Top Banner
ISSN: 1361-8962 Lava Seal L o w Too Low Unexplored Bouldery Passage The Maze 0 25 50 100m Scale Climb 2m L o w N GRID Cave surveyed to BCRA Grade 4 by members of the Shepton Mallet Caving Club during the Laki Underground Expedition, July 2001. Climb 2m Daylight visible through boulders Blue & White Formations Visual Connection Main Entrance Lava Formations Surveying Icelandic lava tubes Describing survey quality Aerial photos from the Millennium Atlas The Journal of the BCRA Cave Surveying Group July 2003 Issue 31
12

Issue 31 July 2003 - CHAOS.ORG.UK

May 22, 2022

Download

Documents

dariahiddleston
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Page 1: Issue 31 July 2003 - CHAOS.ORG.UK

ISSN: 1361-8962

Lava Seal

L o w

Too Low

Unexplored Bouldery Passage

The Maze

0 25 50 100m

Scale

Climb 2m

L o w

NGRID

Cave surveyed to BCRA Grade 4 by members ofthe Shepton Mallet Caving Club during the LakiUnderground Expedition, July 2001.

Climb 2m

Daylight visible through boulders

Blue & WhiteFormations

VisualConnection

Main Entrance

LavaFormations

Surveying Icelandic lava tubes

Describing survey quality

Aerial photos from the Millennium Atlas

The Journal of the BCRA Cave Surveying Group

July 2003

Issue 31

Page 2: Issue 31 July 2003 - CHAOS.ORG.UK

COMPASS POINTS INFORMATIONCompass Points is published three times yearly in March, July andNovember.TheCaveSurveyingGroupis a SpecialInterestGroupof theBritish Cave ResearchAssociation.Information sheetsabout the CSGareavailableby post or by e-mail. Pleasesendan SAE or PostOfficeInternational Reply Coupon.

NOTES FOR CONTRIBUTORSArticlescanbeon paper,but thepreferredformatis ASCII text files withparagraphbreaks.If articlesareparticularlytechnical(i.e. containlots ofsums) then Latex, OpenOffice.orgor Microsoft Word documentsareprobablybest.We areableto copewith manyother formats,but pleasecheckfirst. We can acceptmost commongraphicsformats,but vectorgraphicformatsaremuchpreferredto bit-mappedformatsfor diagrams.Photographsshould be prints, or well-scannedphotossuppliedin anycommonbitmapformat.It is theresponsibilityof contributingauthorstoclear copyright and acknowledgementmatters for any materialpreviously published elsewhereand to ensure that nothing in theirsubmissions may be deemed libellous or defamatory.

COMPASS POINTS EDITOR Anthony Day, 11 Claire Court, St Martins Drive, Walton on Thames,Surrey, KT12 3BW. Tel: 01932 226099E-mail: [email protected]

SUBSCRIPTION & ENQUIRIESAndrew Atkinson, 31 Priory Avenue, Westbury-on-Trym,BRISTOL,BS9 4BZ Tel: 0117 962 3495E-mail: [email protected]

PUBLISHED BY The Cave Surveying Group of the BCRA. BCRA is a registered charity.

OBJECTIVES OF THE GROUP The group aims, by meansof a regularJournal,otherpublicationsandmeetings,to disseminateinformationabout,anddevelopnewtechniquesfor, cave surveying.

COPYRIGHT Copyright© BCRA 2003. The BCRA ownsthe copyright in the layoutof this publication. Copyright in the text, photographsand drawingsresideswith the authorsunlessotherwisestated. No material may becopied without the permission of the copyright owners. Opinionsexpressedin this magazineare those of the authors, and are notnecessarily endorsed by the editor, nor by the BCRA.

SUBSCRIPTION RATES (FOUR ISSUES)U.K. - £4.50 Europe - £6.00 World - £8.00Theseratesapply regardlessof whetheryou area memberof theBCRA.Actual “membership”of theGroupis only availableto BCRA members,to whomit is free.You canjoin theBCRA for aslittle as£5.00– contactBCRA directly for details.Sendsubscriptionsto the CSGsecretary(see“subscriptionsandenquiries”).Chequesshouldbe drawnon a UK bankand payable to BCRA Cave Surveying Group. EurochequesandInternationalGirobankpaymentsare acceptable.At your own risk youmay sendUK banknotesor US$ (add20% to currentexchangerateandcheckyou don’t haveobsoleteUK banknotes).Failing this your bankcan“wire” direct to our bankor you canpay by credit card,if overseas.In boththesecaseswe haveto paya commissionandyou shouldpay5%extra to cover this.

BACK ISSUESPast issuesof Compass Points are available from the secretary(see“subscriptionsandenquiries”)subjectto availability. Cost is £1.25 perissue,pluspostageandpackingat ratesof £0.50(UK), £1.50(Europe)or£3.00(world). Publishedissuesare also accessibleon the Web via theCSG Web pages at http://www.bcra.org.uk/csg/

DATA PROTECTION ACT (1984)Exemption from registration under the Act is claimed under theprovision for mailing lists (exemption6). This requiresthat consentisobtainedfor storageof the data,and for eachdisclosure.Subscribers'namesand addresseswill be storedon computerand disclosedin anaddresslist, availableto subscribers.You must inform us if you do notconsent to this.

COMPASS POINTS LOGOcourtesy of Doug Dotson, Speleotechnologies.

CAVE SURVEYING MAILING LISTTheCSGrunsane-mail list for cavesurveyorsaroundtheworld. To joinsend a messagecontaining the word ‘subscribe’ in the body text [email protected]

CONTENTSof Compass Points 31

The journal of the BCRA Cave Surveying Group

� Editorial..............................................................................2

� CSG Admin.........................................................................3

� Forthcoming Events.........................................................3Hidden Earth 2003

� Snippets..............................................................................3Instrument Problems at Altitude: Update

Wookey

� Press Round-up................................................................3Compass and Tape issue 52 (Dec 2002)

Reviewed by WookeyCREG Journal 52

Reviewed by David Gibson

� Letters.................................................................................4Compass Sight Errors

Bob ThrunBCRA Grade Definitions

Peter Cousins & Dave Irwin

� Laki Underground Expeditions to Iceland....................5Phil Collett, Martin Ellis & Ed Waters

A casestudy of a surveying project in Icelandic lavatubes, including the use of GPS to overcome some of theproblemsof surveyingin regionswith largemagneticanomaliesand integrating the surveying work with geophysical observations.

� Describing Survey Quality...............................................8Bob Thrun

Following the recent debateon the definition of theBCRA surveygrades,Bob presentsa discussionof thestatisticsofloop closure errors and offers his thoughts on the practicalapplication of grading schemes.

� Printable Mapping from the Millennium Atlas............12Andy Waddington

How to obtain aerial photographsof speleologicallyinteresting parts of Britain for free.

Cover image: Example of a CAD generated survey from the LakiUnderground Expeditions (see article starting on page 5).

EditorialOnceagain,this editorial is a grovelling apologyfor the latearrivalof an issue of CompassPoints. On this occasion,it was 99%completewhen I disappearedoff on expeditionfor threeweeks-hencethedelay. I really will try to bemoretimely in future,thoughproducing the next issue is likely to be a little fraut (see admin. item)– soif anyoneis sitting on anypotentialarticles,pleaselet meknowabout them in good time.

2 BCRA Cave Surveying Group, Compass Points 31, July 2003

Page 3: Issue 31 July 2003 - CHAOS.ORG.UK

CSG AdminThe editor is on the move. The editor's postal addressin themastheadwill ceaseto work on 13th September,andat the time ofgoingto pressI don'tknow exactlywhereI will beafterthat. Pleasedirect any paper correspondenceto the “Subscriptions andEnquiries”addressafter this date. However,email canstill be sentto the [email protected] address.

Forthcoming Events

Hidden Earth 2003The UK's annualnationalcavingconferencewill be held at HanleyCastleHigh School, Upton on Severn(near Worcester)over theweekend 3-5 October. Further details can be found at theconference website, http://www.hidden-earth.org.uk

The Arthur Butcher award is judged and presentedby BCRAannually at the conferencefor, broadly speaking,“excellenceincavesurveying.” Thereis a cashprize anda trophy to be kept bythe winner(s) for a year. To be considered,individuals or cavingclubsmustbring their work to the attentionof the judges. This canbeachievedby displayingyour work at theconference.If you wantotherwork to be considered– suchasa reportor publicationon asurveying topic, or other more generalachievements– then youshouldcontactthejudgesin advance.Full detailsof therulesof theawardandnominationprocedurecanbe found on the HiddenEarthwebsite.

Snippets

Instrument Problems at Altitude: UpdateWookey

In thelast issueof CompassPoints,I describedtheproblemsvarioussurveyorshavehadwhenusingtypical cave-surveyingcompassesathigh altitude. In summary,basedon practicalexperienceit appearsthat Silva Clinomasterand Sightmastercan developbubbleswhenusedat altitudesin excessof ~1700mfor a period of more than aweek or two, rendering them unusablein extremecases. Suchbubblesarise becausethe capsuleexpandsslightly in the lowerpressurebut the volume of liquid inside remainsalmost constant,hence a bubble forms. Suunto instrumentsand Silva Type 80compasses (the plastic-bodiedprismatic ones)appearto be lesssusceptibleto this problem. It is also strongly advisedthat theinstrumentsbecarriedin cabinbaggageonaeroplanesto avoidthembeing subjected to extremely low pressures in the hold.

Since this article was printed, some more information has beenreceivedfrom the manufacturers.The Silva technicalsupportstaffstatethattheacceptedaltitudeat which bubblesoccuris 5000m,andthat such bubbles will only becomepermanentif taken above12000m(in an unpressurisedaeroplanehold, for example). Theyaccept that our experiencediffers significantly from the designspecifications. Suunto have also been contacted,but the onlyinformationcurrentlyavailableis for their wristopcomputers,whichshouldbe ok up to 9000m. A requestfor information abouttheircompassesis working its way throughthe system. As I saidin thelast issue,any feedbackon your experiencesof the susceptibilityofvarious compasses to this problem would be appreciated.

Press Round-up

Compass and Tape issue 52 (Dec 2002)Reviewed by Wookey

Call for papers at the 2003 NSS convention, Porterville,California.

Minutes of 2002 Survey and Cartography Section meeting

Held at 2002NSSconvention- 28th June. The sectionhasnearly$5000and211 members.Therewere4 talksat theconferenceand14 entriesin the CartographicSalon. They had not producedtheintended Cart Salon special issue of C&T as intended. Putting Salonmaps online, the south-easternregion salon, and the possibledefinition of computer-drawnmaps was discussed. Next year'sconventionwill havea computerworkshopof digital mapcreationand the possibility of a children'ssurveyingclasswas considered.Robin Barber was voted secretary.

2002 Cartographic Salon Results - Steve Reames

20 mapswere entered- 6 being “display only”. 8 mapsreceivedawards. Cavesof the SnakeWell Complex,by Brent Aulenbachwon the Best of Show Medal.

Development and testing of three components of the processof transferring Digital Cave Survey Data - Mike Yocum

Mammoth Cave National Park (MCNP) neededto integratetheirFlint Ridge Mammoth Cave survey data (collected by the CRF(CaveResearchFoundation)over 40 years)with the Arcview GISsystemfor park/karstmanagement.Thusa projectwasstartedto: 1)develop and test proceduresfor converting CRF survey data toarcviewformat;2) providemetadatafor this info. (i.e. how accurateit is); 3) develop a framework and proceduresfor recording thecontent and statusof CRF data as it is conveyedfrom CRF toMCNP.

A teamof expertswas assembledanddigital dataassembledfor atest areain suitableformatsand datumsfor Arcview. The surveydatawasin Compass,CML, SMAPSandWalls usinga specialCRFco-ordinatesystem. The test areaincludedboth undergroundandsurface benchmarks which provided suitable control points.

Walls exportsArcview Shapefilesdirectly, Compassfiles can beconverted to Arcview using the CaveTools Arcview extension.SMAPSandCML dataneedto beconvertedinto Compassor Wallsfirst. Themostcompletedatasetwasin SMAPSso it wasimportedinto both Walls andCompass.Versionswerealsocreatedwith theCRF co-ordinatesystemconvertedto UTM, NAD27 metres. Allfour datasets were saved as shapefiles.

Meta data were generatedfrom discussionamongstthe team andfrom the resultsof registeringthe dataagainstthe Arcview surfacedata. The Shapefileformat was extendedto include extra surveyattributes. The framework for tracking data is to referenceeachstation to the relevant map sheet in the CRF Mammoth survey.

A discussion of the process and results follows. The mostsignificantbeingthat theSMAPS->ShapefileconversionprocessviaWalls worked OK, but via Compassand Cavetoolsthe historicentrancemovedby over3000ft,dueto a differencein the constantsusedin SMAPS,CompassandCavetoolscombinedwith a roundingerror in Cavetools.

BCRA Cave Surveying Group, Compass Points 31, July 2003 3

Page 4: Issue 31 July 2003 - CHAOS.ORG.UK

It alsobecameclearthat thevariousco-ordinatesystemconversionsand different surfacedata feature formats (DOQQ, DRG, DLG,DEM) werenot in agreementandtherewerevariousdiscrepanciesof up to 100ft. You need to be very careful when converting data.

Oneresultof this work hasbeenanimprovedshapefileformat- thisis published as an appendix.

Obituary - Mike Yocum

Mike diedshortlyaftersubmittingtheabovearticle. He wasa keencave-surveyorand project caver in Kentucky and Tenessee,beingparticularly active with the CRF and working for MCNP in thenineteen nineties.

CREG Journal 52Reviewed by David Gibson

Although it was not billed as a “special” on radiolocationor cavesurveying, the latest journal from the BCRA's Cave Radio andElectronicsGroup devotes18 of its 36 pagesto thesetopics in sixarticles.

Aerial photographyof the entireUK is increasinglyavailable,bothelectronicallyand as hard copy. Imagery from other countriesisalsoavailable,eitherfrom aircraftor satellite. In A Caver'sViewofRemoteImaging Mike Bedford looks at what is on offer and thepossiblebenefitsto cavers. Unfortunately,a gremlin in the wordprocessormeantthat two of thephotographswerere-sizedandtheircaptionsconsequentlydisappeared.However,acorrectedversionofthe article is available at:http://bcra.org.uk/cregj/j52-pp0507.pdf

At the CREG field meeting in March, Wookey gave a talk onElectronic Cave SurveyingInstrumentshighlighting, in particular,how CREGmemberscouldhelpto designequipment.His article inCREGjournal 52 is an extendedversionof that talk. It givessomebackgroundfor non-surveyors,and outlines the pros and cons ofseveraltypesof electronicinstruments. It alsogivesa “time line”showingwho hasbeenworking on whatover the years,andgivesalist of further reading. CREG membersare scoutingaround forprojectsat the moment,andtherewould seemto be plentyof waysfor peopleto get involved in a project to help to “automate”cavesurveying.

The currentstateof the art in professionalsurveyingequipmentisthe 3D laser scanner, described by Mike Bedford in FirstImpressionsof a 3D LaserScanner. Theheartof a 3D scanneris alaser-baseddistancemeasurementdevice. However, instead ofmeasuringthedistanceto singlepointsasdefinedby the user,a 3Dscannerautomatically builds up a three-dimensionalimage byscanning the scene both vertically and horizontally. Data arecollectedautomaticallyasa so-called3D point cloud oncethe areaof interestandthescanningintervalhavebeenspecifiedby theuser.Applications include quarrying and architecture.

An obviouscaving applicationof a 3D laserscanneris surveyinglargecavechambers.With conventionalcavesurveyingtechniquesit may be possible to do little more than survey the perimeter. Usinga 3D scanner,though, it is possible to build a complete three-dimensionalrepresentation.Mike describeshow heusedthe LaserAce Scanner from MDL (Measurement Devices Limited)(http://www.mdl.co.uk, http://www.laserace.com)in Yordas Cave. The article includes a graphic showing therendered view of one wall of the chamber.

The MDL instrument is a combined 3D laser scanner andreflectorless“total station” (to usetheprofessionaljargon). It costs£20,000,including the softwarefor viewing 3D point clouds. Thiscompares favourably with a typical price of £15,000 for areflectorlesstotal station and between£60,000and £100,000formost3D laserscanners.Mike alsodescribesMDL's LaserAce300which is a hand-helddevice that takesa single readingof range,bearingandelevation. Its useasa cavesurveyinginstrumentis self-

explanatory,andat around£2400it is morewithin reachof a cavinggroup, although not by much.

David Gibson presentsa ground-breakingpaper [well, it is mewriting this review!] on radiolocation showing how we candetermine the bearing, elevation and distance of a beacon transmitterby making measurementsat a single radiolocationreceiverstation.Conventional radiolocation requires two stations in the samehorizontalplane,with oneof thembeingat “groundzero”. Clearlyfor a generalised3D method(i.e. a “global positioning”method)wecannotrely on this beingpossible,andso the single-stationmethodwould seemvery attractive. In 3D RadiolocationUsing a SingleStationDavid describesthe limitationsof themethod,andtheerrorsthat are introducedwhen skin depth (a measureof the electricalconductivity of the ground) is taken into account. Very fewdiscussionsof radiolocationconsiderskin depth, which can be asignificant cause of loss of accuracy. David also explains why “timeof flight” methodsthatrely on timing thearrival of radiowaves(i.e.like true GPS) will not work in a sub-surface setting.

The “single station” methoddescribedin the abovearticle dependson an accuratelycalibratedtransmitter,unlessa ratiometricmethodis used(David suggestsmeasuringthe gradientof the vertical fieldcomponent). However, this complication can be avoided by taking asecondreceiverreading. This techniqueis describedby RichardRushton in Towards 3D Radiolocation Using Two Stations, in whichhe explains how readings from two receiver stations can becombined to give position information without an accuratelycalibratedtransmitter,providedthat the relativepositionof the tworeceiver stations is known and they are in the same horizontal plane.

As a footnote to this article David Gibson then observesthatRichard'smethodcanbe generalisedto any two receiverlocations,and he suggestsa further logical step,which is to usethe “singlestation” methodto surveyan entire system,and then to scaletheresulting 3D survey in order to fix some known points. Thistechniqueavoids the complicationof a calibratedtransmitterandreceiver,and it also avoidsthe needto fix pairs of stationsduringthe “two station” method.

Both thesearticles describea techniquethat soundsattractive intheoryalthough,in practice,therearelimitationson how usefulit islikely to be. Clearly further work is needed.

David Gibson'sBibliographyof CaveRadiolocationis an extendedversionof the article that appearedin CompassPoints30 and lists30 referencesto the subject,most of them by David and most ofthem from the CREG journal.

The remainderof CREGjournal 52 containsseveralarticleson theHeyPhonesub-surfaceradio system,aswell as the usualcollectionof news,lettersandshortarticles. A contentslist for CREGjournalsis availableon their web site at http://bcra.org.uk/creg/and back issues can be ordered from:http://caves.org.uk/payments/creg/

Letters

Compass Sight ErrorsBob Thrun

JosBurgerspresenteda nicely balancedsetof compassreadingdata(in CP30– ed.) It is easyto separatethe effectsof compassesandcompassreadersbecauseeverybodyread every instrument. Themain reasonthat compasserror was much more than readererrorwasthatCompassC wasdifferent from theotherfive compassesbyabout 3 grads. If you remove it from the mix, there is about as muchvariation in readersas there is in compasses.Reader2 had everyreadingbelow average. Reader5 was generallyhigh. Reader11was very closeto the averageand very consistent. Reader6 wasinconsistent.Thereaderswho readto a tenthof a gradshowedlessvariation than those who usually read to integer values.

4 BCRA Cave Surveying Group, Compass Points 31, July 2003

Page 5: Issue 31 July 2003 - CHAOS.ORG.UK

BCRA Grade DefinitionsPeter Cousins & Dave Irwin

We get the strongimpressionon readingissueCompassPointsNo.30 thatbothJohnStevensandtheauthorsof therecentCaveStudiesbooklethavemisunderstoodByranEllis' useof theterm“accuracy”in relation to survey grades.

Although we often felt that Bryan's definition of the differencebetween “precision” and “accuracy” (the nearness of the result to thetrue value)satuncomfortablywhen usedfor magneticcavesurveyreadings;thedistinctionwasneverthelessa usefuloneand,sofar aswe know, thereis no otherword in the English languagewe coulduse.

However,we areconfidentthat the distinctionwhich Bryan wishedto makein introducingtheterm“accuracy”for GradeV andVI wassimply that the surveyinstrumentshad beencalibratedso that thereadingscould bereferredto somelocal absolutestandard[1]. It isno coincidencethat both Bryan Ellis and Denis Warburtonweretrainedas laboratorychemistsand usedcalibratedinstrumentsandglassware on a daily basis in their work.

In part the introduction of “calibration” for Grade V was alsoprobablya reactionto the low standardof manyof the thencurrentGrade IV surveys, which were often published with a simplemagneticNorth arrow and without any elevationsor real altitudedata.

Unfortunately John Stevens has also misunderstoodthe entiregradingschemeby suggestingthat the gradingshouldrefer to theaccuracyof the entiresurvey. It doesnot. The graderelatesto theprecision with which the instrumentshave been read (we areincluding tape and station position as an instrument here) andadditionally, by suffix, to the quality of detail recorded (see below).

Implicitly thesurveygradingtells usnothingaboutthefrequencyof“blunders” or other unpredictabledisturbances;although in thehigher gradesadequateattentionto “before and after” calibrationwill allow the pooling of datafrom severalinstrumentsandlargely

eliminate the possibility of errors due to magneticdisturbanceorinstrument damage during the survey (e.g. Cousins [2])

JohnStevensseemsunawareof the long history of work by (forexample) Denis Warburton and Dave Irwin regarding therelationshipbetweenobservedloop closuresandsurveygrades(e.g.Irwin & Stenner[3]), but we aresurprisedthat hefalls into the trapof thinking that it is alwaysdesirableto surveyeverypassageto thesamegradeasthe main surveyso asto introduceasmanyloopsaspossible. The traditional OrdnanceSurveypractisewas to reservetheir main 36 inch theodolitesfor the primary triangulationandfillin thesecondaryandtertiarynetworkswith smaller,but lessprecise,24 or 18 inch instruments.

Oneproblemwith all thesediscussionsis thatauthorsrarely,if ever,definewhat is “the finishedsurvey”. Until thedigital ageit wasthepublishedplan,elevationsandotherinformationasthatwasall thatcould generallybe available. Now it sometimesseemsthat a tableof numberson someessentiallytransientwebsiteis all that matters.Either way the scaleof final plotting largelydefinesthe “accuracy”of the finished map(e.g. Irwin [4]); althoughwe shouldnot forgetthatuntil recentlytheOrdnanceSurveyonly claimed30mfor minorfeaturesin rural areas- so even the most accuratecave surveyswould not necessarilysuperimposecorrectlyon OSmappedsurfacefeatures.

References

[1] Ellis, B.M. (1976).Surveyingcaves,[Bridgwater:British CaveResearch Association.] 88pp.

[2] Cousins,P.R. (1980). Calibration of cave survey instruments,Transactions of the British Cave Research Association, Vol.5(2), 85-90.

[3] Irwin, D.J. & Stenner,R.D. (1976). Accuracy and closuresoftraversesin cave surveys, Transactions of the British CaveResearch Association, Vol. 2(4), 151-166.

[4] Irwin, D.J. (1970). The presentation of cave surveys,Transactions of the Cave Research Group, Vol. 12(3), 155-164.

Laki Underground Expeditions to IcelandPhil Collett, Martin Ellis & Ed Waters

This article describes the surveying activities on two expeditions to Iceland undertaken by BournemouthUniversity with the assistance of the Shepton Mallet Caving Club. This work earned the LakiUnderground team, and Chris Woods and Ed Waters in particular, the Arthur Butcher Award for 2002.

The Laki UndergroundExpeditionsof September2000 and July2001exploredandsurveyedover 11km of lava tubecavepassagesin the Eldhraunlava flow of southernIceland. In addition to this“traditional” rationalefor theproductionof cavesurveys,surveyingactivitieswerecarriedout to supportsomeof the specificscientificaims of theseexpeditions,both in the Eldhraunlava flow and thewell known caves of the Hallmundahraun in western Iceland.

A key featureof thework undertakenwasthesuccessfulintegrationof undergroundsurveyingwith GPSmeasurements.Therefore,thisarticle beginswith a generaldiscussionon thepracticaluseof GPSfor cavingexpeditions,beforegoing on to describethe work of theLaki expeditions.

GPS and Caving ExpeditionsIn the last few yearssalesof handheld GPSsystemshavesoared.Virtually every high streethasat least one shop that stocksunitsmanufacturedby thelikes of GarminandMagellan. Typically thesesystemsarecreditedwith an accuracyof about15m, thoughthis issubject to a degraded performance of 100m when the USDepartmentof Defence Selective Availability (SA) program isrunning. It is only sinceSA wasswitchedoff in 2000thathandheld

GPSbecamea viabletool for cavingexpeditions.In addition,manyhandheldGPSsystemshaveaveragingfunctionsthatallow themtogeta higheraccuracyin areaswith poorsatellitecoverage,which isa most valuablefunction and is highly desirableon units usedforcave survey work. In thesecasesthe accuracyof the locationimprovesthelongertheunit is left stationary– if theunit canbeleftfor severalhoursanEPE(EstimatedPositionalError – anestimatedroot mean square error in position) of below 5m can be obtained.

At its simplest,GPSis usefulto a cavingexpeditionasa quick andeasy way of recording the location of cave entrancesand otherspeleological features as waypoints. Experience has shown,however, that care needsto be taken when using a GPS in thismanner. In areaswith many cave entrancesit is vital that anotebookis usedto recordwhatis at eachwaypoint. Onceyou havelogged40 or 50 featuresin a day, it canbecomeincrediblydifficultto rememberwhy you recordednumber23 for instance.Noting thewaypoint name(a coding for waypoint namescan be useful also),locationanda descriptionof the featurecanhelp prioritise the nextday’s work and also guardsagainstbattery failure and accidentalwaypoint deletion.

BCRA Cave Surveying Group, Compass Points 31, July 2003 5

Page 6: Issue 31 July 2003 - CHAOS.ORG.UK

As a minimumthis techniqueshouldbeaccurateenoughto relocatethefeatureloggedat a laterdate. In caseswherethe featuresaresoclosetogetherthat it is difficult to determinewhich is which, thewritten descriptioncanbea big help. Themoralto this storyis that,if you do havesomefeaturesvery closeto eachother (say20m orless), it is well worth making a more comprehensiveset ofdescription notes about them.

GPStrackfunctionscanbeusedto aid determinationof thesizeandshapeof surfacefeatures. For instancea depressioncanbe quicklymeasuredby walking aroundits rim. This canbea very usefulaidto help drafting a survey,but this techniquerequiresa GPSwith alargememoryor a laptop to downloadthe informationonto. GPStracking functionscan also be usedto recordpathsand tracksnotmarkedon maps,simply by carryingthe unit with the party, whichcan be a useful aid for other parties venturing to the cave inquestion. However, some accuracy problems were encounteredwhen mappinglava channelsin Icelandusing the track facility onthe Garmin GPS76. It transpiresthat the GPS76 “throws away”trackpointswhenthe track is savedasa namedtrack (maximumof128 points in a savedtrack), but maintainsall track information inits current track. The workaroundsolution for the GPS76 whenusing it to surveysurfacefeaturesusingtracking is to neversaveatrackusingtheGarminfacility, but to downloadthecurrenttrack toa computer. This may work with other Garmin models,and iscertainly the case with the Etrex.

For GPS-derivedentrancefixes to be of anyusein thecavesurvey,they must be properly tied in. In order for this to occur the bestmethod is to make the place that the best GPS fixes are obtained intoan externalsurveystationandthentreat it exactlythe sameway asyou would treat a survey leg underground. For multi-entrancesystems, the GPS fixes can be useful as a check that theundergroundsurveyis not grosslyin errorsinceit is quitelikely thata competentparty will producesurvey data that is more accuratethan the entrancefixes. For example,a 1% error (a good rule ofthumbindicatorfor a fair cavesurvey)over1000mof cavepassagegives a positional uncertaintyof 10m whilst a best possibleGPSEPEof about5m alsogivesa positionaluncertaintyof at least10m.Similarly, if two or morecaveentrancesareclosetogether,saylessthan300m apart,it will be more accurateto surveybetweentheseentrancesthan to use GPS readingsat eachentrance. If suitablecomputing facilities are availableon expedition,thesetechniquescanexposeerrorsin the field astheyoccurwhich allows themto berectified if necessary or the dodgy surveying redone.

In caseswherethecaveentranceis in acliff or in undergrowth(botharevery commonin the authors'experience),it is often well worthtaking the GPS reading at the closestpoint to the cave that theminimum EPEcanbe obtained. The exactlocationof the entrancecan then be determinedby the use of normal cave surveyingtechniques.

In theopinionof the authorsit is bestpracticeto marka permanentexternalstationon the surface. Not only is this availablefor useinsurfacesurveysor repeatsections,but it allows eachparty enteringthe caveto pauseto takea GPSfix. If thesefixes are recordedaseriesof themcangive a betterideaof the accuracyof the surfacecoordinates.

Finally, if the GPSdataareto be integratedwith surfacemaps,it isimportantthat you recorddetailsof the datumin which your GPSco-ordinateswere recordedand the datum of the surfacemap toassistyou in any co-ordinatetransformationthat may be required.These issueshave been discussedin detail in recent issuesofCompass Points and will not be repeated here.

Differential GPSEven if Selective Availability is turned off errors in indicatedposition can be causedby severalfactors, the greatestone beingdisturbancesto theGPSsignalby theionosphere.Errorsin theGPSsystem,includingSA, canbegreatlyreducedby usinga differentialGPS(DGPS)system. A fixed basestationGPSanda roving GPSare used. The samesatellite errors apply to both GPSs,so the

difference of position betweenthe two stations is known to anaccuracythat eliminatesmosterror. A DGPSsystembasedon thetechnology employed in hand-held units would typically beexpected to give an accuracy of between 1 to 3 metres.

One advantageof the cooperationwith BournemouthUniversityduring one of the expeditionsto Iceland was accessto a surveyquality DGPS system,the Leica GPS system500, togetherwithpeopleskilled in its use. This systemis ableto processandstoretheGPSsignalsin a far more sophisticatedway than hand-heldunits.Although having a precisionan order of magnitudegreaterthanhandheld units (measurementsto the nearestcentimetreshouldbepossible),thesesystemsare many times moreexpensive. As withotherdifferentialsystemstheaccuracyis relativeto thebasestation.It is importantto beableto replacethebaseunit in exactlythesameposition to within one centimetre.A typical Leica systemwouldconsistof a baseunit on a tripod (Figure1) anda roving unit with abatterypackin a small rucksackandtheGPSunit fixed to a surveystaff. In addition to very accuratelocation of cave entrancestheLeicawasusedto lay out gridsfor geophysicalsurveys,for exampleto searchfor lava tubesusingthe magnetometer.The Leica canbeused to make very accurate 3D representations of small areas in verygreatdetail. For examplea strangelava doughnutformation about15m diameterand1.5m high wasmeasuredandconvertedto a 3Dimagewith little problem.To completethe task with conventionalmeansto the sameaccuracywould haverequireda metalstructureto have been erected to use as a reference point to measure to.

Figure 1: A DGPS Base station in use. As can be seenthe base station is a large tripod mounted piece of

equipment. The mobile unit is of similar size but fittedwith carrying straps. [photo: Phil Collett]

6 BCRA Cave Surveying Group, Compass Points 31, July 2003

Page 7: Issue 31 July 2003 - CHAOS.ORG.UK

Surveying to Record the CavesLike all cavingexpeditions,the primary aim of the majority of thecavesurveyingundertakenon the Laki expeditionswasto makeasaccuratea recordof the cavesthat wereexplored. The techniquesusedfor theactualsurveyingwereprettystandard.Equipmentusedundergroundwas Suunto/Silvacompassesand clinos with Fibrontapesusedto measureleg lengths. Onenovel pieceof equipmentusedwasa Leica Disto-Prolaserdistancemeasurer.This wasnotgenerallyusedfor leg length measurement,but was ideal for thecollection of passage cross-sectional data, and for continuations.

The techniquesused were generally in accordancewith BCRAGrade5, though the instrumentswere not calibratedin the field.Grade4 is thusthehighestgradeclaimedfor anyof theexpedition’ssurveys.

A major potential problem in the surveyingof lava tubes is thatconsiderablelocal magnetic deviations can be present. In oneincidentan expeditionmembernoticeda c.60°swing of a compassneedlewhilst moving only 5m. This should not havecome as asurpriseastheSheptonexpeditionsto Icelandin the1970shadbeenaware of this problem, and carried out non-magneticsurveystocompensate.Sometimeswe fail to learnfrom thosewho go beforeus!

Onemethodof mitigating this problemwould havebeento assumethat any magnetic deviation was constant at a point, and takeforwardandbackbearingsat eachsurveystation. This would haveallowedan “included angle” to havebeencalculatedbetweeneachleg. Unfortunately the problem with this approachis that therewould havebeenno certaintyof the direction of the initial leg onwhich to basethe subsequentangleson, without correctingto gridvia a surfaceresectionexercise. With hindsight it may havebeenpossibleto determinethe bearingof this initial leg by having it onthe surface,makingit very long andusingGPSto calculatethe co-ordinates of the two stations.

Insteadtheapproachtakenwasto usetheGPSloggingof entrancesto provideconfidencein the survey’saccuracyandcorrectthemtoGrid North. This was an ideal situation for this techniqueas themajority of the caveshave multiple entrancesand Iceland is justabout ideal GPS territory with its flat treeless lava flows.

The standardpractice used was to plot the cave survey usingtraditional techniquesandthenplaceit over a pieceof graphpaperwith the entrancelocationsobtainedfrom GPSmarkedon it. Theplotted surveywas then rotateduntil the entrancepositionstalliedwith the GPSlocations. In the southernpartsof the flow it waspossible to use a differential GPS system to log cave entrancelocations,whilst in the more remoteareashand-heldunits had tosuffice.

In mostcasesthe fit wasastonishinglygood,in thecaseof thecaveBlámi with eight entrances,it was possibleto rotatethe surveysothatall of theentranceswerewith 5m of the GPSplottedpositions.This is of similar accuracyto thehandheld GPSusedto recordtheentrancelocations. Overall it proved possible to close surveytraversesto less than 2% error with thesetechniques,the closedtraversein Iðrafossarhaving an error of 8m over a 400m closedtraverse.

This useof GPSalsoshowedthat theunderlyingmagneticdeviationwas not constantover the entire flow, but changedlocally, withmagneticNorth varying from location to location. However, thedataobtainedover the wide rangeof the flow allowed reasonableestimatesof local magneticvariation to be madefor thosecaveswith single entrances (such as Rauðsteinshellir).

A final “proof of the pudding” was madeon return to the UK byoverlaying the combinedsurveysof the Eldhraun caveswith anaerialphotographof the lavaflow. ThecavesurveydataalongwithGPSentranceco-ordinateswereusedto constructanareamapwitha cheapCAD package(Total CAD, costsc.£10from thecheapshelfof most computer shops). It was then necessaryto rectify thephotographto a regular scale. This was done by identifying anumberof known featureswhose location had been recordedby

GPS. The photowas then scaledonto a set of points in the CADpackage and the cave area map overlaid.

In the field it hadbeennoticedthat someof the shallowcaveshadinflated the lava flow surfaceabovethem,so producinga humpedridge above their passages.When the map was overlaid on thephotoit becameapparentthat theseridgeswerevisible on thephoto.This means that this photograph provides a very powerfulendorsementof the techniquesused,asit waspossibleto matchthesurveyed passage to the ridges overlying the passages.

Of coursesome(minor) discrepancieswere noticed,primarily thelocation of Þjönappahellir. Since the cameradoesn’t lie theseaspectsof the surveywere correctedto the photograph. However,thecorrelationof thephotoandthemapdid showsomeremarkablecorrelation. Perhapsthemostimpressivewasthediscoverythat thechamberat the far endof Rauðbogahellir,from which daylight wasvisible through a boulder choke, lay directly beneath a depression onthe surface.

Surveying for Scientific AimsAs well astheexploration,the expeditionshada seriesof scientificaims. Oneof thesewasto relatethecavesto themorphologyof theEldhraun lava flows. Another was the use of geophysicaltechniques(principally magnetometry)to detectthepresenceof lavatube caves from the surface,both on the Eldhraun and on theHallmundahraun lava flow in western Iceland.

For bothof theseaspectsit wasvital to accuratelyrelatethecavestothe surface. Some of the aims of the former aspect weresatisfactorilydealt with by the combinationof cave surveysandaerialphotographs.However,for someof the moredetailedworkother techniques had to be used.

In the caseof geophysicalexperimentsit was vital to ascertainwhetheranomaliesfound weredueto the cavesbelow. For this tobe the case the relationship of the cave to the surface surveytraverseshadto be knownwith someconfidence. This wascarriedout with detailedDGPS survey areasmarkedout on the surface,which in turn weretied into detailednon-magneticsurveysof smallparts of the underlying Surtshellir/Stefansellir system.

The datafor the surveyweretakenfrom that obtainedby the 1972SheptonMallet CavingClub Expedition. Althoughthis datasetwascomplete it had never been drawn up or published. However, for thepurposesof the experimentscarriedout it was invaluableas it wasgeneratedusing non-magneticmethods, in contrast to all othersurveys of these caves (before and since).

Theequipmentusedin 1972comprisedsimpletripod mountedcavetheodolite,tripod mountedsighting targetsand 30m fibron tapes.Backsights and foresights were taken at each station and theinstrumentsreadto the nearest¼° in azimuthand½° in elevation.Analysis of thesedata showeda misclosureof about 1% on the490mclosedtraversein thecentralsectionof Surtshellir. This errorwas shared between all survey legs for drafting.

During the 2000 expeditionthe 1972 datawas supplementedwithmoreaccurateandfrequentmeasurementsof passagecross-sectionaldetails,takenalonga simpletraverseline generatedby compassandclinometer. Thesemeasurementsweretakenwith a LeicaDisto-Prolasermeasurer. Thesedatawere later correctedto the line of the1972 survey.

The line of the relevantcave passageswere then locatedon thesurfaceby repeatingthe 2000 survey traversedata from points oforigin of the undergroundsurvey (using plumb bobs from thesurfacedown into the entrancecollapses). Thoughthis work wascarriedout usingmagneticequipmentit provideda “good enough”starting point for the geophysical work.

BCRA Cave Surveying Group, Compass Points 31, July 2003 7

Page 8: Issue 31 July 2003 - CHAOS.ORG.UK

Geophysicaldatawerethencollectedalongtraverselines andgridsthat weregeo-referencedand locatedrelative to the caveentranceswith the useof a Leica GPSSystemSR530. This instrumentwasalsousedto plot the relationshipsbetweenthe traverselines/grids,the edgesof the surfacecollapsesand the points of origin of thecave survey line traverses (plumb bob positions).

The caretakenwith the measurementsin the field thenallowedthegeophysicaldata and cave surveysto be overlaid. This showed

unambiguouslythat theunderlyingcavepassageshadbeendetectedby the magnetometrytechniques. In addition the survey showedthatthereis morecavepassage,asyet un-enterable,upflow from theterminal lava seal in Stefánshellir, as shown in Figure 2.

Thisarticle is basedon a longerarticle by thesameauthorsentitled“Developmentsin ExpeditionCaveSurveying” which is due to bepublished in the forthcoming Shepton Mallet Caving Club Journal.

Figure 2: A survey of part of Stefánshellir with magnetometry survey results overlaid. The survey shows the magneticanomaly associated with the known cave, the lava seal that terminates the cave, and an anomaly consistent with further

cave passage beyond the lava seal. [figure courtesy of Chris Woods]

Describing Survey QualityBob Thrun

StatisticsIn his article, "ProvingSurveyAccuracy",JohnStevensoverlookedsome aspectsof the way that random errors accumulate. Therandomerror in the sumof measurementstendsto increaseasthesquareroot of the numberof measurements.This occursbecausethe errorssometimesaddand sometimessubtractfrom eachother.It canbehardto find this explicitly statedin a statisticsbook. Thishas been pointed out by some caving authors:Heinz Schwinge,Denis Warburton,Mike Luckwill, and Irwin and Stenner. Manycave surveys have a large loop where the percentageerror isparticularly low and the surveyorsall brag about it. I found thatmany of my short loops, with only threeor four shots,had largepercentageerrors. Both of these situations are simply to beexpected.

Schwingeshowedthat, when there are only length and compasserrors, there is an optimum surveyshot length that minimizes theerror when surveying a given distance.Schwinge did not haveclinometer or station position errors in his derivation. The minimumerror occurswhen the error due to the anglemeasurementin eachshot is the sameas the error due the lengthmeasurement.For the1976 BCRA Grade 5, this shot length is 5.73 meters. Stationposition error would favor somewhat longer shots. The average shotlengths in the loops from the Ogof Draenensurvey that Stevenspresented are close to the optimum.

Someauthorscalculatea standarddeviationfor a surveyandequateit with "probableerror". The term shouldbe "most probableerror"sinceanysizeof error hassomeprobabilityassociatedwith it. Thedefinition of standarddeviationinvolvesonly onevariable,so it isone-dimensional.Themostprobableerrorin onedimensionis zero.Standarddeviationrefersto thewidth of theprobabilitydistribution.

8 BCRA Cave Surveying Group, Compass Points 31, July 2003

MAP OF UPFLOW END OF STEFÁNSHELLIR AND MAGNETIC ANOMALIES REVEALED BY SURVEY IN 2000

Scale

Plan

Long Section

0 10 20 30 40 50m

Page 9: Issue 31 July 2003 - CHAOS.ORG.UK

Some authors also confuse the one-dimensional and the three-dimensional error distributions. If there are independent and equalprobability distributions in more than one variable, then we have achi-square distribution, which has a non-zero most probable error.The errors in a single survey shot of optimum length are anexample. If the errors on a traverse are the same in all directions,we have a good approximation of a chi-square distribution. The chi-square distributions in two and three dimensions have the specialnames of Rayleigh and Maxwell distributions. The chi-squaredistributions can be expressed in units of the one-dimensionalstandard deviation. These distributions are shown in Figure 1.Since we are concerned with the absolute value of the error, thevalues for the one-dimensional normal distribution are doubled. Thecumulative probability, the probability that an error is less thansome amount, is shown in Figure 2. The maximum, median, andmean values for the three distributions are given in Table 1. InFigures 1 and 2, and in Table 1, it is assumed that the errors areindependent and the same in all directions.

0 1 2 3 4Error (standard deviations)

0.00

0.25

0.50

0.75

1.00

Pro

ba

bili

ty D

en

sity 1D

2D 3D

Figure 1: Probability densities in one, two, and threedimensions

0 1 2 3 4Error (standard deviations)

0.00

0.25

0.50

0.75

1.00

Cum

ulat

ive

Pro

babi

lity

1D

2D

3D

Figure 2: Cumulative probability distributions in one, two,and three dimensions.

Maximum Median Mean1-D 0.000 0.674 0.7982-D 1.000 1.177 1.2533-D 1.414 1.538 1.596

Table 1: Maximum, median, and mean values for one-,two-, and, three-dimensional probability distributions.

I did computer simulations for the loops that Stevens showed in hisTable 3. I constructed loops with the same length and number ofshots that he had. On each survey shot I added random errors fromuniform distributions as specified in the 1976 BCRA Grade 5standard. I did 1000 simulations for each set of loop conditions andmade bar graphs showing the distributions from the simulations. Oneach of the bar graphs I added a symbol showing the closure errorfrom the actual survey. These bar graphs are shown in Figures 3 to11. For most of the loops, I constructed N-sided regular polygons,

giving nearly circular loops. This ensured that the errors were thesame in the two horizontal directions. If I did more simulations foreach set of loop conditions, the bar graphs would more closelyresemble the 3-D distribution in Figure 1.

I tried loops where half the shots went out in one direction and thendoubled back along the same route. The doubling back had a slighteffect on the overall error distribution of the loop. One of thedoubled-back loops is shown in Figure 4.

All the survey shots in the simulated loops were of equal length. Ifboth the total length and the number of survey shots are specified,the minimum error occurs when all the shots are the same length.The worst case is one long shot and a bunch of very short shots. Mysimulations predict less error than would more realistic simulationswith varying length shots and the proper angles. Better simulationscould be done by adjusting the loops and then adding random errorsto the adjusted measurements.

0 2 4 6 8 10 12Closure error (meters)

0

40

80

120

160

Fre

qu

en

cyFigure 3: Probable loop closure error distribution for a

circular loop 5083 meters long with 614 shots.

0 2 4 6 8 10 12Closure error (meters)

0

40

80

120

160

Fre

qu

en

cy

Figure 4: Probable loop closure error distribution for adoubled-back loop 5083 meters long with 614 shots.

0 1 2 3 4 5 6 7 8 9 10Closure error (meters)

0

40

80

120

160

Fre

qu

en

cy

Figure 5: Probable loop closure error distribution for acircular loop 4158 meters long with 481 shots.

0 2 4 6 8 10 12Closure error (meters)

050

100150200250

Fre

qu

en

cy

Figure 6: Probable loop closure error distribution for acircular loop 2197 meters long with 292 shots.

BCRA Cave Surveying Group, Compass Points 31, July 2003 9

Page 10: Issue 31 July 2003 - CHAOS.ORG.UK

0 1 2 3 4 5Closure error (meters)

0

40

80

120

160

Fre

qu

en

cy

Figure 7: Probable loop closure error distribution for acircular loop 1314 meters long with 159 shots.

0 1 2 3 4 5Closure error (meters)

04080

120160200

Fre

qu

en

cy

Figure 8: Probable loop closure error distribution for acircular loop 850 meters long with 139 shots.

0 1 2 3 4Closure error (meters)

04080

120160200

Fre

qu

en

cy

Figure 9: Probable loop closure error distribution for acircular loop 570 meters long with 102 shots.

0 1 2 3 4Closure error (meters)

0

40

80

120

160

Fre

qu

en

cy

Figure 10: Probable loop closure error distribution for acircular loop 610 meters long with 71 shots.

0 0.5 1 1.5 2 2.5 3Closure error (meters)

0

40

80

120

160

Fre

quen

cy

Figure 11: Probable loop closure error distribution for acircular loop 395 meters long with 38 shots.

I examined about 40 cave surveys for a talk I gave at the 2000 NSSConvention. None of these surveys met BCRA Grade 5 standards,though some came close enough that I feel there might be a Grade 5survey somewhere. Stevens presented eight loops that he claimed tobe Grade 5 or 6 quality. Of these, only three of the four claimedGrade 6 loops meet Grade 5 quality, while none of the claimedGrade 5 loops do. Two of the loops look like they might meetGrade 6 quality. Or they might just be some better-than-averageloops from a lower quality survey. We would need to examinemany more loops to be sure.

Finding LoopsFor the first cave survey data reduction program I wrote, I had tospecify the route of each loop. I found that even a small networkhas a large number of possible loops. As an example, consider asimple network with nine small squares making up one large square.If we count just rectangular loops, we get the numbers of loopsshown in Table 2. There are other possible shapes of loops andmany more loops in this network. It is easy to get over a hundredloops.

Number ofloops

Size ofloop

9 1x112 1x26 1x34 2x24 2x31 3x3

Table 2: Counts for some loops in a simple network,consisting of nine small squares making up one large

square.

Irwin and Stenner described a survey adjustment method where theypick some key junctions in the survey and then average a fewdifferent routes from the entrance to each of these junctions. Thismethod is best suited for a person working with a calculator. Itwould be hard to program a computer to find just a few distinctroutes.

Larry Fish, in describing his COMPASS program, said it finds allthe loops in a survey. I puzzled over this because I realized that avery large number of loops can be constructed. COMPASS finds aminimal set of loops that are called fundamental cycles of a graph ingraph theory. Any loop may be constructed by adding cycles from aset of fundamental cycles. Where parts of two cycles coincide, theycancel out. There are many possible sets of fundamental cycles forany graph. Figure 12 shows the same network with two differentsets of fundamental cycles. The set that is found by COMPASS isdetermined by the order in which the data are presented. WinKarst,by Garry Petrie, attempts to find small loops. Other cave surveyprograms use least-squares and do not explicitly find loops.

Figure 12: Two different sets of fundamental cycles forthe same graph.

The problem of finding all the loops resembles the classic TravelingSalesman Problem. However, the salesman can go directly fromany city to any other city, while the cave surveyor can go to only afew other survey junctions from any given junction. I have come tothe conclusion that if there are N junctions in a cave survey network,there are between 2N and 3N loops in the network. Consider twopoints with N junctions on the route between them. At eachjunction there are 2 or 3 route choices. You can come to the sameconclusion by considering the effect of an addition to a network.The exact number of loops will depend on the network. It willusually be more than I want to count, even with a computerprogram. I once suggested a random-walk program to gather loopstatistics.

10 BCRA Cave Surveying Group, Compass Points 31, July 2003

Page 11: Issue 31 July 2003 - CHAOS.ORG.UK

Assigning GradesBryan Ellis and otherscontendedthere were three basic types ofsurveys:the rough sketch,the fast survey,and the propersurvey.Theywantedto reducethenumberof gradesto three,but put in thenon-favoredgradesbecausesomecaversdid not want their surveysdowngradedfrom a 5 to a 3! I can't think of any instrumentsgraduatedin 5-degreeincrementsto usefor a Grade3 survey.If thesame surveyorsuse the same instrumentsfor the two types ofsurvey, they should get the sameaccuracy. The typical surfacesurvey would be an exampleof a fast survey. The differencesbetweenthe two types of survey are in the level of detail andthoroughness in going down every passage.

Surveysdonewith compassesthataremarkedin one-degreeor half-degreeincrementshaveerrorslargerthanwould be indicatedby theprecisionof the markings. I would not be surprisedif compassesmarkedin two or five-degreeincrementsproducelesserror thantheprecisionof marking indicates. I seevery few British newsletters.Are thereanyGrade3 surveysdone? If so,could somebodycheckthe closureerrors on someof these? It may be that Grade3 isobsolete because no Grade 3 surveys are done.

All the surveysI looked at for my 2000 talk would have to beclassified as 1976 BCRA Grade 4. Leaving out the very worstsurveys,which had a lot of blunders,therewere factorsof 2 or 3betweenthe bestand worst surveys,which brings us into Grade3territory. There was no obvious place to divide the surveysintogoodandbadcategories.The spreadin the valuesof the individualclosureadjustmentswithin any one surveywas at leastas wide asthe error simulationsI show here. It looks to me like thereis noGrade4 gapbetweenGrades3 and5. Instead,they all mergeintoone broad grade.

I have a copy of the Ogof DraenenGrade 2 survey data. Thecompassis alwaysreadto thenearestdegreeandthedistanceis readto 0.1 meter. The 0.1 meteraccuracywould qualify for Grade5 inthe 1976standard,but not the 2002standard.Therewould be littledifference in the overall accuracybetween0.1 and 0.01 meteraccuracy becausethe larger errors due to angle measurementsdominate. A declination correction is made, although the compassesand readersare not individually calibrated. The main differencebetweenthe data I have and a Grade 5 survey is the lack ofclinometerreadings. Perhapssomeonecould comparethe Grade2and 5 maps.

Somesurveyorspublishmapswith a magneticnorth arrow. Somegetthedeclinationfrom a mapor a magneticfield model. Somegeta combineddeclinationand compasscorrectionby a surfaceshot.Somedo surfaceshotsin multiple directionsto get a correctionforcompasseccentricity error. I will leave it to others to argue ifcalibration or the lack of it is enough for a separate grade.

Thewordingof theBCRA gradesgivestheimpressionthatbaselineaccuracyis the most importantaspectof a cavesurvey.We all tryfor accuracyin our surveys,but doesit really matter? For mostuses,theamountof detailandthecompletenessof themaparemoreimportant. An inaccuratesurveyis adequateasa roadmapaslongasthereareno blunders. On theotherhand,eventhemostaccuratemagneticsurveyis not adequatefor drilling a newentrancefar fromthe original entrance.

William E. Daviesproducedmany mapsfor his book, CavernsofWestVirginia. He often mappedsmall caveson consecutivedays,so they must have beencompass-and-pacemaps. In comparisonwith modernmaps,his anglesmay be off by up to 10 degreesandhis distancesmaybeoff by 30 percenteitherway. Theonly way totell is to overlayoneof his mapswith a new map. He wasa goodobserverand his mapsoften have as much passagedetail as themodern maps. He did not push down every small passage.

The BCRA detail gradingsare worded awkwardly. Rather thanbeingbasedon how muchdetail thereis in a map,theyarebasedonwhetheror not the detailsshownweremeasured.A mapwith lessdetail couldhavea higherdetail gradethana mapwith moredetail,provided that all the details shown were measured. It should be

obviousto anyonelooking at a map how much detail it has. Andwhat about small versions of large maps?

British cavers are accustomedto gradesand regard them as ameasure of quality. Others use BCRA grades with the mistaken ideathat they arescientific. The useof a gradeis an implicit claim ofaccuracy,but there have beenvery few efforts to determineif asurvey meets the specification. There were Irwin and Stenner,myself, and now John Stevens. Any others? Most British cavesurveyorssimply calleda Compass,Tape,andClinometersurvey(aclearerterm that I prefer) Grade5. The 2002 revision brings theBCRA gradingsystemin line with actualpracticeandchangesthesystemfrom a failed attemptto quantify the accuracyof a surveytojargon for the sake of jargon.

If you wantto describethequality of a survey,usingasystemthat isclear, meaningful, and unambiguous, I suggest the following:

� No map.� Sketch with no measurements.� Compass and Pace.� Compass and Tape� Compass, Tape and Clinometer.

Somethingis eithermeasuredor not measuredin this classification.And thereis no needto usecode. This classificationsaysvery littleabout the accuracyof the survey,but the BCRA gradesdon't saymuch either.

Somecaversmay want to assigna numericalvalueto the accuracyof their surveys. Two possiblemeasuresfor a single loop areError / N , whereN is the numberof surveyshotsin a loop, orError / length . The first of thesewill be roughly proportionalto

the error per shot. The secondwill be roughly proportionalto theangularerror on the longershots. Neitherof thesemeasurestakesinto account the lengths and directions of the individual surveyshots. For a givensurvey,therewill bea mixtureof goodloopsandbad loops,long loopsandshort loops. It is possibleto display thevaluesas a point cloud, similar to the way I showedleast-squaresadjustmentsin my 2000 talk. I can't think of a goodnon-pictorialway to summarizethe statisticsfor an entiresurvey. To useeitherof these two loop measures, it is necessary to explicitly find loops.

I can'tthink of anyway of evaluatingthesurveyaccuracythat takesinto accountthe survey shot lengthsand directions that doesnotmakea comparisonto anassumedstandard.If that is done,thereisstill the problem of summarizing the results.

ReferencesDavies, William E., Cavernsof West Virginia (2nd. ed.), West Virginia

Geological and Economic Survey, Vol. XIX A, July 1958Ellis, B. M., "Changesin the Systemof GradingSurveys", CaveResearch

Group of Great Britain Newsletter, No. 132, January 1973, pages 15-17Ellis, Bryan M., "Changesin the System of Grading Surveys", Cave

ResearchGroupof GreatBritain Newsletter,No. 134,June1973,pages4-6

Ellis, Bryan,SurveyingCaves,British CaveResearchAssociation,1976,88pages

Fish,Larry, "The ProblemWith LeastSquaresLoop Closures",Compass&Tape, Vol. 13, Issue 1, No, 41, April 1997, pages15-18, also inCompass Points 15, March 1997, pages 8-10, and athttp://www.fountainware.com/compass/compart.htm(mentions "sorting all of the loops in the cave")

Irwin, D. J. andR. D. Stenner,"AccuracyandClosureof Traversesin CaveSurveying", Transactionsof the British Cave ResearchAssociation,Vol. 2, No. 4, December 1975, pages 151-165

Luckwill, Mike, "A Brief Review of the Theory Available to the CaveSurveyor", Belfry Bulletin, Vol. 23, No. 4, pages 44-51

Schwinge, Heinz T., "The Accuracy of Cave Survey", The NationalSpeleological Bulletin, Vol. 24, Part 1, January 1962, pages 40-47

Stevens,John, "Proving Survey Accuracy", CompassPoints 30, March2003, pages 11-13

Thrun, Robert, "How Accurate are our Cave Surveys", (abstractonly),Journal of Cave and Karst Studies, Vol. 63, No. 1, April 2001, page 54

Warburton,Denis, "The Accuracyof a CaveSurvey", WessexCaveClubJournal, No. 89, Vol. 7, April 1963. pages 166-181

BCRA Cave Surveying Group, Compass Points 31, July 2003 11

Page 12: Issue 31 July 2003 - CHAOS.ORG.UK

Printable Mapping from the Millennium AtlasAndy Waddington

You may have heardabout the "Millennium Atlas" - an effort toprovidea consistentsetof aerial photographyof the UK at a highresolution. By "consistent",theymeanthatall thephotographyhasbeenflown in similar weatherconditionsandfrom a similar height,suchthatphotographsfrom distantareasmaybecomparedlike-for-like. The resultof all this is thatyou canbuy (for really quitea lotof money)a photographof (almost)any500x500mareain EnglandandWalesat a resolutionof 0.25mper pixel on the ground. Thereare also various spins-off like large-scalemapsof popular touristareaswith grid, contoursandotherinformationsuperimposed,andabig expensive Atlas on paper.

What is perhapsless well-known is that, at 2 metresper pixelresolution,all this photographyis availablefor free on the web attheGetMapping.com site, providedaspart of the interfaceusedfor choosing the photos they would like to sell to you.

Why might you want aerial photographyrather than a good old-fashionedOrdnanceSurvey (OS) map with all the advantagesofnamesfor features,definitive symbolsfor thingsthatyou might notrecognisefrom the air andso on? You will still want mapsfor allthe information that aerialphotosdon't contain,andwhich the OS,andothers,havekindly gatheredby surveying"on theground",suchas rights of way, altitudes, names and "tourist information".However,aerialphotosshowmuchdetail that is not translatedintomapping, some of which is particularly useful to the caverproducing surface surveys, searchingfor new cave, or simplywalking the area. Things markedon the map simply as "areaofshakeholes",for example,appearin detail as individual shakeholeson an aerial photo. Whilst OS mapsshow broad-brushvegetationtypes,thereis muchmoredetail on an aerialphoto,andthingslikesheeppathsthroughheatheror brackenmayshowup, which canbeuseful. More subtlechangesin colour of vegetationcan indicateunderlyingsoil type, wetnessor even the geology- the limestoneshaleboundaryis not markedon topographicmaps. Norman &Waltham suggestthat, in their study, a limestone/shalecontactshowsup asa changein colouror tonein 30%of cases,a changeofvegetationin 12%of cases,anda similar proportionasa changeoftexture. 36%of casesshoweda stepchangein level, which canbevery clear in the right lighting.

Whenyou go to theGetMapping.com website,you arerequiredto entera grid referenceor postcodefrom which you get back a500x500mmapwhich you canscroll N/S/E/Win 167mincrements.This is the 2m free dataset,which is presentedas83x83pixel tiles,each covering an area of 167m square. The tiles are jpegs,compresseddown to typically 1 to 1.5 kb, which is rather severecompression,sotherearequitea lot of compressionartefactsvisibleon relatively featurelessareaslike water. However, featuresliketrees, buildings and shakeholesshow up pretty well, and goodfootpathsareclearly visible. Someof the rural areasseemto havebeenmappedfrom a singleflight, suchthatsmallareasareobscuredby cloud,or havegapsin coverage.However,by mid-2002,prettymuch everywhereI have looked in Englandand Wales seemstohavepretty goodcover,thoughhardly any of Scotlandseemsto bemapped.

Theoriginal air photoshavebeenscanned,scaledandtweakedto bea fairly good and consistentmap. In feature-richareas,the geo-referencingseemsto bepretty good,thoughin moorlandareas,lesscarehasbeentaken(lessreferencepointsareavailable),andin myown local area(Teesdale),for example,I canseethe samepatchofburntheatherin two adjacenttiles, suggestingsomethinglike a 20minaccuracyin the geo-referencingof the two adjoiningphotos. Tobefair, however,this doesn'tseemto bevery commonandI suspectthat for the most part the georeferencing is good to 5m. Feeding thismappingto a 1200dpi colour printer at 125 pixels/cmandrelyingon the printer driver softwareto deal with the mismatchbetweenpixel sizesin the imageandon thepaperproducesquite respectablecompositemappingat 1:25000,on which thingslike shakeholescanclearly be seen, and grid references read off.

The mapping is copyrighted, of course, and among variousrestrictionsplacedon your useof thedata,republicationon thewebappearsto be specificallyprohibited. However,for personalstudyandother"fair use",thevariouscopyrightactsgive you a fair bit ofleeway.

Thecatch,of course,is thatwith tiles this size,you need36 tiles tomake up a kilometre square, and a lot of files need to be downloadedto makeup a reasonablemappingarea. Doing this by handsoonbecomestedious,but this is exactlythesortof thing thatcomputersare good at doing, and a simple set of scriptscan very quickly belashedup to utilise commoncross-platformtools like wget andImageMagick to downloadthe tiles and glue them togetherintorather bigger and more convenient units.

The tiles have URLs like

http://www2.getmapping.com/isapi/gettile.dll?Dataset=2mFree&level=0&i=<i>&j=<j>

where the numbers<i> and <j> are derived from the full gridreference. By "full grid reference"I meana grid referencewhichincludesnot just the two-digit numbersalongtheedgeof your map,but alsothe small leadingdigit which canbe seenat the cornersofthe map, and which are normally representedby using the gridletters instead. As a concreteexample,take Gaping Gill, at SD751727. Thekm grid squareis SD 75 72, andlooking at thecornerof the map you can see that this translatesinto 375 472. Thenumbers<i> and<j> areobtainedsimply by multiplying by six, inthis casegiving i=2250,j=2832. Thatis thetile at theSW cornerofthe grid square,and,with six tiles per kilometre,the tile in the NEcorner is i=2255, j=2837.

If you montage36 83x83 pixel tiles into a kilometre square,theresultis a 498x498bitmap. I havefoundit convenientto put a one-pixel blue border round each such kilometre square,making it500x500. Stackingtheseup togetherthen gives a neat500 pixelsper kilometre, and a kilometre grid with lines two pixels thick,which looks about right.

What is neededis a small bit of codethat will takea specificationfor an areayou want mapped,then fetch the tiles, assembletheminto kilometre tiles with a border,thenblock theseup into eitherabitmap for your whole areawhich you can print, or perhapsas awebpagewith the kilometre squaresdisplayedin a table,which isconvenientfor desktopviewing. Since the tiles are alreadya bitdegradedby being jpegged, its probably best to avoid highlycompressingthe kilometre squares,causingfurther degradation. Ihaveusedpng for the kilometresquaresandlargerblocks,astheserenderquickly, but they do useup a lot more disc spacethan theoriginal tiles (300-500Kb perkilometresquare,asopposedto c 50kfor the 36 tiles).

As with all programming tasks, a multitude of solutions are possible.I have hackedtogethera set of small programswhich generatescripts to accomplish the task. This has a somewhatbaroquearchitectureasa consequenceof an interestingmixtureof hardwareon this site and a desire to be able to split the task acrosstwomachinesso that the actual downloadscan go on in backgroundwithout impacting my desktopmachine. However, the softwarehangstogetheron (a) Linux box(es),andcouldserveasanexamplefor someonewantingto write somethingbetter.I'm reluctantto putthestuff up on a webpagein casegetmapping.comdecidedthat thisis misusing their resource,so if you would like a copy of thesoftware, please email [email protected].

Reference

Norman, J.W. & Waltham, A.C. (1969). Aerial Photographyand theinvestigation of karst features, in Symposium on Cave Photography,Transactions of the Cave Research Group of Great Britain, Vol. 11(4),245-254.

12 BCRA Cave Surveying Group, Compass Points 31, July 2003