Archaeological Predictive Modelling Research Articles

Automated Detection of Hillforts in Remote Sensing Imagery With Deep Multimodal Segmentation

ABSTRACTRecent advancements in remote sensing and artificial intelligence can potentially revolutionize the automated detection of archaeological sites. However, the challenging task of interpreting remote sensing imagery combined with the intricate shapes of archaeological sites can hinder the performance of computer vision systems. This work presents a computer vision system trained for efficient hillfort detection in remote sensing imagery. Equipped with an adapted multimodal semantic segmentation model, the system integrates LiDAR‐derived LRM images and aerial orthoimages for feature fusion, generating a binary mask pinpointing detected hillforts. Post‐processing includes margin and area filters to remove edge inferences and smaller anomalies. The resulting inferences are subjected to hard positive and negative mining, where expert archaeologists classify them to populate the training data with new samples for retraining the segmentation model. As the computer vision system is far more likely to encounter background images during its search, the training data are intentionally biased towards negative examples. This approach aims to reduce the number of false positives, typically seen when applying machine learning solutions to remote sensing imagery. Northwest Iberia experiments witnessed a drastic reduction in false positives, from 5678 to 40 after a single hard positive and negative mining iteration, yielding a 99.3% reduction, with a resulting F1 score of 66%. In England experiments, the system achieved a 59% F1 score when fine‐tuned and deployed countrywide. Its scalability to diverse archaeological sites is demonstrated by successfully detecting hillforts and other types of enclosures despite their typical complex and varied shapes. Future work will explore archaeological predictive modelling to identify regions with higher archaeological potential to focus the search, addressing processing time challenges.

Predicting rock art sites in the Pajeú watershed, Brazil

This paper presents an Archaeological Predictive Model (APM) to predict rock art archaeological sites in the Pajeú Watershed, a semiarid region in Pernambuco, Brazil. The model uses Machine Learning (ML) algorithms and re-sampling techniques to account for the unbalanced data set of rock art sites and test different inductive methods for predicting site location. The results show a satisfactory statistical evaluation, with high true positive rates with all ML algorithms and resampling techniques used, indicating a high potential for predicting rock art site locations. The predictive maps generated from the model output, show that certain features, such as aspect, elevation and the distance to different lithologies, are particularly important. The overall model's performance could be corroborated with a test in another semi-arid region, next to the Pajeú watershed, where areas with high favorability of finding rock art sites are predicted near to already known archaeological sites.

Changing with the times: From agricultural potential to spatially explicit reconstructions of past land use

Past land-use reconstructions are a key tool for studying long-term human ecodynamics and addressing pressing questions about the origins and evolutionary dynamics of the Anthropocene. In particular, agricultural landcover reconstructions are vital for understanding long-term human-environment dynamics. Most past agricultural land-use models, however, rely heavily on modelling assumptions, make limited use of known archaeological site locations to constrain or inform their estimates and tend to be limited to general estimates of percentages of plant types within catchments around pollen trapping lakes. The lack of information outside catchment areas and low spatial resolution even within catchments constrain the utility of these models. To address this problem, we propose a new approach that combines archaeological predictive modelling with pollen-based agricultural landcover reconstructions to produce more accurate, spatially explicit past landcover estimates. Here, we present the results of a case study deploying the new approach to produce improved past landcover maps for a region in the Western Taurus Mountains, southwestern Turkey. The study area surrounds Sagalassos, an antique urban centre with a regional settlement history encompassing nine millennia. We produced five archaeological predictive models for the study region using the ‘Locally Adaptive Models of Archaeological Potential’ (LAMAP) method spanning the Hellenistic through Late Ottoman period. We then combined those predictive surfaces with ‘Regional Estimates of VEgetation Abundance from Large Sites’ (REVEALS) pollen landcover reconstructions for the same periods based on pollen from sediment cores extracted from three catchments within the study area. Lastly, we compared the resulting hybrid landcover models to the archaeological record using data not used to make the predictions. We found that the hybrid landcover model aligned very well with the known extents of agricultural land-use from the study area. These results indicate that the proposed approach is a viable way to combine pollen-based landcover models with archaeological data and produce more accurate, empirically-based landcover reconstructions reflecting real human activity in the past.

Mapping landscape in Longshan period’s hierarchical society (3000–2000BCE) of North Loess Plateau: from archaeological predictive model to GIS spatial analysis

On the North Loess Plateau of China, city civilization, social complexity, and stratification emerged during the Longshan period (3000-2000BCE). Based on Geographic Information System (GIS) analysis and archaeological predictive model, we conducted a comparative analysis of environmental characteristics between sites and non-sites, ordinary and walled city sites, as well as large and smaller city sites. Initially, we developed a Binary Logistic Regression (BLR) model to predict the locations of archaeological sites from this period. Our findings indicate a high predictive accuracy of the model, demonstrating a clear environmental preference by the people of the Longshan Period. The presence or absence of the site was found to be influenced by various factors, including temperature, elevation, river distance, and precipitation. Furthermore, we discovered that walled cities had higher environmental requirements compared to ordinary sites. Terrain and land use played a more significant role in shaping prehistoric cities than climate. Lastly, the landscape in the Shimao site, which served as a most crucial and largest settlement centers on the North Loess Plateau (NLP), resembled that of other minor walled cities. Due to its abundant grassland, Shimao relied more on animal husbandry rather than agriculture. The combination of agriculture and animal husbandry has promoted the urbanization processes.

Utilizing the MaxEnt machine learning model to forecast urban heritage sites in the desert regions of southwestern Algeria: A case study in the Saoura region

AbstractThe Saoura region, a renowned oasis in North Africa with heritage and archaeological significance of both national and universal importance, has witnessed a gradual deterioration over time. This research involves archaeological predictive modelling, aiming to create models capable of predicting the likelihood of discovering archaeological sites, cultural resources or evidence of past landscape use within a specific region. The study specifically focuses on predicting the locations of historical sites in the Sahara Desert, employing the maximum entropy (MaxEnt) model and six geo‐environmental criteria, including slope, elevation (digital elevation model [DEM]), distance from water, normalized difference vegetation index (NDVI), fertility and proximity to palm groves. The research is based on data from 58 historical sites and includes an assessment of the model's accuracy. The study highlights the remarkable significance of the fertility variable, which accounts for 94.1% of the predictive influence, making it the most crucial geo‐environmental factor in forecasting the location of historical sites in the Sahara. This underscores its pivotal role in shaping settlement patterns and subsistence strategies within the region, followed by the distance variable from the palm cove (3.2%) and the distance variable from the river (2.3%). The MaxEnt model proves to be suitable for predicting historical site positions, with an impressive average area under the ROC curve (AUC) score of 0.859, reflecting its effectiveness. Notably, areas with a high prediction probability are predominantly situated near the Saoura Valley. The study's findings hold the potential to assist planners in safeguarding archaeological sites by avoiding areas where historical sites are likely to be present.

Identification Liangan Site: Archeological Predictive Modeling

This research was conducted in Liangan Site, an archeological Hindu temple site situated on the eastern slope of Mt. Sundoro, Temanggung District. The discovery of Liangan Site was in the area of sand mining. This research’s interest is trying to disclose of the spatial component of the site that are important to the history, science and culture. The objectives of this research: Mapping the width of Liangan Site that still has the potential to contain of archaeological artifacts. The method that used for this research analyze the distribution of archaeological remains at the Liangan Site with the scoring method through field surveys. While cultural resource management is done by knowing the predictions of the length area of the site using Archaeological Predictive Modeling. Results from this research include the entire potential zone that contains archaeological remains at the site of 20,71 ha. The zone sensitivity archaeological remains are divided into zones of high sensitivity (3.007 ha), medium sensitivity zones (4.369 ha) and low sensitivity zones (13.34 ha).

Open-Access Archaeological Predictive Modeling Using Zonal Statistics: A Case Study from Zanzibar, Tanzania

This paper presents a case study using zonal statistical analysis for archaeological predictive modeling with open-access software and free geospatial datasets. The method is applied to the rural landscapes of Zanzibar, Tanzania on the Swahili Coast. This study used QGIS (version 3.28) to perform zonal statistical analyses of environmental datasets weighted by settlement classes digitized from a 1907 historical map, to create predictive models for settlement across the island. These models were compared against the locations of major precolonial archaeological sites on the island and site data from a random stratified archaeological survey in an environmentally diverse region of northern Zanzibar. The results show strong correspondences between larger permanent site locations and areas of high likelihood for site detection in the predictive model. Additionally, there were correspondences between areas of lower likelihood for site detection and smaller, ephemeral sites related to agricultural production in swidden field plots. These results attest to Swahili rural complexity and enable an understanding of the specific environmental affordances that structured settlement and land use over the last millennium, in ways that shaped colonial contact in rural areas and altered the sociopolitical development of Zanzibar and the East African coast. The methods described here may be applicable for researchers and heritage managers in Africa and the Global South, where funding for large-scale field projects, expensive satellite imagery, or software licensing is limited.

Archaeological Predictive Modeling Using Machine Learning and Statistical Methods for Japan and China

Archaeological predictive modeling (APM) is an essential method for quantitatively assessing the probability of archaeological sites present in a region. It is a necessary tool for archaeological research and cultural heritage management. In particular, the predictive modeling process could help us understand the relationship between past human civilizations and the natural environment; moreover, a better understanding of the mechanisms of the human–land relationship can provide new ideas for sustainable development. This study aims to investigate the impact of topographic and hydrological factors on archaeological sites in the Japanese archipelago and Shaanxi Province, China and proposes a hybrid integration approach for APM. This approach employed a conditional attention mechanism (AM) using deep learning and a frequency ratio (FR) model, in addition to a separate FR model and the widely-used machine learning MaxEnt method. The models’ outcomes were cross-checked using the four-fold cross-validation method, and the models’ performances were compared using the area under the receiver operating characteristic curve (AUC) and Kvamme’s Gain. The results showed that in both study areas, the AM_FR model exhibited the most satisfactory performances.

Sampling methods for archaeological predictive modeling: Spatial autocorrelation and model performance

With a case study of a direct detection model (DDM) designed for Fort Irwin, in southern California, we compare models that are identical except for the sampling methods that draw the data used to train and test them. We find that models trained and tested with a random cell sampling strategy perform better than do models that implement a kind of leave-one-out cross-validation (LOOCV) that focuses on discrete sites and non-site areas. We argue that this difference in measured predictive performance is due to spatial autocorrelation, and that it underscores the importance of clarity and specificity in describing the sampling methods used in archaeological predictive modeling and site detection. Different sampling methods are not necessarily superior or inferior, but they generate models that may be more or less appropriate for different tasks. Different sampling methods can also yield calculations of predictive ability that over- or understate a model’s performance at the task for which it was designed.

Предиктивное археологическое моделирование методом максимальной энтропии (на примере поселений позднеримского времени в бассейне Верхнего Дона)

The paper is devoted to an archaeological predictive modeling technique with GIS and machine learning based on maximal entropy algorithm (MaxEnt). Archaeological sites of the late roman time presented by settlements of Kashirka-Sedelok type (mid-3rd — early 4th century AD) in the Upper Don basin were the modeling objects. We found out that there are 169 known sites of such type in the study region, and collected all available information on their locations into a geodatabase including our own field survey data. The coordinates of these sites split into training and testing subsets served for developing a predictive model with MaxEnt. Environmental GIS-layers based on a digital elevation model and thematic maps were the model predictors. We made such predictors to reflect the paleogeography of the Upper Don basin in the study time period by choosing relatively stable landscape features of relief, bedrock, climate, hydrography and soils of the area. The trained model analyzed through more than 7 million cells of 90x90 m with overall area of 59 400 sq. km to identify potentially suitable areas for the Kashirka-Sedelok settlements and found 29 860 ha the most suitable areas (less than 1 % of the study area) as the result. The efficiency of the conducted modeling, evaluated on the testing sites, is high based on metrics of receiver operating characteristic (AUC = 0,915) and Kvamme’s Gain (KG = 0,97). The results of modeling serve as an agenda for the upcoming archaeological surveys in the region.

Using UAV Survey, High-Density LiDAR Data and Automated Relief Analysis for Habitation Practices Characterization during the Late Bronze Age in NE Romania

The characterization of prehistoric human behavior in terms of habitation practices using GIS cartography methods is an important aspect of any modern geoarchaeological approach. Furthermore, using unmanned aerial vehicle (UAV) surveys to identify archaeological sites with temporal resolution during the spring agro-technical works and automated mapping of the geomorphological features based on LiDAR-derived DEM can provide valuable information about the human–landscape relationships and lead to accurate archaeological and cartographic products. In this study, we applied a GIS-based landform classification method to relief characterization of 362 Late Bronze Age (LBA) settlements belonging to Noua Culture (NC) (cal. 1500/1450-1100 BCE) located in the Jijia catchment (NE Romania). For this purpose, we used an adapted version of Topographic Position Index (TPI) methodology, abbreviated DEV, which consists of: (1) application of standard deviation of TPI for the mean elevation (DEV) around each analyzed LBA site (1000 m buffer zone); (2) classification of the archaeological site’s location using six slope position classes (first method), or ten morphological classes by combining the parameters from two small-DEV and large-DEV neighborhood sizes (second method). The results indicate that the populations belonging to Noua Culture preferred to place their settlements on hilltops but close to the steep slope and on the small hills/local ridges in large valleys. From a geoarchaeological perspective, the outcomes indicate a close connection between occupied landform patterns and habitation practices during the Late Bronze Age and contribute to archaeological predictive modelling in the Jijia catchment (NE Romania).

A Maxent Predictive Model for Hunter-Gatherer Sites in the Southern Pampas, Argentina

The following paper presents the results of a Species Distribution Model (SDM) for grassland hunter-gatherer archaeology sites in the southern Pampas region of Argentina. The goal of this exploratory model is to provide a complementary survey model for the detection of archaeological sites in this region, which will also help characterize and discuss site locations and regional distribution patterns of hunter-gatherer occupations. Even in this largely homogenous and highly dynamic landscape, SDMs can help guide archaeological surveys by identifying some environmental variables affecting hunter-gatherer decisions, and can provide insights into mobility and archaeological settlement patterns. Among the available tools for SDM, Maximum Entropy Modeling (Maxent) is one of the most widely used approaches in archaeological predictive modelling. After controlling for bias and adjustment of several modifiable parameters, the Maxent software provided a potentially effective predictive model to direct future archaeological survey and heritage management projects. The results of this research suggest that watercourses and slope were the key environmental factors influencing the distribution of hunter-gatherer archaeological sites in the southern Pampas region.

An Explorative Application of Random Forest Algorithm for Archaeological Predictive Modeling. A Swiss Case Study

The present work proposes an innovative approach to surveys and demonstrates the effectiveness of bringing together traditional archaeological questions, such as the exploration and the analysis of settlement patterns, with the most innovative technologies related to Machine Learning. Namely, we applied Random Forest, an ensemble learning method based on decision trees, to perform archaeological predictive modeling (APM) for the Canton of Zurich, in Switzerland. This was done based on a dataset of known archaeological sites dating back to the Roman Age. The APM represents an automated decision-making and probabilistic reasoning tool that is relevant for archaeological risk assessment and cultural heritage management. Machine learning-based approaches can learn from data and make predictions, starting from the acquired knowledge, through the modeling of the hidden relationships between a set of observations, representing the dependent variable (i.e. the archeological sites), and the independent variables (i.e. the geo-environmental features prone to influence the site locations). The main objective of the present study is to assess the spatial probability of presence for Roman settlements within the study area. As results, we produced: 1) a probability map, expressing the likelihood of finding a Roman site at different locations; 2) the importance ranking of the geo-environmental features influencing the presence of the archeological sites. These outputs in our results are of paramount importance, not only in verifying the reliability of the data, but also in stimulating experts in different ways. Also, these results help evaluate the benefits and constraints of using such innovative techniques and, ultimately, help explore the performance of machine learning-based models in processing archaeological information.

Towards an Operative Predictive Model for the Songshan Area during the Yangshao Period

The literature in the field of archaeological predictive models has grown in the last years, looking for new factors the most effective methods to introduce. However, where predictive models are used for archaeological heritage management, they could benefit from using a more speedy and consequently useful methods including some well-consolidated factors studied in the literature. In this paper, an operative archaeological predictive model is developed, validated and discussed, in order to test its effectiveness. It is applied to Yangshao period (5000–3000 BC) in the Songshan area, where Chinese civilization emerged and developed, and uses 563 known settlement sites. The satisfactory results herein achieved clearly suggest that the model herein proposed can be reliably used to predict the geographical location of unknown settlements.

Object-Based Predictive Modeling (OBPM) for Archaeology: Finding Control Places in Mountainous Environments

This contribution examines the potential of object-based image analysis (OBIA) for archaeological predictive modeling starting from elevation data, by testing a ruleset for the location of “control places” on two test areas in the Alpine environment (northern Italy). The ruleset was developed on the western Asiago Plateau (Vicenza Province, Veneto) and subsequently re-applied (semi)automatically in the Isarco Valley (South Tirol). Firstly, we considered the physiographic, climatic, and morphological characteristics of the selected areas and we applied 3 DTM processing techniques: Slope, local dominance, and solar radiation. Subsequently, we employed an object-based approach to classification. Solar radiation, local dominance, and slope were visualized as a three-layer RGB image that was segmented with the multiresolution algorithm. The classification was implemented with a ruleset that selected only image–objects with high local dominance and solar radiation, but low slope, which were considered more suitable parameters for human occupation. The classification returned five areas on the Asiago Plateau that were remotely and ground controlled, confirming anthropic exploitation covering a time span from protohistory (2nd-1st millennium BC) to the First World War. Subsequently, the same model was applied to the Isarco Valley to verify the replicability of the method. The procedure resulted in 36 potential control places which find good correspondence with the archaeological sites discovered in the area. Previously unknown contexts were further controlled using very high-resolution (VHR) aerial images and digital terrain model (DTM) data, which often suggested a possible (pre-proto)historic human frequentation. The outcomes of the analysis proved the feasibility of the approach, which can be exported and applied to similar mountainous landscapes for site predictivity analysis.

Connecting the Dots: Integrating Cultural and Natural Resource Management in the United States

Landholding agencies in the United States are under increasing pressure to integrate cultural and natural resource management approaches at a landscape level and to do so earlier and more comprehensively in planning processes. How to integrate management practices is poorly understood, however. An impediment to integration is that the laws, methods, and tools used in cultural and natural resource management differ significantly. Natural resource management protects or rehabilitates habitats and ecosystems that support endangered species, while cultural resource management focuses on identification and protection of individual sites. Agencies need to shift the focus from managing sites to defining cultural landscape elements and their relationship to natural resource management units and concerns. We suggest that agencies use archaeological predictive modeling, resource classes, and paleoenvironmental and cultural historical information to geospatially define cultural landscapes, predict resource distributions and values, and identify opportunities and protocols for collectively managing cultural and natural resources. As the United States faces increasing deregulation and limited preservation funding, we believe an integrated approach will be critical in preserving and protecting both cultural and natural heritage.

Integrating Point Process Models, Evolutionary Ecology and Traditional Knowledge Improves Landscape Archaeology—A Case from Southwest Madagascar

Landscape archaeology has a long history of using predictive models to improve our knowledge of extant archaeological features around the world. Important advancements in spatial statistics, however, have been slow to enter archaeological predictive modeling. Point process models (PPMs), in particular, offer a powerful solution to explicitly model both first- and second-order properties of a point pattern. Here, we use PPMs to refine a recently developed remote sensing-based predictive algorithm applied to the archaeological record of Madagascar’s southwestern coast. This initial remote sensing model resulted in an 80% true positive rate, rapidly expanding our understanding of the archaeological record of this region. Despite the model’s success rate, it yielded a substantial number (~20%) of false positive results. In this paper, we develop a series of PPMs to improve the accuracy of this model in predicting the location of archaeological deposits in southwest Madagascar. We illustrate how PPMs, traditional ecological knowledge, remote sensing, and fieldwork can be used iteratively to improve the accuracy of predictive models and enhance interpretations of the archaeological record. We use an explicit behavioral ecology theoretical framework to formulate and test hypotheses utilizing spatial modeling methods. Our modeling process can be replicated by archaeologists around the world to assist in fieldwork logistics and planning.

A multidisciplinary study of an exceptional prehistoric waste dump in the mountainous inland of Calabria (Italy): Implications for reconstructions of prehistoric land use and vegetation in Southern Italy

The mountainous inland of northern Calabria (Southern Italy) is known for its sparse prehistoric human occupation. Nevertheless, a thorough multidisciplinary approach of field walking, geophysical survey and invasive research led to the discovery of a major archaeological archive. This archive concerns a rich multi-phased dump, spanning about 3000 years (Late Neolithic to Late Imperial Roman Age) and holding two Somma-Vesuvius tephra. Of these, the younger is a distinct layer of juvenile tephra from the Pompeii eruption, while the older concerns reworked tephra from the Bronze Age AP2 eruption (ca. 1700 cal. yr BP). The large dump contains abundant ceramics, faunal remains and charcoal, and most probably originated through long-continued deposition of waste in a former gully like system of depressions. This resulted in an inversed, mound-like relief, whose anthropogenic origin had not been recognized in earlier research. The tephras were found to be important markers that support the reconstruction of the occupational history of the site. The sequence of occupational phases is very similar to that observed in a recent palaeoecological study from nearby situated former lakes (Lago Forano/Fontana Manca). This suggests that this sequence reflects the more regional occupational history of Calabria, which goes back to ca. 3000 BC. Attention is paid to the potential link between this history and Holocene climatic phases, for which no indication was found. The history deviates strongly from histories deduced from the few, but major palaeorecords elsewhere in the inlands of Southern Italy (Lago Grande di Monticchio and Lago Trifoglietti). We conclude that major regional variation occurred in prehistoric land use and its impacts on the vegetation cover of Southern Italy, and studies of additional palaeoarchives are needed to unravel this complex history. Finally, shortcomings of archaeological predictive models are discussed and the advantages of truly integrated multidisciplinary research.

GIS-Based and Statistical Approaches in Archaeological Predictive Modelling (NE Romania)

Archaeological predictive modelling (APM) is an important method for archaeological research and cultural heritage management. This study tests the viability of a new statistical method for APM. Frequency ratio (FR) is widely used in the field of geosciences but has not been applied in APM. This study tests FR in a catchment from the north-eastern part of Romania to predict the possible location(s) of Eneolithic sites. In order to do that, three factors were used: soils, heat load index and slope position classification. Eighty percent of the sites were used to build the model, while the remaining 20% were used to externally test the model’s performance. The final APM was made with the help of GIS software and classified into four susceptibility classes: very high, high, medium and low. The success rate curve and the prediction rate curve reported values of the area under curve (AUC) of 0.72, and 0.75 respectively. The Kvamme’s Gain value for the model has a value of 0.56. Therefore, the final APM is reliable, so FR is a viable technique for APM. The final map can be successfully used in archaeological research, cultural heritage management and protection, preventive archaeology and sustainable development.

Understanding Cumulative Hazards in a Rustbelt City: Integrating GIS, Archaeology, and Spatial History

We combine the Historical Spatial Data Infrastructure (HSDI) concept developed within spatial history with elements of archaeological predictive modeling to demonstrate a novel GIS-based landscape model for identifying the persistence of historically-generated industrial hazards in postindustrial cities. This historical big data approach draws on over a century of both historical and modern spatial big data to project the presence of specific persistent historical hazards across a city. This research improves on previous attempts to understand the origins and persistence of historical pollution hazards, and our final model augments traditional archaeological approaches to site prospection and analysis. This study also demonstrates how models based on the historical record, such as the HSDI, complement existing approaches to identifying postindustrial sites that require remediation. Our approach links the work of archaeologists more closely to other researchers and to municipal decision makers, permitting closer cooperation between those involved in archaeology, heritage, urban redevelopment, and environmental sustainability activities in postindustrial cities.