1
Mapping he Coas al Zone in No h
Eas G eenland
De ec ion o coas al mo phology, idal la s, and
shallow wa e ba hyme y using high- esolu ion
sa elli e image y co e ing si es in Wes and No h
Eas G eenland
Mikkel Høegh Bojesen, Rasmus Eske od Bo gs øm, Ki s y Langley, E a
Mä zle , Alexand a Messe li, Mikkel Lydholm Rasmussen, and Ka l B ix
Zingle sen
Technical Repo no. 112
G eenland Ins i u e o Na u al Resou ces
In collabo a ion wi h Asiaq G eenland Su ey and DHI GRAS A/S
2
Ti le:
Coas al Mapping. De ec ion o coas al mo phology, idal la s, and
shallow wa e ba hyme y using high- esolu ion sa elli e image y
co e ing si es in Wes and No h Eas G eenland
Au ho s:
Mikkel Høegh Bojesen, Rasmus Eske od Bo gs øm, Ki s y Langley,
E a Mä zle , Alexand a Messe li, Mikkel Lydholm Rasmussen, and
Ka l B ix Zingle sen
Publishe :
G eenland Ins i u e o Na u al Resou ces
Final e sion:
29-05-2019
Da e o publica ion:
Financial suppo :
This p ojec is ca ied ou unde a coope a ion ag eemen be ween
he Minis y o Mine al Resou ces & Labou and Asiaq; including a
sub coope a ion ag eemen be ween Asiaq, G eenland Ins i u e o
Na u al Resou ces and DHI GRAS.
P ojec :
P ojec 0043: Mapping he Coas al Zone in NE G eenland
ISSN:
1397-3657
ISBN:
87-91214-90-4
EAN:
9788791214905
Web:
h ps://na u .gl/
Ci a ion:
Zingle sen KB, Ga de E, Langley K, & Mä zle E (2020) Remo eID.
Iden i ica ion o A lan ic Wal us a haul ou si es in G eenland
using high- esolu ion sa elli e images. Technical epo no. 112,
G eenland Ins i u e o Na u al Resou ces, G eenland. ISBN
3
English Summa y 4
G eenlandic summa y – Kalaalissu Naalisa neqa ne a 4
Danish summa y – sammen a ning på dansk 4
1. In oduc ion and Backg ound 5
1.1 Sa elli e based analysis 5
1.2 Field Si es 5
1.3 Fieldwo k ac i i ies 7
2. Da a and Me hods 9
2.1 Sa elli e de i ed ba hyme y 9
2.2 In e idal zones 10
2.3 Coas al classi ica ion 11
2.4 Field su ey me hods 13
3. Resul s 14
3.1 Sa elli e de i ed ba hyme y 14
3.2 In e idal zones 16
3.3 Coas al classi ica ion 18
3.4 Valida ion 21
3.5 Findings and lessons lea ned 22
4. Conclusions & Pe spec i es 24
Acknowledgemen s 25
Re e ences 26
Appendix 1: O ganiza ion 28
Appendix 2 Da a Dis ibu ion 29
Appendix 3: Decision ee o inal coas al classi ica ion 30
4
English Summa y
This epo highligh s he indings o he mapping o he coas al en i onmen on wo loca ions in
G eenland based on analysis o sa elli e image y. The s udy includes sa elli e de i ed ba hyme y,
mapping o he in e idal zones, and classi ica ion o he coas al a eas, which we could map
success ully. The analysis in ol ed algo i hms applied o high- and medium- esolu ion sa elli e
image y, alida ion wi h ield measu emen s, and mo phological analysis om coas line da a. The
s udy success ully demons a ed he easibili y o using sa elli e-based analysis o emo e a eas, bu
also highligh ed he logis ical complexi ies o ield wo k conduc ed he e. Ne e heless, he
echnologies hold high po en ial o e icien la ge-scale mapping in he A c ic.
G eenlandic summa y – Kalaalissu Naalisa neqa ne a
Uani nalunaa usiami Kalaalli Nunaanni sumii inni ma lunni sine issap qani uani immap
assilio ne i ine ne i e se sinnia pagu , aakkua sa elli ’ ikku assilisa a o lugi nalile so neqa ne i
aallaa igalugi suliaappu . Misissuine mi pineqa pu immami ikka umi immap i issusaa
sa elli sikku paasissu issa aallaa igalugi suliaq, immap ini a ne a a uli a ne a alu
nalunaa so neqa ne i aammalu sine issami sumii ii immikkoo i e ne i, ilua sillua umik
nalunaa usio neqa sinnaasimappu . Ma ema ikkimik suliassa naammassisa nissaannu maligassa
a o lugi sa elli sikku assilisa pi saalluinna u naammagina umillu pi saassuseqa umik assilisa
a o neqa pu , naa so suine i sumii inni a o sinnaassusii pineqa ullu ilusai isikkuilu
ine ia o ne minnilu ullumimu isikko ile simasaa misissui igineqa simappu . Misissuine i
akussu issaqa ile paa sumii inni ikikkuminaa suni ungasissumii unilu nalunaa suissagaanni
sa elli ’ i a o neqa sinnaallua u , aammalu sumii innu o nigulluni sulinia aanni qanoq
a aqa igiissaa neqa nissaasa pisa iu iginissaannu akussu issaqa i sile poq. Taamaakkalua oq
Issi umi nalunaa suinissamu eknologiikku pe ia issaqa lua oq e se sinneqa poq.
Danish summa y – sammen a ning på dansk
I denne appo emhæ e i esul a e ne a ko lægningen a kys miljøe o s ede i G ønland
base e på analyse a sa elli billede . Unde søgelsen indbe a e ha dybde på la and be egne a
sa elli da a, ko lægning a ide andszonen og klassi ice ing a kys om åde , som kunne ko lægges
med success. Analysen bes od i algo i me an end på sa elli billede a høj og medium opløsning,
alide ing a målinge i el en og mo ologisk analyse a kys linjens o løb. Unde søgelsen
em is e mulighede ne o a b uge sa elli base e analyse i je n liggende om åde , men
emhæ ede også de logis iske kompleksi e e a el a bejde på s ede . Ikke des o mind e ha
eknologie ne e s o po en iale o e ek i ko lægning i s o skala i A k is.
5
1. In oduc ion and Backg ound
Mo i a ion
The coas line o G eenland is long and complex, bo h in i s o m and in composi ion. Since he
p edominan pa o he coun y consis s o wild na u e wi hou oads, he coas al a eas a e he main
anspo a ion ou es. The e a e a mul i ude o ske ies, a chipelagos, islands and isle s, which a e
uncha ed, and hese p esen a signi ican haza d and isk o acciden s. The lack o na iga ional
cha s and accu a e opog aphic maps a e a pa icula p oblem in he emo e a eas o G eenland,
such as No h Eas G eenland. Wi h hyd oca bon explo a ion expanding e e u he no h i is
essen ial o ill-in c ucial missing in o ma ion on he coas al en i onmen .
The ecen ad ances in sa elli e emo e sensing and image analysis ha e inc eased he po en ial o
ob ain de ailed insigh s, and upda ed in o ma ion abou ulne able and emo e coas al en i onmen s
in he A c ic.
F eely a ailable Sen inel sa elli e image y om he Eu opean Space Agency (ESA) Cope nicus
p og amme, supplemen ed wi h ad anced comme cial sa elli e image y ep esen s an ob ious da a
sou ce o upda ing hese maps.
In his epo we highligh he indings o he coas al mapping p ojec s based on sa elli e image y. I
includes sa elli e de i ed ba hyme y, mapping o he in e idal zones, and classi ica ion o he
coas al a eas, which we could map success ully in wo egions o G eenland.
1.1 Sa elli e based analysis
In his s udy, we ha e demons a ed how sa elli e image y can be used o mapping he coas al
zones in he emo e a ea o No h Eas G eenland. Addi ionally, we de eloped and applied me hods
o de i ing shallow wa e ba hyme y, in e idal zones, and classi ying di e en coas al ypes. The
analysis was ca ied ou using 1) high esolu ion, comme cial sa elli e image y o selec ed a eas,
and 2) eely a ailable medium esolu ion sa elli e da a, which has a la ge spa ial co e age and
he e o e allowed o an upscaling o he a ea o in e es . Field measu emen s we e ca ied ou in
o de o alida e he sa elli e-de i ed p oduc s.
This p ojec se ed as a easibili y s udy ha success ully demons a ed he use o sa elli e-based
me hods o mapping he coas al cha ac e is ics o he s udy a ea.
The challenges encoun e ed du ing he p ojec (such as sea ice co e age, u bidi y, and small idal
le el a iabili y) ha e helped us o de ine he limi s o he me hods, he eby educing he isks o
u u e applica ions.
These posi i e esul s open up he possibili y o applying he newly de eloped me hods o o he
coas al a eas in G eenland, and hus holds an eno mous po en ial o a cos -e icien la ge-scale
mapping o he coas al zone in A c ic.
1.2 Field Si es
Two a eas o in e es (AOI) - one in Wes G eenland and ano he in No h Eas G eenland (Figu e
1.2.1) we e chosen o his pilo s udy.
6
Figu e 1.2.1. Map o G eenland indica ing he wo
AOIs used in his s udy, Nipisa in Wes G eenland
and Vega Sund in No h-Eas G eenland.
Wes G eenland AOI – Nipisa :
Nipisa is a shallow-wa e inle loca ed close o he
capi al o Nuuk (Figu e, 1.2.2). The a ea is an inle o
Nuup Kange lua (Nuuk Fjo d), which has a dep h
ange om 0.5m o 20m and a idal ange o 3-4m.
This AOI was used o es ablish he me hodological
amewo k, da a managemen and analysis p ocedu es
o he s udy. Addi ionally, he dep h measu emen s
acqui ed in Nupisa we e used o calib a e and
alida e he sa elli e-de i ed ba hyme y. The
p oximi y o he AOI o Nuuk allowed o e ec i e
and sa e collec ion o g ound-based alida ion da a
used du ing he de elopmen phase o he p ojec .
Nipisa Sund was isi ed using a boa p o ided by he
G eenland Ins i u e o Na u al Resou ces.
Figu e 1.2.2. Field si e Nipisa Sund, a
shallow-wa e inle close o Nuuk
No h Eas G eenland AOI – Vega Sund:
Wi hin No h Eas G eenland, we ha e
ocused on especially ulne able coas al
a eas, o example b eeding g ounds o
bi d colonies, a eas wi h in e es ing
seaweed deposi s o a chaeological si es
ha a e h ea ened due o coas al e osion.
These AOI’s we e selec ed wi h inpu
om biologis s and a chaeologis s (Fig.
1.2.3). High- esolu ion sa elli e image y
o he AOI’s we e examined in
conjunc ion wi h nea -sho e ba hyme y
da ase s. The Vega Sund a ea was
highligh ed as he mos app op ia e
loca ion due o i s ulne able and
aluable seaweed deposi s, which would
be a isk in case o an oil spill inciden . Vega Sund is a shallow a ea be ween Fos e Bug bay and
he inne jo ds, loca ed no h o he own o I oqqo oo mii (Sco esbysund). A chi e sa elli e
images show ha he a ea has been ee o pa ially ee o sea ice co e du ing he las yea s om
beginning o augus , deeming i a sui able AOI o sa elli e-de i ed ba hyme y analysis.
7
Figu e 1.2.3. The h ee ield si es o No h Eas G eenland
T a elling and logis ics o ield su eys in No h Eas G eenland a e challenging, equi ing
expensi e cha e s o anspo and collabo a ion wi h o he ope a o s o minimize cos s and
enhance sa e y. The su ey eam collabo a ed in ensely wi h NANOK, The No h-Eas G eenland
Company, a non-p o i o ganiza ion es o ing cul u al he i age cabins in he Na ional Pa k. Toge he
we sha ed logis ical pla o ms: cha e o ai planes, sha ed cabin accommoda ions, and as
passenge s on hei boa a elling owa ds he s udy a ea. The locally based mili a y uni s o he
Mes e s ig S a ion and he Ella Ø S a ion (Si ius Dog Sled Pa ol) o he Join A c ic Command, he
Danish De ence, p o ided logis ical suppo o d i ing, accommoda ion and boa anspo . Such
collabo a ion and kind assis ance wi h exis ing logis ical esou ces in he a ea highly minimized he
cos s and suppo ed he ope a ions bu canno be aken o g an ed in u u e cases.
1.3 Fieldwo k ac i i ies
A wo kshop wi h all p ojec pa icipan s was held in Nuuk du ing Sp ing 2018. The goal o he
wo kshop was o p epa e ield ac i i ies, and o es he p elimina y nea -sho e ba hyme y analysis.
The wo kshop included collec ion o g ound u hing da a in a shallow idal zone close o Nuuk.
Fieldwo k was conduc ed in No h Eas G eenland o ou weeks in July-Augus 2018, planned o
coincide wi h minimal snow co e and low sea ice ex en . Howe e , abno mally high snow co e
and a sho mel season c ea ed un a ou able condi ions o bo h ae ial and boa logis ics. Thus, as a
esul o he abundan sea ice co e in Vega Sund, access o Vega Sund was limi ed and he
majo i y o he ieldwo k was ins ead ca ied ou a Ella Ø and Mes e s ig.
The Ella Ø S a ion ha bou was accessed ia a nea by landing s ip on T ail Ø a Holms Bug wi h
suppo om he ci ilian and mili a y ope a o s. F om he e, Ella Ø was accessed by boa wi h
8
assis ance om he Si ius Dog Sled Pa ol. A ieldwo k ac i i y p og amme was subsequen ly
es ablished, which included se ing up a GNSS base s a ion o con inuous posi ioning and
ins alling wo idal gauge s a ions o con inual idal baseline measu emen s. The eam co e ed he
nea sho e wa e wi h an in la able boa o dep h mapping wi h a Ga min echo sounde , posi ioned
using he GNSS base s a ion. Exposed ocks and ske ies we e mapped, and physical dep h
measu emen s we e conduc ed using a lead weigh . On land, he eam documen ed he coas al
mo phology, geological p ope ies and ege a ion co e wi h geoloca ed pho os. The g ound-
u hing da a collec ed he e was used o alida e he shallow ba hyme y and coas al zones. The
coas al wa e s a ound Ella Ø a e deepe han hose in he o iginal AOI a Vega Sund. This esul ed
in a mo e limi ed nea -sho e ba hyme y su ey and idal zone classi ica ion han in ended. Despi e
he challenging logis ical condi ions, his did no ha e a nega i e impac on he o e all ou come o
he p ojec .
A sho ip o Vega Sund was possible a he end o he ield campaign hanks o logis ical suppo
om he Si ius Dog Sled Pa ol. The AOI could be isi ed o 3-4 hou s, wi hin which echo sounds
measu emen s and pho og aphy o he coas line we e comple ed.
9
2. Da a and Me hods
2.1 Sa elli e de i ed ba hyme y
Ad ances in image analysis, mean i is now possible o es ima e he dep hs o clea , shallow wa e s
using op ical sa elli e image y.
The ad an age o sa elli e-de i ed ba hyme y (SDB) a e ha i is cos e ec i e, has a la ge spa ial
co e age o emo e and dange ous egions, a a high empo al esolu ion. The main disad an age o
he me hod is ha i elies on he use o op ical sa elli e da a, which has a limi ed wa e dep h
pene a ion (a ound one secchi dep h). The SDB me hod is mos e ec i e in clea wa e s, as he
pene a ion dep h dec eases in ela ion o inc eased wa e u bidi y, low ligh , and he p esence o
ice.
The SDB’s in his s udy we e p oduced using Sen inel-2A & Wo ldView-2 sa elli e image y. The
high spa ial esolu ion o Sen inel-2 and Wo ldView-2 image y a 10m and 2m espec i ely, allows
o de ailed ba hyme ic mapping. This is o a highe spa ial esolu ion han exis ing ba hyme ic
da ase s.
Theo e ical amewo k o SDB analysis
The e a e a numbe o app oaches ha can be used o de i e SDB. The backg ound o some key
me hods a e b ie ly ou lined below.
Fo his p ojec , a DHI GRAS p op ie a y physical adia i e ans e model (ex ended e sion o
Guzinski e al. 2016; Klonowski e al, 2007; Lee e al. 1998, 1999, 2001) was applied. O he
me hods o no e used o de i ing SDB a e empi ical models, based on he ea ly wo k o Lyzenga,
1978, e ined by S ump , 2003. Addi ionally, adia i e ans e -based models, such as he
SAMBUCA model de eloped by CSIRO (B ando & Dekke , 2003; We le & B ando, 2006) can
also be applied. The undamen al di e ence be ween he empi ical models, and he adia i e
ans e models, is ha he empi ical models calcula e an index ha co ela es wi h dep hs, bu has
o be ans o med in o wa e dep hs, while he adia i e ans e me hods model he wa e column as
p ima ily a unc ion o he wa e dep h and he seabed e lec ance.
The model used in his s udy minimizes he di e ences be ween an obse ed sa elli e image and a
modelled sa elli e image, which is c ea ed as a unc ion o six pa ame e s – dep h, bo om ype,
backsca e ing, chlo ophyll-A, gelbs o , and he slope o he backsca e ing unc ion. Th ough
minimizing he di e ence be ween he obse ed and he modelled sa elli e images, accu a e wa e
dep hs can be e ie ed in op ically shallow wa e s, meaning ligh e lec ing om he seabed is
obse ed.
Essen ially, he SDB can be educed o he ollowing se ies o pseudo equa ions:
𝑟𝑟𝑠𝑚𝑜𝑑𝑒𝑙𝑙𝑒𝑑 = 𝑓(𝑐ℎ𝑙𝑜𝑟𝑜𝑝ℎ𝑦𝑙𝑙, 𝑔𝑒𝑙𝑏𝑠𝑡𝑜𝑓𝑓, 𝑏𝑏, 𝑏𝑥, 𝜌, 𝐻)
Whe e 𝑟𝑟𝑠𝑚𝑜𝑑𝑒𝑙𝑙𝑒𝑑 is he modelled sa elli e image, 𝑏𝑏 is he backsca e ing, 𝑏𝑥 is he slope o he
backsca e ing unc ion, 𝜌 is he summed up bo om e lec ance, and 𝐻 is he dep h.
The nex s ep is o minimize he di e ence be ween he modelled and obse ed sa elli e images:
𝜒2=1
𝑁 [∑(𝑟𝑟𝑠𝑜𝑏𝑠𝑒𝑟𝑣𝑒𝑑 − 𝑟𝑟𝑠𝑚𝑜𝑑𝑒𝑙𝑙𝑒𝑑)]
Wi h N being he numbe o spec al bands, whe e he sa elli e image p o ides in o ma ion, and
𝑟𝑟𝑠𝑜𝑏𝑠𝑒𝑟𝑣𝑒𝑑 is he obse ed sa elli e image.
16
The co ela ion be ween he su ey and he sa elli e da a in Figu e 3.1.3 is high in a eas whe e he
wa e is clea . Howe e , a la ge pa o he wa e in he sa elli e image has e y high sedimen
loads, which obscu es he seabed and makes SDB impossible. As a esul , he a eas a ec ed by
hese sedimen s ha e been emo ed om he da ase . The sedimen co e obscu ed he o e lap
be ween he wes e nmos sa elli e image and he in-si u da a, and he e o e he SDB was no
ex ended o he eas e n ee s a he ip o Geog aphical Socie y Island.
3.2 In e idal zones
The esul o he in e idal zone wo k low is a as e showing land (pixel loca ion always classi ied
as land), wa e (pixel loca ion always classi ied as wa e ), and a eas ha ha e some imes been
classi ied as land and some imes as wa e (Figu e 3.2.1). These la e egions, when loca ed along
he sho eline, a e in e p e ed o be in e idal zones. A mo e p ecise delinea ion o he in e idal a ea
can be ob ained by de ining land and wa e h esholds based on low and high ide scenes.
Figu e 3.2.1. Vega Sund egion. Uppe ow: Sen inel-2 scenes showing wa e co e age du ing low
and high ide. Lowe le : o each pixel loca ion, he numbe o pixels included in he mosaic.
Lowe igh : esul ing mosaiced as e indica ing a eas o land, wa e and in e idal zones.
The NE G eenland si es o Vega Sund and Ella Ø ha e a ela i ely small idal each o abou 1m.
The Vega Sund a ea has a numbe o i e del as ha ha e c ea ed shallow del as along he
coas line and which esul in na ow in e idal zones a he del a mou hs, and in some inle s (Figu e
3.2.1).
17
A ound he island o Ella Ø, he s eep sided deep jo ds esul ed in only a ew in e idal zones being
iden i ied wi h he sa elli e image y (Figu e 3.2.2). I was clea du ing he ieldwo k ha he e a e
na ow beaches and mul iple ske ies on he no he n cen al coas o Ella Ø, howe e he pixel
esolu ion o 10m o he sa elli e image y likely hinde ed he de ec ion o hese.
Figu e 3.2.2 Ella Ø si e. Le : Mosaiced as e indica ing a eas o land, wa e and in e idal zones.
No e ha sea ice can be misclassi ied as “land”, which can be seen in he cen al pa o he igu e.
Righ : Sa elli e image o Ella Ø a ound he ime o ieldwo k, no e he pa ches o b oken-up sea ice
isible in he igh sec ion o he image (Sa elli e image: SnapPlane ).
In con as o NE G eenland, he es a ea in Nipisa Sund, Nuuk Fjo d, has a much la ge idal
ange (3-4m). The shallow, gen ly sloping sound has a la ge in e idal zone and nume ous small
islands ha a e isible (Figu e 3.2.3).
Figu e 3.2.3 Nipisa Sund, Nuuk Fjo d. Righ : Mosaiced as e indica ing a eas o land, wa e and
in e idal zones. Le : A sa elli e image, o Nipisa Sund a ound he ime o ieldwo k (Sa elli e
image: SnapPlane ).
18
3.3 Coas al classi ica ion
Complexi y index
To explo e he spa ial dis ibu ion o he complexi y o he coas line, a mul i a ia e g ouping
analysis, and a clus e and ou lie analysis was ca ied ou . The g ouping analysis showed ha 4
g oups bes ep esen ed he a ia ion in he da a, bu also ha he a ia ion was conside able wi hin
each o hese ou g oups. This made i impossible o dis inguish he ou g oups s a is ically a he
a ying dis ances (0,1km; 1km and 10km). In he clus e and ou lie analysis we ound ha he
spa ial au oco ela ion was signi ican , meaning ha he a eas o high and low complexi y a e
signi ican ly clus e ed. We used he Anselin Local Mo an’s I me hod o u he in es iga e he
loca ion o he clus e ing. In clus e and ou lie analysis one e y impo an aspec is o de ine
neighbou ing poin s, since hese will be used in he s a is ical analysis compa ing neighbou ing
obse a ions. Fo his analysis, neighbou hood dis ances we e calcula ed o be 2750m using Global
Mo an’s I. The esul ing Figu e 3.3.1, b, shows clus e s and ou lie s. High alue clus e s means ha
high alues a e su ounded by o he high alues, ela i e o each o he (pink do s) and low alue
ou lie s, whe e low alues a e su ounded by high alues (da k blue do s).
Figu e 3.3.1 a & b. Complexi y Index measu emen s (a) and clus e and ou lie analysis (b) in
Vega Sund.
In e ec Figu e 3.3.1, a, shows he absolu e complexi y a a de ined scale (1km) whe eas Figu e
3.3.1, b, shows he complexi y o each obse a ion ela i e o i s neighbou s. In combina ion hese
wo igu es allow us o compa e he signi icance o complexi y in he case oil spill eme gency
esponse planning, a a egional le el.
a
b
19
The complexi y index analysis is scale dependan , and he e o e he scale selec ion o he analysis
depends on he le el o de ail needed. Fo example, in he case o an oil spill eme gency esponse,
in o ma ion on he size and sp ead o he oil spill would be impo an ac o s in selec ing he co ec
scale. I is equally impo an o de e mine how much a ia ion exis s along he coas line in o de o
se an app op ia e scale o analysis.
A chipelagos
As seen below in Figu e 3.3.2, la ge pa s o Vega Sund a e ega ded as an a chipelago. This
classi ica ion is no excluded, since impo an in o ma ion abou coas line shape, i.e. complexi y
index, and coas line con en , i.e. su ace ype in o ma ion om he supe ised classi ica ion, can
also eed in o he oil spill in o ma ion sys em.
Figu e 3.3.2. A chipelagos de ined as g oup o islands whe e he a io o he a ea o he wa e o
he a ea o he land is as high as 9:1.
20
Supe ised classi ica ion
Ella Ø has a complex coas line wi h a a ie y o coas al ypes. As seen in Figu e 3.3.3 a ocky (and
s eep) coas domina es he sou he n side o Ella Ø, whe eas much mo e di e se coas al su aces a e
ound on he wes e n side o he island. He e a combina ion o sand, lu ial lood plains,
unconsolida ed ock and o he geology/ ege a ion a e ound. This esul s in a complex and di e se
en i onmen , which is e y di icul o clean up in an oil spill esponse scena io.
Figu e 3.3.3. Final classi ica ion o coas al composi ion a Ella Ø.
21
3.4 Valida ion
Valida ion o he supe ised classi ica ion is impo an in o de o assess he accu acy, plausibili y
and obus ness o he classi ica ion. In o de o assess he classi ica ion, e i ica ion pho os we e
aken a he ield si e, as desc ibed in sec ion 2.4. Th ee example alida ion pho os a e shown in
Figu e 3.4.1.
Figu e 3.4.1. Valida ion o classi ica ion, using e i ica ion pho os. A, shows semi ege a ed a ea,
B shows an example o a ocky coas and C, shows an unconsolida ed ock beach.
Pho o A in Figu e 3.4.1 shows a ege a ed a ea wi h small a eas o clea s anding wa e . Acco ding
o he classi ica ion his loca ion is classi ied as wa e . Despi e also being ege a ed his a ea is
clea ly in luenced by wa e . Pho o B shows how he ocks di ec ly mee he sea, o ming a s eep
ocky coas , which is also depic ed on he classi ied map.
One challenge when making a supe ised classi ica ion is ha e en wi hin each 10m pixel mo e
su ace classes a e ep esen . In e ec each pixel on he map depic s which o he possible classes is
he mos abundan wi hin ha a ea o 100m2.
Pho o C is aken in an a ea cha ac e ised by a loose g a el su ace wi h a ce ain deg ee o mois u e
due o he ege a ion. The a ea is classi ied as lu ial lood plain, sand, ock and snow. This loose
s uc u e o he su ace could also be e med as unconsolida ed ock. This example demons a es
ano he challenge wi h a supe ised classi ica ion. One has o de ine classes o classi ica ion
de ailed enough o ep esen he deg ee o a ia ion in he en i onmen in ques ion, and a he same
ime de ine classes b oad enough o end up wi h an applicable classi ica ion o manage ial
pu poses. Thus, he ques ion is how o de ine obus and applicable class bounda ies.
22
Despi e he unce ain ies discussed abo e, he classi ica ion gene ally i s well wi h he alida ion
pho os and a isual inspec ion o he VHR image y. Fo a u he quan i a i e alida ion o be made,
a e e ence da ase is needed. Un o una ely, such a e e ence da ase is no a ailable o his s udy.
A mo e widesp ead alida ion campaign was planned in Vega Sund, howe e , due o poo wea he
condi ions, only a small ac ion o his was ca ied ou . In u u e s udies he planning o alida ion
sampling should ake in o accoun sampling a ea size and spa ial dis ibu ion ela i e o he size o
he AOI since his will in luence he s eng h o alida ion da a, and o e all end p oduc .
3.5 Findings and lessons lea ned
Many aluable expe iences ha e been gained h oughou he en i e p ojec ; om he g ound
u hing ieldwo k, o sa elli e da a acquisi ion and da a analysis. We hope ha hese expe iences
will enable us o adap ou me hods o u u e p ojec s.
The main indings can be lis ed as ollows:
P oduc
Nipisa, W
G eenland
NE
G eenland
sa elli e de i ed nea -sho e ba hyme y
map up o 8 m dep h
X
X
in e idal zone map
X
X
coas al complexi y
X
eigh coas al su ace classes
X
obus alida ion se up (see epo )
X
X
De ails o some o he challenges encoun e ed in di e en a eas o he p ojec a e ou lined below.
Challenges in classi ica ion
In he classi ica ion o he coas al zone we paid a en ion o a a ie y o su ace ma e ial classes. In
doing so we encoun e ed some challenges ha can be accoun ed o and a oided in u u e wo k.
The su aces we mapped a e o some deg ee in luenced by wa e and o he biophysical ac o s.
Since we a e wo king wi h op ical da a, he spec al signal a ia ion will a y ac oss he s udy a ea.
One issue o conce n is whe he he aining a eas and alida ion a eas a e ep esen a i e o he
a ia ion ac oss he s udy a ea, since his will in luence he accu acy o he inal classi ica ion.
Challenges wi h SDB
In he p oduc ion o SDB o he s udy a ea, he challenges ela ing o he complica ed geog aphy o
G eenland became appa en . The high sedimen load om mel ing glacie s obscu es he seabed in
many a eas. This made i e y di icul o de i e SDB in hese egions, as well as limi ing he
po en ial ex en o he G eenlandic coas line ha can be mapped wi h his me hod. Ano he
challenge ied in o he complex geog aphy is he apid changes in wa e dep h, caused by he ocky
coas line in pa s o G eenland. This was highligh ed a he Ella Ø s udy si e. In hose a eas, he
23
10m spa ial esolu ion o Sen inel-2 da a may no be su icien o accu a ely ep esen he
ba hyme y, and he e o e a la ge sca e is likely o be obse ed in alida ion da ase .
Challenges wi h ield wo k ope a ions
In o de o assess and alida e he sa elli e da a o ba hyme ic mapping, in e idal a ea and e ain
classi ica ion, in-si u da a collec ion is s ill cu en ly p e e ed. The ieldwo k ca ied ou in No h
Eas G eenland was bo h expensi e and ime-consuming. Like all ieldwo k in G eenland, i was a
imes somewha unp edic able due o di icul wea he condi ions, localised sea ice condi ions and
hus lead o some challenges wi h logis ics o he da a collec ion. Whils he su ey eam adap ed o
he dynamic en i onmen al condi ions and was e icien in da a collec ion when condi ions allowed,
ne e heless only a ac ion o he o iginal planned da a was collec ed. This highligh s he need o
a mo e lexible da a acquisi ion plan o bo h he g ound-based da a and sa elli e da a o e he s udy
a ea.
Fo he in-si u da a collec ion, i is ecommended o use high quali y ins umen s, such as a mul i
beam echo sounde in conjunc ion wi h GNSS GPS sys ems o allow o mo e comple e mapping o
he seabed and accu a e geog aphic posi ioning.
24
4. Conclusions & Pe spec i es
As demons a ed du ing his p ojec , sa elli e de i ed in o ma ion holds an eno mous po en ial o
ob aining geo-in o ma ion o emo e and challenging egions, such as No h Eas G eenland. By
aking ad an age o eely, and/o comme cially a ailable sa elli e image y, ad anced image
analysis and local knowledge i is possible o p o ide a de ailed cha ac e isa ion o he coas al zone
in a c ic wa e s. The demons a ed me hods ha e he ad an age o being cos -e icien , objec i e
and wi hou he isks associa ed wi h adi ional su ey me hods.
The de i ed map p oduc s ha e all been de eloped wi h sa elli e-based in o ma ion and a e
p oduced wi h gene ic and ans e able image analysis me hods. This acili a es a la ge upscaling
o he maps o ex end he co e age o e la ge and poo ly mapped egions o he a c ic.
The de eloped me hods and he de i ed esul s o he p ojec will be included in he p oduc ion o
Oil Spill Sensi i i y A lases in he A c ic. An impo an elemen o he exis ing Oil Spill Sensi i i y
A lases o coas al G eenland, ocuses on he analysis o he oil spill esis ance o he coas .
T adi ional analysis is based on manual in e p e a ion o he gene al coas al mo phology and
geology om a ailable opog aphic maps, a ailable ae ial pho og aphy and low- esolu ion sa elli e
images, and how hese ac o s de e mine how oil om a spill will be abso bed by he coas al
ma e ials o washed o . In his ega d, he in o ma ion o in e es is he p edominan ly he slope o
he coas ’s e ain, s aigh coas lines o small beaches, gene al geology o ock, boulde s, pebbles,
sand o sedimen s.
The classi ica ion p ocedu es ou lined in his s udy, using medium esolu ion Sen inel-2 (10m) and
Wo ldView-2, 2 m esolu ion sa elli e images, p o ides he possibili y o apply quan i a i e and
au oma ed analysis me hods on la ge image da ase s wi h inc eased empo al esolu ion.
Addi ionally, his allows o au oma ed ime-se ies analysis, he eby educing he biases in oduced
by adi ional manual digi isa ion me hods.
We expec he coas line complexi y index, he su ace classi ica ion and idal zone p oduc s o he
analyses o be use ul in he p epa a ion o u u e Oil Spill Sensi i i y A lases. The complexi y index
can de ec s aigh o con ou ed coas lines and iden i y he isk o concen a ions o oil in pocke
beaches o o he complex mo phologies. The su ace classi ica ion can play a i al ole in
de e mining he ypes o geology ypes ha abso b o epel oil.
The idal zone mapping and SDB would be new ea u es o an Oil Spill Sensi i i y A las in
G eenland. These p oduc s ha e no p e iously been in es iga ed wi hin he a las amewo k. By
applying empo al analysis o he idal zone mapping and no el algo i hms o shallow wa e
ba hyme y mapping, new c i ical a eas o impo an ma ine o idal habi a s ha a e highly
sensi i e o oil spills a e iden i ied. These highly sensi i e a eas would o he wise equi e cos ly and
de ailed ieldwo k in he emo e a eas in o de o asce ain he same le el o de ail. To suppo he
in as uc u e o ope a ions in he a ea, he SDB can be u ilised o designa ing possible na u al
ha bou s and bays o cap u e oil esidues. This new in o ma ion would be bene icial o he a las, as
oil spill ma e ials a e o en sa u a ed in o he idal soils, depending on he geology. Consequen ly,
we ecommend he implemen a ion o high- esolu ion sa elli e image analysis in he mapping o oil
spill sensi i e a eas and spill emo al in as uc u e.
25
Acknowledgemen s
The wo k ca ied ou in his p ojec has been inanced ia he G eenland Mine als Au ho i y. The
p ojec pa ne s a e hank ul o he wo k oppo uni ies ha has been made possible by his unding.
The p ojec pa ne s also wan o exp ess hei g a i ude owa ds a numbe o ins i u ions who ha e
p o ided assis ance and knowledge sha ing, which has been o g ea alue o he ou come o his
p ojec . These ins i u ions a e:
● No døs g ønlandsk Kompagni Nanok
● Join A c ic Command Head Qua e Nuuk
● Join A c ic Command S a ion Danebo g and Ella Ø (Si ius Dog Sled Pa ol)
● Join A c ic Command S a ion Mes e s ig
● Royal Danish Na y, 1. Squad on HDMS H idbjø nen (F360)
A special hank goes o Nanok Kompagnie , which p o ided us esidence a Ella Ø, and helped us
wi h in aluable ad ice du ing he ield wo k and blessed us wi h g ea s o ies om hei ad en u es
as Si ius pa ols. Also, a big hank you o Join A c ic Command and Si ius, who helped us wi h all
kinds o logis ics. Thanks o you we succeeded wi h ou p ojec .
