24
views
0
recommends
+1 Recommend
1 collections
    0
    shares
      • Record: found
      • Abstract: found
      • Article: found
      Is Open Access

      Automated detection of archaeological mounds using machine-learning classification of multisensor and multitemporal satellite data

      research-article

      Read this article at

      Bookmark
          There is no author summary for this article yet. Authors can add summaries to their articles on ScienceOpen to make them more accessible to a non-specialist audience.

          Significance

          This paper illustrates the potential of machine learning-based classification of multisensor, multitemporal satellite data for the remote detection and mapping of archaeological mounded settlements in arid environments. Our research integrates multitemporal synthetic-aperture radar and multispectral bands to produce a highly accurate probability field of mound signatures. The results largely expand the known concentration of Indus settlements in the Cholistan Desert in Pakistan ( ca. 3300 to 1500 BC), with the detection of hundreds of new sites deeper in the desert than previously suspected including several large-sized (>30 ha) urban centers. These distribution patterns have major implications regarding the influence of climate change and desertification in the collapse of the largest of the Old-World Bronze Age civilizations.

          Abstract

          This paper presents an innovative multisensor, multitemporal machine-learning approach using remote sensing big data for the detection of archaeological mounds in Cholistan (Pakistan). The Cholistan Desert presents one of the largest concentrations of Indus Civilization sites (from ca. 3300 to 1500 BC). Cholistan has figured prominently in theories about changes in water availability, the rise and decline of the Indus Civilization, and the transformation of fertile monsoonal alluvial plains into an extremely arid margin. This paper implements a multisensor, multitemporal machine-learning approach for the remote detection of archaeological mounds. A classifier algorithm that employs a large-scale collection of synthetic-aperture radar and multispectral images has been implemented in Google Earth Engine, resulting in an accurate probability map for mound-like signatures across an area that covers ca. 36,000 km 2. The results show that the area presents many more archaeological mounds than previously recorded, extending south and east into the desert, which has major implications for understanding the archaeological significance of the region. The detection of small (<5 ha) to large mounds (>30 ha) suggests that there were continuous shifts in settlement location. These shifts are likely to reflect responses to a dynamic and changing hydrological network and the influence of the progressive northward advance of the desert in a long-term process that culminated in the abandonment of much of the settled area during the Late Harappan period.

          Related collections

          Most cited references97

          • Record: found
          • Abstract: not found
          • Article: not found

          Abrupt weakening of the summer monsoon in northwest India  4100 yr ago

            Bookmark
            • Record: found
            • Abstract: found
            • Article: found

            Fluvial landscapes of the Harappan civilization

            The collapse of the Bronze Age Harappan, one of the earliest urban civilizations, remains an enigma. Urbanism flourished in the western region of the Indo-Gangetic Plain for approximately 600 y, but since approximately 3,900 y ago, the total settled area and settlement sizes declined, many sites were abandoned, and a significant shift in site numbers and density towards the east is recorded. We report morphologic and chronologic evidence indicating that fluvial landscapes in Harappan territory became remarkably stable during the late Holocene as aridification intensified in the region after approximately 5,000 BP. Upstream on the alluvial plain, the large Himalayan rivers in Punjab stopped incising, while downstream, sedimentation slowed on the distinctive mega-fluvial ridge, which the Indus built in Sindh. This fluvial quiescence suggests a gradual decrease in flood intensity that probably stimulated intensive agriculture initially and encouraged urbanization around 4,500 BP. However, further decline in monsoon precipitation led to conditions adverse to both inundation- and rain-based farming. Contrary to earlier assumptions that a large glacier-fed Himalayan river, identified by some with the mythical Sarasvati, watered the Harappan heartland on the interfluve between the Indus and Ganges basins, we show that only monsoonal-fed rivers were active there during the Holocene. As the monsoon weakened, monsoonal rivers gradually dried or became seasonal, affecting habitability along their courses. Hydroclimatic stress increased the vulnerability of agricultural production supporting Harappan urbanism, leading to settlement downsizing, diversification of crops, and a drastic increase in settlements in the moister monsoon regions of the upper Punjab, Haryana, and Uttar Pradesh. The collapse of the Bronze Age Harappan or Indus Civilization remains an enigma ( 1 ). The Harappans inhabited the Indus plain at the arid edge of the monsoonal belt and developed one of the earliest urban civilizations over a territory larger than the contemporary extent of Egypt and Mesopotamia combined ( Fig. P1 A ). Between the Indus and Ganges watersheds, a now largely defunct drainage system, the Ghaggar-Hakra, was also heavily populated during Harappan times. Controlled by the monsoon and the melting of Himalayan snow, the variable hydrologic regime must have been a concern for Harappans, as it is today for a billion people living on the Indo-Gangetic Plain. We investigate how climate change affected this civilization by focusing on fluvial morphodynamics, which constitutes a critical gap in our current understanding of the Harappan environment. Shuttle Radar Topography Mission (SRTM) data combined with field surveys and radiocarbon and optically stimulated luminescence dating offer us a way to analyze fluvial landforms and date deposits of the Indo-Gangetic Plain. We reexamine archaeological site distribution in relation to climate changes and argue for a gradual decrease in flood intensity leading to a cumulative process of settlement shift eastwards, rather than a sudden collapse linked to a single arid event. Climate Change and the Harappans. Inundation agriculture during the winter was dominant along the Indus from the Harappan era until modern times. Hydroclimate reconstructions for South and Central Asia (Fig. 4) show that precipitation from both monsoon and westerly sources was at its lowest after approximately 4,000 y BP ( 3 ). Aridification would have diminished flood intensity and allowed agriculture to expand along rivers. However, our analysis of fluvial landscapes suggests that further drying was detrimental for the Harappans, who relied on annual floods to sustain their economy. Preservation of archaeological sites at locations seemingly vulnerable to flooding, erosion, or burial suggest that, as aridity intensified, monsoon-augmented floods became less frequent and/or less intense. The most spectacular case of climate-controlled landscape transformation is the Ghaggar-Hakra system, which became ephemeral and was largely abandoned after approximately 4,000 y ago. Posturban settlements are preferentially located near the regularly flooded region at the Indus confluence with the Punjabi rivers, or in eastern regions with more reliable rains. Diversification of agriculture towards summer rain-based crops and the increase in drought-tolerant crops at the end of the urban phase (see SI Text ) reveal efforts to adapt to hydroclimatic stress. Although snowmelt continued to provide water to the Indus and its Himalayan tributaries, the Harappans did not develop canal irrigation. In contrast to inhabitants of Mesopotamia and Egypt, which were surrounded by arid lands, the Harappans had the option to migrate toward more humid regions of the Indo-Gangetic Plain. Migration toward the periphery could have contributed to the decline of urban centers in the core region of the Harappan domain. Without striving for deterministic explanations for the Indus Civilization collapse, it is likely that the unprecedented scale of hydroclimatic stress must have increased the vulnerability of floodwater farmers in Harappan society. Settlement Dynamics in the Harappan Domain. Occurrence of Harappan sites within the incised valleys of Punjab ( Fig. P1 B and C ) indicates that rivers were already entrenched by approximately 5,200 y ago. Numerous sites are located at the confluence zone of the Indus with the Punjabi rivers, where backwater flooding would have been common. Settlements on the Punjabi interfluves, including Harappa itself, tend to occur near their edges, close to fertile, annually flooded areas. Farther to the east, Harappan sites occur in the upper region of the Ghaggar-Hakra domain (Haryana and upper Punjab), but also along and within the incised valley of the Yamuna and on the Yamuna-Ganga interfluve. Settling of the Ghaggar-Hakra took advantage of the smaller floods typical for non-Himalayan monsoonal rivers. However, the largest agglomeration of mature Harappan sites, including the urban Ganweriwala, occurs on the lowermost Ghaggar-Hakra palaeochannel system in modern Cholistan. Proximity to both Ghaggar-Hakra and the well-watered confluence of the Indus with its Punjabi tributaries provides the best explanation for the continuity and high-density occupation there. Harappan discoveries in the Thar Desert adjacent to and along the Nara valley support our reconstructions of a better-watered past for this dry region. In upper Sindh ( Fig. P1 B and C ), good preservation of Harappan sites on the alluvial plain suggests that the Indus mega-ridge has been relatively stable. In contrast, in the southwestern Sindh, any settlement on the alluvial plain that may have existed is now buried under fluvial deposits. Morphodynamics of the Indo-Gangetic Plain. Our analysis reveals a palimpsest of fluvial forms and deposits in the western region of the Indo-Gangetic Plain ( Fig. P1 A ); however, one constant trait that is evident across the entire Harappan landscape is the change from an energetic fluvial regime earlier in the Holocene to increased stability of alluvial forms by Early Harappan times, as the climate became more arid ( 3 ). In contrast to the alluvial megafans of the eastern Indo-Gangetic Plain, we document that rivers in Punjab incised after approximately 10,000 y ago, but before approximately 3,900 y ago, probably driven by sediment load declining during centuries-long weak monsoon periods ( 4 ). Speculations have advanced the idea that the Ghaggar-Hakra, at times identified with the lost mythical Sarasvati River, was a large glacier-fed Himalayan river. Potential sources for this river include the Yamuna River, the Sutlej River, or both rivers, but the absence of wide, deeply entrenched valleys between the Sutlej and the Yamuna ( Fig. P1 A ) demonstrates that large, glacier-fed rivers did not flow across the Ghaggar-Hakra region during the Holocene. However, our evidence shows widespread fluvial redistribution of sediment, suggesting that monsoon rains were able to sustain perennial rivers and explaining why Harappan settlements flourished along the entire Ghaggar-Hakra system. Further south, we document fluvial deposition during Harappan times both at the confluence region between the Indus and Ghaggar-Hakra systems and along channels running through the desert toward the Nara valley. The Nara itself, which runs parallel to the Indus, had active fluvial sedimentation approximately 2,900 y ago ( Fig. P1 A ). Downstream, in Sindh, the Indus River built a unique distributive-type fluvial system that we term the Indus fluvial mega-ridge ( Fig. P1 A ). The alluvial plain here is convex-up, showing maximum aggradation near the modern channel belt and tapering out toward the plain edges. Radiocarbon-dated deposits of old channel belts indicate that aggradation was minimal during the last approximately 3,000 y compared to earlier in the Holocene, which indicates that the development of the Indus fluvial mega-ridge was the direct consequence of late Holocene aridity ( 3 ). The reduction in sediment load compared to water discharge linked to monsoon weakening can explain the channel incision and stabilization. Brief Harappan History. The Harappan cultural tradition ( 1 , 2 ) evolved during an Early Phase (5,200–4,500 y ago) from antecedent agricultural communities of the hills bordering the Indus alluvial plain to the west and reached its urban peak (Mature Phase) between approximately 4,500 and 3,900 y ago. The Harappans were agrarian, but developed large, architecturally complex urban centers and a sophisticated material culture coupled with a robust trade system. In contrast to Egypt and the neighboring hydraulic civilization of Mesopotamia, Harappans did not develop large-scale canal irrigation. Deurbanization after approximately 3,900 y ago was characterized by increasingly regional artifact styles and trading networks and the disappearance of the Harappan script ( 1 , 2 ). While these changes are often referred to as “collapse,” archaeological evidence indicates a protracted and regionally varied process ( 1 , 2 ).
              Bookmark
              • Record: found
              • Abstract: found
              • Article: not found

              Holocene Variations of Monsoon Rainfall in Rajasthan

              Two reconstructed histories of the monsoon rainfall in Rajasthan show that the monsoon was weak or absent in latest glacial time. With the advent of Holocene climatic patterns, fresh water lakes formed in dune fields and the pollen rain preserved in these reservoirs provides a basis for the reconstruction of the monsoon history. The two reconstructions, separated by only 150 km, have some features in common and some striking differences. Both show maximum monsoon amounts in the early Holocene, with a roughly two-thirds decrease to the present. Both show salinization in sub-Boreal time. Both show long intervals of near complete desiccation in the last four millennia. The shorter term variations, however, are not closely parallel. These shorter term variations may be explained in terms of the behavior of the present day interannual variability.
                Bookmark

                Author and article information

                Journal
                Proc Natl Acad Sci U S A
                Proc. Natl. Acad. Sci. U.S.A
                pnas
                pnas
                PNAS
                Proceedings of the National Academy of Sciences of the United States of America
                National Academy of Sciences
                0027-8424
                1091-6490
                4 August 2020
                20 July 2020
                20 July 2020
                : 117
                : 31
                : 18240-18250
                Affiliations
                [1] aLandscape Archaeology Research Group (GIAP), Catalan Institute of Classical Archaeology , 43003 Tarragona, Spain;
                [3] bInstitute of Earth Sciences Jaume Almera, Spanish National Research Council , 08028 Barcelona, Spain;
                [2] cMcDonald Institute for Archaeological Research, University of Cambridge , CB2 3ER Cambridge, United Kingdom;
                [4] dCulture and Socio-Ecological Dynamics, Department of Humanities, Universitat Pompeu Fabra , 08005 Barcelona, Spain;
                [5] eCatalan Institution for Research and Advanced Studies, 08010 Barcelona, Spain;
                [6] fSchool of Geography, Archaeology and Environmental Studies, The University of the Witwatersrand , Johannesburg 2000, South Africa;
                [7] gDepartment of Archaeology, University of Cambridge , CB2 3DZ Cambridge, United Kingdom
                Author notes
                1To whom correspondence may be addressed. Email: horengo@ 123456icac.cat or fconesa@ 123456icac.cat .

                Edited by Elsa M. Redmond, American Museum of Natural History, New York, NY, and approved June 25, 2020 (received for review April 2, 2020)

                Author contributions: H.A.O. and C.A.P. designed research; H.A.O. and F.C.C. performed research; H.A.O. contributed new analytic tools; H.A.O., F.C.C., and A.L. analyzed data; and H.A.O., F.C.C., A.G.-M., A.L., A.S.G., M.M., and C.A.P. wrote the paper.

                Author information
                https://orcid.org/0000-0002-9385-2370
                https://orcid.org/0000-0002-4026-7266
                https://orcid.org/0000-0001-5416-2986
                https://orcid.org/0000-0002-6689-2908
                https://orcid.org/0000-0002-3324-5165
                https://orcid.org/0000-0002-9324-1545
                https://orcid.org/0000-0002-2926-7230
                Article
                202005583
                10.1073/pnas.2005583117
                7414161
                32690717
                7f427de2-b148-41ac-a9b7-e4d08675eb25
                Copyright © 2020 the Author(s). Published by PNAS.

                This open access article is distributed under Creative Commons Attribution License 4.0 (CC BY).

                History
                Page count
                Pages: 11
                Funding
                Funded by: EC | H2020 | H2020 Priority Excellent Science | H2020 European Research Council (ERC) 100010663
                Award ID: 648609
                Award Recipient : Hector A. Orengo Award Recipient : Adam S. Green Award Recipient : Cameron Petrie
                Funded by: EC | H2020 | H2020 Priority Excellent Science | H2020 Marie Skłodowska-Curie Actions (MSCA) 100010665
                Award ID: 794711
                Award Recipient : Francesc C. Conesa Award Recipient : Arnau Garcia-Molsosa
                Funded by: EC | H2020 | H2020 Priority Excellent Science | H2020 Marie Skłodowska-Curie Actions (MSCA) 100010665
                Award ID: 794711
                Award Recipient : Francesc C. Conesa Award Recipient : Arnau Garcia-Molsosa
                Funded by: Ministerio de Ciencia y Tecnología (MICYT) 501100006280
                Award ID: RYC-2016-19637
                Award Recipient : Hector A. Orengo
                Funded by: RCUK | Biotechnology and Biological Sciences Research Council (BBSRC) 501100000268
                Award ID: BB/P027970/1
                Award Recipient : Adam S. Green
                Categories
                Physical Sciences
                Earth, Atmospheric, and Planetary Sciences
                Social Sciences
                Anthropology

                multitemporal and multisensor satellite big data,machine learning,archaeology,indus civilization,virtual constellations

                Comments

                Comment on this article