Understanding the biosphere and howhuman activities affect it nearly al-ways requires the efforts of many in-vestigators, usually from various sci-
entific disciplines. The collective analysis of dataoriginally gathered by individuals but subse-quently stored in shared databases can yield in-sights beyond inquiry of a single data set. To putit all together, ecologists need large, complex datawarehouses and data-mining facilities.1,2 A ma-jor barrier to such data warehouses is scientists’inadequate documentation of field data so thatothers can use that data. Early integration ofdatabase technology into the research processwould enable more efficient data documentation,but the payback to researchers for any additionalwork must be real and immediate.
The Canopy Database Project is one of sev-
eral national efforts building prototype systemsthat deepen our understanding of how field ecol-ogists could use database technology. It focuseson forest canopy research, an emerging ecolog-ical subfield. The canopy is one of the richest butmost poorly studied habitats in the biosphere(see Figure 1).3,4 The field’s relative youth, withits lack of entrenched methods, legacy data sets,and conflicting camps of competing groups, pro-vides an excellent opportunity to integrate datamanagement and analysis tools into the researchprocess. And, because canopy research is inher-ently multidisciplinary, the work is generalizableto other fields of ecology.
This article presents one aspect of our approachto building a data archive for canopy researchers.We show how small, ecologist-centered projectsthat produce immediate short-term value to par-ticipating researchers are essential to achievinglong-term ecosystem informatics research and de-velopment goals. Such projects keep ecologistsinterested and involved, provide experience withreal data and problems, and increase our abilityto use effective software engineering techniquesto construct larger systems iteratively.
Obstacles for Ecologists
Our project began in 1993 with a planning grant
HOW TREES AND FORESTSINFORM BIODIVERSITYAND ECOSYSTEM INFORMATICS
JUDITH BAYARD CUSHING AND NALINI NADKARNI
The Evergreen State CollegeBARBARA BOND
Oregon State UniversityROMAN DIAL
Alaska Pacific University
To solve critical biosphere-level problems such as global warming, decreased biodiversity,and natural resource depletion, scientists must integrate data from many researchers.This, in turn, requires better data infrastructure and informatics tools than are currentlyavailable. The Canopy Database Project brings together computer scientists andecologists to develop informatics tools for forest canopy research that meet suchecosystem informatics challenges.
from the US National Science Foundation. Thefirst step involved surveying 240 canopy re-searchers to identify major obstacles that impedetheir research.5 The most commonly cited ob-stacle was not difficulty of physical access to thecanopy (as we had expected), but problems inmanaging, using, and sharing data sets. Re-searchers associated this with a lack of uniformityin collecting, processing, and analyzing canopydata, a dearth of data archives, and an inability tolink data for comparative research. Two yearslater, with a second grant from the NSF, welaunched the Canopy Database Project, pairingecologists with computer scientists to developdatabase tools that support data sharing and eas-ier access to analysis tools for canopy research.6
Creating data-warehousing and data-miningtools for tomorrow’s ecologists requires many im-provements to current archival systems.7,8 Wemust populate these archives with more and bet-ter-documented data than now but this is a daunt-ing task for scientists who are not “power com-puter users.” Although ecologists often consultWeb-accessible information, they typically stillenter data by hand into private data stores that arerarely published or archived. Despite increasingpressure from funding agencies, availability ofseveral ecological data archives, emerging toolsfor recording metadata,9 and opportunities forpublishing data in the Ecological Society ofAmerica’s archives (www.esapubs.org/esapubs/archive/archive_main.htm), ecologists still per-ceive documenting data for archival purposes tobe a time-consuming process and many don’teven attempt it.10,11
Documentation of field data sets is known asscientific metadata and is essential to retrospec-tive or application use of data set information.These metadata are typically recorded, if at all, atthe end of a scientific study, usually after re-searchers have lost their intimate familiarity withdata set details. Using database systems couldhelp, but ecologists tend to prefer flat files,spreadsheets, or (at best) nonrelational (flat)database systems. Individual researchers rarelyuse modern database management systems, al-though those who are good programmers tendto use sophisticated statistical programs or writecomplex mathematical models.12,13
Project Goals and Objectives
Our initial objective was to integrate databaseuse very early in the research cycle—ideallystarting with a study’s initial design. Specifically,
we wanted to develop an end-user database-design tool that could reuse domain-specificdatabase components. A field database, designedin this way, and used during the entire researchcycle, would include ways to note metadata iter-atively as the project evolved. Such metadata-marked data are easier to validate, document,transform, analyze, and archive than data fromflat files or spreadsheets. Moreover, the fielddatabases developed with a coherent collectionof components would be easier to integrate forcollaborative and retrospective study than thosebuilt idiosyncratically. This vision is being car-ried forward in our lab with the development ofa database-design tool (called DataBank; http://canopy.evergreen.edu/databank).14,15
The project’s initial focus on end-user data-base-design encountered three problems. One,we built a tool that generates field databasesfrom off-the-shelf, domain-specific components,but evolving it as ecological studies grow is be-yond the technical skill or inclination of typicalend-user ecologists. Current database manage-ment systems, even Microsoft Access, are not yeteasy enough for most nonprogrammers to ma-nipulate. Two, finding the “right” reusable com-ponents is not easy and requires effective toolsfor the user community to publish, maintain, andorganize components. Due to the field’s relativeyouth, articulating the components will requirecareful documentation of field protocols andmore time than we originally envisioned. Finally,
Figure 1. The forest canopy of the Pacific Northwest and the WindRiver Canopy Crane Research Facility. The WRCCRF is one of severalresearch sites specialized for studying the forest canopy. Aconstruction crane lets researchers study the canopy from multipleviewpoints. Several researchers working with the Canopy DatabaseProject use this crane site. (Photo courtesy of J. Franklin.)
34 COMPUTING IN SCIENCE & ENGINEERING
using a database system means changing the wayecologists work. Without clear and immediatebenefit, few researchers will spend extra timeadopting new technology.
To better understand how to provide immedi-ate benefit from early integration of databasetools with DataBank, we needed deeper insightinto the research process. To accomplish this, wedeveloped several small, immediately useful toolsapplicable at various phases of the researchprocess. Two of these smaller efforts, presentedin this article, have given particular insight intowhat a larger tool should look like and what pro-ductivity gains might motivate researchers to usedatabase systems and provide metadata.
The Ecology Research Cycle
To learn what ecologists need, we first studiedtheir research cycle. We represent the typicalsteps in ecology research as a linear process forexplanatory purposes, but note that any phase ncan progress to phase n + 1 or return to any pre-vious phase n – x (see Figure 2).
In the first phase (study design), canopy scien-
tists find studies similar to theone they’re designing. Theymay want scientific articles orfield data relevant to the re-search sites they’re consideringor accident reports associatedwith equipment under consid-eration. To help with this phase,we developed a preliminary do-main-specific research refer-ence tool (see http://canopy.evergreen.edu/bcd). Research-ers at this stage also typicallydraft preliminary research pro-tocols, articulate hypotheses,and design preliminary fielddata intake forms. This phaseusually culminates in writing aproposal for funding.
In the second phase (fieldwork), a scientist typically usesthe research protocols and fieldintake forms developed in thefirst phase. Because field charac-teristics often differ from thoseinitially envisioned, ecologistsoften alter the protocols andforms (and hence the databaseschema) in the field. Thus, anydigital data collection and data-
base systems used in the study’s design phase musthave easy update functionality.
The third phase (data entry) is usually a sepa-rate step for the researcher, with data transferredfrom paper to electronic intake forms, some-times months after the data’s initial acquisition.If data errors are discovered, the researcher can-not always return to the field to gather new data,so data analysis strategies or even research ob-jectives might be rethought or data extrapolated.There is currently little automated data valida-tion at this phase, although early data validationcould enhance the research.
During the fourth phase (data analysis), re-searchers often reformat data for analysis or mod-eling or place intermediate results into data sets.This is time-consuming, especially if scientists arenot experienced with the analysis or modelingprogram. At this point, the researcher could stilldiscover that key parameters or data are missingand return to the field, use another researcher’sdata (collected for the same or a different pur-pose), extrapolate or interpolate existing data, ornot conduct an originally envisioned analysis.Database technology would make these problems
Studydesign
Fieldwork
Dataentry andverification
Dataanalysis
Data sharingwithin group
Journalpub
Dataarchive
Datamining
• Validate data against metadata
• Visualize data
Use predefined queries across common names, types, codes •
• Archive data within-lab
• Visualize data analysis
• Capture data digitally in-field; enter onsite into database
Visualize cross-study changes •
Redesign database for semantic integration •
Validate data against metadata •
• Use research reference tool; design preliminary database
Figure 2. The ecological research process annotated with the Canopy DatabaseProject’s efforts to improve researcher productivity. In the first phase of the researchprocess, for example, a research reference tool could facilitate study design and apreliminary database for field data could clarify field protocols. For field work, digitaldata capture should be made available, and these data should be uploaded onsite toa database. At later stages, visualization, validation and query tools, and processesfor intermediate data archiving should be available.
MAY/JUNE 2003 35
obvious earlier and facilitate data transformationand sharing of macros, parameters, or scripts forapplications. We are developing visualization pro-grams using the Visualization Toolkit (http://pub-lic.kitware.com/VTK) that can display the datarepresented in our database components. We alsoare experimenting with a spreadsheet applicationthat finds tree physiology data outside local min-ima or maxima range.
The fifth phase (collaborator data sharing)brings the research group as a whole into thepicture. Subsequent to data verification and pre-liminary analysis, some laboratories require re-searchers to save and document data sets to acommon within-lab data store in formats readilyunderstood by others in the laboratory. Thisavoids potential loss of field data if a researcherleaves, for example. Documenting data, even foran audience familiar with the research objectivesand protocols, takes time, however. Databasetechnology could facilitate this process, espe-cially in larger laboratories where many re-searchers focus on related problems and havecommon research protocols, vocabulary, fieldprocedures, and instrumentation.
The sixth phase (journal publication) is howresearchers receive feedback from peers, advancethe state of the art, and acquire continued fund-ing. Database tools can help in manuscriptpreparation in several ways—for example, byeasily generating graphics or by gaining accessto scientific citations.
Data documentation and validation are themajor bottlenecks to the seventh phase (dataarchiving).11,16 Deciding which data to archive,writing data descriptions, and validating the dataagainst those descriptions seems disconnected toindividual researcher goals. These tasks often areso overwhelming that researchers usually dothem at the end of the research cycle, eventhough data documentation could provide usefulartifacts if applied earlier.
Another major bottleneck to data archiving isthat some scientists hoard data—they hesitategiving up their data for others to use. Concernsabout being appropriately credited and havingtheir data properly used are sociological prob-lems being addressed by funding agencies, long-term research sites, and some journals. The Eco-logical Society of America, for example, publishespeer-reviewed data sets, but many scientists feelsuch publications do not carry the same weightas archival journal publication in terms of recog-nition and career advancement. Although tech-nology cannot change the sociology of data
archiving, applying database technology early inthe research cycle could make data documentingconsiderably easier and turn data archiving into amatter of pushing a button that says, “publish thisdatabase.”
The final stage (data mining) is still usually ac-complished on a person-to-person basis, with somenotable exceptions such as weather data, satellitemaps, or data sets published for permanent sites suchas the Long-Term Ecological Research Network’s(LTER; http://lternetwork.edu).Even where data are publishedelectronically, though, few datastandards exist in ecology. Twodata sets gathered from a singlearchive might have different dataformats or worse, different se-mantics (even with identical datavariable names). Community-maintained common vocabular-ies, common data formats orcomponents, and data integra-tion tools could help, and several other promisingprojects also address these problems; see, for exam-ple, the Science Environment for Ecological Knowl-edge Project (http://seek.ecoinformatics.org).
The next two sections of this article describethe Canopy Database Project’s efforts at in-creasing researcher productivity with smalldatabase-like tools. We first describe the devel-opment and use of a handheld field data acqui-sition tool, which would find straightforward in-tegration with our database design tool,DataBank. Then we describe an effort to help alab use data management tools and best prac-tices to carry out within-lab data documenta-tion and archiving. This initial documentationfor close collaborators, we postulate, could helpautomate data validation and render later meta-data provision for archives significantly less in-timidating. The ideas emanating from this sec-ond effort will also be integrated into futureversions of DataBank.
A Handheld Data Acquisition Tool
Because ecological field data are typically ac-quired by hand and fraught with numerous tran-scription errors, we experimented with digitaldata collection. Our aim was to determine, first,whether such a tool would be useful to canopyresearchers and, second, whether the technologydeveloped for one project could be cost-effec-tively applied to other projects. Here, we describeour efforts to use a handheld data acquisition tool
Data documentation
and validation are the
major bottlenecks to
data archiving.
36 COMPUTING IN SCIENCE & ENGINEERING
to partially automate the collection of field data.We also wanted to integrate the use of databaseswith the handheld tool.
Ideally, a canopy field project would stream-line data handling to integrate all phases of theresearch process. Study design might includefield methodology and technology, which wouldin turn make data entry amenable to immediatedata analysis and sharing. Imagine, for example,field protocols that use instruments to recorddata as they are collected and provide access fordata analysis and real-time communication to adata warehouse. In practice, however, canopyscientists treat research phases as separate tasks,so there are impediments to such sharing or evenanalysis. One of us (Roman Dial) had an idealproject for applying database technology, in partbecause he had already used digital devices, butmore importantly because he recognized thepossibility of improving productivity.
Dial’s work quantified the forest canopy’s struc-ture, particularly the air-space gaps between trees.Forest canopy data are intrinsically three-dimen-sional: 3D locations are typically defined usingCartesian coordinates (x, y, z). In canopy science,the x and y values give the planar projection onthe ground, with the z value representing theheight above ground. In principal, these valuesshould be easy enough to collect, but in practice,locating a canopy element’s specific position canbe difficult. (A canopy element is a vegetativestructure located above ground, such as a treestem, tree foliage, epiphyte, and so forth.) Insteadof Cartesian coordinates, Dial used cylindrical co-ordinates (r, θ , z), which measure z as the dis-tance above ground of an observer looking at acanopy element (itself located at eye level) and incompass direction θ but a distance r away fromthe observer.
Given available technology, cylindrical coordi-nates are easier to collect than Cartesian ones.First, the distance above ground, z, requires a tapemeasure to be stretched from the ground to theobserver. Second, the direction θ, taken from theobserver to the canopy element, requires only amagnetic compass (Dial used a digital MapStarmodel from Laser Technology). Finally, a laserrange finder measures the distance r to the canopyelement from the observer (Dial used a digital Im-pulse 200LR model from Laser Technology). TheImpulse laser and MapStar compass each reportdigital measurements in a downloadable form,and a PDA could, if properly configured, auto-matically receive the data.
Prior to this project, Dial’s team used digital
laser range finders, but like most ecologists, theyrecorded data manually, with a pencil and note-book. Three different individuals handled eachpiece of data: the observer, the recorder, and thedata typist. In the canopy, the observer aimed andfired the laser, then called down to the ground-based recorder the observed element identity andits position values r, θ , and z. The recorder wrotethe information onto a data sheet. Data entry per-sonnel later typed this data into a computer. Tran-scribing numbers in the canopy can be awkward:usually only one hand is available to record datawhile the other uses an instrument to measurestate variables or steadies an observer dangling inspace. Moreover, recording data by hand and later(perhaps weeks or months later) transcribing theminto a digital format is error-prone.
Capturing the data digitally would, we rea-soned, eliminate two steps from the process:manual recording and data entry (both the Map-Star compass and the Impulse laser range finderproduce digital data). Thus, the one person hold-ing the laser and compass could measure, record,and store in a digital database each observationwith one push of the button as it was observed.This would increase productivity, decrease errorpropagation, and eliminate the need for arecorder and a data typist, thereby reducing tra-vel, field costs, and time.
Dial approached Nalini Nadkarni and JudyCushing (the Canopy Database Project’s direc-tors) with his idea of recording compass andrange finder data digitally to a PDA. Together,we designed and developed an “electronic datasheet” for the PDA that seamlessly integratedthe cylindrical coordinates and canopy elementdata with off-the-shelf database tools. Dial usedthe materials and methods we describe nextduring fieldwork in 2002, when he collected 3Ddata in two forest canopies: one a tropical rainforest characterized by high heat and humidity,the other a temperate Eucalyptus forest char-acterized by strong winds (up to 30 mph), andboth characterized by great height (Dial rou-tinely sampled to 250 feet above ground). Heused the instruments to collect spatial data (de-fined as the distance and area between canopyelements) and statistical frequencies of canopyelements in both forests. He also used thePDA’s database structures to record dates,times, and locations of sensors that measuredlight, temperature, and relative humidity, aswell as the dates, times, and locations of nylontrays positioned to capture the rain of arthro-pods killed with insecticidal fog.
MAY/JUNE 2003 37
Research Methodology, Access, and Instrumentation To collect his data, Dial first positioned himself inthe canopy by stretching horizontal ropes be-tween tall trees (termed emergents) that extendabove the level of neighboring crowns. Often theemergents were so tall that Dial’s horizontalropes were suspended above the level of the for-est between them. From these horizontal lines,Dial attached vertical climbing ropes, whichserved as transects (linear sampling units) thatreached to the ground, allowing him to samplethe full forest height. The vertical transects werelocated at 5- to 20-meter intervals along the tra-verses. In essence, Dial’s samples consist of a sys-tematic, vertical cross section through a forest.
Palmtop Computer (PDA)Several PDAs are currently available and mostare quickly improving in terms of features andpower. Dial used the Palm m105. It was rela-tively inexpensive (US$150), provided an 8-Mbyte memory, and used replaceable batteriesand plastic instead of metal and glass for its con-struction. The m105 also offered Palm’s Grafittihandwriting recognition software. AlthoughDial never lost data, changing batteries on PDAunits in the field sometimes reset those units,which could have meant lost data and programs.A spare unit reset itself three times during twomonths, even without removing the batteries.Dial had no moisture problems or breakage, de-spite hundreds of hours in the field and exposureto high winds and humidity. Because the PDAwas enclosed in a padded, waterproof case with aclear vinyl window, neither the screen nor thevinyl window fogged, even when used duringtropical downpours.
PDA SoftwareDial used the Palm PDA’s installed software andspecial-purpose PDA software. Of the Palm’s in-stalled software, he used DateBook to plan andrecord tasks by time and date, NotePad forsketches and other freehand notations, andMemoPad for longer observations. He auto-matically uploaded the resulting data to his lap-top, usually once a day for integration into largerdata structures or reports after hot-synching.The advantage of using packaged software wasthat it eliminated the step of entering hand-recorded data into the computer; the data wererecorded electronically and directly on the PDA.
Dial also used AppLaser, a special-purpose ap-plication developed by the Canopy Database
Project team jointly with Dial, which recordedand uploaded canopy data directly to an MS Ac-cess database. AppLaser’s requirements includethe ability to
• upload digital location data from the MapStarcompass and Impulse laser to the Palm alongwith height above ground, thereby recordingcylindrical coordinates (r, θ, z),
• date-stamp data automatically, and • record hierarchical location data and insert it
into an MS Access database.
Because canopy data collection involves canopyelement identification and location information,Dial required entry of string text in the applica-tion. The composition of canopy elementschanges through the canopy, so he also needed adrop-down menu that he could modify with a sty-lus as he ascended through the canopy (see Fig-ure 3). The data also needed to fit into a commondatabase program and thus be amenable for useby other canopy researchers involved in theCanopy Database Project.
AppLaser Data and Data EntryAppLaser has five menu screens (denoted hereby italics): The main menu lets users select
Figure 3. Ecologist Roman Dial collecting canopyelement data in a 250-foot tall Eucalyptus forest inAustralia. Dial is suspended by a vertical transectrope as he collects and records data using awaterproof PDA. The PDA is connected to a digitalcompass and digital laser range finder. Thesesurveying instruments automatically downloadlocation measurements into the PDA while Dialuses the PDA’s handwriting recognition software torecord the located object’s identity. (Photo courtesyof Bill Hatcher Photography.)
38 COMPUTING IN SCIENCE & ENGINEERING
among four subscreens. Each subscreen corre-sponds to a data record; the screens together de-fine a hierarchical database in which aStudy_Area contains one or more Transects, eachof which contains one or more Obs_Points, eachof which contains one or more Obs_Measure-ments. Each of the four subscreens has a variablenumber of fields, some of which are offspring ofparent screens. Values of the parental screens areautomatically recorded as the user descends thehierarchy. Thus, for example, as the user navi-gates down to Obs_Measurement, the Study_Area,Transect, and Obs_Point values are automaticallyapplied. It is not possible to navigate down with-out defining the parent record.
The lowermost of the hierarchy, the Obs_Measurement screen, includes the two fields au-tomatically filled by the surveying instruments.The field “distance” (to canopy element, r) isthe measurement sent to the PDA by the Im-pulse laser and “azimuth” is the compass direc-tion to the element sent by the MapStar com-pass. The “date” field is a time stamp of date andtime information provided by the native PDAoperating system. All other non-offspring fieldsare filled by the observer using Graffiti or viapop-up menus (offspring fields are populatedusing values from ancestor screens). A user canadd, edit, browse, or delete any data record us-ing screen buttons and Graffiti. Offspring fieldscannot be altered.
Fundamentally, the data structure records thecanopy element’s identity and its location infor-mation in time and space. The canopy element’sposition is located globally via the UniversalTransverse Mercator (UTM) grid system (incontrast to latitude and longitude). In addition,there are opportunities to record opportunisticobservations as comments.
According to Dial, after a day in the field, thePDA feels extremely precious—and it is. It holdsan entire day’s worth, or more, of hard-won data.Ecologists raised on yellow notebooks and pen-cils will feel particularly naked. Eventually thisfeeling passes, but only after hot-synching thePDA to an office, lab, or laptop computer afterevery data collection session and then copying theresulting databases onto backup media. Hot-synching AppLaser data populates four relationaldatabase tables in MS Access: Study_area, Plot_lo-cation, Observation_point, and Observation_mea-surement. Each record in an Access table corre-sponds to a data record that the PDA records.Each hot-synch updates Access tables by append-ing new records at the end of each database table.
Serendipitous Uses of AppLaserDial discovered that AppLaser was robust andflexible enough to handle data other than thespatial data, as originally envisioned. In tandemwith spatial data in forest canopies, light data,temperature, relative humidity, and arthropodabundance and diversity are of interest to canopyresearchers. Thus, in addition to receiving dis-tance and azimuth signals, he recorded micro-climate data and names of arthropod foggingtrays by using the “comment” field in Obs_Point.
Dial used the common Study_Area name forfour variable types (space, data logger location,light, and arthropod tray location). However,although each was collected physically alongthe same vertical rope, each represents a differ-ent transect. This means that even though theyshare the same “bearing,” “X,” and “UTM”fields, they must have different transect names.Thus, the user might record for transect“T7”— where spatial data were collected at abearing of 27º and 35 meters from the origintree (the one whose UTM coordinates definethe Study_Area’s location)—transects “T7light,”“T7temp,” and “T7trays.” Under Obs_Pointwithin transect “T7light,” Dial recorded heightin “Z field” (where the light measurement wasrecorded) and the amount of light in the “com-ment” field. Similarly, in transect “T7temp,” herecorded height in the “Z field.” The benefitwas that these additional observations were au-tomatically recorded into the study’s databasewhen hot-synched.
Work Remaining and Lessons LearnedConceptually, AppLaser proved satisfactory forDial and his team. A few design errors should berectified before the software is made available toothers:
• Inclination. If the “distance” field of Obs_Measurement were filled by the laser rangefinder’s slant distance variable, then each obser-vation point would become the origin in aspherical coordinate system. A canopy workercould use this to observe all around, not just inan assumed flat plane with fixed height aboveground.
• Ability to add a new element to the pop-up menuon the fly. The researcher must currently backout of the Obs_Measurement screen to themain menu, then go into a pop up to add anew element.
• Change of the delete key’s position. When using thelocation page, research assistants occasionally
MAY/JUNE 2003 39
tapped Delete when they wanted to tap New,and as a result had to retake the lost data.
• An index that counts the number of measurementstaken at a given Obs_Point and displays this counton the Obs_Measurement screen automatically. Of-ten a researcher knows enough to take n obser-vations per observation point, but won’t oftenknow if he or she actually did so. Dial countedmeasurements using Next and Prev buttons,but this is awkward, slow, and error-prone.
The following new AppLaser features would beadvantageous:
• On the Obs_Point screen, each line of the“comment” field could be its own field, allow-ing for multiple entries at a given height thatcould be downloaded separately to MS Access.
• There should be a separate pop-up menu forthe “Found On” field on the Obs_Measurementscreen. “Found On” would be useful for habi-tat studies, and it needs its own pop up.
• A Delete key on the main page that com-pletely purges the database could improvespeed immensely. The database on the Palmneeds periodic cleaning because as it fills, eachnew observation takes longer to appear, andthe extra data slow the application. Right now,the user must press Delete repeatedly to purgethe database.
• Better data management on the PDA and bet-ter coordination with the MS Access database.
• Improving the quality of the cables connect-ing the laser finder, compass, and recordingdevice via the RS232 ports would improve thedevice overall. Although waterproof, attach-ing or removing them while either the laser orcompass is turned on could damage them.
Currently, AppLaser is only programmed tohot-synch on a Windows PC. Dial never experi-enced problems with transferring the PDA datainto MS Access on his laptop PC. Although heusually exported data from MS Access to MS Ex-cel and then into SPSS (a statistics package) oras raw text files for use in Mathematica, we seesignificant opportunities for improved data val-idation and management using MS Access data-base capabilities.
Equipment, contract programming, and test-ing in the field cost approximately US$50,000(including Dial’s salary). Dial’s primary canopynotebook is now a PDA. Although other re-searchers are less eager to use PDA technology,Dial is confident that canopy science will grow
into an extra-arboreal-centered activity and thatsuch tools will make data more productively col-lected, less error-prone, more easily analyzed,and more readily shared among researchers.
Including an experienced ecologist such asDial, who acted as a software designer andchange agent, has been a valuable asset to theCanopy Database Project. We learned how tointegrate field data iteratively into a databaseduring a field session and gained insight intodata validation. Dial’s use of general data struc-tures to record serendipitous measurements suchas microclimate data and names of fogging trayswas particularly useful to our DataBank work ona generalized observation data structure. Suchgeneralized structures allow needed flexibility toaccommodate changes in field data collectionprotocols, and let researchers specialize genericsoftware for their own use, thus decreasing costsof future systems.
Within-Lab Metadata Acquisition and Archiving
Unless metadata provision becomes less oner-ous and more obviously helpful, scientists willcontinue to balk at archiving their data. To ad-dress this issue, we investigated metadata anddata-archiving process within one laboratory.We used inexpensive database and spreadsheettools (MS Access and Excel) to ease the burdenof documenting field data sets, which are up-loaded to a shared store (an in-lab archive) at keytimes during research projects. A long-term ad-vantage is that the in-lab metadata will be a firststep in archiving the data.
Barbara Bond’s lab at Oregon State Universityconducts research categorized as forest eco-physiology. Bond and her group study howspecies, community structure, climate, and de-velopmental age affect exchanges of matter andenergy between plants and the environment.These interactions occur at many different scalesof time and space, ranging from the subcellularlevel at time scales of a few seconds to the wa-tershed or regional level at time scales of cen-turies. The data collected are important to non-physiologists, because researchers investigatingfundamental questions of global climate changeand biocomplexity rely on physiological infor-mation. Thus, timely archiving of ecophysio-logical data is critical, and, although researchersoften ask Bond for data, the documentation re-quired before distributing those data is difficultand time-consuming for those in her lab.
40 COMPUTING IN SCIENCE & ENGINEERING
Another reason why data documentation is acritical aspect for Bond’s lab is that to answer aquestion in tree physiology, her researchersmust piece together several types of informa-tion from different sources. Many data sourcesare used in a typical study, and lots of studies on-going in the laboratory at any given time, sodata sources for concurrent studies typicallyoverlap. A typical small project might use thefollowing information:
• leaf area and biomass information for grassesand other small plants, shrubs, and trees, eachwith different sampling designs due to differ-ences in scale;
• meteorological data from three measurementstations, each with five instruments of differ-ent accuracies, sampling frequencies, and site-specific details;
• measurements of sapflow in trees of differentsizes, ages, and species;
• measurements of stomatal conductance per-formed periodically through growing seasons;
• measurements of species composition and dis-tribution in small watersheds, obtained fromvegetation sampling plots; and
• measurements of soil water content using twoinstruments, one with continuous output andthe other sampled manually and periodically.
A complex study could have more than 20 dif-
ferent sets of measurements continued over sev-eral years, with modifications in measurementprotocol as well as personnel. Some of these dataare collected digitally via field instrumentation;others are collected by hand.
In most cases, the data require complex pro-cessing before they are useful. For example, bio-mass information involves combining publishedallometric equations from other locations withonsite field measurements. Another examplecomes from sapflow measurements. To measuresapflow, you place thermocouples and heatingsensors in trees; the raw data is a stream of tem-perature differentials between heated and un-heated probes (see Figure 4). To interpret thisinformation, you first use algorithms to convertfrom temperature to sap-flux density. Usually,some information is missing due to faulty sen-sors; this information is “filled in” statistically.Unless researchers can track changes to the dataset, the final mean values for these continuousmeasurements will show odd abrupt blips as theunderlying sample set changes over time. Hav-ing a permanent record of this kind of data ma-nipulation is important, but in reality, the detailsare sometimes lost. Procedures are needed todocument the steps of data processing for inter-pretation months or years later without creatinga huge burden for the student or technician do-ing the initial work.
After filling in missing data, algorithms aredevised to “scale up” from the individual treeto the stand level to convert from the amountof water flowing through a tree to the amountof water flowing through a group of trees cov-ering a given ground area. At each step, smallbut difficult-to-document errors are intro-duced. In much of the currently published eco-physiological work, these potential errors areseldom acknowledged.
A question this ecophysiology lab faced is howto document data collection and processing, of-ten unique to each data set, without writing pro-hibitively large volumes of support material.How can researchers be sure that the data theyarchive are used appropriately? How can they fitdata management activity into an already ex-tremely tight schedule? Sharing data about thatcould require many hours just to explain dataidiosyncrasies.
We wanted to explore how to better archivedata in the lab and how to capitalize on localsharing to make it easier to later document thosedata for the external world.
Using an LTER metadata standard (the H.J.
Figure 4. Part of a field installation for a study of environmentalcontrols on ecosystem-scale physiological processes. A student isinstalling a sensor to continuously monitor the flow of water throughthe sapwood of a Douglas fir tree. (Photo courtesy of B. Bond.)
MAY/JUNE 2003 41
Andrews LTER at www.fsl.orst.edu/lter/data/metadata/guide.cfm?topnav=115), we wrote anMS Access database application to recordmetadata for Bond’s laboratory. Source tablesfor many of the metadata fields are allowed togrow so that data sets subsequently docu-mented can use those previous descriptions.The database application divides metadata intofour kinds:
• Personnel information. Information about thepeople involved in the conception, design, im-plementation, and documentation of a studyand its data tables.
• Study-level information. General informationabout a study such as its title, abstract, pur-pose, dates, methods, and site characteristics.When a study is not part of a project, thestudy-level information will also include gen-eral information such as use constraints.
• Entity- and attribute-level information. De-tailed information about individual data ta-bles or files that contain GIS layers or im-ages. This is typically what computer sci-entists call metadata, but ecologists might callit table-level metadata.
Some existing tools (in particular MetaCatand Morpho17) allow metadata documentationand browsing of ecology data sets, but theywere not yet available when we needed them.We intend either to phase out our MS Accessdatabase application and use MetaCat or tohave our application produce data in the Eco-logical Markup Language, which MetaCat’s de-signers have defined.9
We developed an MS Excel spreadsheet pro-gram that lets users highlight a table in a work-sheet and then reads column headers as vari-able names from this table. Once a userhighlights a table, a new worksheet is created,and the user is queried for study- and table-level metadata. Metadata thus becomes avail-able in the spreadsheet with the data, and thespreadsheet metadata can later be uploaded tothe MS Access database. Another spreadsheetprogram uses these metadata to look for possi-bly erroneous data.
These applications are currently installed inBond’s lab, where they are being used by gradu-ate students. Working with a laboratory of co-operating researchers has let us experiment withmaking simple mechanisms for metadata provi-sion available early in the research cycle, yet didnot require researchers to drastically change how
they deal with their data. We gained importantinsights from working with Bond’s lab onwithin-lab metadata acquisition and archiving.However valuable database technology might befor documenting a data set for posterity, or evenfor linking it to collaborators’ data sets in an in-tegrative study, scientists will not use that tech-nology unless it increases individual researcherproductivity or (as in Bond’s lab) provides per-ceived benefit to a close-knit group of peers.Even in the latter case, data documentation toolsshould be specialized to the particular sciencepracticed. For example, in Bond’s situation, westage the metadata provision from very simpleand informal to more complex and generally ap-plicable. We now believe it is possible to con-duct rudimentary data validation using prelimi-nary metadata. We are helping the lab developlab-specific source tables for research informa-tion, keywords, research sites, and instrumenta-tion that are consonant with long-term archiv-ing. In short, this work has helped us betterunderstand how to specialize database tools forrelated ecological studies.
We have described the need fornew ecosystem informaticstools, the ecology research cy-cle, and two small projects at
different stages of that cycle. The first project, ahandheld (PDA) data acquisition tool, has nu-merous benefits and seems sustainable. Nextsteps would be to connect the tool to a databaseapplication that performs validation, visualiza-tion, and analysis at further stages of the re-search cycle, and to build tools that specializePDA forms for particular studies. The secondproject, an effort at within-lab metadata acqui-sition and archiving, shows that metadata pro-vision could be less onerous if accomplished instages. An obvious next step would be data vali-dation and cleaning at the lab level using thosemetadata, and transferring metadata from ourtool directly to data managers at longer-termdata archives, such as H.J. Andrews LTER. Bothof these enhancements to the within-lab projectare under way.
The two pilot projects have convinced us thatcurrent technology can help solve short-termproblems, but it can’t produce the integrateddatabase systems ecologists need for the future.Furthermore, these “one up” applications aregenerally not cost-effective for single researchstudies. The research that will deliver this future
42 COMPUTING IN SCIENCE & ENGINEERING
technology will require that ecologists and com-puter scientists work together—alternating be-tween development that uncovers and definesproblems, development that solves those prob-lems in particular contexts, research that gener-alizes key applications, and further work thattests research prototypes in new contexts.
We believe other scientific disciplines wouldbenefit from using the kinds of database tech-nology we describe but emphasize that thestate of technology is currently such that re-searchers need to learn database design be-forehand. Because few scientists want to be-come database programmers, we believe thatend-user database design tools and turn-keyapplications should be made available at thesubdiscipline level. A separate project underCushing’s direction at Evergreen is consider-ing ancillary problems of using database tech-nology to make easier the use of compute-in-tensive applications.
Finally, although improved researcher pro-ductivity is a necessary condition before ecolo-gists will use database tools, it might not be suf-ficient for widespread adoption of those tools.Moving systems such as those proposed hereinto the research cycle will inevitably involvesome changes in the way ecology is practiced.Although such sociological changes are beyondthe Canopy Database Project’s scope, our worksuggests that both ecologists and computer sci-entists will play key roles as these rewards are in-troduced in the scientific arena.
AcknowledgmentsWe acknowledge significant contributions of CanopyDatabase Project programmers and research staff ErikOrdway, Mike Ficker, Steve Rentmeester, Abraham Svoboda,Alex Mikitik, Youngmi Kim, Janet Rhoades, Peter Boonekamp,James Tucker, Brook Hatch, and Neil Honomichl. InformationManagers Don Henshaw and Gody Spycher at the NSF LTERH.J. Andrew site, as well as its former director Susan Staffordand the national LTER information management team,provided considerable help by providing metadata standardsand understanding of ecological information management.We thank our research collaborators (including, but notlimited to) Lois Delcambre, Jerry Franklin, Mark Harmon,Hiroaki Ishii, Betsy Lyons, Dave Maier, Robert Mutzelfeldt,David Shaw, Steve Sillett, Akihiro Sumida, and Robert VanPelt for freely sharing insight and data.
Ben Bloodworth, Jeff Heys, Patrick Boyne, Andrew Lee,Steve Sillett, Jim Spickler, Betsy Young, Emily Bearnhardt,and Matt Dunlap helped Roman Dial with design andfield implementation. Starling Consulting programmersand project managers Eugene Ryser, Erica Frandsen,
Porsche Everson, Jay Turner, and Bonnie Moonchild didmost of the programming on this device and madenumerous contributions to the project. Kate George andGeorgianne Moore are working with the in-lab metadataacquisition tool at Oregon State and providing helpfulmetadata, data, and suggestions for improving thesoftware. Travis Brooks ably programmed the MS Accessand Excel applications.
NSF grants BIR 9975510, 9630316, and 9300771,INT 9981531, and EIA 131952 and 75066 supportedthis work. Dial’s work was partially supported by theGlobal Science Society (GF 18-2000-114), and Bond’sby the H.J. Andrews Long Term Ecological Researchprogram and the US Department of Energy, through theWestern Regional Center of the National Institute forGlobal Environmental Change under CooperativeAgreement DE-FC03-90ER61010.
References1. US Nat’l Research Council, Finding the Forest for the Trees: The
Challenge of Combining Diverse Environmental Data (Selected CaseStudies), Nat’l Academy Press, 1995.
2. US Nat’l Research Council, Bits of Power: Issues in Global Accessto Scientific Data, Nat’l Academy Press, 1997.
3. M. Moffett, The High Frontier: Exploring the Tropical Rain ForestCanopy, Harvard Univ. Press, 1993.
4. M. Lowman and N. Nadkarni, Forest Canopies, Academic Press,1995.
5. N. Nadkarni and G. Parker, “A Profile of Forest Canopy Scienceand Scientists—Who We Are, What We Want to Know, and Ob-stacles We Face: Results of an International Survey,” Selbyana,vol. 15, 1994, pp. 38–50.
6. N. Nadkarni and J. Cushing, Final Report: Designing the ForestCanopy Researcher’s Workbench: Computer Tools for the 21st Cen-tury, Int’l Canopy Network, 1995.
7. D. Maier et al., eds., Report on a NSF, USGS, NASA June 2000Workshop on Biodiversity and Ecosystem Informatics, 2001;http://evergreen.edu/bdei2001, http://bdi.cse.ogi.edu, andwww.nsf.gov/cgi-bin/getpub?nst0199.
8. J. Cushing et al., Summary of VLDB Panel Database Research Issuesin Biodiversity and Ecosystem Informatics, Morgan Kauffman, 2002;www.cs.ust.hk/vldb2002/program-info/panels.html.
9. R. Nottrott, M.B. Jones, and M. Schildhauer, “Using XML-Struc-tured Metadata to Automate Quality Assurance Processing for Eco-logical Data,” Proc. Third IEEE Computer Society Metadata Conf., IEEECS Press, 1999; http://computer.org/proceedings/meta/1999.
10. W. Michener et al., “Non-Geospatial Metadata for the EcologicalSciences,” Ecological Applications, vol. 7, 1997, pp. 330–342.
11. G. Spycher et al., “Solving Problems for Validation, Federation,and Migration of Ecological Databases,” EcoInforma, vol. 11, 1996.
12. W. Michener, J.H. Porter, and S. Stafford, eds., Data and Infor-mation Management in the Ecological Sciences: A Resource Guide,LTER Network Office, Univ. New Mexico, 1998.
13. W. Michener and J. Brunt, eds., Ecological Data—Design, Man-agement and Processing, Blackwell Science, 2001.
14. J. Cushing et al., “Template-Driven End-User Ecological DatabaseDesign,” Proc. 6th World Multiconference Systemics, Cybernetics andInformatics, vol. 7, Int’l Inst. Informatics and Systemics, 2002, pp361–366; http://216.72.45.230:1081/ProceedingSCI/index98.htm.
15. J. Cushing et al., “Designing Ecological Databases with Compo-nents,” to be published in a Special Issue of J. Intelligent Infor-mation Systems, J. Schnase and J. Smith, eds., 2003.
16. J.H. Porter, D.L. Henshaw, and S. Stafford, “Research Metadata inLong-Term Ecological Research (LTER),” Proc. IEEE MetadataConf., IEEE CS Press, 1997.
17. P. McCartney and M. Jones, “Using XML-encoded Metadata asa Basis for Advanced Information Systems for Ecological Re-search,” Proc. 6th World Multiconference Systemics, Cyberneticsand Informatics, vol. 7, Int’l Inst. Informatics and Systematics,2002, pp. 379–384.
Judith Bayard Cushing is a member of the faculty incomputer science at The Evergreen College. Her re-search interests include software engineering anddatabase infrastructures for scientific applications. Shereceived her MA in philosophy from Brown Universityand a PhD in computer science and engineering fromthe Oregon Graduate Institute. She is a member of theACM and the IEEE. Contact her at The Evergreen Col-lege, 2700 Evergreen Parkway NW, Olympia, WA98505-0002; [email protected].
Barbara Bond is an associate professor in the Depart-ment of Forest Science at Oregon State University. Herresearch interests include studies of physiologicalprocesses associated with aging in forest trees and theuse of carbon isotope discrimination by ecosystems asa tool for understanding ecosystem function. She hasan MS in terrestrial ecology and a PhD in plant physi-
ology and forest science, both from Oregon State Uni-versity. She belongs to the Ecological Society of Amer-ica. Contact her at the Dept. of Forest Science,Richardson 313, Oregon State Univ., Corvallis, OR97331; [email protected].
Roman Dial is an associate professor in the Depart-ment of Environmental Science at Alaska Pacific Uni-versity. His research interests include canopy scienceand theoretical ecology. He holds an MS in mathe-matics from the University of Alaska Fairbanks and aPhD in biological science from Stanford University. Hebelongs to the Ecological Society of America. Contacthim at the Dept. of Environmental Science, Alaska Pa-cific Univ., 4101 University Dr., Anchorage, AK 99508;[email protected].
Nalini Nadkarni is a member of the faculty in envi-ronmental studies at The Evergreen State College. Herresearch interests include the ecology of tropical andtemperate forest canopies and the ecosystem process.She received a PhD in forest ecology from the Univer-sity of Washington. She belongs to the Ecological So-ciety of America. Contact her at The Evergreen Col-lege, 2700 Evergreen Parkway NW, Olympia, [email protected].
2003 MRS FALL MEETING
Exhibit and
Integrated Device Technology
A: Micro- and NanosystemsB: Materials, Integration, and Packaging Issues for
High-Frequency DevicesC: Ferroelectric Thin Films XIID: Materials and Devices for Smart SystemsE: Fundamentals of Novel Oxide/Semiconductor Interfaces
Organic, Soft, and Biological Materials
F: Biomaterials for Tissue EngineeringG: Molecularly Imprinted MaterialsH: Biological and Bio-Inspired Materials AssemblyI: Biomaterials for Drug DeliveryJ: Interfaces in Organic and Molecular ElectronicsK: Functional Organic Materials and Devices
Nano- to Microstructured Materials
L: Continuous Nanophase and Nanostructured MaterialsM: Nontraditional Approaches to PatterningN: Quantum Dots, Nanoparticles, and NanowiresO: Nanostructured Organic MaterialsP: Dynamics in Small Confining Systems VIIQ: Mechanical Properties of Nanostructured Materials
and Nanocomposites
Inorganic Materials and Films
R: Radiation Effects and Ion Beam Processing of MaterialsS: Thermoelectric Materials 2003—
Research and ApplicationsT: Self-Organized Processes in Semiconductor
HeteroepitaxyU: Thin Films—Stresses and Mechanical Properties X
Photonics
V: Critical Interfacial Issues in Thin Film Optoelectronic and Energy Conversion Devices
W: Engineered Porosity for Microphotonics and PlasmonicsY: GaN and Related AlloysZ: Progress in Compound Semiconductor Materials III—
Electronic and Optoelectronic Applications
Energy Storage, Generation, and Transport
AA: Synthesis, Characterization, and Properties of Energetic/Reactive Nanomaterials
BB: Materials and Technologies for a Hydrogen EconomyCC: Microbattery and Micropower SystemsDD: Actinides—Basic Science, Applications, and TechnologyEE: Frontiers in Superconducting Materials—
New Materials and Applications
Information Storage Materials
FF: Advanced Magnetic NanostructuresGG: Advanced Characterization Techniques for
Data Storage MaterialsHH: Phase Change and Nonmagnetic Materials for
Data Storage
Design of Materials by Man and Nature
X: Frontiers of Materials ResearchII: The Science of Gem MaterialsJJ: Combinatorial and Artificial Intelligence Methods in
Materials Science IIKK: Atomic Scale Materials Design—Modeling and SimulationLL: QuasicrystalsMM:Amorphous and Nanocrystalline Metals
SYMPOSIA MEETING ACTIVITIES
Symposium Tutorial Program
Available only to meeting registrants, the symposiumtutorials will concentrate on new, rapidly breakingareas of research.
Exhibit and Research Tools Seminars
A major exhibit encompassing the full spectrumof equipment, instrumentation, products, software,publications, and services is scheduled forDecember 2-4 in the Hynes Convention Center,convenient to the technical session rooms.Research Tools Seminars, an educational seminarseries that focuses on the scientific basis andpractical application of commercially available,state-of-the-art tools, will be held again this fall.
Publications Desk
A full display of over 775 books, plus videotapesand electronic databases, will be available at theMRS Publications Desk.
Symposium Assistant Opportunities
Graduate students planning to attend the 2003MRS Fall Meeting are encouraged to apply for aSymposium Assistant (audio-visual assistant)position.
Career Center
A Career Center for MRS meeting attendees will beopen Tuesday through Thursday.
www.mrs.org/meetings/fall2003/
The 2003 MRS Fall Meeting will serve as a key forum for discussion of interdisciplinary leading-edge materials research from around the world.Various meeting formats—oral, poster, round-table, forum and workshop sessions—are offered to maximize participation.
Abstract Deadlines — In fairness to all potential authors, late abstracts will not be accepted.June 5, 2003: for abstracts sent via fax or mail � June 19, 2003: for abstracts sent via the MRS Web site
