Periodic Reporting for period 1 - UnderTheSands (Ancient irrigation detection and analysis using advanced remote sensing methods)
Période du rapport: 2022-09-05 au 2024-09-04
Ancient irrigation systems were the first large-scale human landscape modifications, creating and sustaining the first urban civilizations. Although there has been an enormous amount of research, a consistent methodology for identifying large-scale irrigation networks has yet to be developed.
The UnderTheSands MSCA project at the Catalan Institute of Classical Archaeology focused on applying advanced remote sensing and machine learning methods to studying ancient irrigation systems in arid environments. The project's objectives were to locate and reconstruct the irrigation networks of the study areas, to obtain their relative chronology by relating them to archaeological settlements, and to understand the long-term socio-economic and historical circumstances that created and maintained these irrigation networks. UnderTheSands aimed to develop a novel workflow for large-scale analysis of irrigation networks, which anyone could later apply to any other area with similar characteristics.
The research included different test areas in Iraq, Iran, Turkmenistan, Afghanistan and Morocco, each representing different landscape types and environmental conditions. This interdisciplinary approach allowed us to investigate methods for mapping the remains of ancient canals (levees) and natural or partially modified paleochannels in Iraq, Iran and Turkmenistan. Special emphasis was placed on the study of qanat subsurface irrigation systems using selected test areas in Iran, Afghanistan and Morocco. The proposed methodology is not confined to a single discipline, but it draws from remote sensing, terrain analysis, hybrid machine/deep learning methods, archaeomorphology, spatial correlation indices, historical studies and various data sources.
The development of a novel workflow can enable large-scale mapping of irrigation features, allow the creation of standardized irrigation datasets, provide a tool to study landscape evolution by comparing images from different time periods, and facilitate the development of large-scale AI and computer simulation models that address settlement development and evolution. The results of the project will not just benefit the field of archaeology, but they will have applications beyond the broader field of heritage management and conservation and to contemporary issues of climate change and aridification.
The UnderTheSands MSCA project at the Catalan Institute of Classical Archaeology focused on applying advanced remote sensing and machine learning methods to studying ancient irrigation systems in arid environments. The project's objectives were to locate and reconstruct the irrigation networks of the study areas, to obtain their relative chronology by relating them to archaeological settlements, and to understand the long-term socio-economic and historical circumstances that created and maintained these irrigation networks. UnderTheSands aimed to develop a novel workflow for large-scale analysis of irrigation networks, which anyone could later apply to any other area with similar characteristics.
The research included different test areas in Iraq, Iran, Turkmenistan, Afghanistan and Morocco, each representing different landscape types and environmental conditions. This interdisciplinary approach allowed us to investigate methods for mapping the remains of ancient canals (levees) and natural or partially modified paleochannels in Iraq, Iran and Turkmenistan. Special emphasis was placed on the study of qanat subsurface irrigation systems using selected test areas in Iran, Afghanistan and Morocco. The proposed methodology is not confined to a single discipline, but it draws from remote sensing, terrain analysis, hybrid machine/deep learning methods, archaeomorphology, spatial correlation indices, historical studies and various data sources.
The development of a novel workflow can enable large-scale mapping of irrigation features, allow the creation of standardized irrigation datasets, provide a tool to study landscape evolution by comparing images from different time periods, and facilitate the development of large-scale AI and computer simulation models that address settlement development and evolution. The results of the project will not just benefit the field of archaeology, but they will have applications beyond the broader field of heritage management and conservation and to contemporary issues of climate change and aridification.
The UnderThesands project tested the applicability of MSRM (Multi-Scale Relief Model), SMTVI (Seasonal Multi-Temporal Vegetation Indices), TCT (Multi-Temporal Tasselled Cap Transformation), and Principal Component Analysis using Sentinel-2 and LANDSAT to identify paleochannels and traces of ancient canals within the chosen study areas. New algorithms were developed using Synthetic Aperture Radar Sentinel-1 data. The paleochannel network was then mapped, and the composite of SMTVI and MSRM was created, laying the groundwork for the project's upcoming phase involving machine/deep learning techniques.
UnderTheSands project developed the machine learning model for detecting levees in Google Earth Engine (GEE). The model, which combined SMTVI, TCT, and raw Sentinel-2 reflectance data with MSRM in the Murghab fan area, had insufficient applicability to other areas because of the limitations of the GEE platform. However, the team's extensive work on a deep learning U-Net model using training data from various locations resulted in the highest precision and recall of about 0.66 with nine selected Sentinel-2 bands or the SMTVI. To further enhance the model's performance, the team incorporated Swin UNETR and Attention UNET architectures based on Visual Transformers into the research, achieving a precision of 0.7 and recall of 0.78 with one of the selected areas. We plan to continue increasing the dataset's size and adjusting parameters to enhance the model's performance further.
The project focused on detecting the ancient underground water distribution systems called qanat (kārīz, kariz, or falaj), visible in satellite imagery as linear groups of holes. First, the Gorgan Plain region bordering Turkmenistan was selected as a training area. Several approaches to labelling, dataset splitting, deep learning techniques (Instant Segmentation and Object Detection), and architectures were tested (YOLOv5 and YOLOv7-9). A new approach to the dataset was created. This included marking individual shafts and their pairs. The training data was created to best represent qanats from Gorgan and Maywand areas. Part of the dataset consisted of augmented and artificial representations of qanats, which have been created. After completing the central core of the model, the post-processing and pre-processing techniques have been developed.
Two detailed study test areas, Murghab Fan and Serakhs / Geoksyur Oasis, have been selected for detailed archaeomorphological analysis of hydrological networks and autocorrelation techniques. The hypothetical dating of the paleochannels and remains of ancient canals (levees) has been proposed. The results were further integrated into a broader discussion of landscape and water management evolution in the selected studied areas. An intensive literature review has been conducted to cover all the analysed areas.
UnderTheSands project developed the machine learning model for detecting levees in Google Earth Engine (GEE). The model, which combined SMTVI, TCT, and raw Sentinel-2 reflectance data with MSRM in the Murghab fan area, had insufficient applicability to other areas because of the limitations of the GEE platform. However, the team's extensive work on a deep learning U-Net model using training data from various locations resulted in the highest precision and recall of about 0.66 with nine selected Sentinel-2 bands or the SMTVI. To further enhance the model's performance, the team incorporated Swin UNETR and Attention UNET architectures based on Visual Transformers into the research, achieving a precision of 0.7 and recall of 0.78 with one of the selected areas. We plan to continue increasing the dataset's size and adjusting parameters to enhance the model's performance further.
The project focused on detecting the ancient underground water distribution systems called qanat (kārīz, kariz, or falaj), visible in satellite imagery as linear groups of holes. First, the Gorgan Plain region bordering Turkmenistan was selected as a training area. Several approaches to labelling, dataset splitting, deep learning techniques (Instant Segmentation and Object Detection), and architectures were tested (YOLOv5 and YOLOv7-9). A new approach to the dataset was created. This included marking individual shafts and their pairs. The training data was created to best represent qanats from Gorgan and Maywand areas. Part of the dataset consisted of augmented and artificial representations of qanats, which have been created. After completing the central core of the model, the post-processing and pre-processing techniques have been developed.
Two detailed study test areas, Murghab Fan and Serakhs / Geoksyur Oasis, have been selected for detailed archaeomorphological analysis of hydrological networks and autocorrelation techniques. The hypothetical dating of the paleochannels and remains of ancient canals (levees) has been proposed. The results were further integrated into a broader discussion of landscape and water management evolution in the selected studied areas. An intensive literature review has been conducted to cover all the analysed areas.
The UnderTheSands project has developed an innovative workflow for identifying ancient canals, watercourse remnants, and qanat underground irrigation systems in arid regions. This unique approach sets our project apart and promises exciting new insights into archaeology and environmental science.
The results of UnderThesands indicate no universal method for mapping all irrigation features. The most widely applied application for studying the remains of ancient canals (levees) and paleochannels is analysing local terrain relief. The utilisation of the TanDEM-X resolution in the MSRM (Multi-Scale Relief Model) calculation enabled the mapping of local terrain relief with excellent detail. Differences in vegetation and soil, specifically differences in their ability to store water, can also be used to map the location of significant levees. Those characteristics can be used to enhance the detection of the levees using machine learning and deep learning. Paleochannels can be found by combining vegetating differences with local relief.
New methods based on seasonal composites of Sentinel-1 SAR backscatter were developed. The VH band can enhance moisture difference, leading to medium-large and extensive features even in transformed areas by modern agriculture. The VV band, on the other hand, is primarily valuable for completely deserted areas; it enhances the visibility of subtle relief changes. By exploring each designated test area, paleochannels and levees were identified, many of which were previously unknown.
The UnderTheSands project has developed a global deep learning model for detecting qanat underground irrigation systems from grayscale HEXAGON imagery. This model, which can detect almost 63% of all qanat shafts with high accuracy using different grayscale imagery (CORONA and HEXAGON), is not limited to a specific region. It can be applied in various areas, as demonstrated by its high precision (0.881) and recall (0.627) when tested in Afghanistan, particularly in desertic areas where it achieved the best results (precision: 0.993 and recall: 0.739).
The results of UnderThesands indicate no universal method for mapping all irrigation features. The most widely applied application for studying the remains of ancient canals (levees) and paleochannels is analysing local terrain relief. The utilisation of the TanDEM-X resolution in the MSRM (Multi-Scale Relief Model) calculation enabled the mapping of local terrain relief with excellent detail. Differences in vegetation and soil, specifically differences in their ability to store water, can also be used to map the location of significant levees. Those characteristics can be used to enhance the detection of the levees using machine learning and deep learning. Paleochannels can be found by combining vegetating differences with local relief.
New methods based on seasonal composites of Sentinel-1 SAR backscatter were developed. The VH band can enhance moisture difference, leading to medium-large and extensive features even in transformed areas by modern agriculture. The VV band, on the other hand, is primarily valuable for completely deserted areas; it enhances the visibility of subtle relief changes. By exploring each designated test area, paleochannels and levees were identified, many of which were previously unknown.
The UnderTheSands project has developed a global deep learning model for detecting qanat underground irrigation systems from grayscale HEXAGON imagery. This model, which can detect almost 63% of all qanat shafts with high accuracy using different grayscale imagery (CORONA and HEXAGON), is not limited to a specific region. It can be applied in various areas, as demonstrated by its high precision (0.881) and recall (0.627) when tested in Afghanistan, particularly in desertic areas where it achieved the best results (precision: 0.993 and recall: 0.739).