Finding Fossils with Remote Sensing

| |

Advertisement

In the field of paleontology, finding fossils is often a substantial undertaking. Researchers usually have to assemble crews of up to 15 people, each one carrying their own food and shelter, and head out into remote areas where making good time means driving 10 miles an hour. On top of that, the isolated areas where paleontologists work may or may not contain fossils. Finding them is often a matter of luck. With increasingly limited time and money for fieldwork, the solution may be to use remote sensing to help uncover fossils, and new technology is showing some promise.

Satellite images and some sophisticated technology have the potential to help paleontologists find more reliable fossil sites. For work in the Great Divide Basin of Wyoming, Professors Robert Anemone and Jay Emerson, colleagues at Western Michigan University, developed a system that analyzes images from satellites and tags potential areas that could contain fossils. Researchers still have to go out into the field to dig them up but it is one step closer to having a quicker way to identify possible locations rich in fossils.

This new system is based upon images from NASA’s series of Landsat missions. These satellites have been observing the different environments of the Earth since the 1970s, and even though satellites cannot detect fossils, they are still able to distinguish between different types of rock and pinpoint where fossils might be. This is because the special equipment onboard these satellites can detect wavelengths of electromagnetic radiation that humans cannot detect.

As an example, the Landsat 7 satellite’s Enhanced Thematic Mapper Plus (ETM+) detects infrared radiation in addition to blue, green, and red wavelengths. This allows scientists to use its images to identify different types of land cover because rock layers are made up of minerals that reflect different wavelengths. These differences occur in both visible and infrared wavelengths, and analysts can then identify different types of land cover.

Combinations of different wavelengths of light—including invisible wavelengths—reveal different properties of the Earth’s surface. (NASA image by Robert Simmon, using Landsat data.)
Landsat 7 satellite’s Enhanced Thematic Mapper Plus (ETM+) detects infrared radiation in addition to blue, green, and red wavelengths. This allows scientists to use its images to identify different types of land cover because rock layers are made up of minerals that reflect different wavelengths. (NASA image by Robert Simmon, using Landsat data.)

Using technology to find fossils is not new. In the mid-1980s, Richard Stucky of the Denver Museum of Nature and Science started working with NASA to take advantage of Landsat satellite technology for finding fossils. He was able to produce false terrain images intended to distinguish between different types of rock. By identifying rock layer inconsistencies through satellite images and then confirming them in field, Stucky believed he could create a computer program to help paleontologists.

The development of technology since then could see his dream come true through the work of Anemone and Emerson. The two have painstakingly developed a computer program that analyzes hundreds of images from satellites using neural networks that function like nerve cells in the human brain. These neural networks are made up of input nodes, hidden nodes, and output nodes. The input nodes are made up of the spectral bands from the Landsat ETM+, the hidden notes manage the inner workings, and the output nodes are the different types of land cover, including known fossil sites.

Anemone and Emerson did not stop there, however. They wanted their program to analyze satellite images but also learn how to distinguish between sites that could hold fossils and those that would not. To do this, they had to train their network to recognize mistakes, a process called back-propagation. Trainers give the network examples of what it should produce and any discrepancies are considered errors. When the errors travel backward through the network, the computer learns how to calculate the correct results.


Advertisement


The two scientists used satellite images of the Great Basin for their lesson plan. Just like any digital photo, satellite images are made up of pixels, and as Anemone and Emerson analyzed these images, they tagged pixels that corresponded to known fossil sites. The network was fed pixels that were both fossil-rich and fossil-poor in order to train the computer to differentiate between land cover types including forests, wetlands, scrubland, and areas that could hold fossils. When the duo fed the computer images of the Great Basin, it accurately detected 80 percent of the pixels that held fossils.

Using Landsat images analyzed by a neural network, Anemone and his colleagues created a land cover map of the Great Divide Basin. Potential fossil locations are light red, and likely locations are dark red. (Map adapted from Anemone, et al., 2011.)
Land cover map of the Great Divide Basin from the results of using a neural network to analyze Landsat images. Potential fossil locations are light red, and likely locations are dark red. (Anemone, et al., 2011.)

This new method does not guarantee results, though. A trip back to Wyoming back in 2012 did not yield the results Anemone and Emerson expected. Still, this new analysis method holds some promise for more efficient methods of fossil detection as opposed to relying solely on luck. Even if it does not guarantee positive outcomes, using remote sensing for finding fossils can help stack the deck in the paleontologists’ favor.

References

Anemone, R., Emerson, C. and Conroy, G. (2011), Finding fossils in new ways: An artificial neural network approach to predicting the location of productive fossil localities. Evol. Anthropol., 20: 169–180. doi: 10.1002/evan.20324

“Hunting Fossils from Afar.” http://earthobservatory.nasa.gov/Features/WyomingFossils/




Previous

Congressman Calls for a Subcommittee with Jurisdiction over Federal Geospatial Activities

GIS in Urban Planning

Next