Computational archaeology
File
Vienna_Manifesto_on_Digital_Humanism_Farsi.pdf · 593 KB · click to show
Vienna_Manifesto_on_Digital_Humanism_Farsi.pdf · 593 KB · click to show
Photo
Computational archaeology
Photo
Preservation bias
Preservation bias represents a foundational vulnerability in computational archaeology because algorithms inherently treat the absence of data as evidence of absence in the past. When machine learning models or predictive mapping tools ingest archaeological databases, they process skewed realities; durable materials like stone, pottery, and bone survive to become digital data points, while organic components like wood, textiles, and hides rot away entirely. Because code cannot spontaneously account for what has vanished, computational models map patterns, trade routes, and demographic densities based solely on these durable remnants. This creates a dangerous feedback loop where automated tools mistake the geological durability of an object for its cultural significance, ultimately generating highly precise but fundamentally flawed reconstructions that structurally erase organic-reliant societies from the historical record.
60 · Computational archaeology
Photo
Imagery and the electromagnetic spectrum
Electromagnetic radiation is energy that moves through space at the speed of light as a combination of electric and magnetic fields. All objects reflect, absorb, and emit electromagnetic radiation. Using technology, we can detect this energy and capture it as data. The resulting visualization of this data is called imagery.
Because of differences in how objects on Earth's surface interact with electromagnetic radiation, imagery enables the identification of features and reveals information about surface conditions. Understanding electromagnetic radiation and how it is detected using remote sensing is foundational for effectively mapping, analyzing, and interpreting imagery.
36 · Photo
Multispectral imagery (MSI)
is the result of processing and visualizing electromagnetic radiation from within or across one or more specific ranges of targeted wavelengths—or discrete bands—of the EMS.
The bands can be visualized individually; they also can be combined using a three-color band combination, in which any band can be assigned to the red, green, and blue display channels. For example, a natural color imagery visualization combines the red, green, and blue imagery bands into the red, green, and blue display channels, respectively.
However, you can choose other band combinations to perform different types of visual analysis. The image will appear brighter or darker corresponding with higher and lower levels of reflectance within a specific region of the EMS.
37 · Computational archaeology
Photo
Hyperspectral imagery (HSI) is a type of remote sensing technology that captures and analyzes information from across the EMS with more spectral granularity than multispectral imagery.
HSI data consists of numerous, narrowly defined, continuous bands. Hundreds, or even thousands, of spectral bands may be available within a single image.
Like MSI, HSI can be displayed by assigning any three imagery bands to the three display channels to visualize the data in a three-color band combination.
32 · Computational archaeology
Photo
A spectral profile
is a visual representation of the spectral signature of objects visible within imagery. Every physical object—as a result of differences in the way it can reflect, absorb, and emit energy—has a unique spectral signature.
In this example of a spectral profile, the x-axis represents the wavelength range of the EMS in nanometers (nm), and the y-axis is the reflected energy values. Reflectance values for specific features in the imagery, such as water, are plotted as curves along the range of the EMS. The unique reflectance patterns help you identify and understand how various features absorb and reflect energy in different regions of the EMS.
33 · Link
click to show
click to show
A new study revisits one of the central questions in Neanderthal archaeology:
Did climate change fragment Neanderthal habitats before their disappearance?
Yousefi et al. reconstruct Neanderthal habitat suitability and landscape connectivity across Eurasia, from the Last Interglacial around 130,000 years ago to the period preceding their disappearance around 40,000 years ago.
Their results challenge a simple extinction narrative. Although suitable habitat declined modestly toward MIS 3, the authors do not find a corresponding increase in habitat fragmentation or a major collapse in connectivity. Neanderthal landscapes apparently remained relatively extensive and connected.
Archaeologically, this weakens the idea that Neanderthals were simply pushed into isolated climatic refuges.
But it does not prove that climate was irrelevant. Nor does it demonstrate that Homo sapiens alone caused the extinction. The model does not directly reconstruct prey availability, demographic decline, competition, or social isolation.
The most important contribution of the study may therefore be conceptual: it narrows one explanation without replacing the extinction debate with a single alternative.
Climate, ecology, demography, prey, and interaction between human populations must still be examined together.
Paper: https://doi.org/10.1038/s43247-026-03960-8
#Neanderthals #Palaeoanthropology #Archaeology #HumanEvolution
26 · Link
click to show
click to show
From a computational-archaeology perspective, this paper is especially valuable because it makes a large-scale archaeological argument through a chain of spatial models.
The workflow combines Neanderthal occurrence data with six environmental predictors, five modelling algorithms (GLM, GAM, random forest, GBM, and MaxNet) and an ensemble species-distribution model. The resulting suitability surfaces are then used in circuit-theory connectivity analysis to estimate potential movement pathways. This is a strong framework for testing whether climatic suitability became more fragmented through time.
However, the supplementary code also shows why model outputs must remain open to scrutiny. As supplied, the script creates a five-fold cross validation object with five repetitions, but the modelling call appears to use two repetitions and a 70/30 data split. The supplied loops also execute the LIG stage unless adapted. Suitable area is calculated by summing binary raster cells, and the connectivity model uses specific choices :: four-neighbour movement, raster aggregation, a fixed parameter, and an implicit suitability-to-conductance relationship.
These are not minor technical details. They can influence the resulting corridors and landscape metrics. The lesson is not that the model is invalid, it is that a corridor is a model-dependent hypothesis, not an observed Neanderthal route. Future work can add spatial-block validation, resistance-surface sensitivity tests, uncertainty propagation, independent archaeological validation, and explicit modelling of prey, water, seasonality, and demography.
Supplementary Data: https://media.springernature.com/original/springer-static/esm/art%3A10.1038%2Fs43247-026-03960-8/MediaObjects/43247_2026_3960_MOESM2_ESM.pdf
#ComputationalArchaeology #GIS #SpeciesDistributionModels #ReproducibleResearch
34 · Computational archaeology
Photo
Synthetic aperture radar (SAR) is a type of imagery collected by a sensor using the microwave and radio frequencies. It is a type of radar (a term derived from "radio detection and ranging"), a technology that is used to measure the distance, direction, and velocity of objects.
SAR's longer wavelengths can penetrate clouds, smoke, and vegetation, making it a useful tool for planetary observation where optical sensors—such as multispectral or hyperspectral instruments—cannot otherwise see. SAR is typically represented using frequency in gigahertz (GHz) or wavelength in meters (m). For example, SAR imagery that is captured by the sensors on Sentinel-1 satellites is within the C-band of the EMS with a center frequency of 5.405 GHz (or at a wavelength of 0.05546576 m).
40 · Computational archaeology
Photo
The Landsat satellites
are a series of spacecraft orbiting Earth that have collected multispectral imagery of the planet's surface continually since 1972. The Landsat missions are managed jointly by the National Aeronautics and Space Administration (NASA) and the United States Geological Survey (USGS), and they have returned millions of images since the program's inception.
The Landsat satellites have evolved over the years:
Sensors on the early Landsat missions collected only visible and near-infrared imagery.
Landsat 4 (launched in 1982) introduced shortwave infrared and thermal bands, and it marked the beginning of mission-to-mission data continuity.
Landsat 7 (launched in 1999) became the program's first satellite with a panchromatic band.
Landsat 8 (launched in 2013) and Landsat 9 (launched in 2021) collect data across 11 bands using the Operational Land Imager (OLI) and Thermal Infrared Sensor (TIRS).
Landsat 10 (a planned future Landsat mission) is designed to collect even more data than Landsat 9
41 · Computational archaeology
Photo
GIS ... wtf is a Raster Function?
A raster function is an operation that processes one or more input layers—raster, imagery, or feature data, depending on the function—using algorithms that are customized for a specific purpose. The operation uses on-the-fly processing to quickly and efficiently perform analysis directly to pixels of imagery and raster datasets, generating a raster layer (called a function raster layer) with the results of the analysis. Raster functions can detect change over time, extract features like trees, and classify objects in images, among other uses.
36 · Computational archaeology
Photo
A raster function template enables you to build and customize a raster analysis workflow by chaining together the results from multiple raster functions. Raster function templates are helpful when one raster function requires the output from another raster function as input. Chaining together results is also more efficient than running each raster function individually.
30 · Photo
A geoprocessing tool is an operation that performs analysis on one or more layers and then generates the output as a file on disk (feature class, raster dataset, imagery, or table). Depending on the type of analysis that you are performing, the tool can accept imagery, features, or both as input. Geoprocessing tools that are used with raster data in ArcGIS Online are called raster tools.
Similar to how you can chain raster functions, you can combine multiple geoprocessing tools into a processing chain with ModelBuilder, a visual programming language for automating workflows.
33 · Photo
Hyperspectral analysis
In imagery and remote sensing, spectroscopy is the measurement of how features and materials on the Earth's surface interact with, absorb, or emit electromagnetic energy. Hyperspectral analysis is a type of spectroscopy that uses hyperspectral imagery (HSI)—data with hundreds, or even thousands, of wavelength bands—to identify the composition of soils, vegetation, and human-made objects on Earth. The numerous individual wavelength bands in HSI data make it possible for the unique spectral signatures of objects—such as trees, or the paint on a car—to appear when plotted on a spectral profile chart, like in the following example.
36 · Computational archaeology
Photo
Deep learning analyses in ArcGIS follow a generalized workflow. Regardless of the deep learning task—whether you are extracting features, classifying pixels, detecting change, or detecting objects—the workflow for applying deep learning has five steps:
Prepare training data
Train the model
Run the model
Evaluate the results
Postprocess (if necessary)
These steps are iterative and may be repeated as necessary. After a model is adequately trained, it can be used on large datasets and then used again and again, saving valuable time and resources overall.
27 · Computational archaeology
Photo
Prepare training data
To prepare training data, you manually label objects in your data. These labeled objects are then converted into a format that can be used by the deep learning neural network. Preparing data can be time-consuming, and it requires a significant amount of user-created data. However, a larger and more accurate input dataset results in a more successful deep learning model.
25 · Computational archaeology
Photo
Train the model
When you train the model, the model's neural network is learning how to accurately label an image according to the sample training data for a specific task. To avoid requiring user input, the model also learns to identify patterns while analyzing the data.
Training a model can be time-consuming and may require a lot of processing power. If you do not have the time or processing power to train a model, you can use a pretrained deep learning model instead.
27 · Computational archaeology
Photo
Run the model
Whether you run a model that was trained by you or someone else, it is important to know how the model was trained so that it can be applied appropriately. Ensure that you review the training metrics that outline how the model was trained and its expected accuracy.
With a pretrained deep learning model, the first two steps of the deep learning workflow were already completed using a sample dataset; the model includes output from the first two steps of the workflow, enabling you to use the output with your own data. Before you run a pretrained deep learning model, you must verify that the specific deep learning task, geographic region, and data type on which it was trained are appropriate for your use case.
31 · Computational archaeology
Photo
Postprocess (if necessary)
The output of a deep learning model may be an interim dataset that needs further action or postprocessing to achieve the desired result. For example, you might need to regularize polygons from an instance segmentation output, or compute the accuracy of object detection.
33 · Photo
Evaluate the results
After running a deep learning model, it is important to evaluate the results to verify that the model worked as expected and to determine whether further action is required. You may want to repeat the first two steps of the workflow to improve model accuracy or modify parameters during model inferencing. You might also decide that the results need postprocessing.
35 · Computational archaeology
شمارۀ جدید مجلۀ باستانشناسی که توسط پژوهشکدۀ باستانشناسی پژوهشگاه میراث فرهنگی و گردشگری به صورت فصلنامه به انتشار میرسد، منتشر شد. شمارۀ نخست از دورۀ پنجم این نشریۀ تخصصی شامل 6 مقاله علمی با عناوین زیر میشود، که در وبسایت مجله به آدرس archj.richt.ir در دسترس است.
- «باستانشناسی منطقه لارستان در جنوب فارس»، نوشته علیرضا سرداری
- «نگاهی به روند نوسنگیشدن در زاگرس مرکزی در پرتو کاوشهای اخیر در چغاگُلان»، نوشته حجت دارابی
- «زیرساختهای هیدرولیکی در پارس دوره هخامنشی بر پایه پژوهشهای جدید»، نوشتۀ کورش محمدخانی، سباستین گندت و ماری- لور شامبراد
- «زیرساختهای آبی و الگوهای بهرهبرداری از آب در محوطه دهقاید»، نوشتۀ زهره زهبری، مهرداد پارسایی برازجانی، نصرالله ابراهیمی، نغمه رفیعی و سامران آسیابانی
- «پژوهشی در سلادنهای مکشوفه از بندر تاریخی بَردستان (بُردو)»، نوشتۀ حسین توفیقیان
- «محوطه چخا: احیای مجدد و شکوفایی اقتصادی، اجتماعی و بازرگانی هرمز قدیم در دورة صفوی»، نوشتۀ سیامک سرلک، شیرین عقیلی نیاکی، عباس نوروزی و ثریا افشاری
از پژوهشگران گرامی دعوت میشود مقالات علمی خود را جهت انتشار در مجلۀ باستانشناسی براساس شیوهنامه، از طریق سامانۀ نشریه ارسال نمایند.
25 · File
richt-jour-v5n1p77-fa.pdf · 3.2 MB · click to show
30 · richt-jour-v5n1p77-fa.pdf · 3.2 MB · click to show
Computational archaeology
Photo
Oriented imagery
ArcGIS also supports imagery collected in other ways, such as horizontal, spherical, and close-range inspection imagery. When images are associated with metadata that can relate pixels in the raw source image to the ground, it is called oriented imagery. Oriented imagery is supported by a special data model—the oriented imagery dataset—and tools that enable you to visualize and explore the unprocessed source imagery from a first-person perspective in the context of a map or scene. Nadir and low-oblique images are also supported by the oriented imagery data model and tools when they include the necessary metadata.
Sensors on satellites, aircraft, and drones capture imagery that is often managed using a mosaic dataset. Oriented imagery is captured when sensors collect images with metadata that relates pixels to the ground, such as sensors aboard aircraft, drones, and automobiles, as well as handheld cameras. Sensors that are used to collect oriented imagery might also be found mounted.
26 ·