Aeolian Research 3 (2011) 229–234

Contents lists available at ScienceDirect

Aeolian Research

journal homepage: www.elsevier .com/locate /aeol ia

Observations of wind ripple migration on an Egyptian seif dune usingan inexpensive digital timelapse camera

Ralph D Lorenz ⇑Applied Physics Laboratory, Johns Hopkins University, Laurel, MD 20723, USA

a r t i c l e i n f o

Article history:Received 13 October 2010Revised 10 January 2011Accepted 16 January 2011Available online 17 February 2011

Keywords:Aeolian ripplesBedform dynamicsTimelapse camera

1875-9637/$ - see front matter � 2011 Elsevier B.V. Adoi:10.1016/j.aeolia.2011.01.004

⇑ Tel.: +1 443 778 2903; fax: +1 443 778 8939.E-mail address: Ralph.lorenz@jhuapl.edu

a b s t r a c t

I report experiments on the observation of sand ripples on the Quattani dunes in Egypt using an inexpen-sive ($150) digital timelapse camera. Simple ripples in the fine sand had a wavelength of 10 cm and prop-agated downwind (windspeed �10 m/s) at a speed of �3 cm/min, and Y-junction defects moved onewavelength forward in the propagating ripple pattern over �6 min. This defect propagation speed of�0.5 times the ripple speed is rather slower than suggested in idealized models. Compound ripples ata nearby site were also observed – the megaripples moved at�0.2 cm/min while smaller superposed nor-mal ripples moved �3–4� faster: the small ripples did not propagate through the larger ones. Experimen-tal technique and some aspects of the design of a windsock/gnomon for use in further studies arediscussed. A similar experimental setup should be able to observe fast-moving barchans.

� 2011 Elsevier B.V. All rights reserved.

1. Introduction

A fundamental distinction and appeal of Aeolian landforms isthat they move and evolve. While systematic measurements allowquantification of these processes, they take place at rates that aregenerally too slow for the human attention span to allow directobservation. However, sequences of images acquired at regularintervals can be displayed in a timelapse movie, accelerating real-ity and allowing an observer to perceive the dynamic nature ofthese processes.

Digital or video timelapse observations have been straightfor-ward to implement in the laboratory for some years, since a videorecorder or computer can operate in secure, dry and sand-free con-ditions nearby, and electrical power is readily available. Somestriking recent examples of observations recorded in this way arethe evolution of bedforms in a water tank by Reffet et al. (2010).

Observations under field conditions are more challenging, buttwo recent technological developments now make matters muchmore straightforward. First is the development and plummetingcost of flash memory storage, which makes it possible to store hun-dreds or thousands of images in a tiny memory card or stick. Sometimelapse studies have been performed with ‘conventional’ digitalSLR cameras in weatherproof housings (e.g. Bogle et al., 2010),although the equipment costs here are still of order $1000. Sec-ondly, compact CMOS imagers have similarly fallen in cost (dueto their application in cellphone cameras, etc.) and size. The resultis that compact, low-power digital cameras with ample storage are

ll rights reserved.

now small compared with their mounting structures (makingunobtrusive installation possible) and inexpensive enough to beconsidered ‘expendable’ – the hardware cost (of order $100) isnow small compared with the logistical costs of deploying to thefield (often >$1000). This opens new possibilities for field observa-tions which might not have been attempted before, either becauseof uncertain prospects of success, or prohibitive costs, or both.

Automatic timelapse observations can be implemented withmodification to a commercial camera to allow its periodic trigger-ing by a timing circuit or microntroller (e.g. Lorenz et al., 2010):observations using such camera equipment have been applied tothe dust devil activity and to monitoring the flooding and freezingof Racetrack Playa. More recently, self-contained timelapse unitshave become commercially available for wildlife photography(with infrared motion triggers) or for making timelapse moviesof horticultural processes. Here we use the latter equipment, andit is the purpose of this note to demonstrate the capabilities of suchmodest hardware and to encourage further experimentation: thesetimelapse movie data can be a powerful supplement to more con-ventional documentation of process rate measurements in the field(e.g. Yizhaq et al., 2009; Andreotti et al., 2006; Zimbelman et al.,2009) and, indeed of geomorphic change in general.

2. Materials and methods

2.1. Equipment

A single Brinno Gardenwatchcam ™ was used in the experiment(www.brinno.com). This unit (Fig. 1 – �0.5 kg, 5 � 10 � 20 cm,�$200) is marketed for horticultural timelapse observations and

Fig. 1. Brinno Gardenwatchcam (two camera units shown to give view of front andback), a 15 cm ruler for scale: the black ABS plastic sections of the monopod/stakeare 3 cm in diameter. AA batteries and a USB memory stick are mounted in a backpanel (cover removed, not shown), which also has a dial to select the timelapseinterval. The unit is sealed hermetically and has a single power button, status LEDsand a two-position lens.

Fig. 2. Location map (an annotated Band 70 Landsat 4 TM image from 1989)

230 R.D. Lorenz / Aeolian Research 3 (2011) 229–234

records images acquired with a 1.3 megapixel CMOS imager to anAVI file on a USB memory stick (the 2 GB stick supplied can holdover 15,000 frames). The unit is nominally powered by four alka-line AA cells, although these do not last long enough to fill thememory stick. The frame rate can be selected from 1 min to1 day by a rotary switch, or a custom rate can be specified in a con-figuration file on the memory stick. The unit has a hermeticallysealed case which makes it attractive for outdoor observations.The unit has a 54� field of view, with a focus that can be set forabout 20–50 cm closeup, or 50 cm to infinity.

Images are timestamped automatically, although the absoluteclock setting is set rather imprecisely via a configuration file onthe memory stick which relies on setting the time on a laptopand transferring the file quickly to the camera. However, the rela-tive timing is reasonably accurate, and the timestamping is partic-ularly valuable in long-term applications where observations mayspan one or more nights. (This is because the camera does not re-cord images in darkness, so consecutive frames in the file can havea nondeterministic time spacing.)

showing the location of the study site, a few km north of the Bawiti-Cairo road.

Fig. 3. Setup on the crest of the Quattani dunes, looking north. A section of themounting stake with a small plastic bottle is used as a windsock; in this instancethe camera is mounted on a conventional tripod (which subsequently toppled over.)

2.2. Field site

The Quattani dunes are longitudinal (seif) dunes west of Cairo(Fig. 2), about 50 km north of the Birket Qarun lake. They lie onan extensive gravel and sand plain of Pleistocene deltaic deposits.They are described as follows in the memoirs of the IntelligenceOfficer of Bagnold’s Long Range Desert Group (Kennedy Shaw,2000): ‘‘A treeless, plantless, waterless, manless world, almost fea-tureless too save for poor, nondescript Gebel Hamid ahead, appear-ing and disappearing over the rolling gravel, and further on thelong dune lines of Qatania and Rammak, their saw-toothed sandpeaks like a string of battleships in line ahead at sea.’’ The sitewas visited in September 2010 in order to study the dune internalstructure with Ground Penetrating Radar and to ground-truth thedetailed geomorphology for the interpretation of spaceborne radarimagery. Linear dunes of this type are interesting planetary analogssince these dunes resemble those found in Cassini radar imagery ofSaturn’s moon Titan (Lorenz et al., 2006; Radebaugh et al., 2010).

The site was on the easternmost dune of the group, and thus theclosest to the Cairo-Bawiti road. The terrain is generally flat gravelapart from the dunes and is easily trafficable by four-wheel-drive

(although one must cross a pipeline near the tip of the dune) – in-deed the area is popular for recreational off-roading. The specificsite was at the rounded crest of the dune at 26�510N, 30�170E. Itwas noted in the radar experiments that the sand was exception-ally dry.

2.3. Field operations

Four movie sequences were acquired as opportunities arose inthe field, the sequences being between 45 min and 3 h in length,and were acquired between about 10am and 4pm local time on27th and 28th September 2010. In three instances the camerawas set up on the monopod stake supplied with the camera. Inone instance, a conventional camera tripod was used (Fig. 3), andalthough the legs were buried to �10 cm depth in the sand for sta-bility, after 1 h the camera was found to have toppled over. Thecamera was mounted facing northwards (i.e. away from the noonsun) with its boresight 10–20� below horizontal, and the imageacquisition interval was set to 1 min or to 20 s.

Fig. 4. Nine successive frames, each one minute apart, of the simple ripple pattern on the dune crest. The pattern moves one wavelength over about four frames. A defect(black arrow in frames 1 and 4) begins in frame 1 as a Y-junction, the downwind leg breaking away (frame 4) to form a free termination which then links to the next rippledownwind to form a new Y junction (frame 8). Contrast has been stretched to highlight the ripple pattern. White arrow in frame 1 shows the wind and ripple migrationdirection.

R.D. Lorenz / Aeolian Research 3 (2011) 229–234 231

Desirably, ripple movement conditions would be documentedcontinuously by saltation sensors and anemometers. However,since the experiment was an opportunistic one conducted in paral-lel with geomorphological mapping and GPR studies, no contextinstrumentation was available (although we did make a contextmeasurement of windspeed with a handheld anemometer at thebeginning and end of the measurements, indicating about 9 m/s ± �1 m/s at shoulder height). But much as cine cameras wereused as datalogging systems in the 1930s and 1940s before elec-tronic data acquisition became widespread, the camera itself canbe used to record contextual data if suitable instrumentation witha visual display is set up in the field of view. Such an approach hasthe advantage of providing intrinsic timetagging – the contextmeasurements recorded in the image are necessarily contempora-neous with the processes being observed without the need for anysynchronization of datafiles. Inspired by camera measurements ofwind on Mars landers (Sullivan et al., 2000; Holstein-Rathlou,2010), an improvised ‘telltale’ wind indicator was installed by sus-pending a 100ml plastic bottle via a length of string from a stake.The thinking behind using a bottle was that the mass of the bottle(and thus the tilt angle at which the moments due to its weight anddrag would balance for a given windspeed) could be adjusted eas-ily in the field by partially filling with water.

3. Results

The video files were examined in a playback application sup-plied with the camera (that allows stepping through frame-by-frame). Additionally, a utility (A4 Video Converter) was used tosplit the video file into individual images for more quantitativemeasurements using the image analysis tools ImageJ, Adobe

Photoshop and ITT’s Interactive Data Language IDL. In addition toinclusion in the online version of this paper, the video files areavailable at http://www.lpl.arizona.edu/~rlorenz.

3.1. Simple ripple migration

A simple and obvious result at the first site (video 1) is the reg-ular motion of the simple, regular ripple pattern (Fig. 4). Successivecrests pass the marker stake (gnomon) every third or fourth frame(i.e. every 3–4 min). Since the ripple crests are 10+/1 cm apart, thecorresponding propagation speed is �2.5–3.3 cm/min.

3.2. Defects

It could be seen that defects in the ripple pattern migrated in aconsistent manner. The forward branch of a Y-junction would dis-connect, and then connect with the ripple ahead of it. This processusually took �6–7 min (see Fig. 4). In our observations, the defectmigration was always forwards (downwind) in the pattern, andwas usually associated with little or no loss of crest length and insome cases the defect migration led to an increase of total crestlength.

3.3. Superposed patterns

For a few moments in the image sequence, an additional set ofsmaller ripples appear, superposed on the more typical set. It is as-sumed that these form during a gust of wind from a slightly differ-ent direction but quickly decay when the wind returns to its usualdirection.

Fig. 5. The second, complex ripple field, on the upwind side of a saddle betweentwo dune crests. The small ripples superposed on the larger set are clearly seen. Thedark line indicates the image line extracted from each frame to generate thewaterfall chart in Fig. 6.

Fig. 6. Waterfall chart assembled from the same line extracted from each of 150images, each spaced by 20 s. The crests of major ripples are seen prominently (A)and propagate slowly to the right (by about 10 cm over 50 min). The camera isrocked by the wind, causing jitter in the pattern (e.g. B) – the camera motion isalmost entirely in the vertical plane which is why vertically-oriented crests likethose at A are unaffected, but diagonal crests appear to jitter horizontally. The smallripple pattern is seen to propagate several times faster (C) than the large pattern,but when small ripples encounter the large ripple, they merge with it.

232 R.D. Lorenz / Aeolian Research 3 (2011) 229–234

3.4. Telltale/windsock

The telltale wind indicator was hastily improvised and was notas useful as might be hoped. Specifically, rather than settle at anequilibrium angle determined by the wind, it tended to move cha-otically. This effect was in fact worse when the bottle was filled toincrease its mass – the resultant system had lower sensitivity, butlower damping too. A further complication was that the cameramounting stake used to suspend the bottle was wide enough toshed vortices which caused the bottle to be continuously in motioneven in steady winds. Thus the bottle’s instantaneous position re-corded in an image does not relate directly to the instantaneouswind. However, over long time intervals it might be possible to re-late the windspeed to the peak or mean displacement of the bottlein an image sequence, but that was not attempted here.

3.5. Camera motion

The entire scene occasionally moves due to displacement of thecamera by the wind. This is generally distracting for image analysis(especially automated analysis – see later) and so should beavoided by mounting the camera as stiffly as possible (this was amotivation for trying the tripod rather than the monopod stake).On the other hand, movement of the camera might be employedas a crude measurement of windspeed.

3.6. Saltation layer

In a number of image frames, evidence of active saltation can beseen. Specifically, a ‘streaky’ pattern can be seen in the direction ofthe wind (the crosswind variations in the pattern perhaps coinci-dentally being approximately the same as the ripple wavelength),and the optical contrast of the ripple pattern is temporarily re-duced due to scattering and absorption by the lofted sand. Sincecontrast drops due to change in illumination due to cloud, etc.can also occur (noting that the camera implements a variable expo-sure, so a drop in illumination does not necessarily lead to darken-ing of the recorded image) no attempt has been made to quantifythis effect. A calibration target with which to separate illuminationfrom saltation layer effects would be useful.

3.7. Complex ripple migration

A second nearby ripple field was also observed (video 2) about30 m to the southeast of the first site. This area had a notably dif-ferent morphology (see Fig. 5), with more megaripples (�30 cm)with sharp irregular crest lines and smaller rounded ripples be-tween the megaripple crests. The morphology of this site, whichwas on a sloping upwind saddle between and to the east of twodune peaks, is rather similar to that in the Negev desert describedby Yizhaq et al. (2009).

Ripple motion was both slower and more intermittent in thistimelapse sequence than in that of the regular ripples, either dueto lower winds or a higher threshold or both. When motion oc-curred, it could be seen that the small ripples moved much fasterthan the megaripples, as would be expected. The small ripplesmoved only in regions bounded by the larger ones – they did notcross the crests of the megaripples. In the upper part of the image,frequent contrast drops were seen, indicating active saltation. Latein the sequence, contrast declined over the entire image as the sunset over the crest of the dune, casting the site into shadow makingfurther measurements difficult.

Clearly, there is a large range of arbitrarily elaborate analysesthat can be applied to timelapse data, which can be voluminous.However, by way of example, we can illustrate this differential mo-tion rather simply by extracting a horizontal line of image data and

assembling these lines into a ‘waterfall’ chart (Fig. 6). Specifically, aline from a desired position in the first frame of the sequence is in-serted as the top line in the waterfall image. Then the line from thesame position in the second image of the sequence is inserted asthe second line in the waterfall, and so on. Thus the waterfall chartis a representation wherein the horizontal coordinate is the hori-zontal coordinate in the original images, and the vertical coordi-nate corresponds to time. Thus a feature in the sequence thatremains fixed can be seen in the chart forming a vertical line,whereas an object that moves will form a slanting line.

To generate the chart, the AVI movie file was converted into asequence of separate JPEG image files using A4 Video Convertersoftware; these were then read by an IDL script which assembleda line of each grayscale image into a line in the chart. A more inten-sive analysis might shift each frame to correct for wind motion ofthe camera, although Fig. 6 shows the original data in order toillustrate the camera motion effect.

R.D. Lorenz / Aeolian Research 3 (2011) 229–234 233

4. Discussion

The principal results here are the movies themselves, whichmay be useful in pedagogical or outreach applications. However,some remarks may be made on the quantitative results that canbe obtained. Additionally we note some potential improvementsfor future observations.

4.1. Process measurements

The rate of simple ripple migration (3 cm/min) we record isstrikingly fast. The windspeed we measured (at shoulder heightwith a hand-held anemometer at the beginning and end of theobservation) was similar to that recorded by Zimbelman et al.((2009), 9 m/s) who observed motion of a 10 cm-high granule rip-ple crest of 10.5 cm over 1380 min at Great Sand Dunes NationalPark and Preserve (thus a speed of 0.007 cm/min), and of a 3 cm-high ripple of 2.1 cm in 109 min (0.019 cm/min). The small size(1–1.5 cm high) of the ripples here would account for a factor of16–27 faster movement: the remaining factor of �5–10 is presum-ably due to the smaller particle size of our sand and thus a lowersaltation threshold and hence higher volume transport rate for agiven windspeed.

The migration of Y-junction defects in a ripple pattern has beendocumented in field observations of ripples (Anderson and McDon-ald, 1990) and in computer simulations (Landry and Werner,1994). Because the tapering tip of the defect is thin, it propagatesfaster than the regular ripples, and thus bends forward to attachto the ripple ahead of it, see e.g. Werner and Kocurek (1997,1999) and Werner and Gillespie (1993). Specifically, in an idealizedmodel Werner and Kocurek (1999) suggest that the defect propa-gation speed is �3 times the ripple speed. In our observation, thedefects moved through the pattern at about half the speed thatthe pattern itself propagated past a fixed obstacle, a considerablyslower speed than suggested in the model.

4.2. Lessons in technique

With the exception of the toppled tripod, the imaging experi-ments were generally successful, and the camera appeared unaf-fected by operation in full sunlight with air temperatures of40 �C. However, some clear areas for improvement were noted.

It might be desired to mount the camera as high as possible toobtain a ‘plan view’ of the ripple pattern, with minimal foreshort-ening. However, a higher mounting point will typically lead tomore wind movement of the camera, which can be distracting dur-ing analysis. Thus a tradeoff exists which is difficult to assess byany means other than trial and error. Tethering the monopod/tri-pod to stakes may be a straightforward way to improve the stiff-ness of the mounting.

Although at least approximate navigation of images can beaccomplished with a single gnomon or scalebar as used here, givenknowledge of the angular image scale and the height of the cameramount, rather easier and more precise measurements could be ob-tained with better photogrammetric control. Specifically, installa-tion of a set (four or more) fixed markers with some suitabletwo-dimensional regular spacing (e.g. 1 m squares) would be help.

Some obvious improvements in the windsock design would beto have a more heavily-damped suspension (perhaps by using awire instead of a string) and to suspend it from a narrower struc-ture or at least out of the wake of the structure. A trade-off existsbetween having a wide arc (allowing precise measurement of thetilt angle from imagery) and a narrow one (such that the objectis being dragged does not move substantially through the velocitygradient in the boundary layer). It may be that a rigid hinged

structure like the Mars Pathfinder windsock (Sullivan et al.,2000) is more accurate than a suspended object.

Finally, a photometric target (i.e. a fixed object with patches ofdifferent reflectivity to allow calibration of illumination conditionsand thus detect contrast reduction due to a saltation cloud) wouldbe a useful augmentation to a scalebar.

Timelapse imagery can of course be used to study settings otherthan the movement of extant ripples. Other experiments (see e.g.Andreotti et al., 2006; Yizhaq et al., 2009) might include flatteningpatches of sand and observing the emergence of a ripple pattern, orto introduce artificial ripples of different wavelength and observetheir subsequent evolution.

5. Conclusions

We have measured ripple migration rates and styles under fieldconditions with modest effort using novel timelapse camera obser-vations. The generation of timelapse movies of ripple movement inthe field is, to the author’s knowledge, novel: improved measure-ments with longer duration, better wind documentation and pho-togrammetric control can be readily obtained in future work.Motion of megaripples and even of barchan dunes should be feasi-ble to observe: with barchan migration rates of 15 m/yr, observa-tions over a few months (an operation period alreadydemonstrated with these cameras – Lorenz et al., 2010) shouldshow perceptible motion.

The instrumentation is inexpensive compared with the costs offield deployment and personnel labor, so offers new possibilitiesfor both quantitative measurement of the intermittent process ofAeolian movement and for illustrating these processes to thepublic.

Acknowledgements

We thank Tarek ElMahady of Dabuka Expeditions for transpor-tation and logistics and the field team for participation in a visit toa remarkable place. Professor Jani Radebaugh of Brigham YoungUniversity is thanked for remarks on the aesthetics of gnomonplacement. Camera experiments have been supported by the NASAApplied Information Systems Research program. The expedition toEgypt was supported by the Cassini program.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at doi:10.1016/j.aeolia.2011.01.004.

References

Anderson, R.S., McDonald, R.R., 1990. Bifurcations and terminations in eolianripples. EOS 71, 1344.

Andreotti, B., Claudin, P., Pouliquen, O., 2006. Aeolian sand ripples: experimentalstudy of fully developed states. Physical Review Letters 96, 028001.

Bogle, R. C., Vogel, J. M., Velasco, M., Redsteer, M. H., 2010. Sand Dune MigrationMonitoring on the Navajo Nation, Southwestern, Abstract #2012, SecondInternational Planetary Dunes Workshop, Alamosa, CO, May 2010.

Holstein-Rathlou, C., Gunnlaugsson, H. P., Merrison, J. P., Bean, K. M., Cantor, B. A.,Davis, J. A., Davy, R., Drake, N. B., Ellehoj, M. D., Goetz, W., Hviid, S. F., Lange, C. F.,Larsen, S. E., Lemmon, M. T., Madsen, M. B., Malin, M., Moores, J. E., Nørnberg, P.,Smith, P., Tamppari, L.K., Taylor, P.A., 2010. Winds at the Phoenix landing site,Journal of Geophysical Research 115, E00E18, 2010.

Kennedy Shaw, W. B., Long Range Desert Group, Greenhill Books, 2000.Landry, W., Werner, B.T., 1994. Computer simulations of self-organized wind ripple

patterns. Physica D77, 238–260.Lorenz, R. D., Wall, S., Radebaugh, J., Boubin, G., Reffet, E., Janssen, M., Stofan, E.,

Lopes, R., Kirk, R., Elachi, C., Lunine, J., Paganelli, F., Soderblom, L., Wood, C., Wye,L., Zebker, H., Anderson, Y., Ostro, S., Allison, M., Boehmer, R., Callahan, P.,Encrenaz, P., Ori, G.G., Francescetti, G., Gim, Y., Hamilton, G., Hensley, S.,Johnson, W., Kelleher, K., Mitchell, K., Muhleman, D., Picardi, G., Posa, F., Roth, L.,Seu, R., Shaffer, S., Stiles, B., Vetrella, S., Flamini, E., West, R., 2006. The Sand Seas

234 R.D. Lorenz / Aeolian Research 3 (2011) 229–234

of Titan Cassini RADAR observations of longitudinal dunes, Science 312, 724–727.

Lorenz, R.D., Jackson, B., Barnes, J., 2010. Inexpensive timelapse digital cameras forstudying transient meteorological phenomena: dust devils and playa flooding.Journal of Atmospheric and Oceanic Technology 27, 246–256.

Radebaugh, J., Lorenz, R., Farr, T., Paillou, P., Savage, C., Spencer, C., 2010. Lineardunes on titan and earth: initial remote sensing comparisons. Geomorphology121, 122–132.

Reffet, E., Courrech du Pont, S., Hersen, P., Douady, S., 2010. Formation and stabilityof transverse and longitudinal dunes. Geology 38, 491–494.

Sullivan, R., Greeley, R., Kraft, M., Wilson, G., Golombek, M., Herkenhoff, A., Murphy,J., Smith, P., 2000. Results of the imager for mars pathfinder windsockexperiment. Journal of Geophysical Research 105, 24,547–24,562.

Werner, B.T., Gillespie, D.T., 1993. Fundamentally discrete stochastic model forwind ripple dynamics. Physical Review Letters 71, 3230–3233.

Werner, B.T., Kocurek, G., 1997. Bed-form dynamics: does the tail wag the dog?Geology 25, 771–774.

Werner, B.T., Kocurek, G., 1999. Bedform spacing from defect dynamics. Geology 27,727–730.

Yizhaq, H., O. Isenberg, R. Wenkart, Tsoar, H., Karnieli, A., 2009. Morphology anddynamics of Aeolian mega-ripples in Nahal Kasuy, southern Israel. IsraeliJournal of Earth Science 57, 149–165.

Zimbelman, J.R., Irwin III, R.P., Williams, S.H., Bunch, F., Valdez, A., Stevens, S.,2009. The rate of granule ripple movement on Earth and Mars. Icarus 203,71–76.