PFG 2014 / 1, 0041–0054 Report Stuttgart, February 2014 Modification of High Resolution Airborne Laser Scanning DTMs for Drainage Network Delineation MICHAEL VETTER &GOTTFRIED MANDLBURGER, Vienna, Austria Keywords:Hydrology,flowaccumulation,opensource,GRASS-GIS,anthropogeniceffects, street networks, DTM, ALS Summary: In this paper a method is presented to Zusammenfassung: Anpassung eines hochauflö- adapt a 1 m ALS (Airborne Laser Scanning) DTM senden Airborne Laser Scanning DTMs zur Be- (DigitalTerrainModel)fordrainagenetworkde- rechnung von hydrologischen Modellen. In diesem lineation. In such DTMs small natural topographic Artikel wird eine Methode vorgestellt, mit der ein features as well as anthropogenic structures like hochauflösendes 1 m ALS (Airborne Laser Scan- roads and artificial embankments are contained, ning) DTM (Digital Terrain Model) für die Ablei- both influencing the drainage network delineation tung eines potentiellen Gerinnenetzes aufbereitet process in different ways. While natural topo- wird.In1mGeländemodellensindnatürlicheGe- graphic features lead water to drainage systems, ländemerkmale sowie Bereiche mit anthropogener anthropogenic structures can deflect water along- Reliefüberformung(StraßenundkünstlicheBö- side artificial surfaces, which are not flow-active schungen) abgebildet. Der natürliche Geländever- under regular hydrological conditions and, there- lauf und die anthropogenen Strukturen beeinflus- fore,resultinwrongdrainagesystems.Wepresent sendieAbleitungvonGerinnenetzwerkenbasie- aworkflowtosubtracttheinfluenceofroadsinthe rend auf Fließakkumulation unterschiedlich. So DTM, replacing the actual road profile by an aver- werdenGerinnenetzwerkeentlangdesnatürlichen agehillslopeoftheneighbouringterrain.This Geländeverlaufs in der Regel ein realistisches Ge- modified DTM is the basis for drainage network rinnenetzwerk ergeben. Entlang von künstlichen delineation using standard flow accumulation. The Strukturen können die Berechnungen in eine Rich- drainagenetworksarederivedforfourDTMvari- tung abgelenkt werden, in welche unter normalen ants(1mDTM,1mmodifiedDTM,5mDTMand hydrologischen Bedingungen kein Abfluss stattfin- 5 m modified DTM) and the results are compared det. Eine solche Ablenkung resultiert daher in ei- to a reference dataset. It is shown that the accuracy nem unrealistischen Gerinnenetzwerk. Mit dem of the derived drainage network can be increased vorgestellten Ansatz, bei dem der Einfluss der Stra- by 3–5% using the modified instead of the original ßen aus dem Geländemodell herausgerechnet und 1mDTM.Thegaininaccuracyamountsupto12% durch eine angenäherte natürliche Neigung entlang whenusingthemodified1mDTMcomparedto der Straßen ersetzt wird, ist eine deutliche Verbes- anyofthe5mDTM.Therefore,ourconclusionis serungderGerinneableitungmöglich.Fürvierver- that high resolution and modified 1 m DTMs with schiedene DTM-Varianten werden Gerinnenetz- anthropogenic structures such as roads strictly re- werke abgeleitet (1m DTM, modifiziertes 1m movedshouldbeusedfordrainagenetworkdeline- DTM, geglättetes 5 m DTM und modifiziertes 5 m ationinsteadofacoarsespatialresolutionoran DTM). Die Ergebnisse werden mit einem Referenz- original1mDTM. datensatzverglichen,umdieGenauigkeitderAb- leitungzudokumentieren.DieendgültigenResul- tate der Ableitung und der Qualitätskontrolle erge- beneineVerbesserungderGenauigkeitum3–5% desmodifiziertenimVergleichzumoriginalen1m DTMundeineVerbesserungvonbiszu12%des modifizierten 1 m DTM im Vergleich zum geglät- teten bzw. modifizierten 5 m DTM. Folglich ist un- sere Empfehlung, für die Ableitung von Gerin- nenetzwerken anstatt eines geglätteten 5 m DTM oder eines originalen 1 m DTM ein modifiziertes 1 m DTM zu verwenden, bei dem anthropogene Strukturen wie Straßen rigoros entfernt wurden. © 2014 E. Schweizerbart'sche Verlagsbuchhandlung, Stuttgart, Germany www.schweizerbart.de DOI: 10.1127/1432-8364/2014/0208 1432-8364/14/0208 $ 3.50 42 Photogrammetrie • Fernerkundung • Geoinformation 1/2014 1Introduction The differences between drainage net- work delineation based on high spatial reso- Forthelasttwodecades,AirborneLaserScan- lutionandcoarseDTMsareshowninLI & ning (ALS) has become the prime acquisition WONG (2010).Thequalityofadrainagenet- technique for collecting topographic data in work based on high spatial resolution data, highspatialresolution(>1point/m2)witha e.g. a 1 m DTM grid, is mainly influenced by height accuracy of less than 15 cm (WEHR & man-made objects such as streets and dams, LOHR 1999, BALTSAVIAS 1999),whichisusedin wheretheflowdirectionisdeflectedalong many different fields of application, e.g. HÖF- the gradient of the street. Remaining brid- LE & RUTZINGER (2011), MANDLBURGER et al. gesormissingobjectslikeculvertsorpipes (2009), and SOFIA et al. (2011). While country- actingasflowbarriers,whicharenotrepre- wideALSdataacquisitionisalreadyfinished sentedintheDTMandyieldunrealisticflow in some European countries, e.g. The Nether- paths (VIANELLO etal.2009).ADTMfreeof lands, Denmark, and Switzerland, some other the above mentioned flow barriers, which ad- countries will complete ALS data acquisition ditionally guarantees a monotone height pro- inthenearfuture,e.g.AustriaandFinland. gressing along the streams, is referred to as a Almost all European countries maintain riv- “hydrologicallyenforcedDTM”.SuchaDTM ernetworkdataderivedfromlowresolution isaprerequisiteforobtainingcorrectdrainage Digital Terrain Models (DTMs) or topograph- network results. icmaps.ThisisalsotrueforAustria,butthe planimetric accuracy and the completeness of the Austrian river network are poor, especial- 2 Objectives ly for small catchments. The European Water FrameworkDirective(WFD,EU2000)ob- We present a study to map the potential river ligates the member countries to provide de- drainagenetworksystemusingahighresolu- tailed, up-to-date river network data in high tionALSDTMinordertoincreasethelev- planimetricaccuracywithadditionalattri- el of detail, correctness and completeness of butes per river reach, e.g. length, bed slope, thefinaldrainagesystem.Duetothefactthat width, and stream ordering. Currently, activi- streets produce massive errors for drainage tiesareundertakentostandardizethedataex- network delineation (VIANELLO et al. 2009) it change on a transnational level. Guidelines for isouraimtoshowthepositiveimpactofma- basic datasets have already been implemented nipulating artificial structures in a high spatial (INSPIRE, EU 2007). resolution ALS DTM on the delineation qual- The standard drainage delineation meth- ity.Thegoalofthiscontributionisthederiva- ods implemented in proper GIS-Software are tion of potential river drainage systems. Here- basedeitheronsingle-neighbourflowalgo- by, the term “potential river” describes the wa- rithms (D8, O’CALLAGHAN & MARK 1984) or tercourse following the terrain gradient under multiple-neighbour flow algorithms like mul- regular hydrological conditions. At locations tipleflowdirection(MFD,QUINN et al. 1991). where streets block the potential flow path, Both flow algorithms are used to compute pipes and culverts are usually built to guar- drainage networks which indicate the poten- antee a continuous watercourse along ditches. tial watercourses. A comprehensive overview Thosepipesandculvertsaredimensionedto about flow algorithms and their differences cope with regular run-off volumes. As regular are given by GRUBER & PECKHAM (2009) and hydrological conditions are presumed in this WILSON et al. (2008). Various implementations article, no deflections alongside the streets are areavailableinstandardGISsoftwareprovid- expected. As subsurface man-made structures ingspecializedtoolsforindividualenviron- under roads, e.g. pipes, are undetectable in ments. Besides flow algorithms, other meth- ALSdata,theroadsactaswallsorbarriersif ods exist to derive drainage networks and fea- theyarenotremovedfromtheALSDTMprior tures related to hydrology (PASSALACQUA et al. tothedrainagedelineation.As1mDTMsare 2010, 2012). availableinmanycountries,thegoalistoex- ploitthefullpotentialofthosedataassource MichaelVetter&GottfriedMandlburger,ModificationofHighResolutionALSDTMs 43 fordrainagenetworkdelineation,whichisnot covers an area of 93 km2 and is divided into thecaseuptonow(MANDLBURGER et al. 2011). an upper (Bolgenach) and a lower (Weißache) Therefore, we present a method to modify a sub-catchment (Fig. 1). The terrain elevations DTMbysection-wiserecalculatingthenear- rangefrom450mattheconfluenceofthe naturalslopebeforetheroadwasconstructed WeißachandtheBregenzerAchto1645mat andbyreplacingtheactualroadprofilebased the Feuerstätterkopf. The dominant geological ontheaveragehillslope.Thisyieldsamodi- formation is Molasse with a dense drainage fied 1 m DTM without streets. The results are system (OBERHAUSER & RATAJ 1998). comparedtoareferencedatasetprovidedby ForthetestsiteALSdataareavailablefrom the local mapping authority. We assume that differentepochs.ForthisstudytheDTMde- themodificationoftheDTMleadstoabet- rivedfromthepointcloud,usingthehierarchic ter drainage network mapping accuracy when robust filtering approach described by KRAUS applying automatic delineation based on flow & PFEIFER (1998),oftheNovember2003cam- accumulation. paign, was chosen. The average point density is 1.6 points/m2 (last echoes). The data were collected with Optech’s ALTM 2050 scanner 3 TestSiteandData in discrete echo recording with a maximum of four reflections per laser shot. The grid width Thechosentestsiteisasub-catchmentofthe oftheDTMis1m.Inaddition,aresampled Bregenzer