ClimSpecVis:Visualizing the Distribution of Specimens from Natural

History Collections and Their Relation with ClimateData

Miguel de Oliveira Martins Melıcia Cruzmiguelcruz@tecnico.ulisboa.pt

Instituto Superior Tecnico, Lisboa, Portugal

October 2018

Abstract

Museums have ever-growing collections of specimens found during expeditions. The study of thesecollections and their species ecology can be used to understand the state of the environment where thespecimens are found. The study of Phenology, the effect of climate in the distribution of these speciesand its change over time, can provide further insight into climate change and the need for efforts inspecies conservation. We analyze current methods and approaches to this problem and researchemethods to address their shortcomings and limitations, resulting in the creation of an innovative solutionbased on the principles of information visualization. With support from the National Museum of NaturalHistory and Science in Lisbon, Portugal, our solution was turned into a functional prototype. We describethe development of this prototype as well as the various stages of evaluation with domain experts andinexperienced users. We then present our final evaluation’s results in terms of usability and utility.Usability is satisfactory though users could benefit from some level of expertise in the domain. Whileutility will depend on the necessities of an investigator and what type of research they do on theirspecimen records, it can still be useful to convey important messages to casual users such as museumvisitors.Keywords: Information Visualization, Species Distribution, Climate, Phenology, Natural History Speci-men Collections

1. IntroductionEvery day, there are on-field collections of speci-mens of animals, plants or any other living beingsdone either on expeditions or by chance. Theseare then retrieved to be identified, cataloged andstored. The continuous growth in size of these col-lections improves our awareness of the biodiversityand the richness of the research that can be doneover these specimens.

Studying the historical distribution of a speciesand the period at which its specimens were col-lected can provide insight into changes in climatevariables. Certain animal and plant species areclimate indicators in the sense that certain peri-odic life cycle events are influenced by variationsin climate happening seasonally or over a periodof years.

The study of these climate events and their in-fluences on life-cycle events of a species is calledPhenology [13]. A change in the geographical dis-tribution or in the occurrence of period lifecycle

events of a certain species could in turn be a goodindicator of climate change and provide insight intohow to conserve this biodiversity [12].

The growing size of these collections can leadto difficulties for researchers if they wish to exam-ine their current status and draw conclusions. Cur-rently, using the collections and tools they have athand, researchers can for example represent thedistribution of a species, make statistical analysisto determine their habitat and make predictions forwhere it may be possible to collect them. Howeverpowerful they may be, these methods can take timeto execute and lack means to quickly interact withthe findings. From a different point of view, anotherproblem is that it becomes difficult to present thiscollection data to viewers outside of the domainwhile maintaining its attractiveness.

This is a problem that the MUHNAC – the Na-tional Museum of Natural History and Science, in

1

Lisbon, Portugal1 – currently faces. Applying Infor-mation Visualization techniques could provide bothresearchers and more casual viewers, such as vis-itors of the MUHNAC, with ways to further analyzethe specimen collections in conjunction with histori-cal climate data by introducing the sense of sight tochannel more information to the viewer, while alsoenabling a higher degree of interaction and explo-ration.

The main objectives of this study are: to ana-lyze current and modern approaches and methodsthat can apply to our problem in regards to theirfeatures, strengths and weaknesses; to researchsolutions to our problem that address the short-comings of existing work and overcome the chal-lenges we may face; from our research, to createan innovative solution to visualize information fromNatural History collections of records and histori-cal climate data and in ways that are meaningful toviewers inside and outside the domain; to establisha standard to evaluate the success of our solutionin terms of usability and utility, to develop a proto-type based on our solution and with it conduct theevaluation process.

2. Related WorkThe increasing amount of digitalized geotaggedrecords of specimens, either from more moderncapture methods or from intensive work involv-ing records in paper, has facilitated the creationof visualizations to observe the temporal evolutionof the geographical distribution of species. Fur-thermore, it also opens up opportunities for morecomplex visualizations that show relations betweenweather and climate changes on the distributionand abundance of different species.

We will present the more recent approachesto subjects like these or related, be they purelygrounded on InfoVis principles or not, and that aremost relevant to our solution. As the main sub-ject is the geographical distribution of events andits variation over time, coupled with the geographi-cal and temporal aspects of climate, our focus willmainly be on on geographical distribution, temporaldistribution and geotemporal evolution.

2.1. GeographicalColgan et al. [3] is a study to estimate the amountof biomass of vegetation species on recording lo-cations and allows users to view those estimatesover a map. By assigning a color to a species itgives the perception of which species are domi-nant in different locations. Colgan et al. [3] alsocombine those records with environmental vari-ables, such as elevation or geologic substrate, toallow the user to see on which conditions a species

1http://www.museus.ulisboa.pt/

thrives or falls, as seen on Figure 1. Though idealto spot dominant species, it is also poorly scalablein the sense that less densely packed species willbe overshadowed by others.

Figure 1: Colgan et al. [3] – On the left, the percentage ofbiomass a species of tree amasses to geographically; on thetop right ground elevation (meters), slope (degrees), profile cur-vature, aspect (degrees).

Similarly, oBis-seamaP[5] offers additional lay-ers for environmental variables such as sea sur-face temperature and ocean color for the selectedtime period to provide context for the animal obser-vations.

GeoTemCo [10] provides a visualization of cli-mate events, such as periods of heat, cold anddrought, over a map. It can display cooccurringevents in the same location and on the same dateusing a circular mark for each type of event and po-sitioning them adjacently to one another and avoid-ing occlusion in these cases.

2.2. TemporalKitamoto et al. [9] bring together linear and cyclictime to show off patterns between the bumblebeesand flowers that interact throughout the months ofthe year, while also allowing comparison betweenplant species in terms of their interactions, as seenon Figure 2.

Figure 2: Kitamoto et al. [9] – Cyclic temporal distribution ofvarious species.

Approaches such as GAV [8] (CO2 levels), Tax-iVis [4] (taxi pickups, dropoffs, fares and tips),GeoTemCo [10] (number of natural hazards and

2

their victims) also use time-series to show trendsof variables. oBis-seamaP [5] uses a bar chart forthe same purpose in the context of number of ob-servations of animals.

2.3. GeotemporalCartografia de Briofits 2 allows viewers to pick aspecies of Bryophyte and view its records over amap. Their coordinates are limited to the UTM Co-ordinate System and their temporal description iseither pre-1970 or post-1970, thus lacking preci-sion.

HerbariaViz [2] can use Coxcomb symbols laidout over a map to represent the distribution of ob-servations through the months of a year on eachlocation, at lower levels of map zoom. While otherapproaches aim for linear time, this one makes useof cyclic time, as flora observations can be cyclicevents depending on weather and climate, allow-ing the user to explore phenology by region.

GTdiff [6] proposes a map visualization withadditional small multiples pertaining to differentranges of time. A time range is picked and subdi-vided into five sub-ranges of years, with each yearbeing coded by a color in a gradient. Each bub-ble from the main visualization has two channels:abundance in that location as size, and the year ofthe record as the color from the gradient.

Andrienko et al. [1] display median speeds inspatial compartments by days of the week andhours of the day, with each compartment laid overits correspondent geographic location. An exampleof such a small multiple is seen on Figure 3.

Figure 3: A small multiple gridded heatmaps on Andrienko etal. [1].

2.4. Other approachesUnipept [11] takes several approaches towards thematter of diversity of species, all of them relyingon InfoVis techniques to represent hierarchies -– atreemap, a Sunburst, and a treeview with elementsof a Sankey diagram. Not only do they representhierarchies, they also show the relevance of thebranches on each taxonomical rank,

2http://briofits.iec.cat/

3. DevelopmentThe development of ClimSpecVis was necessaryto study the potential of these techniques and itwas based on the strengths of the approacheswe’ve discussed, while at the same time attempt-ing to overcome their weaknesses. It was dividedin incremental iteration cycles where on each werefined the requirements, improved our prototypeand evaluated its performance.

3.1. DataThe two departments of MUHNAC have providedus with geotagged data of bryophytes – an informalgroup of land plants such as mosses – and insectsfound on national grounds (Portugal), with tempo-ral, geographical and taxonomic descriptions up toa certain level of detail.

The data is not uniformly distributed as it existsonly for locations that have been explored and itdepends on which species there was an effort tocollect specimens of during each of those excur-sions. This skews the information we’re displaying,which introduces limitations into our solution.

We worked with approximately 2700 records ofBryophytes and 3300 records of Insects.

Regarding environment variables, we have gath-ered historical climate records from WorldClim andNOAA for various points of the Portuguese terri-tory, including precipitation and maximum, mini-mum and average temperature levels, as these of-ten determine the success of a species especiallyin the case of Bryophytes.

3.2. RequirementsThe requirements evolved throughout the develop-ment process of our prototypes. The final state ofthe requirements includes the questions that follow.

Q1 - On which locations have specimens oftaxon G been found?

Q2 - Was there change in the locations wheregenus G could be found, over time?

Q3 - On which locations haven’t there been col-lected specimens, thus needing to be explored?

Q4 - How many occurrences of taxon G werethere in location L in time range T?

Q5 - What is the evolution in abundance of spec-imens of taxon G, before and after 1996, in locationL?

Q6 - What is the evolution in abundance of spec-imens of genus G, over the months and years?

Q7 - How do the abundance and locations wheretaxa A and B were found compare?

Q8 - On which months are there more sightingsof specimens of taxon G?

Q9 - Could the change in maximum temperaturebe affecting the collection period of genus G, usingclimate station S, through the months and years?

3

Q10 - Could there be a relation between the pre-cipitation levels and the abundance of specimensof genus G? Q11 - What is the geographical evolu-tion of the state of knowledge throughout time?

Q12 - Which taxa where found in location L?Q14 - How do the children taxa of a taxon R rank

up in terms of number of collected specimens?What is the most collected taxon?

3.3. Low-Fidelity PrototypesWe conceived non-functional prototypes in con-junction with the curators of MUHNAC depart-ments we partnered with in order to better estab-lish their requirements and needs and to adjust ourproposal for a solution. Low-fidelity Prototypes canbe produced quickly and with few resources whilestill covering various parts and possible states of asystem [7, p. 389].

The Low-fidelity Prototypes we created stronglyinfluenced the functional prototype we created andwhich we will showcase later.

Figure 4 shows the taxonomic tree in the taxo-nomic view, with two selected taxa, one a child ofthe other, each in a different color assigned fromselected gradients from ColorBrewer3.

Figure 4: Low-fidelity prototype of taxonomic view.

Figure 5 shows an initial version of the geotem-poral view. Three levels of detail existed that variedby zoom level. Low levels zoom had marks indicat-ing presence of a taxon, while higher levels fea-tured more complex small multiples of marks withtemporal dimensions.

Figure 5: Low-fidelity prototype of geotemporal view.

3http://colorbrewer2.org/

Figure 6 shows the temporal view with a cli-mate variable matrix as background layer and aspecimen abundance bubble grid. Rows pertainto months and columns pertain to years.

Figure 6: Low-fidelity prototype of temporal view.

3.4. First Iteration Development

We implemented our solution as a Web browserapplication that the user can run locally or ac-cess through a website, based on Javascript libraryD3.js4.

3.4.1 Taxonomic View

There were concerns regarding the arrangementof the tree nodes and edges and how they wouldaffect the navigation and the existence of text la-bels for taxa. Initially, speaking in terms of SVGcoordinates, each parent node was displayed bycomputing the average between the y coordinateof its topmost child node and the y coordinate ofits bottommost child node. This made navigationmore difficult as it required more scrolling to finda parent node before moving to its children nodes.Furthermore, it could require the positioning of thetaxon name labels to be inconsistent in order to re-duce occlusion with the edges.

This was addressed by displaying the tree moresimilarly to an operating system’s user interface fordirectory navigation, where a parent directory ap-pears with its name beside it and its children nodesare listed starting with a vertical offset, as seenon Figure 7 where the labels rarely intersect withedges.

4https://d3js.org/

4

Figure 7: Improved arrangement of tree nodes and edges ontaxonomic view.

The selections are maintained as a list, with taxanever changing their order. The first taxon is theone whose records are considered in the temporalview. If two or more taxa are selected and the firsttaxon is deselected, then the second taxon takesits position. Through experiments with each avail-able color, it was agreed that the color that wasmost visible against the temporal view’s matrix wasorange and that remained as the color for the firsttaxon in the selection list.

3.4.2 Geotemporal View

This first iteration of the functional prototype onlysaw implemented the low and high levels of zoom.It already featured the final look of the groupedmarks at low levels of zoom, as explained on Fig-ure 8 and seen on Figure 9.

Figure 8: Diagram showing which quadrant each taxon’s col-ored square would be represented.

Figure 9: Marks of the geotemporal view at a high zoom level,for multiple taxa.

An addition on this first iteration was the yellow

circles over the map, corresponding to each of theavailable climate stations from which data is usedon the temporal view.

3.4.3 Temporal View

The functional prototype’s first version of the celland bubble matrix closely followed the Low-fidelityPrototype and it can be seen on 10.

Figure 10: A complete look at the temporal view.

3.5. First Iteration Formative EvaluationThe first formative evaluation session for the func-tional prototype was performed with the domain ex-perts who had supported our development of thelow fidelity and functional prototypes. At this mo-ment, we had implemented features only sufficientfor the first twelve questions to be answered. Thesession followed a guide describing the prepara-tion of the session, how the collection of feedbackwas to be made and listed.

In the case of the Zoology department person-nel, they requested a collapse and expansion fea-ture on the taxonomy view’s tree nodes.

Upon discussing the medium and high levels ofzoom on the geotemporal view, it was decided thatthose were not intuitive enough and would be dis-carded. It was suggested that we add a single highlevel of zoom where we would compare betweenonly two selectable time instances and not through-out a continuous range of years. These two yearinstances would be separated by a splitting year.

It was suggested that information was shownabout climate variables when clicking on a point onthe map on the geotemporal view, such as by usinga tooltip.

Some criticisms were made regarding the tem-poral view’s matrix use of a single station’s climatevariable values while its bubbles show specimencounts spanning the whole Portuguese territory.Not all specimens are collected in locations withconditions like those reported by the selected cli-mate station. This was a limitation that remainedthroughout the life-time of our prototype and solu-tion.

3.6. Second Iteration ImplementationThis iteration served to perfect the state of the func-tional prototype was it was then.

5

3.6.1 Taxonomic View

The collapse and expand feature was added to thetaxonomic tree as well as tooltips.

3.6.2 Geotemporal View

At the new high level of zoom, the marks remainsimilar, also following the separation into quad-rants. In more densely packed regions, severalinstances of these quadrant-organized marks canbe visible, so we are maintaining the small mul-tiples technique of visuzalization. Each quadrantcan now contain up to two bars instead of a singlesquare. A bar’s height grows with more specimenscollected in that temporal interval. In each quad-rant, the bar for specimens before the splitting yearis to the left of the bar for specimens after the se-lected year. On the two lower quadrants, the barsgrow downwards, with their base at the horizon-tal axis that separates the quadrants. Figure 11shows an example of this small multiple at a highlevel of zoom.

Figure 11: Multiple taxa before and since a splitting year, at ahigh zoom level.

A map scale and a table for additional informa-tion about specimens collected in a location wereadded.

3.6.3 Temporal View

The selection of the year range was made morepowerful. The year range can now be dragged anda user can also choose the splitting year.

Figure 12: A close up of the temporal view.

4. Final EvaluationWith the last iteration of implementation havingcome to an end, we began preparing for the finalevaluation of the functional prototype. Once again,we constructed guides for each kind of test: utilityand usability.

Utility tests were run over a few hours with thesame domain experts that guided us along the de-velopment of this solution. Usability tests were runwith fifteen users not familiar with the domain butwith varying levels of experience using computersand interfaces of this kind. Several quantitative andqualitative measures were taken and spoken feed-back was recorded as much as possible.

4.1. Utility EvaluationEach of the two departments showed special in-terest in different parts of the solution as their re-search is, at current times, in different phases: theZoology department’s efforts are mostly centeredaround cleaning up their records, many of whichwere not considered in our solution, and learn-ing general information about their collection; theHerbarium is at a further phase where they alsouse their records for analysis and to produce pre-dictive models for species distribution based on en-vironmental variables.

The Zoology department personnel mostlypraised the different types of view one can haveover the distribution of specimens of their collec-tion, stating the solution would be a good comple-ment to a record database navigation system.

The Herbarium, though not seeing potential asfar as discovering new knoledge, praised our solu-tion’s ability to display and confirm knowledge.

In addition, the Herbarium suggested the use ofour solution in a future demonstration to the Lisbontown hall regarding climate change and the expan-sion of species throughout Lisbon.

4.2. Usability EvaluationUsability tests provided quantitative and qualitativemeasures for the success of our solution in termsof usability.

Quantitative measures were obtained by mea-suring the time spent, errors made and clicks doneto answer each of the questions in our require-ments, which we will now call a task.

To standardize our qualitative measures asmuch as possible, we resorted to two typesof questionnaires: System Usability Scale5 forglobal satisfaction, and Single Ease Question6 forquestion-level ease of use.

These measurements underwent a statisticalanalysis and such results were analyzed as well.

5https://measuringu.com/sus/6https://measuringu.com/seq10/

6

4.2.1 Quantitative Measures

In general, the completion of these tasks takes be-tween one and two minutes and over four clicks.Some of this time, as well as clicks made, is spentnavigating through the taxonomic tree until the userreached the taxon indicated for the task.

Most of the issues arose due to the users beingoverwhelmed by both the amount of interface ele-ments they could interact with and by the perceivedcomplexity of the domain. The concept of speciestaxonomy was difficult for some to grasp. Therewas also some difficulty to be able to tell patternsthat could exist between the temporal distributionof species and the temporal variation of climatevariables. The high levels of zoom also proved dif-ficult to interpret, in some cases.

Other issues included the difficulty in handlingsmall interface elements in the selection of yearranges and the lack of a territorial division layer.

Figure 13: Distribution of time spent answering questions Q1to Q7.

Question 2 proved problematic early on as userswere still not used to the features the prototype pro-vided, in this case being the use of the year rangethat can be dragged. In general, the year range barand ticks also proved difficult to use for some usersas these elements could be too small for the userto easily click on.

Question 4 highlighted the problem that arisesfrom the lack of a territorial division layer, as theusers had trouble understanding where the givenlocation was.

Figure 14: Distribution of clicks done to answer questions Q1to Q7.

Question 5 also proved difficult as some userstook time to understand the concept of the splittingyear, its effect on the geotemporal view marks athigh levels of zoom, as well as how to interpret theresults they saw.

Question 6, while it did not in fact seem to gatherpoor results, still required some reflection by theusers as they could not immediately recognize theuse of the temporal view’s matrix and bubble gridand first assumed they would use the year range.

Figure 15: Distribution of errors made answering questions Q1to Q7.

Question 7 confused some users as it introducedthe selection of more than one taxa, which also ledto more clicks for the increased number of selec-tions to make. In general, the concept of orderbetween selected taxa also confused users. Thisextended into understanding which taxon was in fo-cus in the temporal view’s matrix.

7

Question 9 was the most problematic. A greatdeal of users could not complete it as, once theyhad applied the needed filters, they could not pro-ceed with the interpretation of the results. Thiswas, in part, due to data we are using being scarce.However, it was also not obvious that certain pat-terns could be highlighted when comparing the cli-mate variable’s matrix and the taxon’s temporal dis-tribution bubble grid. Only five users were success-ful in this task.

Figure 16: Distribution of time spent answering questions Q8to Q14.

Question 10 brought up somewhat similar prob-lems, though users could better understand whatwas asked after the previous question and under-stood.

Question 11 suffered from similar problems asQuestion 2, since the navigation with the yearrange bar and blips wasn’t immediately obviousand some users struggled with accurately drag-ging the correct interface elements. Otherwise, thisquestion would not have required a large amount ofclicks or time.

Figure 17: Distribution of clicks done to answer questions Q8to Q14.

Question 12, while it appears to not be com-plex from the results we recorded, required slightlymore time than expected. Users didn’t seem to re-call that further information about a location couldbe seen by clicking on its marks in the geotemporalview to fill the details table.

Question 14 is another question that had lessgood results in part due to users not completely un-derstanding the concept of species taxonomy andhow to apply the sorting features our taxonomicview offers. Furthermore, the large number of in-terface elements that are involved in answering thisquestion is high and seemed to overwhelm users.

Figure 18: Distribution of errors made answering questions Q8to Q14.

8

Questions 5, 9, 10 and 11 are those that seemedto be more difficult to some users. The responsesto the Single Ease Questions match the issues thatoccurred in the more problematic questions, be itin terms of errors made, time taken, clicks made oreven the reflection moments users needed to haveto interpret what the visualization showed them.

4.2.2 Qualitative Measures

While, in general, results are positive, usually tend-ing to the positive side in each question’s scope– that of being easier and more satisfying to use– there were of course some cases where someusers had less than ideal experiences.

Figure 19: Distribution of the answers to Single Ease Questionspertaining to questions Q1 to Q17.

Some users were less experienced with systemssimilar to ours which created difficulty in using theprototype. At times, these were related to smallobjects in the views which were difficult to inter-act with or see. In other instances, the complex-ity of the visualization and domain, along with howhelpful the initial demonstration had been, led tothese users never fully understanding what theyshould look for to answer the questions. Neverthe-less, these test users seemed aware of their lackof experience compared to other people which ex-plained the results of question 7 being higher.

Figure 20: Distribution of the answers to Single Ease Questionspertaining to questions Q8 to Q14.

One of the Herbarium experts, upon navigatingin our prototype, seemed to very quickly adapt to itsinterface. As we saw in our results of the evaluationof usability per task, the user interface sometimescaused confusion to users and they did not seemto adapt as quickly. This suggests that expertise inthis domain influences not only how our visualiza-tion is interpreted by them but also how usable itis.

5. ConclusionsSome of the results of the evaluation sessionswe held were not as positive as expected, mainlythose that depended on more domain-specificknowledge, on certain unresolved user interactionissues, and on the quality of the data we had avail-able.

However, we do consider our solution to be suc-cessful at meeting the requirements we set outinitially, even considering how some aspects mayrequire higher levels of expertise in the domain.While usability is at satisfactory levels, especiallywhen users gain some practice, utility can be highfor some domain experts, depending on the type ofstudies they are doing with the data they possess.

5.1. LimitationsAs mentioned during the description of the de-velopment cycle and during evaluation sessions,there are certain limitations that our solution faces,mainly due to the nature of the data we use, itsquality, completeness and high-dimensionality.

5.2. Future WorkOther techniques could be explored to attempt tovisualize the phenology and relation between cli-mate variables and species distribution.

Some fine-tuning to user interface elements

9

could be made in order to reduce the errors usersmake and the time they spend navigating to get ananswer to their question.

Future work will also include further collaborationwith the museum personnel to produce a presen-tation for the Lisbon town hall, should they makea commitment to their proposal in late October2018.

References[1] G. Andrienko and N. Andrienko. Spatio-

temporal aggregation for visual analysis ofmovements. In Visual Analytics Science andTechnology, 2008. VAST’08. IEEE Sympo-sium on, pages 51–58. IEEE, 2008.

[2] T. Auer, A. M. MacEachren, C. McCabe,S. Pezanowski, and M. Stryker. Herbariaviz:A web-based client–server interface for map-ping and exploring flora observation data.Ecological Informatics, 6(2):93–110, 2011.

[3] M. S. Colgan and G. P. Asner. Coexistenceand environmental filtering of species-specificbiomass in an african savanna. Ecology,95(6):1579–1590, 2014.

[4] N. Ferreira, J. Poco, H. T. Vo, J. Freire, andC. T. Silva. Visual exploration of big spatio-temporal urban data: A study of new york citytaxi trips. IEEE Transactions on Visualizationand Computer Graphics, 19(12):2149–2158,2013.

[5] P. N. Halpin, A. J. Read, E. Fujioka, B. D. Best,B. Donnelly, L. J. Hazen, C. Kot, K. Urian,E. LaBrecque, A. Dimatteo, et al. Obis-seamap: The world data center for marinemammal, sea bird, and sea turtle distribu-tions. Oceanography, 22(2):104–115, 2009.

[6] O. Hoeber, G. Wilson, S. Harding, R. En-guehard, and R. Devillers. Exploring geo-temporal differences using gtdiff. In PacificVisualization Symposium (PacificVis), 2011IEEE, pages 139–146. IEEE, 2011.

[7] Y. R. Jenny Preece, Helen Sharp. Interac-tion Design: Beyond Human-Computer Inter-action. Wiley, 4 edition, 2015.

[8] M. Jern and J. Franzen. Integrating infovisand geovis components. In Information Visu-alization, 2007. IV’07. 11th International Con-ference, pages 511–520. IEEE, 2007.

[9] A. Kitamoto, M. Nakahara, I. Washitani,T. Kadoya, M. Yasuwaka, and M. Kitsure-gawa. Information visualization and organi-zation for participatory monitoring of invasive

alien species. In Database and Expert Sys-tems Application, 2009. DEXA’09. 20th Inter-national Workshop on, pages 345–349. IEEE,2009.

[10] O. Kolditz, K. Rink, and G. Scheuermann. Uti-lizing geotemco for visualizing environmentaldata.

[11] B. Mesuere, G. Debyser, M. Aerts, B. De-vreese, P. Vandamme, and P. Dawyndt. Theunipept metaproteomics analysis pipeline.Proteomics, 15(8):1437–1442, 2015.

[12] C. Sergio, C. Garcia, C. Vieira, H. Hespanhol,M. Sim-Sim, S. Stow, and R. Figueira. Con-servation of portuguese red-listed bryophytesspecies in portugal: promoting a shift inperspective on climate changes. PlantBiosystems-An International Journal Dealingwith all Aspects of Plant Biology, 148(4):837–850, 2014.

[13] C. G. Willis, E. R. Ellwood, R. B. Primack,C. C. Davis, K. D. Pearson, A. S. Gallinat, J. M.Yost, G. Nelson, S. J. Mazer, N. L. Rossing-ton, et al. Old plants, new tricks: Phenologicalresearch using herbarium specimens. Trendsin ecology & evolution, 2017.

10