Remote Sensing Final Project: Multispectral Image Analysis in Archaeological Applications.

 

 

 

 

Remote Sensing Final Project

Multispectral Image Analysis in Archaeological Applications.

Tristan Haroldsen

University of West Florida

Photo Interpretation and Remote Sensing

Brian Fulfrost

12/4/2024


Introduction:

The consensus of Archaeologists is that we have discovered and documented less than three percent of available Archaeological sites worldwide. This is due in part to the remote nature of many sites, the ever-changing landscape as well as numerous other factors. To cut through these factors Archaeologists are implementing new and improving technologies in the field of remote sensing to rediscover our collective past.

While the utilization of remote sensing technologies in the field of archaeology is commonplace the implementation of aerial photogrammetry as well as other remote sensing procedures is often made difficult due to thick forest or vegetation that blankets the terrain. In such cases, technologies such as LiDAR may be implemented to penetrate the thick forest canopy allowing researchers to render detailed depictions of the forest floor once obscured by the canopy.

This, however, is not without its challenges and limitations. Due to the limited number of laser points which ultimately penetrate the forest canopy many archaeological sites may be inaccurately classified simply as bare ground rather than identified as an archaeological feature due to the region’s topography and other mitigating circumstances.

In such instances I propose that it may be possible to identify features or patterns of the vegetation itself to discern the location of potential archaeological sites. For example, could multispectral imaging be used to identify patterns in the vegetation such as its health, density, and or species diversity to locate archaeological sites hidden beneath?

  

Case Studies:

An analysis of archaeological literature does indicate that these techniques have been used successfully in similar contexts. For example, in a study of Crop Mark Analysis in Northwestern Iberia researchers utilized “multispectral imaging via Unmanned Aerial Systems (UAS) and digital image processing to identify archaeological sites obscured by complex vegetation in the Iberian Peninsula. Techniques such as NDVI and Principal Component Analysis (PCA) helped distinguish crop marks associated with ancient settlements from natural vegetation patterns.” (Peña-Villasenín et al., 2024)

 

Background:

However, in order to further test this hypothesis, it is first necessary to identify a known archaeological site obscured by thick vegetation. Coba, a documented Mayan site located in the Yucatan peninsula was an important cultural site to the Mayan people of the region. The site itself consists of numerous causeways spanning out from a central megalithic site consisting of several identified structures as seen below.

Coba Central Megalithic Site (Weyleb)

 

It is important to note that while this central location and the outstretching causeways are well documented and mapped the location and boundaries of the outlying suburban groups have yet to be determined as they are hidden under dense vegetation and have not yet been identified. The locations of these suburban groups have however been estimated by archaeologists based on proximity to the causeways and are depicted in pink on the map below providing areas of interest to focus on with our multispectral analysis.

Coba Basemap 2024


Methods:

By removing the estimated locations of the suburban groups from the map leaving only the documented causeways we can run our multispectral satellite images through multiple processes to decern the location of potential anomalies. Anomalies such as an increase or decrease in plant health or any identifiable growth patterns may then be used to ascertain the location of structures under the forest canopy paying special attention to any anomalies located at the ends and / or intersections of the known causeways where these suburban group sites are estimated to exist.

Note: Remote sensing imagery sources.

Satellite Imagery: GloVis Landsat 4-5 Thematic Mapper / Multispectral Scanner Collection 2 Level 1 which features 7 bands sourced with a opto-mechanical sensor with a spatial resolution of 30 m and a spectral range of 0.45 – 12.5 µm.

Coba Urban Groups and Causeways KML: “Among the sources used to create this atlas are various GPS measurements from Karl Herbert Mayer, the Atlas Arqueológico de Guatemala, 2008, Guatemala City: Ministerio de Cultura y Deportes; and the information from Oscar Quintana & Wolfgang Wurster (2001): Ciudades Mayas del noreste del Petén, Guatemala, Mainz: Phillip von Zabern.” (Weylab GIS Atlas, 2017)

 

Filters and Band indices (like NDVI)

               Using a multispectral image composite of the location and implementing a variety of filters and band indices such as False Color Infrared and NDVI or Normalized Difference Vegetation Index which are often utilized in forestry and agricultural applications to assess crop health as well as vegetation density of an area we can ascertain if any anomalies are present in the image.

               False Color IR which readily detects vegetation by visualizing colors outside of what the human eye can detect and displaying it in shades of Red with dense or healthy vegetation appearing darker and sparse of unhealthy vegetation appearing lighter allows for a clear view of the vegetative health of the site as seen below. Using this filter, it is easy to visualize the separation of healthy forest and several deforested or sparkly forested areas surrounding the site.

Upon careful inspection it is evident that several anomalies can be seen within the image. Several square shaped growth patterns can be seen within the central intersections of the Nohoch Mul causeway system (blue) as well as the Coba system depicted (red) which may indicate potential structures at these sites.

 

Coba TM False Color IR Map 2024

In order to collaborate these findings NDVI or the Normalized Difference Vegetation Index which uses a multispectral image and calculates the relation between the Red and Infrared Bands to assess the photosynthetic capacity of vegetation can be used which allows researchers to assess biomass, growth density and plant health allowing by displaying vegetation on an index of 1 to -1 with bare ground displayed as 0 and water displayed as -1 allowing for slight variations in photosynthetic properties to be identified. In the map below healthy or dense vegetation is displayed as bright white while unhealthy vegetation will be displayed as darkening shades of grey.

In this image we again can see that several anomalies are present in the form of healthy vegetative growth appearing in square patterns within the central intersections of the Nohoch Mul causeway system depicted in on the map below in light blue which may indicate archaeological activity at the location while also confirming the anomaly found within the Near Infrared Image.

 

Cobe NDVI Map 2024

 

Unsupervised Classifications.

                                         By utilizing the location of the anomaly identified and corroborated with both the IR and the NDVI indices as a reference point it is then possible to isolate and classify the location based on its specific spectral signature to identify any features sharing that identical signature which may be difficult to discern using IR and NDVI. By performing an Unsupervised Classification of the image, each pixel displayed on the image is broken down into unique classifications allowing us to then isolate the area of interest and display it in red alongside all other locations emitting the same frequency ultimately providing us with the locations of all sites on the map displaying those unique qualities.

                                         On the map below these locations are depicted in red and by doing so it becomes obvious that many of these features do indeed align themselves with the various causeway systems as well as fall within the estimated locations of the suburban groups (pink) identified by archaeologists on the ground.

Coba Unsupervised Classification Map 2024


Ground Truthing

While ground truthing these locations requires boots on the ground in the form of archaeological experts it is possible to implement high resolution aerial imagery sourced from the Esri imagery basemap to get a closer look at the location perhaps providing detail which may be unavailable with satellite imagery. Using these high resolution arial images of the location it is evident that an obvious archaeological anomaly is indeed located at the center of the Nohoche Mul Causeway System as seen below providing us with some level of verification to our findings.

 

Coba Aerial Image Map 2024


Results and Conclusion

               In conclusion it appears evident that through the usage of multispectral image processing archaeological sites may be able to be identified based solely on vegetation density and plant health. In the case of Coba, it appears that when combined with the known locations of the Mayan causeways and by ground proofing against known sites these processes provide reasonable evidence of structures hidden underneath vegetative cover within out designated areas of interest.

While it is impossible to definitively identify archaeological locations without traditional archaeological excavations, multispectral image analysis can be a useful tool to visualize the probable locations of sites and should be considered in specific instances such as those detailed in this journal.

                

References

1)     Peña-Villasenín, S., Gil-Docampo, M., & Ortiz-Sanz, J. (2024). Hidden archaeological remains in heterogeneous vegetation: A crop marks study in fortified settlements of northwestern Iberian Peninsula. Remote Sensing, 16(21), 3923. https://doi.org/10.3390/rs16213923

2)     Atlas of Maya Archaeological Sites. Wayeb. (2017, December 6). https://www.wayeb.org/resources-links/wayeb-resources/wayeb-gis-atlas/#1511961556518-9b8a4bab-1001


Comments

Popular posts from this blog

About Me Blog Post

Module 1: Map Design & Typography

Choropleth Mapping Lab