n1

sG I S&

REMOTE SENSING NEWSFIJI USER GROUP, Number 11 (9502), June 1995

N E W S

This is the second newsletter for 1995 and apologiesare in order for the small number of issues thisyear. The main excuse is always the lack ofcontributions and that it is put together in spare

time and of course who has spare time! We have a faircross section of articles this issue touching both GIS andRemote Sensing. MSD has focused on remote sensing forFiji’s forests drawing on more than a years work and MSDalso explains data handling which covers geometric correc-tion as well as file format translation. The two user meetingssummarised here also addressed this type of data handlingand there has been an ongoing discussion image projectionand correction. While the MRD articles have also addressedDTM and data translation.

USP has provided an article which covers the humanresource development in the GIS sector and while this is thefirst time we have received reports on this subject it isexpected that there will be followup.

There has been a further contribution from Europe whichshows first experience with the new MOMS satellite data.

Again we welcome contributions from national, regionaland international users. Yes, you can reach us by Internet –see back page for the addresses!c

There were GIS & Remote Sens-ing User Meetings in March andApril where the March meetingwas held at MSD-Forestry inColo-i-Suva on 7 March 1995while the April meeting was heldat Mineral Resources Depart-ment (MRD), 11 April 1995.

During the March meeting:

Osea Tuinivanua (MSD-For-estry) gave a briefing on FAO/RAPA1 �Expert Consultation onForest Resources Monitoring Sys-tems� in the Asia/Pacific Re-gions. The meeting was held on27/2/1995 - 3/3/1995, in Bang-kok, Thailand. The group of ex-perts came from China, India,Bangladesh, Burma, Nepal, Bhu-tan, Philippines, Malaysia, Laos,Cambodia, Sri Lanka, Thailand,Vietnam and Fiji. It was part ofthe continuous global effort ofFAO/Forestry Sector to stand-ardise the forest type classifica-tion, improve accessibility andreliability of information, in-crease the period of forest re-sources assessment and enhancecountry capacity for plannersand decision makers in forestryand related development sectors,which includes scientific commu-nity, Global Change communityof Universities, national as wellas international research teamsand public at large.

Unlike FAO´s two global devel-oping countries of tropical forest(Africa and Latin America), theAsia/Pacific Region has the leasttropical forest cover areas. Fijiforest resources inventory hasimproved remarkably in recentyears using satellite images andadopting some recent technolo-gies to enhance its monitoring ca-pacity. Fiji (PICs2) commented thevariability of standards used forforest classification and proposeda mobile receiving station forregular satellite data receipt inthe Pacific.

The consultation endorsed: thenational level classification to bemaintained, non-forest land for-est products to be monitored,

sub-regional and regional levelsof classification of forest types toconsider the ecofloristic zones,regional scale maps to be stand-ardised at scale 1:250 000, stand-ard terminology and methodol-ogy to be adopted, exchange ofexpertise to be promoted, na-tional planning and funding tobe supported, and FAO to pro-mote national biodiversity inven-tory and monitoring.

Don Forbes (SOPAC) explainedthe features of GRASS (Geo-graphic Resource Analysis Sup-port System) which is a publicdomain GIS and image analysispackage produced and availablefrom the US Army. It is a UNIX-based commandprocessor thatincludes capabilities for digitalrectification (rubber sheeting orfull ortho-rectification) of rasterimages, extraction and overlay of

vector and point data, surfacegeneration and shaded reliefvisualisation, among others. Ap-plications of GRASS at the Geo-logical Survey of Canada haveincluded analysis of historicalshoreline changes from digitised,rectified and geographically ref-erenced aerial photographs andshaded relief seafloor imagesfrom swath and sweep bathy-metric surveys. GRASS has beeninstalled at SOPAC, where it willbe used for coastal process stud-ies, with possible future applica-tion to SOPACMAPS and otherswath bathymetry in the SouthPacific.

Robert Smith (SOPAC) illus-trated the advantages of griddingand contouring using QuickSurfunder AutoCAD. This softwarewas compared to Surfer for Win-dows which was demonstratedat a GIS & RS meeting last year.QuickSurf, which is a LISP rou-tine and runs inside bothAutoCAD for DOS and Win-dows, allows the user to contourASCII data sets right insideworking base maps, readily de-fine boundaries, interpolate inareas between surveys and de-velop digital terrain models ofdelineated aggregate resourceareas in the marine environment.At the same time, the user has theuse of AutoCAD�s editing anddrawing capabilities at his dis-posal. Cut and fill volumes,beach modelling for accretion orerosion are easily integrated intobase maps and can be clearly il-lustrated and easily interpretedin a pictorial sense yet still retaina true to scale orthogonal map-ping projection (eg UTM) whichwould be lost if imported into apresentation program such asCorelDRAW.

Wolf Forstreuter and SamuelaUluikadavu (MSD-Forestry)presented the Forestry EconomicData Base. The results have al-ready been described in an arti-cle by Osea Tuinivanua in the lastGIS & RS Newsletter. Each vil-lage closest to a the current per-manent forest sample plots wasasked questions about forest us-age. In Fiji, the natural forest stillplays an important role for foodand medicine supply. In addition,forest is used extensively as graz-ing ground for cattle. Quantita-tive data is necessary to monitorsuch types of forest use. A dis-cussion in the GIS & RS UserGroup started about the need of

1 FAO = Food and Agriculture Or-ganisation of United NationsRAPA = Regional office for Asiaand Pacific

2 PICs = Pacific Island Countries

2n2

further data and additional investigationsof the interaction between villages and for-ests. The survey also noted a change in wa-ter quality which is related to the amount offorest cover and to the forest density andforest type within a water catchment. Therewere indications that the quality of water isalso influenced by tree species. Suggestionswere made by the GIS & RS User Group tolink the data base to other department suchas PWD or USP which have ongoing inves-tigations in this field.

DiscussionESCAP Workshop on GIS/Remote Sensing. 13-17 February, 1995, Tradewinds Hotel, SuvaFLIS advised that the final record of proceed-ings and recommendations from this work-shop would be received shortly. Questionswere raised as to the value of this workshopand it was concluded that the two reasonsfor it being held were raising levels of aware-ness and identification of needs and crea-tion of a �shopping list�. This �list� is com-prised of possible projects for submission topotential donors. There did not appear tobe any avenues for ESCAP themselves tosatisfy the beneficiaries by funding such�shopping lists� and the question was raisedof the likelihood of donors supporting anysuch projects as ESCAP may expect tobeacting as project managers for which theywould need a portion of these funds.

UpdatesNLTB announced that the first maps havebeen produced showing the mataqaliboundaries at 1:25 000 scale on FMG (WGS72) spheroid. This information will now betranslated to DXF3 file format and deliveredto FLIS. The data conversion is done by asoftware called �shift translator�. FLISchecked already the first maps. NLTB is buy-ing InfoCAT software which imports infor-mation stored in internal Synercom format(NLTB GIS). NLTB has plans to scan photo-graphs taken in the field which is also pos-sible with InfoCAT. This photo documenta-tion of development will be stored in a database and can be included digitally into re-ports.

FLIS reported from the last FLIC Meetingthat FLIC had requested a written report ofthe GIS & RS User Meeting. This is becauseFLIC members are very interested in thedevelopment and discussions at the techni-cal level. FLIS will revise its charging policy,there will be an article in the next GIS & RSNewsletter. The stereo plotters at LandsDepartment have been upgraded and arenow able to provide digital data sets. RobinPickering a short term consultant from NewZealand visited all important institutionsrelated to Remote Sensing and GIS in orderto investigate the needs and the format of

3 DXF = Data Exchange Format is the file struc-ture of AutoCAD software. It is a very commonformat which was agreed as standard for ex-change of spatial data in Fiji. It is different fromSIF which is the System Interchange Format.

the digital topographic data base. As men-tioned by NLTB, the first maps showingmataqali boundaries are translated to DXFfile format and will now be checked by FLIS.As a result of an investigation of and modi-fication to the wide area network, the re-sponse time for data access has improvedto an acceptable level.

At the Department of Environment instal-lation of ARC-INFO (PC) has started. Thehardware consists of two PCs, an A0 sizeplotter and a digitising table of same dimen-sions. The department bought a portablePinnacle Micro 3.5� optical drive to be ableto store information safely and to importdata from MSD-Forestry in a fast and sim-ple way. In-house training will be providedby a consultant from UNEP for two weeks.

MRD provided information on the follow-ing projects:

Namosi DEM (Digital Elevation Model)A DEM concerning the Namosi copper mineprospect area was produced using raw data(profile with elevation with a spacing ofapproximately 20 m between each profile)provided by Placer and using GDM (Geo-logical Data Management) software atMRD. The DEM is available as a grid (20 x20 m) and also as a vector file (contours)which has been incorporated into MRD GISMapgrafix.

Vanua Levu Geological MapThis project is aimed at producing an up-to-date 1:250 000 digital geological map forthe assessment and promotion of mineralresources exploration. This map will bebased on a compilation of existing 1:50 000geological map and others relevant avail-able information including open-file re-ports, geological notes, unpublished work.8 out of 14 maps have been digitised so far.Information digitised are the geologicalunits and boundaries, faults, fossils loca-tions, rock samples. The major part of thisproject is performed by a consultant.

Mines Tenements DatabaseThe development of the database applica-tion for the management of mines tenementshas progressed. A first version should beavailable by end of April.

MSD-Forestry has digitised the water catch-ments from the old LRD4 forest type mapsat 1:50,000 scale. Water catchments (west-ern part of Viti Levu) not covered by LRDmaps were digitised from topographicalmaps from PWD. MSD started to correctLandsat TM and SPOT data geometricallyreferenced to the new Lands Departmentmap series 1:50 000. Two new hand heldGPS instruments arrived for documentation

4 LRD = Land Resource Division

n by Dr Bruce Davis, Director, USP GIS Unit, University of the South Pacific

GIS/RS Human Resource Devel-opment in the South PacificThe topic of Human Resource Devel-opment (HRD) was given emphasis atthe February ESCAP GIS/RS workshopbecause it appears to be one of themajor elements that is too often over-looked when considering the technol-ogy. Several speakers underscored thenecessity of viewing the “people” partas a paramount concern in starting andmaintaining GIS/RS in the region (hereinjust GIS for convenience). Understand-ably, new users are awed by the flashand wonder of hardware and software,and in fact it may be these attributes thatprovide initial sparks in adopting thetechnology (actually GIS is considereda technology and a methodology). Sadly,however, the human part of a GIS infra-structure typically is given the weakestattention, before and even after estab-lishment. We hope to avoid this syn-drome in the Pacific.

This brief report summarizes a few ma-jor points made at the workshop in termsof:• GIS Human Resource Development• Needs of the region• Special considerations• Current situation• Development and strategies

GIS Human Resource Developmenti) GIS without people has no value; GIS

exists for people, not people for GIS:GIS is a tool used for a wide varietyof purposes, but it is only a tool andnot a solution (despite the hyperbolicterminology of computer vendors).Humans are not only part of theequation of use, they are the core,the central intent. GIS will not oper-ate in any sense of the word withouthumans and certainly will have nopurpose without them. Therefore, it

n3

stands to reason that the“people” componentshould receive primaryconsideration.

ii) GIS is an enabling technol-ogy: The major function ofGIS is to “enable” humansto achieve goals that incor-porate spatial and associ-ated non-spatial data. It isa technology that helps usto gather and managedata, to analyse andmodel, and to present in-formation.

iii) Why is HRD typically last,or even absent, in thechain of planning and con-siderations when begin-

ning or maintaining GIS?This has no convincinganswer, but one premisehas been that people aretaken for granted and canbe counted on to achievewhatever is necessary.There are too many con-tradictions for acceptanceof such thinking, however,e.g., certain qualificationsare needed for every pro-fession and people are notexpected to arrive unpre-pared; also, continuingtraining and education areuniversally appreciated.Obviously, a major mes-sage is that indeed, GIS

HRD should be *first*, notlast.

Regional NeedsIn brief, the major needs of theregion for HRD are:

i) Dedicated programme:without a designed strat-egy, specialty training andeducation become ad-hocand a disservice. Theremust be recognition thatthe GIS user communityhas grown and evolved outof the initial explorationstage and is in need ofstructured advancement.

ii) Properly trained profes-

of logged over areas. The loggingplanning continues and the GPSbase station is working withoutproblems. A new EU fundedproject started to map hardwoodplantations using GPS and GIStechnique as well as remote sens-ing data. Wolf Forstreuter willwork under this project and willstay in Fiji for another 13 months.

SOPAC announced that: The CDwriter or CD recorder (CDR) isnow fully utilised creating off-shore bathymetry data inMapInfo format from Europeanfunded surveys in the region andthe recent acquisition of an A3scanner will enable the collectionof aerial photos to be archivedand accessible in a digital library.

Bathymetry maps for the regionwere being produced usingGebco Digital Atlas 5th edition,DMA World Vector Shoreline andofficial EEZ boundaries providedby Forum Fisheries Agency. Themaps will be produced in vari-ous sheet sizes using MapInfo forCook Islands, Federated States ofMicronesia, Fiji, Marshall Is-lands, Niue, Tokelau, Tonga andSolomon Islands at scales rang-ing from 1:2 000 000 to 1:4 000000. The maps will be providedin MapInfo data format whererequested.

Resource maps of lagoonal ag-gregate in a digital format havebeen prepared from recent sur-vey data in Kiribati. This data setcan be incorporated into MapInfoand used for coastal manage-ment needs. In Fiji nine CTD(Current Temperature and

Depth) casts measuring waterquality conditions to a depth of225 metres were obtained off-shore Navua using SOPAC�s�Seabird CTD�. Location of castswas with GPS. Recent hardwareand software upgrades havegiven us depth capabilities to 1.2kilometres. With the newprocessing software this data cannow be stored in a database inan ASCII format with the dataconverted to engineering units.This means that from a GIS pointof view this data can be retrievedand used for modelling purposeswith software other than thatprovided by the instrumentmanufacturer. In the earlier ver-sion of the seabird software athird party user interested inmodelling such data requiredaccess to this software.

The Japanese Government sent amission to the SOPAC Secretariatin Suva, early March to finaliseand launch Phase III of the Ja-pan/SOPAC Deepsea MineralResources Survey Program.Phases I and II of the programwhich spanned ten years from1985 to early 1995, had the MetalMining Agency of Japan�s re-search vessel Hakurei Maru No.2 carrying out deepsea surveysin the exclusive economic zonesof Cook Islands, Kiribati, PapuaNew Guinea, Solomon Islands,Tuvalu, Vanuatu and WesternSamoa. With strong petitions bythe governments of Fiji, MarshallIslands, Federated States ofMicronesia and Tonga who wantsimilar work done in their exclu-sive economic zones and praisefrom the affected countries on the

excellent results of the first twophases of the Program, the Japa-nese Government agreed to en-tering a third five-year phase tobegin in 1995 and go on until1999.

During the April GIS And Re-mote Sensing User Meeting:

PresentationsHervé Dropsy and Andre Vial,MRD gave a briefing on theNamosi DEM (Digital ElevationModel). A DEM (Vector andRaster form) concerning theNamosi copper mine prospectarea and associated geographicalinformation were integrated intoa single GIS map.

Hervé Dropsy, MRD presentedthe Vanua Levu Digital Geologi-cal Map 1:50 000. These havebeen scanned and digitised onthe monitor screen. The endproduct will be a 1:250 000 mapshowing the geological features inaddition to the existing soil1:127,000 map.

Michel Larue, SOPAC presentedsome of the new geographicaltools of MapInfo 3.0. In particu-lar the capacity to use �cookiecutting of a polygon�. This al-lowed users to split the objects ofthe Digital Chart of the Worldand later, with the help of a smallMapBasic utility, to realign all thelayers of this product between 0and 360 degrees. Now Fiji is nolonger split in two parts andunity prevails!

Michel Larue, SOPAC explained

sionals should not dependupon the first generation ofGIS users. A recent surveymentioned in the MarchGIS World notes that morethan two-thirds of GIS us-ers in natural resourcesand environmental indus-tries are self-taught (withover half receiving GIStraining from short coursesand 44.3% from vendortraining; only 25% havegraduate degrees in GIS).These users deserve ac-colades and may be idea“mentors” in specific appli-cations, but they are notnecessarily the best onesto teach the next genera-tion. In the Pacific, we arestill relying upon that firstgeneration, “recycled” pro-fessional group to keepGIS moving, but the sec-ond generation shouldhave more substantial for-mal training. The evolution-

the capabilities of the recentlypurchased flatbed A3 colourscanner where the resolution canbe up to 1200 DPI5. When set tothe highest resolution, it wouldgenerate files as big as 800 MB.A first evaluation its capacitieswas presented together with itsanticipated use, which includedmaking mosaics of aerial photo-graph as well as digitising mapson the screen. The latter is of par-ticular interest for the Pacific Is-land Countries which do notpresently have any digitising ta-ble.

Wolf Forstreuter, MSD-Forestry,explained the activity of geomet-ric correction of satellite data atthis division. Satellite images,unlike aerial photographs, havea nearly orthogonal projection.However, they cannot fit onto theFMG orientated maps of the newLands Department series andtherefore a geometric correctionis necessary to rubber sheet theseimages to this map projection.The necessary GCPs6 are storedin a dBASE file, from which aspecial program selects the rel-evant points for the correspond-ing map sheet. The geometric

5 DPI = Dot Per Inch6 GCP = Ground Control Points

...wrapped up on page 6

4n4

ary states of the technol-ogy and applications de-mand it.

iii) Continuing training, educa-tion, and support: HRD isboth a maintenance and anevolutionary paradigm.Training (specific hard-ware, software, or applica-tions) and education(broader concepts, inte-grated coursework) are notone-off processes; each isan on-going strategy thatensures the latest tech-niques and methodologies,particularly for neophytes.Further, current usersneed to keep up with ad-vances, learn new “tricks,”update software and hard-ware capabilities, enhancecurrent and new applica-tions, and refresh conceptsand operations. Today’sGIS specialist has onefood in the applications,one in computer science,and one in spatial meth-ods—a three-leggeddance is required forproper operations. HRDsupport is necessary.

iv) Distance learning: The re-gion is composed of 15-20% of the Earth’s surface,a small amount of highlyfragmented land, low anddispersed populations,usually a primate city, lim-ited technologic resources,and basically it is out of theworld informational main-stream. A centralised sitesuch as Suva cannot pro-vide adequate regionalservice. A system ofoutreach HRD is essential.Fortunately, there is someinfrastructure from whichto build a distance learningstructure, e.g., USP’s Ex-tension Centre services,developments in the USPGIS Unit, and the estab-lishment of the globalInternet. There is increas-ing demand in many Pa-cific nations for GIS edu-cation and only a substan-tial outreach programmecan answer the call.

Special ConsiderationsEstablishment of a GIS HRD

i) A dedicated GIS HRD pro-gramme has been sub-stantiated by regional us-ers and MUST be institutedin the very near future.Dedicated means thatsuch an approach cannotbe a by-product of someother activity nor can it bea substandard part ofsome other programme,lest it become subservientto other needs. The regioncannot afford this type ofsubordinate development.

ii) It must be a sustained pro-gramme to provide spe-cialised and general pro-fessional development andeducation; it must be con-tinuous, work at both lowand high GIS technical lev-els, and should includeOJT (On-the-Job training).

iii) There should be special-ised training workshops,dealing in such matters assoftware, introduction toGIS for administrators andmanagers as well as thetechnical operators, andprofessional support ofvarious applications.

iv) A professional certificationprogramme is desired, of-fering specialised certifi-cates for selected coursesand a diploma for a sub-stantial programme. This isunder development at USPand awaits human re-source enhancement, e.g.an additional GIS facultyperson.

v) In the future, there shouldbe dedicated educationalcourses, such as a full ar-ray of GIS technology sup-port classes (remote sens-ing, applications, program-ming, etc.), perhaps lead-ing to a specialised bach-elor degree and post-graduate specialties.

vi) Substantial distance train-ing and education are re-quired. As has been dis-cussed, there is much todo in this topic, but there isa good infrastructure inplace at USP to begin theprocess.

The last points to be made aresummary, but essential:

programme in the Pacific re-quires additional considera-tions from the traditional para-digms of the western world.Lack of space prevents ad-equate discussion of the moreimportant, but some should beisted. Each is very importantand must be considered in aolistic approach to implement-ing GIS in the region:• Multiple cultures and various

“world perceptions.”• Multiple languages. English

as a second language(ESL).

• Diverse educational ap-proaches; combinations oflocal and etropolitan sys-tems.

• Diverse man-land relation-ships; different land tenureystems; variety of culturallandscapes.

• Diverse technology tradi-tions; low computer experi-

ence; out of the world tech-nology and informationalmainstream (requires extraeffort and expense to par-ticipate); low technology re-source base.

• Low available economic re-sources; external supportrequired for high-tech pro-grammes.

• Desire for less dependenceon outside expertise; prideand capabilities to achieveself-support and regionallyappropriate expertise.

Current SituationWhile the current situation isexcellent for some parts of

GIS technology, there are con-cerns for others.

i) There is no regional trainingstrategy at present.SOPAC has a small pro-gramme to bring a severalregionals for an extendedtime and USP has signifi-cant plans, resourcespending. The universitynow offers an introductoryGIS course and a few sup-port activities are under de-velopment.

ii) Training is ad-hoc, depend-ing on various one-off pro-grammes. It is almost al-ways externally aided, usu-ally with donor resources.Trainers are visitors, whodespite their expertise andaltruism, are unavailablefor follow-up support.Workshops are closed,limited to aid recipients,

and normally not part of aregional scheme for GISHRD.

iii) The Pacific is maturing rap-idly, however, and there areincreasing demands forHRD that is more substan-tial and more regionallyappropriate.

Development StrategiesAgain, with limited space dis-cussion is restricted to listingitems, but it is important tonote that various ideas havebeen discussed, some plansare active and others await de-velopments. At the risk of rep-etition to some of the above:

GIS is a growing and very dynamicprofession globally as well as in thePacific. The first generation of self-starters will be unable to sustain thedemands; there should be continu-ous production of �new blood� in theprofession!

n5

• GIS HRD has to be accessible by the entire GIS/RS usercommunity, not just the lucky few in Suva.

• HRD should be a continuous process, including periodicenhancement of capabilities; an ad-hoc unsystematic ap-proach is unsatisfactory.

• GIS is a growing and very dynamic profession globally aswell as in the Pacific. The first generation of self-starters willbe unable to sustain the demands; there should be continu-ous production of “new blood” in the profession.

• There are various levels of HRD needed: technician, man-ager, administrator, and users for the many and diverse ap-plications. Each has specific needs and limitations; eachshould be perceived as an integral part of HRD develop-ment.

• An HRD programme must be local, national, and regional;various components have specific demands. A distant learn-ing component is critical.

• GIS is establishing a new profession and a new set of para-digms. It is not merely a computer and special software tomake maps, but should be thought as a very flexible andpowerful systems, a methodology (an approach, a set ofprocedures and techniques), and an efficient way to achieveold and new goals. We MUST support it or the loss is ours.

Only the fundamentals of GIS HRD are outlined above andmany aspects have not been discussed, yet there is muchmore to consider and much to accomplish. Success will takea long time, will be dynamic and evolutionary, will often en-counter steps back as well as forward, but ultimately it willprovide the region with appropriate developments that areneeded to achieve its goals. GIS is only a tool, but a veryimportant one that can help to achieve and sustain regionalintegrity and self-reliance. For that we must dedicate ourefforts.c

Both cameras were based on the Komos satellites. KVR-1000images cover 34 x 57 km on the ground and TK-350 approxi-mately 175 x 257 km. Radiometrically corrected digital data willcost $US 3,500 if the data is also geometrically corrected $US4,000 per scene. (EOSAT Notes 4/1994)

RADARSAT, the Canadian satellite, will be launched end of thisyear. As described in Newsletter 9407 the satellite will carry aSynthetic Aperture Radar (SAR). It will produce images with upto 10m resolution. The frequency/wavelength used by the ac-tive sensor will be 5.3 GHz/C-band 5.6 cm. The polarisation willbe HH. (for explanation of these terms see article by ThomasKremmers, Newsletter 9407) In February, Shawn Burns the Ca-nadian sales representative for the South Pacific and Asia wason a promotion tour in Fiji and visited MSD-Forestry and SOPAC.MSD asked for data to investigate the potential of this remotesensing information for mangrove mapping and forest type strati-fication. Radar data is weather independent because the wavespenetrate clouds. It records the plant water content and thecanopy structure. It could have potential in forest application.

EOSAT is marketing Russian space photography and data fromthe Japanese satellite. The American satellite Landsat 5 is stillworking without any problems, but the data purchase is restrictedto ground antennas. The new satellite Landsat 7 (Landsat 6 waslaunched but did not reach its orbit) will probably be launchednext year. There is no further news to date. Further, EOSAT signedan agreement that makes the company the exclusive distributorof data generated during the next 10 years by a planned 10-sat-ellite constallation of Indian Remote Sensing (IRS) satellites (GISWorld April 1995).

EOSAT captured data from IRS-1B since June 1994. This satel-lite has two on board sensors named LISS-I and LISS-II. LISSstands for Linear Self-scanning System. LISS-I has a spatial reso-lution of 73 m while that of LISS-II is 36.5 m. The spectral bandscovered are blue, green, red and near infrared (further informa-tion see Newsletter 3 from January 1994).

IRS-P2 was launched in India 16 October 1994. The satellite candownload data in Hyderabad and in Norma, USA. The satelliteis equipped with a LISS-II (Linear Self Scanning Sensor) whichhas a spatial resolution of 32 x 36 m. The spectral bands coveredare blue, green, orange red and near infrared. Every 24 days thesatellite can take an image from the same area.

IRS-1C will be launched in mid 1995. The satellite carries threesensors: LISS-III, with 20m resolution, a panchromatic camerawith a sub 10m resolution (GIS User April 1995) and a Wide Field

Sensor (WiFS), with approximately 188 m resolution (EOSATNotes 4/1994). This satellite will have a on boad data recordingfacility (GIS World, April 1995). This will allow to order datafrom Fiji!

JERS-1 was launched on 11 February 1992 in Japan for designedlife of two years (see Newsletter 9407). This satellite has an opti-cal sensor on board and a Synthetic Aperture Radar (SAR) witha 75 km swath width, a resolution of 18 x 18 m. The RemoteSensing Technology Center of Japan (RESTEC) announced thatthere are space borne radar data available XXXX which has beenstored on the onboard tapes. It can be purchased from RESTECor EOSAT as a sales representative of RESTEC. In addition,RESTEC advised of the following correction of the JERS-1 de-scription in Newsletter 9407: the SAR has a frequency of 1275MHz not GHz!c

SATELLITE NEWSSPOT is still working well. MSD-Forestry bought a new scene(436-383) from western part of Vanua Levu recorded at 24 Janu-ary 1995. The scene is stored on CD-ROM like the other scenesavailable at MSD.

MOMS-02 is a push broom scanner which will be placed in theRussian space station MIR. The sensor provides panchromaticdata with 4.5m spatial resolution (details see article by Dr BarbaraKoch in Newsletter 9407 and article in this newsletter). DrWolfgang Steinborn from DARA, the German Space Agencywhich will do the marketing for MOMS data, visited MSD-For-estry. He confirmed that data from Fiji will be recorded after in-stallation of the sensor in the space platform. He will keep closecontact with Osea Tuinivanua during his study in Germany.

KVR-1000 and TK-350 images can be purchased through EOSAT.EOSAT, PADCO an international consulting company from USAand Kiberso a Russian geographic information system companysigned an agreement to process and distribute Russian spaceborne photography world wide. Kiberso situated in Moscow hasaccess to archives of the Russian Military Cartographic Depart-ment, where KVR-1000 images (2m resolution) and TK-350 im-ages (5 - 10m resolution) are stored. This photographic productswill be scanned and radiometrically and geometrically corrected.

6n6

and software are required which are cur-rently not available in Fiji.

UpdatesAt MRD, a GIS awareness week was organ-ised 10-13 April. This function was avail-able to all MRD staff and consisted of a se-ries of videos on GIS and Remote Sensing(Videos from Bruce Davis and also sometaken during the February ESCAP regionalGIS/RS workshop).

The database application (first version)MINTENET for the management of minetenements has been completed.

FLIS is currently initiating moves to haveat least one candidate from each FLISagency, to undergo a introductory paper oncomputing and information technology, inrecognition of the need for computing ex-pertise in existing FLIS agencies.

FLIS announced that the final report on theFebruary 1995 ESCAP workshop has stillnot been received.

FLIS is currently working on a chargingstructure for all FLIS data, and will be con-ducting a survey of all existing GIS/IS agen-cies that are providing GIS data to the pub-lic. The purpose of the survey is to ascer-tain the charges that are being made in eachdepartment, and obtain their responses tothe proposed charging structure.

MSD-Forestry checked some plantationareas with GPS and estimated the inaccu-racy of areas shown with the earlier sketchmaps. A new survey of plantations will startsoon.

Satellite data was corrected to the new topo-graphic maps series of the Lands Depart-ment. For six sheets map corrected imagesare available.

The file transformation from NLTB via FLISto MSD does not work yet. Further investi-gation is necessary.

SOPAC provided ongoing MapInfo train-ing for the South Pacific Island CountryNationals. As a part of the Earth Sciencecertificate, a two days training course hasbeen carried out on beach profiling.

SOPAC also developed new utilities underMapBasic.c

correction uses these points and corrects asub-image of the satellite scene covering thearea of this map. The map-image N26(Rakiraki) was displayed as a TIFF file onthe monitor in MRD´s conference room.

DiscussionThe discussion focused on the limitationsin the use of scanned aerial photographs ina similar manner to digital satellite imagesand the following constraints were identi-fied:

For a geometric correction a three dimen-sional model must be created to convert thecentral projection of an aerial photographinto a orthogonal projection. Otherwise ar-eas on high elevation are larger than thesame area in the valleys.

Scanned aerial photographs have large filesizes. Geometric correction which may besuitable for flat terrain (coastal areas) wouldneed very large disk space and excessiveprocessing time.

The illumination within an aerial photo-graph is not uniform as it decreases fromthe centre to the edges and in addition thereis a bright area (hot spot) where the sun rayshave the same angle from the sun to thesurface and from there to the camera.

To accomplish the above, special scanners

GEOMETRIC CORRECTIONOF SATELLITE DATA FOR THE FIJI MAP GRIDIntroductionDigital satellite images are recordedrelative to the direction of flight of thesatellite and due to the track inclinationof satellites with sun synchronous or-bits every scan line is shifted progres-sively to the right (see Figure 1). In ad-dition, the sensor records every line innadir direction which results in the im-age having near orthogonal projection.However, the image is not fully orien-tated to north and does not fit on thelocal map projection.

It was impossible to get the satellite datageometrically corrected from Germanywhere the satellite data analysis wascarried out. The image data is availablemostly in original form (some are geo-metrically corrected to UTM which is notapplicable in Fiji). The image data hasto be geometrically corrected to:

• visualise forest change by overlayingSPOT data recorded 1994 and 1995

with Landsat data recorded 1991 and1992;

• overlay image data over plantation ar-eas which are in FMG maps in order

Figure 1. Geometry of a Landsat or SPOT image.

to analyse the status of areas plantedwith introduced species, areas cov-ered with natural forest and areaswithout tree cover.

Normally, the geometric correction forsatellite data is done for a completescene. Image analysis software con-tains common projections such as UTM,Transverse Mercator along with com-mon spheroids. Fiji Map Grid, which isonly used in Fiji, is not included in mostsoftware and MSD decided to do thegeometric correction for each map sheetseparately. This reduces the error toan insignificant scale.

The Ground Control Point Data BaseAt MSD, Landsat TM data from 1991and 1992 is stored as well as SPOT data

n7

from 1994 and 1995 coveringthe same area and additionalTM data will arrive. See Fig-ure 2. The geometric correc-tion has to be done for everyimage separately even if itcovers the area only recordedat a different day. In order touse the same Ground ControlPoints (GCP) these are storedin a dBASE data base whichcontains the GCP name, ashort description, the FMG Xand Y value and the file posi-tions (line and column num-bers within the different imagefiles) as shown in Table 1.

The establishment of a GCPdata base allows this data toused for geometric correctionof different satellite images.The data base can be updateddirectly from the ERDASmenu by calling a MS-DOSbatch file which starts dBASEand an associated programwhich opens the GCP databank base in and browsemode. The operator can in-clude additional GCPs or cor-rect FMG values or image filepositions.

Creating a TransformationMatrixA transformation matrix isnecessary for the geometricalcorrection of digital images.Transformation matrices canbe used to calculate the posi-tion for every pixel in a newraster file which will containthe rectified or corrected im-age. The transformation ma-trix consists of coefficientswhich are used in polynomialequation to convert the co-or-dinates and are calculatedfrom the GCP image file posi-tions and FMG co-ordinates.The coefficients are stored ina separate file with the CFNextension.

To establish the transforma-tion matrix all GCPs of asubarea covering one mapsheet have to be selectedfrom the GCP data bank andstored in a Ground ControlPoint file. This can be donesemi automatically at MSD.The user starts a MS-DOS

Table 1. Ground Control Point data bank. The field �MAP� contains the map sheet identification and thenumber of the GCP within this map sheet. The description makes the identification of the GCP easier. The FMGco-ordinates are very precise because they are digitised from the corresponding map sheet. The fields �T03_X,T03_Y, T04_X etc. indicate the line and column number of the pixel within the satellite image file (image fileposition).

MAP DESCRIPTION FMG_X FMG_Y T03_X T03_Y T04_X T04_Y

N26-07 Penany Junction 1935009 3960311 1143 3424 0 0N26-08 Mabua Island 1945776 3963351 1501 3261 0 0N29-04 Wainiburu creek M 1954547 3875178 0 0 1956 1488N29-05 Nanuku Island 1937623 3851573 0 0 2416 580

74/71

75/72 74/72 73/72

75/73 74/73

world reference date of coverage MSD internalsystem (path/row) acquisition scene ID

075/072 25/12/91 full scene 175x180 km T01075/072 15/05/91 full scene 175x180 km T02074/072 27/07/91 full scene 175x180 km T03074/073 06/03/92 quarter 1 90 x 90 km T04073/072 06/12/92 quarter 1 90 x 90 km T05074/071 16/04/92 full scene 175x180 km T06

Figure 2. Landsat TM scene coverage. A similar coverage is available with SPOT scenes.

batch file from the ERDASmenu which calls a dBASEprogram. This program asksthe user for the map identifi-cation and the MSD internalsatellite image identification(see Figure 2). Using this in-formation the program selectsall GCPs from the data base

which have the selected mapidentification and image filepositions. The program thenconverts the GCP informationinto an ERDAS GCP file.

The transformation matrix inthen calculated by a theCOORDN program which

also calculates the error of thetransformation in the X and Ydirection. Using the inverse ofthe transformation, it is possi-ble to recalculate the GCPcoordinates in the original im-age file. The distance (inpixels) between the originallocation and the retransformed

8n8

location is called the RMS error. Theprogram also shows a list of error con-tribution by GCP. This indicates whichGCP position has to be checked. Theuser can correct detected errors quicklyin the way explained above.

This interactive process continuesuntil the RMS error is minimised to anacceptable level (for map sheets cover-ing landmass it should be smaller thanone). Then, the operator starts the trans-formation process for the subset of theoriginal data file. To use only a subset isappropriate because a geometric correc-tion of image files is a time consumingprocess. The operator cuts a part out ofthe satellite image (subset) which cov-ers a frame of approximately 10 km toeach side of one map sheet.

This subset then is geometrically cor-rected by applying the transformationmatrix (see Figure 3).

Creating Map Sheet Image FilesBecause of the overlap, the correctedimage subset is larger than the corre-sponding map sheet. The user has toidentify the part covering exactly the mapsheet. Therefore the user has to locatethe pixel indicating the upper left mapcorner. All pixels (one pixel covers 25 x25 m on the ground) of the correctedimage have a FMG value indicating theircentre. The user has to identify the pixelwhich FMG co-ordinate is most closedto the upper left map co-ordinate plus12.5 m in X and minus 12.5 m in Y-di-rection1. Knowing the file co-ordinate ofthis pixel he is able to identify the pixelindicating the lower right map co-ordi-nate by adding 1600 columns and 1200lines. With the information of the upperleft and lower right pixel position the op-erator cuts out the part which coversexactly the corresponding map sheet.

A geometric correction of satellitedata (30 m spatial resolution) at 1:50 000scale cannot have the precision of atopographic map. However, for furtherGIS analysis it is necessary to have theexact reference. This can be done byreadjustment of the FMG grid. A ERDASprogram (FIXHED) corrects the FMGposition connected to the pixels. Theoperator tells the program that the up-per left pixel has the FMG coordinate itwould have in the corresponding map.The other pixels are corrected corre-

Figure 3. Steps of geometric correction.

2 In reality the FMG co-ordinates are only stored in the header of a ERDAS raster file, the pixels only contain8 bit for every spectral band

spondingly2.This shift is always less than 12.5

m. After this process all pixels fit ex-actly to the corresponding position onthe 1:50 000 topographic map.

Results and RecommendationsCorrected image data is available for16 map sheets of Viti Levu. Further datawill be corrected in the same way. Forevery file the file name indicates themap identification, the satellite imageidentification, the spectral band(s) usedand the type of rectification. The corre-sponding GCP and CFN files are alwaysstored on the same optical disk. In ad-dition information about the RMS error,the shift and other parameter is writtenin an ASCII file with the extension TXT.

The image data is stored in a de-fined way which allows other users towork with it. In an island country like Fijithis data should be used for other pur-poses and satellite data should beshared to for cost reduction. Other us-ers who are starting to rectify satelliteimage data should follow the procedureas described.c

1 The operator has to add 12.5 in X-direction and sub-tract 12.5 m in Y-direction because the FMG valueindicates the centre of the 25 x 25 m pixel and not theupper left corner.

ENLARGEMENT OFSUBSET FOR IDENTIFI-CATION OF GCPs

RECTANGLE OFGEOMETRICALLYCORRECTED DATACOVERS MAPSHEET

SUBSET OF SATELLITE SCENE

n by Asesela Wata, Management Services Division, Forestry Department

LAND SUITABILITY ANALYSIS FORPOTENTIAL HARDWOOD PLANTATIONSAbstract�Eighty percent of all decision requires some use of land related data. They are tools whichhelp people to answer locational problem. For example, where to site a new school, whereto release building land, how to manage natural resources, whether or not to close down abus service and how to plan for the impact of an industrial accident.� Professor Kent Lyons,University of Queensland

Land Suitability AnalysisDetermining suitable land for forest establishment has been an on going process for forestdevelopment for the past years. The lack of an information base and also the lack of equip-ment/technique to derive and formulate scientific and meaningful plans has been resolved.

The build up of GIS within the Management Services Division has improved the capa-bility of the information system and helps to decide ome of future plans on land to beacquired for forest establishment, that in the end will support a sustainable timber produc-tion within the forest sector.Mahogany species is currently the dominant tree species planted to date. Based on pastassessment, it grows extremely well on soil type Humic Latosols and Latosolic soils.

Slope ranged from 00-320 are regarded ideal for reafforestation and above 320 are avoideddue to steepness. The logged out information which is yet to be updated were digitisedfrom 1:50 000 Lands Department Base Map transferred from the 1:50 000 DOS map. To-gether with the existing forest cover (derived from satellite image) we are in a position tocalculate the estimated forest area removed up to 1986. It is hope that with the Introductionof GPS (Global Position System); all logged out data will be captured and updated so thatland suitability study can continue. *

t

t

n9

LAND SUITABILITY FLOW CHART

STEP I SOIL SLOPE FOREST LOG OUT TYPE COVER AREAS

R R I E E M C C P GISMO O O O Retain Plantation D D R area E E T

STEP II REC SOIL TYPE REC SLOPE FOREST SUITABLE (With loggout)

Values Values Values

1 - 4 ==> 100 0 0 - 10 0 = > 1 6 - 8 ==> 100 5 - 5 ==> 1 11 0 - 31 0 ==> 2 9 - 9 ==> 6 < === RECODE 6 - 6 ==> 2 > 32 0 ==> 3 11 - 11 ==> 8 7 - 99 ==> 100

GISMO PROGRAM

STEP III LAND ACQUISITION (LA) RECODE RECOLA Result in Appendix I(a) Appendix I(b)

GISMO PROGRAM

LAND SUITABLE (LS)STEP IV RESULT IN Appendix I(c)

STEP V BSTATS Area Calculation

RESULTS

Appendix I(a) Land Acquisition (LA)_ GIS VALUES DESCRIPTION

11 Moderate slope on Latosolic soils 12 Moderate slope on Humic Latosolic soils 21 Steep slope on Latosolic soils 22 Steep slope on Humic Latosols soils 31 Very steep slope on Latosolic soils 32 Very steep slope on Humic Latosols soils 100 Background

(b)

RECODE VALUES FOR LAND ACQUISITION (LA)

11 ==>1, 12 ==>2, 21 ==>3, 22 ==>4, 31 ==>5, 32 ==>6

(c) LAND SUITABLE (LS) DESCRIPTION

11 - 19 Moderate slope with latosolic soils in (a) 21 - 29 Moderate slope with Humic latosolic soils in (a) 31 - 39 Steep slope with latosolic soils in (a) 41 - 49 Steep slope with Humic latosols soil in (a) 51 - 59 Very steep slope with latosolic soils in (a) 61 - 69 Very steep slope with Humic latosols soil in (a) (a) - Information within non-forest, hardwood, softwood, scattered forest, medium dense forest, dense forest and logged out areas.

Statistics can be produced for various conditions under description....continues overleaf

10n10

RecommendationsFor better yield production, it is recommended to acquire accessible areaswith moderate/steep slope on Humic Latosols/Latosolic soils in arealoggout. Proposal could be made for unlogged areas in moderate/step slopeon Humic Latosols/Latosolic soils which are currently within a loggingconcession.

Inaccessible areas which are yet to be logged could also be highlighted ifthey fall on adaptable soils with moderate to steep slopes.

The production of area statistics for suitable area will help us to prepare areliable and acceptable budget for future forest development. Without doubtconstraints will be easily detected and can be avoided as all small opera-tions are well monitored.

Likewise, logging licenses and allocation of concession can also be moni-tored and controlled, using the land suitability data. Together with the for-est reafforestation plan, we are in a position to cooperate and control all logremoval within our forest resources.c

n by Thomas Kremmers, Barbara Koch, Freiburg, andChristel Lingnau, CNPq, Brazil

ested in high resolution imagedata is the forestry commu-nity, in order to integrate re-mote sensing data more in-tensively into inventory andmonitoring tasks.

Unfortunately the dataquality of the evaluated

AbstractMOMS-02 NIR and HR datawere investigated in compari-son with LANDSAT TM datato evaluate their informationcontent for forestry inventory-ing and monitoring tasks. Dueto the missing information inthe mid infrared, the spectralseparation of different vegeta-tion covers is less successfulfor MOMS-02 data. The spa-tial performance of MOMS-02HR proved to be excellent andprovides forestry related infor-mation superior to compara-ble earth observation sensors.

IntroductionWithin the next future differentnew earth observation sys-tems will provide the usercommunity with high resolu-tion digital image data. One ofthese systems is the ModularOptoelectronic MultispectralStereo Scanner MOMS-02.

The technical concept ofMOMS-02 was explained indetail many times [1-3], andthe major system parameterswere described in the GIS andRemote Sensing News, issuenumber 9, December 1994.

It is expected that one ofthe groups especially inter-

MOMS scene was unsatisfac-tory due to a low sun eleva-tion and strong detector strip-ing in band 2 and 3. The re-sults presented will thereforedemonstrate the possibleperfomance of MOMS-02data under worst case condi-tions.

Test siteThe test site is located insouthern Brazil and is charac-terised by extended pine plan-tations (Pinus taeda) covering

the moderately undulatedparts of the area. Steeperslopes, especially those fac-ing the river are dominated bygallery or secondary forest(Figure 1).

The mature pine treeshave a mean age of 25 yearsand are planted row-wise witha high stand density (left partof the shown subscene, Fig-ure 1). Therefore, pine standsthat have not been treated byselectiv logging show a veryhigh degree of canopy clo-sure. Selective logging is car-ried out in the form of row-thinning. Every third or secondrow of trees is cut out, so thatthe pine stands appear ‘stripy’

when seen from the bird’s-eyeview, e.g. on aerial photo-graphs.

A regeneration area is lo-cated at the upper right handside of the mature pinestands. The plants have amean age of three to fouryears (1994) and the canopyis far away from closure. Oldroot plates and other woodresidues from the last clear-cutting have been pushed to-gether to form long straightrows, crossing the regenera-

tion area vertically. The regionbetween the regenaratingplantation an the secondaryforest following the river ismostly covered with bare soilor dry grass.

Data baseThe MOMS-02 data used forthis analysis are a subsceneof data take 43. They wererecorded in the afternoon ofthe 4th of May 1994 usingimaging mode 6, which is acombination of the bands 2, 3,4 and the high resolution (HR)panchromatic band 5. Duringthe data take the sun eleva-tion changed from 21.84 degat the beginning to 16.09 deg

MOMS-02 DATA FOR FORESTRY PURPOSES� A FIRST EVALUATION �

Figure 1.MOMS-02high-resolutionpanchromaticimage of thetest site.

n11

n by Osea Tuinivanua, Forest Management OfficerForestry Department, Fiji

at the end. This fact causesthe strong and very distinctshadows in the MOMS-02 im-age. Unfortunately the dataquality of band 2 and band 3was not satisfactory becauseof a very strong row- and col-umn-wise striping. So only theHR band 5 and the NIR band4 were incorporated in thisstudy.

The LANDSAT 5 TM datautilized are a subset of scene220_79 and were recorded inthe morning hours of the 10thof April 1993 with a sun eleva-tion of approximately 36 deg.Aerial color-photographstaken with a non metric 35 mmcamera and a scale of ap-proximately 1:8000 were pro-vided for this analysis.

Data pre-processingThe MOMS-02 typical imagestriping was eliminated quitewell for the HR and NIR bandduring the preprocessing car-ried out by DLR. Neverthelessthe image quality of the HRband was slightly reduced bythe randomly distributed ap-pearance of single picture el-ements with relatively highgrey values within uniformdark sorrounding areas. Tominimize this noise effect theHR image was filtered with a3x3 low pass convolution fil-ter. After that a rotation wasperformed to orientate the im-age approximately to the northdirection. For the comparativeanalysis and the later on com-bination of MOMS and TMdata, the MOMS NIR bandwith 13.5 meter pixel size andthe TM bands with 30 meterpixel size were registered tothe MOMS HR band.

Data analysisA signature analysis incorpo-rating 13 test areas was car-ried out to evaluate theseperability of the main landcover classes. The test areaswere collected according tothe exposition dependent illu-mination conditions and werethen compared with respect totheir mean value and theirstandard deviation.

To to give an estimation for

RS and GIS application forMonitoring Fiji's Natural ForestAbstractA national forest inventory was carried out for Fiji�s natural forest and forest plantations. Thedetailed data base for all hardwood plantations and natural forest are now available locally.Fiji�s natural forest inventory also included the production of forest type and forest functionmaps. The mapping required satellite data analysis for forest type stratification and GIS tech-nology for the definition of forest functions. There were some unexpected challenges andproblems which undoubtedly other South Pacific Countries will face. These were the acqui-sition of cloud free satellite data and the inclusion of local knowledge. A practical solution isthe installation of remote sensing and GIS technique in Fiji, operated by local foresters on apermanent basis and supported by regional organisations in the region.

IntroductionThe lack of up-to-date information about Fiji�s natural forest resources development wasexpressed by Fiji in Geneva and a natural forest inventory was decided along with the Tropi-cal Forestry Action Plan. The German Agency for Technical Co-operation (GTZ) then con-tracted the Consultant Companies GOPA/SIGNUM to complete this task in 24 months.

The expected output was:� statistical report on current status of natural rain forest� 1:50,000 scale map showing forest types and forest functions

The forest inventory covered all the main islands forest area and forest tree species inde-pendent of its social and economic values. It took also into account regeneration potential,minor forest products and environmental parameters. The field survey was completed in1992 whereas the map production is not fully completed yet.

Components of Fiji�s Natural Forest Inventory

The Terrestrial SurveyOver 500 sample plots have been measured during the field work. These plots have beenrandomly distributed throughout the country. The three four men team for the inventoryconsist of a forester, a treespotter and two labourers carrying out measurement of all trees,natural regeneration counting and the identification of plants used by the villages as food or

the accuracy of a Bayesianclassification, a land coverseperability matrix was com-puted for the MOMS data us-ing the separability index S fortwo classes, i and j,

where m and s are the meanand standard deviations [5].For Gaussian distributions,index values S 2.0 indicate acorrect classification for thetraining areas of 95.44% orbetter.

The matrix (Table 1, p.12)shows the separability indicesfor the HR band left of the di-

agonal and those for the NIRband on the right. Bold framedareas highlight indices ofthose training areas belongingto the same illumination class.

Looking at the separabiltymatrix for the HR band, it isobvious that spectral discrimi-nation between mature pinustaeda stands and secondaryforest is very poor in all illumi-nation classes (indices rang-ing from 0.28 to 0.32).

Grassland can be sepa-rated quite well from the pinusstands and from secondaryforest, whereas regenerationmixes up moderately withpinus, secondary forest andgrassland. Even the separa-tion between water and thosevegetated areas showing lowreflection is not significant.

As the spectral behaviourof the different land covertypes is similar for all illumina-tion classes, an exposition de-pendent pre-stratification willnot improve the differentiationamong the land cover typeswhen looking only at the pan-chromatic band, but it will helpto avoid mixing between the il-lumination classes.

Due to the larger dynamicrange in the MOMS NIR band,the higher indices indicate animproved separation betweensome of the classes. Regen-eration can be distinguishedsignificantly from pinus andsecondary forest when lookingat the illuminated and shad-owed classes. Some mixingoccurs between regenerationand grassland for shadowed

mi – m

j(1) S =s

i + s

j

12n12

flat terrain Illuminated shadowed

pt1 re3 sf5 gr2 pt2 re2 sf1 gr1 pt3 re5 sf4 gr3 w1

pt1 - 1.10 3.24 0.34 1.62 6.95 0.60 2.42 4.66 1.64 4.07 2.44 5.74

re3 1.07 - 3.81 0.59 0.46 5.05 0.48 1.17 5.14 2.41 4.61 3.11 6.10

sf5 0.32 1.29 - 3.01 4.36 9.44 3.54 5.09 1.41 1.91 0.87 0.45 2.32

gr2 1.96 0.96 2.09 - 1.01 5.25 0.16 1.67 4.14 1.66 3.66 2.33 5.00

pt2 0.24 0.79 0.53 1.67 - 4.54 0.96 2.24 5.29 3.12 5.07 3.55 6.57

re2 2.23 1.23 2.33 0.25 1.93 - 5.79 3.76 10.59 8.88 10.01 8.09 11.76

sf1 0.52 0.45 0.76 1.30 0.28 1.54 - 1.70 4.85 2.13 4.30 2.77 5.85

gr1 3.68 2.48 3.68 1.22 3.26 0.92 2.71 - 6.29 3.95 5.77 4.21 7.28

pt3 1.15 2.30 0.68 3.17 1.37 3.46 1.58 5.30 - 3.44 0.51 1.73 0.85

re5 0.07 1.00 0.38 1.90 0.17 2.16 0.45 3.60 1.23 - 2.83 1.21 4.54

sf4 0.78 1.85 0.38 2.70 1.00 2.96 1.22 4.56 0.30 0.85 - 1.24 1.38

gr3 0.96 0.03 1.18 0.93 0.71 1.18 0.39 2.33 2.10 0.90 1.7 - 2.58

w1 1.56 1.76 1.07 3.62 1.76 3.92 1.96 5.95 0.39 1.64 0.68 2.51 -

pt = pinus taeda; re = regeneration; sf = secondary forest; gr = grassland; w = water

areas whereas for illuminatedareas the discrimation is quitegood. Unfortunately differen-tiation between pinus and sec-ondary forest is very poor inthis classes. Except for shad-owed pinus and secondaryforest there is no mixing upwith the water training areastatistics.

As result it can be statedthat reliable separation of dif-ferent vegetation formations isnot possible when using theMOMS HR and NIR channelsalone under such worse illu-mination conditions.

Based on the experienceswith LANDSAT TM data it canbe expected that the incorpo-ration of the MOMS band 2(visible green) and band 3 (vis-ible red) will only slightly im-prove multispectral vegetationdiscrimination.

The combination ofLANDSAT TM band 7, whichis sensitive in the mid infraredregion, with MOMS HR andNIR bands (Figures 2a and 2b,

next page), improved theclass separation significantly.Except the signature overlapbetween water (ellipse 13) andthe pinus stands (ellipses 1, 2,3), the different vegetationcoverages can be separatedfrom each other quite well.The sequence is from pinus(ellipses 1, 2, 3) and second-

ary forest (ellipses 7, 8, 9) toregeneration (ellipses 4, 5, 6)and grassland (ellipses 10, 11,12).

The results of the spectraldata analysis indicate that nei-ther the MOMS-02 NIR bandnor the HR band prove suffi-cient radiometric dynamic fora spectral classification of for-

est types. Even though it is toassume that under compara-ble illumination conditions theradiometry of MOMS-02 band4 is superior to LANDSAT TMband 4 [6-7], a distinct im-provement of spectral separa-tion is not expected becauseimportant information in themid infrared is missing. The

local medicine. The Depart-ment of Environment enu-merated birds in the vicinityof plots. All data are stored ina relational data base (dBASE)for linkage with spatial infor-mation.

Digital Satellite ImageAnalysisRecent aerial photographswere not available, incom-plete and too expensive toproduce. Landsat TM datawas chosen because of itshigher radiometric contentand low cost compared toSPOT. Many cloud coveredareas were replaced by cloudfree parts of a correspondingscene. The separation of for-est cover area from non forestarea processed in Germanyincludes the forest stratifica-tion into scattered, mediumand dense forest.

The Geographic InformationSystemA PC raster data based GIS(ERDAS) was an acceptable

software for the task. The in-formation layers are availablein form of digital maps at theForestry Department. It can beshared by other GovernmentDepartments. This informa-tion includes:� soil map of Fiji� digital terrain model and

slope map� declared areas (declared re-

serves, water catchmentsetc.)

� areas with high Biodiversity� mean annual rainfall� seasonally rainfall� district and province

boundaries� road network� major rivers� hardwood, coconut and

pine plantations� forest cover of last inven-

tory (1969)� forest types (forest cover

1991/1992)The �forest function map �has been the resulting combi-nation of these overlays. Theforest function map shows theforest cover (dense forest,

medium forest and scatteredforest) and a management cat-egories indicating ReservedForest, Protection Forest andMultiple Use Forest.

The �forest cover (1969)�was compared with the �for-est types (1991/1992)� to es-tablish a �forest change� layer.

Installation of Forest Moni-toring SystemThe inventory design wasupgraded to be able to moni-tor the forest change perma-nently and for any changedetection in the tropical Fiji.Therefore, a modification wasforced on all three compo-nents including the terrestrialsurvey, the satellite dataanalysis and the GIS analysis.

Satellite Image AnalysisA PC base digital imageanalysis system was installedby the project in the Manage-ment Services Division (MSD)of the Forestry Department.This system includes inputfacilities to import all types of

satellite data. User friendlyimage analysis software alongwith training allows the for-estry staff to stratify the for-est cover using their field ex-perience. Apparently, furtherstratification is possible withadditional test areas1, 2.

GIS Analysis and Data CaptureUpgrading of software andhardware made map produc-tion possible within the MSDand enabled the direct dataexchange by network to otherprojects in the Division. Partsof the hardware now improvethe map production. Theproblems of recalculationfrom one co-ordinate systemto other is solved by an expertof an AIDAB funded projectwithin MSD. His locally de-signed software package was

1 see Tuinivanua O., Possibilities of MangroveMapping with Landsat TM Data,GIS&Remote Sensing Newsletter, October19932 see Tuinivanua O., Raintree as a New ForestType, GIS&Remote Sensing Newsletter, De-cember 1994

Table 1. Landcover separability matrix for MOMS-02 HR band (left of the diagonal) and NIR band (right of thediagonal).

n13

decisive advantage ofMOMS-02 data is thereforethe high spatial resolution andthe on track stereo capabil-ity.

Visual InterpretationA visual image interpretationwas carried out based on sin-gle band images of theMOMS-02 NIR and HRbands as well as on a colorcomposite of MOMS-02 NIRband (assigned to red),MOMS-02 HR band (as-signed to green) and TMband 7 (assigned to blue).Theinterpretation proved a cor-rect deliniation of all differentlandscape elements in thetest area. The imagesshowed the road netwokdown to forest paths (approxi-mately 5m in width) and gavea good overview on the dis-

available to other Depart-ments.

Terrestrial SurveyOne team continues to estab-lish permanent sample plotsnation-wide. The plots wereconcentrated on secondaryforest including forest not ini-tially enumerated. The MSDalso established an addi-tional data base containingsocio-economic data. Thefield team collecting data toinvestigate the socio-eco-nomic benefits of Fiji�s forest.

Problems with Fiji�s Environ-mentThe inventory was planned inGermany by experts who haveexperienced the tropical coun-tries. However, the South Pa-cific countries have its ownunique features and character-istics making it different fromother countries with tropicalrain forest.

Satellite Data Acquisitionand AnalysisThe project data acquisitionwas based on two Landsat sat-

ellites that were fully func-tional in orbit, Landsat 4 andLandsat 5. However, onlyLandsat 4 was able todownload data, recordedfrom Fiji, via relay satellite tothe ground antenna in USA.SIGNUM in Germany had topurchase 19 scenes (instead of6) to mosaic a cloud free cov-erage of Fiji. Unexpected hazeover forest cover which wasnot visible in the quick lookimages created other difficul-ties for the digital image clas-sification. The forest stratifica-

tion using satellite data wasdelayed due to the mentionedreasons.

The satellite data receivedfrom Germany had been geo-metrically corrected to UTM.However, the image data hadto be corrected to Fiji MapGrid. Further, all pre imageenhancement modules suchas destriping do not workwith geocoded data.

In other tropical areas suchas West Africa or SouthAmerica, there is often a struc-tural difference in canopy be-

Figure 2. Signature ellipses with ó = 2, for numbering of the ellipses see text. a: MOMS HR and TM7; b: MOMS NIR and TM 7.

42

0

22 50MOMS-02 PAN

TM 7

42

0

61 158MOMS-02 BAND 4

TM

7

Figure 2bFigure 2a

closure of the wooded area.Extraction lines in forest landwere discernable in mostcases as well as the linea-ments of old root plates andwood residues in former clear-cuttings. Large single trees,tree rows and groups ofbushes were distinguishable.An estimation of forest crownclosure seems to be possibleand openings within standswere displayed clearly.

According to image texture,a reliable differentiation of pineplantations (uniform smoothtexture) and secondary forest(‘cauliflower’-like texture) waspossible. Differences in colorallowed the seperation be-tween young regeneratingplantations and older parts withmature trees.

This first visual interpreta-tion does not allow any state-

ments on tree species or ageclass assessment, but it canbe confirmed, that the infor-mation content according togeometric details is significantfor forestry applications, e.g.mapping logging roads, forest/non-forest boundaries andmonitoring logging schedules.

ConclusionThis study was based onMOMS-02 data with low qual-ity and was restricted to asmall test area and a singleforest management system.Therefore conclusions on theinformation content of MOMS-02 data for forestry purposesare limited. However, for thespectral separation of forestclasses the NIR and HRbands alone were not suffi-cient, because the importantmissing information from the

mid infrared region cannot becompensated by these chan-nels. The integration of the vis-ible channels 2 and 3 willhardly improve the separabiltyof vegetation coverages as theexperiences with SPOT andLANDSAT TM show. There-fore an improvement in thespectral separability of vegeta-tion classes is only achievablein combination with mid infra-red channels, as it was dem-onstrated with LANDSAT TMband 7.

Even though it has to betaken into account that thecombination of different sen-sor data can cause a lot ofproblems (illumination condi-tions, data integrity, geomet-ric properties) data fusiontechniques will become moreand more important to exploitthe advantages of differentsensor systems. The decisiveadvantage for incorporatingMOMS-02 data in such datamerging procedures is its ex-cellent geometric resolution,especially in the HR band.

AcknowledgementsThe MOMS data set was pro-vided and system corrected bythe DLR, Oberpfaffenhofen.The GFZ, Potsdam, sup-ported this evaluation study bytheir advices in data handlingand pre-processing.

Ground truth informationand aerial photographs werekindly put at disposal byMODO - Battistella

14n14

Reflorestamento, Rio Negrinho, S. C., Brazil.

References[1] KAUFMANN, H., MEISSNER, D., BODECHTEL, J. &BEHR, F.J. (1989): Design of Spectral and PanchromaticBands for the German MOMS-02 Sensor.- PhotogrammetricEngeneering and Remote Sensing, Vol. 55, No. 6, pp 875-881.[2] MEISSNER, D. (1990): MOMS-02 Radiometric Perform-ance Analysis.- MBB Doc. No.: MOMS-02, EA.2101.074p.[3] SEIGE, P. & MEISSNER, D. (1993): MOMS-02: An Ad-vanced High Resolution Multispectral Stereo Scanner for EarthObservation.- Geo-Informations-Systeme, Jahrg. 6, No. 1, pp4-11.[4] GERMAN SPACE AGENCY, DARA (ed.,1994): MOMS-02-D2 Data Catalogue[5] SWAIN, P.H. & DAVIS, S.M. (ed., 1978): Remote Sens-ing : The Quantitative Approach.[6] BODECHTEL, J., LÖRCHER, G., SOMMER, ST.,AMMER, U., KOCH, B., & SCHNEIDER, TH. (1993):Geoscientific Objectives and Expected Thematic Performanceof MOMS-02 Imager.- Geo-Informations-Systeme, Jahrg. 6,No. 1, pp 11-15.[7] BERGER, M. & KAUFMANN, H. (1994): PerformanceVerification of Spectral and Panchromatic Modules of MOMS-02 Sensor flown on Board STS-55 D-2 Mission.- Proc. IEEE,International Geoscience and Remote Sensing Symposium1994, pp 2301-2304.c

tween virgin or dense forest and logged forest, because thelogged forest has a rough and shadowy canopy. The canopyclosure is often a good indicator for the forest density. This ishighly irrelevant in Fiji. The canopy closure is not related toforest density in terms of standing biomass per hectare. Dueto frequent cyclones the forest has a rough canopy structure.Furthermore, tree ferns as a pioneer species covers the landslides, have no woody biomass but exhibit tree reflections inthe satellite image.

The geographical distance between the digital image proc-essor and the foresters with local knowledge has been an on-going problem. It�s also an economic challenge in developedcountries to do all computer work by foresters with experi-ence in tropical rain-forest. Such an operator will not realisemany phenomena and he will not ask the foresters in the SouthPacific Country, even if a good communication link such asfax and e-mail is established. For example the seasonal effect(defoliation) with African Tulip (Sparodae campanulata),Raintree (Samanea saman) or Mahogany (Swietenia macrophylla)will not be recognised by an off shore operator.

Terrestrial SurveyThe review of the forest legislation carried out by an FAOexpert did not include the forest definition. The boundarybetween forest and non forest was not defined. The plotswere checked in Germany with satellite images if they werelocated within or outside the forest cover. A non local opera-tor will always underestimate the forest area. However, therewere secondary forest mapped as none forest in the last in-ventory3.

Spatial Data Collection and AnalysisMany spatial information already available in the country hadto be redigitised because it was impossible to obtain these maplayers due to administrative and technical reasons. Many do-nor countries have kindly donated computer hardware andsoftware to various Government Departments and institutions.Quite often, these products are not compatible which makesan information transfer impossible.

The mentioned use of different spheroids and projectionscreated an enormous amount of additional work because itwas not always possible to transfer differently digitised mapsto the other grid system.

RecommendationsBefore starting an inventory related to nation wide forest cover,the legal definition of forest should exist.

Within the Management Services Division of the ForestryDepartment, several technical and financing projects are work-ing harmoniously together. This was possible because all ex-perts form an integral part of the Division and Department ofForestry. In the future, more emphasis should be put intoproject planning to be aware of full integration in the admin-istration, rather than creating an own independent projectinfrastructure outside the administration.

It is necessary to get agreement from the satellite data dis-tributing agencies to share data between different Govern-ment Departments. Such an agreement has been reached withEOSAT for Landsat data, but it is also necessary for digitalSPOT images.

Cloud cover will be a continual problem in South PacificCountries until a break through in space borne Radar data. Itwill be an important step to have a Mobil ground station able

3 see Tuinivanua O., Change Detection of Fiji National Forest Cover Using Landsat TM Dataand GIS Techniques, GIS&Remote Sensing Newsletter, September 1994

The Digital Elevation Model has been produced for the Namosi areaof approximately 5 by 7 km with the pixel size of 20 m x 20 m. Wewill see the different steps in the production of this DEM from thebase data to the end-product, the GIS map.

The map produced (Figure 1) contains :� a raster image derived from the DEM� the contour lines produced from the DEM� geographical information such as rivers, village...

This exercise is part of MRD assessment of the resources of thepossible Namosi Copper Mine. A Digital Elevation Model is one ofnumerous information required to calculate the volume of ore min-able and its content in copper and by-product such as gold. Theviability of the prospect is currently being explored by Placer com-pany.

The base data set which was provided by the exploring com-pany, is a tabular ASCII file. It contains also position of villages,roads, rivers, flood areas. The elevation values, called Z, are classi-cally associated with easting and northing value and with a linenumber. These original elevation data were produced using classicphotogrammetry method. The data are scattered along parallel lines

NAMOSI DIGITALELEVATION MODEL

n by Hervé Dropsy & André Vial, Mineral Resources Department

n15

to receive Landsat TM, SPOT and ERS-1 data.Forest mapping which includes forest stratification is a

complicated issue in the South Pacific. It requires the involve-ment of local foresters. Therefore satellite data analysis shouldbe carried out in the Forestry Department, which makes itnecessary to provide necessary hardware, software and train-ing.

However, the Forestry Department will not be able to up-date this modern technology and will be incapable in carryout all its own maintenance. This could be shared with a re-gional organisation in the South Pacific, as SOPAC4 is doingthis. Such an organisation should be extended to a regionalcentre for central data register and for performing jobs suchas high quality image output, which requires expensivehardware.c4 SOPAC = South Pacific Applied Geoscience Commission

NAMOSI COPPER MINE PROSPECT AREAScale 1:30 000

Figure 1.

spaced at approximately 25 m asshown in Figure 2, and every 10to 20 m along the lines them-selves (total number of points:86,511).

The first part of the exercisewas to separate the geographicinformation from the elevationdata. A database applicationMicrosoft FoxPro was used toimport the ASCII file, then isolatethe geographical information (riv-

ers, village) using the data typeidentificator, and export it into aCGDEF format (interchange for-mat used by Mapgrafix). Fromthis CGDEF file, a first GIS mapcontaining geographical vectorinformation was produced.

The elevation data were im-ported from FoxPro to the GDM(Geological Data Management)system in order to create a gridof elevation and elevation con-

tours.Kriging was used to estimate

elevation values at nodes of thisregular WE-NS grid (size :365X242 spaced 20 m in eastingand northing). 67818 values werecalculated. The rest of the nodeswas given the value ��undefined��by the software (where data werelacking or insufficient, cf the ir-regular shape of the original datain Figure 2). The GDM softwareoffers different griding (=modeling) algorithms, the krigingtechniques giving the most accu-rate estimation in such a case.(as it is also the case for manynatural sciences applications).Furthermore these techniquesallow to calculate at same timethe standard deviation value foreach evaluated altitude, valuewhich can be stored in the eleva-tion model. It represents the de-gree of accuracy of the estima-tion and can be contoured to in-dicate where the evaluations areof good quality and where moredata are required to get correct

values. A contoured map of theelevation was produced with theGDM software from the elevationmodel, using at new time thekriging technique, and then wasexported to a DXF file. The modelitself was exported as an ASCIIfile (one pixel value per node).

The integration of the varioustype of data (geographical, vec-tor DEM, raster DEM) presenteda few challenges.

The vector DEM (contour linein DXF format) was convertedinto CGDEF (ASCII interchangeformat used by MapGrafix) us-ing the module MapLink. Theresulting file was imported intoMapgrafix without difficulties.However contour in the DXFwere coded as a series of seg-ments (line linked two points).The CGDEF was processed us-ing a small program (usingFoxPro) to aggregate the seg-ments into line. This alone de-crease the size of the CGDEF filefrom 12 Mb to less than 4 Mb.

The grid was first classified

16n16

into 5 m interval classes, and then Idrisi 4.1was used to import the grid ASCII file pro-duced from GDM and to export it as a TIFFfile. Due an incompatibilities between TIFFfile format with MapGrafix and Idrisi. A im-age processing shareware was used to pro-duce a compatible TIFF format, the lookuptable was also changed.

The TIFF file was then imported intoMapGrafix GIS as a background image toproduce the final document.

A contoured map of the elevation wascalculated from a relatively small number ofpoint (250) on a part of the studied area. Itwas compared with the one produced fromseveral tens of thousand of measured points(approximately 30 000). Results are of suf-ficient good quality where points are regu-larly placed, are not too far apart from eachothers, and are also measured onsignificative position. Associated with thevalues representing the accuracy of evalu-ations, this DEM can be used in natural re-sources estimations which do not requireda too high degree of precision. If this is thecase the number of measured values mustbe increased taking in consideration the in-formation given by the contour map of thedegree of accuracy. It is then easy to dem-onstrate that above a certain number, extrameasures do not increase the exactness ofthe calculus.

So the combination of a griding & con-touring software with a GIS can provide aquick and cost-effective way of producingmaps representing in vector and raster forma digital model of elevation or other type ofdata (rain, mineral content, ...) from a scat-tered set of measures.c

Figure 2.

Requests for inclusion in the mailing list for this newsletter as well as the submission ofarticles for publication should be sent to:

GIS AND REMOTE SENSING NEWSSOPAC, PRIVATE MAIL BAG, GPOSUVA, FIJIAttention: Les Allinson

Tel: 381377Fax:370040E-mail: LES@SOPAC.ORSTOM.FR

or

MANAGEMENT SERVICES DIVISIONFORESTRY DEPARTMENTPO BOX 3890, SAMABULASUVA, FIJIAttention: Wolf Forstreuter

Tel: 322635Fax: 320311E-mail: MSD@SOPAC.ORSTOM.FR

It would be appreciated if contributions could be sent on floppy disk in Word for Windows(preferred), Wordperfect for Windows or Wordperfect for DOS format.c

FIJI User Group, GIS & Remote Sensing NewsNumber 11 (9502), June 1995

SEND TO:

The Fiji Times, April 1995