- Email: [email protected]
Uncovering the ancient canal-based tuntian agricultural landscape at China's northwestern frontiers
G Model
ARTICLE IN PRESS
CULHER-3077; No. of Pages 10
Journal of Cultural Heritage xxx (2016) xxx–xxx
Available online at
ScienceDirect www.sciencedirect.com
Uncovering the ancient canal-based tuntian agricultural landscape at China’s northwestern frontiers Lei Luo a,b , Xinyuan Wang a,b,∗ , Jie Liu a,b , Huadong Guo a,b , Rosa Lasaponara c , Wei Ji a , Chuansheng Liu a,b a
Key Laboratory of Digital Earth Science, Institute of Remote Sensing and Digital Earth, Chinese Academy of Sciences, Beijing 100094, China International Centre on Space Technologies for Natural and Cultural Heritage under the Auspices of UNESCO, Beijing 100094, China c Institute of Methodologies for Environmental Analysis (IMAA), CNR, C.da Santa Loja, 85050 Tito Scalo (PZ), Italy b
a r t i c l e
i n f o
Article history: Received 3 February 2016 Accepted 30 April 2016 Available online xxx Keywords: Tuntian Agricultural landscape Canal Remote sensing GIS Archaeological
a b s t r a c t The tuntian system was a state-promoted system of military–agriculture, which originated in the Western Han dynasty (206 BC–9 AD). All the imperial dynasties in Chinese history adopted the practice of tuntian to cultivate and guard frontier areas as an important state policy for developing border areas and consolidating frontier defense. This paper describes the use of satellite remote sensing data to uncover an ancient canal-based tuntian system located in an oasis agricultural landscape adjacent to the ancient Kingdom of Loulan at the southern margin of the Tarim Basin. The remote sensing data examined include Chinese Gaofen-1 (GF-1) VHR imagery, Landsat-8 (LS-8) OLI data and ASTER Global Digital Elevation Model Version 2 (ASTER GDEMV2) products. The effective irrigated tuntian area was estimated to be 2800 ha and the maximum irrigated tuntian area was found to be more than 8000 ha during the area’s most prosperous period. The overall spatial structure of Milan’s tuntian agricultural landscape was explored using the patch–corridor–matrix model. By detailed analysis of satellite remote sensing data, this study reconstructed a 3D view of Milan’s tuntian agricultural landscape in a GIS. © 2016 Published by Elsevier Masson SAS.
1. Research aims The aim of the research presented hereafter is the application of satellite remote sensing and GIS to tuntian system in order to explore the ability of an integrated prospection approach to investigating, mapping and reconstructing cultural landscape and provide new insights for agricultural archaeology in arid area. 2. Introduction During the Han Dynasty (206 BC–220 AD) the ancient Silk Road from Dunhuang to Central Asia [1,2] was extended through the Tarim Basin between the Kunlun Mountains and the Tianshan Mountains (Fig. 1). To establish a military presence along these frontiers and to ease the problem of transporting food grain to these remote areas, the Han government sponsored military–agricultural colonies [1,3,4], known as the tuntian system, in northwest China
∗ Corresponding author at: Key Laboratory of Digital Earth Science, Institute of Remote Sensing and Digital Earth, Chinese Academy of Sciences, Beijing 100094, China. E-mail address: [email protected] (X. Wang).
in order to ensure the safety of traders along the Silk Road. The tuntian system grew out of the Han Dynasty policy of having garrison troops or newly settled peasants bring undeveloped land under cultivation [2]. The Han government first implemented the tuntian system in the Hexi Corridor [1–3], and soon extended it to the empire’s western regions (including today’s Xinjiang and Central Asia) after it proved to be a great success. It is known from Han records written on bamboo slips discovered at Xuanquan in 1990 that ancient Milan was one of the major headquarters of this system [4]. In the second year of the Han Emperor, Zhaodi (77 BC), Weituyan, the King of Shanshan (the state of Loulan) asked the Han government to send troops to develop the wasteland and to plant grain [1]. A sima (minister of war in ancient China) and 40 soldiers were then dispatched to garrison Milan and to farm there. Following the initial success of the Milan tuntian system, the Han government wasted no time in extending it to all of its frontiers; as a result the positive effects of this organized military farming were soon felt all over the ancient Xinjiang, which the Han dynasty unified for the first time [2]. In 1907, the British archaeologist, Stein, first unearthed the Milan archaeological site [5] and discovered a large number of sculptures and frescoes. In the 1950s, a Chinese exploration team discovered an intact system of irrigation canals as well as vast fertile fields buried under the desert [6]. In
http://dx.doi.org/10.1016/j.culher.2016.04.013 1296-2074/© 2016 Published by Elsevier Masson SAS.
Please cite this article in press as: L. Luo, et al., Uncovering the ancient canal-based tuntian agricultural landscape at China’s northwestern frontiers, Journal of Cultural Heritage (2016), http://dx.doi.org/10.1016/j.culher.2016.04.013
G Model CULHER-3077; No. of Pages 10
ARTICLE IN PRESS L. Luo et al. / Journal of Cultural Heritage xxx (2016) xxx–xxx
2
Fig. 1. The location of ancient Milan and the main routes of the ancient Silk Road. Pink lines indicate the Silk Road; the base map is the ETOPO1 Global Relief Model, which can be downloaded from http://maps.ngdc.noaa.gov/viewers/bathymetry.
1973, Xinjiang archaeologists unearthed Milan Castle at the center of the Milan site [6,7]. Early Buddhist sculptures and frescoes excavated from the site show stylistic similarities to the traditions of Central Asia and North India [8] and other artistic aspects of the paintings found there suggest that Milan had a direct connection with ancient Rome [9,10]. Studies that involve the investigation of cultural heritages and archaeological features increasingly employ aerial photographs and satellite remote sensing imagery [11–23]. The combined application of remote sensing and GIS have made it possible to realize agricultural archaeological detecting [24], mapping [19,25] and monitoring [26,27] in landscape investigation, but they’re far from meeting the demands of past landscape reconstructing [28–30]. It’s a great challenge for researchers (e.g. environmentalists, agriculturalists, ecologists, geographers and archaeologists), not only should they have a better understanding of satellite remote sensing and GIS, but they should know the environmental and geographic conditions [31], agricultural and soci-cultural characteristics [32] and past human activities [33]. As one of the most important state-promoted polices in ancient China, tuntian system was viewed as the symbol of arid agricultural development. It is important to try to map the tuntian system so that the defensive and agricultural policies employed along the northwestern frontiers of ancient China can be better understood. Section 3 introduces the materials and methods; the results and discussion are presented in Section 4 and Section 5, respectively and conclusions are drawn in Section 6. 3. Materials and methods 3.1. Archaeological site description Located at the eastern end of the Tarim Basin in Xinjiang Uyghur Autonomous Region, China, Milan was formerly one of the main stops on the ancient Silk Road (Fig. 1). The Milan site is adjacent to the new oasis of Milan town (Fig. 2), which was established by
the Xinjiang Production and Construction Corps (XPCC) after the foundation of the People’s Republic of China in 1949. The Milan site is located approximately 75 km east of Ruoqiang County and 165 km southwest of ancient Loulan (Fig. 2a). The Milan site is now one of the most important archaeological sites in the greater Lop Nor areas [6,7,34–36]. The current conservation area of the Milan site consists of Milan Castle together with cultural relics from the Han–Tang period (206 BC–907 AD), which are scattered around the surrounding areas [35,36], as well as the tuntian canals that were used for irrigating farmland and water conservation when the area was garrisoned during the Han Dynasty (Fig. 2b). Archaeological excavations since the Stein’s expedition have uncovered an extensive Buddhist monastic site that existed between the 2nd to 5th centuries AD [5], as well as Milan Castle (Fig. 2c), which was a Tibetan fort during the 8th and 9th centuries AD [6]. Based on the available historical records [1,2] and archaeological discoveries [5–7,34–36], archaeologists consider that Milan’s tuntian systems were continuously used from the Han Dynasty to Tang Dynasty (618–907 AD). 3.2. Satellite remote sensing data Gaofen (GF) images, acquired by the GF-1 satellite, were obtained from the China Centre for Resource Satellite Data and Applications. GF-1 is equipped with two panchromatic–multispectral (PMS) cameras, which can generate 2 m panchromatic (PAN) data and 8 m multispectral (MS) data (Table 1) across the same 60 km swath. Because Chinese GF-1 data have a very high-resolution, they are ideal sources for mapping the tuntian landscapes found at the Milan archaeological site. The GF-1 data used in this study were acquired on September 18th, 2014 at 5:22:54 UTC. LS-8 Operational Land Imager (OLI) data were acquired from the Institute of Remote Sensing and Digital Earth of the Chinese Academy of Sciences (http://ids.ceode.ac.cn/). This imagery is composed of nine bands covering the visible to the short-wave infrared
Please cite this article in press as: L. Luo, et al., Uncovering the ancient canal-based tuntian agricultural landscape at China’s northwestern frontiers, Journal of Cultural Heritage (2016), http://dx.doi.org/10.1016/j.culher.2016.04.013
G Model
ARTICLE IN PRESS
CULHER-3077; No. of Pages 10
L. Luo et al. / Journal of Cultural Heritage xxx (2016) xxx–xxx
3
Fig. 2. a: locations of Milan archaeological site and Lop Nor areas; b: the conservation area of the Milan site–ancient tuntian canals can be seen in the image; c: image of Milan Castle. The base map is MODIS satellite data saved from ArcGIS10.1; the images used for (b) and (c) are Gaofen-1 (GF-1) panchromatic (PAN) data with a histogram equalization enhancement.
Table 1 Characteristics of the GF-1 PMS and LS-8 OLI data used in this study. Sensors Wavelength/m
Spatial resolution/m
PAN MS
PAN MS
GF-1 PMS
LS-8 OLI
0.45–0.90 Blue 0.45–0.52 Green 0.52–0.59 Red 0.63–0.69 NIR 0.77–0.89
0.500–0.680 Coastal 0.43–0.45 Blue 0.45–0.52 Green 0.53–0.60 Red 0.63–0.68 NIR 0.85–0.89 SWIR 1 1.56–1.66 SWIR 2 2.10–2.30 Cirrus 1.36–1.39 15 30
2 8
GF-1: Gaofen-1; LS-8: PMS: panchromatic–multispectral; Landsat-8; PAN: panchromatic.
parts of the spectrum (Table 1). Each band has a spatial resolution of 30 m. These data also included a panchromatic image with a spatial resolution of 15 m. The LS-8 OLI pansharpened data were used as a supplement in the identification of the dry (buried) channels and trunk canals. In this study, the two LS-8 images used were acquired on October 17th, 2013. A full scene of a LS-8 image has a coverage area of about 185 km by 185 km. The ASTER GDEM, with 30 m grid postings and generated from data collected by the ASTER satellite, has been used to obtain topographic information by many users around the world [37]. ASTER GDEMV2 was released as an upgrade to ASTER GDEMV1 on October 17, 2011. The ASTER GDEM covers land surfaces between 83◦ N and 83◦ S and is composed of 22,600 1◦ by 1◦ tiles. In this study, ASTER GDEMV2 products were used to extract the drainage system of the Milan River by using the GIS hydrological analysis tools, and to produce 3D views by integrating these products with GF-1
and LS-8 OLI data. ASTER GDEMV2 products can be downloaded from the Geospatial Data Cloud (http://www.gscloud.cn/).
4. Methods 4.1.1. Remote sensing data processing The GF-1 and LS-8 images were first atmospherically and geometrically corrected in order to minimize the interference due to signal distortion caused by atmospheric and geometric factors. Data fusion was then carried out in order to combine the very high spatial resolution of the PAN imagery with the high spectral resolution of the MS images [38]. Both the GF-1 and LS-8 images were fused by Gram Schmidt pansharpening method, since it can yield less spectral distortion as compared with other fusion methods [39]. The pansharpened products were only for extracting dry and buried channels of the Milan River. Normalized difference vegetation index (NDVI) is a satellite-derived index form the Near-infrared (NIR) and Red channels, and it was found to be more capable to better enhance crop marks observed for green vegetation [15]. In this study, we discuss the use of GF-1 NDVI data to estimate the spatial extent of the ancient tuntian landscape at Milan. NDVI was expressed as follows: NDVI = (NIR − Red ) /(NIR + Red )
(1)
where NIR and Red were the reflectance of NIR and Red band, respectively. In this area, the NDVI not only reflects the growth of vegetation but also the distribution of paleochannels and springs, which are relatively rich in underground water.
Please cite this article in press as: L. Luo, et al., Uncovering the ancient canal-based tuntian agricultural landscape at China’s northwestern frontiers, Journal of Cultural Heritage (2016), http://dx.doi.org/10.1016/j.culher.2016.04.013
G Model CULHER-3077; No. of Pages 10 4
ARTICLE IN PRESS L. Luo et al. / Journal of Cultural Heritage xxx (2016) xxx–xxx
Fig. 3. Four pansharpened Gaofen-1 (GF-1) MS images and panchromatic (PAN) image for Milan site: a: NIR; b: red; c: green; d: blue; e: PAN image; and f: field photo of tuntian canal.
4.1.2. Extraction of tuntian canals and drainage system The tuntian canals were identified from GF-1 VHR PAN data based on visual interpretation in a GIS. Visual interpretation method has the advantage that the naked eye can identify subtle differences between agricultural archaeological features and the backgrounds that a computer cannot. The automatic feature extraction techniques [24,40–42] are not very successful except at an extremely limited range of spatial scales and spectral contrasts. For instance, in the study by Figorito and Tarantino [24], each image covered only 0.1 km2 and was chosen for the great spectral variability of the crop marks and the marked contrast between them and their surroundings. Based on a ground water model provided in the
hydrological analysis tools under the ArcGIS10.1 environment, the drainage system of the Milan River basin was extracted from ASTER GDEMV2 products showing the mountainous area. 4.1.3. Landscape structure analysis Landscape has been defined as a heterogeneous land area made up of a group of interacting ecosystems. The study on landscape pattern was an important field of land use and ecological change. In this study, the well-known patch–corridor–matrix model [43] was used to analyze the structure of coupling ecosystem found at Milan. Within this model, the landscape is considered to be a mosaic consisting of three major elements: patches, corridors, and matrix.
Please cite this article in press as: L. Luo, et al., Uncovering the ancient canal-based tuntian agricultural landscape at China’s northwestern frontiers, Journal of Cultural Heritage (2016), http://dx.doi.org/10.1016/j.culher.2016.04.013
G Model
ARTICLE IN PRESS
CULHER-3077; No. of Pages 10
L. Luo et al. / Journal of Cultural Heritage xxx (2016) xxx–xxx
5
Fig. 4. Results of the visual extraction of tuntian canals from Gaofen-1 (GF-1) panchromatic (PAN) image.
From an ecological perspective, patches represent relatively discrete areas of relatively homogeneous environmental conditions. Corridors are linear landscape elements that can be defined on the basis of their structure or function. Forman and Godron [43] define corridors as narrow strips of land that differ from the matrices to their two sides. The matrix is the most extensive and most connected landscape element and therefore plays the dominant role in the functioning of the landscape.
5. Results 5.1. Extraction results for tuntian canals Fig. 3a–d displays all four pansharpened GF-1 MS images individually over the Milan site. Despite differences in how each pansharpened image accentuates surface characteristics [23,38], little difference can be seen. However, the tuntian canal traces are clearly visible in GF-1 PAN image due to the obvious textural characteristics (Fig. 3e). Image texture gives us information about the spatial arrangement of color or intensities in an image or selected region of an image. Textural information has been an important factor in visual image interpretation. To the GF-1 PAN imagery, it takes into consideration the distribution and variation of neighborhood pixel values. The Milan’s tuntian canals, which have no distinctive spectral signature, are completely buried by a layer of aeolian sand deposits and gravels (Fig. 3f), which come from the alluvial fan of the Milan River. It is thus very difficult to distinguish the ancient canals from the background using spectral classification. In such situations, the GF-1 PAN image alone can be usefully employed to extract canal traces. The map was made in ArcGIS10.1 by digitally drawing polyline vector features over target canals (as shown on Fig. 4).
5.2. Extraction of drainage system Fig. 5a shows the MS, PAN and pansharpened GF-1 MS images of a buried channel: the pansharpened image allows better discrimination of the dry (buried) channels than the original MS image. Fig. 5b shows the MS, PAN and pansharpened LS-8 OLI images of tuntian canal and channels of the Milan River. These images allow easier visual identification of the canal and river than the original multispectral image; in particular, the spectral characteristics of water bodies are obvious in the pansharpened LS-8 image. Fig. 6 shows the drainage system of the Milan River in the mountainous area as extracted by using hydrological analysis tool in ArcGIS10.1. LS-8 OLI images provided the basis for mapping dried-up river channels in the plain area, where the morphological aspects of the dry channels can be observed exceptionally well due to their spectral signature and textural characteristics [44]. Most of these dry channels were once the main channel of the Milan River but have now disappeared beyond the southern margin of the Taklimakan Desert. 6. Discussion 6.1. Uncovering the canal system in ancient Milan’s tuntian agricultural landscape Based on the hydrological positions and the relationships among different tuntian canals, and by developing Morehart’s [19] typology, the canals in the Milan area can be considered to represent a hierarchical irrigation system that consists of four canal types: in this paper, these are labeled trunk, primary, secondary and tertiary. Fig. 7 shows a map of the canal-based tuntian irrigation system. Only one trunk canal existed. The main body of the trunk canal has disappeared due to shifting sand dunes and wind erosion. The
Please cite this article in press as: L. Luo, et al., Uncovering the ancient canal-based tuntian agricultural landscape at China’s northwestern frontiers, Journal of Cultural Heritage (2016), http://dx.doi.org/10.1016/j.culher.2016.04.013
G Model CULHER-3077; No. of Pages 10 6
ARTICLE IN PRESS L. Luo et al. / Journal of Cultural Heritage xxx (2016) xxx–xxx
Fig. 5. The left, middle and right panels are the original MS images, panchromatic (PAN) images and Gram Schmidt pansharpened images, respectively for (a) the Gaofen-1 (GF-1) data (RGB342 false color composite) and (b) the Landsat-8 (LS-8) data (RGB654 false color composite). Black arrows indicate channels.
pansharpened LS-8 image shows that the traces of the trunk canal lead from the middle reaches of the Milan River at Longkou Village (Fig. 7a). There are seven primary canals leading from the trunk canal through the Milan area. For the convenience of analysis, the primary canals were numbered P1, P2, P3, P4, P5, P6 and P7 from left to right (Fig. 7b). These primary canals form a fan-shaped structure and constitute the basic skeleton of Milan’s canal-based irrigation system. The primary canals, which are analogous to the central arteries of the human body, were used to distribute water from the trunk canal. Secondary canals, which were branches of the primary canals, were built to assist the distribution of water from the primary canals to the hinterland of the ancient Milan area. From a hydrological point of view, the primary and secondary canals constituted a basis irrigation network. The tertiary canals were “feeder” canals [19] that distributed water from the primary canals to the central region of the ancient Milan area. Compared with the primary canals, the tertiary canals are the shorter, narrower canals lying along the same direction that give the tuntian landscape its characteristic appearance. Most of these canals led directly from primary canals, although some were connected to secondary canals. The tertiary canals appear to be
organized in a more regular way; they were usually parallel to other tertiary canals and perpendicular to the higher-level primary and secondary canals. The visual extraction method was able to extract 754 tertiary canals of various sizes. 6.2. Estimating the spatial extent of ancient Milan’s tuntian agricultural landscape Fig. 8a shows an NDVI map of the Milan site. Based on the observed paleochannel patterns and the NDVI values in combination with visual interpretation of the GF-1 data, the greater Milan oasis was manually divided into five parts. These were labeled Area I, Area II, Area III, Area IV and Area V (Fig. 8b). Area I, today’s Milan archaeological site, has barely any vegetation cover. The extent of this site is around 1800 ha. Area II, part of the new Milan oasis, is covered by dense crops with a high NDVI value. Based on the paleochannel patterns and the boundaries of Area I, we speculate that the core area of ancient Milan included both Area I and Area II. This guess was confirmed by both historical Landsat-1 MSS (Fig. 8c) data and Stein’s [5] archaeological map (Fig. 8d). Area II was cultivated by the XPCC in 1970s, as detailed
Please cite this article in press as: L. Luo, et al., Uncovering the ancient canal-based tuntian agricultural landscape at China’s northwestern frontiers, Journal of Cultural Heritage (2016), http://dx.doi.org/10.1016/j.culher.2016.04.013
G Model CULHER-3077; No. of Pages 10
ARTICLE IN PRESS L. Luo et al. / Journal of Cultural Heritage xxx (2016) xxx–xxx
7
Fig. 6. Drainage system in the mountainous area with a flow accumulation threshold of 2500; dry and buried channels were extracted from the pansharpened Gaofen-1 (GF-1) and Landsat-8 (LS-8) MS data.
Fig. 7. Map of Milan’s canal-based tuntian irrigation system: (a) the canal structure in the core area, (b) a fan-shaped complex including trunk, primary, secondary and tertiary canals, photos of modern cement canal (c) and ancient canal (d).
Please cite this article in press as: L. Luo, et al., Uncovering the ancient canal-based tuntian agricultural landscape at China’s northwestern frontiers, Journal of Cultural Heritage (2016), http://dx.doi.org/10.1016/j.culher.2016.04.013
G Model CULHER-3077; No. of Pages 10 8
ARTICLE IN PRESS L. Luo et al. / Journal of Cultural Heritage xxx (2016) xxx–xxx
Fig. 8. a: Gaofen-1 (GF-1) normalized difference vegetation index (NDVI) map of greater Milan oasis depicting the canal system but with the tertiary canals omitted; b: map of greater Milan oasis divided into Areas I to V based on NDVI and paleochannel patterns; the base map is the GF-1 MS data (RGB432); c: Landsat-1 MSS image [acquired on July 1st, 1973 (RGB321)] of the Milan oasis revealing the interesting, oval tuntian agricultural landscape; d: Stein’s archaeological map.
in the chorographical records of the XPCC. The total area of Area I and II is about 2800 ha. Area III not only exhibits good continuity and gradual variation in the NDVI values but also contains archaeological vegetation marks that are clearly related to the tuntian landscape. Thus, it is concluded that Area III was a peripheral region of Milan’s tuntian agricultural landscape. The area of Area III is approximately 5200 ha. Area III combined with Areas I and II forms one region with an oval shape, which completes Milan’s tuntian agricultural landscape. The total area of this oval-shaped region is about 8000 ha. Area IV comprises the new Milan oasis, which has been reclaimed since the 1950s. Due to the sprawling nature of the new Milan oasis, the archaeological conservation area at the Milan site was partly included in the new area. Area V is covered with sparse vegetation and used to be part of the old Milan River. The Milan River once flowed northwards to the Tarim River through Areas IV and V (Fig. 8d) but has, instead, flowed northeastwards to the same river through Areas I, II and III since the flood control dam on the Milan River was constructed in the 1960s. Areas IV and V cover 12,500 ha and 4500 ha, respectively. 6.3. Exploring the spatial patterns in ancient Milan’s tuntian agricultural landscape The ancient tuntian agricultural landscape described in this paper was formed in the middle part (oasis area) of the Milan River, but developments in the upstream (mountainous area) and downstream (desert area) parts of the river jointly decided the eventual fate of the tuntian landscape. The Milan River, the basis of the
tuntian landscape, originates in the Altyn Mountains and is fed by glaciers and snow meltwater from these mountains located to the south on the Qinghai–Tibet Plateau. One of the main reasons why the tuntian landscape at Milan was abandoned was the invasion of the Taklimakan Desert. Thus, it is important to explore the spatial patterns in the tuntian landscape of ancient Milan at a wider scale. Milan’s tuntian agricultural landscape was co-controlled by the Altyn Mountain Ecosystem, the Milan Oasis System and the Taklimakan Desert System, which together constituted a coupled mountain–oasis–desert ecosystem (MODES) (Fig. 9). In this study, there were two relevant matrixes–the Altyn Mountains and the Taklimakan Desert. The Altyn Mountain range constitutes a matrix because it is greatest in areal extent, is mostly connected and exerts a dominant influence on the mountain ecosystem and its ecological processes. The Taklimakan Desert constitutes the second greatest matrix element; it is also mostly connected and exerts a dominant influence on the desert ecosystem and its ecological processes. Proposing the Milan River as a linear corridor supports the existence of the MODES. At a larger scale, the tuntian canal system can be regarded as an extension of the Milan River that also, therefore, belongs to the linear Milan River corridor. At the local scale, each individual tuntian canal is a linear corridor in the irrigated landscape. In Milan’s tuntian landscape, the Milan oasis and the Altyn glaciers are the most important patches in the Taklimakan Desert Matrix and the Altyn Mountain Matrix, respectively. Milan’s tuntian landscape depended on the Milan oasis, especially in ancient times. Although the Milan oasis constitutes a patch, the ecosystem of the
Please cite this article in press as: L. Luo, et al., Uncovering the ancient canal-based tuntian agricultural landscape at China’s northwestern frontiers, Journal of Cultural Heritage (2016), http://dx.doi.org/10.1016/j.culher.2016.04.013
G Model CULHER-3077; No. of Pages 10
ARTICLE IN PRESS L. Luo et al. / Journal of Cultural Heritage xxx (2016) xxx–xxx
9
Fig. 9. 3D view of the Mountain–Oasis–Desert Ecosystem (MODES) around Milan as viewed from the north. In this view, Gaofen-1 (GF-1) panchromatic–multispectral (PMS) (RGB432) and Landsat-8 OLI (RGB654) pansharpened data have been laid over an ASTER GDEMV2 DEM; the drainage system of the Milan River is also shown.
Milan oasis depends on the presence of human activity. The Altyn glaciers, which supplement the fresh water used for irrigation, were also crucial to the existence of Milan’s tuntian landscape. Largescale changes have impacted the spatio-temporal distribution of the glaciers on the Tibet–Qinghai Plateau, as shown by the retreat of the Altyn glaciers and associated rise in the snow line [34], and the occurrence of extreme climate events (blizzards and sandstorms) and disasters (snowmelt flood and drought) in the Milan oasis. The pattern of the Altyn mountain system is complicated and the complicated plateau environment creates abundant biodiversity, frequent hydrothermal exchanges and unique vertical natural zones. The Milan oasis system is a compound artificial one and high productivity, complicated structure and sensitivity constitute its basic characteristics. The Taklimakan Desert system also has some particular characteristics, such as a simple structure, high ecological vulnerability and low productivity. There are great differences between these three systems but the systems are not isolated–MODES is a compound system, and the three systems are connected by the Milan River.
7. Conclusions The 14 C ages for samples from Milan sites fall within the range 1450–1900 cal BP (about 50–500 AD) [36], i.e. a span of ∼ 500 years from the Eastern Han dynasty, through the Wei and Jin, to the Northern and Southern dynasties. The tuntian system was to have far-reaching effects, both for the Han government and for the overall economy and geopolitics of China. Barren land was reclaimed and hundreds of irrigated oases together with dozens of ancient towns were established in the frontier areas. Satellite remote sensing data were used to extract and map Milan’s tuntian system and to uncover traces of an ancient Chinese military–agriculture landscape. Based on the GF-1 extraction results and historical records, the hierarchical irrigation structure of Milan’s tuntian landscape was revealed and the functions of the different types of canal were discussed. It should be pointed out that paleoenvironment changes with time may influence the landscape and these changes may mask the true archaeological features. In this study, the Milan’s tuntian canals are completely buried by a layer of aeolian sand deposits and gravels due to the environmental risks and responses to global changes (drought, desertification and erosion).
The information about the locations of the branches and nodes of canals, the distribution of freshwater and the agricultural output of the tuntian system obtained for the Milan site are important parameters for understanding the ancient agricultural tuntian landscape and the spatial patterns of irrigation canals in the study area, and these aspects need to be studied in more detail in the future. Future work will also incorporate the use of radar and Lidar data to enrich the structural information, including width and height, about the canals that is available. In addition, attempts will be made to extract the ancient canal irrigation systems at other oases (e.g. Loulan, Qiemo and Yutian) in Xinjiang and the spatial characteristics of these canals will be analyzed using integrated GIS tools and field investigations. In particular, knowing the spatial pattern of irrigation canals across these regions assists in understanding the distribution and use of water resources, which is important in the context of the evolution of an irrigated oasis–agricultural civilization and the development of a border defense strategy. Acknowledgements This work was supported by the National Natural Science Foundation of China (No. 41271427), the Directorial Foundation for Doctoral Student Research of the Institute of Remote Sensing and Digital Earth (RADI). Many thanks are due to the anonymous reviewers for their constructive comments and suggestions. References [1] Ban G. 92. HanShu. Zhonghua Book Company, Beijing. (reprint in 1962 [in Chinese]). [2] Sima Q. 91 BC. Shiji. Zhonghua Book Company, Beijing. (reprint in 1962 [in Chinese]). [3] G. Lao, Textual Research and Explanation of the Juyan’s Han Bamboo Slips, Taiwan Press, Taipei, 1984 (in Chinese). [4] D.F. Zhang, Station troops to open up wasteland of Western Region in the Han dynasty based on the Han slips from Xuanquan, Dunhuang Res. 3 (2001) 113–121 (in Chinese). [5] A. Stein, Innermost Asia. Detailed Report of Explorations in Central Asia, Kan-su and Eastern Iran, Carried Out and Described under the Orders of H. M. Indian Government, Clarendon Press, Oxford, 1928. [6] X.C. Xia, B.F. Wang, Y.J. Zhao, Lop Nor in China, Science Press, Beijing, 2007 (in Chinese). [7] S.C. Wang, Historical status of the capital Loulan and ancient Lop Nor region, West Reg. Stud. 4 (1996) 43–53 (in Chinese). [8] H.A. Van Oort, The Iconography of Chinese Buddhism in Traditional China, Vol. 1, Brill Academic Publishers, Leiden, 1986. [9] M.M. Rhie, Early Buddhist art of China and Central Asia, Vol. 1, Brill Academic Publishers, Leiden, 1999.
Please cite this article in press as: L. Luo, et al., Uncovering the ancient canal-based tuntian agricultural landscape at China’s northwestern frontiers, Journal of Cultural Heritage (2016), http://dx.doi.org/10.1016/j.culher.2016.04.013
G Model CULHER-3077; No. of Pages 10 10
ARTICLE IN PRESS L. Luo et al. / Journal of Cultural Heritage xxx (2016) xxx–xxx
[10] L. Lin, A new study of the Buddhist Temples in Milan, Archaeol. Cult. Relics 3 (2003) 47–55 (in Chinese). [11] A. Agapiou, D.D. Alexakis, D.G. Hadjimitsis, Spectral sensitivity of ALOS, ASTER, IKONOS, LANDSAT and SPOT satellite imagery intended for the detection of archaeological crop marks, Int. J. Digit. Earth 7 (2014) 351–372. [12] B. Aminzadeh, F. Samani, Identifying the boundaries of the historical site of Persepolis using remote sensing, Remote Sens. Environ. 102 (2006) 52–62. [13] P.A. Brivio, M. Pepe, R. Tomasoni, Multispectral and multiscale remote sensing data for archaeological prospecting in an alpine alluvial plain, J. Cult. Herit. 1 (2000) 155–164. [14] M. Hejcman, Z. Smrˇz, Cropmarks in stands of cereals, legumes and winter rape indicate sub-soil archaeological features in the agricultural landscape of Central Europe, Agric. Ecosyst. Environ. 138 (2010) 348–354. [15] N. Masini, R. Lasaponara, Detection of archaeological crop marks by using satellite QuickBird multispectral imagery, J. Archaeol. Sci. 34 (2007) 214–221. [16] R. Lasaponara, N. Masini, Image enhancement, feature extraction and geospatial analysis in an archaeological perspective, in: R. Lasaponara, N. Masini (Eds.), Satellite Remote Sensing: A New Tool for Archaeology, Springer, Berlin, 2012, pp. 17–63. [17] R. Lasaponara, N. Masini, Beyond modern landscape features: new insights in the archaeological area of Tiwanaku in Bolivia from satellite data, Int. J. Appl. Earth Obs. Geoinf. 26 (2014) 464–471. [18] L. Luo, X.Y. Wang, C.S. Liu, H.D. Guo, X.C. Du, Integrated RS, GIS and GPS approaches to archaeological prospecting in the Hexi Corridor, NW China: a case study of the royal road to ancient Dunhuang, J. Archaeol. Sci. 50 (2014) 178–190. [19] C.T. Morehart, Mapping ancient chinampa landscapes in the Basin of Mexico: a remote sensing and GIS approach, J. Archaeol. Sci. 39 (2012) 2541–2551. [20] M. Noviello, M. Ciminale, V.D. Pasquale, Combined application of pansharpening and enhancement methods to improve archaeological cropmark visibility and identification in QuickBird imagery: two case studies from Apulia, Southern Italy, J. Archaeol. Sci. 40 (2013) 3604–3613. [21] S. Parcak, Satellite Remote Sensing for Archaeology, Routledge, New York, 2009. [22] G.J.J. Verhoeven, Providing an archaeological bird’s-eye view—an overall picture of ground-based means to execute low-altitude aerial photography (LAAP) in archaeology, Archaeol. Prospect. 16 (2009) 233–249. [23] J. Wiseman, F. El-Baz, Remote Sensing in Archaeology, Springer, New York, 2007. [24] B. Figorito, E. Tarantino, Semi-automatic extraction of linear archaeological traces from orthorectified aerial images, Int. J. Appl. Earth Obs. Geoinf. 26 (2014) 458–463. [25] D. Comer, M. Harrower, Mapping Archaeological Landscapes from Space, Springer, New York, 2013. [26] T.G. Garrison, S.D. Houston, C. Golden, T. Inomata, Z. Nelson, J. Munson, Evaluating the use of IKONOS satellite imagery in lowland Maya settlement archaeology, J. Archaeol. Sci. 35 (2008) 2770–2777. [27] J.X. Wu, X. Cheng, H.S. Xiao, H. Wang, L.Z. Yang, E.C. Ellis, Agricultural landscape change in China’s Yangtze Delta, 1942–2002: a case study, Agric. Ecosyst. Environ. 129 (2009) 523–533.
[28] D. Alexakis, A. Sarris, T. Astaras, K. Albanakis, Integrated GIS, remote sensing and geomorphologic approaches for the reconstruction of the landscape habitation of Thessaly during the Neolithic period, J. Archaeol. Sci. 38 (2011) 89–100. [29] D. Evans, C. Pottier, R. Fletcher, S. Hensley, I. Tapley, A. Milne, M. Barbetti, A comprehensive archaeological map of the world’s largest preindustrial settlement complex at Angkor, Cambodia, Proc. Natl. Acad. Sci. U S A 104 (2007) 14277–14282. [30] D. Gallo, M. Ciminale, H. Becker, N. Masini, Remote sensing techniques for constructing a vast Neolithic settlement in Southern Italy, J. Archaeol. Sci. 36 (2009) 43–50. [31] J.D.H. Lambert, A.H. Siemens, J.T. Arnason, Ancient Maya drained field agriculture: its possible application today in the New River Floodplain, Belize C.A, Agric. Ecosyst. Environ. 11 (1984) 67–84. [32] L. Luo, X.Y. Wang, H.D. Guo, J. Liu, H. Cao, L. Wu, C.S. Liu, VHR GeoEye-1 image reveals ancient watery landscape at the Longcheng site, northern Chaohu Lake Basin (China), Int. J. Digit. Earth (2016) (under review). [33] P.J. Ucko, R. Layton, The Archaeology and Anthropology of Landscape: Shaping Your Landscape, Routledge, London, 1999. [34] J.P. Zhang, H.Y. Lü, N.Q. Wu, X.G. Qin, L. Wang, Paleoenvironment and agriculture of ancient Loulan and Milan on the Silk Road, Holocene 23 (2012) 208–217. [35] Z.Q. Chen, Lop Nor and Lop Desert (in Chinese), Acta Geogr. Sin. 3 (1936) 18–49. [36] H.Y. Lü, X.C. Xia, J.Q. Liu, X.G. Qin, F.B. Wang, A. Yidilisi, L.P. Zhou, G.J. Mu, Y.X. Jiao, J.Z. Li, A preliminary study of chronology for a newly-discovered ancient city and five archaeological sites in Lop Nor, China, Chin. Sci. Bull. 55 (2009) 63–71. [37] P.G. Chirico, K.C. Malpeli, S.M. Trimble, Accuracy evaluation of an ASTERDerived Global Digital Elevation Model (GDEM) version 1 and version 2 for two sites in western Africa, Geosci. Remote Sens. 49 (2012) 775–801. [38] R. Lasaponara, N. Masini, Pansharpening techniques to enhance archaeological marks: an overview, in: R. Lasaponara, N. Masini (Eds.), Satellite Remote Sensing: A New Tool for Archaeology, Springer, Berlin, 2012, pp. 87–110. [39] Yuhendra, I. Alimuddin, J.T.S. Sumantyo, H. Kuze, Assessment of pansharpening methods applied to image fusion of remotely sensed multi-band data, Int. J. Appl. Earth Obs. Geoinf. 18 (2012) 165–175. [40] V. De Laet, E. Paulissen, M. Waelkens, Methods for the extraction of archaeological features from very high-resolution Ikonos-2 remote sensing imagery, Hisar (southwest Turkey), J. Archaeol. Sci. 34 (2007) 830–841. [41] T. D’Orazio, F. Palumbo, C. Guaragnell, Archaeological trace extraction by a local directional active contour approach, Pattern Recognit. 45 (2012) 3427–3438. [42] M. Jahjah, C. Ulivieri, Automatic archaeological feature extraction from satellite VHR images, Acta Astronaut. 66 (2010) 1302–1310. [43] R.T.T. Forman, M. Godron, Landscape Ecology, John Wiley & Sons, New York, 1986. [44] L. Luo, X.Y. Wang, H.D. Guo, C.S. Liu, J. Liu, L. Li, X.C. Du, G.Q. Qian, Automated extraction of the archaeological tops of Qanat shafts from VHR Imagery in Google Earth, Remote Sens. 6 (2014) 11956–11976.
Please cite this article in press as: L. Luo, et al., Uncovering the ancient canal-based tuntian agricultural landscape at China’s northwestern frontiers, Journal of Cultural Heritage (2016), http://dx.doi.org/10.1016/j.culher.2016.04.013
ARTICLE IN PRESS
CULHER-3077; No. of Pages 10
Journal of Cultural Heritage xxx (2016) xxx–xxx
Available online at
ScienceDirect www.sciencedirect.com
Uncovering the ancient canal-based tuntian agricultural landscape at China’s northwestern frontiers Lei Luo a,b , Xinyuan Wang a,b,∗ , Jie Liu a,b , Huadong Guo a,b , Rosa Lasaponara c , Wei Ji a , Chuansheng Liu a,b a
Key Laboratory of Digital Earth Science, Institute of Remote Sensing and Digital Earth, Chinese Academy of Sciences, Beijing 100094, China International Centre on Space Technologies for Natural and Cultural Heritage under the Auspices of UNESCO, Beijing 100094, China c Institute of Methodologies for Environmental Analysis (IMAA), CNR, C.da Santa Loja, 85050 Tito Scalo (PZ), Italy b
a r t i c l e
i n f o
Article history: Received 3 February 2016 Accepted 30 April 2016 Available online xxx Keywords: Tuntian Agricultural landscape Canal Remote sensing GIS Archaeological
a b s t r a c t The tuntian system was a state-promoted system of military–agriculture, which originated in the Western Han dynasty (206 BC–9 AD). All the imperial dynasties in Chinese history adopted the practice of tuntian to cultivate and guard frontier areas as an important state policy for developing border areas and consolidating frontier defense. This paper describes the use of satellite remote sensing data to uncover an ancient canal-based tuntian system located in an oasis agricultural landscape adjacent to the ancient Kingdom of Loulan at the southern margin of the Tarim Basin. The remote sensing data examined include Chinese Gaofen-1 (GF-1) VHR imagery, Landsat-8 (LS-8) OLI data and ASTER Global Digital Elevation Model Version 2 (ASTER GDEMV2) products. The effective irrigated tuntian area was estimated to be 2800 ha and the maximum irrigated tuntian area was found to be more than 8000 ha during the area’s most prosperous period. The overall spatial structure of Milan’s tuntian agricultural landscape was explored using the patch–corridor–matrix model. By detailed analysis of satellite remote sensing data, this study reconstructed a 3D view of Milan’s tuntian agricultural landscape in a GIS. © 2016 Published by Elsevier Masson SAS.
1. Research aims The aim of the research presented hereafter is the application of satellite remote sensing and GIS to tuntian system in order to explore the ability of an integrated prospection approach to investigating, mapping and reconstructing cultural landscape and provide new insights for agricultural archaeology in arid area. 2. Introduction During the Han Dynasty (206 BC–220 AD) the ancient Silk Road from Dunhuang to Central Asia [1,2] was extended through the Tarim Basin between the Kunlun Mountains and the Tianshan Mountains (Fig. 1). To establish a military presence along these frontiers and to ease the problem of transporting food grain to these remote areas, the Han government sponsored military–agricultural colonies [1,3,4], known as the tuntian system, in northwest China
∗ Corresponding author at: Key Laboratory of Digital Earth Science, Institute of Remote Sensing and Digital Earth, Chinese Academy of Sciences, Beijing 100094, China. E-mail address: [email protected] (X. Wang).
in order to ensure the safety of traders along the Silk Road. The tuntian system grew out of the Han Dynasty policy of having garrison troops or newly settled peasants bring undeveloped land under cultivation [2]. The Han government first implemented the tuntian system in the Hexi Corridor [1–3], and soon extended it to the empire’s western regions (including today’s Xinjiang and Central Asia) after it proved to be a great success. It is known from Han records written on bamboo slips discovered at Xuanquan in 1990 that ancient Milan was one of the major headquarters of this system [4]. In the second year of the Han Emperor, Zhaodi (77 BC), Weituyan, the King of Shanshan (the state of Loulan) asked the Han government to send troops to develop the wasteland and to plant grain [1]. A sima (minister of war in ancient China) and 40 soldiers were then dispatched to garrison Milan and to farm there. Following the initial success of the Milan tuntian system, the Han government wasted no time in extending it to all of its frontiers; as a result the positive effects of this organized military farming were soon felt all over the ancient Xinjiang, which the Han dynasty unified for the first time [2]. In 1907, the British archaeologist, Stein, first unearthed the Milan archaeological site [5] and discovered a large number of sculptures and frescoes. In the 1950s, a Chinese exploration team discovered an intact system of irrigation canals as well as vast fertile fields buried under the desert [6]. In
http://dx.doi.org/10.1016/j.culher.2016.04.013 1296-2074/© 2016 Published by Elsevier Masson SAS.
Please cite this article in press as: L. Luo, et al., Uncovering the ancient canal-based tuntian agricultural landscape at China’s northwestern frontiers, Journal of Cultural Heritage (2016), http://dx.doi.org/10.1016/j.culher.2016.04.013
G Model CULHER-3077; No. of Pages 10
ARTICLE IN PRESS L. Luo et al. / Journal of Cultural Heritage xxx (2016) xxx–xxx
2
Fig. 1. The location of ancient Milan and the main routes of the ancient Silk Road. Pink lines indicate the Silk Road; the base map is the ETOPO1 Global Relief Model, which can be downloaded from http://maps.ngdc.noaa.gov/viewers/bathymetry.
1973, Xinjiang archaeologists unearthed Milan Castle at the center of the Milan site [6,7]. Early Buddhist sculptures and frescoes excavated from the site show stylistic similarities to the traditions of Central Asia and North India [8] and other artistic aspects of the paintings found there suggest that Milan had a direct connection with ancient Rome [9,10]. Studies that involve the investigation of cultural heritages and archaeological features increasingly employ aerial photographs and satellite remote sensing imagery [11–23]. The combined application of remote sensing and GIS have made it possible to realize agricultural archaeological detecting [24], mapping [19,25] and monitoring [26,27] in landscape investigation, but they’re far from meeting the demands of past landscape reconstructing [28–30]. It’s a great challenge for researchers (e.g. environmentalists, agriculturalists, ecologists, geographers and archaeologists), not only should they have a better understanding of satellite remote sensing and GIS, but they should know the environmental and geographic conditions [31], agricultural and soci-cultural characteristics [32] and past human activities [33]. As one of the most important state-promoted polices in ancient China, tuntian system was viewed as the symbol of arid agricultural development. It is important to try to map the tuntian system so that the defensive and agricultural policies employed along the northwestern frontiers of ancient China can be better understood. Section 3 introduces the materials and methods; the results and discussion are presented in Section 4 and Section 5, respectively and conclusions are drawn in Section 6. 3. Materials and methods 3.1. Archaeological site description Located at the eastern end of the Tarim Basin in Xinjiang Uyghur Autonomous Region, China, Milan was formerly one of the main stops on the ancient Silk Road (Fig. 1). The Milan site is adjacent to the new oasis of Milan town (Fig. 2), which was established by
the Xinjiang Production and Construction Corps (XPCC) after the foundation of the People’s Republic of China in 1949. The Milan site is located approximately 75 km east of Ruoqiang County and 165 km southwest of ancient Loulan (Fig. 2a). The Milan site is now one of the most important archaeological sites in the greater Lop Nor areas [6,7,34–36]. The current conservation area of the Milan site consists of Milan Castle together with cultural relics from the Han–Tang period (206 BC–907 AD), which are scattered around the surrounding areas [35,36], as well as the tuntian canals that were used for irrigating farmland and water conservation when the area was garrisoned during the Han Dynasty (Fig. 2b). Archaeological excavations since the Stein’s expedition have uncovered an extensive Buddhist monastic site that existed between the 2nd to 5th centuries AD [5], as well as Milan Castle (Fig. 2c), which was a Tibetan fort during the 8th and 9th centuries AD [6]. Based on the available historical records [1,2] and archaeological discoveries [5–7,34–36], archaeologists consider that Milan’s tuntian systems were continuously used from the Han Dynasty to Tang Dynasty (618–907 AD). 3.2. Satellite remote sensing data Gaofen (GF) images, acquired by the GF-1 satellite, were obtained from the China Centre for Resource Satellite Data and Applications. GF-1 is equipped with two panchromatic–multispectral (PMS) cameras, which can generate 2 m panchromatic (PAN) data and 8 m multispectral (MS) data (Table 1) across the same 60 km swath. Because Chinese GF-1 data have a very high-resolution, they are ideal sources for mapping the tuntian landscapes found at the Milan archaeological site. The GF-1 data used in this study were acquired on September 18th, 2014 at 5:22:54 UTC. LS-8 Operational Land Imager (OLI) data were acquired from the Institute of Remote Sensing and Digital Earth of the Chinese Academy of Sciences (http://ids.ceode.ac.cn/). This imagery is composed of nine bands covering the visible to the short-wave infrared
Please cite this article in press as: L. Luo, et al., Uncovering the ancient canal-based tuntian agricultural landscape at China’s northwestern frontiers, Journal of Cultural Heritage (2016), http://dx.doi.org/10.1016/j.culher.2016.04.013
G Model
ARTICLE IN PRESS
CULHER-3077; No. of Pages 10
L. Luo et al. / Journal of Cultural Heritage xxx (2016) xxx–xxx
3
Fig. 2. a: locations of Milan archaeological site and Lop Nor areas; b: the conservation area of the Milan site–ancient tuntian canals can be seen in the image; c: image of Milan Castle. The base map is MODIS satellite data saved from ArcGIS10.1; the images used for (b) and (c) are Gaofen-1 (GF-1) panchromatic (PAN) data with a histogram equalization enhancement.
Table 1 Characteristics of the GF-1 PMS and LS-8 OLI data used in this study. Sensors Wavelength/m
Spatial resolution/m
PAN MS
PAN MS
GF-1 PMS
LS-8 OLI
0.45–0.90 Blue 0.45–0.52 Green 0.52–0.59 Red 0.63–0.69 NIR 0.77–0.89
0.500–0.680 Coastal 0.43–0.45 Blue 0.45–0.52 Green 0.53–0.60 Red 0.63–0.68 NIR 0.85–0.89 SWIR 1 1.56–1.66 SWIR 2 2.10–2.30 Cirrus 1.36–1.39 15 30
2 8
GF-1: Gaofen-1; LS-8: PMS: panchromatic–multispectral; Landsat-8; PAN: panchromatic.
parts of the spectrum (Table 1). Each band has a spatial resolution of 30 m. These data also included a panchromatic image with a spatial resolution of 15 m. The LS-8 OLI pansharpened data were used as a supplement in the identification of the dry (buried) channels and trunk canals. In this study, the two LS-8 images used were acquired on October 17th, 2013. A full scene of a LS-8 image has a coverage area of about 185 km by 185 km. The ASTER GDEM, with 30 m grid postings and generated from data collected by the ASTER satellite, has been used to obtain topographic information by many users around the world [37]. ASTER GDEMV2 was released as an upgrade to ASTER GDEMV1 on October 17, 2011. The ASTER GDEM covers land surfaces between 83◦ N and 83◦ S and is composed of 22,600 1◦ by 1◦ tiles. In this study, ASTER GDEMV2 products were used to extract the drainage system of the Milan River by using the GIS hydrological analysis tools, and to produce 3D views by integrating these products with GF-1
and LS-8 OLI data. ASTER GDEMV2 products can be downloaded from the Geospatial Data Cloud (http://www.gscloud.cn/).
4. Methods 4.1.1. Remote sensing data processing The GF-1 and LS-8 images were first atmospherically and geometrically corrected in order to minimize the interference due to signal distortion caused by atmospheric and geometric factors. Data fusion was then carried out in order to combine the very high spatial resolution of the PAN imagery with the high spectral resolution of the MS images [38]. Both the GF-1 and LS-8 images were fused by Gram Schmidt pansharpening method, since it can yield less spectral distortion as compared with other fusion methods [39]. The pansharpened products were only for extracting dry and buried channels of the Milan River. Normalized difference vegetation index (NDVI) is a satellite-derived index form the Near-infrared (NIR) and Red channels, and it was found to be more capable to better enhance crop marks observed for green vegetation [15]. In this study, we discuss the use of GF-1 NDVI data to estimate the spatial extent of the ancient tuntian landscape at Milan. NDVI was expressed as follows: NDVI = (NIR − Red ) /(NIR + Red )
(1)
where NIR and Red were the reflectance of NIR and Red band, respectively. In this area, the NDVI not only reflects the growth of vegetation but also the distribution of paleochannels and springs, which are relatively rich in underground water.
Please cite this article in press as: L. Luo, et al., Uncovering the ancient canal-based tuntian agricultural landscape at China’s northwestern frontiers, Journal of Cultural Heritage (2016), http://dx.doi.org/10.1016/j.culher.2016.04.013
G Model CULHER-3077; No. of Pages 10 4
ARTICLE IN PRESS L. Luo et al. / Journal of Cultural Heritage xxx (2016) xxx–xxx
Fig. 3. Four pansharpened Gaofen-1 (GF-1) MS images and panchromatic (PAN) image for Milan site: a: NIR; b: red; c: green; d: blue; e: PAN image; and f: field photo of tuntian canal.
4.1.2. Extraction of tuntian canals and drainage system The tuntian canals were identified from GF-1 VHR PAN data based on visual interpretation in a GIS. Visual interpretation method has the advantage that the naked eye can identify subtle differences between agricultural archaeological features and the backgrounds that a computer cannot. The automatic feature extraction techniques [24,40–42] are not very successful except at an extremely limited range of spatial scales and spectral contrasts. For instance, in the study by Figorito and Tarantino [24], each image covered only 0.1 km2 and was chosen for the great spectral variability of the crop marks and the marked contrast between them and their surroundings. Based on a ground water model provided in the
hydrological analysis tools under the ArcGIS10.1 environment, the drainage system of the Milan River basin was extracted from ASTER GDEMV2 products showing the mountainous area. 4.1.3. Landscape structure analysis Landscape has been defined as a heterogeneous land area made up of a group of interacting ecosystems. The study on landscape pattern was an important field of land use and ecological change. In this study, the well-known patch–corridor–matrix model [43] was used to analyze the structure of coupling ecosystem found at Milan. Within this model, the landscape is considered to be a mosaic consisting of three major elements: patches, corridors, and matrix.
Please cite this article in press as: L. Luo, et al., Uncovering the ancient canal-based tuntian agricultural landscape at China’s northwestern frontiers, Journal of Cultural Heritage (2016), http://dx.doi.org/10.1016/j.culher.2016.04.013
G Model
ARTICLE IN PRESS
CULHER-3077; No. of Pages 10
L. Luo et al. / Journal of Cultural Heritage xxx (2016) xxx–xxx
5
Fig. 4. Results of the visual extraction of tuntian canals from Gaofen-1 (GF-1) panchromatic (PAN) image.
From an ecological perspective, patches represent relatively discrete areas of relatively homogeneous environmental conditions. Corridors are linear landscape elements that can be defined on the basis of their structure or function. Forman and Godron [43] define corridors as narrow strips of land that differ from the matrices to their two sides. The matrix is the most extensive and most connected landscape element and therefore plays the dominant role in the functioning of the landscape.
5. Results 5.1. Extraction results for tuntian canals Fig. 3a–d displays all four pansharpened GF-1 MS images individually over the Milan site. Despite differences in how each pansharpened image accentuates surface characteristics [23,38], little difference can be seen. However, the tuntian canal traces are clearly visible in GF-1 PAN image due to the obvious textural characteristics (Fig. 3e). Image texture gives us information about the spatial arrangement of color or intensities in an image or selected region of an image. Textural information has been an important factor in visual image interpretation. To the GF-1 PAN imagery, it takes into consideration the distribution and variation of neighborhood pixel values. The Milan’s tuntian canals, which have no distinctive spectral signature, are completely buried by a layer of aeolian sand deposits and gravels (Fig. 3f), which come from the alluvial fan of the Milan River. It is thus very difficult to distinguish the ancient canals from the background using spectral classification. In such situations, the GF-1 PAN image alone can be usefully employed to extract canal traces. The map was made in ArcGIS10.1 by digitally drawing polyline vector features over target canals (as shown on Fig. 4).
5.2. Extraction of drainage system Fig. 5a shows the MS, PAN and pansharpened GF-1 MS images of a buried channel: the pansharpened image allows better discrimination of the dry (buried) channels than the original MS image. Fig. 5b shows the MS, PAN and pansharpened LS-8 OLI images of tuntian canal and channels of the Milan River. These images allow easier visual identification of the canal and river than the original multispectral image; in particular, the spectral characteristics of water bodies are obvious in the pansharpened LS-8 image. Fig. 6 shows the drainage system of the Milan River in the mountainous area as extracted by using hydrological analysis tool in ArcGIS10.1. LS-8 OLI images provided the basis for mapping dried-up river channels in the plain area, where the morphological aspects of the dry channels can be observed exceptionally well due to their spectral signature and textural characteristics [44]. Most of these dry channels were once the main channel of the Milan River but have now disappeared beyond the southern margin of the Taklimakan Desert. 6. Discussion 6.1. Uncovering the canal system in ancient Milan’s tuntian agricultural landscape Based on the hydrological positions and the relationships among different tuntian canals, and by developing Morehart’s [19] typology, the canals in the Milan area can be considered to represent a hierarchical irrigation system that consists of four canal types: in this paper, these are labeled trunk, primary, secondary and tertiary. Fig. 7 shows a map of the canal-based tuntian irrigation system. Only one trunk canal existed. The main body of the trunk canal has disappeared due to shifting sand dunes and wind erosion. The
Please cite this article in press as: L. Luo, et al., Uncovering the ancient canal-based tuntian agricultural landscape at China’s northwestern frontiers, Journal of Cultural Heritage (2016), http://dx.doi.org/10.1016/j.culher.2016.04.013
G Model CULHER-3077; No. of Pages 10 6
ARTICLE IN PRESS L. Luo et al. / Journal of Cultural Heritage xxx (2016) xxx–xxx
Fig. 5. The left, middle and right panels are the original MS images, panchromatic (PAN) images and Gram Schmidt pansharpened images, respectively for (a) the Gaofen-1 (GF-1) data (RGB342 false color composite) and (b) the Landsat-8 (LS-8) data (RGB654 false color composite). Black arrows indicate channels.
pansharpened LS-8 image shows that the traces of the trunk canal lead from the middle reaches of the Milan River at Longkou Village (Fig. 7a). There are seven primary canals leading from the trunk canal through the Milan area. For the convenience of analysis, the primary canals were numbered P1, P2, P3, P4, P5, P6 and P7 from left to right (Fig. 7b). These primary canals form a fan-shaped structure and constitute the basic skeleton of Milan’s canal-based irrigation system. The primary canals, which are analogous to the central arteries of the human body, were used to distribute water from the trunk canal. Secondary canals, which were branches of the primary canals, were built to assist the distribution of water from the primary canals to the hinterland of the ancient Milan area. From a hydrological point of view, the primary and secondary canals constituted a basis irrigation network. The tertiary canals were “feeder” canals [19] that distributed water from the primary canals to the central region of the ancient Milan area. Compared with the primary canals, the tertiary canals are the shorter, narrower canals lying along the same direction that give the tuntian landscape its characteristic appearance. Most of these canals led directly from primary canals, although some were connected to secondary canals. The tertiary canals appear to be
organized in a more regular way; they were usually parallel to other tertiary canals and perpendicular to the higher-level primary and secondary canals. The visual extraction method was able to extract 754 tertiary canals of various sizes. 6.2. Estimating the spatial extent of ancient Milan’s tuntian agricultural landscape Fig. 8a shows an NDVI map of the Milan site. Based on the observed paleochannel patterns and the NDVI values in combination with visual interpretation of the GF-1 data, the greater Milan oasis was manually divided into five parts. These were labeled Area I, Area II, Area III, Area IV and Area V (Fig. 8b). Area I, today’s Milan archaeological site, has barely any vegetation cover. The extent of this site is around 1800 ha. Area II, part of the new Milan oasis, is covered by dense crops with a high NDVI value. Based on the paleochannel patterns and the boundaries of Area I, we speculate that the core area of ancient Milan included both Area I and Area II. This guess was confirmed by both historical Landsat-1 MSS (Fig. 8c) data and Stein’s [5] archaeological map (Fig. 8d). Area II was cultivated by the XPCC in 1970s, as detailed
Please cite this article in press as: L. Luo, et al., Uncovering the ancient canal-based tuntian agricultural landscape at China’s northwestern frontiers, Journal of Cultural Heritage (2016), http://dx.doi.org/10.1016/j.culher.2016.04.013
G Model CULHER-3077; No. of Pages 10
ARTICLE IN PRESS L. Luo et al. / Journal of Cultural Heritage xxx (2016) xxx–xxx
7
Fig. 6. Drainage system in the mountainous area with a flow accumulation threshold of 2500; dry and buried channels were extracted from the pansharpened Gaofen-1 (GF-1) and Landsat-8 (LS-8) MS data.
Fig. 7. Map of Milan’s canal-based tuntian irrigation system: (a) the canal structure in the core area, (b) a fan-shaped complex including trunk, primary, secondary and tertiary canals, photos of modern cement canal (c) and ancient canal (d).
Please cite this article in press as: L. Luo, et al., Uncovering the ancient canal-based tuntian agricultural landscape at China’s northwestern frontiers, Journal of Cultural Heritage (2016), http://dx.doi.org/10.1016/j.culher.2016.04.013
G Model CULHER-3077; No. of Pages 10 8
ARTICLE IN PRESS L. Luo et al. / Journal of Cultural Heritage xxx (2016) xxx–xxx
Fig. 8. a: Gaofen-1 (GF-1) normalized difference vegetation index (NDVI) map of greater Milan oasis depicting the canal system but with the tertiary canals omitted; b: map of greater Milan oasis divided into Areas I to V based on NDVI and paleochannel patterns; the base map is the GF-1 MS data (RGB432); c: Landsat-1 MSS image [acquired on July 1st, 1973 (RGB321)] of the Milan oasis revealing the interesting, oval tuntian agricultural landscape; d: Stein’s archaeological map.
in the chorographical records of the XPCC. The total area of Area I and II is about 2800 ha. Area III not only exhibits good continuity and gradual variation in the NDVI values but also contains archaeological vegetation marks that are clearly related to the tuntian landscape. Thus, it is concluded that Area III was a peripheral region of Milan’s tuntian agricultural landscape. The area of Area III is approximately 5200 ha. Area III combined with Areas I and II forms one region with an oval shape, which completes Milan’s tuntian agricultural landscape. The total area of this oval-shaped region is about 8000 ha. Area IV comprises the new Milan oasis, which has been reclaimed since the 1950s. Due to the sprawling nature of the new Milan oasis, the archaeological conservation area at the Milan site was partly included in the new area. Area V is covered with sparse vegetation and used to be part of the old Milan River. The Milan River once flowed northwards to the Tarim River through Areas IV and V (Fig. 8d) but has, instead, flowed northeastwards to the same river through Areas I, II and III since the flood control dam on the Milan River was constructed in the 1960s. Areas IV and V cover 12,500 ha and 4500 ha, respectively. 6.3. Exploring the spatial patterns in ancient Milan’s tuntian agricultural landscape The ancient tuntian agricultural landscape described in this paper was formed in the middle part (oasis area) of the Milan River, but developments in the upstream (mountainous area) and downstream (desert area) parts of the river jointly decided the eventual fate of the tuntian landscape. The Milan River, the basis of the
tuntian landscape, originates in the Altyn Mountains and is fed by glaciers and snow meltwater from these mountains located to the south on the Qinghai–Tibet Plateau. One of the main reasons why the tuntian landscape at Milan was abandoned was the invasion of the Taklimakan Desert. Thus, it is important to explore the spatial patterns in the tuntian landscape of ancient Milan at a wider scale. Milan’s tuntian agricultural landscape was co-controlled by the Altyn Mountain Ecosystem, the Milan Oasis System and the Taklimakan Desert System, which together constituted a coupled mountain–oasis–desert ecosystem (MODES) (Fig. 9). In this study, there were two relevant matrixes–the Altyn Mountains and the Taklimakan Desert. The Altyn Mountain range constitutes a matrix because it is greatest in areal extent, is mostly connected and exerts a dominant influence on the mountain ecosystem and its ecological processes. The Taklimakan Desert constitutes the second greatest matrix element; it is also mostly connected and exerts a dominant influence on the desert ecosystem and its ecological processes. Proposing the Milan River as a linear corridor supports the existence of the MODES. At a larger scale, the tuntian canal system can be regarded as an extension of the Milan River that also, therefore, belongs to the linear Milan River corridor. At the local scale, each individual tuntian canal is a linear corridor in the irrigated landscape. In Milan’s tuntian landscape, the Milan oasis and the Altyn glaciers are the most important patches in the Taklimakan Desert Matrix and the Altyn Mountain Matrix, respectively. Milan’s tuntian landscape depended on the Milan oasis, especially in ancient times. Although the Milan oasis constitutes a patch, the ecosystem of the
Please cite this article in press as: L. Luo, et al., Uncovering the ancient canal-based tuntian agricultural landscape at China’s northwestern frontiers, Journal of Cultural Heritage (2016), http://dx.doi.org/10.1016/j.culher.2016.04.013
G Model CULHER-3077; No. of Pages 10
ARTICLE IN PRESS L. Luo et al. / Journal of Cultural Heritage xxx (2016) xxx–xxx
9
Fig. 9. 3D view of the Mountain–Oasis–Desert Ecosystem (MODES) around Milan as viewed from the north. In this view, Gaofen-1 (GF-1) panchromatic–multispectral (PMS) (RGB432) and Landsat-8 OLI (RGB654) pansharpened data have been laid over an ASTER GDEMV2 DEM; the drainage system of the Milan River is also shown.
Milan oasis depends on the presence of human activity. The Altyn glaciers, which supplement the fresh water used for irrigation, were also crucial to the existence of Milan’s tuntian landscape. Largescale changes have impacted the spatio-temporal distribution of the glaciers on the Tibet–Qinghai Plateau, as shown by the retreat of the Altyn glaciers and associated rise in the snow line [34], and the occurrence of extreme climate events (blizzards and sandstorms) and disasters (snowmelt flood and drought) in the Milan oasis. The pattern of the Altyn mountain system is complicated and the complicated plateau environment creates abundant biodiversity, frequent hydrothermal exchanges and unique vertical natural zones. The Milan oasis system is a compound artificial one and high productivity, complicated structure and sensitivity constitute its basic characteristics. The Taklimakan Desert system also has some particular characteristics, such as a simple structure, high ecological vulnerability and low productivity. There are great differences between these three systems but the systems are not isolated–MODES is a compound system, and the three systems are connected by the Milan River.
7. Conclusions The 14 C ages for samples from Milan sites fall within the range 1450–1900 cal BP (about 50–500 AD) [36], i.e. a span of ∼ 500 years from the Eastern Han dynasty, through the Wei and Jin, to the Northern and Southern dynasties. The tuntian system was to have far-reaching effects, both for the Han government and for the overall economy and geopolitics of China. Barren land was reclaimed and hundreds of irrigated oases together with dozens of ancient towns were established in the frontier areas. Satellite remote sensing data were used to extract and map Milan’s tuntian system and to uncover traces of an ancient Chinese military–agriculture landscape. Based on the GF-1 extraction results and historical records, the hierarchical irrigation structure of Milan’s tuntian landscape was revealed and the functions of the different types of canal were discussed. It should be pointed out that paleoenvironment changes with time may influence the landscape and these changes may mask the true archaeological features. In this study, the Milan’s tuntian canals are completely buried by a layer of aeolian sand deposits and gravels due to the environmental risks and responses to global changes (drought, desertification and erosion).
The information about the locations of the branches and nodes of canals, the distribution of freshwater and the agricultural output of the tuntian system obtained for the Milan site are important parameters for understanding the ancient agricultural tuntian landscape and the spatial patterns of irrigation canals in the study area, and these aspects need to be studied in more detail in the future. Future work will also incorporate the use of radar and Lidar data to enrich the structural information, including width and height, about the canals that is available. In addition, attempts will be made to extract the ancient canal irrigation systems at other oases (e.g. Loulan, Qiemo and Yutian) in Xinjiang and the spatial characteristics of these canals will be analyzed using integrated GIS tools and field investigations. In particular, knowing the spatial pattern of irrigation canals across these regions assists in understanding the distribution and use of water resources, which is important in the context of the evolution of an irrigated oasis–agricultural civilization and the development of a border defense strategy. Acknowledgements This work was supported by the National Natural Science Foundation of China (No. 41271427), the Directorial Foundation for Doctoral Student Research of the Institute of Remote Sensing and Digital Earth (RADI). Many thanks are due to the anonymous reviewers for their constructive comments and suggestions. References [1] Ban G. 92. HanShu. Zhonghua Book Company, Beijing. (reprint in 1962 [in Chinese]). [2] Sima Q. 91 BC. Shiji. Zhonghua Book Company, Beijing. (reprint in 1962 [in Chinese]). [3] G. Lao, Textual Research and Explanation of the Juyan’s Han Bamboo Slips, Taiwan Press, Taipei, 1984 (in Chinese). [4] D.F. Zhang, Station troops to open up wasteland of Western Region in the Han dynasty based on the Han slips from Xuanquan, Dunhuang Res. 3 (2001) 113–121 (in Chinese). [5] A. Stein, Innermost Asia. Detailed Report of Explorations in Central Asia, Kan-su and Eastern Iran, Carried Out and Described under the Orders of H. M. Indian Government, Clarendon Press, Oxford, 1928. [6] X.C. Xia, B.F. Wang, Y.J. Zhao, Lop Nor in China, Science Press, Beijing, 2007 (in Chinese). [7] S.C. Wang, Historical status of the capital Loulan and ancient Lop Nor region, West Reg. Stud. 4 (1996) 43–53 (in Chinese). [8] H.A. Van Oort, The Iconography of Chinese Buddhism in Traditional China, Vol. 1, Brill Academic Publishers, Leiden, 1986. [9] M.M. Rhie, Early Buddhist art of China and Central Asia, Vol. 1, Brill Academic Publishers, Leiden, 1999.
Please cite this article in press as: L. Luo, et al., Uncovering the ancient canal-based tuntian agricultural landscape at China’s northwestern frontiers, Journal of Cultural Heritage (2016), http://dx.doi.org/10.1016/j.culher.2016.04.013
G Model CULHER-3077; No. of Pages 10 10
ARTICLE IN PRESS L. Luo et al. / Journal of Cultural Heritage xxx (2016) xxx–xxx
[10] L. Lin, A new study of the Buddhist Temples in Milan, Archaeol. Cult. Relics 3 (2003) 47–55 (in Chinese). [11] A. Agapiou, D.D. Alexakis, D.G. Hadjimitsis, Spectral sensitivity of ALOS, ASTER, IKONOS, LANDSAT and SPOT satellite imagery intended for the detection of archaeological crop marks, Int. J. Digit. Earth 7 (2014) 351–372. [12] B. Aminzadeh, F. Samani, Identifying the boundaries of the historical site of Persepolis using remote sensing, Remote Sens. Environ. 102 (2006) 52–62. [13] P.A. Brivio, M. Pepe, R. Tomasoni, Multispectral and multiscale remote sensing data for archaeological prospecting in an alpine alluvial plain, J. Cult. Herit. 1 (2000) 155–164. [14] M. Hejcman, Z. Smrˇz, Cropmarks in stands of cereals, legumes and winter rape indicate sub-soil archaeological features in the agricultural landscape of Central Europe, Agric. Ecosyst. Environ. 138 (2010) 348–354. [15] N. Masini, R. Lasaponara, Detection of archaeological crop marks by using satellite QuickBird multispectral imagery, J. Archaeol. Sci. 34 (2007) 214–221. [16] R. Lasaponara, N. Masini, Image enhancement, feature extraction and geospatial analysis in an archaeological perspective, in: R. Lasaponara, N. Masini (Eds.), Satellite Remote Sensing: A New Tool for Archaeology, Springer, Berlin, 2012, pp. 17–63. [17] R. Lasaponara, N. Masini, Beyond modern landscape features: new insights in the archaeological area of Tiwanaku in Bolivia from satellite data, Int. J. Appl. Earth Obs. Geoinf. 26 (2014) 464–471. [18] L. Luo, X.Y. Wang, C.S. Liu, H.D. Guo, X.C. Du, Integrated RS, GIS and GPS approaches to archaeological prospecting in the Hexi Corridor, NW China: a case study of the royal road to ancient Dunhuang, J. Archaeol. Sci. 50 (2014) 178–190. [19] C.T. Morehart, Mapping ancient chinampa landscapes in the Basin of Mexico: a remote sensing and GIS approach, J. Archaeol. Sci. 39 (2012) 2541–2551. [20] M. Noviello, M. Ciminale, V.D. Pasquale, Combined application of pansharpening and enhancement methods to improve archaeological cropmark visibility and identification in QuickBird imagery: two case studies from Apulia, Southern Italy, J. Archaeol. Sci. 40 (2013) 3604–3613. [21] S. Parcak, Satellite Remote Sensing for Archaeology, Routledge, New York, 2009. [22] G.J.J. Verhoeven, Providing an archaeological bird’s-eye view—an overall picture of ground-based means to execute low-altitude aerial photography (LAAP) in archaeology, Archaeol. Prospect. 16 (2009) 233–249. [23] J. Wiseman, F. El-Baz, Remote Sensing in Archaeology, Springer, New York, 2007. [24] B. Figorito, E. Tarantino, Semi-automatic extraction of linear archaeological traces from orthorectified aerial images, Int. J. Appl. Earth Obs. Geoinf. 26 (2014) 458–463. [25] D. Comer, M. Harrower, Mapping Archaeological Landscapes from Space, Springer, New York, 2013. [26] T.G. Garrison, S.D. Houston, C. Golden, T. Inomata, Z. Nelson, J. Munson, Evaluating the use of IKONOS satellite imagery in lowland Maya settlement archaeology, J. Archaeol. Sci. 35 (2008) 2770–2777. [27] J.X. Wu, X. Cheng, H.S. Xiao, H. Wang, L.Z. Yang, E.C. Ellis, Agricultural landscape change in China’s Yangtze Delta, 1942–2002: a case study, Agric. Ecosyst. Environ. 129 (2009) 523–533.
[28] D. Alexakis, A. Sarris, T. Astaras, K. Albanakis, Integrated GIS, remote sensing and geomorphologic approaches for the reconstruction of the landscape habitation of Thessaly during the Neolithic period, J. Archaeol. Sci. 38 (2011) 89–100. [29] D. Evans, C. Pottier, R. Fletcher, S. Hensley, I. Tapley, A. Milne, M. Barbetti, A comprehensive archaeological map of the world’s largest preindustrial settlement complex at Angkor, Cambodia, Proc. Natl. Acad. Sci. U S A 104 (2007) 14277–14282. [30] D. Gallo, M. Ciminale, H. Becker, N. Masini, Remote sensing techniques for constructing a vast Neolithic settlement in Southern Italy, J. Archaeol. Sci. 36 (2009) 43–50. [31] J.D.H. Lambert, A.H. Siemens, J.T. Arnason, Ancient Maya drained field agriculture: its possible application today in the New River Floodplain, Belize C.A, Agric. Ecosyst. Environ. 11 (1984) 67–84. [32] L. Luo, X.Y. Wang, H.D. Guo, J. Liu, H. Cao, L. Wu, C.S. Liu, VHR GeoEye-1 image reveals ancient watery landscape at the Longcheng site, northern Chaohu Lake Basin (China), Int. J. Digit. Earth (2016) (under review). [33] P.J. Ucko, R. Layton, The Archaeology and Anthropology of Landscape: Shaping Your Landscape, Routledge, London, 1999. [34] J.P. Zhang, H.Y. Lü, N.Q. Wu, X.G. Qin, L. Wang, Paleoenvironment and agriculture of ancient Loulan and Milan on the Silk Road, Holocene 23 (2012) 208–217. [35] Z.Q. Chen, Lop Nor and Lop Desert (in Chinese), Acta Geogr. Sin. 3 (1936) 18–49. [36] H.Y. Lü, X.C. Xia, J.Q. Liu, X.G. Qin, F.B. Wang, A. Yidilisi, L.P. Zhou, G.J. Mu, Y.X. Jiao, J.Z. Li, A preliminary study of chronology for a newly-discovered ancient city and five archaeological sites in Lop Nor, China, Chin. Sci. Bull. 55 (2009) 63–71. [37] P.G. Chirico, K.C. Malpeli, S.M. Trimble, Accuracy evaluation of an ASTERDerived Global Digital Elevation Model (GDEM) version 1 and version 2 for two sites in western Africa, Geosci. Remote Sens. 49 (2012) 775–801. [38] R. Lasaponara, N. Masini, Pansharpening techniques to enhance archaeological marks: an overview, in: R. Lasaponara, N. Masini (Eds.), Satellite Remote Sensing: A New Tool for Archaeology, Springer, Berlin, 2012, pp. 87–110. [39] Yuhendra, I. Alimuddin, J.T.S. Sumantyo, H. Kuze, Assessment of pansharpening methods applied to image fusion of remotely sensed multi-band data, Int. J. Appl. Earth Obs. Geoinf. 18 (2012) 165–175. [40] V. De Laet, E. Paulissen, M. Waelkens, Methods for the extraction of archaeological features from very high-resolution Ikonos-2 remote sensing imagery, Hisar (southwest Turkey), J. Archaeol. Sci. 34 (2007) 830–841. [41] T. D’Orazio, F. Palumbo, C. Guaragnell, Archaeological trace extraction by a local directional active contour approach, Pattern Recognit. 45 (2012) 3427–3438. [42] M. Jahjah, C. Ulivieri, Automatic archaeological feature extraction from satellite VHR images, Acta Astronaut. 66 (2010) 1302–1310. [43] R.T.T. Forman, M. Godron, Landscape Ecology, John Wiley & Sons, New York, 1986. [44] L. Luo, X.Y. Wang, H.D. Guo, C.S. Liu, J. Liu, L. Li, X.C. Du, G.Q. Qian, Automated extraction of the archaeological tops of Qanat shafts from VHR Imagery in Google Earth, Remote Sens. 6 (2014) 11956–11976.
Please cite this article in press as: L. Luo, et al., Uncovering the ancient canal-based tuntian agricultural landscape at China’s northwestern frontiers, Journal of Cultural Heritage (2016), http://dx.doi.org/10.1016/j.culher.2016.04.013