NL ENG
20231005_100507

Multispectral Imaging

Multispectral imaging is a well-established tool in the field of archaeological remote sensing. Within the AMZ cycle, this technique is particularly useful in the exploratory and mapping phases of a prospection survey. Under favourable conditions, multispectral imaging can reveal valuable archaeological information in terrestrial contexts.

What?

How does it work?

Multispectral sensors are advanced instruments capable of passively capturing visible light and parts of the invisible electromagnetic spectrum in separate bands across different sensors, resulting in various reflection images. These sensors provide a unique ability to analyze the reflective properties of various materials. Multispectral imaging is generally effective in detecting crop marks (figure 1). These are most clearly visible during dry conditions, when soil moisture is low, and crops or vegetation are under stress. This makes differences in growth or color more apparent, and the contrast can be further enhanced using vegetation indices.

Figure 1: Oblique photo of crop marks near Diepenveen, Overijssel, visible after a period of prolonged drought in the summer of 2022 (source: NAR80, © A. Speelman).
Want to know more?

In-depth explanation X

Multispectral imaging provide several reflection images, most commonly in the colors Blue (center wavelength: 475 nm), Green (center wavelength: 560 nm), Red (center wavelength: 668 nm), Rededge (center wavelength: 772 nm), and Near-infrared (center wavelength: 840 nm) (figure 2). Different combinations and slightly varying wavelengths are possible.

Figure 2: An overview of wavelengths in the electromagnetic spectrum. The range measured by multispectral imaging is highlighted with diagonal stripes (adapted from Wikimedia Commons).

The way in which different materials absorb or reflect radiation at various wavelengths allows for valuable insights into their physical composition. This enables observations beyond the human visible spectrum; for example, crop marks can be significantly enhanced because healthier vegetation reflects more near-infrared radiation while absorbing more visible light. By applying photogrammetric techniques similar to those used with optical data, mosaic reflection maps can be created. These maps, often referred to as spectral or, more commonly, vegetation indices (VIs), are mathematical functions that use the reflection values of two or more spectral bands to derive information about the physical properties of materials in an image. The various reflection maps can be part of many comparisons that highlight different aspects of vegetation, such as the Normalized Difference Vegetation Index (NDVI), which is frequently used for agricultural purposes.

In essence, multispectral imaging is a form of spectroscopy: a collective term for techniques that study materials based on their interaction with the electromagnetic spectrum.

What do you need?

To capture images from different spectral bands, specific cameras are used with multiple sensors (typically <10), each filtering incoming radiation and recording only the desired spectral band. These multispectral cameras are sometimes integrated into UAS (drone) systems, allowing for a direct live feed to be displayed on the screen. There are also solutions available where the integration with the drone’s hardware and software is less extensive. While multispectral images can be captured as individual reflection indices, for archaeological purposes, it is more useful to create image sequences. Using photogrammetric techniques, also applied to optical data, mosaic reflection maps can be generated.

Multispectral handheld cameras can be used to study the physical and chemical properties of objects, for example.

Multispectral images are also collected by satellites. These images typically have a coarser resolution compared to those taken with drones, but satellite images allow for the study of large areas, as a single capture can cover a vast surface. Due to the revisit frequency of various satellites, satellite data can also be used for long-term monitoring of objects. In the Netherlands, this data is made available by the Netherlands Space Office through the ‘satellietdataportaal‘.

Want to know more?

In-depth explanation X

Multispectral cameras produce multiple images for each capture, depending on the specific spectral bands they measure. In photogrammetric software, images can be processed using the green spectral band to generate point clouds, 3D models, and orthophotos, which then serve as the basis for projecting information from the other spectral bands. Well-known photogrammetric software packages like Agisoft Metashape or Pix4D can work with this data. These packages allow for the direct creation of various vegetation indices, but this can also be done in GIS software such as QGIS. To work with multispectral satellite data, in addition to GIS software, Google Earth Engine can be used.

Since comparing reflection across different images depends on the total amount of electromagnetic radiation present—such as due to variable cloud cover—it is important to calibrate the images. For drone captures, this can be done using a calibration panel before and after each capture, along with a downwelling light sensor (DLS) mounted on the drone to continuously measure the total amount of electromagnetic radiation.

Can be used with..

By using a handheld camera or a camera mounted on a drone, differences in reflection across these spectral bands can be effectively captured at short distances (Figure 3). With the rise of UAS technology, multispectral imaging for archaeological prospecting has made significant advances, as very high-resolution images can now be collected at precisely determined (and potentially optimal) times of the year. However, there are limitations in geographical coverage due to legislation and the limited battery life of drones.

Figure 3: Preparing for a multispectral drone survey (source: J. Waagen, private collection).
Want to know more?

In-depth explanation X

Multispectral images can also be collected with cameras mounted on satellites and aircraft. However, these images often have a lower resolution, and atmospheric absorption of near-infrared radiation becomes more significant when the sensor is farther from the Earth’s surface. Additionally, the exact timing of the capture is crucial in multispectral imaging, so without control over this, it becomes challenging to accurately interpret the images.

Archaeological Applications

Place in the Dutch archaeological heritage management process

Multispectral imaging, like most other archaeological remote sensing methods, can be applied in the exploratory and mapping phases of an archaeological survey. The collected and analyzed images result in a set of spectral anomalies (Figure 4), complete with descriptions and interpretations. These can be used to gain a deeper understanding of both subsurface and surface archaeological features and may help inform further prospecting or excavation research.

Figure 4: Multispectral image taken in Weesp in the summer of 2023. This NDVI representation shows various spectral anomalies (source: 4D Research Lab).
Want to know more?

In-depth explanation X

Due to current legal (maximum distance from the pilot, limited visibility after sunset) and technical (e.g., battery life) constraints, using drones equipped with multispectral cameras over large areas is often not feasible. For this reason, such research typically takes place in areas where there is at least a strong suspicion of archaeological remains. Because of the limited range, a handheld camera will primarily be used at known archaeological sites and will be linked to very specific research questions (e.g., physical or chemical composition of an object).

What types of archaeological materials/landscapes

Multispectral imaging can be applied in any situation where differences in spectral reflection between archaeological features and the natural landscape are likely to be detectable. Whether these differences can be observed depends heavily on local conditions. Typical scenarios include fields with crops in late summer, where differences in crop health are often at their peak, making them most visible as crop marks.

Want to know more?

In-depth explanation X

It is important to recognize that the ability to detect archaeology in the form of crop marks is highly dependent on the condition of the vegetation, which means it is influenced by the season and the specific growth cycle of the crop. Weather conditions also play a significant role. In certain years, such as those with prolonged drought, crop marks may become more visible and noticeable more quickly.

Limitations/uncertainties

The capture of multispectral images is highly seasonal and should take place around (solar) noon, which limits its use. Additionally, the photogrammetric postprocessing is a relatively complex process, susceptible to many variables that influence the final output.

Want to know more?

In-depth explanation X

Due to the variety of landscapes and climates worldwide, including in the Netherlands, the relative effectiveness of multispectral imaging is a subject of ongoing research. An important aspect of multispectral research is capturing the situational parameters at the time of drone surveys and using complementary and potentially validating prospecting techniques. The potential of multispectral imaging to detect archaeological features is also dependent on vegetation types and growth cycles. This makes it essential to compare workflows and outputs with data obtained from other sources, such as optical sensors.

Casestudies

Casestudies

Curious about how multispectral imaging has been successfully used in archaeological fieldwork? Click on the tiles below to explore the case studies where this innovative sensor technique has been applied.

References/further reading

Kalayci, T., Lasaponara, R., Wainwright, J., & Masini, N. (2019). Multispectral Contrast of Archaeological Features: A Quantitative Evaluation. Remote Sensing, 11(8), 913. https://doi.org/10.3390/rs11080913

Agapiou, A., Hegyi, A., Gogâltan, F., Stavilă, A., Sava, V., Sarris, A., Floca, C., & Dorogostaisky, L. (2023). Exploring the largest known Bronze Age earthworks in Europe through medium resolution multispectral satellite images. International Journal of Applied Earth Observation and Geoinformation, 118. https://doi.org/10.1016/j.jag.2023.103239

Casana, J. (2023). Archaeological Remote Sensing. In Handbook of Archaeological Sciences (eds A.M. Pollard, R.A. Armitage and C.A. Makarewicz). https://doi.org/10.1002/9781119592112.ch50

Rensink, E., Theunissen, L., Feiken, R., Bourgeois, J., Deforce, K., van Doesburg, J., Emaus, R., van der Heiden, M., de Jong-Lambregts, N., Karagiannis, N., de Kort, J. W., Liagre, E., van Londen, H., Meylemans, E., Orbons, J., Stichelbaut, B., Terlouw, B., Timmermans, G., Waagen, J., & van Zijverden, W. (2022). Vanuit de lucht zie je meer. Remote sensing in de Nederlandse archeologie. Nederlandse Archeologische Rapporten (NAR) 80.

Netherlands Space Office. Satellietdataportaal

Seyfried, Simon. (2020). Thermal and Multispectral Monitoring of cropmarks by UAV AARGnews, 61.

Waagen, J. (2023). In search of a castle: Multisensor UAS research at the Medieval site of ‘t Huijs ten Bosch, Weesp. 4D Research Lab report series, 4. https://doi.org/10.21942/uva.23375486.v2