advanced data acquisition and analysis technologies for sustainable development

8/13/2019 Advanced Data Acquisition and Analysis Technologies for Sustainable Development

http://slidepdf.com/reader/full/advanced-data-acquisition-and-analysis-technologies-for-sustainable-development 1/34

Eutrophication management fra m ewo rk for the policy-maker (198 9).

walier Rast, Marjorie Holland and Sven-Olof Ryding.2. Human investment and resource use: a new research orientation al

the environment/economics interface (198 9).

Editors: Michae l Young and Natarajan lshwaran.3. Contributing 10 sustained resource use in the humid and sub-humid tropics:

some researcli approaches and insights (1989) .

Malcolm Hadley and Kathrin Schreckenberg.

4. The role of land/inland water ecotones in landscape managementand restoration: a proposal fo r collaborative research (1989).

Editors: Robert 1. Naiman, Henri Décamps and Frédéric Fournier.5. Management a nd restoration of human-impacted resources: approaches 10

ecosystem rehabilitation (1990).Editors: Kathrin Schreckenberg, Malcolm Hadley and Melvin l. Dyer.

6. Debt-for-nature exchanges and biosphere reserves: experiences andpotential (1990).Peter Dogsé and Bernd von Droste.

7. Carbon, nutrient and water balances oftropical rain forest ecosystemssubject to disturbance: management implications and researchproposals (1991).Jonathan M. Anderson and Thomas Spencer.

8. Economic and ecological sustainability oftropical rain forest management(199 1) .

Editors: Kathrin Schreckenberg and Malcolm Hadley.9. Biodiversity: scientific issues and collaborative researcn proposals (19 91).

Otto T. Solbrig.10. Les systémes agro-sylvo-pastoraux mediterranéens : enjeux et réflexions

pour une gestion raisonnée (1991).

Richard loffre, Bernard Huber el Michel Meuret,11. Long-term monitoring of biological diversity in tropical [orest areas:

methods for establishment and inventory of permanent plots ( 1992).

Editor: Francisco Dallmeier.

A D VA N C E DD A T A A (Q U IS IT IO NA N DA N A L Y S IS T E C H N O L O G IE SF O R

S U S T A IN A B L E D E V E L O P M E N T

10/11 1 E. Es/esMan fred EhlersJean-Paul MalingreauIati R. Noblelonathan RaperAlbert Se limanleffrey L. StarJ im Weber



The designations ernployed and the presentation of material throughout this publieationdo not imply the expression of any opinion what s oe ver on the pan of UNESCOconcerning the legal status of any country, territory, city or area of its authorities. orconcerning the delirnitation of its frontiers or boundaries. The opinions expressed in thisdigest arethose of the authors and not neeessarily those of UNESCO or the authors 'ernployers.

John E. EstesDepartrneru of GeographyUniversity of California, SantaBarbara, CA 93106-4060, USA

Manfred EhlersISPA, Univeristy of OsnabrückVeehta, P.O. Box 1553W-2848 Vechta, Germany

Jean-Paul MalingreauJoint Researeh Centre, Institute forRernote Sensing Applications ofthe Commission of European

Cornrnunities, Ispra, talyIan R. NobleReseareh Sehool o f BiologiealScienees, Australian NationalUniversity, P.O. Box 4, Canberra,ACT 2600, Australia

Jonathan RaperDepartment of Geography,Birkbeek College, 7-15 Gresse SI.London WIP IPA, UK

Albert SellrnanDirector of International Serviees,CIESIN, P.O. Box 134001,Ann Arbor, MI 48113-4001, USA

Jeffrey L. StarDepartrnent of Geography, Universityof California,Santa Barbara

CA 93106-4060, USAJim Weber

Publie Service Satellite Consortiurn,1235 Jefferson Davis Highway,Suite 904,Arlington, VA 22202, USA.

Addresses of the auih ors of this reporteOverall direction: Moharned SkouriSeries edi/or:Malcoírn HadleyComputer assisted layout: Ivette FabbriCover designe Jean-Francis Cheriez

lllustrntion 011 back cover: NASAIJPL Spaee Sh uule Imaging Radar A (SIRA) of faultzones in western China

Suggested citation: Estes. l.E., Ehlers. M. .Malingreau, l.P. , Noble, I.R., Raper, l.,Sellman, A., Star. l . L Weber, l. 1992. Advanced Dota Acq u is iti o n and AnalysisTechnologirsjor Sustainoble Development. MAB Digest 12.UNESCO, Paris.

Published in 1992 by the United NationsEd uca tional , S cicmific and Cultural Organiz.uion7, place de Fon tcnoy. 75352 Paris 07 SI'Printed by UNESCO© UNESCO Novcmbcr 1992Printed in Fr a n cc

P R E F E

A b o u t th is s er ies . ..

The MAB Digest Series was launched by UNESCO in 1989. Severa t ypes of

publications are included: distillations of the substantive findings of MAB activ-ities; overviews of recent, ongoing and planned activities within MAB in partic-ular subject or problem areas; and proposals for new research activities. The tar-get audience varies from one digest to another. Some are designed with plannersand policy-makers as the main audience in mind. Others are aimed at collabora-tors in the MAB Programme. Still others have technica personnel and researchworkers as the target, irrespective of whether or not they are involved in MAB.

... a nd M A B D iges t 1 2

The purpose of this digest is to provide scientific researchers engaged in activi-ties reJated to sustainabJe deveJopment with an overview of several data collec-tion and information technologies, including remote sensing, geographical infor -

mation systerns and advanced modeling tools. The primary audience is not theexperts and pracritioners of the technologies and approaches described in thedigest, but rather members of the broader scientific com rnunity who might wishlO have at hand a synoptic overview of the characteristics and potential of newtechnologies and their possible app licarion 10 contemporary research challenges.Particular ernphasis is given to rernote sensing and geographic inforrna tionsysterns.

The digest has be en prepared by (In ad hoc group on Scientific Informntionfor Sustainable Development sct up in 1989 by the Scierui fic Commiuee on

Problems 01' the Environmenl eSCOPE). with support from its parent body, the



AccuracyStandardsTimeJinessTechnoJogy trendsRoJe of scientists

6J61636364

67o u rc e m ate ria l

S U M M A RY

Throughout human history, technology has been a key factor in faci lita tin g anddriving change. Today s technologies can create environtnental change on spa-tial and temporal scales never before possible. Yet, the wise application oftech-nologies can also facilitate investigations and lead to a more complete under-standing of human. impact on the environment. Througli appropriate applicationof remote sensing, geographic information systems and modeling, we can movea significant way towards an~wering questions concerning the spatial and tem-poral dimensions of variations in environmental resources. These technologiesfor data collection, manipulation, analysis and information extraction can beused by the scientific community 10 : improve our understanding of the globalenvironment; measure, map , monitor and model changes in that environment;and provide decision-makers with the information which they require 10 summa-rize the options available to decision-rnaking bodies.

Remate sensing is used to gather information from a distonce, with sensorsrecording electro-magnetic energy emitted 01 reflected from the Earth s su rface.

Different types of vegetation, soils and other features emit and reftect energy dif-ferently. This characteristic makes ir possible 1 0 measure, map and monitorthese [eatures using remote sensing svstenis. Sotellite sensors collect the bulk ofremate sensing data C/I the present lime. Most o.f these data are ovailoble in [ WO

formats: electronic ami analog (i.e. image). Electronic data I1 1U .I I generally beenhanced andfiltered by a computer and compored with limited ground truth data before thev can be used for resource munagctnent.

Geographic ln jo r mati o n Sys tems can be defined os comput er systems [ o r

integrating and analysing spatiat injormotion in a decision-making context.Recent advances in GIS technologv include: nel hardware and s oftw a re [o r

digilizing and matiaging dora, extracting information [rom coniplex databusesand re p ro d uci n g 11I0p S . making these tools more powerful mil i e a s ie r 1 0 use;

9



desk-top work stations. with increased data-processing speed and dat a storagecapabilities; inexpensive p lo tt e rs lowering the CO SI and improving the quality o]produc ing maps, charts and sables; easier 10 use GIS software; and improvedgraphics and visualiiation 10015.

Models can take manv [orms and are built Jor a variety oJ purposes. A I/UI1l-be; of factors have restricted the application of models in the management ofnatural resources, bUI these problems are being overcome. The tasks required ofmodels are being stoted more clear/y (e.g. prediction of crop yie lds , structural

changes in vegeta/ion, land suitability¡ and a suite of suc cess ful generic modelsis being developed in such fields as fores/ growth, crop /rajectories and speciesranges.

Expert systems (sometimes called knowledge based systems 01 advisory sys-tem) are computer programs able 10 simulate the problem-solving ability oJ anexpert, and are among the 1II0s1 prominent products of the revival in artificialintelligence research. Al expert system can be thought of as a model composedof rules and these rules are o fte n qualitative (e.g. if the amount oJ monoculturecropping in the district is higher than the normfor the region tnen.i.}. This is amajar area of application in remote sensing where rule based models may assistin the initial classification of images. The other maja r app lication of expert sys-tems is in assisting users harness the forces of the very powerful, but complextools, provided in a comprehensive GIS.

These new information technologies have many uses in resource manage-ment, and are increasingly being used in combination, as illustrated by exam-pies on land use planning in Senegal, determination of natural hazards risk toagricultural production in Ecuador, crop yield estima/ion in Ita ly and habitaranalysis for preservation of species richness in the United Sta/es.

Key issues in (he [u r the r use and development of these new informa/ion tech-nologies include the question of scale in scientific analysis and policy-making,infrastructural investments for the successful adopti on of tec hno logies , accuracyin physical nieasurement and in interpreta/ion and transformation, standardiza-tion of transfer formats for spatial data, timeliness of access to data, trends intechnological progress. aud clearer and wider understanding of the role o]science and technology in hu man affairs .

10

INT R O D UC T I O N

As scientists we strive constantly to keep abreast of new developments in ourareas of specialization. This is not an easy task in today's increasingly informa-tion-oriented society. As the rate and volume of output of basic data anelresearch results in our own and cognate fields increase, many of us find it

increasingly difficult to follow developments in those fields seeking to i rnproveour understanding of our global resources. Advances in climate and atmosphericmodeling, improvements in hydrologic forecasting and our understanding of thecarbon cycle are significant. Equally significant are developrnents in those fieldswhich facilitate the acquisition and analysis of dataJinformation. These are so-called hi-tech areas. In recent years there has been an ex plos ión in the areaof information technologies. These technologies offer the potential to facilitatethe condu ct of research, resource management and policy decision-making atscales frorn local to global. The technologies which we believe are critical toimproving our understanding of the best paths towarel sustainable econornicdevelopment include: geographic information systems (GIS). rernore sensing,anel moeleling with expert systern- ancl artificial intelligence-assisted

me rhodo logie s .

Back gro u n d

In 198 7 the World Cornmi ss ion 011 Environment and D cvc loprn cnt p ublishcd theresults of its two ycars 01' eleliberation in a rcport c nt i t lcd O ur Co m m o n

Future . A key element 01' this repon wns irs cndorscmcnt 01' thc concept 01'

s u s ia in a b le dcvcloprn e ru , d efi n ed as d cvc lo p m c nt rh .u Il lCC IS rhe n c c d s of thc

present wiihout com prornis ing lile ahilii y 01' fururc gcncr.uion s to meet thcrn .

Jl



To carry out this concept of resource development with due consideration ofthe mid-term and long-term effects of such development on the environmentwill require that researehers, planners, and decision-makers have aceess 1 0

unpreeedented quantities and types of resouree and environmental information.Measurements from specific sites need to be eombined with data and informa-tion on eonditions al other locations. Maps depieting the spatial dimensions ofobjeets and phenomena need lO be produced and updated to keep traek of (i.e.monitor) ehanges at varying time intervals. Models depieting the range of alter-natives whieh could result from actions ha ve to be tested.

Many of the systems required to produce these large quantities of data cur-rently exist and data are being eolleeted now. Other advaneed data eollectionsystems are eurrently in the planning and development stages and will produceeven larger quantities of data in digital format within the next ten years.Likewise, improved tools for eonverting these data to relevant environment andresouree and poliey-oriented information either exist or are in varying stages ofdevelopment.

As scientists it behoves us to keep abreast of relevant teehnologies which canenhanee our investigations and which can be employed to produce results whiehare readily understandable by individuals in the resouree planning, management

and pol iey areas.

P urpose of O o c u men t

The purpose of this document is to provide seientists, researehers, resource man-agers and publie poliey deeision-makers engaged in sustainable developmentaetivities with an overview of several technologies for data eolleetion, manipula-tion, analysis and information extraction. Such technologies can be used by thescientific eommunity to improve our understanding of the global environment,measure, map, monitor and model changes in that environment, and provide

deeision-makers with the information whieh they require to summarize theoptions available 1 0 decision-rnaking bodies.

A suite of technologies that has the potential to improve our ability toachieve the goal of sustainable development is eurrently available, yet in differ-ing staies of maturity. This suite of technologies includes GIS, rernote sensing,and advanced modeling tools. Applications of sorne of these technologies haveexpanded rapidly in recent ycars, while some trace their progression from devel-oprnents dating back into the last century. Use of these technologies canimprove our ability:T to acquire measurernents of key state variables ancl irnages from whieh spa-

tial information on environmental constitucnts can be derived;

2

T to access data we require in our investigations more efficiently;T to store, retrieve, manipulate and analyze these data more effectively; andT to provide a wider range of output products to suit user requirements (be the

user a seientist, environmentalplanner, developer or decision-maker).We are modifying our environment at unprecedented rates and scales. We can,however, debate the specific spatial c1imensions, rates and significance of thesechanges. As humankind has progressed from the Stone Age through the Bronzeand Iron Ages to the agricultural and industrial revolutions, technology has beena key factor facilitating change. Today's technologies can crea te environmentalchange on spatial and temporal scales never before possible. Yet, the wise appli-cation of technologies al so offers us the ability to facilitate our investigationsand leads to a more complete understanding of human impact on our environ-ment. Through appropriate application of remo te sensing, GIS and modeling wecan move a significant way towards answering questions conceming the spatialand temporal dimensions of variations in environmental resources.

There are still barriers to the adoption and application of these technologies.These barriers are no longer primarily technical. They are institutional in manycases, sociological in some and legal in others. Some cite the costs; others the

time required to leam; others believe that capabi ities are still no sufficient fortheir purposes. These reasons may all be valido Yet, costs of sharing data andcomputing today continue to rapid y decrease. Systems are easier to leam, per-fonnance is rapid y increasing. What has not been fully realized by the scientificcommunity at large is the scientific ancl cornmercial potential of these technolo-gies. Examples of the use of these technoJogies to enhance the understanding ofkey science issues exist; severa examples are discussed in the body of this doc-ument, yet more can and should be done.

We have not even come c1ose, we believe, to exhausting the potentiaJ ofthese technologies. A great deal needs to be done; we have only scratched thesurface. What is required is innovative thought and a willingness to work withindivicluals in cognate fields seeking to achieve the goal of sustainable develop-

ment. Ir is our eonsidered opinion that working together we can achieveimproved insights, by employing GIS as an integrating resource. Through theuse of remote sensing systems, we can begin to collect required c1ata which caneither be translated directly into information or input into models which improveour understanding of key environmental processes related to sustainable devel-oprnent.

W ho t Fo llo ws

The following material presents a brief ovcrview of sorne important data typcs

3



available 10 the science cornmunity today which could be used as input lO sus-tainable developrnenr studies. This is followed by a discussion of the current sta-tus of GIS, image processing, modeling, artificial intelligence and expert sys-tems. The paper concludes with a review of some issues involved in the use ofthese technologies, needs for further research, and tre nds iowards futuredevelopment.

Information

analysis

for

management

decision

making

Figure1. Tronsformingdolo lo informolion

14

D A T A

It is important to understand specific characteristics of data types used in envi-ronment and development programmes. Too many of us often accept data at face value ; often it may be the only available data. Today, however, thepowerful too1s we have to col lect , integrate, manipulate, and analyze data makeit necessary that we understand both the nature and accuracy of our data and itssources as never before, including recognition of the distinction between dataand information, and the distinction between spatiaI and non-spatiaI data andinformation,

Data V5. In form a tion

An essential goal of the use of inforrnation systems and models is to convertdata into information: converting measurements and observations into knowl-edge which may be used for evaluating options as a step toward making deci-

sions. As used in this paper. data can be thought of as any input to a processwhere information is created. As such, measurements could be directly analyzedto provide information. e.g., daily rainfall lo produce a rnonthly mean averagefor a given location. Measurements could also be used 10 model a process, Themodel's output could then be employed as information in a management deci-sion process which would in turn gcnerare oiher information types (see Fig. 1).

In this process of making measurerncnts and observations which will be usecl 10generate inlormation, one 1l111stbe extrcmely careful lo evaluate the quality anclvalidity 01 the conversión process. In many instances, the costs of verifying thequality of datasets can be greater than the original C OSl S 0 1 acquiring the data.

As a par of our concern to produce high quality data/inforrnation upon whichdecisions on the best paths toward susuiinable developmcnt can be based, we 1l111S1

15



distinguish between accuracy and precision, not only in the raw data, but also inthe derived information. Accuracy is concerned with fidelity to a standard, andlack of bias. Precision, in contrast, regards the ability lO make fine distinctions.For example, one could measure annual rainfall lO the nearest 10cenrimetres in asimple rain gauge, which rnight provide highly accurate (or unbiased) data whichare rather imprecise. Such imprecisión, if inappropriately extrapolated over spaceand lime or across scales, could invalidate the use of these data and lead to incor-rect decisions. On the other hand, use of precise but inaccurate (or biased) datacan also create significant problems for managernent decision-makers.

Sp o t iol v s. N o n-S po tio l D o ta a nd In fo rm o tio n

In applications which involve environmental analysis and sustainable develop-ment, we must necessarily consider both spatial and non-spatial data and infor-mation. By spatial data, we mean data that can be located on the Earth's surface:a water well, a river, a polítical or adntinistrative district, By non-spatial data,

we incJude such information asthe date that the wa ter well was dug, the pH andtemperarure of the river at a monitoring station, and the name of the politicaldistrict. In a variety of applications, the volume of the non-spatial data may bemuch larger than that of the spatial.

There is a cJass of non-spatial data associated with spatial datasets as a wholewhich are called metadata. Metadata essentially describethe characteristics ofa given data set. Metadata are often described as data about data or data sets. Forexample, the date at which the dataset was verified for accuracy, and the mea-surement scale of a dataset (e.g. elevations were recorded in metres, rather thanfeet) are both forms of rnetadara. Metadata are often key elements of catalogueswhich can facilitate access to an understanding of spatial data and inforrnation.

In the context of this document, we distinguish two broad c1asses of spatialdata which are commonly used in the analysis of environmental problems and inthe consideration of issues related lOsustainable development. We discuss thesedata which come from remote sensing systerns separately frorn those ihat do not.Remote sensing iechnology (as described in the fol1owing secuon) provides themeans to consistenlly collect data over large areas of the earth Ior a number 01'applications. In addition, rhere are a variety of other data types and thernesdescribed subsequently in this docurnent, some of which may even represent aproduct derived frorn ihe analysis of remorely sensecl data that are importanlinput lO a wide variety 01' sustainable development relaied studies. Thcse sec-tions provide definitions Ior and examples 01' c1atatypes which are being usedwithin geographic information systerns and advanced models LO improve ourknowledge of the global enviroument.

16

S PAT I A L D AT A

R e m o te S en sin g

Since the early days of the various space programmes, there has been an increas-

ing public awareness of the ability of space-derived imagery to capture a senseof our fragi1e planet. While remote sensing dates from a century before the eraof space ftight, there is probably no more familiar remote sensing product thanthe daily view of clouds and weather pattems presented in weather forecasts onthe television news programmes in many countries.

Defi nitio n

Remote sensing is the use of systems to gather information on objects and phe-nomena from a distance. The human eye can be considered a remote sensingsystern as can an ordinary camera. Yet, contrary to the popular expression, with

rernote sensing what you get is not necessarily what you see. A great deai ofwhat exists in the world around us cannoi be perceived by hurnan sensoryorgans (see Fig. 2). While one can assurne that distance is the major factor limit-ing hurnan visión, the spectrum of energy we can see with the unaided eye isalso an importan¡ lirnitation. The human eye responds lOlight from o~llya smallportion of the electromagnetic spectrurn from bJue lo red, while energy frorngamma rays lOradio waves swirl around li S unseen. Through the years, however,we have consturuly siriven to expand our ability lo see energy fielcls beyondour direct perccprion. 111 this process we have constructed a variety of spccial-ized insirurnents 10 record infonnation and present it to us in picture-Iike forrnscal le d images.

17



1()1Q 1018 1016

F r eq u en cy in cy eles seco nd·1

1014 1012 1010 108 102

xM

~------

E-co

xM

,,

, S ide look ing , E lectro magn elic ,: a ir bo rn e r ad ars . : p ulse lechn iq u es ',, , ,, rf receivers in , ,, scanners and ' ,, ra diom elers' ,, , ,, , ,, , ,,

EE

eo> (

o

ci

---: S ci ntillation, counlers.,, Gomma·rays: spectomelers

1,,,

, Counlers' ' Fil m in c am era s. ' C am era s w it h: with f illered :: • : in frared, photo· Scanners with ,sensitive film., m ultipliers. , f i ltered photo- ,, , rnulfipliers ' Solid slate: I mag e o rtho con ~ : d et ec lo rs i n, amI- cameras , , scanners and, with filtered , , radiometers, in frared fi lÍ n ' , , .

'~.,,-, ' , ,:~; ..

'.Solid·stote¡d etec to rs in, s cc nn er s a nd, rqdiometers,,,,

F ig u re 2. T h e f ronsm iss ion0 1e le c fr o m og n ef ic e n er g y t h ro u q h t h elo r th's o f m osp her e (below)

ond exomp le s0 1s en so rs t hn t im o g e i n i f ( o bo ve ) ( od op fe d I ro m E s fe s, 1 9 74 )

R em o te S en si ng S ys te m s

Today, a wide variety of systemsare used to detect energy patterns(see Fi g. 2 ) which provide infor-mation about the earth, its atmos-phere, oceans, and land surfaces(see Table 1 and Fig. 3).

The bulk of satellite remotesensing data is currently acquiredby multispectral scanners and lin-ear array devices in t he visible andnear infrared positions of the elec-tromagnetic spectrum. These arepassive systems recording solarradiation reflected from the earth'ssurface. Data derived from multi-spectral scanners provide informa-tion on (among other things): veg-etation type, distribution and con-d ition; geomorphology; soils; sur-face waters and river networks.

These remote sensor systemscan be broadly categorized intoactive systems and passive sys-terns. A camera is an example of a passive system as its film records light energyreflected from an existing source: the sun (see Fig. 3). Add a flash attachment tothe camera and it becomes an active system, since it now provides its ownsource of illumination. Active m ic rowa ve (radar) systems are commonly used ingeological and hydrological applications.

Active systerns in the optical range using laser technologies are being devel-oped for oceanographic and Iorestry applica rions. Because active systerns do notdepend on ref lec te d energy frorn the s un for image forrnarion they may be ableto acquire data during the day or night. In addition. as sorne microwave radiationis not disturbed by the atmosphcre ro as great a degree as shorter wavelengthenergy, radar sysicms allow data acquisition in cloucly and rainy weather (seeTable 1). This ability lO opcrure day or nighl, and to cre.ue irnages of the surfac ein spite 01 cloud cover. ma ke s racial' sensors parucularl y auructive tools -forirnprovin g our unde rsta nding 0 1 boih tropical a nd polar areas. Thermal intrareddata which represent a record of cmittcd energy Irom surfuces ha ve becn partic-ular ly usefu l in monitoring tires and in irnproving our understanding 01 arcas 01'volcanic and gcorhennat acuvuy. Surluc e tcmpcrarure 01 the occan is also

Dust cloud 25 At mosphericabsorption = 19

F ig u re 3 . E n er gy Ilow m o d el lo r r em o fe

s ens ingsystems

19



-:-v .:-.._~ ~'.~+ ~-:, ;; 1:t;~ <kP..< ~v : r : ; i . : . c _ -J . - f •. 0 - • r.. . ? l - . - ~'~~ '-?~ : .5ysÍem chorocteristics inClud ed or e hos 'most comrTlon¡Y,empl?yed in resource pJonning

. ond monogement cpplicotions todóy: . ,< ',.' ~ ' ' '' '',' . .2 , 6enot~s the otmospheric co~ditio~~ hot can be pene tr qt ed b y~~ergy in this portion

.h ¡. : otthe electrornoqnetic spectrurn where: .;,'\ c:·¡:?~ . .

= haze; 5 =\moke; 5g ~'smog; F = fog or clouds; R ~·rq~n. .Discounting the use oF active opticol systems su ch as f la sh u nit s, loser linetrocers, or lig ht amplification systems. •

Toble 1. General rh nru derisfic s 0 1r e m ate sensor sy stems

related to the dynamics of coastal waters and currents, Over land, plant waterstress also induces changes in canopy ternperatures which can be detected bytherrnal sensors. Other rernote sensing systerns detect the earth's rnagnetic andgravitational fields; these tools are used extensively in oil and mineral explo-ration. Figure 4 is an ex ample of irnagery 01' the arca around the carnpus 01 ' theUniversity of California, Santa Barbnra, frorn various ponions of the electro-magnetic spectrurn over the sarne general location. Analyses 01' such multispec-tral data can. if propcrly designed, incrcase both the quantity and quality 01 '

inforrnation for given applications.

Reso lu t ion

Resolution is at the core of the successful use of remote sensing data, Yet, reso-lution is a very difficult concept to understand. Resolution can be measured inline pairs per millimetre or modeled using modulation transfer functions as atool. Resolution, as typically applied by a user, however normally involves thequestion can 1 identify what 1 need to identify on this imagery? Often thisinvolves the use of the term ground resolved distance (GRD). GRD is used todescribe the smallest object that can be unambiguously identified on the groundgiven good object to background contrast. Object to background contrast isimportant here. While it may be possible to see white lines on a black road onone set of images, it may not be possible to see the same size white lines on aconcrete stretch of the same road. Resolution is thus dependent upon manythings.

Indeed, the level of matching between the spatial, spectral and temporal reso-lutions of the measurements and the needs of the investigation wil condition theusefulness of the derived data/inforrnation. Spectral range and resolution as used

here refer to the portíon of the electromagnetic spectrum which is being used inthe measurement. It must be appropriate to the questions being addressed, e.g.thermal range for fire detection, red band for vegetation absorptions. Spatial res-olution conditions the level of detail which can be extracted concerning objectsin a given scene. As a general rule it must correspond to the size of a typicalobject which must be separately identified (Le. a resolution of a metre or les s isneeded to identify a single tree, while a resolution of 500 m to a kilometre maybe enough for studying ocean surface features). A necessary compromisebetween area to be covered and ground resolution of the measurement must takeinto account the volume of data which will be generated by, say, a large areacoverage - millions of km2 - using high resolution data - 10-30 m (see alsoFig. 16, page 60).

Finally, the frequency of the data acquisition must also match the natural fre-quency of change in the landscape. Daily observations may be needed for anassessment of plant evapotranspiration while observation once every year willbe needed for land cover change assessment,

H isto ry0 1R e m o t e S e n si ng

The first of what would now be calJecl airborne remotely sensed clata wereacquired from baJloons over Paris in the 1850's. The first pictures were more artand curiosities than data for scientific 01 ' resource management. Yet, by the late1870's, foresters in Germany had begun lO use aerial photographs taken fromballoons lO help rnanage timber harvests. By the 1920's, drarnatic increases in

21



Figur e 4 .Four view s0 1íh eoreo oround th e campus 01th eUniv ers ily 01C olilorn io S onl oB orbo ro : ( o )B l o c k o n dwh ile ponchromo ic p h o logrophy

the uses of aerial phorography began to occur as individuals with militaryexperience gained during World War 1 began to apply air photos to peace timeuses. Agriculture. water resources, urban and regional planning, civil engineer-ing, soil scienee, and mineral exploration applications were all pursued. By the1930's, colour aerial phorography was in use. and in the 1950's and 1960's wemoved beyond the visible ponion of the spcctrum to the precursors of the clcc-tronie sensor systerns of today. Indeed, the term remote sensing was coined inthe late 1950's because picture-Iike images that could not be called photographsanyrnore were beginning to come into wider use.

22

Figure 4I h ). B lock o ndwhite inlrored ph otogroph y

Those who have been involved in the application of rernote sensing technol-ogy over the years have watched the increasing cornplexity of sensor systems,data types and the analyses undertaken in ernploying these data. Figure 5 repre-sents an atternpt t o illustrate graphically the increasing cornplex ity associatedwith various aspects of the developmeru of remo te sensing. We have progressedto a point where we are employing m ul iis e n s o r data in solving cornplex prob-lerns associated with global surveys.

Ac(ess lo Remate Sensing Doto

The applications 01 global satellitc covcruge move forward rapidly. However.there is still no cornprchcnxive treaty 011dal;¡ g;lIhcril1g which has bccn r.uificd

23



F i g ure 4 (c).l-b and ac tive miu o w ove im agery

·24

F i g u r e4 (d) ,A e ro -m a g ne lic s u rv ey im a g er y

F ig ur e 5.

H is lo r ica l

dev el o p m e n l0 1

rem ole sensin g,em p h asiz ing

inueasing

com p l ex ily

0 1plntlcrrns,

sy slems

u nd losk s

• • • • ~ ~



• • • • ~ ~

by UN member nations regarding rernote sensing from space. The open skiesproposal first put forward by the Lnited States to the Soviet Union has neverbeen officially adopted. Today the United Nations operates under the 1986 reso-lution entitled: Principles on Rernote Sensing of the Earth from Space. Keypoints of this resolutíon include:

.••• Remote sensing ís defined as being for the purpose of improving naturalresources management, land use. and the protection of the environment, thisexcludes surveillance satellites from these principies;

.••• Worldwide collection and distribution of data is pemútted while ensuringthat sensed states have access to data of their country;

.••• Countries operating remo te sensing systems bear intematíonal responsibili-ties for their activities, irrespective of whether such actívities are carried for-ward by govemment agencies or prívate cornpanies; and

.••• Further cooperative activitie s to benefit as many countries as possible are

stronglyencouraged.In April 1991, the United Natíons went further. Legal aspects related to theapplicatíon of the principie that the exploration and utilization of outer spacewould be carried out for the benefit and in the interest of all states, taking intoparticular account the needs of peveloping countries, were put forward. .TheUnited Nations proposed that states should concentrate their efforts in the fol-lowing areas among others:.••• Continued exchange of information, data, materials and equipment on space

science a nd technology;.••• Promotion of easy and low-cost accessibility and availability of remote-

sensing data, the ground receiving stations and digital image processing sys-

terns; and.••• Technical cooperation to promote and facilitate the transfer of technology

and expertise in space scienceand rechnolog y, particularly with the develop-

ing countrics.lssues of data must certainly continué to be addressed with some urgency. Thisurgcncy is driven by the realization that a growing number of operational andrcsearch instruments are being planned and that potential users are increasinglyIac ed with difficult questions related to th echoice of data sources, the availabili-ty of ap propriate data, the COS¡of th e r eq uired coverage, and the av ailability ofthe data in a timely fashion. By th e end of this cenrury, we will be receiving ter-abytes of data daily from space. Indeed, if we examine only the approved andproposed United States missions to th e year 2002 these systems alone can

l. /\ terubyic (10 ) of data is the equivalent of o competer tape approxirnately 5000 km long, givcn1992 da ta storagc iechnology.

To bl e 2 . E x om p le s o f r u r re nt s ole lli te o nd s en so rsy s te m p a r am e te rs

A c ro ny m s f or Tob le 2

IF OV In stontoneous Field 01View (reso lu lionl

MS S Mu ltispectr ol S co nner

1 M 1 h em a tic M o p p er

AVH R R A d v onced V ery H ig hR e s olu lion Rad io meter

H R V-P Hig h Resol ution Visib le P onchrom olic

HR V·XS Multispectrol M ode 01SP OTSensorM ES SR Multispec tro l Electronic Sellsconning R o d i o m eter

S AR Synt hetic Aper ture Rod ar

produce data in the terabyte range (see Tables 2, 3 and 4 and Fig. 6). These datawill be of value to earth resource scientists, managers and policy-rnak ers as weauernpt toimprove our understanding 01' the factors that sustain life on the earth.They will, for example, help us measure key aunospheric constituenrs. map landcover, monitor the expansiOIl of agricultural land and model future we arh e r pat-ie rn s. Su ch clata can provide information which ma y improve our ma na gc me nt ofth c rcsources of our earth.

27



.' li~ h l n in ~ o c l iv i ty1; eS l i;;bie s 0 1precip i lo l ion• ~_ • • • 7 f

~I~)_~ ~. : ; j~ ~ ;J i5: f - ó ¡~M u lii-On g l e r o d i o n ce~ ; o l b ed o; o l m o s p h e r ic o e r o s o ls/.,,::& .:4 _ , , . , . ,,:. 1

Ver ti ca lp ro liles0 1e ao nd p oss ible C H-4

. F OJo n sí¡ fl.oc ew in d7peed'. . .

7 G H z d o lo u plin k c op obility

To b le 3. E O S· A S e ries Instruments. Th einst ru ments lis te d w e re pro p o s ed lor f1ig hton the E O S · A

seri es p lo tlo r m s

28

T o b le 4. E O S - B S e r i es E o rth S d en c e I ns lru m e n l s. T h e i n s lr u m e n ls l is l e d o r e c o n d i d o tes l o r f li g h t

i n t he E O S-Bseries

O t he r S pa tial D ata

T h e m e s

It is well bey ond the scope of th is paper lo provide a description of all spatialdata types which could be used in an analysis related to sustainable econ omicdevelopment. W hat follows are examples of more co rnrn only employed datatypes. For each of the following corn rnonly-used geo graphic data, a sarnple ofthe derivation of thedata is provided .

9



Elevo ÍonData about elevation may be acquired by making measurements at specifiedlocations using Global Positioning Satellite (GPS) receivers plus conventionalsurveying techniques. Advance in hardware and software for the satellite-basedGPSs have revolutionized the field of geodetic surveying and navigalion. lt isnow possible lO precisely determine X-, y-, and z-coordinates of any location onearth using a pocket-size GPS receiver. Cornbining digital image processing

techniques and GPS technology, remotely sensed information can be quicklyand accurately related to absolute .coordinare systems, thus creating improvedand efficient ways for creatíng topographic inforrnation. Elevation data can alsobe produced using photogrammetric techniques from overlapping pairs ofremotely sensed imagery. From these sets of original source data one may derivecontour lines for creating printed maps, and regular arrays of estimated eleva-tions through a mathematical interpolation operator. Interferometry usingmicrowave data is also a promising field of research.

Lo nd ove r, Ve ge lolio n o nd S oilsData about vegetation and soils may be created through fie ldsurveys. These sur-vey datasets may then be used as a sampling frame from which data on vegeta-tion and soil are extrapola ted across space by incorporating remotely sensedimagery. Biomass and primary productivity can be derived from such databas esusing appropriate models.

Lo ndU seLand use in this context means human interaction wíth the land. Land usesinclude transportatíon, structures, agriculture, urban (residentíal vs. recreation),utilities. Data conceming human use of the land surface may come from manysources, including tax records from a regional govemment agency, maps fromplanning agencies, engineering drawings and databases from civil engineers andarchitects, and aerial and satellite data coupled with or collected solely by field

surveys.

M e l eo ro logyData on rainfall, temperature, pressure, humidity and air quality (e.g. percentN02, S02' 03) may come from surface, airborne and space borne instruments,mathematically interpolated lO derive estimated values where measurements areunav ailable.

Fig u re 6 , r ig h t. Exomples 01 turren í so lellile m issions ond th ose b eing plon ned into th e next

cenlu ry

30

) ' - : :

l. Joinl wi th Itoly2 J oint with Fro nce

3. Russ ian s o tellite

4 J apa n ese s o l ellile

5. osedoto purchase6 . J oint with Jopon

b ~ e ~Y4 ~*. HPossib/e extended

1 m J IBm I m J I B I m JI m 1. Í l I f 1 Z J É I D t l f . m m mmE

I IProposed missionpproved,

u nd e r development,orop e ro ting missicn

m ission

Fu ture eorth probéca ndidotes indu h .1Aristoteles/Grofity ,MFE/Mogno/io TO PO M issions



D e m o graph icInfonnation on the characteristics of people, as a function of their loeation, is ,commonly derived from survey data sueh as a periodic census of the population.These are typically recorded as statistieal summaries, aggregated over variousspatial scales from pareel to eensus traer to city, county, state, nation and globe.

C a d a stralProperty ownership. Information about land parcels and their boundaries andcharacteristics is typically gathered using conventional surveying practices andcompiled inanalog or digital map format as well as documents created whenproperties are sold or exchanged.

Geolog icLithology of formation and fossil records along with infonnation or structuresare often measured at a small number of locations, and spatiaI pattems in thesedata are then inferred from ancillary data such as remote sensing observations.

Hydro logic

Drainage patterns, flow rates, size of catchments, depth to groundwater.Hydrological data on drainage pattems, flow rates and catchment basins areoften derived from ground observations coupled with an analysis of the spatialpattems of lithology and elevation either from existing maps or using remotelysensed data. Data on groundwater are frequently measured at a smalJ number oflocations and generalized in space based on correlative infonnation.

Form a ts a nd S tr uc tu res

For any set of spatial data, there is no single way to record the data values and

their variation through space and time. For example, elevation in a region eouldbe recorded as:T Contour lines on a map - estimates of the set of loeations which correspond

to a predetermined list of elevations above daturnT Entries in a cartographic database - digital data which is designed to store

elevation data in a manner which may be used to create prínted rnaps

F ig ure 7, right. Ele volio ns on o reg u l ar g ríd . T h e u p per p ortion 01the lig ure sh ow s o se r01con lour lin e s I ro mo m op, in whichloceíions olong o gív en line a re the som eelev o lion ob ove

seo level. T he orroy 01n u m bers i n t he low er por tíon 0 1Ih e l igure rep r esen l íhe est imoled

elevolion 01 the centre 01o set 0 1s qu ore rosler rells,

3

2 5

26

2 7

2 8

2 9

3 031

32

3 334

3 536

37

38

3 9

40

4142

43

44

45

4 6

4 748



.•••Values in a raster array - a regular array of adjacent cel s where a representa-tive elevation value is stored for each cel ; one imagines a sheet of graphpaper placed on the earth surface and the elevation at the centre of each celin the sheet is recorded (see Fig. 7)

.••• An irregular network - the shape of the earth surface is captured by creatingan irregular set of 2-dimensional facets located in 3-dimensional space;where the terrain has linle curvature the facets may be large.

The following section provides additional details of some of these alternatives.In general, printed data suchas maps and photographs may be con verted to

digital databases, albeit with significant costs if high precision is required. Becautious of cartographic generalization: amap is not reality but a cartographer'srepresentation of reality. Indeed map accuracy and the accuracy which resultsfrom integrating or deriving inforrnation from the overlay of various maps andor ot her spatial data types of either known or unknown accuracy is a very activeresearch area with many unresolved questions. However, it should be noted thatmany users rank currency before accuracy when using maps of variousresources. This is often done with the presumption that standard cartographicpractice wil be fol owed in creating the map product, and that since use of suchproducts represents the current state-of-the-practice , decisions made and

actions taken wil l conforrn to accepted procedures.

3 4

T E C H N O L O G I E S

{¡t,

In the last decade an important set o f new technologies tohandle spatial data hasdeveloped and matured. Geographic Information Systems (GIS) are an impor-tant example of these technologies. The emergence of this suite of new tech-nologies has been made possible by rapid progress in advances in disparatefields. Tbese advances include:.• •• the massive i ncrease in the power of computer hardware at a given pri ce;

.••• the rapid deveIopment in digital spatial data col ection technologies, data-bases, and i nformation systems;

.••• the significant developments in the techniques of spatial dat a modeling andanalysis; and,

.••• the emergence of highly real is ti c forms of visualization for spatially distrib-uted phenomena.

Today the technologies resulting from this decade of rapid development offer ameans to encapsulate problems with a spatial dimension in models which canrepresent, as never before, much of the real world's complexity and d iversity.These models can then be populated with spatial data so as to incorporate spatialinter-relationships and dynamic behaviour. Such models, once vaJidated, canthen be explored by the use of analysis tools. These systems are already beingwidely used to address many of the challenges associated with finding the opti-mum path towards sustainable dcvelopmern: however, we believe many moreexciting applications for t hese systems exist.

The material which follows briefly introduces these key technologies, con-cisely profiling their procedures and potential.

Geo graphic In formati on Systems

Certainly one of the most imporuuu technologies for handl ing cnvironmental

35



data, the Geographic Inforrnation Systern (GIS), is now highly developed andthere is a sophisticated rnarket-place where al present over 100 commercialproducts compete. The last few years have seen an enormous investment in GISworld-wide, ranging from systerns which cost less than $200 to those which areimplernented on large computers and are accessible via networking across wholeorganizations. GIS can be defined as Cornputer systerns for integrating and ana-Iyzing spatial information in a decision-making context

GIS have been described as the biggest step forward since the introductionof the map , The map is the customary means of communication for observa-tions and analyses of spatial phenomena. The paper map is, however, by defini-tion, a static and interpreted record of a particular phenomena at a given timeand place. As such, the traditional map is a useful but limited tool for communi-cating spatial information. Typically, map users employ paper maps as a back-drop for plotting other kinds of information. As users of spatial data have cometo make maps by computers, a shift in the role of the map has become apparent.When stored in a computer, maps take on a new and dynarnic form: their con-tent, symbolization and the area covered can now be changed relatively easilyand rapidly. For the first time, users of spatial data have truly flexible, interactive

access to their data, any part of which can be updated in real time.It can be argued therefore, that such maps formed within a GIS are better

thought of as databases, or collections of information linked by their spatial rela-tionships. In this conception a map can be seen simply as one unique expressionof a spatial database, for example, showing the area-by-area values of a spatiallyvarying variable where the categories chosen refIect a spatial model of somekind. In this new spatial data computing environment, the GIS offers new poten-tial for spatial analysis, and is a much more efficient place to s to re spatialinformation.

The functions of a comprehensive GIS can be traced to a range of early spa-tial data handling systems which evolved in several related fields such asAutomated Cartography (CARTO), Computer Aided Design (CAD), Data BaseManagement Systerns (DBMS). and Image Processing (IMPROC).

A comprehensive GIS today can be defined as the union of the functicnalityof each of these systems to form a true multi-function, multi-purpose spatial datahandling system. The figure (above right) can be used to visualize this develop-ment, and lO chart the origin and strengths of particular software package s soldas GIS. Depending on the functionality offered, most packages can be plotted onthis diagram : many will plot in the overlap of CAD and CARTO, or IMPROC,DBMS and CARTO, but Icw can actually be placed near ih e central GIS catego-ry when defined in this way (see Fig. 8).

To complete a description 01 the components of a GIS, it is al so necessary tomention: hardware, spaiial data ano [he necessary human skiIls and management

required lO operaré the system. Estirnates of the typical costs of a GIS projccr

R g ure 8 . V enn d i og ro m of over lo p sin sp o liol dolo processing lech nologies

indicate that hardware and software costs only account for around 20% of totalsystem implementation costs, whereas the staff and data costs associated withdatabase development actually amount to 80% of the total. This implies that thecost of the data and the expertise required to run a GIS are a key part of projectcosting; and that data and expertise are thus the most irnportant assets for anorganization to protect in the long termo

AII uses of a GIS should begin with a model of the phe no me na under study.These data models may be specified in various ways, for example, as a set ofpoint samples of reality approximated by interlocking regions (whose internalvariation may be described by simple functions); or as a se r of identified spa tia l

objects which litter the plane. The data models used by GIS are usuaIly phe-nornenon-based using the features in, or derived frorn, the subjects under studyto specify the model used to represent thern. The best exarnple of such a datamode l which uses the object view is the one c1erivecl from a topog ra phic map ,

where a set of features on the lanclscape sueh as the coast. rivers, roads andbuildings are idemified and classified by convention.

To create a cornputer representation 01' this model. georncrric prirnitives suchas points, lines and arcas are typically uscd as thc building blocks for thecreation 01' spatial objects. Such reprcsentations are crcated by first measuringlocations using coordinates based on a spatial rcfercncing sysiem: the simples:Iorrn is analogous to an x,y graph. This is known as mctric inlormation and is

srorcd in the Iorrn 01' characters, nurnbers and boolcan valucs (true or [a lxc).

,HB L 1 OfE C A D ES Y W SD EL A U I.C H3 6 3 7



since the cornputer cannot handle spatial data types such as lines. To completethe representation, the spatial components of the model must be ordered andindexed to encode spatial inter-relationships (referred to as topological informa-tion): this is known as a 'data structure'. The two alternative methods which aremost often used to create data structures, are known as the vector and rasterapproaches.

Vector data structures employ the principies of coordinate fixing as in navi-gation. A point can be described by its distance along each of two axes of mea-surement, e.g. north and east, while a line can be described as the shorteststraight line distance between two such points - a vector in the mathematicalsense. Areas can be represented by any number of lines joined together in a loopand arranged to reftect the specific shape of the area. In all cases, the only infor-rnation needed is the coordinates in number form, which can easily be stored inthe computer and recalled in the right order when needed.

Raster data structures, however, use a different approach altogether, usingregular building blocks to build up the shape of a feature on a grid. Using such agrid of very small squares, a point, line or area can be approximated by sets ofsquares or pixels, the resolution of the rep resentation being directly controlled

by the size of the grid squares. In a raster data structure, the computer keeps arecord of which pixels have which values and therefore form part of one of thespatial phenomena being represented.

Using one or another of these data structures allows the storage of spatialphenomena in the computer. While raster based systems typically allow formore effi cient data storage and retrieval, vector systems offer the potential formore specificity in location when precise coordinates are included. To completethe geographical data base requires the addition of a unique ident ifi er to all thepoints, lines and areas represented so that they al can be linked LO attribute datasuch as a name or other non-spatial data stored in relation to the spatial data.Many GIS currently in use permit a user to group all geometric and attributedata together for a single subject of study in a single theme such as 'transport' or

'pollurion'. This can be implemented as a set of thematic layers in the database(see Fig. 9). The cornpleted daiabase is an integrated spa tial and non-spatialinformation systern: however, its potential can only be unlocked by applyingspatial tools which operate on this structured spatial data.

Functions av ailablc in a typical GTS can easily be frarned by the sequence ofoperations carried out in the crcation and analysis of a geographical database.Hence typical operarions would include querying the spatial database for givenloca tion s or Ieatures mee ting ,1 particular cr ite ria; analyzing the characteristics ofa surface or network: or the intcgr.uion of spatial data frorn different sources.These ope ra rions o ffe r key ud va rua ge s in the use of a GIS as they enable theuser to examine a model of sornc spatial pheno rne na and gain new insights.

38

F i g u re 9. Age o g r a p hi( inlorm ofio n sy slem « m b eroncep í u oliz ed o s o b ase m o p b y n um erou s

regis tered overloy s. lnlo r m otio n inth ed ata base wos derived Ir o m avoriety 0 1s ources

I m o ge P r o cessin g

While images frorn remote sensing sysrems on platforms from aircraft to satel-lites rnay be fascinating, they ha ve 1 0 be processecl and analyzed to acq uire suit-ab1e information for applications such as monitoring the paihs of scvere storms.estirnating agricultura1 yields, ancl ihe prcdicrion 01 runolT Irorn mo unta in snowpacks. We have moved frorn a reliance on airborne or more reccntly satelliteda ta alone ro a cco mplis h an objcc tive LO thc use 01' me thodolog ics which rcquireincorporati ng rernorely sensccl data a nd ma ny othe r iypcs 01' information in sys-

rerns which can examine quesiions al sc a lcs frorn loc a l lo global. The use 01'

geographic inforrnntion sysrems has grcatly Iaciliuucd rhis trend as sccn in

3 9



Figure 10. In this progression we have also moved from using remotely senseddata for simple identification of forest species to their application in complexproblem solving such as their use in models to estimate runoff from given water-sheds within specified time intervals. and the use of this information in the pro-jection of available hydropower generator potential (see Fig. 5, page 25).

In the pas t, aerial and space photography was primarily analyzed throughhuman photo-interpretation. However, progress in digital data acquisition andcomputing technologies has led to computer-assisted and interactive digitalimage analysis systems. As in many ot~er areas, remote sensing applicationshave benefitted from the explosion of computing capabilities in the past twodecades. Where earlier systems required room s full of computers, technicalstaff, and custom development of hardware and software, desktop computersand commerciaJ software are now capable of serious project work. Where earliersystems required years of research and development to tailor new tools for users,there are now practical tools for a range of applications.

Digital image processing systems can be seen as an example of a raster-based system where the remo te sensing image is tiled into small (normallysquare) picture elements or pixels. lmage processing techniques are used in alarge number ofapplication areas such as industrial fabrication and control, sur-

veillance systems, television, archiving, and desktop publishing. Specific tech-niques to process and analyze remotely sensed images include image enhance-ments (i.e. filtering to smooth images, or contrast stretching), georeferencing andrectification (i.e. registering rernotely sensed data with a geographic coordinatesystem), automated classification and feature extraction for larid usemapping/land co~er mapping, and change detection and habitat analysis.

In their early stages, digital image processing techniques were primarily con-cerned with single irnage evaluarion and local applications, and relied to a largedegree 011 algorithms based on statistical and mathernatical algorithms or visualanalysis. Recently, these techniques have been extended to address topics such asmultisensor, multitemporal and multispatial image analyses, integration with geo-graphic information systems, interfacing geoscientific and socio-economic modeling,

and the use of expert system technology for knowledge guided image interpretation.For example, aerial photographs have been in operational use for decades to

inventory and manage resources in forestry applications. With the advent of clig-ital photogrammetry. orthoimage maps in digital formal are increasingly beingused for this purpose. For countries Iike Callada or Brazil, there is no other waythan to rely almost cornpletely on remote sensing and digital image processingtechn iques lo rout ine ly monitor ancl manage the i r vast forested areas.Monitoring and quantifying tropical deforestation on a global scale is only nowpractica with the availability of modern remate sensing techniques. Recently.colour infrareclairphotos and scanner imagcs have also been used to analyze anclassess vegeration stress. cspecially for arcas affecicd by acid rain.

~( M AtlAGEME NT

F i g u r e10. Diog rom01o GIS os o d erision su p p orf s ys fe m

40 41



The critical importance of rernote sensing is of course not restricted to onespecific application area . The ability to use digital image processing techniquesto extraer rneasurernents (and as such. spatial inforrnation) frorn remotely senseddata has been used in a number of fields concerned with rnapping, monitoring,modeling and managing our environment (see Fig. 10). For exarnple, usingsophisticated atmospheric correction models, researchers have been able torelate digital pixel values to energy received from the sun at specific wave-lengths at a specific ground location. This has helped users to irnprove their

models describing heat flux at the boundary layer between earth and atrnosphere.How else if not with synoptic rernote sensing image data could atmospheric

physicists have detected the Antarctic ozone hole or could oceanographersdetect and analyze sea surface currents and eddies to assess and refine oceanicglobal circulation models. It has been shown that ocean colour values as record-ed on the Coastal Zone Color Scanner (CZCS) on Nimbus can be correlatedwith biomass and consequently provide productivity figures for the upper oceanlayers.

Using SPOT and Landsat satellite data, GPS measurements, and digital pho-togrammetric and image processing techniques, digital image maps can be pro-duced for desert areas in the Sahara. For the first time up-to-date, detailed, andaccurate maps are available for these areas - a basic requirement for planningand decision-rnaking.

Expert systems are also being developed 1 0 improve accuracies of landuselland cover maps. Whereas earlier image analysis techniques were almostsolely based on statistical analysis of spectral values at every pixel position,newer developments take into account parameters such as neighbourhood infor-mation, texture parameters, or surface shape analysis: researchers are also tryingto formalize human interpretation skills into rule-based expert systems in anattempt toimprove our ability to extract information from remotely sensed data.

Whereas some of the techniques and interfaces are still in the research anddevelopment stage, these interdisciplinary approaches already provide promisingtools Ior efficient storage, query, information extraction and integration of

remotely sensed data (see Fig. I Il.

Fig u re 1 1 . C o mb ining rem ole se nsing ond G I S le ch no lo gy in lo so-co l led im oge-

i nl eg r ol ed G I S (IG I S l

M o deli ng

A model is a caricature of reality that emphasizes those parts of a system ofmost relevance to the user while ignoring those components of less importance.The skill in model building is to decide which detail is essential and which canbe omitted.

Models can take many forrns. Physical models are widely used in areas ofengineering such as fluid dynamics. However, the main types of models relevantto sustainable development are analytical models (i.e.a series of mathematicalequations that can be sol ved) and sirnulations (i.e. a series of functionsexpressed as a computer program). Analytical models often oversimplify thesystern being modeled to achieve a formal rnathernatical representation, but havethe advantage that their behaviour can be efficiently and comprehensivelydescribed. Simulation rnodels can capture the cornplexity of many real systerns

bUIalways a l the risk that a large cornplex systern will be replaced with a largecomplex model - neither of which can be understood.Models are built for a variety 01' purposes. Many are built as experimental

tools that explore cornplex hypotheses about the mechanistic basis of systerns.These models are probably bette r never released from the laboratory, since theyoíten lead to Irusuation anel disillusionrnent when planners and decision-rnakersattempt ro use them in practice. Here we are more interested in modcls that canbe populated with spatial data and used lO explore the cornplex spatinl amidynamic temporal behaviour 01' systerns.

Most models used in resource and environmcntal management are dynamicmodels. Dynamic rnodels folIow the chunge ihrough lime of particular variables

Humans havc always used models. The first were physical models, ranging frorndolIs la model boats and gliders. The garne i-go (wei-hai¡ originated in Chinabefore 600 BC :15 a simularion of the tactics 01' warfare. and in the Thirteenthceruury AD data frorn ruin and snow gauges were being conveyed to the con-trollers of river systerns in China to help them preelict flood peaks.

4 43



describing a point or points in space or an aggregated area. Examples of dynarn-ic models include models of the yield of a crop, the growth and populationdynamics of a forest stand, or the production of a herd of animals grazing aspecified field.

A few models predict the changes of variables in space. These are often usedin association with spatial databases to predict the values of variables that aredifficult to measure by other means. Gradient modeling is perhaps the bestknown example of these approaches. In gradient modeling the distribution ofspecies or community properties is correlated directly with more readily measur-able descriptors such as altitude, aspect, geological substrate and the like. Thesecan then be used to predict among other things the potential distribution of acrop or species of natural vegetation.

There have been very few attempts to combine both model forrns of predic-tion in time and space. The computational demands of such spatío-temporalmodels were once asevere limitation. However, computing power has increasedand more efficient computational algorithms are being developed. Someprogress is also being made with the automatic simplification of models. Someforms of models can also be used efficiently in standard analytical and optimiza-tion packages, thus allowing the user to examine a wide range of options inseeking better management strategies.

An important l imitation in developing spatio-temporal models is our inexpe-rience in coupling subsystems of different spatial scales. We have too littleexperience with this c1ass of models to be certain just which details to includeand which to leave out. Mistakes can lead to cumulative errors and spuriouschaotic behaviour.

A number of factors have limited the more widespread application of modelsin the management of natural resources. Most models are built for a particulartask, often with the mixed goals of prediction and exploration (or explanation)of the system itself. This means that they are often curnbersorne to run and areusable only by the original development team. These problems are being over-

come. The tasks required of models are being stated more clearly (e.g. predic-tion of crop yields, structural changes in vegetation, land suitability). Successfulgeneric models are being developed (e.g. the crop trajectory models of Hall andBadhwar, the JABOW A/FORET suite of forest growth models, BIOCLIM forpredicting species ranges) and are currently available in well tested softwarepacka ges, Also a largor cadre of biologists, agriculturalists, planners and otherusers with skills in modeling and computing is emerging each day. The users arepushing for more development and better models. This bodes wel for furtheradvances in modeling.

E x pe rt S ys te m s

Expert systerns (sornetirnes called knowledge based systerns or advisory sys-tems) are among the most prominent products of artificial intelligence research.An expert system is a computer program designed to sirnulate the problem solv-ing ability of an experto The program asks questions and, based on the responsesof the user, eventually reaches, or validates, a oonclusíon and is able to explainhow and why it reached that conclusion.

The first expert systerns were developed in the late 1960' S . Many of theseearly expert systems were designed 1 0 assist in diagnosis of diseases or in identi-fication or c1assification. A very large literature on the methodology of develop-ing expert systerns has been created and there are numerous software packagesto assist in the task of building and using expert systems. Much of the theorybehind expert systems draws upon the larger debate about artificial intelligence;however, questions about how humans think or whether a particular programdoes show true intelligence are not of prime importance in this area. The goalin developing expert systems should be to design and implement useful advisorysysterns. What may be of importance are the modes of logic and explanation thatpeople find intuitive so that expert systerns can more readily comrnunicate withtheir users.

The construction of an expert system is a time-consuming process both for thebuilder (sornetimes called the knowledge engineer) and the experts whose experi-ence is being captured. A cornmon fallacy runs, We don't know enough aboutthis topic to build a real model so let's build an expert system . This can be thestart of a wasteful process. Expert systems should only be built when humanexpertise is scarce but available. This expertise may be in the forrn of a series ofrules or ernpirical relationships, as long as the knowledge encapsulated in theserules or relationships can be cornmunicated in text or pictures. The experts musthave the time and willingness to cooperate in the venture. Expert systerns workbest in diagnostic or classificatory tasks, especially where the understanding of the

task is empirical and the input data are noisy . They also work most effectivelywithin a wel prescribed dornain. It is very difficult to capture all the pieces ofknowledge that go to make up cornmon sense in a widely defined task.

A promising developrnenr in the building of expert systerns is the applicationof the techniques of rnachine learning or rule induction. In developing a quanti-tative model, a scientisr may use statistical software to unravel relations andderive mathematical functions frorn data. Machine learning plays a similar rolein developing expert systems. New algorithms and software packages arebecoming available to assist in extracting information Irom large data sets. Thealgorithms are especially suitablc for handling qualitative relationships betweenvariables and discontinuities within the data. The information is exprcssed in the

forrn of rules 01' decision trees.

4 4 4 5



The field of advisory systerns has merged with the more general problern ofdeveloping systerns of knowledge handling and delivery. This is becorningincreasingly important as the quantity of information and the cornplexity of itsinterconnectedness increases. Even when information is available electronically.simple searching approaches can still be slow and unsuccessful. The main prod-uct in this area is better structured knowledge bases and intelligent retrievalsysterns available on dernand and on-site.

Another related area is rule based models and qualitative modeling. Anexpert systern can be thought of as a model cornposed of rules and these rulesare often qualitative (e.g. if the amount of monoculture cropping in the dislrict ishigher than the norrn for the region, then ...). This is a major area of applicationin rernote sensing where rule based models may assist in the initial c1assificationof irnages (e.g. if the ratio of Landsat Thematic Mapper bands three and two islow, then the land cover class is water).

Another major application of expert systems is as advisory systems assistingusers as they attempt to master the force of the very powerful, but complex,tools provided in a comprehensive GIS (see Fig. 11). Similarly, expert systemscan be employed to guide users in the intricacies of reformulating generic mod-els for their particular application, in running the models and in interpreting theirmodels outputs.

The fusion of remote sensing and GIS technology into so called image-inte-grated GIS (IGIS) is especially seen as one of the key research fields in spatialdata handling over the next decade. National and intemational research pro-grammes are underway to study the degree to which these two technologies canbe combined. It is important that advanced expert systems be developed whichcan aid users as the anernpt to understand both the potentials and the pitfallsinherent in the integration of these complex technologies.

AP P L I C AT I O N SA N D E X A M P lE S

Each of the technologies discussed in this digest - remote sensing, geographicinformation systems and environmental modeling - has a variety of currentapplications both within the research as well as the operational resource man-agement communities. More and more often we find, however, that they arecombined togreater or lesser degrees depending upon the type ofanalysis beingconducted. Results are available on the use of remote sensing to provide mea-surements of environmental parameters such as net primary productivity, atrnos-pheric concentrations of ozone, or forest stand densities. The use of GeographicInformation Systems for facilities management or locating facilities is also read-ily documented, as is the use of modeling for the prediction of the economicimpacts of development. We have chosen to stress applications that bridge thesetechnologies, in providing a perspective on how integration can achieve syner-gisms that facilitate our understanding and improve our ability to manage ourenvironmental resources.

The four exarnples discussed here are drawn frorn four different regions and

concern:'Y Land-Use Planning in Senegal:'Y Crop Yield Estimation in lraly:'Y Habita Analysis for Preservation of Species Richness in the United Stares; andV Deterrninarion of Natural Hazards Risk to Agricultural Production in

Ecuador.The first and fourth application exarnples are based on work undertaken withinthe Monitoring of Tropical Ve getarion and Tropical Ecosystern EnvironmentMonitoring by Satellitc (TREES) projects of the Joint Research Centre of theCommission of the Europea» Cornmunities (lspra. lra ly). The second exarnple isa joiut projcci of (he Deparuncnt of Agromereorolog y of the Region of the

Vene ro, ltaly, ami (he Uni versity of California, Santa Barbara. The third

4 6



example is based on cooperation among a range of govemmental (Federal andState) and non-governrnental agencies in the United States'.

R esul ts

L on d-U s e P lc nn in g in S en eg cl

The analysis concluded that Senegal could significantly increase its productionof rain-fed cereals, without exrensive costs in terms of natural resources, byapplying improved technologies and by expanding cultivation. Specific findingsincluded:T Existing rain-fed cereals alone could sustain 3.9 million people (56% of the

Senegalese population).T By applying improved agricultural technologies, 5.4 million people (79% of

the current population and 37% of the country's projected population for theyear 2010) could be sustained from these crops.

T By applying improved agriculture technology and expanding cultivation topotentially arable lands, approximately 7.7 million people (more than thecurrent population and some 56% of the projected population for the year2010) could be sustained, while preserving some natural vegetation.

This information was used to def ine two strategic development goals: (1) sus-tained efforts in family planning; and (2) an increase in crop productivitythrough the use of n ew technology in zones of reliable rainfall.

A number of otherAfrican countrie s facethe same challenge ofincreasingfood production to feed growing populations. In many cases, the naturalresource and socio-economic data needed to carry out GIS analyses similar tothe Senegal study are available.

lntrod uttion

Senegal is experiencing a long-term decline in per capita food production as popula-tion growth out-paces agricultural growth. In 1988, Senegal's population was esti-mated to be 6.9 million. Population is likelyto double by 2010. Planners in Senegalface tough land-use decisions. They must weigh the need toincrease food productionagainst costs associated with the loss of natural environments to cultivated lands.

A p p r o o c h

In 1990, the US Agency for International Development, in contact withSenegalese specialists, sought information on options for increasing rain-fed cere-al production in Senegal. The EROS Data Center of theUS Geological Surveyperformed the entire analysis on desk-top and mainframe computers. A GIS wasused because of the complex spatial component of the data, the need to integratenumerous natural resource andsocio-econornic data layers, and the need to ana-lyze a wide range of production scenarios , wi th the Iikelihood that these analyseswould be repeated in future years. The major cost involved was the time analystsspent entering data and maps defined bythe GIS model. All of the data came fromexisting sources, including maps prepared from LANDSAT remotely sensedimages, Senegal's national census, c limate data and reports. The GIS containeddata on natural resources, clirnate, demographics, poli tic al subdivisions, infra-structure, agricultural production and nutrition. Natural resources data included

information on soils. vegetation, land-use and water. Agriculture data covered dis-tribution of cropland, major crops, yields and production figures by crop, thecaloric value of the major food crops and land-use rotat ion practices.

E stim ctin g rop Y ie ld s in Ita ly

In t roduc tion

The Po River Valley of northern Italy is the breadbasket of the country, withcrop production in the area exceeding US$ 20 billion per year (see Figs. 12

and 13). The Po River itself forms the southern boundary of the Regione delVeneto (RdV) before it flows into the Adriatic Sea. In 1985, RdV officialsbecame convinced that they would need improved methods for compiling andanalyzing agricultural statistics for the Regione in order to make effective agri-cultural and economic policy decisions in an increasingly complex ecologicaland economic environment._ The exarnples which tollow in this section draw on material comp iled by the authors for UNESCO

Environment and Devclopmcnt Br ief 3. This bri ef e ntitled New Technologies: Remate Sensíng andGeogrophic tnfonnation Systems, \\ :1S pr odu ce d as 3 joint e ffort by UNESCO and the ScicnrificCommitte e on Problerns of the Environment (SCOPE ). Sorne support of p ersonnel wa s a ls o providcd

by the US Nauonnl Acronautics an d Space Adm inistrations Earth S c ien ce a n d A pp lica tions

División. The l ó-page colourcd brief waspublished in m id 199 2, and is available in English, Frenchand Spanish fro rn rhe Bureau Ior thc Co ord ina tion of Environrnent Prograrnmes, UNES CO 7 p lacede Fontcnoy. 7535 2 Pan, 07 SP (France }.

A p p r o Q (h

In 1987, RdV officials institutecl a co-operative research programme with the

4 8 4 9



Vene to oreo : 1 836 8 19 ha

cartographic data and environmen.al data frorn many other sources as input into aGIS. To produce the crop estimaies for the RdV, a mínimum of four LANDSATTM image acquisitions are required per year. The project has made use of oorn-mercially available rernote sensing and GIS technology. The initial investment forproject hardware, software and data input has been in the order of US$ 45,000.

Extensively processed LAl\DSA T TM data are used to provide primaryinformation on the five most proritable crops grown in the RdV: com, soybeans,sugar-beet, small grains and orchards. The systern al so uses meteorological,agronomic, economic and field validation data. The collection of ground truthdata represents a signi ficant cost factor in the project.

University of California, Santa Barbara, to develop better crop productionestimates. The systern that was developed uses digital satellite image data,

Fig u r e12. V ene to r eq io n o ndits p rovinces Fig ure 1 3. Lo nd u s eb yprov ince

.... g ~ J fWS D E50 51



Resu l ts Gap analysis places resource managers in a proactive rather than reactivemode in the preservation of biological diversity. Gap analysis builds from aprernise that unprotected species and cornmunities represent gaps in our con-servation safety net.

When completed, the system \ViIIprovide estirnates of the total agricultural areawithin the RdV that is under cultivation for each of the five major crops. Theseestimates will be available on a biweekly or monthly basis at 75-95% accuracy.

The information wil be used on a regular basis to upgrade region-wide agri-cultural production forecasts which will become increasingly important as rnem-bers of the European Community develop closer economic co-operation. In

addition, RdV officials will use the system 10 provide farmers with tirnely infor-mation on crop production, yields and profitability.Environmental engineers and scientists working with the RdV wil also ben-

efit from ready access to accurate spatially referenced information on chemicalherbicide, pesticide and fertilizer use within the region. Forecasts of ag~iculturalproduction are a key economic indicator. GIS using remote sensing data foragricultural monitoring are being developed in a number of countries, includingItaly and Egypt, and for regional areas such as the European Community.

In the future, systems Jike that developed for the RdV are likely to becomernuch more widely used. Their accuracy should continue to increase with experi-ence and as the models they employ are fine-tuned to local conditions.

Approoc h

l n t red u d ien

\York in Hawaii demonstrates the potential of gap analysis in assisting decision-makers in their efforts to preserve diversity. This work involved cooperation byt he United States Fish and Wildlife Service (USFWS) and the HawaiiDepartrnent of Land and Natural Resources. Other cooperating groups includedprivate citizens, the Sierra Club, Hawaii Audubon Society, ConservationCouncil, Wildlife Federation, and the Nature Conservancy.

In the sumrner of 1975, ornithologists in Hawaii were asked to identify thebest area in the islands for an endangered forest bird reserve. Hawaiian ornithol-ogists could not provide an adequate answer that would convince decision-mak-ers to cornmit funds for the acquisition of the preserve. While there were severallocations where endangered species were known to occur, there were far larger

areas where there was no species distribution information.To correct this situation, the Hawaii Forest bird survey was begun in 1976.

During the survey, 538 biologists and technicians surveyed 9940 stations andcounted nearly 135,000 birds. These studies wereconducted along 1401 km oftransects systematical y distributed over the forested areas of five islands. Inassessing the need for new preserves, the first step was to plot species richnessfor endangered forest birds on each island. The next step was lo overlay on thisbase map a map of existing reserves (see Fig. 14). In doing this simple GISanalysis it was obvious that despite the large surface areas set aside on bothMaui and Hawaii, there was little overlap with the ranges of endangered species.Prior lo establishing reserve boundaries, research and rnanagemenr biologists

carefully reviewed the floral and faunal distribution and abundance data.Reserve designs were developed and forwarded lO the US Fish and WildlifeService personnel, based on a nurnber of factors includinz: the distribuiion ofcurrent and potential habitats, future land-uses inthe gener¡¡ region. informationon rhe prevalence and abundance of suspected limiting factors, and the need fora minimum viable size for the species of concern.

Biod iver s ity /G a p A n aly ses in th e U n ite d S t a te s

Loss of species has become an issue attracting world-wide attention. Manytimes, however, we have focused our efforts on individual, highly visible flag-ship species, possibly at the expense of broader issues. Researchers andresource managers are coming to believe that it is more effective and cost effi-cient to focus on the preservation of intact functioning ecosystems with theirmyriad species than to continue to practise ernergency room conservation onone endangered species after another or wait until common species becorneendangered before acting to protect them.

A challenge to the preservation of biodiversity is to plan future patterns ofgrowth and rnodifications in land use to insure the survival of remaining biologi-cal diversity. This goal will not necessarily be reached by rescuing specificendangercd species but by keeping enough of the living world around to supplyus with disease-resistant strains of crops. new medicines to fight disease andfunctioning watersheds that supply water for drinking and for irrigation. Gapanalysis is esseruially a GIS based approach to facilitate the achievement 01'

these objectives.

Resu lts

To date tive reserves have been established on various islands at a total cost 01'

more than forty mili ion dollars. An cxample frorn the big islands of Hawaii is

55 3



illustrative of the process. After analysis of the map of species richness overlainby the existing reserves, it was found that the largest nurnbers and highestdensities of endangered forest birds occurred in the Hakalau Forest, in an area ofprivately owned parcel on the mid elevation slopes of Mauna Kea. Densities ofthe more common birds at Hakalau were arnong the highest in the state. Thearea also included the habitar of the endangered Hawaii hoary bat and a numberof rare and endemic plants including several species considered for the endan-gered list.

After a thorough review of the project and a complete public review of theproposed refuge boundaries and alternatives for accomplishing habitar protec-tion goals, the director of USF\VS approved the acquisition of some 113,360 haof privately owned land in the Hakalau region. Individuals associated with thisproject credit the graphic depictions presented in maps (such as that shown inFig. 14), for helping to convince their management of the need for new reserves.

Such reserves are needed in many areas of the world. Wise use of the prod-ucts of gap analysis is one way we may be able to reduce the projected losses inplant and animal species. By the year 2000, every state in the United States isnow scheduled to complete a statewide gap analysis.

Risk Assessmen t in Ecuad o r

In t roduction

Ecuador's agricultural sector, which provides a high percentage of the country'sGross National Product (GNP), is vulnerable to natural hazards, particularlyfloods and droughts. During 1982-83, ñoods caused losses of more than US$110 million in damage to crops and the coumry's agricultural infrastructure.

By 1990, Ecuador's Ministry of Agriculrure had become increasingly con-cerned about the frequency and extent to which natural hazards were affecting

agricultural proe uction and relaied income, employment and investment.Although officials were aware of the problcm, they did not know which roadsand other facilities used for agricultural distribution were most vulnerable tohazards or where the government should implernent mitigation measures to pro-teet secroral incorne. cmploymeru and investrnent.

A p prooch

Frorn the outset, it was clear an autornatcd sysiem was requircd beca use of thesheer nurnber 0 1' variables and qu.uuitv 01' data to be analyzed. In 1991. with

5 4

Is lan d o f

Hawaii«; ~ ~

KOh: · ;~-tt.

o 10 2 0

Enda ngere dSpec ies Richn ess

1 Spec ies

2 Sp e cie s

113Sp ecies.4S pec ies......~L....1 1972 Pr e ser ve s

D 1987 Pre ser ve s

Hilo

3 0 km

Naale hu

F ig u r e14. Dist ribution 01th e end ongered lores t b ird s ond pres e r v es on the Islond 0 1Ho w o ii

in 1 9 87, s h owin g th e exten t 01th e preserves in1 9 7 2o nd th e newp reserv e o n th e eostern

slo pes 0 1Mou n o K eo

.55

N E E D S



I S S U E S , N E E D SA N D T R E,N D Ssupport frorn the Department of Regional Development and Environment andthe Organization of American States, the Government acquired a desk-top GISwork station and spreadsheet software to analyze data for Ecuador's 20provinces. For each province, the GIS cornbined data on 26 production systems,14 crops and related infrastructures - storage facilities, irrigation systems, roadnetworks - with data on droughts, erosion, volcanoes, f1oods, landslides and seis-mic hazards. Data on five economic indicators - income, employment, foreign

exchange earnings, investment and food security - were also introduced into thesystern.

R esu lts

The analysis performed on this database identified 49 critical situations thatrequired action to mitigate hazard vulnerability, including the following:T Erosion hazards were likely to damage a potato-growing area that, at the time of

the study, accounted for 43% of the national income frorn potato production, and40% of the ernployment and 80% of the income in one province (Carchi).

T Flood hazards in one province (Guayas), and erosion hazards in another

(Tungurahua), posed the greatest threat to agricultural employment in Ecuador.T Drought impacts on banana production in El Oro province posed the greatest

threat to Ecuador's foreign exchange earnings from agriculture.Based on th is information, the Ministry prepared a US$ 317,000 technical coop-eration proposal which several agencies are now evaluating to finance activitiesto reduce the risks to agricultural production from natural hazards. In addition,the Ministry forrnulated new investment policies and technical assistance activi-ties in the agricultural sector. Since natural hazards pose a major threat to theagricultural sector in many Latin American countries, analyses similar to thisone could be used throughout the region based on GIS technology.

A growing number of uses of the technologies discussed in this digest haveachieved either acceptance within various areas of the scientific community oroperational status within resource management agencies and the commercialarena. This does not mean that there are not issues yet to be addressed andresearch needed. The trends, however, are very positive and these technology

trends will, we feel, continue to enable even more scientific utilization of thesetechnologies. In this closing section, we address some of the key issues in thefuture development and application of data collection and information technolo-gies including the role of the scientist in the development and wise use of thesetechnologies.

Scales

Issues related to the use of information systems technologies for the studies ofsustainable development cannot be dissociated from the question of scale.Indeed, scientific analyses and policy making processes are intimately linkecl ro[he level of organization ro which they apply (see Fig. 15).

Concerns with the Earth systerns and with linkages between its parts requirethat analyses be made at global levels and that actions be taken under theurnbrella of international agreemenrs (i.e. Montreal Prorocol. [PCC, lGBP, etc.).Regional ques iions arising in the framework of transpon 01' water, gases, pcllu-tarns or of intercountry agreements with respecl lO resource exploitation or pop-ulation policies (rnultinationa] watershed plans, forest concessions, migration,etc.) are most often examinecl al supra-narional scales despire territorial divi-sions of auihority and irrespective of cornperences. Local needs for inforrnation

5756



about resources are more directly linked to well identified projects and the defin-ition of precise courses of action. At each scale, requirernents for data and infor-mation will be very specific. Clearly, if the examination of the factors thatenhance sustainable developrnent by their very nature integrates economic,political, institutional, legal, ethical and ecological considerations, then the rela-tive importance of each of those factors will vary according to the adopted per-spective. The perspectives of the individual - planners, resource managers, poli-

cy-makers - are coloured by the acculturation processes operating in the nationsand regions involved. The role of science and technology will, sirnilarly, assumea variable irnportance depending on local, narional, regional and globalimperatives.

The question of scale also bears heavily upon the resolution or grain of theanalysis and thus of the necessary data sets required to accomplish adequately agiven task. As a rule, local investigations will need finer resolution data than ascientist, resource manager, or policy maker involved in a global analysis.Furthermore, although rapid advances are being made in this area, technologyand limitations in our current ability to handle largeamounts of data still preventthe systematic application of higher resolution observations and analyses tolarge regions, continents, hemispheres and the entire globe. However, there maybe specific needs to dissociate scale and resolution when, for exarnple, a highlyaccurate rneasurernent of a specific chemical compound must be obtained over alarge area (ozone, pollutant, etc.). Thus, today, information technologies muststill be managed through a set of compromises between scale and detail, and uti-lize data structures and algorithms which facilitate flexibility in anaJyses.

A major advantage of the new information technologies is that the sametechniques and instrurnents can, within some limits, work at different scaleswhich can be nested into each other (see Fig. 16). This provides new opportu-nities for examining specific problems across scales ; a local analysis may thusbe easily set in a more regional context and, vice-versa, regional or continentalissues can be substantiated using linked local analyses. Together, GIS, remate

sensing and advanced modeling offer a means to extend significantly the breadthand depth of considerations brought into the decision-making process.

In fra structu re

lnfrasrructural investrnerus represent a key elernent in the successful adoption 01

new information technologies. Such investmenrs, which can be made al differentlevels 01' sophistication. are neee ed to Iacilitate access to the tools, networks,systems and concepts discussed in this digest. Again. apart Irom obvious finan-cial considcrations, the scale of concern and the nature 01 ' the problem to be

58

Ge omo rph ic \phen omen a

Forestma nagement

F i g ure 1 5. T h esp o t io l o n dtemporolsc oles 0 1vor iou senviron m e nt ol p h en om e n o

tackled wil govern the choice of technologies acquired by institut ions.

Essentially, investrnents in inforrnatics are nceded to acquire, display ane ana-Iyze data corning Irorn sources as varied as saiellite sensors, existing maps, fieldsurveys. sratistical tables .

The rapid advance 01' cornputing capabilities over the last two decae es hasgrcatly facilitnted access lO the new information technologies. Desktop cornput-crs and conunercial software are now available to support most types of projcciwork. High data ratc communications nctworks are becoming more cornmon.Yct, training, maintenance ancl updaring of the hardware and software requirecarelul considcration as they involve rccurring cosis not always rcadily availablein institutions. Until a Iew years ago, the balance of invcstments was hcavilyslarued towards rhe purchase 01 basic harclware and software. Data acquixitionand rraining now rcpresent the bulk of ihe necessary invcstmcrus involvcd in

5 9

1- '--

ACQUISITION



ACQUISITION PREPROCESSINGM PLEMENT ATION

flE'.Il IlUll lll1 llj~ SP ECIALlZEDPR OCESSING

Figure16. Neltedsroles0 1 o bservofien . Sompled esiq ns employingthisconceptore being usedtoeffectivelyextropolotelield dotoorrossspo re

Meosuremenl

P reeision 01 rn eosurementDensity01 observotionsílbserver bias.(omp letenes s

VolidilyRelevonceStotilticolsignilicooceDotomixingccnstruints

Formal

DigitolrepresentationGeometriceqist m f i o n

1nterpolotion/extropolotionNormolizotion

SimulolionSimplilicotionSu b-gridsro le proressesBilurcotion

ConfidenceSenlitivity01 obiedive[undions

Unduetrust in technology

Conlent

Mothemolicollog icSpoticl convolutionClolsificotionystemSub iectivity

PerceplionReodobilityApp rop rioteproienionInterprefotion

6

implementing the technologies discussed here. Training is particularly importantwithin the context of the needs of both the scientific and operational applicationsof these technologies. As both remote sensing and GIS software systems bec omemore sophisticated, the need to understand their potential increases, even asatrernpts are made to make these systems less user hostile. Yet, institutionswhich must do such training are faced with the complex issues related to main-

taining state of the art cap abilities in areas of rapidly advancing technologies.

A ecu raey

In remote sensing, two components of the accuracy evaluation can be identified.The first one is related to the physical measurement itself, such as the one madeby a satellite sensor recording the radiation reflected from the earth surface. Thesecond refers to the accuracy ofinterpretation or transformation ofsuch ph ysicalme asurements into information related to either features or processes on thegroun d. Obviously, both the physical measurements and the derived information

are linked. Yet, different types of research are required to improve our knowl-edge of these processes. Methodological developments in image analysis try toimprove constantly the accuracy of interpretation. This accu racy depends, how-ever, upon the type of information sought. For example, land cov er maps cangenerally be produced with an accuracy of more than 90 . National assessmentsof forest changes can be produced with similar accu racy.

Er ror assessment still represents a major research item in the dev elopmentand application of the GIS technology. Data used in GI S-based analyses typical-Iy must be collected from a variety of sources. Each of these data produ cts willhave been compiled to address specific and perhaps conflicting goals withrespect to locational and thema tic accuracy. Methods must be developed toassess the uncertainty, or error, associated with specific types of data products.

Furthermore, a sophisticated understanding must be developed for the accumula-tion of various types of error as multiple data sources are combined into infor-mation products (see Fig. 17). What one must constantly strive to ensure is thatthe accur acy imp roves with respect to traditional field methods of datacollection.

Stcn d n rds

Standards are neecled for the transfer of sp atial data in electronic formo As cur-rcntly structured, spatial data sers typically carry an unaccept able overhead

61



L EVE L 1 :G lo b al A V H RRIle so / u /io n: 1 .1km

L E V E L 1 1 :C O N T I N E N TA LAV H R R

Londsof m ultispecfrol scanner

Il eso/u/ion: 1 .1km 8 0 m

L E V E L11 1:B IOME

L on ds of m ultisp ec frol s ca nner

them otk m op p er

Ile so / u /io n: 8 0 m 3 0 m

L EVE LIV: R E GIO N

T hemofic m o p p er hig h oltifude

oiruoft

Il eso / u / i on: 30m 3 m

L EVE LV :P LOT

High ond low oltifude oiruoft

Ile so/ u / ion: 3 m 1m+

LEVE LVI: S AM PlE S IT E

Surfme meos urements

on d observofions

when rapid error-free exchange between users is needed. More research isrequired here to facilitate distributed cooperative scientific investigations andsharing of data and information. Standardization of spatial data transfer formatswill further encourage the combined use of remote sensing and GIS data sets.Finally, although the problems of operating in heterogeneous computing envi-ronments remain, this issue is being seriously addressed by major hardware sup-

pliers. It is our belief that problems currently associated with the exchange ofdata and processing between different vendor systems will be substantiaJly miti-gated within the next few years.

Timel iness

Timeliness of access to data is a crucial characteristic of an information system.Requirements will obviously vary from one application to another (see Fig. 15).Crop monitoring or a search for locust breeding grounds will lead to more strin-gent temporal requirements than a general land use survey. Measurements offlooded surfaces will require more frequent data than a national forest inventory.In its early stages, remote sensing may have appeared to respond slowly to oper-ational needs, and requirements of agencies, institutions and comrnercial enter-prises. Today, advances in communication or in real-time local satellite dataacquisition are constantly improving the potential for the real enough time useof these data. Commercialization of portions of the remote sensing sector hasalso contributed to improve thetimeliness of data access even if the pricingstructure currently applied to data by the commercial sector may strongly limitaccess to data for manyapplications. The problem of data dissernination is thusseen as two-fold: on the one hand, it is up to the users to rnake sure that cornmu-nication channels with data sources are optirnized for their specific needs. Onthe other hand, it is also the responsibility of nations and agencies which holdthe key to satellite systerns and data archives to reduce as much as possible poli-cy or practical restrictions lo tirnely data access.

Another aspect of timeliness in producing the required analyses for sustain-able development necessitates the mastering of techniques. This leuds us toernphasize the need for the training of scientists ami technicians able to copewith evolving technologies. This can also be seen as an infrastructure issue.

Gen erolize dVe ge lolionClos sif ic o tio n

/

~~Bourdor y Wote rs Con oe Are o

/ \

F ig u r e 17 (Left). A((u mu lofion 01 err or in on infegroted re rnote sensing G IS processing flow

6 3



Te h n o J o g yT re n d s

Information technologies of the kind proposed in this digest may appear to manyas being intensively high-tech driven. Human ingenuity can thus be seen asapplied first to the design of instrurnents and tools and Iess to the nurturing ofskills and know-how to use them. The trend is clear and field practitioners mayquickly feel overtaken by events. Such general feelings of incapacity to keep up

with advances, let alone master their application, can only be mitigated by amajor and concerted educational and developmental effort to reverse the currentstate of affairs. At this stage, it may not be so much the technology that iswanting, but perhaps more our ability to put it properly to work.

The situation is further eompounded by constant changes in the technicalbase. Current trends in the three major technology areas discussed here (Remotesensing, GIS and mo deling) are surnmarized in Table 5. Techniques in rhesethree areas have amply dernonstrated their capability to aid researchers andresource managers as they monitor the earth surface, detect changes, and mea-sure and evaluate trends. The combination of these technologies represent apowerful means of verifying that environmental commitments made withrespect to the conservation of resources have been rnet. Whether we like it or

not, those techniques offer a possible verificationmechanism for holding nationsaccountable to those commitments. The quality of intemational data which canhelp chart the path toward sustainable development is a shared responsibilitywhich requires the free circulation of information. This topic and these technolo-gies have far-reaching scientific, social, economic, political and ethicalimplications.

R ole o fSd en t ists

It is important that the role of seienee and technology in human affairs be more

widely known and beuer undcrstood. A corollary to that statement is ihat the sci-eru ific co m m uni ty holds a special responsibility for communicating what thebusiness of science is about. and kceping up wi th advances in a wide variety officlds of inves tiua tinn. Currcntty. the da ta and inforrnation technologies dis-cussed in this dige st are liulc known and, 1 0 date, are truly understood by a rela-tivc ly s mull num be r 01' scicmisrs . Cross dísciplinary a pp roa ches are. a s a rule,not dirccily a ppcaling 10 more cla s s ica lly uuined scientists. This ignorance 01'

the cupa bility a nd limit.uions o f tools by rhe broader scientific a nd rcsourccma na gcmcnt commu nity is oücn the source of unclue s ce pticism or cxce ss ivc

optimism with respect lO rcality. Thc scicntific comm uuity involvcd in variouselernents 0 - sus ta inab lc d cvclopm e nt has, thcrelore, the particular rcsponsibility

64

T

T

T a ble 5. Cu r r en r ends in rem oJe sensin g, GIS o nd modeling

to significantly upgrade its familiarity with tcchniques which have been showntobe mos t efficient a nd effective in co llec ring, organizing a nd applying appro-pnaie data to a parl~cuJar field of investigation. One migh: cven venture 10 sayrhat the problerns of de ve lopm e nt and environment will increasingly beco me so

intractable that their resolution \ViII necessitaie the integ ra rion and the use of sci-ence and technology at scales which. as yct, have reccived liule consider.uionThe scicntific community has an additional rolc lO play in Inaking sure that rhcscrcchniques are properly tuned ancl adjustcd lO problerns of xustainahlc c1evclop-mcnt a l all scales. Finally, scientists I11USlendcavour, as a co mmu niry, to steera dva nce d technology programmes of the furure in dircctions which truly SC I've

human necds.

65

With such a perspective technology transfer assumes a particular signifi- S O U R C E M AT E R I AL



With such a perspective, technology transfer assumes a particular significance. Indeed, inforrnation systems 100ls now permit significant decentralizationin the data analysis and decision-rnaking process. Technological advances arenot reserved 10 a particular cou ntry or a specific class of scientists any more;they are increasingly within the reach of individual scientists or institutionsaround the globe. A true contribution by the world scientific community to sus-tainable development objectives would thus derive, among others, frornimprovements in the accessibility of those techniques discussed herein to the

globally distributed science community as a whole. In retum, farniliarization ofscientists and technicians with a diversity of scientific concems and field situa-tions can only work toward irnproving the appropriateness of proposed technicalaltematives to the solution of problerns arising out of our attempt to achieve thegoal of sustainable development.

American Society of Photograrnmetry and Rernote Sensing. 1991. Special Issueon Integration of Rernote Sensing and GIS. Photograrnmetric Engineeringand Remate Sens ing 57 6):641-697 .

Burrough, P.A. 1986. Principles o/ Geographical Information Systems for LandResources Assessment. Clarendon Press, Oxford.

Cohen, P.R. & E.A. Feigenbaum (Eds). 1982. The Handbook of Artificiallntelligence. Heuris Tech Press, Stánford.

Co lw e ll, R.N. (Ed.). 1983. The Manual of Remate Sensing, Second edition.American Society of Photogrammetry and Remote Sensing, Falls Church,Virginia.

Estes, J.E. 1985. The need for improved information systems. Canadian Journalof Remate Sensing 11(2): 124-131.

Goodchild, M. & S. Gopal. 1989. Accuracy of Spatial Databases. Taylor andFrancis, London.

Gennany. Gesellschaft für Technische Zusammenarbeit. 1991. Remote Sensing:Tools fo r Development. Deutsche Gesellschaft fü r Technische

Zusammenarbeit (GTZ), Eschbom.Jensen, J.R. 1986. Int roductory Di gital lmage Processing . Prentice Hall,Englewood Cliffs, New Jersey.

Monmonier, M.S. 1982. Com put er Assisted Cartograpliy. Principies an dProspects. Prentice Hall, Englewood Cliffs, New Jersey.

Richards, J.A. 1986. Remate Sensing Digital Image Analysis: An lntroduction,Springer- Verlag. Berlin.

Star, J.L. & J.E. Estes. 1990. Geographic lnformation Systems: Anlntroduction,Prentice Hall, Englewood Cliffs, New Jersey.

United States Geological Survey - Mapping Science Commiuee. 1990. SpatialDota Needs: The Future o/ th e National Mapping Prog ram, NationalAcademy Press, Washington, D .e .

66

Fecha de Venc imiento



United States Geologieal Survey - Mapping Seienee Comrnittee. 1991. Researchand Development in the National Ñapping Division, USGS: Trends andProspects. National Academy Press, Washington, ~.c.

Weiss, S.M. & C.A. Kulikowski. 1984. A Practical Guide to Designing ExpertSystems. Rowman and Allanhead, Totowa, New Jersey.

Winston, P.H. 1984. Artificial Intelligence. Seeond edition. Addison- Wesley,Reading, Massachussetts.

S ug ge ste d R e ad in g

Aronoff, S. 1989. GIS: A Management Perspective. WDL Pub1ieations, Ottawa.Cowan, DJ. 1988. GIS versus CAD versus DBMS, what are the differenees?

Photogrammetric Engineering and Remote Sensing 34:15551-55.Eeologieal Society of Ameriea. 1991. The future of remote sensing in eeologieal

studies. Special feature. Ecology72(6): 1917-1945.Fritsch, D. ; Bill, R. 1991.Geoinformationsysteme, Wichman Verlag, Karlsruhe.Guptill, S.C. (Ed.). 1988. A Process for Evaluating Geographical Information

Systems. USGS Open File Report 88-105. Federal Interageney CoordinatingCommittee on Digital Cartography Teehnieal Report l.

Mounsey, H.M. (Ed.). 1988. Building Databases for Global Science. Taylor andFraneis, London.

Parker, H.D. 1991. International GIS Sourcebook 1991-1992. GIS World Ine.,Fort Collins, Colorado.

Peuquet, DJ. & D.F. Marble (Eds). 1990. Introductory Readings in G/S. Taylorand Francis, London.

Raper, J.F. (Ed.). 1989. Three Dimensional Applications in GeographicalInformation. Systems. Taylor and Francis, London.

Rhind, D.W. & N.P.A. Green. 1988. Design of geographieal information sys-

terns for a hererogcncous scieruific eommunity. lme r na tio no l Ioúrnal ofGeographical Injorrnatipn Systems 2:171-190.. c

Rouet, P. 1991. Les données dans les systémes d information géographique.Editions Hcrrnes, París.

Seanvie. J.Y. 1993. Utilisation de la télédétection dans les sciences de la [erre.Manuels el Méthodes, 7. BRGM, Orléans.

SIUlh, J.W. & B.G. Lyons (Eds). 1993. Decisión Support Systerns for theManagement of Grazing Lands: Emerging lssues. Man and the BiosphereSeries 11. UNESCO, París, and Parthenon Publishíng, Carnforth.

UNESCO. 1992. New Technolo g ies: Re m ote Sen sin g an d Geo graphi clnjormation Syst ems. UNESCO Environment and Development Brief 3.

UNESCO. París.

68

• 'I~~O I

2;M D ~ f ; n-.\

0<

~ ()(i Jc ;;Ur, 1 H n

~ i u M U : :. ,.•. . •.v ... • .

e o M I 1 I

i

:II

I

1

,

1

i

1

advanced data acquisition and analysis technologies for sustainable development

Documents