Prince Edward Island
Wind Assessment
Demonstrating the potential of SAR for Sites classification
Audrey Lessard Fontaine, Thomas Bergeron, Monique Bernier, Karem
Chokmani et Gaétan Lafrance
31/03/2010
1
ISBN : 978-2-89146-785-8
Table of Contents
Introduction
L.t
Background.
1..2
Objectives
1.3
P r o j e ct te a m . . . . . . . . . . .
7.4
Defiverables
1.5
Report's
structure.....
. . . . . . . . . . . . . . . . . .1. . . . . . .
. . . . . . . . . . . . .1. . . .
....................2
.........................2
.................2
.......................
2
How is it possibleto retrievewind speedfrom radarsatelliteimagery?
...........
3
........................
3
2.1
First,what is Radar?
2.2
signal?.........
Howa wind mapis obtainedfrom a backscatter
2.3
yone?..........
W h a ti st h eV a l i d i t Z
. . . . . . . . . . . . . . . . . .4. . . . . . .
2.4
Whatarethe benefitsof SARwith respectto met masts?
4
.........................
...................
3
to get a precisepictureof the wind
would be necessary
2.5 How manysnapshots
Weibull'sparameters?
andassess
speedfrequency
analysis
Methodology..................
3.1
A v a i l a b lSeA Rd a t a
3.2
Correlation
offshore-onshore.............
3.3
D i r e c t i o n a l c o m p o smi taep s . . . . . . . . . . .
3.4
R a n k i nogf s i t e s . . . . . . . .
Results
........................
5
. . . . . . . . . . . . . . . . . .7. . . . . . .
.......7
.................
8
. . . . . . . . . . . . . .1. .0. . .
. . . . . . . . . . . . . . . .t.2. . . . .
...........
15
4.7
for Wei Candata............
SARMet mastcorrelation
.............
15
4.2
for Environment
Canada
SARmet mastcorrelation
.............
18
4.3
D i r e c t i o n a l c o m p o smi a
t ep s . . . . . . . . . . .
4.4
R a n k i nogf s i t e s . . . . . . . .
. . . . . . . . . . . . . .1. .8. . .
. . . . . . . . . . . . . . . .2. .0. . . .
5
D i s c u s s i ocno,n c l u s i oann dr e c o m m e n d a t i o. .n.s. . . . . .
6
Aknowledgements..
...........25
7
Bibliography............
...........27
8
Appendix.................
...........29
8.1
9
....................23
Howto obtainbetterresultswhenextrapolating
seameasures
on land?...
................29
A p p e n d iAx . . . . . . . . . . . . .
. . . . . . . . .3.3.
9.1
A.1Howto makeRoughness
mapwith GIS: ............
9.2
A.2Stepsto makea roughness
map:...........
.....33
9.3
A.3 Howto useelevationdatawith digitolelevotionmodels(DEM)?
.....37
10
A p p e n d i8x. . . . . . . . . . . . . . . . . .
10.1 Sitesdescriptions
:..........
tO.2 Unvisited
sites............
............
33
. . . . . . . . . . . . . . . .3
. .9. . . . .
...............
39
...................
48
1
Introduction
for
This documentis the final report of the first phaseof the SARwind resourceassessment
communityrinkswithin the collaborationframeworkbetweenINRSand WElCan.This report
the potential
providesthe resultsobtainedfrom a limiteddatasetof SARimagesto demonstrate
of SAR imagesand the type of resultsobtainedfrom this technique. This project is an
applicationof WESNetProjectNumberP1.1c- High ResolutionSurfaceWinds Mappingin the
CoastalZonefrom SARSatellitelmagery
1 . 1 B a ckg ro u n d
of
ln 2008 the PEIgovernmentinitiateda projectto supportsmallwind turbine installations
capacityup to 100kW.Among25 communityrinksinterestedin the project,five siteshavebeen
selectedto be funded. For any wind turbine project,the knowledgeof wind resourcesis a
criticalelementof decisionmaking. However,for smallwind turbine projects,traditionalwind
programsmay appeartoo expensive.
ln addition,the shorttime framefor
resourceassessment
executingthe completeproject,aboutsix monthsfor this particularprogram,requiresa means
and rankingthe wind potentialat thesesites. One of the main goalsis to
of quicklyassessing
estimatethe wind resourceat the lowestcostin the shortestperiodpossible.
Underthe WESNetgroup,INRSis pursuingresearchinto the use of SyntheticApertureRadar
in marineand coastal
(SAR)satelliteimagesof wave patternsfor wind resourceassessments
projectsand data validation,a currentproject
areas. Followingpreviousresourceassessment
appliedto the Magdalenelslandsregionis givingpromisingresultsfor determiningwind speed
of the wind direction,in particularspecificpatternsat
distributionand some understanding
givenlocationsfor differentwind directions.With regardsto the criteriapreviouslymentioned,
isthe SARapproacha goodcandidate?
To answerthis question,an agreementhas been made betweenWElCanand INRSto test the
SARtechniquefor the PEIcommunityrink program.lt is an opportunityto test the applicability
method.ln the
of the SARtechniquein a real life projectas an independentwind assessment
first phaseof collaboration,
the SARtechniqueis testedwithin certainlimits,namelyINRSactual
SARimagesdatabase.
In the casewhere INRSand WElCanwould chooseto extendtheir collaboration,
therewould be
a secondphaseof the SARproject where new imagesmight be acquiredand be used to
compareresultswith a wind monitoringprogramfor selectedsites.
1.2 Objectives
Theobjectiveof usingthe SARapproachfor PEIcommunityrink projectis to: test the
applicability
of the SARtechniquein a reallife asan independent
wind assessment
technique.
To achievethis mainobjective,threesub-objectives
havebeendetermined:
1) To test the correlationbetweeninlandand SARdata
2) To create,usingonly remotesensingdata,wind mapsof prevalentwind directions
3) To rankthe varioussitesconsidered
with respectto the wind criteria.
1.3
Proiect team
L4
Deliverables
With regardsto the first phaseof the project,the deliverables
are the presentreport and a
videoconferencepresentationin order to discussthe SAR techniques,its benefits and
limitations,
andthe resultsto the partnersinvolvedin the PEIRinksproject.
1 . 5 R e p o rt'sstru ctu re
The reportstartswith a reviewof what the SARtechniqueis.Then,the methodologyemployed
in this project(chapter3) and the results(chapter4) are presented,followedby a conclusion
(chapter5) andthreeAppendices.
AppendixB presentedthe description
of eachpotentialsite.
2
How is it possible to retrieve wind speed from radar satellite
imagery?
2.I
First, what is Radar?
(EM) signal.In
A radar is an active instrumentemitting and receivingan electromagnetic
general,the EM signalis linearlypolarizedand its wavelengthmay vary from the millimeterto
the 102meter scale.For wind retrievalpurposes,radarsoperatingin the Ku, X or C-bandare
used(between1.65and 7.5 cm).
2.2 How a wind map is obtained from a backscattersignal?
The
relationship between
radar
and wind speedis an indirect
backscatter
one. The radar signalis reflectedat the
wate/s surfaceand is influencedby the
surface'sroughness.
Wind blows over water, modulatesthe
surface.lt createsgravitywaves(-m) and
capillarywaves(-cm) (figure1). Capillary
wavesform almost instantaneously
and
Figure1 Gravityand Capillarywaveson Water surface
vary rapidlywith any changeof wind regime
(windspeedor direction).Sincetheir wavelengths
are similar,the backscattered
signalis mainly
related to capillary waves because of Bragg
resonantscattering.
modelfunctionCMODSrelateswind
Geophysical
(1):
speedand directionto signalbackscatter
oo = fovoos(0,
0, Uro)x PR(0) (1)
Where 0 is the satelliteincidenceangle,$ the
wind directionrelativeto radarlook direction,Uro
Figure2 raw SARimage
the wind speed at 10 m and PR is Kirchhoff
polarization
ratio.
3
An instantaneous
wind mapof the regionunderstudyis then obtainedby invertingthe CMOD-S
function(seefigure3 asexample).By instantaneous,
we meanthat the wind mapobtained
represents
the wind speedof the wholeregionat a specificmomentin time.
F i g u r e 3 S A Rw i n d s p e e d e s t i m a t e f r o m C M O D - 5 f u n c t i o n
2.3 What is the Validity Zone?
The wind map obtainedis valid both offshoreand up to a few hundred metersfrom the
coastlinewhen usingSyntheticApertureRadar(SAR)imagery.The resolutionof sucha map is
400m.lf the data useis from a scatterometer,
then the resolutionis muchcoarser(-50km)and
no data is availablecloseto the coastline.In this specificcasestudy,two satelliteswere used:
SeaWindsScatterometeraboard QuikSCAT(referred to as QuikSCATfrom now on) and
RADARSAT-I
SyntheticapertureRADAR.
They both givethe possibility
to retrievewind speeds
over a largeregionthroughsea surfacestate analysis.Many studieshavedemonstratedthat
wind mapsderivedfrom space-borne
instrumentshavea goodaccuracylevel:a smallbiasand a
standarddeviationof -1.5 m/s (Bourassa,
Legleret al. 2003).Previouswork done within the
INRSresearchgroupevendemonstrated
that SARsatellitescan retrievewind speedsaccurately
in complexcoastalregion(Choisnard,
Bernieret al. 2004;Beaucage,
Glazeret al. 2007).
2.+ What are the benefits of SARwith respect to met masts?
Oneof the mainbenefitsof SARimagerywith respectto METmastsis that it allowswind speed
estimationovera largerregionwhereasMETmastsonly providelocalinformation.Havinglocal
informationis often problematicbecauseit doesn'tgivea globalpictureof the wind distribution
over a region.In the wind power industry,wind flow analysisis usedto extrapolatethe wind
informationfrom the nearestMETmastto the site (or region)understudy.However,significant
errors tend to arise from such an extrapolation,especiallyin complexterrain where wind
patternsare very complex(Bowenand Mortensen1995;Su6rez,Gardineret al. 1999;Ayotte,
Davyet af. 2001;Bechrakis,
Deaneet al. 2004).Havingthe opportunityof gettinga snopshotof
the wind speeddistributionover a regiongivesa better idea of its spatialdistribution.When
generaltrendsof wind patternsusuallycomeout, enabling
workingwith numeroussnapshots,
the determinationof sub-regionswhere wind is more favorable.lt is even possible,by
comparingthe wind mapsto the surrounding
topographicand roughness
maps,to get a better
understanding
of the surroundingobstaclesand of the site'squalitywith respectto the wind
resourceaspect.
2.5 How many snapshotswould be necessaryto get a precise picture of
the
wind
speed frequency analysis and
assess Weibull's
parameters?
A simulationwork from Pryoret al. established
that, usingonly SARimages,a minimumof 75
(!7oo/o,9
imageswere necessary
timesout of 10)the scaleparameterand
to estimateaccurately
the mean wind speedof a site, L75 for the shapeparameterand 500 imagesfor the power
densityestimation(Pryor,Nielsenet al. 2004).Work realizedhere at INRSon the St. Lawrence
Gulf regiondemonstratethat it is possibleto obtaina greateraccuracy,
offshore,by usingSAR
imagesin combinationwith the coarserbut more denselysampledQuikSCAT
data.The study
was done using80 SARimagesand gave promisingresults(Lessard-Fontaine,
Beaucage
et al.
2009).As for extrapolation,
to determinethe precisionup to the coastlineand inland,further
studiesare necessary.
3
Methodology
3.1 Available SARdata
In the presentcase,at best,40 imagesare coveringthe region(Figure4). Sincethis work is
mostlya demonstration
of what can be done usingsatelliteimagery,no imageswere acquired
for the study.The SARimagesemployedfor this work were originallyintendedfor
specifically
project,which explainswhy they are not centeredon
anotherEnvironment
Canada(EC-ISTOP)
the regionof interest,i.e. PEl. In an idealsituation,allthe imageswould be centeredon Prince
Edwardlsland,but sinceit is quickerto buy archivedpicturesthan to acquirenew ones,it is not
uncommonto have picturesthat are imperfectlyoverlappingand not exactlycenteredat the
desiredposition.Toolscanbe developedin orderto take into accounttheseissues.
Figure4 AvailableSARlmagedensityover PE.
With regardswith what hasbeenexplainedin the previoussection,havingat best40 imagesis
not sufficientto completelyestimatethe wind speeddistributionand Weibull'sparameters.
On
the other hand, the spatialinformationin those few imagesis still very interestingfrom a
qualitativepointof view.
3 . 2 C o rre l a ti o no ffsh o re -onshor e
PrinceEdwardislandmet mastand projectedsites
m.Lrnat En !!fmd
Ca|r(h
Prqdrbr
fl,**=
0
i0
20
-i
dfin d-dbs
I
mal_me6l
l ,br_cdl
Nrd 63 Utm Z@c 20N
Soure l^bran
Author: Tho|E
Ber9€ton Audey{esiard
Per Gov
FonlEru
Figure5 Locatlonof MET mast, and projects at PEI
To estimatethe correlationbetween offshoreand onshorewind for PrinceEdwardlsland
region, the instantaneousoffshore wind maps, obtainedfrom the 40 SAR images,were
correlatedwith MET mast data on the island.Two MET mast datasetswere used in this
research:
Environment
Canada10 m METmastsproviding10 minutesaveragewind speedand
directioneveryhour.Datahasbeendownloadedfrom the Environment
Canada
database
at thiswebsitehttp://climate.weatheroffice.ec.sc.calWelcome
f.html.Three
METmastswere used:NorthCape,Eastpointand SaintPeters.Theywere chosen
because
of their goodSARimagecoverage,
data
their locationwith regardto QuikSCAT
andtheir proximityto sea,
WElCanMETmastsproviding10 minutesaveragewind speedand directionevery10
minutes.Most of the WElCanMETmastsare 10 metersin height,but METmastsof
Souris,SouthLake,Hermanville
and Eastpoint are at 30m height;they havebeen
broughtbackto 10mheightusingthe logwind profileformulabelowfor neutral
conditions:
( 21
WhereV2 is the wind speedat heighZ2,Vt is the speedalreadyknownat heightzl and z0 the
roughnesslength, extractedfrom the roughnessmaps. For more informationabout the
roughness
map,seeAppendixA.
To optimizethe correlationbetweenSARand on site data,the MET mast closestto the SAR
imagetime were selected.As mentionedearlier,data obtainedfrom WElCanand Environment
Canadaare respectively
at 10 minutesand one hour intervals.Thus,for the WElCandataset,
data within 5 minutes of the SAR image, were extractedfrom the database.For the
Environment
Canadadatase!datawithin 30 minutesof the SARimagewasused.
0
tuac,dc,lF.
F.c6
USGS@61
F i g u r e6 P o s i t i o n o f s i n g l e p i x e l ( s t a r s y m b o l ) a n d Mean zone(squaresurroundingthe star) for w i n d speed
e x t r a c t io n
To insureno inlandpixel
Foreachmet mast,the closestavailableseameasures
were calculated.
or pixelinfluencedby bottomtopographywas selectedNOAAworld vectorshorelinerwas used
to createa 1 Km bufferzonealongthe coast.Two siteswere outsidethe shorelinebufferzone
according
to the available
data:Lennoxlslandand NorthCape.The NOAAworld vectorshoreline
hasa scaleof 1:250000 and hasbeensmoothedin orderto createthe buffer.Errorof precision
liesin the localscaleand smoothingoperation.Forthosetwo sites,a two kilometerbufferwas
usedto insurethat no inlandpixelswere taken into account.To look at the Gustwind effect
when fast windsare blowing,a 1.2 km*1.2km (seeFigure6), meanwind speedwas alsotaken
aroundeachadjacentsite.We correlatedboth one singlepixeland the 1.2 Km *1.2 Km mean
wind to METmastdatato verifywhichone is mostaccurate.
3 . 3 D i re cti o n a lco mp o si tem aps
As mentioned,the data in our possession
are not sufficientto providepower density,Weibull
SARdatacanstill
turbines.Nevertheless,
statisticsand informationon powercurvesfor specifics
providean initialestimateaboutwhichsitesare mostpromisingwith regardsto the wind quality
criterion.Two methodswere tested to estimatea site qualitywith respectto other sites,a
method called the directionalmean wind speed classification,
and the other being the
directionalrelativewind speedclassification.
For both methods,SARsceneswere dividedinto
four subgroups
dependingon the directionfrom whichthe wind wascoming:NEif the wind was
blowingfrom an anglebetween0 and 90', SEif between90 and 180', SW if between180 and
270" and NW if between2TOand 360".
D i r e c t i o n oI n t e o n c I os s i f i c o t i o n
and the meanwind
For this method,all SARscenesfrom a samedirectionwere superimposed
speedper directionwas computed.We then had an averagewind speedper direction(NE,SE,
SWandNW)for the wholePrinceEdwardlsland.
To get a globalpictureof the wind qualityat each site under evaluation,all directionswere
takentogetherusinga proportionalmeanof the four directions.First,the SARpixelclosestto
sitewasselected.Notethat a 3 km bufferzonewas usedto insurethat no landor highreefpixel
was used in the analysis.
Then we used QuikSCAT
data to determinethe proportionof wind
pixelclosestto sitewasselectedandfrom the 10 years
alongeachmaindirection.TheQuikSCAT
'
.
NOAAworldvectorshorelines
were used,see: http://shoreline.noaa.gov/data/datasheets/wvs.html
10
from NE,SE,SWand
and usedto calculatethe proportionof wind respectively
of dataavailable,
data is only availableon the Northernside of the
NW. lt is importantto note, that QuikSCAT
lsland(beingavailableonly in opensea).The proportionscomputedfor sitessouthof the lsland
must be analyzedwith extra caution.Once,for everysite, the proportionof wind from each
directionwas determined,a proportionalmean of the mean wind speed per directionwas
taken.
Directional relative classificotion
Forthis secondmethodcalled,the treatmentis almostthe sameas for the first methodexcept
of scenesfrom a samedirection.Thisextrastep
an extrastepis addedpriorto the superposition
was addedin orderto work in relativewind speed.We definerelativewind speedas beingthe
speeddifferencebetweenthe speedat a specificlocationandthe meanspeedof the regionat a
specificmoment.To obtainthis relativewind speed,maska 10 km regionaroundthe islandwas
isolatedto be analyzedfirst. The 10 Km zonewas chosenbecausewe are only interestedin
coastalwind patternin the PEIstudycase(seethe secondmap of Figure7). The instantaneous
meanwind speedwas computedfor the coastalregionand that valuewassubtractedfrom the
instantaneous
wind map.The map thus obtainedis of the localvariationsof coastalwinds(see
third mapof Figure7).Thisgiveinformationaboutthe localwind comparedto the regionalwind
for a specificdate. lf the wind is weakerat a specificlocationthan in the region,then the
relativevaluewill be negativeand if it is stronger,the valuewill be positive.The advantageof
workingwith relativewind speedscomparedwith true wind speedis to minimizethe effectof
havingimagesnot perfectlycenteredat the samelocation.
11
v
&"
)r
Figure7 Steps to work in relative wind speed
Thefirst imageat the top is the Novemberl1th 2005completescene.A zoneof strongwindsis
observedin the St-Lawrence
Gulf,NorthWest of PrinceEdwardlsland.lf the meanwind speed
wastakenfrom this dataset,then it would not be representative
of PEIcoastalzonesincewinds
in the Gulf are muchstrongerthan aroundthe coast.The middleimageshowsthe data within
the bufferzone.For that specificdate,the meanwind speedin the bufferzonewas of 4.0375
m/s.Thethird imageshowsthe relativewind speed.Notethat the patternsobservedare exactly
the samein the secondandthird image,onlythe scalevaries.Thefourth imageis the composite
mapof relativewind speedfor windscomingfrom the North-West.
3.4 Ranking of sites
To estimatewhichsiteshavethe better potentialaccordingto our data,we rankedthe various
sites accordingto the results of both the directionalmean and the directionalrelative
classification.
To rankthe sites,we usedtwo techniques:
1) We rankedthe sitesin order,according
to our values,1 beingthe siteshavingbest
winds.
2l We dividedthe sitesinto five classesof pre-determined
wind by dividingthe rangeof
resultsinto five equalsub-groups.
Oncethe siteswere rankedaccordingto our data,we comparedthose resultsto the Canadian
wind Atlas.TheAtlasis one of the two methodsusedby WElCanto rankthe siteswith respectto
the wind criterion.We usedthe samerankingasfor our data,exceptthat for the five classes,
we
usedWElCanclasses
insteadof dividingthe rangeof meanwind speedinto five equalclasses.
13
4
Results
4.1 SARMet mast correlation for Wei Candata
The correlation has been calculatedfor each MET mast in comparisonwith SAR sea
(Figure8). As Figure8 shows,all MET mastsare within a five kilometerrange
measurements
from SARmeasurements
exceptSpringValleywhichis morethan 7 kilometersaway.
8
1
^e 6
E5
;U 4
=
E3
2
1
0
-4")*"t'"""1."-|".""."1..*d
Figure8 DistancebetweenWElcanmet mastand SARseameasurements
The correlationobservedvariedfor singlepixelversusmeanzone.One possibleexplanationis
the variabilityof wind especiallyin fast wind conditionsand gust. Some MET mastsshow a
better correlationwith singlepixelother with meanzone.Mean zonetendsto attenuategust
effects,but seemsto havea greaterchanceto includepixelbiasedby bottom topographyand
reef.Variationis sometimehighbetweenthe two methods.Forexample,Lennoxlslandhasa R2
Averagevariationis 0.11for all sites.Figure5
differenceof 0.2852betweenthe two techniques.
showsthe bestresultsof both techniquesfor eachsite.Sincenoneof the two methodsappears
to showsystematically
better correlation,we decideto work with the singlepixeltechniquefor
the remainderof the study.
Somesitesdid not showsignificantcorrelation.CapeWolfe (0.1085),Souris(0.1885)and South
Lake(0.1838)havevery low correlation.The lackof collocateddata and the extrapolationof
heightfrom L0 to 30m for SouthLakeand Sourisare the main concerns.The distancesfrom
15
data can also be a sourceof error. The correlationfor Lennoxlsland(0.5557)and
QuikSCAT
(0.3109)also
Stanhope(0.9375)appearsto be concluding.Eastpoint (0.337)and Hermanville
show quite good correlationeven if tower heightswere at 30 m. Anotherexplanationfor low
correlationmay be the topography.For example,CapeTryonhasa steepcliff of 34m highwith
9.7o/o
of slopeand CapeWolfe hasa 20m one with a slopeof 8.4%.Wind changeabruptlyand
cantake longdistancebeforeattaininga new equilibriumafter strikingandobstacleof that size.
16
7 data
11 data
l0
r:l
30
u
t
1{}
I. ro
J
7
0
0,F
Sr
tEB
i:l.F
lo,s
l!
ut
10
,
!r
LJ
FO
TE
trr
o,m
14 data
*
u
zg,fr
t:t
TI,IF
lll,qt
51!
t.{xt
J
s
.l--r:-
osl
*G
,*o
16 data
NF=D.
il = q3??t
Tdata
10,{xl,
e0,m
u.tr
*
f 1'5,so
ulJxt
l,0.ql
s-i--'
t.In
t,s
0,txt
0,m r
orto
tlr
5 data
ng
# =t.l!ir7)
ID
r,o
J
ol
ro
F i g u r e9 C o r r e l a t i o n l a n d v e r s u s s e a f o r W e i C a n d a t a ( d a t a a r e i n m / s )
L7
4 . 2 S A Rme t ma st co rre l a tionfor Envir onm entCanada
After lookingat the previousresults,it is importantto test if better correlationcould be
observedwith a greaternumberof collocateddata.SomeWElCanMETmastsdon't havelong
time seriesand are not alwayscollocated
with SARimages.
The comparisondone with EnvironmentCanada
D
Ernolrt
E
E
f=051
ltt
containedmore comparisonpoints (figure 9).
I
North Cape has 26 co-locateddata, East Point
T
t
5
-
data were performedon longertime seriesand
and Saint-Peters
21 each.Thosesitesare all very
o
closeto sea, North cape is 0.815 Km from the
5tDts
$InIII
$
sfthtl|
sea,SaintPeters2.012Km and EastPoint0.115
]=oJs
Km. Roughness
is 0.75 for North Cape,0.01 for
Saint-Petersand 0.005 for East Point. High
I'
roughnesslengthdue to Forestterrain of North
I'
t
Capecan be a sourceof variationsin land/sea
E6
5tots'o
data.Topographyis smoothfor everysite,North
$lrEl
Cape lies at an altitude of 9m, Saint-Peters
at
n
E-
27m and EastPoint7m. Correlationwas greater
I,
than 0.58 for every site. Those resultsare in
I
data conductto better correlation.lt alsoshows
T,
agreementswith our hypothesis,largersets of
o
that extrapolationof speed at different height
o5101!tn
$lE-i
Figure 10 Correlationof EnvironmentCanadamet
mastversusSARmeasurements
at sea
may induce another bias, East Point is an
example.ThecoastalzonesurroundingEastPoint
seemsto be subjectedto shoal.Obviously,
wind is not alwaysorthogonalto coastand is subject
to frequentdirectionalvariations.
4.3 Directional compositemaps
Thereis a much strongerpresenceof linearelementsand abruptvariationsin the composite
meanwind speedmap (Figure11) comparedto the relativewind speedcompositemap (see
graphic4 of Figure7) eventhough both are of NW direction.Theseeffects,mainlydue to the
sceneslimitsvaryingfrom one date to another,are smoothedin the relativecompositemap
18
becausethe data are in relativewind speedand not in true wind speed.Eventhough,similar
globalpatternscanbe observedin both figures,they strikeout muchmoreclearlyin the relative
wind speedmap.
Figure11 Compositemeanwind map for windsfrom NW
Sincerelativewind speedmapsseemsto givea clearerpictureof PEIwind patterns,let us focus
winds on the southern
on those (Figure12).For both windsfrom North-Westand North-East,
sideof the lslandare muchweaker.The North-Westcompositemap alsoshowsa shadowfrom
sideof the lsland.The mapfrom
Tignishto RichmondBayand strongerwind on the North-East
the South-Westshowsslightlyslowerwind aroundNorth Capeand on the lsland'sNorth-East
side.A higherwind speedzone is alsovisiblearoundthe RichmondBay,as if the winds were
map is lessinformativeas it containsonly one SAR
canalized
throughthat Bay.The South-East
scene;it is thus not reallya compositemap. Furthermore,the map doesn'tcoverthe entire
lsland.Thishasto be takeninto accountin the globaldataanalysis.
By this visualanalysis,we can make the hypothesisthat winds on the north shore,east of
RichmondBay,shouldbe beston PEl.On the other hand,windson the lsland'ssouth-eastern
sideshouldbe weaker.
19
NdttFwcst
I
i
:
a'
lr\
2
1
{
it
,\
16\
t6\
i:
1
I
6J' W
.l
a5 \t
6: \\'
6: \\'
t
t'
I
{'\
tr}i
.1
i
f"
1;
''l
i\
t(\
6:
\lf
Figure12 Relativewind speed composite maps
+.4 Ranking of sites
First,it is importantto note that somesitesare very far from SARmeasurements
as situated
inland.Theanalysiscan be performedforthose sitesbut accuracydecreases
as distanceinland
(seeFigure13).O'leary,Montagueand Kensington
increases
sitesare more than 13 Km away
from measurements,
addinguncertainties
to the resultsof thosesites.Borden,Alberton,North
RusticoandTignishare situatedcloseto the coast:resultsshouldbe more representative
of the
realresources,
especially
the lastthree locatednorthwhereQuikSCAT
dataare alsoavailable.
20
Figure13 Distanceof projectedsitesfrom SARdata
Our resultsshowthat sitessituatedon the North coastof the islandhavethe better potential.
North Rusticohavethe better potentialfollowedby Morell,Inverness,
Tignishand Albertonfor
relativewind speed classification.
The south east cities have the lower potential, Pownal,
Belfastand MurrrayRiver.Inlandcitieslike Montague,Kensington
and O'Learyshow medium
results,but their distances
from the coastmakethe resultslessaccurate.
Comparisonbetweenthe mean wind speed,the relativewind speedand the Canadianwind
energyAtlas show mitigatedresults.Data situatedsouth of the islandand far inland were
particularly
(class4) appear
under-estimated.
Pownal(class4), Cornwall(class3)andKensington
to havebeenlargelyunderestimated,
classof 1 or 2 were attributeinsteadwith the relativeand
meanwind speed.Distancefrom shore,distancefrom the diffusiometerand lackof southern
data (7) are main concernsfor those results.Alberton,Inverness
and TyneValleyalsoseemto
be far from Canadian
wind energyatlasresults.Thosesitesare situatedin a closeto eachother;
there might be localphenomenonthat modifieswind patterns.The sitesin North Rusticoand
Georgetownhaveperfectmatchesbetweenthe three techniques.For O'leary,Borden,Belfast,
Souris,Morell and Tignish,the Canadianwind atlasdata matcheswith one of the techniques.
The relativewind speedtechniquesmatch 5 times and the mean zone4 times. Relativewind
speedresultsseemto be closerthan meanmethod.
27
i t ' , : t i L r ; t ' l . i r t t r' i i l r i l i t , t l ; r l i t i
Perfectsitesfor wind qualityare Bordenand North Rustico.Kensington,
Alberton,Morell,
Tignishand Pownalfollowand are in the 6.5 to 7 m/s rangeof wind speed.lt is importantto
noticethat Radardataare extractedfor a 400mpixel,and Canadian
wind energywork with 4
Kmgrid.Results
can be verydifferentbecause
one techniqueis at a meso-scales
and the other
at microscales.
Also,the lackof directional
information
in the Northumberland
Straitpotentially
biasedthe resultssouthof the island.
C||r! otwind (l-5)
J z,s
!
I
Meanwindspeed
Rehtivswindsp€ed
Canadian wiod
INRS
o
125
25
AL,thor Thomas Bergeron Audrey-Lesserd Fonlarnc
Nad 63 Ulm Zone 20N
Sourcc Canadran space agency Canadran wind energy allas
t i g L t r e1 4 ( o m p a r i s o no f p r o J e c t e sd i t e sc l a s s ri ft a t i c ' nn i e t h c i d s
5
Discussion,conclusionand recommendations
It is possibleto extrapolatedata from seato land,but with certainlimitations.Thereis a 25 to
30 km transitionzoneat the interfacebetweenseaand landwherethe wind is subjectto both
seaand landinfluence(Beaucage,
Glazeret al. 2007).Thewind on landis thus correlatedto the
wind at sea in that transitionregion.Most of PrinceEdwardlslandlies within the transition
zone,thusthe onshorewind shouldbe correlatedto the offshorewind at somedegree.But this
correlationis far from beingperfectsincethe wind is highlyinfluencedby upstreamobstacles,
from landto
suchas landroughness,
topography,
etc. (Ricard,Bernieret al. 2005).Extrapolation
CanadaMET
sea is thus a feasiblething,and is far from beingtrivial.Resultsfrom Environment
mastsshow that when the sufficientdata is available,correlationsare good.Sitessituatedon
with WElCandata.
the NorthCoastof the lslandshowthe bestcorrelations
The siteschosenby WElCanfor the net meteringinitiativeare Alberton,CrapaudKensington,
MurrayRiverandTignish.Fromthosesites,onlyTignishand Albertonhavegoodwind potential,
while Murray Riverreceiveda note of zero for wind quality.For a total of 130 points,wind
qualityonly hasa weightof 20 in the total balances.Thatmakesthe presentstudymore limited
potentialsitesfor the net meteringinitiative,but for future studywere wind hasa
in assessing
strongerweights,it couldbe more helpful.The GIStechniquepresentedin the appendixcan be
usedto assesswind turbine sittingand obstacles,visibilityfrom the communityrink and the
areawheresoundlevelsexceedthe criteria.
As
Thisreportis a first assessment
of what couldbe done usingSARwind resourceassessment.
it has been shown throughoutthis report, it seemsclear that SARtechnologyhas a great
potentialin helpingdecisionmakersin their wind resourceanalysis,but further developments
Wind
are neededto betterunderstandthe discrepancies
betweenSARresultsandthe Canadian
Atlas. Furthermore,more SAR imageswould be neededto be able to make a thorough
evaluationof the region'spotential.
23
For a future assessment
of wind at PrinceEdwardlslandthe followingconsiderations
can be
takeninto account:
Havinga minimumof 40 SARimagesdnd a constantdensityof imagesthroughoutthe
island: Thatmeans,acquiringimagesfor the purposeof mappingwind at PrinceEdward
lslandandsubsequently
usinga SARapproachto derivewind direction,localgradient,
fast Fouriertransformor wavelet.In that case, directionalinformationwould be
available
all aroundthe islandat a Kilometers
scale;
Usinggeostatiscal
approachandvariogramto knowthe distances
rangeof validityof
seadata;
Takingroughness
andtopographyinto account.
24
6
Aknowledgements
The authorsthank the PrinceEdwardlslandEnergyCorporationfor providingthe MET mast
data, as well as the Wind EnergyInstituteof Canadafor their interestin the projectand for
providingaccommodationin North Cape.They also wish to acknowledgethe help of Paul
Dockrill,for meaningfuldiscussions
and for sharinghis time very generously.
Finally,thanksto
WESNetfor fundings.
25
Bibliography
Ayotte,K.W., R.J. Davy,et al. (2001)."A simpletemporaland spatialanalysis
of flow in complex
terrainin the contextof wind energymodelling."Boundarv-Laver
Meteorologv98(21:275-295.
Beaucage,
P.,A. Glazer,et al. (2OO7l.
"Windassessment
in a coastalenvironmentusingsynthetic
apertureradarsatelliteimageryand a numericalweatherpredictionmodel."Canadian
Journal
of RemoteSensins33(5):358-377.
Beaucage,P., M. Bernier,et al. (2008)."RegionalMappingof the OffshoreWind Resource:
Towardsa Significant
ContributionFromSpace-Borne
SyntheticApertureRadars."leeeJournal
of Selected
Topicsin AppliedEarthObservations
and RemoteSensine1(1):a8-56.
Bechrakis,
D. A., J. P. Deane,et al. (2004)."Wind resourceassessment
of an area usingshort
term data correlatedto a longterm data set."lglgfE!.gIgy 76(61:725-732.
Bourassa,
M. A., D. M. Legler,et al. (2003)."SeaWinds
validationwith researchvessels."
Journal
ofGeophvsical
Research
C:Oceans108(2):1-1.
Bowen,A. J. and N. G. Mortensen(1996)."Exploringthe limitsof the wind atlasanalysisand
applicationprogram."Proceedinss
European
Wind EnergvConference:
584-587.
Choisnard,
J., M. Bernier,et al. (2004)."Mappingof the windsin the Gulf of St. Lawrenceusing
RADARSAT-1
images."Canadian
Journalof RemoteSensins30(a):504-616.
Lessard-Fontaine,
A., P. Beaucage,
et al. (2009).OffshoreWind Ressource
Assessment
Usins
(SAR).
SvntheticApertureRadar
AWEA,Chicago.
Manwelf,J.F.,J.GMcGowan,.and A.L.Rogers(2OO2l,Wind energyexploined-Theory
designond
opplication,JohnWiley & SonsLtd,Amherst.
Pryor,S.C.,M. Nielsen,et al. (2004)."Cansatellitesamplingof offshorewind speedsrealistically
represent wind speed distributions?Part ll: Quantifyinguncertaintiesassociatedwith
distributionfittingmethods'.'@43(5):739-750.
Ricard,8., M. Bernier,et al. (2006)."Statistical
relationsbetweenthe measurements
of wind in
situ and the estimations
of windsin a coastalregionobtainedby the RSOimageryof RADARSAT1."
32(2):65-73.
Su6rez,J. C., B. A. Gardiner,et al. (1999)."A comparisonof three methodsfor predictingwind
speedsin complexforestedterrain."MeteorologicalApplications
6(4l:329342.
27
B
Appendix
8.1 How to obtain better results when extrapolating sea measures on
land?
A few techniquesexist that can improve correlationbetween land and sea measures.
Nevertheless,
they were not appliedin the presentcase,becauseof limitedtime and resources
were
availablefor the study.The followingtechniquescould be appliedif time and resources
sufficient.
Seporotewindsby main directions: (Ricard,Bernieret al. 2005)had separatedwind in
threeclass:wind parallelto coast,seawind andwind comingfrom land.
Moke o topogrophicclossification
for eochsite: (Ricard,Bernieret al. 2006)suggestthe
and slope.Thisclassification
followingclassification:
bank,inclination,
escarpment
can
be possibleby makingtopographictransecton a digitalelevationmodel.To seehow it
works,fook up the appendixhow to retrieveelevotionfrom o digital elevotionmodels
(DEM).Theformulais simplefor correctionsS=
ZtfACfif, is site height,zd the distance
from shoreand s the slopecorrection.Transectmustbe in the wind direction.In the
caseof windscomingfrom landthe transectstartsat the site'slocation.
Correctmeosurementswith roughnesslengthcoefficient:dependingon wind directions.
Thebestwayto do so is by makinga mapandtransect,the sameway asfor the DEM.
Hereare someformulascommonlyused:
'5/Ah)
R o u g l r r t cl esrsr g t lZt := 0 , 5 ' ( H
C o rcr c t i o rfra c t o rc o r= l n { Z o l Z o
il2
rr(h/zOiI
(ZolZo
lrr
1)lrt(h/202)
|| A(sitczo1)' co r
5r.t
Wcill uII A(sr
tc zo2)=yy..I
29
Table1: Roughness
length
L
0.9- 0.6
0.5
0.3
0.2
0.1
0.0s
0.03
0.01
5*10^-3
10^-3
3*10^-4
LO^-4
city
forest
suburb
shelterbelts
treesand shrubs
closeagriculture
fields
openagriculture
fields
veryopenagriculturelands
herbs
exposedland
snow surfaces
sandsurfaces
water
Roughncsslength
Thesriclength teWffientingthe
hedgfrt(rt wieh winrl his null
stl.dst$tt ts inEreqseltl $
logafitlunic way.. (Trsenet
petersr'tl,
lg8gl
In the roughnesslengthformula,H is the height,S the cross-section
in front of wind, Ah the
mean of the horizontalsurfaceof eachhomogeneous
element.Note that tablesof roughness
lengthalreadyexistfor manyspecificlanduses.Forcorrectionfactor,zois turbineheightor met
mast,ze1the sea roughnesslength,zo2the land sitesroughnesslength.The h in the formula,
which is the boundarylimit layer,is calculatedthis way. Within boundarylimits layer,wind
speeddependson roughness
lengthof terrainandobstacleheight.
7.t
WhereX is the distancefrom the discontinuity
lengthbetween
and ze'the maximumroughness
the two sites.lt is importantto noticethat if the Weibulldistributionis not known,correction
factorcanbe appliedon raw wind speed.
30
(Ricard,Bernieret al. 2006),who worked in the Gasp6sie
region,Provinceof Quebecdivided
windsin 12 class,3 for wind regimeand 4 for topographic
situationsof sites.Theyrejectedsites
locatedmore than 3 km inland,becauseof terrestrialwind patterns.They also rejectedsites
closeto river,becauseof possibleeffectof canalization
of windsin valley.The resultsobtained
for the entiredataareshownin Figure8.
Conrparaisonclircctc(R2= 0,.5227)
20
20
l8
t6
Itt
(') t 2
r0
a
8
()
t)
t)
6
4
2
0
l6
?
1t
Ir+
a
Suilc i I'c'xtrapolation(R2= t1,t1266)
a
te
v
1)L
,a
'^itr
l2
l0
t ^l e-Foc
or'|'
.,tiT.
.rgr
^&rl
8
6
lfF.}{.
Itrir?^.. - -
.trfr
4
g
2
0
0246810121416I[t20
0
Vitcsscdc vcntcstimicparRSOd l0 m 1m.s-r
)
2
4
6
8
r012t4
t61820
Vitcsscdc vcnt cxtrapolccsur tcrrc (m.s'')
F i g u r e 1 5 C o m p a r i s o n o f ( R i c a r d ,B e r n i e r e t a l . 2 0 0 6 ) d a t a b e f o r e c o r r e c t i o n s ( l e f t ) , a n d a f t e r c o r r e c t i o n s ( r i g h t )
Thoseresultsshow well the potentialof extrapolating
SARoffshoreestimatedwind speedto
inshoresites.In the caseof PrinceEdwardlsland,a majorityof sites lies in coastalzone and
corresponds
to good candidatefor SARextrapolation.That situationmakesthe presentstudy
feasibleand resultsshow it well. For better results,more constraintshaveto be taken into
account,standarddataat L0m and a wider rangeof scenecollections
collocatedwith met mast
data.
31
9
Appendix A
9 . I A .1 H o w to ma ke R o ughnessm ap with GIS:
Geographic
informationsystemsis softwaremadeto stock,manageand manipulategeographic
information.The three functions of GIS are: Spatialanalysis,numericalcartographyand
databasemanagement.CommonGISare Arc GlS, Map info and ldrisi.Some GISare open
sources:
QuantumGlS,Grassand others.Thesoftwareusedin the presentcaseis Arc GlS.
9.2 A.2 Stepsto make a roughnessmap:
www.geobase.caand www.geogratis.ca.Six images were downloaded on the
http://scf.rncan.gc.ca.The images are part of the project Earth Observotionfor
SustoinobleDevelopmentof Forest(EOSD).Landsatimagescan also be acquiredand
classified
with remotesensingsoftwarelike PCIGeomatica,Enviand Erdas.Download
NationalRoadnetworksof PEIon Geobase,and the MARITIMESTRATEGIC
IAND USE
DATA BASE(SLUD)on Geogratis.Noticethat for more precisionyou can download
topographicdata insteadof Slud,BNDTor new Canvec1:50 000 on Geogratis.Layers
BatimP and BatimS give you the localization
of each building,if you want to work in
urbanzoneor at finer scales.
systems.The projectionchosehis NAD83 UTM zone 20N. This projectioncoversthe
entirePrinceEdwardlsland
analyze.
33
IrMd
!
I
I
I
I
!
fl
I
o
CoYGr Clatr
no 0.."
I
un.t.rrirr.o
tr
crouo
reongrrc{rture
ra
st"oo"
r2r tr.-croprrnd
!
ae ,"t..
!
se
rcn-to.t"t.o
or
sno", I."
rz
ro.r r rroot"
r"no
|
!
I
I
!.:,
ro a.vo,os
I
so st"uor"no
s r s r " u or " r t
I
8". -::.""""
!
S
I
rrc orrsstano
/ rorie.
raa
"s.-."'.,r.
2"" ro".t, t ,,rrt
are conir..ou'
ari conir..ous-d.n..
!""n Lihd
.. ,:.;",..
I
roe r""o'
ro
r
I
Elqad
or
rctt.no-t.rro
ae
r"tt.no-rn.ro
rr
mtr.no-n..1
I
"":,,.:":,-,,":,.
e a or " o " o r . " r
e2! !rord]..f-drns.
a a ar . o " c t . " r - o l . n
f;...::,.":""""'""
[
E
!
23r Fixeduood-oais.
232 nrKoFod-oD.n
a". n,*"0*oo-ro..r.
t i g u r e 1 5 C l a s so l t O 5 D i m a g e s
Reclassification
hasbeendonethisway :
0 - 1 0 -ll- 1 2
0- No data
20
l- water
30-33
2- exposed
3l
3- Snow-ice
32
4- Rock
34
5-Urban (different sources)
40-50-51-52
6- Shrub
80-81-82-83
7-Wetland
1 0 0 -ll0
8-grassland
t20-t2l
9-Agriculture
200-210-2rt-212-213
lO-Coniferous
220-221-222-223
I 1-Broadleaf
230-231-232-233
l2-Mixedwood
l3- Roads
F i g u r e1 6 R e c l a s s i cf a t i o no { t h e E O S Di m a g e s
34
Transformyour vector data to raster.Take only the urban zone of The S/ud
databaseand exportit to raster,For road networks,makea bufferof 50 m to
transformyour datafrom lineto polygon,similarto a realisticroadwidth. Most
GIShavea Bufferfunctions.Thenconvertit to raster.lt's very importantto put
map.
the pixelssizeat 17m,the samethanthe EOSD
Reclassify
the new rasterlayer,givinga zeroto the informationand one for no
datavalue.Thisstepwill be important,when combiningthe layers.
Combineyour layers.Eachlayerhasto be multipliedwith the EOSDlayer.Value
of zerowill representthe top layervalue.Youhaveto replacethe zerowith the
class,5 for the urban zone,13 for the road networks,with the
corresponding
function.
reclassify
) Younow havea nicelandusemaps.lt's importantto saveit and conserveit.
F Reclassify
for the lasttime, givingthe appropriateroughnesslengthcoefficient
to eachlanduseclass.
F i g u r e1 7 R o u g h n e s sl e n g t h sf o r s p e c i f i cl a n d u s e
35
) Enjoyyour mapsand giveit nicecolors.
Roughnesslengthand landuse for Lennoxisland
INRS
Rooehmr.
bngh
t.,",
f- ow
@u*
r"'
Ioorwr
!orm*rr
!onwr
IOA
oawt*v
f
I'
Lrnd u$
I*.
@c",*,
I",.
Ik.
ls.c.
l**r
!o,-u*
les'x"I6e.r
I
"o*r
lu'oa
[llr*,
0
2 250 4 500
9 000
13 500
16 m0
M€lers
Aulhor Thoros Eerg€ron,Audrey+essard Fonlerft
Soorce G@base Geogaatrs
) Youcando transecteasilyif your softwarehavea 3D analysttoolbox
36
9.3 A.3 How to use elevationdata with digital elevationmodels(DEM)?
DEMare commonlydistributedby data providersas open source.Thereis wide availabilityof
DEMon the internetwith differentresolution.DEMfor all of Canadacan be downloadedon
Geobase.The
UnitedStatesGeological
Survey(USGS)
are good providersof DEM.The DEMuse
in the presentcaseis the ShuttleRadartopographymission(Srtm),which is a NASAplatform
and can be downloadedon www.USGS.sov.
The resolutionof DEM is 90m. SomeDEM have
betterresolutionbut are harderto acquire.TheNewAsterplatformhasa resolutionof 25m,but
data are still hard to acquireand is primarilyrelevantfor sites in the United States.After
acquiringthe appropriateDEM,it canimmediatelybe usedwith GISsoftware.
degrees-if this is attempted,the resultsof the calculations
will be erroneous.UTM or
MTM projectionis appropriatein that case.
are mostlymadefor scaleof 1:50000 and higher.
DEMhasmanyapplications,
in Aeolianindustry,it canhavethe followingapplications:
Seeingthe relief in crosssectionand viewingthe shelteringeffectsof specificsites.
Transects
madein the presentstudyaregoodexamples.
Calculating
the slopeof an environmentand applyingtopographiccorrectionfactorson
winds.
whichit canbe viewedcanbe determined..
37
10
AppendixB
1 0 . 1S i t e sd e s c r i p t i o n:s
CapeTryon
-+
Fk.dbn olr.Dos.phi(lrN.d
0
30
28
2E
24
t-
rg
fre
'14
o
€
0
0
50
100
150
209
250
300
Dislancefrom shore
fopograph;c
lran!'ecf
39
!5
!10
:Xke4
Cape_wolfe
r+
rir.{rion ot topoFht
ir.ENf,
C
&
12!
Lr--rf-r
:rl
f,rJ
30
2S
_?0
.9 tE
-T
10
5
0
100
150
Distance
fromshore
Topographielrensecl
40
200
l!:.4
EastPoint
->
Dlrillon
ol roPoCqhk rr**d
0
t!
Eastpoint
?n
t6
fr
21
22
18
E,ia
'Dd ' -
T"
ra
ll
1C
o
t
n
0
100
{En
Distance
fromshore
Topographic
fr8n$-ecf
4I
lrA
@ wFr
Hermanville
->
![€ilron
ohoporrThli
ilin+d
Hermanville
1n
25
tn
E
6l rc
oJ-
I
IU
n
0
100
200
r00
30c
D i s t a n c fer o ms h o r e
Topograph;c
treos'ecl
42
Lennoxisland
-}i
D[.dhn
ot roPomnk
!rrEed
0
6!
rS
Lennox island
;30
l6
2e
;24
u
i20
I
18
E-16
i '6
14
'E
'1?
,
:E
10
6
2
:0
0
50(
1000
1500
2 000
Distancefrom shore
Topogrephic
fransecf
43
NMsr
+>
!l'edroh 0t roP.rhri
firn+d
C
2n
tl
'24
I
20
18
, $re
o.l{
ri-
10
5
4
?
0
100 120 140
E0
Dislancefromshore
Iopograpiic
tren-cect
44
a3
llc
S0 len{!
South Lake
g
l0
i&
tC ^nr.r.
South lake
30
2c-
20
E
El ac
10
n
1 500
0
Iopographictrersecf
45
SpringValley
oi'.dbn
ot roP.Dhk
t'ffi.d
+
0
S
!W
2@kFrt
Springvalley
50
45
40
35
E30
EI
'd
25
I
2A
15
10
;
0
1000
2 000
3 000
Distanceftom shore
ProfileGraphSubtitl€
46
4 000
5000
Stanhope
r+
Dlr.db. ot toPo34hk rr*.d
0
275
S
rr6\h
Stanhope
?n
IJ
20
E
EI {E
o)
-
10
5
0
0
1 000
1 500
Distance
fiomshore
Topographic
harsec!
47
2 000
10.2 Unvisited sites
Krnrlng[on
60
g0
a0
20
t0
0
o
g 000
1000
0
T4,er'Fhtcm'3,cl
fw|f#tr.'|'/'!xct
2 000
3 000
llurny Rlvrr
Nonh c.p.
fwt+lticffi#
48
| 000
© Copyright 2026 Paperzz