large-scale particle image velocimetry for measurements in riverine

Large-scale particle image velocimetry for measurements

in riverine environments

M. Muste,1 I. Fujita,2 and A. Hauet3

Received 27 February 2008; revised 31 July 2008; accepted 17 September 2008; published 30 December 2008.

[1] Large-scale particle image velocimetry (LSPIV) is a nonintrusive approach tomeasure velocities at the free surface of a water body. The raw LSPIV results areinstantaneous water surface velocity fields, spanning flow areas up to hundreds of squaremeters. Measurements conducted in typical conditions in conjunction with appropriateselections of parameters for image processing resulted in mean velocity errors of less than3.5%. The current article reviews the background of LSPIV and the work of three researchteams spanning over a decade. Implementation examples using various LSPIVconfigurations are then described to illustrate the capability of the technique tocharacterize spatially distributed two- and three-dimensional flow kinematic features thatcan be related to important morphologic and hydrodynamic aspects of natural rivers.Finally, results and a critique of research methods are discussed to encourage LSPIV useand to improve its capabilities to collect field data needed to better understand complexgeomorphic, hydrologic, and ecologic river processes and interactions under normaland extreme conditions.

Citation: Muste, M., I. Fujita, and A. Hauet (2008), Large-scale particle image velocimetry for measurements in riverine

environments, Water Resour. Res., 44, W00D19, doi:10.1029/2008WR006950.

1. Introduction

[2] Flow discharge is the most common riverine hydrau-lic measurement being the primary parameter for character-izing river dynamics. Until recently, discharge measurementshave relied on mechanical velocimeters. These instruments,applied extensively on a global scale produced a vastamount of data that still serves as reference for more modernriver flow measuring instruments. The advent of a newgeneration of acoustic, radar, and image-based velocitymeasurement methods in the 1980s, has improved hydro-logic and hydraulic measurement efficiency, performance,and safety. The new instruments are fast, automated andcomputerized. They have no moving parts, require fewercalibrations, and are less intrusive than their predecessors.The superior efficiency of this new generation of instru-ments extends conventional applications beyond dischargemeasurements, such as documenting river hydrodynamicfeatures, previously obtained only in laboratory conditions[Dinehart and Burau, 2005].[3] This technological development is timely as new con-

cerns with rivers warrant additional data. Issues of channelreconfiguration, bank stabilization, floodplain reconnection,in-stream habitat improvement and dam removal requirehigh-resolution estimates of flow velocity, duration, timing,and rate of change of total stream discharge [Poff et al.,1997]. For example, stream restoration projects demand

spatial characterization of flow distribution and regimeswithin the river reach subjected to retrofitting. Inferenceson ecological habitats’ health conditions can be made byexamining velocity gradients that promote dynamically sta-ble channel morphologies. These new data needs are expen-sive or impossible to obtain with conventional techniques.[4] Currently, the new generation of instruments is

replacing mechanical instruments at a considerable rate.Most notable is the steady spread of acoustic velocimeters,significantly assisted by aggressive industrial grade produc-tion and distribution [Christensen and Herrick, 1982].Radar-based techniques have also advanced with the directsupport of the USGS’s Hydro 21 Committee [Costa et al.,2000]. Image-based techniques are less frequently used forfield work in the hydrologic community, despite early atten-tion received from the same USGS committee [Melcher etal., 1999]. The acoustic methods measure along verticals inthe water body and the latter two techniques measure at thestream free surface. These instruments are frequently char-acterized as nonintrusive, though ‘‘quasi-nonintrusive’’ maybe a more accurate description. Indeed, the instrumentsmeasure along lines or on surfaces away from their physicallocation. However, acoustic-based instruments requiredeployment of a probe under the water surface and soundwave-scattering particles suspended in the water to bemeasured while image and radar-based tools require distin-guishable tracer or patterns on the free surface to capture theunderlying water body movement. Extent of intrusivenessfor all newer instruments is, however, minimal compared toconventional instruments.[5] Historically, the image-based technique was the first

and continues to serve as an important tool for flowinvestigation. The intricate flow patterns depicted inLeonardo da Vinci’s sketches suggest that the human eye

1IIHR–Hydroscience and Engineering, University of Iowa, Iowa City,Iowa, USA.

2Research Center for Urban Safety and Security, Kobe University, Kobe,Japan.

3DTG, Electricite de France, Toulouse, France.

Copyright 2008 by the American Geophysical Union.0043-1397/08/2008WR006950$09.00

W00D19

WATER RESOURCES RESEARCH, VOL. 44, W00D19, doi:10.1029/2008WR006950, 2008ClickHere

for

FullArticle

1 of 14

http://dx.doi.org/10.1029/2008WR006950

can sense important qualitative aspects of a river’s flow.Transfiguring these visual impressions into quantitativeriver flow information, however, has only recently becomepossible. Developments over the last three decades inoptics, lasers, electronics, and computer-related technolo-gies have facilitated implementation of image-based techni-ques for flow visualization and quantitative measurementsin laboratory studies. The first image-based quantitativeinstruments, generically labeled as particle image velocimetry(PIV), have greatly enhanced measurement techniques ofinstantaneous velocity vectors in a variety of laboratoryflows [e.g., Adrian, 1991; Raffel et al., 1998]. Despite theincreased PIV popularity in laboratories, image velocimetrywas rarely applied to natural-scale flows. Some earlyattempts to investigate natural flows with image velocimetrywere those of Leese et al. [1971] using satellite imagery totrack atmospheric cloud movements, Collins and Emery[1988] sea ice, and Holland et al. [1997] for swash flowquantifications in coastal areas.[6] The first image velocimetry measurements in rivers

were made in Japan in the mid-1990s [Fujita and Komura,1994; Aya et al., 1995; Fujita et al., 1997]. The techniquehas subsequently undergone continuous development andtesting in anticipation of hydraulic applications [Muste etal., 2004a]. As most of the measurements were taken oversurfaces much larger than those in traditional PIV, thetechnique was dubbed large-scale PIV (LSPIV). The presentpaper introduces LSPIV to the hydrologic community bybriefly reviewing the methodology, synthesizing its evolu-tion, providing implementation examples, sharing findings,and formulating research needs for further technique optimi-zation for a variety of riverine environment investigations.

2. LSPIV System Components

[7] Conventional PIVentails four components: flow visu-alization, illumination, image recording, and image process-ing. Given that LSPIV images cover large areas usuallyrecorded from an oblique angle to the flow surface, anadditional step is customarily involved: image orthorectifi-

cation. The LSPIV measurement sequence is illustrated inFigure 1.

2.1. Flow Visualization, Illumination, and Recording

[8] In general these technique components are stronglyinterrelated, such that selection of one approach for acomponent imposes the types of devices or approachesavailable for the remaining components. The selection ofthe components and their integrated operation in the con-ventional PIV is driven by established rules of thumbregarding the concentration of particles, their size withrespect with the image processing parameters, and thedesirable particle displacement in a series of images[Adrian, 1991]. Use of these rules is common practice forPIV measurements in the laboratory environment. Unfortu-nately, except possibly for sufficiently small channels andstreams [Bradley et al., 2002; Jodeau et al., 2008], less thandesirable laboratory recording conditions require proceduraladjustments when LSPIV is implemented in field measure-ments. These include finding recording position(s) thatmitigate two pervasive problems that impact flow visuali-zation in the field. The first is poor or strong illuminationthat might occur when only natural light is used. Glare andshadows on the water surface significantly degrade imagequality [Hauet et al., 2008b].[9] A second problem is insufficient flow seeding. A

favorable situation is when the free surface is visualized bynaturally occurring tracers/patterns floating at the freesurface (e.g., light floating debris or foam or boils createdat the free surface by turbulence). These tracers are, how-ever, not always available or in sufficient quantities innatural streams, therefore they may need to be added atthe free surface. Another favorable situation is when spec-ular reflection formed by incident light interacting with thefree-surface deformations can be used as seeding surrogate.These deformations, with typical wavelengths in the 2 to4 cm range are also used to generate the backscattering inthe measurements with radar-based velocimeters. The free-surface waviness is generated by wind or large-scale turbu-lence structures intersecting the free surface. Using the lightintensity variation associated with the free-surface deforma-

Figure 1. LSPIV measurement sequence: (a) imaging the area to be measured (white patterns indicatethe natural or added tracers used for visualization of the free surface), (b) the distorted raw image, and(c) the undistorted image with the estimated velocity vectors overlaid on the image.

2 of 14

W00D19 MUSTE ET AL.: LARGE-SCALE PARTICLE VELOCIMETRY W00D19

tions, field measurements have been successfully obtainedusing this tracer substitute [e.g., Creutin et al., 2002; Fujitaand Hino, 2003]. When none of the above favorablesituations occur, artificial seeding is needed. For both‘‘naturally occurring’’ and artificial seeding the key require-ments is that they have to accurately follow local flowmovements. Tracer inertia and submergence are primaryfactors determining flow visualization suitability. Adversefactors for flow seeding can be strong winds at the freesurface or aggregation of the seeding particles induced byparticle-to-particle electrostatic forces or high-velocity gra-dients in the flow.[10] The framing of the flow during the recording is

decided by the availability of light and tracers at the freesurface. The size of the image is commensurate with itsresolution and the capability to distinguish movement of thewater body in image pairs. There are situations when severalimages are acquired successively from various locations andsubsequently assembled to cover with measurements thearea of interest. Extensive LSPIV measurements acquired ina wide range of laboratory and field measurement condi-tions indicate a 30 Hz sampling rate of the conventionalvideo systems is adequate for capturing velocities encoun-tered in hydraulic and hydrologic applications. This is incontrast to the complex and sophisticated laser-based sys-tems that are employed in conventional PIV. The use ofconventional video systems for LSPIV is advantageoussince the imaging devices continue to improve in spatialand temporal resolution.

2.2. Image Orthorectification

[11] River surface images are usually recorded from abridge or river bank using an oblique angle to the freesurface plane (see Figure 2a). In order to extract accurateflow data from such images, they have to be rectified by anappropriate image transformation scheme [Mikhail andAckermann, 1976]. Generally, a conventional photogram-metric relation is applied to produce orthoimages usingknown coordinates of ground control points (GCPs) in thereal (X, Y, and Z) and the image (x and y) coordinatesystems, as shown in Figure 2. The mapping relationshipsbetween the two systems is [Fujita et al., 1998]

x ¼ A1X þ A2Y þ A3Z þ A4

C1X þ C2Y þ C3Z þ 1; y ¼ B1X þ B2Y þ B3Z þ B4

C1X þ C2Y þ C3Z þ 1;

ð1Þ

where the eleven mapping coefficients A1–C3 can bedetermined by the least square method using the knownGCPs coordinates. A minimum of 6 GCPs are needed forconducting the transformation. The control points aresurveyed in the field using specialized equipment. TheGCP selection is often dictated by what is accessible out inthe field (e.g., trees, power line poles, building corners, etc.)rather than what is desirable. The effects of radial lensdistortion throughout an image must be corrected beforeestablishing the above relations. As a rule, the size of thenon-distorted image should be nearly the same as the size ofthe original image.[12] In addition to the geometrical transformation applied

to homologous points in the two coordinate systems, areconstruction of the pixel intensity distribution is simulta-neously made to obtain the orthorectified (nondistorted)image. Intensity reconstruction at a point in the transformedimage is obtained using a cubic convolution interpolation ofthe intensity in 16 neighboring points of the original image[Muste et al., 1999]. The nondistorted image contains theflow image to be analyzed and possibly regions surroundingthe flow which are not needed for analysis. Increasedcomputational efficiency and processing accuracy aregained if these regions are discarded (masked) from theanalysis before processing.

2.3. Image Processing

[13] The LSPIV algorithms for estimating velocities arethe same with those used in conventional high-densityimage PIV [Adrian, 1991]. In essence, a pattern matchingtechnique is applied to image intensity distribution in aseries of images, as illustrated in Figure 3. The similarityindex for patterns enclosed in a small interrogation area (IA)fixed in the first image is calculated for the same-sizedwindow within a larger search area (SA) selected in thesecond image. The window pair with the maximum valuefor the similarity index is assumed to be the pattern’s mostprobable displacement between two consecutive images.Once the distance between the centers of the respectivesmall window is obtained, velocity can be calculated bydividing it with the time difference (dt) between consecutiveimages. This searching process is applied successively to allIAs in the image.[14] Our image velocimetry algorithm uses the cross-

correlation coefficient as a similarity index [Fujita et al.,1998]. Cross correlation is computed between an interroga-tion area (IA) in the first image and interrogation areas

Figure 2. Relationship between the camera and field coordinate systems.

W00D19 MUSTE ET AL.: LARGE-SCALE PARTICLE VELOCIMETRY

3 of 14

W00D19

located within a search area (SA) in the second image. Thepair of particles showing the maximum cross-correlationcoefficient is selected as a candidate vector. In this methodthe cross-correlation coefficient, Rab, is defined as

Rab ¼

PMX

x¼1

PMY

y¼1

axy � �axy� �

bxy � �bxy� ��

PMX

x¼1

PMY

y¼1

axy � �axy� �2 PMX

x¼1

PMY

y¼1

bxy � �bxy� �2( )1=2

ð2Þ

where MX and MY are the sizes of the interrogation areas,and axy and bxy are the distributions of the gray-levelintensities (ranging from 0 to 255 for an 8-bit image) in thetwo interrogation areas separated by the time interval dt (seeFigure 3). The overbar indicates the mean value of theintensity for the interrogation area. For improving themeasurement accuracy, subpixel peak detection methodsusing Gaussian fitting or parabolic fitting is applied to thecross-correlation distribution [Fujita et al., 1998].[15] Our image processing algorithm is similar to the

correlation imaging velocimetry of Fincham and Spedding[1997]. Both algorithms use a variance normalized correla-tion, in which each pixel in the IA is equally weighted, suchthat the background is just as important as the particleimages. Consequently, the algorithm can estimate velocitiesfrom low-resolution images, such as those captured bystandard video cameras. Another important feature of ouralgorithm is the decoupling of the interrogation area from itsfixed location in the first image to any arbitrary location inthe second image (see Figure 3). This process completelyeliminates the velocity bias error [Adrian, 1991]. It alsogreatly improves the signal-to-noise ratio in the presence oflarge displacements, significantly extending the dynamicrange of the velocity measurement. More importantly, thealgorithm allows the use of relatively small sampling areas,which significantly increases the available spatial resolutionand reduces the errors encountered when measuring high-vorticity flows.

3. Measurement Outcomes and Accuracy

[16] The discussion in this section uses for illustrationpurposes, results obtained with LSPIV in a laboratory model

[Muste et al., 2004b]. The raw LSPIV measurements areinstantaneous vector fields (see Figure 4b). Each IA encom-passed in the original free surface image (see Figure 4a) hasa vector attached. The technique is the only available thatprovides instantaneous velocity measurements on a plane.The LSPIV vector field so obtained makes it possible toconduct Lagrangian and Eulerian analysis for determiningspatial and temporal flow features such as the mean velocityfield, streamlines, and vorticity (see Figures 4c, 4d, and 4e)as well as other velocity-derived quantities (strain rates,fluxes, dispersion coefficients due to shear, etc).[17] The LSPIV surface velocity in conjunction with

bathymetry can provide flow rates in streams. The methodused for estimation of the discharge is the velocity areamethod (VAM), as illustrated in Figure 5. The channelbathymetry can be obtained from direct surveys usingspecialized instruments (e.g., sonars or acoustic Dopplercurrent profilers). The channel bathymetry can be surveyedat the time of the LSPIV measurements or prior to themunder the assumption that bathymetry is not changing in thetime interval between the bed and free-surface measure-ments. Surface velocities at several points along the sur-veyed cross section (Vi in Figure 5) are computed by linearinterpolation from neighboring grid points of the PIV-estimated surface velocity vector field (Vs). Assuming thatthe shape of the vertical velocity profile is the same at eachpoint i, (see Figure 5) the depth-averaged velocity at each ivertical is related to the free-surface velocity by a velocityindex. The discharge for each river subsection (i, i + 1) iscomputed following the classical VAM procedure [Rantz,1982].[18] The index velocity value was, and continues to be, a

subject of research [Polatel, 2005]. The index is dependenton the shape of the vertical velocity profile, which isaffected by the flow aspect ratio, Froude and Reynoldsnumber, micro and macro bed roughness, and relativesubmergence of the large-scale roughness elements. Anattempt to articulate this intricate dependence was madeby Polatel [2005] in a series of laboratory experiments withvarying velocity flows over smooth bed and bed roughenedwith dunes and ribs. For these experimental conditions, thevelocity index varied between 0.789 and 0.928. The resultsshowed that the velocity indices are higher for smooth bedand larger flow depths. Considering the substantial changes

Figure 3. Conceptualization of the LSPIV image processing algorithm (the patterns in the images aboveare usually formed by clustering of smaller particles of the same nature, i.e., foam, leaves, or artificialseeding added to the surface for collecting the measurements).

4 of 14


in the roughness conditions, it was, however, concluded thatthe range of variation of velocity indices was fairly small. Itis obvious that more research is needed to further explorethe variation of the indices for other ranges of conditions

and assemble a matrix of indices covering the range ofnatural flow situations. For the LSPIV results presentedherein, a value of k = 0.85 for the index velocity is used.This value is generally accepted for river flows by the

Figure 5. LSPIV-based discharge measurement procedure.

Figure 4. LSPIV results [from Muste et al., 2004b] (with permission from ASCE): (a) video frame ofthe upstream reach of a 5 m � 40 m hydraulic model, (b) instantaneous vector field superposed on anundistorted video frame, (c) comparison of LSPIV velocities with ADV velocities in a cross section,(d) mean vector field, (e) streamlines established on the mean vector field, and (f) vorticity field establishedfrom the mean vector field.


5 of 14

W00D19

hydraulic community and used in conjunction with othermeasurement techniques [Costa et al., 2000].[19] The spatial nature of the LSPIV measurement com-

plicates the LSPIV uncertainty analysis. For instance,because image perspective distortion is always affectingthe field recorded images, transformed image quality is notuniform; objects in the near field are better resolved thanthose in the far field. Further, nonuniform seeding densitiesover the area to be measured will result in inadequate flowvisualization. Consequently the accuracy of the velocityobtained with LSPIV varies spatially, depending on theobliqueness of the image distortion, seeding density anddistribution, local illumination and other factors. A total of27 elemental error sources have been identified that affectthe LSPIV measurements [Kim, 2006]. The errors aregenerated in all stages of the LSPIV measurement process,i.e., illumination, seeding, recording, transformation, andprocessing. The sensitivity analysis for the LSPIV velocityuncertainty, conducted by Kim [2006], indicates that therelative contribution of the elemental errors to the finalresults is mostly affected by (listed in order): seedingdensity, identification of the GCPs, accuracy of flow tracingby the seeding particles, and sampling time.[20] The present authors attempted to estimate the LSPIV

measurement accuracy using both standardized uncertaintyanalysis methodology [American Institute for Aeronauticsand Astronautics, 1995] as well as by comparing LSPIVwith alternative instrument measurements. Most of thetwenty seven elemental error sources needed for conductingthe standardized uncertainty analysis have yet to be esti-mated because of the prohibitive degree of processing andexpense required for assessment. Use of the AmericanInstitute for Aeronautics and Astronautics uncertainty anal-ysis in conjunction with the best available information onelemental error sources to a LSPIV measurement situationin adverse field conditions (low visibility) led to an averagetotal error in velocity of 10% and maximum error of 35%[Kim, 2006]. Several assessments of the LSPIV velocityaccuracy were made through direct instrument comparisons.Comparison of LSPIV velocity was obtained in the labora-tory by moving a cart over a fixed surface containinggraphical patterns with the cart velocity resulted in averagedifference of 3.5% [Muste et al., 1999]. Comparisons ofvelocities obtained with LSPIV in field conditions andacoustic Doppler velocimeters at the same measurementlocation displayed differences up to 10% [Muste et al.,2004b]. The same comparison against mechanical currentmeters showed a 16% difference [Bradley et al., 2002].[21] The accuracy of the discharge measurements using

LSPIV is slightly better compared to the velocity because ofthe inherent spatial averaging involved in the estimation ofdischarges with the velocity area method [Muste et al.,2004a]. For example, the LSPIV estimated discharges for arelatively small creek (width of about 12m) were 2% higherthan those conducted simultaneously with an acousticStream PRO velocity profiler (TeledyneRDI Inc). TheUSGS rating curve located at the measurement site indicated3.5% higher discharge than the LSPIV estimate. Concurrentmeasurements in a larger river (width of about 70 m),displayed LSPIV discharges 5.6% lower than the readingof the USGS gauging station located at the measurement

site and 1.4% higher than the discharge provided by anADCP (TeledyneRDI Inc).

4. LSPIV Evolution

[22] The second author placed the foundation of PIV forhydraulic applications in the late 1980s [Fujita and Komura,1988]. Since 1994, three research institutions, KobeUniversity, The University of Iowa’s IIHR–Hydroscienceand Engineering (IIHR), and the Institute National Poly-techniqueGrenoble (INPG), have been actively collaboratingon LSPIV developments. Efforts at IIHR and Kobe Univer-sity were initiated by the first two authors, respectively.Developments at INPG were initiated by Creutin [2001]and by the third author, an alumna of INPG. For more thanone decade, the authors have collaboratively addressed themultifaceted aspects of LSPIV. The first decade of LSPIVdevelopment and implementation was dominated by theadjustment of the conventional PIV techniques and algo-rithms to measure large-scale flows specific to hydraulic andhydrologic applications and the transfer of the laboratoryexperience to field conditions. Over the years, areas from100 to 5000 m2 have been mapped nonintrusively withLSPIV to provide instantaneous surface velocity vectorfields, document flow patterns, and measure river discharges[Fujita et al., 1998; Fujita and Aya, 2000;Muste et al., 2000;Bradley et al., 2002; Muto et al., 2002: Creutin et al., 2003;Muste et al., 2004b; Hauet et al., 2006; Hauet, 2006]. Theinitial success of this research has attracted the interestof other researchers to LSPIV [e.g., Muller et al., 2002;Admiraal et al., 2004, Harpold and Mostaghimi, 2004,Weitbrecht et al., 2007; Chen et al., 2007].[23] Subsequently, the LSPIV flow visualization capabil-

ity and reliability was compared to other measurementinstruments through laboratory investigations [Muste et al.,2000, Kim et al., 2007, Hauet et al., 2008a]. Error analysisdue to imaging from oblique angles was estimated in acontrolled laboratory environment [Muste et al., 1999, Kim,2006]. The coupling of image velocimetry and numericalsimulation for inference of flow information outside themeasured area has been investigated through severalapproaches [Muste et al., 2000; Bradley et al., 2002; Jodeauet al., 2008]. More recently, LSPIV was fitted with newimage enhancement and processing algorithms developed byFujita et al. [2007a] and Hauet et al. [2008b]. Concurrent,ongoing hardware improvements are improving LSPIVconfigurations and operational capabilities. Most notable isthe advancements of commercial computers, digital cameras,and surveying equipment. These upgrades have been incor-porated in new LSPIV configurations as soon as theybecame available. A summary of recent LSPIV configu-rations developed by the present authors is presented inFigure 6. Each of these LSPIValternatives was developed toaddress a specific purpose (see Figure 6 for more details).[24] 1. Space-time image velocimetry measures without

seeding.[25] 2. Large-scale adaptive PIV is used for verification of

image processing and transformation robustness.[26] 3. LSPIV simulator is used for assessment of the

measurement accuracy.[27] 4. Real-timeLSPIVprovides continuousmeasurements.[28] 5. Mobile LSPIV uses the technique at ungauged

sites or during floods.

6 of 14


Figure 6. Configurations developed for the improvement of LSPIV performance and capabilities.


7 of 14

W00D19

Figure 6. (continued)

8 of 14


[29] 6. Controlled surface wave image velocimetry pro-vides measurements without seeding.[30] 7. River digital mapping provides quantitative infor-

mation about the stream velocity and waterway surrounding(bank, floodplain) using the same instrument.

5. Implementation Examples

[31] The following are sample LSPIV measurements thataim at illustrating the capabilities of the technique toquickly and remotely measure whole-velocity fields overlarge flow areas. Two types of measurement are portrayed inthis section: measurement and mapping of the flow distri-bution (1) during floods and (2) in the vicinity of hydraulicstructures. While the examples are not exhaustive, they areintended to illustrate that LSPIV can quickly and safely takemeasurements in natural-scale streams for providing com-prehensive, quantitative flow information over a wide rangeof flow types (uniform, nonuniform) and measurementconditions (e.g., floods, low, shallow flows) with minimumor no site preparation.

5.1. Floods

[32] In most cases, flow velocity measurements duringfloods cannot be conducted because of the danger posed byhigh water velocities on equipment deployment and opera-tion. A safer alternative is to record images of the flow freesurface from shore or from the air (see Figure 7). Videorecordings of the flooded areas was often the only rawinformation needed for measuring flow, as the seeding andGCPs were readily available in those images [Fujita andKomura, 1994; Fujita et al., 2007b]. Seeding of varioustypes can be generated and enhanced by the large velocitiesand turbulence occurring during flooding. First, there isreasonable probability that vigorous kolks, boils, and ripplesgenerated by the large-scale turbulent eddies moving overthe bed forms and roughness will be observable at the freesurface. These free surface perturbations act as reflectors ofthe sun or other dominant ambient light playing the role of

natural seeding for the LSPIV measurements [Fujita et al.,2007b, Figure 6a]. A second type of tracer that can occurnaturally during floods is created by large-scale turbulenteddies entraining sediment throughout the flow depth[Fujita and Hino, 2003, Figure 6b]. The entrainment resultsare the sediment ‘‘clouds,’’ patterns that can be distin-guished by their color gradients. The third tracer type,ubiquitous during floods, is the floating debris conveyedby the stream. The latter is easily observable during daylightirrespective of the illumination situation [Fujita and Komura,1994].

5.2. Hydraulic Structures

[33] An increasing number of ecohydrological applica-tions (such as studies of stream restoration, the design offish-friendly structures, and creation of suitable riverineecohabitat) requires documentation of the flow field alongthe river banks, around and through river structures, or overthe entire river reach. LSPIV can provide these flow fieldsas it is the only measurement techniques capable of simul-taneously measuring two-component velocities on a surface,rather than at isolated points or along lines. Illustrativesample measurements are provided in Figure 8, wherevelocity fields within and over groins in Uji River (Japan)are measured for a range of flow conditions [Fujita et al.,2003].[34] The area encompassed between two consecutive

groins is approximate 40 m by 20 m. To capture details ofthe flow structure within the groins required additionalseeding besides the one provided by the naturally occurringripples at the free surface. For this purpose, biodegradablepackaging particles were strategically released at few pointsin the vicinity of the groins. Recordings of the seeded flowareas were taken for about five minutes. The measurementsclearly document changes that occur in the flow distributionwithin a groin pair when the river flow depth changes (seeFigures 8b and 8c). Assembling multiple recordings of theflow over consecutive groin pairs allows documenting the

Figure 7. Mean flow distribution during floods measured from helicopter (Japan): (a) cross section inthe Katsura River (river width 90 m) and (b) flow distribution measured during a levee breach on theShinkawa River (river width 80 m).


9 of 14

W00D19

flow distribution over a set of groins, as illustrated inFigure 8d. The vectors fields plotted in Figures 8e and 8fclearly illustrate the change in the free surface flow distri-bution with the change in the river stage.

6. Lessons Learned and Future Research

[35] This section will first review key findings regardingtechniques and later offer research recommendations.Research and testing conducted by the present researchteam strongly suggests that LSPIV is a promising techniquefor documenting the hydrodynamics of riverine environ-ments. Our cumulative experience with digital imageryfacilitated an appreciation of both the capabilities andlimitations of LSPIV as well as increased awareness ofresearch challenges regarding further applications in fieldconditions.[36] Although the new acoustic and radar-based technol-

ogies can nonintrusively provide velocities at a point on or

along a line, respectively, the key advantage of imagevelocimetry is that it instantaneously measures velocitiesin a flow plane. The use of images instead of transduceroutput such as signals, makes image velocimetry more userfriendly. The technique does not require calibration andallows reprocessing of the raw information with variablespatial and temporal resolutions to obtain flow data. Themean vector field, turbulence characteristics, flow patterns(streamlines, pathlines), vorticity, and discharges can all bereadily obtained from the raw image-based measured ve-locities. The grid-attached nature of the measurementscomplements efficiently the requirements for the calibra-tions/validations of the numerical simulations. The abovecapabilities have rapidly established image velocimetry as apreferred choice for documenting detailed turbulence oftwo- and three-dimensional laboratory flows.[37] Selected characteristics of LSPIV comparedwith other

contemporary river measurement techniques are summarizedin Table 1. Regarding practical LSPIV implementation, the

Figure 8. Mean flow distribution documented with LSPIV: (a) raw image of the river flow near groins,(b) velocity distribution within a nonsubmersed groin pair, (c) velocity distribution over a submersedgroin pair, (d) orthorectified image of the flow area encompassed by four groin pairs, (e) velocitydistribution within three nonsubmersed groin pairs, and (f) velocity distribution over three submersedgroin pairs.

10 of 14


present authors conclude that if the human eye can detectthe movement of a water body, LSPIV can capture andquantify it. The relatively low flow velocities in hydrologicapplications allow for the use of simple image registrationdevices. Specifically, standard video recording equipmentand natural illumination suffice for acquiring images. Thesefeatures contribute to an efficient and inexpensive techniquecompared to existing point or line velocity instruments. Thedigital domain of the components considerably facilitatesdata management and makes LSPIV feasible for real-timesystem configurations.[38] A unique LSPIV feature among the new generation

of river instruments is swift and convenient estimation ofwhole-field velocity for stream segments during extremeflows (i.e., droughts or floods) without flow contact. Duringdroughts, rivers are shallow and velocities are low, such thatthere are few (if any) alternative measurement instrumentsto use [Weitbrecht et al., 2007]. For such situations, acousticinstruments cannot be deployed because of the small flowdepths while radar will not operate in the absence of thefree-surface waviness. LSPIV is the desirable alternative inhigh flow measurement situations because the fast velocitiespose risks to equipment and staff operating the acousticinstruments on one hand and the radar measurements arecompromised by nonuniformity of waviness on the other.[39] The visualization of the free surface for the presented

LSPIV measurements was accomplished with a variety oftracers. A favorable measurement situation is when naturallyoccurring seeding or patterns are floating at the free surface.They can consist of foam, light debris or small free-surfacedeformations created by turbulent eddies intersecting thefree surface. In the absence of ‘‘natural seeding’’ adequatemeasurement conditions must be created. Synthetic materialcan be chosen to visually contrast against the backgroundwater body’s color, accurately trace the flow, and be largeenough to be detected at the individual particle level or inparticle clumps. Convenient, natural seeding alternatives are

dry, light vegetation, wood debris or other ecologicallyharmless materials such as the biodegradable ecofoam pea-nuts (an off-the-shelf granular packaging material contain-ing 99% corn syrup). For accurate data collection, theparticles must closely trace the flow features of interest,be partially submersed to avoid possible complications dueto wind interference, and be uniformly dispersed onto thefree surface within the measured area.[40] Properly aiming the camera at the flow area of study

is critical in several respects. Whenever possible, the camerashould be placed at a comparatively high vantage point withthe optical axis perpendicular to the flow direction. Imagingfrom a small tilting angle will compromise image integrityvia distortions that are subsequently difficult to correct withthe image transformation algorithms. A tilting angle of10 degrees was found by Kim et al. [2007] as the acceptablelimit. The presence of glares or shadows on the free surfacewill hamper the image processing resulting in erroneous orabsent vectors. Diffuse light or midday recordings arerecommended for minimizing these detrimental illuminationeffects. Measures must be taken to minimize transmissionsof extraneous physical vibrations to the camera duringexposures (i.e., gusts of wind, nearby heavy construction,vehicle traffic, or inadvertent human camera jostling). Theimage orthorectification is directly related to the accuracy ofthe surveying equipment used for the GCPs. While a totallystationary survey is always recommended where measure-ment accuracy is crucial, fast methods such as those basedon laser and radar ranging or handheld Global PositioningSystems are working alternatives albeit under continuousscrutiny with respect to their accuracy. An alternative imageorthorectification algorithm that does not require a survey ofGCPs is currently being studied by the authors. Thealternative approach uses a three-dimensional conformalcoordinate transformation [Wolf, 1983], whereby the inputdata is the camera orientation, its distance from the watersurface, and the camera optical parameters.

Table 1. Comparison of Selected Characteristics of the Acoustic Doppler, Radar-, and Image-Based Velocimeters for Riverine

Environments

Technique or Characteristics Acoustic Radar Image

Measurement type Profile: along the acousticbeam path (verticals);three-componentvelocity.

Point: at the intersection ofthe beam with the freesurface; one-componentvelocity.

Surface: instantaneous vectorfield at the free surface;two-component velocity.

Flow tracers Small particles usuallynaturally suspended inthe water column.

Small surface wavescreated by wind or byflow turbulence at thefree surface.

Foams, debris, ice floes, andspecular reflections on thefree surface deformations(waves, boils, kolks).Added seeding.

Operating constraints Instrument probe incontact with the flow.The flow assumedhorizontallyhomogeneous.

The ratio between theIncident electromagneticand water wavewavelengths withinspecified range.Instrument aligned withthe dominant velocity.

Survey of minimum six pointswithin the imaged area.Occasionally, additionalseeding and illumination.

Output quality Good spatial and temporalresolution. Inaccuratefor very slow flows.

Limited spatial andtemporal resolution. Noreverse flows. Inaccuratefor very slow flows.

Good spatial andtemporal resolution.


11 of 14

W00D19

[41] The present LSPIV applications were recorded withdigital video (DV) cameras, digital cameras, and high-definition (HD) video cameras. Data suggests the qualityof recordings to be directly related to the size of the imagingdevice sensor. HD cameras facilitate detailed measurementsof large-scale turbulence, but require large computermemory for storage. For most riverine applications, cameraswith approximately 1000 � 1000 pixels are of sufficientresolution. The continuous improvements in temporal andspatial resolutions of the standard image recording equip-ment hold, however, great promise for enhancements ofLSPIV performance and expanding its applicability at nocost to the hydrologic and hydraulic communities.[42] The current LSPIV developmental efforts continue to

be focused on overcoming challenges posed by measure-ments in natural rivers conditions and making the techniquerobust and reliable for a wide range of measurementenvironments and conditions. At the time of this writing,the authors are investigating new visualization means (e.g.,particles, specular reflection) and their respective flowtracing accuracy. Recognizing that the ability to detectmotion from a series of digital images is the most criticalissue for LSPIV field applications, this research team isdeveloping algorithms that will improve image quality forpoor visibility conditions and processing algorithms to morerobustly match image patterns in recordings with lowparticle densities. The index velocity used to determinedepth-averaged velocities for various flow conditions andbed roughness is another subject of research. Additionalefforts are underway to turn the detrimental wind effect intoan advantage: preliminary experiments suggest that a con-stant wind on the free surface can be successfully used astracer for an ‘‘unseeded’’ flow over a wide range of dynamicconditions [Muste et al., 2005]. With the wind effectsproperly identified, the LSPIV wind-affected measurementcan be compensated to provide reliable estimates of under-lying water body movements. Finally, developing LSPIVsystems capable of continuous day or night operation areplanned.[43] In addition to the efforts dedicated to improve the

overall performance of the technique, the ongoing researchwill also target expansion of the LSPIV measurementcapabilities. Preliminary laboratory tests suggest that forflow situations where the shallow water theory framework isvalid, LSPIV can be used to determine changes in thechannel bed geometry on the basis of the divergence ofthe velocity vector field measured at the free surface [Musteet al., 2004b; Hauet et al., 2007]. In general, since thevariations of flow velocities within a river reach are relatedto variations in channel bed bathymetry, additional flowkinematic analyses will be elaborated to diagnose morecomplex changes in bed geometry.[44] Another series of preliminary research have identi-

fied correlations between the free surface texture and thestate of the channel. The correlations are determined by thecombined effect of channel roughness, relative roughnesssubmergence and flow velocity [Polatel, 2006; Polatel etal., 2006]. Investigations of the relationship between free-surface appearance and the velocity distribution in thevertical will continue as this subject is not only critical forLSPIV discharge measurements, but it is important for the

understanding of gas and heat transfer and sediment trans-port processes.

7. Conclusions

[45] The application of LSPIV in a measurement situationrequires a full understanding of the instrument underlyingprinciples and of various parameters involved in imagerecording and processing, as well as of the flow subjectedto measurements and its interaction with the surrounding.Under laboratory conditions the LSPIV approach yieldsreliable and accurate measurements. Field measurementsmay be confounded by poor free-surface illumination,scarce seeding, or adverse conditions acting on the freesurface (such as strong winds). These factors can drasticallyreduce measurement accuracy or prevent measurementsentirely. In typical flow situations, where almost all LSPIVrequirements are met, the technique provides accuratevelocity measurements compared to point-based and profil-ing instruments that require considerable efforts to obtaincomparable data. In some situations, such as measurementsduring extreme flow events (floods, hurricanes) or veryslow and shallow flows (wetlands, small streams), LSPIVmay be the only measurement alternative. LSPIV should notbe construed as the magic instrument. It is most appropriatewhen considered as a complementary alternative of anoverall measurement strategy. In this context, it robustlysupports a variety of measurement purposes.[46] Since its inception more than a decade ago, LSPIV

has continuously capitalized on the symbiosis of imagingtechnology, engineering, and computer science to produce apromising measurement tool for hydraulics and hydrology.The mobility, autonomy, and the expedient measurementprocedures make the LSPIV ideal for intensive measure-ment at sites deemed otherwise difficult to access duringnormal and extreme hydrological events. In addition toconventional one-dimensional river measurements (i.e., dis-charges), this new technology brings in new measurementcapabilities at reduced costs and increased accuracy (par-ticularly in extreme conditions). The LSPIV capacity toprovide 2-D and 3-D river information may shed light oncritically important processes, such as interaction betweenmain channel and overbank (floodplain), floods, the impactof floodplain flows on riparian vegetation and habitat,evolution of meandering streams, and the effect of riverstructures on the river ecosystem. The higher-dimensionalflow component estimates provided by LSPIV may alsolead to advances in river monitoring of channel stabiliza-tion, bathymetry change due to dam removal, bank erosion,stream and wetland ecology, stream corridor restoration, andenvironmental impact.

[47] Acknowledgments. The developments described in this paperwere made possible through important contributions by many researchers,technicians, and shop personnel. LSPIV supporters and developers atuniversity institutions include A. Kruger, A. Bradley, W. Krajewski, andLarry Weber (IIHR): S. Komura (Gifu University); S. Aya (Osaka Instituteof Technology); R. Tsubaki (Nagoya University); Y. Muto (Kyoto Univer-sity); T. Okabe (Tokushima University); and D. Creutin (INPG France).Several Ph.D. and master’s students have also contributed to developments(Y. S. Kim, K. Yu, J. Schone, Z. Xiong, and Z. Li). Funding supportingLSPIV developments and implementation has been provided by IIIHR;Iowa Department of Transportation; University of Kobe’s Grant-in-Aid forScientific Research, River Environment Fund, Tokushima University; INP

12 of 14


Grenoble; Cemagref Lyon; and OHM-CV Observatory. The authors grate-fully acknowledge their contributions.

ReferencesAdmiraal, D. M., J. S. Stansbury, and C. J. Haberman (2004), Case study:Particle velocimetry in a model of Lake Ogallala, J. Hydrol. Eng.,130(7), 509–607.

Adrian, R. J. (1991), Particle-imaging techniques for experimental fluidmechanics, Annu. Rev. Fluid Mech., 23, 261–304.

American Institute for Aeronautics and Astronautics (1995), Assessment ofwind tunnel data uncertainty, Standard S-071-1995, New York.

Aya, S., I. Fujita, and M. Yagyu (1995), Field observation of flood in a riverby video image analysis, Proc. Hydraul. Eng., 39, 447–452.

Bradley, A. A., A. Kruger, E. Meselhe, and M. Muste (2002), Flow mea-surement in streams using video imagery, Water Resour. Res., 38(12),1315, doi:10.1029/2002WR001317.

Chen, C., H. Chen, and H. Tang (2007), Large-scale particle tracing velo-cimetry for modeling free-surface flows, paper presented at 32nd Con-gress, Int. Assoc. of Hydraul. Eng. and Res., Venice, Italy.

Christensen, J. L., and L. E. Herrick (1982), Mississippi River test, Rep.DCP4400/300, vol. 1, Straza Div., AMETEK, El Cajon, Calif.

Collins, M. J., andW. J. Emery (1988), A computational method for estimat-ing sea ice motion in sequential Seasat synthetic aperture radar imagery bymatched filtering, J. Geophys. Res., 93, 9241 – 9251, doi:10.1029/JC093iC08p09241.

Costa, J. E., K. R. Spicer, R. T. Cheng, F. P. Haeni, N. B. Melcher, E. M.Thurman, W. J. Plant, and W. C. Keller (2000), Measuring stream dis-charge by non-contact methods: A proof-of-concept experiment,Geophys.Res. Lett., 27(4), 553–556, doi:10.1029/1999GL006087.

Creutin, J. D. (2001), Local remote sensing of rivers, report, Lab. Etude desTransferts en Hydrol. et Environ., Grenoble, France.

Creutin, J. D., M. Muste, and Z. Li (2002), Traceless quantitative alterna-tives for measurements in natural streams, paper presented at HydraulicMeasurements and Experimental Methods, Am. Soc. of Civ. Eng., EstesPark, Colo.

Creutin, J. D., M. Muste, A. A. Bradley, S. C. Kim, and A. Kruger (2003),River gauging using PIV technique: Proof of concept experiment on theIowa River, J. Hydrol., 277(3–4), 182–194, doi:10.1016/S0022-1694(03)00081-7.

Dinehart, R. L., and J. R. Burau (2005), Averaged indicators of secondaryflow in repeated acoustic Doppler current profiler crossings of bends,Water Resour. Res., 41, W09405, doi:10.1029/2005WR004050.

Fincham, A. M., and G. R. Spedding (1997), Low-cost, high resolutionDPIV for measurement in turbulent fluid flows, Exp. Fluids, 23, 449–462,doi:10.1007/s003480050135.

Fujita, I., and S. Aya (2000), Refinement of LSPIV technique for monitor-ing river surface flows, paper presented at Water Resources Joint Con-ference on Water Resource Engineering and Water Resources Planningand Management, Am. Soc. of Civ. Eng., Minneapolis, Minn.

Fujita, I., and T. Hino (2003), Unseeded and seeded PIV measurements ofriver flows videotaped from a helicopter, J. Visualization, 6(3), 245–252.

Fujita, I., and S. Komura (1988), Visualization of the flow at a confluence,in Refined Flow Modelling and Turbulence Measurements: Proceedingsof the Third International Symposium, Tokyo, Japan, 26–28 July, 1988,edited by A. Wada, Y. Iwasa, and N. Tamai, pp. 611–618, UniversalAcad., Tokyo.

Fujita, I., and S. Komura (1994), Application of video image analysis formeasurements of river-surface flows (in Japanese), Proc. Hydraul. Eng.,38, 733–738.

Fujita, I., S. Aya, and T. Deguchi (1997), Surface velocity measurement ofriver flow using video images of an oblique angle, paper presented at27th Congress, Int. Assoc. of Hydraul. Eng. and Res., San Francisco,Calif.

Fujita, I., M. Muste, and A. Kruger (1998), Large-scale particle imagevelocimetry for flow analysis in hydraulic applications, J. Hydraul.Res., 36(3), 397–414.

Fujita, I., Y. Muto, Y. Shimazu, R. Tsubaki, and S. Aya (2003), Velocitymeasurements around non-submerged and submerged spur dykes bymeans of large-scale particle image velocimetry, paper presented at30th Congress, Int. Assoc. of Hydraul. Eng. and Res., Thessaloniki,Greece.

Fujita, I., H. Watanabe, and R. Tsubaki (2007a), Development of a non-intrusive and efficient flow monitoring technique: The space time imagevelocimetry (STIV), Int. J. River Basin Manage., 5(2), 105–114.

Fujita, I., R. Tsubaki, and T. Deguchi (2007b), PIV measurement of large-scale river surface flow during flood by using a high resolution video

camera from a helicopter, paper presented at Hydraulic Measurementsand Experimental Methods, Am. Soc. of Civ. Eng., Lake Placid, N. Y.

Harpold, A. A., and S. Mostaghimi (2004), Stream discharge measurementusing a large-scale particle image velocimetry prototype, Eos Trans.AGU, 85(47), Fall Meet. Suppl., Abstract H23E-1173.

Hauet, A. (2006), Discharge estimation and velocity measurement in riversusing LSPIV (in French), Ph.D. thesis, Grenoble Inst. of Technol.,Grenoble, France.

Hauet, A., J.-D. Creutin, P. Belleudy, M. Muste, and W. Krajewski (2006),Discharge measurement using large scale PIV under various flowconditions—Recent results, accuracy and perspectives, in River Flow2006, edited by R. M. L. Ferreira et al., pp. 1828–1834, Taylor andFrancis, Lisbon.

Hauet, A., P. Belleudy, and M. Muste (2007), Estimation of the bathymetryof a channel using LSPIV, paper presented at 32nd Congress, Int. Assoc.of Hydraul. Eng. and Res., Venice, Italy.

Hauet, A., J.-D. Creutin, and P. Belleudy (2008a), Sensitivity study oflarge-scale particle image velocimetry measurement of river dischargeusing numerical simulations, J. Hydrol., 349(1–2), 178–190, doi:10.1016/j.jhydrol.2007.10.062.

Hauet, A., A. Kruger, W. Krajewski, A. Bradley, M. Muste, J.-D. Creutin,and M. Wilson (2008b), Experimental system for real-time dischargeestimation using an image-based method, J. Hydrol. Eng., 13(2), 105–110,doi:10.1061/(ASCE)1084-0699[2008]13:2(105).

Hauet, A., M. Muste, and H.-C. Ho (2008c), Digital mapping of riverinewaterway hydrodynamic and geomorphic features, Earth Surf, ProcessesLandforms, in press.

Holland, K. T., R. A. Holman, T. C. Lippmann, J. Stanley, and N. Plant(1997), Practical use of video imagery in nearshore oceanographic fieldstudies, IEEE J. Oceanic Eng., 22(1), 81–92, doi:10.1109/48.557542.

Jodeau, M., A. Hauet, A. Paquier, A. Le Coz, and G. Dramais (2008),Application and evaluation of LS-PIV technique for the monitoring ofriver surface velocities in high flow conditions, Flow Meas. Instrum.,19(2), 117–127.

Kim, Y. (2006), Uncertainty analysis for non-intrusive measurement of riverdischarge using image velocimetry, Ph.D. thesis, Univ. of Iowa, Iowa City.

Kim, Y., M. Muste, A. Hauet, A. Bradley, L. Weber, and D. Koh (2007),Uncertainty analysis for LSPIV in-situ velocity measurements, paperpresented at 32nd Congress, Int. Assoc. of Hydraul. Eng. and Res.,Venice, Italy.

Kim, Y., M. Muste, A. Hauet, W. F. Krajewski, A. Kruger, and A. Bradley(2008), Stream discharge using mobile large-scale particle image veloci-metry: A proof of concept,Water Resour. Res., 44, W09502, doi:10.1029/2006WR005441.

Leese, J. A., C. S. Novak, and B. B. Clark (1971), An automated techniquefor obtaining cloud motion from geosynchronous satellite data usingcross correlation, J. Appl. Meteorol., 10, 118–132, doi:10.1175/1520-0450(1971)010<0118:AATFOC>2.0.CO;2.

Melcher, N. B., R. T. Cheng, and F. P. Haeni (1999), Investigating tech-nologies to monitor open-channel discharge by direct measurement ofcross-sectional area and velocity floe, paper presented at 28th Congress,Int. Assoc. of Hydraul. Eng. and Res., Graz, Austria.

Mikhail, E. M., and F. Ackermann (1976), Observation and Least Squares,Dun-Donnelley, New York.

Muller, G., T. Bruce, and K. Kauppert (2002), Particle image velocimetry:A simple technique for complex surface flows, in River Flow 2002: Pro-ceedings of the International Conference on Fluvial Hydraulics, edited byD. Bousmar and Y. Zech, pp. 1227–1234, Balkema, Lisse, Netherlands.

Muste, M., Z. Xiong, and A. Kruger (1999), Error estimation in PIVappliedto large-scale flows, paper presented at 3rd International Workshop onParticle Image Velocimetry, Univ. of Calif., Santa Barbara, Calif.,16–18 Sept.

Muste, M., R. Ettema, and I. Fujita (2000), Image-based technique formonitoring ice motion in laboratory and field conditions, paper presentedat International Workshop on River Environments Considering Hydraulicand Hydrologic Phenomena in Snowy and Cold Regions, Japan Int. Sci.and Technol. Exch. Cent., Quebec City, Que., Canada.

Muste, M., K. Yu, T. Pratt, and D. Abraham (2004a), Practical aspects ofADCP data use for quantification of mean river flow characteristics: PartII: Fixed-vessel measurements, Flow Meas. Instrum., 15(1), 17–28,doi:10.1016/j.flowmeasinst.2003.09.002.

Muste, M., Z. Xiong, J. Schone, and Z. Li (2004b), Flow diagnostic inhydraulic modeling using image velocimetry, J. Hydrol. Eng., 130(3),175–185, doi:10.1061/(ASCE)0733-9429[2004]130:3(175).

Muste, M., J. Schone, and J.-D. Creutin (2005), Measurement of free-surfaceflow velocity using controlled surface waves, J. Flow Meas. Instrum.,16(1), 47–55, doi:10.1016/j.flowmeasinst.2004.08.003.


13 of 14

W00D19

Muto, Y., Y. Baba, and I. Fujita (2002), Velocity measurements in rectan-gular embayments attached to a straight open channel, in River Flow2002: Proceedings of the International Conference on Fluvial Hydrau-lics, edited by D. Bousmar and Y. Zech, pp. 1213–1219, Balkema, Lisse,Netherlands.

Poff, N. L., J. D. Allan, M. B. Bain, J. R. Karr, K. L. Prestegaard, B. D.Richter, R. E. Sparks, and J. C. Stromberg (1997), The natural flowregime: A paradigm for river conservation and restoration, BioScience,47(11), 769–784, doi:10.2307/1313099.

Polatel, C. (2005), Indexing free-surface velocity: A prospect for remotedischarge estimation, paper presented at 31st Congress, Int. Assoc. ofHydraul. Eng. and Res., Seoul, South Korea.

Polatel, C. (2006), Signature of the roughness and the flow regime on thefree surface, Ph.D. thesis, Univ. of Iowa, Iowa City.

Polatel, C., M. Muste, V. C. Patel, T. Stosser, and W. Rodi (2006), Freesurface response to large-scale bed roughness, paper presented at 7thInternational Conference on Hydroscience and Engineering, Int. Assoc.of Hydraul. Res., Philadelphia, Pa.

Raffel, M., C. Willert, and J. Kompenhans (1998), Particle Image Veloci-metry: A Practical Guide, Springer, New York.

Rantz, S. E., et al. (1982), Measurement and computation of streamflow:Volume 1: Computation of Discharge, U.S. Geol. Surv. Water SupplyPap., 2175.

Weitbrecht, V., M. Muste, J. D. Creutin, and G. H. Jirka (2007), Gesch-windigkeitsmessungen mit Particle-Image-Velocimetry: Labor- und Feld-messungen, inWasserbauliches Versuchswesen,BAWMitt. 90, pp. 79–90,Bundesanst. fuer Wasserbau, Karlsruhe, Germany.

Wolf, P. R. (1983), Elements of Photogrammetry, McGraw-Hill, New York.

��I. Fujita, Research Center for Urban Safety and Security, Kobe

University, Kobe 657-8501, Japan. ([email protected])

A. Hauet, DTG, Electricite de France, F-31000 Toulouse, France.([email protected])

M. Muste, IIHR–Hydroscience and Engineering, University of Iowa,323E C. Maxwell Stanley Hydraulics Laboratory, Iowa City, IA 52242,USA. ([email protected])

14 of 14


large-scale particle image velocimetry for measurements in riverine

Documents