areplaceholders for the map view bounds

Listing 1 SPARQL query template for retrievingplots in the map view

SELECT plot lat lngWHERE

plot a finvPlot

poshasPosition pos pos poshasCoordinateReferenceSystem crs4326

axis1 lat axis2 lng

FILTER (lat gt latsouth) FILTER (lat lt latnorth) FILTER (lng gt lngwest) FILTER (lng lt lngeast)

As patches and provinces have polygons as ge-ometries queries have to be adapted to find thepolygons intersecting with the map view The tem-plate used for patches is included in Listing 2Note that we use the bounding box of the poly-gon (included for all polygons in the dataset)to simplify the detection of patches in the mapview In order to comply with requirement R4the template is parametrized to select a specificlayer and to define a minimum threshold forthe area of polygons ie minarea The Datamanager chooses the most appropriate layer basedon the zoom level of the map view while it sets theminimum area to 10 pixels in the user device13

Listing 2 SPARQL query template for retrievingpatches in the map view

SELECT patch poly west east north south areaWHERE

patch a mapPatch poshasPolygon poly

poly epsghasLeftBound west epsghasRightBound east epsghasUpperBound north epsghasLowerBound south poshasAreaInSquareMeters area posisInLayer layer

FILTER (south lt latnorth) FILTER (north gt latsouth) FILTER (west lt lngeast) FILTER (east gt lngwest) FILTER (area gt minarea)

Once the Data manager obtains the set of fea-tures in the area of interest it retrieves the cor-responding data in a next step Listing 3 showsthe query template used for this purpose it is ex-

13The assumption here is that patches of less than 10pixels are barely visible so it is better to not waste timeand computation resources with them The area of a pixelchanges with the zoom level so a patch discarded for beingtoo small can be later displayed in case of zooming in

tremely generic and easily adaptable to each fea-ture type For example obtaining the height of acollection of trees just requires setting their IRIsand the height property IRI defined in the Cross-Forest ontology As a result gathering feature datais just a matter of selecting the set of properties toextract for each feature type (see Table 1) Classmembership is required in some cases ndash for exam-ple to obtain tree species ndash so we have includedanother generic template for retrieving the classesof a set of individuals

Listing 3 SPARQL query template for obtainingthe values of a property for a set of individuals

SELECT DISTINCT iri valueWHERE

iri propiri value VALUES iri iris

The Data manager stores all retrieved data ina Cache so as to reduce exchanges with the end-point This is quite effective with feature datasince the template query in Listing 3 can be eas-ily parametrized to avoid requests of feature dataalready available in the Cache With respect tofeature locations a similar approach consists oncaching the results for every map view bounds re-quest ie queries built with Listing 1 for plotsListing 2 for patches and so on However this so-lution is too naive as the cache will only get a hitin case of a request with exactly the same mapboundaries as a previous one Instead the Datamanager keeps track of the map regions with lo-calized features and exploits this information torestrict queries to unknown areas Figure 2 illus-trates the idea

(a) At the beginning the Data manager has no in-formation of any part of the map

(b) Upon receiving a feature location request forregion R0 it asks the endpoint about thefeatures in such area ndash the Data managerstores the results in the Cache and updatesits tracked area to the bounding box of R0

(c) A request about region R1 is decomposed intosubregions R1prime and R1primeprime ndash R1prime is the inter-section of R1 and the tracked area so theincluded features can be obtained from theCache Since R1primeprime is unknown to the Datamanager it has to query the endpoint to re-

trieve the features in R1primeprime Then the Datamanager updates its tracked area to R0 cup R1

(d) A request about region R2 does not requirefurther queries to the endpoint as R2 is con-tained in the tracked area

43 User interface

In order to comply with requirement R3 theuser interface of Forest Explorer exposes an inter-active map as its main component Figure 3 showssome snapshots that illustrate the design of theuser interface Since the tool has to work with a di-versity of devices (R1) the map is set in fullscreenmode to easily adapt to different screen sizes Sim-ilar to other map applications panning and zoom-ing are naturally supported for both point-and-click and touch interfaces In this regard zoombuttons are included in the lower-right corner theextra button with a person icon is used to navigateto the user location ndash this latter functionality isespecially useful when employing Forest Explorerwith a mobile device in a field trip

As the user navigates with the map a Featuremanager takes control by obtaining feature infor-mation from the Data manager and then sendingdisplay requests to the Map generator (see Sec-tion 41) For example the Province manager con-trols the view of Figure 3(a) the Patch manageris activated in Figure 3(b) the Patch and the Plotmanagers collaborate to build the view in Fig-ure 3(c) and the Tree manager is active in Fig-ure 3(d)

Beyond map navigation the user interface needsto include controls to allow the user to adjust theinformation to display (R5) For this purpose For-est Explorer includes a form in the upper-left partndash see Figure 3(a) This form can take a significantpart of the screen in mobile devices so it can becollapsed by pushing the lsquo-rsquo button ndash the collapsedform is shown in Figure 3(b)(c)(d)

A key characteristic of the dataset is the treespecies of land cover map and forestry inventoryThus the form includes a lsquoFilter speciesrsquo buttonthat can be pushed to easily browse the taxonomyof tree species (obtained from the Cross-Forest on-tology) and then select one or more filters Theform in Figure 3(a) includes two species filters Pi-nus sylvestris in indigo color and Pinus pinasterin brown Each filter includes buttons for removal(lsquoxrsquo icon) color change (lsquotintrsquo icon) and more info

Figure 2 The Data manager limits query exchanges with the endpoint by keeping track of the map regions with localizedfeatures The map is initially unknown to the Data manager (a) so a feature location request for region R0 (b) requiresquerying the endpoint ndash the Data manager stores the results in the Cache and updates its tracked area to R0 In (c) arequest about region R1 is decomposed into subregions R1prime and R1primeprime ndash the features in R1prime are obtained from the Cache andthe endpoint is queried to retrieve the features in R1primeprime The tracked area now includes R0 cup R1 As region R2 is containedin the tracked area the feature location request in (d) is answered with the Cache

(lsquoinforsquo icon) ndash the latter one displays a popup withadditional information about a tree species such asan image and a localized description obtained fromDBpedia Species filters have a significant impactin the map view the color code of species filtersis applied to the different features while speciesdata is also displayed in the different tooltips andpopovers ndash see the snapshots in Figure 3In addition the form includes a textbox for

searching places (obtained from the GeoNamesdataset14) that sets the view of the map to thelocation of the selected place Unsurprisingly thelsquoScientific namesrsquo switch allows the user to choosebetween scientific and vulgar names of tree speciesOther controls are dependent on the zoom levelthe lsquoShow provincesrsquo switch is displayed with lowzoom levels so as to choose between the province ndashFigure 3(a) ndash and patch ndash Figure 3(b) ndash visualiza-tions With intermediate zoom levels the user canhide or show the plots in the map ndash this appliesto Figure 3(c) although the form is collapsed in

14httpwwwgeonamesorg

this snapshot Color saturation for plots can alsobe adjusted with a range input Finally it is pos-sible to highlight patches by land use eg to findplantation forests dehesas thickets and so on

44 Implementation

Forest Explorer is coded in JavaScript to facil-itate its deployment as a web application As aresult the tool can be used in any device with amodern web browser15

The code is organized in several files that re-flect the logical architecture in Figure 1 The im-plementation effort has been considerably reducedby the integration of a number of JavaScript li-braries We use Leaflet16 for the interactive mapthus fulfilling the role of the Map generator com-ponent (see Section 41) This library offers al-most all the mapping functionality needed for For-

15We have tested Forest Explorer with the latest versionsof Mozilla Firefox and Google Chrome in a variety of mobilephones tablets and desktop computers

16httpsleafletjscom

tremely generic and easily adaptable to each fea-ture type For example obtaining the height of acollection of trees just requires setting their IRIsand the height property IRI defined in the Cross-Forest ontology As a result gathering feature datais just a matter of selecting the set of properties toextract for each feature type (see Table 1) Classmembership is required in some cases ndash for exam-ple to obtain tree species ndash so we have includedanother generic template for retrieving the classesof a set of individuals

Listing 3 SPARQL query template for obtainingthe values of a property for a set of individuals

SELECT DISTINCT iri valueWHERE

iri propiri value VALUES iri iris

The Data manager stores all retrieved data ina Cache so as to reduce exchanges with the end-point This is quite effective with feature datasince the template query in Listing 3 can be eas-ily parametrized to avoid requests of feature dataalready available in the Cache With respect tofeature locations a similar approach consists oncaching the results for every map view bounds re-quest ie queries built with Listing 1 for plotsListing 2 for patches and so on However this so-lution is too naive as the cache will only get a hitin case of a request with exactly the same mapboundaries as a previous one Instead the Datamanager keeps track of the map regions with lo-calized features and exploits this information torestrict queries to unknown areas Figure 2 illus-trates the idea

(a) At the beginning the Data manager has no in-formation of any part of the map

(b) Upon receiving a feature location request forregion R0 it asks the endpoint about thefeatures in such area ndash the Data managerstores the results in the Cache and updatesits tracked area to the bounding box of R0

(c) A request about region R1 is decomposed intosubregions R1prime and R1primeprime ndash R1prime is the inter-section of R1 and the tracked area so theincluded features can be obtained from theCache Since R1primeprime is unknown to the Datamanager it has to query the endpoint to re-

trieve the features in R1primeprime Then the Datamanager updates its tracked area to R0 cup R1

(d) A request about region R2 does not requirefurther queries to the endpoint as R2 is con-tained in the tracked area

43 User interface

In order to comply with requirement R3 theuser interface of Forest Explorer exposes an inter-active map as its main component Figure 3 showssome snapshots that illustrate the design of theuser interface Since the tool has to work with a di-versity of devices (R1) the map is set in fullscreenmode to easily adapt to different screen sizes Sim-ilar to other map applications panning and zoom-ing are naturally supported for both point-and-click and touch interfaces In this regard zoombuttons are included in the lower-right corner theextra button with a person icon is used to navigateto the user location ndash this latter functionality isespecially useful when employing Forest Explorerwith a mobile device in a field trip

As the user navigates with the map a Featuremanager takes control by obtaining feature infor-mation from the Data manager and then sendingdisplay requests to the Map generator (see Sec-tion 41) For example the Province manager con-trols the view of Figure 3(a) the Patch manageris activated in Figure 3(b) the Patch and the Plotmanagers collaborate to build the view in Fig-ure 3(c) and the Tree manager is active in Fig-ure 3(d)

Beyond map navigation the user interface needsto include controls to allow the user to adjust theinformation to display (R5) For this purpose For-est Explorer includes a form in the upper-left partndash see Figure 3(a) This form can take a significantpart of the screen in mobile devices so it can becollapsed by pushing the lsquo-rsquo button ndash the collapsedform is shown in Figure 3(b)(c)(d)

A key characteristic of the dataset is the treespecies of land cover map and forestry inventoryThus the form includes a lsquoFilter speciesrsquo buttonthat can be pushed to easily browse the taxonomyof tree species (obtained from the Cross-Forest on-tology) and then select one or more filters Theform in Figure 3(a) includes two species filters Pi-nus sylvestris in indigo color and Pinus pinasterin brown Each filter includes buttons for removal(lsquoxrsquo icon) color change (lsquotintrsquo icon) and more info

Figure 2 The Data manager limits query exchanges with the endpoint by keeping track of the map regions with localizedfeatures The map is initially unknown to the Data manager (a) so a feature location request for region R0 (b) requiresquerying the endpoint ndash the Data manager stores the results in the Cache and updates its tracked area to R0 In (c) arequest about region R1 is decomposed into subregions R1prime and R1primeprime ndash the features in R1prime are obtained from the Cache andthe endpoint is queried to retrieve the features in R1primeprime The tracked area now includes R0 cup R1 As region R2 is containedin the tracked area the feature location request in (d) is answered with the Cache

(lsquoinforsquo icon) ndash the latter one displays a popup withadditional information about a tree species such asan image and a localized description obtained fromDBpedia Species filters have a significant impactin the map view the color code of species filtersis applied to the different features while speciesdata is also displayed in the different tooltips andpopovers ndash see the snapshots in Figure 3In addition the form includes a textbox for

searching places (obtained from the GeoNamesdataset14) that sets the view of the map to thelocation of the selected place Unsurprisingly thelsquoScientific namesrsquo switch allows the user to choosebetween scientific and vulgar names of tree speciesOther controls are dependent on the zoom levelthe lsquoShow provincesrsquo switch is displayed with lowzoom levels so as to choose between the province ndashFigure 3(a) ndash and patch ndash Figure 3(b) ndash visualiza-tions With intermediate zoom levels the user canhide or show the plots in the map ndash this appliesto Figure 3(c) although the form is collapsed in

14httpwwwgeonamesorg

this snapshot Color saturation for plots can alsobe adjusted with a range input Finally it is pos-sible to highlight patches by land use eg to findplantation forests dehesas thickets and so on

44 Implementation

Forest Explorer is coded in JavaScript to facil-itate its deployment as a web application As aresult the tool can be used in any device with amodern web browser15

The code is organized in several files that re-flect the logical architecture in Figure 1 The im-plementation effort has been considerably reducedby the integration of a number of JavaScript li-braries We use Leaflet16 for the interactive mapthus fulfilling the role of the Map generator com-ponent (see Section 41) This library offers al-most all the mapping functionality needed for For-

15We have tested Forest Explorer with the latest versionsof Mozilla Firefox and Google Chrome in a variety of mobilephones tablets and desktop computers

16httpsleafletjscom

Figure 2 The Data manager limits query exchanges with the endpoint by keeping track of the map regions with localizedfeatures The map is initially unknown to the Data manager (a) so a feature location request for region R0 (b) requiresquerying the endpoint ndash the Data manager stores the results in the Cache and updates its tracked area to R0 In (c) arequest about region R1 is decomposed into subregions R1prime and R1primeprime ndash the features in R1prime are obtained from the Cache andthe endpoint is queried to retrieve the features in R1primeprime The tracked area now includes R0 cup R1 As region R2 is containedin the tracked area the feature location request in (d) is answered with the Cache

(lsquoinforsquo icon) ndash the latter one displays a popup withadditional information about a tree species such asan image and a localized description obtained fromDBpedia Species filters have a significant impactin the map view the color code of species filtersis applied to the different features while speciesdata is also displayed in the different tooltips andpopovers ndash see the snapshots in Figure 3In addition the form includes a textbox for

searching places (obtained from the GeoNamesdataset14) that sets the view of the map to thelocation of the selected place Unsurprisingly thelsquoScientific namesrsquo switch allows the user to choosebetween scientific and vulgar names of tree speciesOther controls are dependent on the zoom levelthe lsquoShow provincesrsquo switch is displayed with lowzoom levels so as to choose between the province ndashFigure 3(a) ndash and patch ndash Figure 3(b) ndash visualiza-tions With intermediate zoom levels the user canhide or show the plots in the map ndash this appliesto Figure 3(c) although the form is collapsed in

14httpwwwgeonamesorg

this snapshot Color saturation for plots can alsobe adjusted with a range input Finally it is pos-sible to highlight patches by land use eg to findplantation forests dehesas thickets and so on

44 Implementation

Forest Explorer is coded in JavaScript to facil-itate its deployment as a web application As aresult the tool can be used in any device with amodern web browser15

The code is organized in several files that re-flect the logical architecture in Figure 1 The im-plementation effort has been considerably reducedby the integration of a number of JavaScript li-braries We use Leaflet16 for the interactive mapthus fulfilling the role of the Map generator com-ponent (see Section 41) This library offers al-most all the mapping functionality needed for For-

15We have tested Forest Explorer with the latest versionsof Mozilla Firefox and Google Chrome in a variety of mobilephones tablets and desktop computers

16httpsleafletjscom

Figure 3 Snapshots of the user interface of Forest Explorer (a) View of the Spanish provinces in a large area (see the mapscale in the lower-right corner) with the form in the upper-left part expanded and showing two species filters Pinus sylvestrisin indigo color and Pinus pinaster in brown inventory tree data for the province of Soria is displayed in a tooltip (b) Viewof the patches in a large area of Spain with the form in the upper-left part collapsed patches are plotted in different colorsdepending on their use (farms in orange water in blue artificial in grey and forests in green) forest patches containing afiltered species use the color filter (indigo and brown in this running example) a pop-up shows the data of a forest patch inthe province of Soria (c) View of a small forest area (see the map scale in the lower-right corner) in the province of Soriaplots are displayed as circles on top of the patches plots and patches employ the same color code as before a tooltip showsinventory data for a plot (d) View of a tiny small forest area corresponding to the center of a plot in Soria tree markersare shown in their actual positions using taxa-dependent icons and corresponding filter colors a tooltip shows the speciesheight and diameter of a specific tree

est Explorer vector layers for plotting featuresmarkers popups tooltips map controls and inter-

action capabilities We use two additional Leaflet

plug-ins LeafletLocate17 to geolocate the userand LeafletCircle-sector18 for drawing pie-shapedplots when filtering multiple tree species ndash see Fig-ure 3(c) We use the custom Mapbox Light map19

as a background for displaying forestry data on topndash this is the Map server component in Section 41The form of the user interface is built with Boot-

strap20 to simplify front-end development acrossdifferent browsers and devices We use jQuery21

for event handling and manipulating the Docu-ment Object Model [34] We also employ the util-ity functions of Underscore22 for handling collec-tions We use Mustache23 for templating SPARQLqueries (see Section 42) Lastly Google Analyt-ics24 is included to keep track of the usage of For-est ExplorerAs for the Data manager component all the

queries are included in a separate file using tem-plating parameters as necessary eg Listings 1 2and 3 We also use a mapping file that assigns keysto ontology IRIs the Data manager only referencesthe keys and this level of indirection decouples theimplementation from the Cross-Forest ontology asa resultSince the tool needs to be localized to English

and Spanish (requirement R6) the Cross-Forestdataset is localized to these languages and all thelabels employed in the user interface are includedin a multilingual strings file When the tool startsit gets the browser language preferences in orderto select the session language that is then appliedto every text shownThe source code of Forest Explorer is available

on GitHub25 We have also set up a live version ofthe tool26 for anybody who wants to use it All inall Forest Explorer can be used with different de-vices (R1) the user interface hides the intricaciesof Semantic Web languages from the user (R2) aninteractive map (R3) is employed to present thedata adapted to different zoom levels (R4) with a

17httpsgithubcomdomoritzleaflet-locatecontrol18httpsgithubcomkluizebergLeafletCircle-sector19httpswwwmapboxcommapslight-dark20httpsgetbootstrapcom21httpsjquerycom22httpsunderscorejsorg23httpsmustachegithubio24httpsmarketingplatformgooglecomaboutanalyti

cs25httpsgithubcomCross-Forestforestexplorer26httpsforestexplorergsicuvaes

number of filtering capabilities (R5) and localizedto English and Spanish (R6) as elaborated above

5 Impact

We submitted a preliminary prototype of For-est Explorer to Desafiacuteo Aporta 201927 an opendata challenge on agrofood forestry and rural ar-eas that was sponsored by the Spanish Ministry ofEconomy This challenge received 40 proposals andours was shortlisted to a final panel although wedid not get an award The communication boardof Universidad de Valladolid prepared a press re-lease about Forest Explorer28 reaching the finalround of Desafiacuteo Aporta 2019 in December 2019At that time the live site of Forest Explorer wasopenly available and we shared the website linkwith our contacts Spanish local media publishedseveral articles about the tool29 In addition twonewspapers published long interviews with us cov-ering Forest explorer30

In January 2020 we used our social networks toannounce the release of Forest Explorer Specifi-cally we sent 8 tweets in Twitter that received over6500 impressions more than 100 likes and about30 retweets We also prepared three short posts inFacebook and one more in Twitter receiving over1000 and 375 impressions respectively

Beyond media coverage and social networks weused Google Analytics to assess the uptake of For-est Explorer Tracked data was obtained from thelive site in October 2020 About usage more than3200 users have accessed the tool and have car-ried out over 5500 sessions with an average sessiontime of 4 minutes and 11 seconds 85 of the traf-fic comes from Spain and the preferred languagewas Spanish in 82 of the sessions ndash this is not sur-prising as the dataset only covers Spanish forests

27httpsdatosgobesesdesafios-aportadesafio-aporta-2019

28httpscomunicacionuvaeses_ESdetalleUna-herramienta-web-desarrollada-por-miembros-de-la-Universidad-de-Valladolid-que-facilita-la-exploracion-del-inventario-forestal-finalista-del-Desafio-Aporta-2019

29For example httpswwwefecomefecastillayleonsociedadcrean-un-explorador-forestal-para-hacer-seguimiento-de-incendios-y-plagas50000473-4138848

30httpwwwdiariodevalladolidesnoticiasinnovadoresarboles-golpe-clic_170743html andhttpleeleconomistaesrtsgo2aspxh=260137amptp=i-H43-Dc-6T8-FyPmJ-1c-1gVE-1c-FyOR8-1NwpSj

plug-ins LeafletLocate17 to geolocate the userand LeafletCircle-sector18 for drawing pie-shapedplots when filtering multiple tree species ndash see Fig-ure 3(c) We use the custom Mapbox Light map19

as a background for displaying forestry data on topndash this is the Map server component in Section 41The form of the user interface is built with Boot-

strap20 to simplify front-end development acrossdifferent browsers and devices We use jQuery21

for event handling and manipulating the Docu-ment Object Model [34] We also employ the util-ity functions of Underscore22 for handling collec-tions We use Mustache23 for templating SPARQLqueries (see Section 42) Lastly Google Analyt-ics24 is included to keep track of the usage of For-est ExplorerAs for the Data manager component all the

queries are included in a separate file using tem-plating parameters as necessary eg Listings 1 2and 3 We also use a mapping file that assigns keysto ontology IRIs the Data manager only referencesthe keys and this level of indirection decouples theimplementation from the Cross-Forest ontology asa resultSince the tool needs to be localized to English

and Spanish (requirement R6) the Cross-Forestdataset is localized to these languages and all thelabels employed in the user interface are includedin a multilingual strings file When the tool startsit gets the browser language preferences in orderto select the session language that is then appliedto every text shownThe source code of Forest Explorer is available

on GitHub25 We have also set up a live version ofthe tool26 for anybody who wants to use it All inall Forest Explorer can be used with different de-vices (R1) the user interface hides the intricaciesof Semantic Web languages from the user (R2) aninteractive map (R3) is employed to present thedata adapted to different zoom levels (R4) with a

17httpsgithubcomdomoritzleaflet-locatecontrol18httpsgithubcomkluizebergLeafletCircle-sector19httpswwwmapboxcommapslight-dark20httpsgetbootstrapcom21httpsjquerycom22httpsunderscorejsorg23httpsmustachegithubio24httpsmarketingplatformgooglecomaboutanalyti

cs25httpsgithubcomCross-Forestforestexplorer26httpsforestexplorergsicuvaes

number of filtering capabilities (R5) and localizedto English and Spanish (R6) as elaborated above

5 Impact

We submitted a preliminary prototype of For-est Explorer to Desafiacuteo Aporta 201927 an opendata challenge on agrofood forestry and rural ar-eas that was sponsored by the Spanish Ministry ofEconomy This challenge received 40 proposals andours was shortlisted to a final panel although wedid not get an award The communication boardof Universidad de Valladolid prepared a press re-lease about Forest Explorer28 reaching the finalround of Desafiacuteo Aporta 2019 in December 2019At that time the live site of Forest Explorer wasopenly available and we shared the website linkwith our contacts Spanish local media publishedseveral articles about the tool29 In addition twonewspapers published long interviews with us cov-ering Forest explorer30

In January 2020 we used our social networks toannounce the release of Forest Explorer Specifi-cally we sent 8 tweets in Twitter that received over6500 impressions more than 100 likes and about30 retweets We also prepared three short posts inFacebook and one more in Twitter receiving over1000 and 375 impressions respectively

Beyond media coverage and social networks weused Google Analytics to assess the uptake of For-est Explorer Tracked data was obtained from thelive site in October 2020 About usage more than3200 users have accessed the tool and have car-ried out over 5500 sessions with an average sessiontime of 4 minutes and 11 seconds 85 of the traf-fic comes from Spain and the preferred languagewas Spanish in 82 of the sessions ndash this is not sur-prising as the dataset only covers Spanish forests

27httpsdatosgobesesdesafios-aportadesafio-aporta-2019

28httpscomunicacionuvaeses_ESdetalleUna-herramienta-web-desarrollada-por-miembros-de-la-Universidad-de-Valladolid-que-facilita-la-exploracion-del-inventario-forestal-finalista-del-Desafio-Aporta-2019

29For example httpswwwefecomefecastillayleonsociedadcrean-un-explorador-forestal-para-hacer-seguimiento-de-incendios-y-plagas50000473-4138848

30httpwwwdiariodevalladolidesnoticiasinnovadoresarboles-golpe-clic_170743html andhttpleeleconomistaesrtsgo2aspxh=260137amptp=i-H43-Dc-6T8-FyPmJ-1c-1gVE-1c-FyOR8-1NwpSj

Google Analytics gives also information about theemployed devices 43 were Windows computers38 Android mobiles or tablets 12 iOS devices6 Mac computers and 1 Linux computers

51 User study

We have carried out a user study of Forest Ex-plorer that was promoted by the Cross-Forest con-sortium ndash we participated in the preparation of thequestionnaire but we did not reach potential re-spondents The invitation to take part in this userstudy included a link to Forest Explorer to test itas well as the link to the questionnaire This wasdivided in four parts (1) User profile (2) Datasetassessment (3) Usability through the standard-ized System Usability Score (SUS) [6] and (4)User feedback28 participants have answered the questionnaire

so far ndash information about their user profiles issummarized in Table 3 Note that most of themcan be classified as forestry domain experts (93)while the remaining 7 correspond to the usergroup of lay users Moving to the data assess-ment section participants rated the usefulness ofthe data exposed with an average figure of 39and a standard deviation of 08 in a scale from1 (not useful) to 5 (very useful) They were alsoasked about the understandability of the data ex-posed (from 1 not understandable to 5 very un-derstandable) the average was 39 and the stan-dard deviation 09 Participants were also asked tooptionally propose new datasets to be integratedfour suggested the inclusion of orthophotos as basemap three proposed the inclusion of additionalforestry data (combustibility biomass carbon se-questration) two proposed the inclusion of altime-try data and another proposed the inclusion ofprevious editions of forest inventoriesRegarding usability the computed SUS score

was 75 in average with a standard deviation of 16This figure is good given that SUS scores rangefrom 0 to 100 According to the grading scale in-terpretation of SUS scores in [26 ch 8] ForestExplorer was graded with a B Interestingly thegroup of Spanish respondents gave higher scores(81 in average and an A in the aforementionedgrading scale) than the participants from othercountries mainly Portugal with a SUS score of 69in average

Table 3Profiles of the participants in the user study (first part ofthe questionnaire)

Way of contact

Email 18 64Colleagues 8 29Project website 2 7

Country

Spain 15 54Portugal 12 43France 1 4

Group sector

Public Administration 17 61Academics 5 18Forest professionals 4 14Citizens 2 7

Role31

Data consumer 17 61Data provider 11 39Service provider 4 14

Expertise level (1ndash5) avg std

Using forestry data 31 13Using geodatabases 37 09

The fourth part of the questionnaire began witha question about the likeliness of recommendingForest Explorer to a friend or colleague (from 1not likely at all to 5 very likely) the averagewas 40 with a standard deviation of 09 Thisblock also contained two open-ended questionsabout what liked most and liked least of ForestExplorer We analyzed the results by identifyingseveral themes and categorizing the answers Themain findings are included in Table 4 with theirlevel of support and a sample comment for eachfinding ndash note that we only include a point if sup-ported by at least two participants so as to avoidirrelevant or spurious judgments

With respect to the tool strengths 43 of therespondents stressed its ease of use and 21 itsusability ndash connected to requirement R2 and one

30It was possible to select multiple options to this ques-tion 86 of the participants chose one of the three optionsgiven and 14 marked two

Table 4Strengths and weaknesses of Forest Explorer (fourth part of the questionnaire)

Point Support Sample comment

+ Easy to use 1228 Easy access to heavy detailed databases [P27]+ Good usability 628 Interface (simple and visually attractive) and optimization (fast data load-

ing) [P6]+ Fast 628 Fast loading of information layers [P3]+ Data integration 428 Unified presentation of all data in MFE and IFN at national level with fast

and exhaustive search The possibility of filtering one two three specieswith a very visual search of plots and patches [P15]

+ Search capabilities 328 Compound search of several species [P16]minus Portuguese data missing 528 Not having information about Portugal [P1]minus Data not downloadable 428 Not possible to download data [P18]minus Lack of background maps 328 There are no different maps you cannot know the coordinates download

data in another format (Excel pdf) [P7]minus Provide more info 328 I miss a user manual [P4]minus Somewhat slow 328 Sometimes it takes time until filters are applied [P28]minus Aesthetics 228 Color of maps [P23]

of the main goals of Forest Explorer 21 consid-ered the tool fast ndash a good sign given the efforton efficient data gathering (see Section 32) Fourparticipants included positive comments about theintegration of forest databases through Forest Ex-plorer Three liked the search capabilities of thetool ndash connected to requirement R5 On the neg-ative side we obtained very useful feedback forimproving Forest Explorer 18 of the respon-dents missed Portuguese data ndash this limitationmay partially explain the lower SUS scores of Por-tuguese participants identified above New featurerequests include a downloading data functionalityproviding background maps (see also the analy-sis of the second part of the questionnaire above)and additional information like a user manual orprovenance data Besides some participants re-ported speed problems and minor aesthetics mod-ificationsAll in all these results are generally positive

supporting the design goals of Forest ExplorerParticipants mainly correspond to the group offorest domain experts They were able to use thetool for exploring a complex semantic dataset in-tegrating forest inventories and land cover mapsand considered Forest Explorer easy to use Us-ability was ranked good ndash this is consistent withSUS scores and answers to open-ended questionsNo important limitations were found they weremostly feature requests that are quite useful to

guide the development of new versions of ForestExplorer

6 Discussion

Forest Explorer is designed to work with theCross-Forest dataset thus benefitting from the useof Semantic Web technologies for data integra-tion Indeed this advantage was identified in theuser study (see Table 4) although participantsstill demanded additional sources to be includedndash it looks data integration is much needed in theforestry domain In this regard we are currentlyworking on the conversion of the Portuguese forestinventory and land cover map into Linked OpenData Once included in the Cross-Forest datasetthey will be automatically browsable through For-est Explorer Despite the schemata of the origi-nal Portuguese and Spanish sources are differentthe overarching ontology homogenizes the termi-nology so no further changes will be required inForest Explorer

Moving on to the technical design of Forest Ex-plorer the Data manager is the component incharge of gathering data from the Cross-Forest andDBpedia endpoints (see Figure 1) The solutiondevised relies on the use of SPARQL query tem-plates to facilitate the extensibility of Forest Ex-plorer As an example the template query in List-ing 3 is employed for gathering every type of fea-

ture data and also for obtaining images and treespecies descriptions from DBpedia Similarly theData manager uses a query template for obtainingthe taxonomy of subclasses of a target class this isemployed with species and soil uses FurthermoreSection 42 already describes the query templatesused for retrieving plots and patches as well as theCache proposed for reducing query exchanges withthe Cross-Forest endpoint Such query templatescan be abstracted away to retrieve arbitrary fea-tures in order to extend the applicability of ForestExplorer to other contextsGeospatial queries in Forest Explorer are han-

dled with the SPARQL templates in Listings 1ndash2 They are simple interval queries that run quitefast and provide the necessary expressiveness forthe retrieval of features in a bounding box Alter-natively GeoSPARQL functions such as sfWithincould be employed for this purpose ndash this approachwas discarded because the Cross-Forest triple-store provides custom geospatial functions butGeoSPARQL functions are not supported We ac-knowledge that GeoSPARQL is quite powerful andconvenient for conducting geospatial operationsand for this reason the Cross-Forest dataset pro-vides geometries expressed in Well Known TextWe envision the use of GeoSPARQL for new forestscenarios in the near futureThe different Feature managers in Forest Ex-

plorer are bound to the corresponding featuretypes ie province patch plot and tree Theycan be easily modified so as to prepare new la-bels for tooltips to request different markers tobe rendered by the Map generator etc Moreovernew Feature managers can be added to the sys-tem the recommended way is (1) to use an exist-ing Feature manager as a base and (2) to updatethe Data manager for gathering feature data ndash ex-isting query templates should be sufficient in mostcasesTo wrap up Forest Explorer is a novel explo-

ration tool of forestry Linked Open Data designedfor non-Semantic Web experts Users can interactwith a map to navigate to the area of interest andto control the information to display To limit dataexchanges with the SPARQL endpoint Forest ex-plorer uses a cache and smart geographic query-ing So far we have attracted the attention of theforestry community not very familiar with Seman-tic Web technologies Our future work includes theintegration of new data sources from other coun-

tries in the Cross-Forest dataset beginning withPortugal Other geolocated data can be integratedsuch as cadastral parcel information digital ter-rain models land use maps remote sensing lay-ers and hazard occurrence or vulnerability (for-est fires pests and diseases or blown down dam-ages) We also plan to introduce data analysis ca-pabilities to support forest management and plan-ning scenarios In this way Forest Explorer couldprovide similar functionalities as BASIFOR butwith strong visualizations and taking advantage ofLinked Open Data for forestry data integration

Acknowledgements

This work has been partially funded by the Eu-ropean Commission through Cross-Forest (2017-EU-IA-0140) and Spanish Ministry of Science andInnovation through REFORM (PCIN-2017-027ERANET SUMFOREST) projects

References

[1] W Beek E Folmer L Rietveld and J Walker GeoY-ASGUI The GeoSPARQL query editor and result setvisualizer The International Archives of Photogram-metry Remote Sensing and Spatial Information Sci-ences 42 (2017) 39ndash42

[2] F Bravo M del Riacuteo and C del Peso (eds) El In-ventario Forestal Nacional Elemento clave para lagestioacuten forestal sostenible Fundacioacuten General de laUniversidad de Valladolid 2002

[3] F Bravo JC Rivas JA Monreal-Nuacutentildeez and C Or-doacutentildeez BASIFOR 20 Aplicacioacuten informaacutetica para elmanejo de las bases de datos del inventario forestal na-cional Cuadernos de la Sociedad Espantildeola de Cien-cias Forestales (2004) 243ndash247

[4] F Bravo M Fabrika C Ammer S BarreiroK Bielak L Coll T Fonseca A Kangur M LofK Merganicova M Pach H Pretzsch D StojanovicL Schuler S Peric T Rotzer M del Riacuteo M Dodanand A Bravo-Oviedo Modelling approaches for mixedforests dynamics prognosis Research gaps and oppor-tunities Forest Systems 28(1) (2019) ISSN 21715068doi105424fs2019281-14342

[5] D Brickley Basic Geo (WGS84 latlong) VocabularyTechnical Report World Wide Web Consortium 2006URL httpswwww3org200301geo last visitedApril 2020

[6] J Brooke SUS ndash A quick and dirty usability scalein Usability evaluation in industry PW JordanB Thomas IL McClelland and B Weerdmeester edsTaylor amp Francis London UK 1996

Table 4Strengths and weaknesses of Forest Explorer (fourth part of the questionnaire)

Point Support Sample comment

+ Easy to use 1228 Easy access to heavy detailed databases [P27]+ Good usability 628 Interface (simple and visually attractive) and optimization (fast data load-

ing) [P6]+ Fast 628 Fast loading of information layers [P3]+ Data integration 428 Unified presentation of all data in MFE and IFN at national level with fast

and exhaustive search The possibility of filtering one two three specieswith a very visual search of plots and patches [P15]

+ Search capabilities 328 Compound search of several species [P16]minus Portuguese data missing 528 Not having information about Portugal [P1]minus Data not downloadable 428 Not possible to download data [P18]minus Lack of background maps 328 There are no different maps you cannot know the coordinates download

data in another format (Excel pdf) [P7]minus Provide more info 328 I miss a user manual [P4]minus Somewhat slow 328 Sometimes it takes time until filters are applied [P28]minus Aesthetics 228 Color of maps [P23]

of the main goals of Forest Explorer 21 consid-ered the tool fast ndash a good sign given the efforton efficient data gathering (see Section 32) Fourparticipants included positive comments about theintegration of forest databases through Forest Ex-plorer Three liked the search capabilities of thetool ndash connected to requirement R5 On the neg-ative side we obtained very useful feedback forimproving Forest Explorer 18 of the respon-dents missed Portuguese data ndash this limitationmay partially explain the lower SUS scores of Por-tuguese participants identified above New featurerequests include a downloading data functionalityproviding background maps (see also the analy-sis of the second part of the questionnaire above)and additional information like a user manual orprovenance data Besides some participants re-ported speed problems and minor aesthetics mod-ificationsAll in all these results are generally positive

supporting the design goals of Forest ExplorerParticipants mainly correspond to the group offorest domain experts They were able to use thetool for exploring a complex semantic dataset in-tegrating forest inventories and land cover mapsand considered Forest Explorer easy to use Us-ability was ranked good ndash this is consistent withSUS scores and answers to open-ended questionsNo important limitations were found they weremostly feature requests that are quite useful to

guide the development of new versions of ForestExplorer

6 Discussion

Forest Explorer is designed to work with theCross-Forest dataset thus benefitting from the useof Semantic Web technologies for data integra-tion Indeed this advantage was identified in theuser study (see Table 4) although participantsstill demanded additional sources to be includedndash it looks data integration is much needed in theforestry domain In this regard we are currentlyworking on the conversion of the Portuguese forestinventory and land cover map into Linked OpenData Once included in the Cross-Forest datasetthey will be automatically browsable through For-est Explorer Despite the schemata of the origi-nal Portuguese and Spanish sources are differentthe overarching ontology homogenizes the termi-nology so no further changes will be required inForest Explorer

Moving on to the technical design of Forest Ex-plorer the Data manager is the component incharge of gathering data from the Cross-Forest andDBpedia endpoints (see Figure 1) The solutiondevised relies on the use of SPARQL query tem-plates to facilitate the extensibility of Forest Ex-plorer As an example the template query in List-ing 3 is employed for gathering every type of fea-

ture data and also for obtaining images and treespecies descriptions from DBpedia Similarly theData manager uses a query template for obtainingthe taxonomy of subclasses of a target class this isemployed with species and soil uses FurthermoreSection 42 already describes the query templatesused for retrieving plots and patches as well as theCache proposed for reducing query exchanges withthe Cross-Forest endpoint Such query templatescan be abstracted away to retrieve arbitrary fea-tures in order to extend the applicability of ForestExplorer to other contextsGeospatial queries in Forest Explorer are han-

dled with the SPARQL templates in Listings 1ndash2 They are simple interval queries that run quitefast and provide the necessary expressiveness forthe retrieval of features in a bounding box Alter-natively GeoSPARQL functions such as sfWithincould be employed for this purpose ndash this approachwas discarded because the Cross-Forest triple-store provides custom geospatial functions butGeoSPARQL functions are not supported We ac-knowledge that GeoSPARQL is quite powerful andconvenient for conducting geospatial operationsand for this reason the Cross-Forest dataset pro-vides geometries expressed in Well Known TextWe envision the use of GeoSPARQL for new forestscenarios in the near futureThe different Feature managers in Forest Ex-

plorer are bound to the corresponding featuretypes ie province patch plot and tree Theycan be easily modified so as to prepare new la-bels for tooltips to request different markers tobe rendered by the Map generator etc Moreovernew Feature managers can be added to the sys-tem the recommended way is (1) to use an exist-ing Feature manager as a base and (2) to updatethe Data manager for gathering feature data ndash ex-isting query templates should be sufficient in mostcasesTo wrap up Forest Explorer is a novel explo-

ration tool of forestry Linked Open Data designedfor non-Semantic Web experts Users can interactwith a map to navigate to the area of interest andto control the information to display To limit dataexchanges with the SPARQL endpoint Forest ex-plorer uses a cache and smart geographic query-ing So far we have attracted the attention of theforestry community not very familiar with Seman-tic Web technologies Our future work includes theintegration of new data sources from other coun-

tries in the Cross-Forest dataset beginning withPortugal Other geolocated data can be integratedsuch as cadastral parcel information digital ter-rain models land use maps remote sensing lay-ers and hazard occurrence or vulnerability (for-est fires pests and diseases or blown down dam-ages) We also plan to introduce data analysis ca-pabilities to support forest management and plan-ning scenarios In this way Forest Explorer couldprovide similar functionalities as BASIFOR butwith strong visualizations and taking advantage ofLinked Open Data for forestry data integration

Acknowledgements

This work has been partially funded by the Eu-ropean Commission through Cross-Forest (2017-EU-IA-0140) and Spanish Ministry of Science andInnovation through REFORM (PCIN-2017-027ERANET SUMFOREST) projects

References

[1] W Beek E Folmer L Rietveld and J Walker GeoY-ASGUI The GeoSPARQL query editor and result setvisualizer The International Archives of Photogram-metry Remote Sensing and Spatial Information Sci-ences 42 (2017) 39ndash42

[2] F Bravo M del Riacuteo and C del Peso (eds) El In-ventario Forestal Nacional Elemento clave para lagestioacuten forestal sostenible Fundacioacuten General de laUniversidad de Valladolid 2002

[3] F Bravo JC Rivas JA Monreal-Nuacutentildeez and C Or-doacutentildeez BASIFOR 20 Aplicacioacuten informaacutetica para elmanejo de las bases de datos del inventario forestal na-cional Cuadernos de la Sociedad Espantildeola de Cien-cias Forestales (2004) 243ndash247

[4] F Bravo M Fabrika C Ammer S BarreiroK Bielak L Coll T Fonseca A Kangur M LofK Merganicova M Pach H Pretzsch D StojanovicL Schuler S Peric T Rotzer M del Riacuteo M Dodanand A Bravo-Oviedo Modelling approaches for mixedforests dynamics prognosis Research gaps and oppor-tunities Forest Systems 28(1) (2019) ISSN 21715068doi105424fs2019281-14342

[5] D Brickley Basic Geo (WGS84 latlong) VocabularyTechnical Report World Wide Web Consortium 2006URL httpswwww3org200301geo last visitedApril 2020

[6] J Brooke SUS ndash A quick and dirty usability scalein Usability evaluation in industry PW JordanB Thomas IL McClelland and B Weerdmeester edsTaylor amp Francis London UK 1996

ture data and also for obtaining images and treespecies descriptions from DBpedia Similarly theData manager uses a query template for obtainingthe taxonomy of subclasses of a target class this isemployed with species and soil uses FurthermoreSection 42 already describes the query templatesused for retrieving plots and patches as well as theCache proposed for reducing query exchanges withthe Cross-Forest endpoint Such query templatescan be abstracted away to retrieve arbitrary fea-tures in order to extend the applicability of ForestExplorer to other contextsGeospatial queries in Forest Explorer are han-

dled with the SPARQL templates in Listings 1ndash2 They are simple interval queries that run quitefast and provide the necessary expressiveness forthe retrieval of features in a bounding box Alter-natively GeoSPARQL functions such as sfWithincould be employed for this purpose ndash this approachwas discarded because the Cross-Forest triple-store provides custom geospatial functions butGeoSPARQL functions are not supported We ac-knowledge that GeoSPARQL is quite powerful andconvenient for conducting geospatial operationsand for this reason the Cross-Forest dataset pro-vides geometries expressed in Well Known TextWe envision the use of GeoSPARQL for new forestscenarios in the near futureThe different Feature managers in Forest Ex-

plorer are bound to the corresponding featuretypes ie province patch plot and tree Theycan be easily modified so as to prepare new la-bels for tooltips to request different markers tobe rendered by the Map generator etc Moreovernew Feature managers can be added to the sys-tem the recommended way is (1) to use an exist-ing Feature manager as a base and (2) to updatethe Data manager for gathering feature data ndash ex-isting query templates should be sufficient in mostcasesTo wrap up Forest Explorer is a novel explo-

ration tool of forestry Linked Open Data designedfor non-Semantic Web experts Users can interactwith a map to navigate to the area of interest andto control the information to display To limit dataexchanges with the SPARQL endpoint Forest ex-plorer uses a cache and smart geographic query-ing So far we have attracted the attention of theforestry community not very familiar with Seman-tic Web technologies Our future work includes theintegration of new data sources from other coun-

tries in the Cross-Forest dataset beginning withPortugal Other geolocated data can be integratedsuch as cadastral parcel information digital ter-rain models land use maps remote sensing lay-ers and hazard occurrence or vulnerability (for-est fires pests and diseases or blown down dam-ages) We also plan to introduce data analysis ca-pabilities to support forest management and plan-ning scenarios In this way Forest Explorer couldprovide similar functionalities as BASIFOR butwith strong visualizations and taking advantage ofLinked Open Data for forestry data integration

Acknowledgements

This work has been partially funded by the Eu-ropean Commission through Cross-Forest (2017-EU-IA-0140) and Spanish Ministry of Science andInnovation through REFORM (PCIN-2017-027ERANET SUMFOREST) projects

References

[1] W Beek E Folmer L Rietveld and J Walker GeoY-ASGUI The GeoSPARQL query editor and result setvisualizer The International Archives of Photogram-metry Remote Sensing and Spatial Information Sci-ences 42 (2017) 39ndash42

[2] F Bravo M del Riacuteo and C del Peso (eds) El In-ventario Forestal Nacional Elemento clave para lagestioacuten forestal sostenible Fundacioacuten General de laUniversidad de Valladolid 2002

[3] F Bravo JC Rivas JA Monreal-Nuacutentildeez and C Or-doacutentildeez BASIFOR 20 Aplicacioacuten informaacutetica para elmanejo de las bases de datos del inventario forestal na-cional Cuadernos de la Sociedad Espantildeola de Cien-cias Forestales (2004) 243ndash247

[4] F Bravo M Fabrika C Ammer S BarreiroK Bielak L Coll T Fonseca A Kangur M LofK Merganicova M Pach H Pretzsch D StojanovicL Schuler S Peric T Rotzer M del Riacuteo M Dodanand A Bravo-Oviedo Modelling approaches for mixedforests dynamics prognosis Research gaps and oppor-tunities Forest Systems 28(1) (2019) ISSN 21715068doi105424fs2019281-14342

[5] D Brickley Basic Geo (WGS84 latlong) VocabularyTechnical Report World Wide Web Consortium 2006URL httpswwww3org200301geo last visitedApril 2020

[6] J Brooke SUS ndash A quick and dirty usability scalein Usability evaluation in industry PW JordanB Thomas IL McClelland and B Weerdmeester edsTaylor amp Francis London UK 1996

[7] S Corlosquet R Delbru T Clark A Polleres andS Decker Produce and Consume Linked Data withDrupal in Proceedings of the 10th InternationalSemantic Web Conference (ISWC 2011) LNCSVol 7031 Springer Bonn Germany 2011 pp 763ndash778

[8] A-S Dadzie and E Pietriga Visualisation of LinkedData ndash Reprise Semantic Web 8(1) (2017) 1ndash21

[9] A-S Dadzie and M Rowe Approaches to VisualisingLinked Data A Survey Semantic Web 2(2) (2011)89ndash124

[10] A de Leoacuten F Wisniewki B Villazoacuten-Terrazas andO Corcho Map4rdf ndash Faceted browser for geospa-tial datasets in Proceedings of the First Interna-tional Workshop on Open Data (WOD-2012) NantesFrance 2012

[11] F Erxleben M Guumlnther M Kroumltzsch J Mendezand D Vrandecic Introducing Wikidata to the LinkedData Web in Proceedings of the 13th InternationalSemantic Web Conference (ISWC 2014) LNCSVol 8796 Springer Riva del Garda Italy 2014pp 50ndash65

[12] S Federhen The NCBI taxonomy database Nucleicacids research 40(D1) (2012) 136ndash143

[13] T Heath J Domingue and P Shabajee User inter-action and uptake challenges to successfully deployingSemantic Web technologies in Proceedings of the 3rdInternational Semantic Web User Interaction Work-shop (SWUI2006) co-located with the 5th Interna-tional Semantic Web Conference Athens GA USA2006

[14] J Kliacutemek P Škoda and M Nečaskyacute Survey of toolsfor Linked Data consumption Semantic Web 10(4)(2019) 665ndash720

[15] M Lefranccedilois A Zimmermann and N Bakerally ASPARQL extension for generating RDF from hetero-geneous formats in Proceedings of the Extended Se-mantic Web Conference (ESWCrsquo17) Portoroz Slove-nia 2017

[16] J Lehmann R Isele M Jakob A Jentzsch D Kon-tokostas PN Mendes S Hellmann M MorseyP Van Kleef S Auer et al DBpedia ndash a large-scale multilingual knowledge base extracted fromWikipedia Semantic Web 6(2) (2015a) 167ndash195

[17] J Lehmann S Athanasiou A Both A Garciacutea-RojasG Giannopoulos D Hladky JJ Le Grange A-CN Ngomo MA Sherif C Stadler et al Manag-ing Geospatial Linked Data in the GeoKnow Projectin The Semantic Web in Earth and Space ScienceCurrent Status and Future Directions T Narock andP Fox eds IOS Press 2015b pp 51ndash77 Chap 4

[18] C Nikolaou K Dogani K Bereta G GarbisM Karpathiotakis K Kyzirakos and M KoubarakisSextant Visualizing time-evolving linked geospatialdata Journal of Web Semantics 35 (2015) 35ndash52

[19] A Perego and M Lutz ISA Programme Location CoreVocabulary Technical Report World Wide Web Con-sortium 2015 URL httpswwww3orgnslocnlast visited April 2020

[20] M Perry and J Herring OGC GeoSPARQL ndash A Ge-ographic Query Language for RDF Data OGC Im-

plementation Standard Open Geospatial Consortium2012 URL httpwwwopengisnetdocISgeosparql10 last visited April 2020

[21] H Pretzsch Forest Dynamics Growth and YieldSpringer Cham Switzerland 2009

[22] Md Riacuteo JC Rivas S Condes J Martiacutenez-MillaacutenG Montero I Cantildeellas AC Ordoacutentildeez V PandoR San-Martiacuten and F Bravo BASIFOR Aplicacioacuteninformaacutetica para el manejo de bases de datos del Se-gundo Inventario Forestal Nacional F Bravo M delRiacuteo and C del Peso eds Fundacioacuten General de la Uni-versidad de Valladolid 2002 pp 181ndash191

[23] J Riofriacuteo M del Riacuteo and F Bravo Mixing effects ongrowth efficiency in mixed pine forests Forestry AnInternational Journal of Forest Research 90(3) (2017)381ndash392

[24] R Ruiz-Peinado A Bravo-Oviedo E Loacutepez-Senespleda F Bravo and M del Riacuteo Forest manage-ment and carbon sequestration in the Mediterraneanregion A review Forest Systems 26(2) (2017)

[25] R Ruiz-Peinado M Heym L Drossler P CoronaS Condes F Bravo H Pretzsch A Bravo-Oviedoand M del Riacuteo Data Platforms for Mixed Forest Re-search Contributions from the EuMIXFOR Networkin Dynamics Silviculture and Management of MixedForests A Bravo-Oviedo H Pretzsch and M del Riacuteoeds Springer Cham Switzerland 2018 pp 73ndash101

[26] J Sauro and JR Lewis Quantifying the user expe-rience Practical statistics for user research Morgan-Kaufmann Amsterdam Netherlands 2012

[27] MJ Schelhaas S Varis A Schuck and GJ NabuursEFISCEN Inventory Database 2006 URL httpwwwefiintportalvirtual_librarydatabasesefiscenlast visited April 2020

[28] MG Skjaeligveland Sgvizler A javascript wrapper foreasy visualization of SPARQL result sets in ExtendedSemantic Web Conference Heraklion Greece 2012pp 361ndash365

[29] C Stadler J Lehmann K Houmlffner and S AuerLinkedGeoData A core for a web of spatial open dataSemantic Web 3(4) (2012) 333ndash354

[30] J Tandy L van den Brink and P Barnaghi Spatialdata on the Web best practices W3C Working GroupNote OGC 15-107 OGC amp W3C 2017 URL httpswwww3orgTRsdw-bp last visited March 2020

[31] E Tomppo T Gschwantner M Lawrence andRE McRoberts (eds) National forest inventoriesPathways for common reporting Springer ChamSwitzerland 2010

[32] B Veenendaal MA Brovelli and S Li Review of webmapping Eras trends and directions ISPRS Interna-tional Journal of Geo-Information 6(10) (2017) 317

[33] G Vega-Gorgojo L Slaughter BM Von ZernichowN Nikolov and D Roman Linked Data ExplorationWith RDF Surveyor IEEE Access 7 (2019) 172199ndash172213

[34] WHATWG (Apple Google Mozilla Microsoft)DOM Living Standard WHATWG 2020 URL httpsdomspecwhatwgorg last visited March 2020