Biodi e si y Da a Jou nal 8: e55959
doi: 10.3897/BDJ.8.e55959
Me hods
A s a egy o digi ise na u al his o y collec ions
wi h limi ed esou ces
Joaquim San os , Paulo Rupino da Cunha , Fá ima Sales
‡ Uni e si y o Coimb a, Cen e o Func ional Ecology - Science o People & he Plane , Depa men o Li e Sciences,
Coimb a, Po ugal
§ Uni e si y o Coimb a, Cen e o In o ma ics and Sys ems o he Uni e si y o Coimb a, Depa men o In o ma ics
Enginee ing, Coimb a, Po ugal
Co esponding au ho : Joaquim San os ([email p o ec ed])
Academic edi o : Vincen Smi h
Recei ed: 29 Jun 2020 | Accep ed: 03 Sep 2020 | Published: 23 Oc 2020
Ci a ion: San os J, Rupino da Cunha P, Sales F (2020) A s a egy o digi ise na u al his o y collec ions wi h
limi ed esou ces. Biodi e si y Da a Jou nal 8: e55959. h ps://doi.o g/10.3897/BDJ.8.e55959
Abs ac
The p esen wo k is a con ibu ion owa ds accele a ing he digi isa ion p ocess o na u al
his o y collec ions, usually a slow p ocess. A wo-s age p ocess was de eloped a he
he ba ium o he Uni e si y o Coimb a: (i) a new wo k low was es ablished o
au oma ically c ea e eco ds in he he ba ium mas e da abase wi h minimum in o ma ion,
while cap u ing digi al images; (ii) hese eco ds a e hen used o popula e a web-based
c owdsou cing pla o m whe e ci izens a e in ol ed in he ansc ip ion o specimen labels
om he digi al images. This app oach simpli ies and accele a es da abasing, educes
specimen manipula ion and p omo es he in ol emen o ci izens in he scien i ic goals o
he he ba ium. The no el ea u es o his p ocess a e: (i) he alida ion me hod o he
c owdsou cing con ibu ion ha ensu es quali y con ol, enabling he da a o in eg a e he
mas e da abase di ec ly and (ii) he ield-by- ield in eg a ion in he mas e da abase
enables immedia e co ec ions o any eco d in he ca alogue.
Keywo ds
accele a ing he ba ium digi isa ion, au oma e da abasing p ocesses, c owdsou cing
pla o m
‡ § ‡
© San os J e al. This is an open access a icle dis ibu ed unde he e ms o he C ea i e Commons A ibu ion License (CC BY
4.0), which pe mi s un es ic ed use, dis ibu ion, and ep oduc ion in any medium, p o ided he o iginal au ho and sou ce a e
c edi ed.
In oduc ion
Biological collec ions a e majo sou ces o aluable in o ma ion wi h po en ial o mul iple
a eas o knowledge (Funk 2018). An exponen ial g ow h has been no ed o publica ions
on
subjec s as di e se as axonomy, global change biology o DNA analyses, o name bu a
ew, based on he collec ions housed in he ba ia (Hebe ling e al. 2019). Such an inc ease
is possible because o he mass digi isa ion ha he ba ia ha e gone h ough o e he las
wo decades, acili a ing he access o specimens and associa ed in o ma ion in online
ca alogues and also in da a agg ega o s, such as GBIF (Sol is 2017). The e a e se e al
app oaches o es ablish a digi isa ion plan, depending on he collec ion cha ac e is ics,
such as i s size, he a ailable esou ces and budge (Vollma e al. 2010, Wal on e al.
2020). In gene al, digi isa ion is a demanding ask, equi ing conside able human labou
and ime. P ima ily, i consis s o c ea ing a eco d in a da abase o each specimen and
hen ansc ibing he ele an da a o speci ic ields (da abasing). A unique iden i ie is
assigned o each specimen in he da abase and, in mos collec ions, a ba code s icke is
placed on he specimen's moun ing suppo . Fo many collec ions, i is desi able o cap u e
images o he specimens (imaging). Ins i u ions ha e es ablished digi isa ion wo k lows
acco ding o hei equi emen s and p e e ences (Tulig e al. 2012, Has on e al. 2012,
Nie a de la Hidalga e al. 2020).
The he ba ium o he Uni e si y o Coimb a (COI, h p://www.uc.p /en/he ba io_digi al) is
digi ising i s plan collec ion o ca. 800,000 specimens and making he da a a ailable online
(h p://coica alogue.uc.p ). Due o he slowness o he me hods only ca. 10% o he
ma e ials is p ocessed so a . Some ins i u es ha e implemen ed c owdsou cing agendas
o bene i om emo e ansc ibing by olun ee s, his p o ing o be a p ac ical solu ion
(Swanson e al. 2016) wi h colla e al bene i s, such as he imp o ed scien i ic li e acy o he
public in ol ed (C onje e al. 2011, Ellwood e al. 2015). Va ious c owdsou cing pla o ms
ha e been designed o build on ci izen collabo a ion, om b oad scope o ien a ed p ojec s
(Zooni e se 2020), o na ow ields o in e es , such as biological collec ions (DigiVol 2020,
DoeDa 2020) and e en o speci ic kinds o collec ions (Les he bonau es 2020). The
wo k low o all hose pla o ms is simila :
1. c ea ion o a p ojec /mission limi ed o a numbe o objec s and/o ime,
2. submission o da a,
3. alida ion and
4. in eg a ion o he collec ed da a in he collec ion da abase. The in eg a ion o da a
occu s only when he p ojec is comple ed.
The so wa e used a COI o da abasing is SPECIFY (Speci y Collec ions Conso ium
2019), a ull sui e o manage biological collec ions, cu en ly used by ca. 500 ins i u es in
mo e han 40 coun ies (Speci y Collec ions Conso ium 2017). Imaging is made using
ei he an in e ed A3 scanne o a ull- ame digi al pho og aphy se -up, depending on he
pu pose. The ilename gi en o he image ile is he specimen's ba code, which enables
he au oma ic associa ion o he image o he co esponding specimen. Images can be
acqui ed in di e en ile o ma s; he e o e, i is necessa y o p ocess hem when uni o mi y
2San os J e al
is equi ed. Image p ocessing is associa ed wi h backup wo k lows; he e o e, he e is a
ou ine ha deals wi h images o comple e all asks needed.
To speed up digi isa ion, COI has es ablished a new wo-s age wo k low ha op imises he
ime spen in he imaging p ocess and associa ed handling o he specimens. When
imaging, he co esponding eco d in he da abase is au oma ically c ea ed wi h minimal
in o ma ion (ba code and axon). This image is p o ided o emo e olun ee s who will
ansc ibe he da a om he specimen labels ( isible in he image) o co esponding ields
in a web o m using a dedica ed c owdsou cing pla o m. This pla o m was de eloped
om sc a ch o ul il speci ic equi emen s, he main one being he possibili y o da a
in eg a ion wi h he mas e da abase in nea eal ime.
Me hodology
S age 1. Au oma ic p ocess o c ea e eco ds wi h minimum in o ma ion in he
mas e da abase om se s o digi al images
The objec i es o accomplish du ing he i s s age o he accele a ing p ocess a e:
• o c ea e eco ds in he mas e da abase and
• o p ocess images o ile o ma ans o ma ion and backup.
Specimens a e s o ed in he he ba ium cabine s acco ding o a axonomic sequence;
he e o e, all hose in he same olde belong o he same axon (scien i ic name). When
imaging he specimens, he ope a o can easily c ea e di ec o ies named a e he
co esponding axon and s o e he images p oduced in he imaging s a ion compu e . Each
specimen shee will ha e one image ile, named using a ba code scanne o ead all he
ba codes on he specimen shee ( he e can be mo e han one specimen pe shee ). A he
end o an imaging session, each di ec o y will con ain one o se e al image iles, each ile
being named acco ding o he co esponding shee ba codes. When images a e p ocessed
o in eg a e he cen alised ile sys em o image s o age, a sc ip can be execu ed o
c ea e a eco d in he mas e da abase o each specimen, assigning one de e mina ion
wi h he espec i e axon name (di ec o y name).
The concep is simple, bu some aspec s mus be conside ed, depending on he local se -
up. The da abase schema o SPECIFY has mo e han one hund ed ables, so c ea ing
eco ds ex e nally o i s own in e ace should be well planned o a oid da a co up ion.
The da abase s uc u e equi es he use o a hie a chically-s uc u ed axon ee (Family ->
Genus -> Species -> …). C ea ing de e mina ions om a s ing being con e ed o a
hie a chical s uc u e mus conside all he anks, including hose below species le el, i.e.
subspecies, a ie y, o m and e en o he less equen ly as mic ogene. Acco ding o
bo anical nomencla u e, an in aspeci ic axon includes a connec ing e m in he ull name
o deno e he ank (subsp. a . e c.) (Tu land e al. 2018). The ope a o will w i e he name
as i is on he specimen olde (au ho names a e no included, as his would inc ease
subs an ially he complexi y o in aspeci ic names ecogni ion and he e will ha dly exis
A s a egy o digi ise na u al his o y collec ions wi h limi ed esou ces 3
homonyms in ou collec ion). The e o e, when he s ing con aining he axon name is ead,
he in aspeci ic ank is s anda dised o mee wha is es ablished in ou da abase, o
example, “ssp.” will be con e ed o “subsp.”. Then, a que y is execu ed in he SPECIFY
da abase o check whe he he ull name al eady exis s. I a ma ch is no ound, he s ing
is spli by spaces and he possible name combina ions a e sea ched and ma ched un il all
he names exis in he hie a chy ee: check and ma ch he genus o c ea e; check and
ma ch he species o c ea e; check and ma ch each in aspeci ic ank o c ea e (Fig. 1).
Thus, when a name does no exis in a ank, i mus be c ea ed. A gi en axon name s a s
wi h he gene ic name. Consequen ly, all in a-gene ic names can be c ea ed as long as
he gene ic name is in he da abase. To enable he c ea ion o a new gene ic name wi hou
knowing i s amily, a new “ amily” was added in SPECIFY o include all such gene a. La e ,
Figu e 1.
C ea e axon name p ocess. This p ocess is called by he main p ocess (Fig. 2) i he ull
axon name is no ound. A axon name can ha e se e al in aspeci ic anks, which a e
p ocessed using he same ou ine in he inal loop.
4San os J e al
hese can be easily alloca ed o he igh amily using SPECIFY ools – wi h no loss o he
de e mina ions c ea ed unde hose gene a.
When he axon name exis s in he da abase, a new eco d is c ea ed o each ba code
ead om he image ilenames inside his one olde . A pa icula ins ance is when mo e
han one specimen is moun ed on he same shee . Since he beginning o ou digi isa ion
p og amme, he policy o hese cases has been o cap u e and s o e only one image o
he he ba ium shee . The ilename gi en o his image ile con ains all he ba codes on he
shee ( o example, COI00057276COI00057277.jpg). To cope wi h his, a ou ine will spli
he ilenames o ob ain all he ba codes in each shee and c ea e he co esponding eco d
in he da abase (Fig. 2).
Figu e 2.
P ocess o c ea e eco ds in he da abase om specimen images. I he e is he need o
c ea e a axon in he da abase, he "c ea e axon name" p ocess is called (Fig. 1). One
specimen can ha e mul iple ba codes, which a e ansc ibed in he image ilename. Each
ba code will co espond o one eco d in he da abase, wi h one de e mina ion.
A s a egy o digi ise na u al his o y collec ions wi h limi ed esou ces 5

O e ime, a specimen can be imaged mo e han once and all e sions a e kep because
he mos ecen one is no necessa ily ha wi h he g ea es de ail. Fo image p ocessing,
he i s s ep e i ies whe he images wi h he same name al eady exis in he inal
des ina ion (no conside ing ile ex ension). In his case, a su ix is added o he ilename
(e.g. _1, _2) o a oid o e w i ing p e iously-cap u ed images. Then, he o iginal ile will be
sa ed o he des ina ion. File o ma con e sions occu i needed and copies will hen be
c ea ed (JPGs, humbnails).
The PHP language was chosen o pe o m hese ope a ions because i is independen o
he ope a ing sys em and includes unc ions o copy iles, manage images and c ea e
eco ds in he da abase. In addi ion, he sc ip can be deployed om clien machines using
only an in e ne b owse . To p ocess Ti images, I an iew (Skiljan 2018), a hi d-pa y
so wa e wi h command line op ions, is used, embedded in he PHP code.
S age 2. C owdsou cing pla o m o specimen label ansc ip ion
This pla o m was designed o enable ci izens-use s o ansc ibe he in o ma ion on
specimen labels in o a web o m (h ps://coica alogue.uc.p /explo a o ). The ools al eady
de eloped wi h his pu pose (e.g. Les he bonau es, DigiVol, Zooni e se) lack some
ea u es ha we conside essen ial. The mos ele an a e:
To allow use s o inse and edi da a on ei he one ield o all ields a a ime. Filling
in all specimen ields is ime-consuming and can be a edious job. I is also po en ially a
di icul ask, equi ing a b oad knowledge o axon names, geog aphical names, habi a
and desc ip i e e ms. P o iding one ield a a ime is a good s a egy o main ain use s’
a en ion. Once a ield is submi ed, he nex one is displayed. On each submission, all
alues a e s o ed in he da abase o allow he use o skip o lea e he ask a any momen .
Ne e heless, an ad anced use may p e e o submi all equi ed ields in a single
ope a ion and a ab selec o p o ides his oppo uni y.
To compa e inpu s o he same eco d om di e en use s and o issue ale s o
misma ches. When use s submi a o m, he alue o each ield is compa ed o
p e iously-submi ed alues. I a disc epancy is de ec ed, an ale is shown indica ing all
submissions. The use has he op ion o e u n o he ield o submi a new answe o o
con inue o he nex ield, keeping he ini ial alue. This immedia e co ec ion has he
ob ious pu pose o accele a ing he alida ion, bu i also has he pu pose o educa ing he
use s and accele a ing hei lea ning p ocess.
To ank use s based on p o iciency (ca ego ies). Regis e ed use s a e assigned a
ca ego y based on hei expe ience, which is assessed h ough he numbe o alida ed
use con ibu ions o he pla o m. Fo a ce ain numbe o alid ields, a new ca ego y will
be awa ded (Table 1). This is use ul o p esen ques ions o di e en di icul y le els o
use s acco ding o hei p o iciency ( o example, collec ion da e is an easie ield han
de e mina ion). Addi ionally, he use ca ego y is used o a ibu e con idence o hei
answe s and alues submi ed by mo e expe ienced use s ha e mo e weigh o alida ion.
6San os J e al
Role Ca ego y Desc ip ion Accep ed
submissions
equi ed
Poin s a ibu ed o
each submission
CONTRIBUTOR Basic Fi s ime use . Fields displayed
a e es ic ed.
0 10
Beginne Mo e ields a e displayed, bu
some a e es ic ed.
10 20
Compe en Mo e ields a e displayed, bu
some a e es ic ed.
50 30
Ad anced Mo e ields a e displayed, bu
some a e es ic ed.
100 40
Expe Can submi all ields. 500 50
ADMINISTRATOR Adminis a o Can pe o m all asks abo e, da a
managemen (submission
app o al)
- 60
ROOT Roo Can pe o m all asks abo e,
adminis a o managemen ,
specimen managemen .
- 60
To au oma ically alida e submi ed alues. An au oma ic ou ine is implemen ed o
e alua e pending submissions o alida ion. The simple way would be o compa e he
alues o he same ield o a specimen submi ed by di e en use s; in he case o a
ma ch, accep hose alues. In ou sys em, use le els a e based on use con ibu ions
(p o iciency) and his is u ilised o alida e da a by assigning a con idence alue linked o
he use s a us (Table 1). Valida ion occu s when he sum o poin s o a alue eaches a
de ined numbe o poin s (Table 1). As an example, a alue is alida ed when he sum is
equal o highe han 60 poin s, i.e., a single answe om an adminis a o is enough o ge
alida ion, bu i would equi e six basic use s o one expe and one basic use (o any
o he combina ion ha sums a leas 60 poin s) o submi he same alue o be accep ed
(Table 1).
To allow submission o any alue o any specimen o he collec ion. Despi e he
pla o m's aim o p o ide use s wi h se s o specimens o ganised in missions/p ojec s, he
edi ing p ope ies can be used o edi any specimen. This is use ul o da a edi ing by any
use consul ing he online ca alogue. Fo he edi ion, a hype link in he online ca alogue
ans e s he use o he c owdsou cing pla o m. Fo ha eason, we popula e he pla o m
wi h all he eco ds in he mas e da abase, ega dless o he ac ha hey a e in eg a ed
o no in a mission/p ojec .
To ully in eg a e wi h he main da abase. Da a supplied by he c owdsou cing pla o m
a e added p og essi ely o he he ba ium mas e da abase, no wai ing o he whole
ansc ibing mission o end, no e en o a specimen o be comple ed. A sc ip ha ob ains
Table 1.
Collabo a i e applica ion: use ca ego ies and oles. Valida ion o submi ed da a conside s use ’s
p o iciency as a c i e ion o con idence.
A s a egy o digi ise na u al his o y collec ions wi h limi ed esou ces 7
he lis o alida ed ields om he c owdsou cing pla o m se e (see abo e o alida ion
me hods) h ough an h p eques in an inc emen al way, i.e, i only ob ains he eco ds
since he las eques , ob aining he esul s in JSON o ma and impo ing hem in o a
able. The da a in eg a ion occu s o each alida ed alue by compa ing each one wi h
ha p esen in he SPECIFY mas e da abase. When a ield is emp y (o absen ), i is
w i en (o c ea ed). Reco d c ea ion/edi ion is c edi ed o a speci ic use in SPECIFY
da abase o allow acking changes. When a ield is al eady illed in and is equal o he new
alue, i is conside ed esol ed. I he e is a con lic , hen i is lis ed o he adminis a o o
esol e. Con lic esolu ion is no a li e p ocess, since i equi es he adminis a o o check
he specimen image o he co ec alue, being execu ed la e om a con ol panel whe e
he adminis a o accep s o ejec s alues.
Discussion
The au oma ion sugges ed in S age 1 sa es conside able human labou and ensu es
consis ency, making he specimens in he online ca alogue a ailable wi h minimal
in o ma ion, which can hen be supplemen ed wi h he help o he c owdsou cing pla o m
desc ibed in S age 2.
The c owdsou cing pla o m was in ensi ely es ed wi h ca. 200 s uden s (Feb ua y o Ap il
2020). Tes use s we e asked o achie e expe le el (500 alida ed answe s) in o de o
e alua e he pla o m o beha iou and pe o mance. The pla o m was made publicly
a ailable in Ap il 2020 and a ew new use s ha e become ac i e since hen. Un il now, ca.
30,000 ields we e alida ed, being mos o he inpu om he es use s. When looking a
he alida ed da a, no majo p oblems we e de ec ed, despi e some inconsis ency wi h ou
in e nal pa e ns (pe son names o ema ks). Ne e heless, hose alues could be
conside ed sa e o in eg a e he mas e da abase and e en ually be bulk-co ec ed in ou
egula da abase check-ups. A mo e de ailed quali y check analysis will be made in he
u u e, along wi h o he social pa ame e s.
As online and local sys ems a e in eg a ed, all he in o ma ion lows ac oss he mas e
da abase, he online ca alogue and he c owdsou cing pla o m, wi h e y li le human
in e en ion (Fig. 3).
The p ocedu es desc ibed abo e will con ibu e o inc easing he a e o digi isa ion o
collec ions. Mo e han a wo k low, his cons i u es a pa adigm shi , de aching many
p ocedu es om he collec ion s a . Ul ima ely, mos digi isa ion p ocesses could be un by
olun ee s, om imaging o alida ion. This is o u mos ele ance in collec ions wi h ew
employees who no mally need o be ocused on cu a o ial p ocedu es. Such
implemen a ion enables con inuous digi isa ion o occu , e en i a low pace, ins ead o
being comple ely dependen on unded p og ammes.
8San os J e al
Conclusions
The p oposed app oach inc eases he a e o digi isa ion o specimens. Ou au oma ic
da abasing p ocedu e educes he ime needed when compa ed o pe o ming he ask
manually, i.e. c ea ing he eco d and hen illing up he axon name.
Valida ion o c owdsou ced da a is a sensi i e issue because o he isk o inse ing
inco ec in o ma ion in o he mas e da abase. Howe e , he alida ion me hod desc ibed
he e, based on he use p o iciency, mi iga es ha isk and we conside ha i will gi e
enough con idence o allow he in eg a ion o da a in he mas e da abase.
We emphasise he unique ea u e o he de eloped sys em ha enables he edi ion a any
ime o any online ca alogue da a using he c owdsou cing pla o m.
The in eg a ion o he mas e da abase, he c owdsou cing pla o m and he online
ca alogue esul s in a no el dynamic en i onmen o da a cons uc ion. The di e se
con ibu ions en ich he inal s uc u e.
Funding p og am
PORBIOTA - Po uguese e-in as uc u e o In o ma ion and Resea ch on Biodi e si y
(POCI-01-0145-FEDER-022127)
Figu e 3.
P oposed da a low. Local da abase (Speci y) is managed by i s own so wa e in e ace.
Reco ds can also be c ea ed au oma ically om image p ocessing (desc ibed in S age 1) and
edi ed wi h he c owdsou cing pla o m. Da a is made a ailable in he online ca alogue. Any
da a in he online ca alogue can be edi ed using he c owdsou cing pla o m.
A s a egy o digi ise na u al his o y collec ions wi h limi ed esou ces 9