It is inevitable that some important specimens will become lost or damaged over time, conservation is therefore of vital importance. The Paluxy River dinosaur tracksite is among the most famous in the world. In 1940, Roland T. Bird described and excavated a portion of the site containing associated theropod and sauropod trackways. This excavated trackway was split up and housed in different institutions, and during the process a portion was lost or destroyed. We applied photogrammetric techniques to photographs taken by Bird over 70 years ago, before the trackway was removed, to digitally reconstruct the site as it was prior to excavation. The 3D digital model offers the opportunity to corroborate maps drawn by R.T. Bird when the tracksite was first described. More broadly, this work demonstrates the exciting potential for digitally recreating palaeontological, geological, or archaeological specimens that have been lost to science, but for which photographic documentation exists.
Citation: Falkingham PL, Bates KT, Farlow JO (2014) Historical Photogrammetry: Bird's Paluxy River Dinosaur Chase Sequence Digitally Reconstructed as It Was prior to Excavation 70 Years Ago. PLoS ONE 9(4): e93247. doi:10.1371/journal.pone.0093247
Editor: Peter Dodson, University of Pennsylvania, United States of America
Received: January 27, 2014; Accepted: February 28, 2014; Published: April 2, 2014
Copyright: © 2014 Falkingham et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This work was supported by a Marie Curie International Outgoing Fellowship within the 7th European Framework Programme awarded to PLF, and a National Geographic grant awarded to JOF. The Laser scans of the AMNH and TMM sections were collected during previous work funded by Jurassic foundation and Palaeontological Association grants awarded to KTB. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
The fossil footprints and trackways preserved in the Glen Rose Formation along the Paluxy River at Glen Rose, Texas (USA) are among the most famous dinosaur tracks in the world, and represent a colossal wealth of information about dinosaur palaeobiology, being exceptionally well preserved and extremely extensive. Although tracks from the Paluxy River were published on as early as 1917 –, it was Roland T. Bird, a fossil collector for Barnum Brown of New York's American Museum of Natural History (AMNH), who recognised unambiguous sauropod tracks for the first time , tracks which would later be designated the holotype of Brontopodus birdi . Among the many diverse trackways that Bird examined in the Glen Rose Formation sites across Texas, were sauropod manus-only trackways, interpreted by Bird to have been produced by a swimming sauropod , . But it is perhaps the Paluxy River's ‘chase sequence’ that is the most infamous from the area. In 1940 Bird returned to Texas on an expedition to collect portions of this ‘chase sequence’ in which a theropod was apparently following a sauropod , , .
The section of trackway collected measured over 9 m in length and 3.65 m in width, and in order to remove the tracks, the entire section was broken into large blocks . The blocks were then transported to separate locations – one part of the trackway was housed at the Texas Memorial Museum (TMM), and one part was sent to the AMNH. The final portion of the trackway has since been lost or destroyed. The trackways remain on display at their respective museums. However, in 1988 deterioration of the trackway section held at the TMM was reported, and that deterioration has continued since .
Given the significance of the tracks located in the Paluxy River, as well as the TMM and AMNH, and because the tracks are subject to the destructive forces of said river, it is imperative that on-going efforts are made to document the tracks in as accurate and systematic way as possible. Since 2008 a team has been doing just that, using overhead photography, LiDAR laser scanning, photogrammetry, and other techniques , .
Of the methods employed in the documentation of the Paluxy River tracks, photogrammetry has seen phenomenal advancement in recent years, to the point where it has attained comparable accuracy and resolution to laser scanning . Where once the generation of a 3D digital model required the transport and set up of a large, heavy, and expensive laser scanner on top of a large platform over the river (and where that laser scanner would often shut down due to the ambient heat), an accurate and high resolution, phototextured, digital outcrop model (DOM) can now be made using photos from a consumer digital camera and free software .
While the application of photogrammetry to the site has considerably increased the ease with which modern documentation can be carried out, it has also raised the exciting opportunity to retrospectively produce digital models from photographs. Much as the significance of the Paluxy River tracks has led to a major documentation effort today, so too did R.T. Bird ensure that efforts were made to record the ‘chase sequence’ prior to, and during, excavation. To this end, R.T. Bird took numerous photographs, as well as a film, during the excavation in 1940, to complement his hand-drawn cartographic maps.
Because modern photogrammetric software is able to account for unknown focal lengths and camera types, the photos taken by Bird over 70 years ago can be used to generate a digital model of the tracks as they were in 1940, in situ prior to excavation. In this paper we have taken Bird's original photographs, and used them to generate a digital model of the Paluxy River ‘chase sequence’. This model is then compared with Bird's original map drawings, as well as with laser scan data of the TMM and AMNH blocks.
Materials and Methods
Seventeen photographs (either 5 inches by 7 inches, or 8 inches by 10 inches, in size) and/or negatives made by Bird during the excavation in 1940 were used (Figure 1). Unfortunately, the photographs come from a number of stages throughout the excavation process, and as such the content of the photographs varies, including images prior to and after damming/flooding of a parallel exposed trackway, the emplacement of sandbags, and the excavation itself, with one photograph recording the breakup of the trackway section. In addition, multiple shots contain people, or tools such as picks and shovels, which do not remain stationary between images. These factors are generally detrimental when producing photogrammetric models, and should always be avoided if possible when collecting data. Obviously in this case however, the data contain these complications and there is little can be done to remove them whilst retaining primary information. It is likely that these issues will be common to historical photogrammetry.
Figure 1. Sixteen of Bird's original photographs used in the photogrammetric reconstruction of the trackway.
Note that the state of excavation (flooded parallel trackways, sandbags, tools etc) varies between images, causing complications for the reconstruction.doi:10.1371/journal.pone.0093247.g001
Bird's photographs or negatives were scanned at a resolution of 5436×7787 pixels (Figure 1). In order to process the photographs and produce a digital model, the freely available software VisualSFM – was used to match the photos and generate a sparse reconstruction. The difficulties associated with the photographs, listed above, meant that it was difficult for the software to match images; only 12 of the original 17 photographs could be matched to each other and used to generate a model.
This sparse point cloud was then processed using PMVS and CMVS (through the VisualSFM GUI) to generate a dense point cloud –. A polygonal surface was generated from the dense point cloud using the Screened Poisson Surface Reconstruction algorithm .
In addition to the digital model produced from photographs, previously collected laser scans of the TMM and AMNH sections of the trackway in their current conditions were used for comparative purposes. These models were produced using a Reigl LMS-Z420i LiDAR laser scanner during previous work ,  and readers are referred to those publications for additional details. Because photogrammetry is an inherently scale-less method, scale must be applied after the photogrammetric model is generated. In order to do this, we used measurements between distinct tracks from both Bird's original maps and the laser scans in the areas that the photogrammetric model displayed highest fidelity.
In order to compare and corroborate the photogrammetric reconstruction, two maps drawn by R.T. Bird were used – the Austin chart and the Rye chart (Figure 2) . These maps were drawn by R.T. Bird around the time the trackways were excavated. However, the drawings are not identical – in the Austin chart the trackways are much straighter, while in the Rye chart the trackways curve gently to the left (Figure 2c).
Figure 2. R.T. Bird's maps of the Paluxy ‘chase sequence.’
a) Bird's Rye chart, b) the Austin chart, and c) the Austin and Rye charts overlaid. Note that the Austin and Rye charts diverge toward the north.doi:10.1371/journal.pone.0093247.g002
The nature of the photographs - low sharpness, mobile features (people, tools) over the site, and the fact that the photographs were all taken from a roughly south facing direction - meant that parts of the resulting model, particularly at the northern end, are severely lacking in detail. In addition, the single general direction of the photographs created linear artefacts running along the length of the reconstructed model. Nevertheless, the entire sequence observable from the photographs was reconstructed in 3D, measuring over 45 m in length. Although the tracks lack fine detail, their locations are obvious on the textured model, and as such an aerial-view of the site, as it was prior to excavation, can be produced (Figure 3, Movie S1). In addition, we were able to replace the phototexture with colour corresponding to height, which accentuates the tracks, and aids in visually locating individual footprints.
Figure 3. Photogrammetric reconstruction of Bird's chase sequence.
Far left, photo-textured and height mapped plan-view of the reconstructed trackway. Track labels according to Farlow et al. (1989). Right, photo-textured and height mapped views, top to bottom; isometric view along trackway, close up of high fidelity southern end, close up of poor quality northern end.doi:10.1371/journal.pone.0093247.g003
In contrast to the northern and central portions of the trackway, the southernmost end, where 6 photographs are clustered (see Figure 1), has been reconstructed in enough detail to observe digit impressions in the theropod and sauropod tracks. This makes comparisons with R.T. Bird's original maps possible. By superimposing the digital reconstruction and the maps, it can be seen that the Rye chart, in which the theropod and sauropod trackways curve to the left, is a much better match than the Austin chart in which the trackways are much straighter.
The digital reconstruction shows high variations in quality along the length of the trackways. Those tracks in the middle section in particular are of poor quality. This portion of the reconstruction suffers because it is captured only in the upper, distant portion of half of the photographs; the remaining images do not show this portion of the trackway at all. However, the distal most portion of the trackway, near to the sandbag barrier, is of excellent quality (Figure 3, Supplementary data), with individual digit impressions of the theropod tracks visible, and variations in the base of the sauropod tracks observable in the height-mapped colouration. This shows that even with poor source photographs, highly detailed reconstructions are at least partially possible.
When R.T. Bird drew his maps of the Paluxy trackways, he did so using lengths of string – fighting against constant flooding from the river . It is unclear as to the timing of the creation of the Austin and Rye charts, but by comparing the maps with the photogrammetric reconstruction (Figure 4), it is clear that the Rye chart is the more accurate representation of the tracks as they were prior to excavation – the trackways curve gently to the left in both the Rye chart and the photogrammetric model, whereas they are straight in the Austin chart.
Figure 4. Overlays of Bird's Rye and Austin charts with photogrammetric and laser scan digital models.
a) The Rye chart, b) portion of the Rye chart reconstructed via photogrammetry with historical photos, c) Close-up of photogrammetric reconstruction, d) location of laser scans of the AMNH and TMM sections in the Rye chart, e) close-up of match between Rye chart and laser scans, f) Austin chart, g) Austin chart and photogrammetric reconstruction, h) Austin chart and laser scans of AMNH and TMM sections.doi:10.1371/journal.pone.0093247.g004
The laser scans of the AMNH and TMM sections indicate a number of theropod tracks that are not recorded in either of Bird's maps (Figure 4e, h), though these occur in an area of the photogrammetric model that is poorly resolved. If the charts and the photogrammetric model or laser scans are aligned based on the locations of the sauropod tracks, there is a slight offset to many of the theropod tracks. This may indicate that R.T. Bird measured and mapped the trackways independently of each other.
Using a world-famous dinosaur tracksite, this work has demonstrated the potential for historical photogrammetry: the reconstruction of 3D digital models of specimens, sites, or exposures which have deteriorated, or been lost to science entirely, but are recorded with photographs taken prior to loss/deterioration. In this case, Bird's ‘chase sequence’ has not existed in a complete form since it was excavated in 1940 over 70 years ago. After excavation, some parts were distributed to several institutions, and others were lost completely. With the photogrammetric reconstruction, we were able to corroborate which of two differing maps of the trackway, drawn by R. T. Bird was the most accurate.
It is an exciting prospect to think that many palaeontological or archaeological specimens that have been lost to science, or suffered irreparable damage, may be digitally reconstructed in 3D using free software and a desktop computer. We envisage that historical photogrammetry will become a powerful, common, tool in the future, particularly as advances in photogrammetric techniques enable software to compensate for the difficulties inherent in using old photographs.
Fly through of the reconstructed trackway.
We wish to express our thanks to reviewers Richard T. McCrea and Matteo Belvedere, as well as the editor.
Conceived and designed the experiments: PLF. Performed the experiments: PLF. Analyzed the data: PLF JOF. Contributed reagents/materials/analysis tools: KTB JOF. Wrote the paper: PLF KTB JOF.
- 1. Shuler EW (1917) Dinosaur tracks in the Glen Rose Limestone near Glen Rose, Texas. American Journal of Science 44: 294–298. doi: 10.2475/ajs.s4-44.262.294
- 2. Shuler EW (1935) Dinosaur track mounted in the band stand at Glen Rose, Texas. Field and Laboratory 4: 9–13.
- 3. Shuler EW (1937) Dinosaur tracks at the fourth crossing of the Paluxy River near Glen Rose, Texas. Field and Laboratory 5: 33–36.
- 4. Bird RT (1939) Thunder in his footsteps. Natural History 43: 254–261.
- 5. Farlow JO, Pittman JG, Hawthorne JM (1989) Brontopodus birdi, Lower Cretaceous Sauropod footprints from the U.S. Gulf Coastal Plain. In: Gillette DD, Lockley MG, editors. Dinosaur tracks and traces. Cambridge: Cambridge University Press. pp. 371–394.
- 6. Bird RT (1944) Did Brontosaurus ever walk on land? Natural History 53: 60–67.
- 7. Falkingham PL, Bates KT, Margetts L, Manning PL (2011) Simulating sauropod manus-only trackway formation using finite-element analysis. Biology Letters 7: 142–145. doi: 10.1098/rsbl.2010.0403
- 8. Farlow JO, O'Brien M, Kuban GJ, Dattilo BF, Bates KT, et al. (2012) Dinosaur Tracksites of the Paluxy River Valley (Glen Rose Formation, Lower Cretaceous), Dinosaur Valley State Park, Somervell County, Texas. Proceedings of the V International Symposium about Dinosaur Palaeontology and their Environment 41–69. doi: 10.1080/02724634.1988.10011689
- 9. Thomas DA, Farlow JO (1997) Tracking a dinosaur attack. Scientific American 277: 74–79. doi: 10.1038/scientificamerican1297-74
- 10. Bird R (1985) Bones for Barnum Brown: adventures of a dinosaur hunter: Texas A&M University Press.
- 11. Bates KT, Falkingham PL, Hodgetts D, Farlow JO, Breithaupt BH, et al. (2009) Digital imaging and public engagement in palaeontology. Geology Today 25: 134–139. doi: 10.1111/j.1365-2451.2009.00714.x
- 12. Bennett MR, Falkingham PL, Morse Sa, Bates K, Crompton RH (2013) Preserving the Impossible: Conservation of Soft-Sediment Hominin Footprint Sites and Strategies for Three-Dimensional Digital Data Capture. PLoS ONE 8: e60755. doi: 10.1371/journal.pone.0060755
- 13. Falkingham PL (2012) Acquisition of high resolution three-dimensional models using free, open-source, photogrammetric software. Palaeontologia Electronica 15: 1T: 15.
- 14. Wu C (2007) SiftGPU: A GPU implementation of scale invariant feature transform, SIFT. http://csuncedu/~ccwu/siftgpu.
- 15. Wu C (2011) VisualSFM: A visual structure from motion system. http://ccwu.me/vsfm/.
- 16. Wu C, Agarwal S, Curless B, Seitz SM (2011) Multicore bundle adjustment;. IEEE 3057–3064. doi: 10.1109/cvpr.2011.5995552
- 17. Furukawa Y, Curless B, Seitz SM, Szeliski R (2010) Clustering Views for Multi-View Stereo. doi: 10.1109/cvpr.2010.5539802
- 18. Furukawa Y, Ponce J (2010) Accurate, Dense, and Robust Multi-View Stereopsis. IEEE Trans on Pattern Analysis and Machine Intelligence 32: 1362–1376. doi: 10.1109/tpami.2009.161
- 19. Furukawa Y, Curless B, Seitz SM, Szeliski R (2010) Towards Internet-scale Multi-view Stereo. doi: 10.1109/cvpr.2010.5539802
- 20. Kazhdan M, Hoppe H (2013) Screened poisson surface reconstruction. ACM Transactions on Graphics 32: 1–13. doi: 10.1145/2487228.2487237