Periodic Reporting for period 2 - archaeoscape.ai (Exploring complexity in the archaeological landscapes of monsoon Asia using lidar and deep learning)
Reporting period: 2022-04-01 to 2023-09-30
The Anthropocene defines the period in which human activity has impacted Earth systems on a planetary scale. Although the definition of this epoch has achieved widespread acceptance in the scientific community and beyond, past human impacts on the natural environment are poorly understood in many parts of the world. This is perhaps especially true of humid tropical environments, where dense vegetation cover typically obscures the traces of human activity that might otherwise be observable on the surface in conventional aerial and satellite imagery. In places like Southeast Asia, what we are left with in many cases are the remains of monumental stone structures that indicate the existence of complex societies with large populations, but very little insight into the nature, scale or intensity of their transformations of regional landscapes.
In recent years, however, the use of lidar technology in archaeology has made an important contribution to resolving this problem. Using this airborne laser scanning technique, scientists can now reveal faint topographic traces over wide areas, even beneath dense tropical forest. In some areas, lidar has revealed previously-undocumented urban and agricultural landscapes surrounding well-known temple complexes, and in other areas filled in critical lacunae. Almost everywhere that lidar scans have been undertaken in Southeast Asia, we have documented new evidence for transformations of natural landscapes in the distant past.
In some respects, however, the ‘lidar revolution’ in tropical archaeology has created a series of important challenges alongside these new opportunities. Our coverage of the landscape is patchy, and important areas remain unstudied. The sheer volume of data now being acquired means that we can no longer continue to rely on subjective, hand-drawn maps and interpretations of lidar imagery, which also hinders quantitative, cross-cultural comparison. Access to massive geospatial datasets and tools to interpret them is uneven and limited, creating inequality and barriers to collaboration.
This project will develop the technical foundations to address these challenges. We use cutting-edge lidar technologies to undertake the largest-ever acquisition achieved by archaeologists in Asia and distribute it to stakeholders. We have established a hub for coordinating and collaborating on applications of AI for feature identification in archaeo-geospatial data. With a range of partners, we are pursuing frontier research in computer vision and deep learning, developing computational models for automatically identifying and analysing traces of human activity on the Earth’s surface. We develop and host web-based infrastructure to make those services available to the research community, in order to address systemic barriers to entry and allow researchers to concentrate on higher-order analytical tasks instead of technical problems.
Ultimately, the archives of past activity we uncover with lidar and deep learning will offer a unique laboratory for evaluating the nature and degree of human impacts on natural systems from the deep past to the present day, and for tracing pathways of innovation and complexity in the tropical world. Moving forward, our aim is to make a major contribution to a pantropical perspective on how humans have reshaped the surface of our planet in the Anthropocene, with a particular emphasis on trajectories of urbanism, contemporary ecological legacies of past human behaviour, and the ways in which engineered landscapes shape and constrain responses to social and environmental change over the long term.
In recent years, however, the use of lidar technology in archaeology has made an important contribution to resolving this problem. Using this airborne laser scanning technique, scientists can now reveal faint topographic traces over wide areas, even beneath dense tropical forest. In some areas, lidar has revealed previously-undocumented urban and agricultural landscapes surrounding well-known temple complexes, and in other areas filled in critical lacunae. Almost everywhere that lidar scans have been undertaken in Southeast Asia, we have documented new evidence for transformations of natural landscapes in the distant past.
In some respects, however, the ‘lidar revolution’ in tropical archaeology has created a series of important challenges alongside these new opportunities. Our coverage of the landscape is patchy, and important areas remain unstudied. The sheer volume of data now being acquired means that we can no longer continue to rely on subjective, hand-drawn maps and interpretations of lidar imagery, which also hinders quantitative, cross-cultural comparison. Access to massive geospatial datasets and tools to interpret them is uneven and limited, creating inequality and barriers to collaboration.
This project will develop the technical foundations to address these challenges. We use cutting-edge lidar technologies to undertake the largest-ever acquisition achieved by archaeologists in Asia and distribute it to stakeholders. We have established a hub for coordinating and collaborating on applications of AI for feature identification in archaeo-geospatial data. With a range of partners, we are pursuing frontier research in computer vision and deep learning, developing computational models for automatically identifying and analysing traces of human activity on the Earth’s surface. We develop and host web-based infrastructure to make those services available to the research community, in order to address systemic barriers to entry and allow researchers to concentrate on higher-order analytical tasks instead of technical problems.
Ultimately, the archives of past activity we uncover with lidar and deep learning will offer a unique laboratory for evaluating the nature and degree of human impacts on natural systems from the deep past to the present day, and for tracing pathways of innovation and complexity in the tropical world. Moving forward, our aim is to make a major contribution to a pantropical perspective on how humans have reshaped the surface of our planet in the Anthropocene, with a particular emphasis on trajectories of urbanism, contemporary ecological legacies of past human behaviour, and the ways in which engineered landscapes shape and constrain responses to social and environmental change over the long term.
The project ties together three separate and interrelated approaches: a geospatial component, revolving around the acquisition of lidar data at landscape scale; a computational component, involving the development of automated approaches to data analysis and tools for dissemination and collaboration; and an archaeological component, which locates this work in a broader historical and ecological context and explores the implications of the findings.
In terms of the specific aims of each stream of the project, and work completed so far:
The geospatial component of the project is anchored by the acquisition of airborne lidar data covering several thousand square kilometres of Southeast Asia, spanning areas of archaeological interest in several countries. At this point, we have participated in the acquisition of more than 4000 km2 of lidar, effectively tripling the coverage of archaeological lidar in the region. The project has also integrated raw data and results from previous lidar campaigns into a coherent geospatial framework, and has worked with local partners in the region to set the stage for further acquisitions in additional countries in the second half of the project.
The computational stream is managed through a purpose-built laboratory in Paris that is funded by this project, and that is co-led by the principal investigator and a computer scientist specialised in AI for automated image analysis. Having established this lab, the project has been focused on pushing the frontiers of machine learning approaches to recognition of material culture in big geospatial datasets. As a first step, the lab has aggregated previous lidar data and archaeological analyses into datasets for training and validation of AI models. The project has also established partnerships in industry and academia to consolidate previous work on automated feature recognition in archaeology and develop a roadmap for new approaches. Work on designing, developing and evaluating algorithms for object recognition in lidar datasets is well underway, with particular attention to the state of the art in computer vision and architectures for processing 3D datasets like point clouds.
In parallel, an archaeological and ecological agenda has revolved around coordinating work in the lab and in the field, the analysis and interpretation of the lidar data, managing the articulation between the geospatial and computational work packages, and evaluating the implications of the project’s findings for broader theoretical perspectives in archaeology and landscape ecology. Technical milestones achieved so far include the aggregation of all existing datasets and other material for the areas of interest from the domains of archaeology, ecology and paleoecology, and integration within a common digital/geospatial framework; consulting with stakeholders across the Southeast Asian region to develop broad and inclusive research partnerships, refine localised research objectives, and outreach to the general public; completing fieldwork in the region to refine models of archaeological topography and establish baseline parameters for the acquisition of archaeological lidar data in the aerial surveys; targeted field verification of lidar results; applying techniques from digital/spatial archaeology to develop quantitative models of socio-ecological change and complexity; and bringing those insights to publication.
In terms of the specific aims of each stream of the project, and work completed so far:
The geospatial component of the project is anchored by the acquisition of airborne lidar data covering several thousand square kilometres of Southeast Asia, spanning areas of archaeological interest in several countries. At this point, we have participated in the acquisition of more than 4000 km2 of lidar, effectively tripling the coverage of archaeological lidar in the region. The project has also integrated raw data and results from previous lidar campaigns into a coherent geospatial framework, and has worked with local partners in the region to set the stage for further acquisitions in additional countries in the second half of the project.
The computational stream is managed through a purpose-built laboratory in Paris that is funded by this project, and that is co-led by the principal investigator and a computer scientist specialised in AI for automated image analysis. Having established this lab, the project has been focused on pushing the frontiers of machine learning approaches to recognition of material culture in big geospatial datasets. As a first step, the lab has aggregated previous lidar data and archaeological analyses into datasets for training and validation of AI models. The project has also established partnerships in industry and academia to consolidate previous work on automated feature recognition in archaeology and develop a roadmap for new approaches. Work on designing, developing and evaluating algorithms for object recognition in lidar datasets is well underway, with particular attention to the state of the art in computer vision and architectures for processing 3D datasets like point clouds.
In parallel, an archaeological and ecological agenda has revolved around coordinating work in the lab and in the field, the analysis and interpretation of the lidar data, managing the articulation between the geospatial and computational work packages, and evaluating the implications of the project’s findings for broader theoretical perspectives in archaeology and landscape ecology. Technical milestones achieved so far include the aggregation of all existing datasets and other material for the areas of interest from the domains of archaeology, ecology and paleoecology, and integration within a common digital/geospatial framework; consulting with stakeholders across the Southeast Asian region to develop broad and inclusive research partnerships, refine localised research objectives, and outreach to the general public; completing fieldwork in the region to refine models of archaeological topography and establish baseline parameters for the acquisition of archaeological lidar data in the aerial surveys; targeted field verification of lidar results; applying techniques from digital/spatial archaeology to develop quantitative models of socio-ecological change and complexity; and bringing those insights to publication.
Previous archaeological lidar campaigns in Southeast Asia have focused on the areas immediately surrounding well-known monumental remains, introducing bias into the spatial coverage of the datasets and the interpretations that derive from them. With the radical expansion of the scope of archaeological lidar in the region, we have begun to address this bias by including large swaths of the landscape distant from monumental remains, including areas of supposed ‘wilderness’. This has created a baseline for a more objective and systematic understanding of anthropogenic change to the landscape in different time periods. The data reveal a complex mosaic of settlement space and early agricultural systems that challenge the conventional divide between ‘natural’ and ‘cultural’ landscapes in the region.
Our project has pioneered the use of UAV technology to acquire lidar data at landscape scale in the region, and a baseline comparison has been completed to evaluate the precision and accuracy of UAV-borne lidar to data acquired using a conventional aircraft. The project has also developed partnerships with commercial entities undertaking wide-area lidar acquisitions in the region for purposes such as geological exploration, with a view to adapting commercial acquisitions to produce datasets that are also of interest to archaeologists.
The project has successfully completed the first application of machine learning approaches to the analysis of archaeological lidar datasets in the Asian tropics, demonstrating the validity of the approach in this environment and cultural milieu. We have advanced the state of the art by focusing on replicability and interpretability of our methods; by developing and using vast datasets for training and validation deriving from decades of hand-mapped archaeological data; by expanding the standard approach to identifying one archaeological feature type to an entire suite of cultural remains across different early societies; and by grappling with applying algorithms to ‘raw’ elevation and rich point cloud data, rather than the degraded, human-interpretable images that are typically used in this domain.
At the half way point the project has begun to explore the broader implications of this work for heritage, environmental history, and archaeology, and some of the results have been published in a peer-reviewed context. Published work so far has focused on tracing and defining the trajectory of early urban and agricultural systems in mainland Southeast Asia, and re-appraising the emergence and ‘collapse’ of regional societies in light of the new, lidar-derived information on early landscape transformations. Published work has also focused the ways in which remote sensing technology in archaeology, and lidar in particular, underpins new perspectives about human-environment interactions over the long term, and of the importance of acquiring data before this material legacy on the Earth’s surface is permanently lost to modern development.
Finally, our project has established a consortium with research teams working in the tropical Americas, including data-sharing agreements, with a view to undertaking comparative, quantitative studies of tropical urbanism on a global scale. For decades, researchers have proposed that a specific kind of low-density urban trajectory emerged in early tropical civilisations in Southeast Asia and the Americas, but evidence so far has been anecdotal. A common and consistent baseline of archaeological lidar datasets, along with AI-driven analytics, has now set the stage for a rigorous, quantitative appraisal of this hypothesis, which seeks to challenge conventional models of urbanism from past to present.
Our project has pioneered the use of UAV technology to acquire lidar data at landscape scale in the region, and a baseline comparison has been completed to evaluate the precision and accuracy of UAV-borne lidar to data acquired using a conventional aircraft. The project has also developed partnerships with commercial entities undertaking wide-area lidar acquisitions in the region for purposes such as geological exploration, with a view to adapting commercial acquisitions to produce datasets that are also of interest to archaeologists.
The project has successfully completed the first application of machine learning approaches to the analysis of archaeological lidar datasets in the Asian tropics, demonstrating the validity of the approach in this environment and cultural milieu. We have advanced the state of the art by focusing on replicability and interpretability of our methods; by developing and using vast datasets for training and validation deriving from decades of hand-mapped archaeological data; by expanding the standard approach to identifying one archaeological feature type to an entire suite of cultural remains across different early societies; and by grappling with applying algorithms to ‘raw’ elevation and rich point cloud data, rather than the degraded, human-interpretable images that are typically used in this domain.
At the half way point the project has begun to explore the broader implications of this work for heritage, environmental history, and archaeology, and some of the results have been published in a peer-reviewed context. Published work so far has focused on tracing and defining the trajectory of early urban and agricultural systems in mainland Southeast Asia, and re-appraising the emergence and ‘collapse’ of regional societies in light of the new, lidar-derived information on early landscape transformations. Published work has also focused the ways in which remote sensing technology in archaeology, and lidar in particular, underpins new perspectives about human-environment interactions over the long term, and of the importance of acquiring data before this material legacy on the Earth’s surface is permanently lost to modern development.
Finally, our project has established a consortium with research teams working in the tropical Americas, including data-sharing agreements, with a view to undertaking comparative, quantitative studies of tropical urbanism on a global scale. For decades, researchers have proposed that a specific kind of low-density urban trajectory emerged in early tropical civilisations in Southeast Asia and the Americas, but evidence so far has been anecdotal. A common and consistent baseline of archaeological lidar datasets, along with AI-driven analytics, has now set the stage for a rigorous, quantitative appraisal of this hypothesis, which seeks to challenge conventional models of urbanism from past to present.