Skip to main content

MAPA: A Linked Open Data Gazetteer of the Southern Babylonian Landscape

Published onDec 09, 2022
MAPA: A Linked Open Data Gazetteer of the Southern Babylonian Landscape

Introduction to the Mesopotamian Almanac Project1

In the year 529 BCE, during the reign of Cambyses, heir to Cyrus the Great of the Teispid dynasty and ruler of the grand Persian empire, a large tithe was collected in southern Mesopotamia (Figure 1).2 While this in itself was not unusual for the period, the small promissory note recording this act is surprisingly descriptive of the geographical horizon of the territory under the control of one of two large tribes in this area. Thanks to this promissory note and another contemporary record, it is possible to reconstruct how one of these tribes controlled much of the pasture lands west of the King’s Canal (Akkadian Nār-Šarri), up until the banks of the Euphrates, and north almost to the city of Babylon. In fact, the extent of the expansive tribal lands east of the King’s canal were rather murky until these two texts were published in 2013 (Beaulieu).

This example shows how little we still know about the historical geography of Babylonia in the first millennium BCE. What we do know is often limited to better documented time frames and areas. Even when the sources are available, historical geography is bound by the fuzziness and interpretative nature of placenames and their contextualization across the heterogeneity of written records: cuneiform sources are sometimes incomplete, placenames and their contexts can be partially or entirely missing, and the identification of placenames with known archaeological features on the ground requires intensive philological and archaeological work. Therefore, it is difficult to reconstruct historical geography without knowledge of the bigger picture—a picture that can only be painted in broad strokes, unless it includes textual data and archaeological surveys side by side.

A first step in this direction is the Mesopotamian Ancient Placenames Almanac (MAPA), which is planned as a global authority resource of Mesopotamian placenames in the Age of Empires (8th–4th centuries BCE). MAPA aims to incorporate both textual and remote-sensing data (archaeological remains identified through satellite and aerial imagery) for large scale relational mapping of the landscape. The core of the textual data is currently made up of a gazetteer of placenames assembled mainly from cuneiform texts, which mention the city of Uruk (biblical Erech, modern Warka) or its hinterland in southern Babylonia. Our data structure has been designed to accommodate the standards of Linked Open Data (LOD), which provides an ontological framework agreed upon by experts, meant to allow for integration of interdisciplinary knowledge through the semantic web (Hyvönen). MAPA is compatible with the ontological framework of the World Historical Gazetteer (WHG), the standards of which are part of the growing initiative to document and link historical geodata (Vitale et al.). In this fashion, MAPA enriches the linked open geodata ecosystem, like Pleiades, with ancient Mesopotamian geography.

Figure 1: Uruk and its location on the Euphrates-Tigris-Karun river delta in southern Iraq, showing palaeo-canal data (courtesy of Jaafar Jotheri), reprinted with permission © 2016 Durham University. Created by Shai Gordin; Design and development by Christoph Forster (Datalino, Berlin).

In this article, we first provide a background on gazetteers and previous work on gazetteers of Mesopotamian placenames specifically. We then present the structure, techniques, and processes used in the assembly of the MAPA gazetteer, the first version of which is now deposited in JSON and TSV files on both GitHub and Zenodo (Gordin and Clark). Two case studies are considered for enriching the new geographical network of the city of Uruk with socio-historical data.

One case study analyzes a network produced from the gazetteer, of the relational proximity of Uruk’s surrounding toponyms (proper names of topographical features) and hydronyms (proper names of bodies of water): villages, fields, canals, and other urban centres. The results are based on the computational analysis of relational data that does not take into account all known connections, but rather gravity and sequence according to the known elements in the network. This already provides a more robust imagined landscape than previous studies done in the 60s and 70s on a chosen subset of the texts.

The second case study is an initial attempt to enrich the gazetteer data with information from ancient documents, in this case tax records which pertain to the movement of people in time and space. The records provide us with distances in days between places that can be visualized as the level of strength between discrete toponyms—the closer they are in days, the closer they will appear in the network. Such a visualization reconstructs the itineraries of Babylonian tax collectors and their implication on the ancient Babylonian economy.

Finally, given that the great majority of ancient placenames recorded in the gazetteer are not currently identified with known archaeological sites or features, we conclude the article with a discussion on the potential for identifying placenames with known archaeological and remote sensing data: palaeo-channels, physical sites, features, and traces in satellite and aerial photography.

Gazetteers and Mesopotamian Historical Geography

The MAPA gazetteer can be placed in the long tradition of accumulating data on geographical locations of the Near East. In fact, the earliest historical examples of lists of places can be traced back to ancient Sumerian texts, such as the Cities List from the early third millennium BCE (Sallaberger 398–400), all the way to the urban description of Babylon in the composition Tintir from the later first millennium BCE (George). Although these are called topographical lists, scholars have long recognised this misnomer. They are arguably better understood as gazetteers with a theological focus, explaining how the temples and cities are related to the pantheon and its cultic topography. Clearly, a gazetteer is a flexible format, which can be used and reused for different purposes.

In its simplest definition, a gazetteer is a knowledge organisation system (KOS), a database of places or placenames including different information about them (Berman et al.; Hill 91–154). The latter part is what makes it not just a toponym list. The information kept could be textual information as well as mathematical, i.e., the georeferencing coordinates of places. In modern gazetteers, these can be used to ask questions and receive useful answers, such as “where is the city X?” “how many malls are in Y?” or “what is the hospital nearest to my current location?” The answers to these questions are driven by the structure chosen for the gazetteer. Thus, a gazetteer should be driven by an ontology constructed in the background.

When choosing the database structure of a gazetteer, one chooses the main points of inquiry, which will be used to compare and contrast between the different toponyms and their features. Accordingly, the research focus of a gazetteer can be geographic, geologic, temporal, or some combination of these or other elements. When moving to historical gazetteers, one can ask historical questions, such as “which places are most frequently attested in the sources?” “Which water sources have the most settlements around them?” or “What are the similarities and dissimilarities between places of the same topographical description?” Furthermore, for gazetteers to remain applicable over the long-term, they should follow certain conventions of syntax and structure, or gazetteer protocols. One of the most effective ways to do that is by maintaining LOD standards.

In this respect, the MAPA gazetteer is the next logical step in a long tradition in the field of Assyriology, of identifying placenames appearing in texts and connecting them with physical features identified by archaeological remains. From the beginning of the field, scholars were inundated by toponyms in cuneiform sources. The nature of such documents usually leaves no explicit information about the actual location of the toponyms, and those must be inferred contextually, preferably from as large a group of texts as possible.

Geographical reconstructions are also vital in order to understand historical processes, a fact which keeps the task of toponym identification a constant desideratum in the field. The project Répertoire Géographique des Textes Cunéiformes (RGTC) aimed to collect every toponym mentioned in cuneiform texts, according to regions and periods (Bagg, “The Geography”). The project led to the publication of the Tübinger Atlas des Vorderen Orients (TAVO), which ended in 1992.

Two more recent projects on ancient Mesopotamian toponyms combine GIS coordinates of ancient sites with ancient placenames in texts, when possible (Boochs et al.; Cancik-Kirschbaum and Hess; Fink; Ziegler and Langlois), and one of them also published a LOD gazetteer (Bruhn). Olof Pedersén collected ancient Near Eastern archaeological sites alongside modern and ancient water systems, like canals, in a KMZ format which can be deployed in google earth (Pedersén et al.; Pedersén). Lastly, the Cuneiform Inscriptions Geographical Site Index (CIGS) collects all the places in which cuneiform tablets were found, including coordinates, in a GeoJSON dataset (Rattenborg et al., “Cuneiform Inscriptions”; Rattenborg et al., “An Open”).

The MAPA Gazetteer: Research Questions and Structure

The MAPA gazetteer contains, at present, toponyms and hydronyms that are textually connected to the city of Uruk, in the Neo-Assyrian, Neo-Babylonian, and Achaemenid periods (ca. 800–330 BCE, with an emphasis on the sixth and fifth centuries BCE). The entries in the gazetteer were recorded and based on volumes 7 and 8 of RGTC, and supporting literature (Bagg, Die Levante, Zentralassyrien, and Babylonien; Ermidoro; Zadok, Geographical Names; Zadok, “New Documents”). Using these systematised collections of Babylonian placenames is necessary because they provide disambiguation of places with the same name, or similar names.

The main purpose of the gazetteer at the moment is to serve as the analytical background to understand the four following research questions:

  1. Understand the naming patterns of ancient place names: which types of locals are named based on nearby topographical features? Which include names of animals or people? Which are named after historical events? And based on this, what is the processes behind the naming patterns?

  2. Investigate the nature of Akkadian terms which describe geographical features: for example, the placename Šaqillatu is qualified in ancient texts by several determinatives, most commonly irrigated farmland or district (Sumerian garim = Akkadian tamirtu), a town (Sumerian uru = Akkadian ālu), and a canal (Sumerian i₇ = Akkadian nāru). This ambiguity leads to problems of geographical reconstruction, in the space and form that the entity takes on the map: a town can be point, a canal can be a line, and a district a polygon. Other designations of placenames are even less clear as to their physical structure: what was the typical size of tribal places named as Bīt-X “house of X,” or small villages named as Ḫuṣṣēti-Y “reed huts/fence of Y?” How large was a ḫarru (“canal, ditch”) as opposed to the typical nāru (river or canal)?

  3. Analyze and visualize the connections between toponyms and hydronyms: the relative spatial proximity and any other connections between different toponyms, such as people, products, etc. attested in more than one place.

  4. Identify, as much as possible, the toponyms in the gazetteer with ground features, i.e., topographical and archaeological features on the ground, and thus provide them with coordinates that can be visualized on modern GIS systems.

Furthermore, we aligned the gazetteer with Linked Open Data standards and protocols. This is vital in order to maintain the applicability of the gazetteer beyond the scope of this project, and enable it to be operable with other gazetteers of the ancient world. The Linked Places format (LPF) of the WHG, adopted by the Pelagios network, require certain fields for each gazetteer entry: “URI,” a persistent link to the toponym; “Site-ID” which is unique for each entry; and “start” the earliest attestation of the toponym. The WHG standard dictates these fields names, as well as their definitions. We also implement two more fields in order to maintain further operability with leading online projects within ancient studies: “CDLI,” the provenance number of the location as kept in the Cuneiform Digital Library Initiative database, and “Pleiades,” the URI of the toponym in the Pleiades project.

In addition to the above-mentioned fields, there are optional fields which we have used as is or with adaptation, in order to answer our four research questions. For the first question, we collect the “title,” the name of the place in academic literature, the “variant_Akkadian” spellings of it, attested in the sources, as well as “variant_Persian,” “variant_Aramaic,” and “variant_Modern” names, when applicable.

For the second question, we collect the “type” of the current toponym, e.g., a hydraulic structure or administrative structure, as well as “sub_types” where applicable, such as canals, provinces, or villages.

For the third question, we keep “Location Notes” to document information on the relationship between the current toponym and the city of Uruk. This is the centre of our inquiry, according to secondary sources which are recorded in “title_source.” In addition, the “Probability” field states the likelihood that the information in “Location Notes” is correct, as described in the same secondary sources. We also maintain “LOC – x” and “Notes” fields, which specify other locations that may be connected to the current toponym. This is only the most basic connection between two places, i.e., that they were mentioned in the same ancient textual source. “Notes” also saves any information that does not fit any of the other fields.

For the fourth question, we keep two fields: “Image(s),” which holds citations to aerial or satellite images where possible physical remains of the toponym are visible, and the “ccode,” which is the two-letter ISO country code of the modern state where the place in question is located.

A continuing issue when collecting and analysing the data gathered in the MAPA gazetteer is the subjective complexity of the data. This is particularly relevant since the gazetteer is currently created from secondary sources, and as such already includes interpretation. One example of this issue is the inclusion of specificity in secondary sources (e.g., the RGTC volumes describe places as on the left bank of a specific canal or a settlement as being in Bit-Amukāni) and not simply that a connection exists between two places. This information is divided and kept in the different fields of the gazetteer, including their levels of certainty. These complications must ultimately be undergirded by textual citations of the primary sources (i.e., geoparsing: Hill 100–03), which will enable the creation of more complex networks and a better understanding of the geographic relationships in and around Uruk (see discussion on geoparsing below).

Figure 2: Assembling sporadic registers into one gazetteer dataset based on the Linked Places Format of the WHG platform (bottom left) in both CSV (top left) and JSON (right) files.

 The gazetteer is currently held in a CSV file, and in a MongoDB document database in JSON format (Figure 2). The gazetteer also exists in a simplified TSV format, without the extraneous fields adapted for our project, in order to align it with the WHG standards (Figure 3). The flexibility of these formats is essential: on the one hand, we maintain the minimal ontology needed for interoperability with established standards, while on the other hand our other formats are flexible enough to organically grow and adapt as we populate the gazetteer with more toponyms, allowing for expressivity without losing basic usability (Freitas et al.)

Figure 3: The MAPA gazetteer in simplified TSV format uploaded to the WHG website

In the next sections, we will bring two examples of how the gazetteer can already be used to answer research questions relating to the connections between different placenames. We will conclude with one of the projects’ upcoming next steps, which is integrating aerial imagery and spatial data into the gazetteer.

A Relational Network of Ancient Placenames

The extent of the city of Uruk is clearly defined by the circumference of its wall, c. 5.5 km2. However, both texts and surface surveys show that during the first millennium BCE the actual size of occupation within the walls was much less dense, fluctuating between c. 50 hectares during Neo-Babylonian times to closer to 300 hectares for the Hellenistic period. The general picture is of a large town with gardens, date orchards, functional production areas near the city wall, and other unused or unbuilt spaces. The geographical horizon of the land around Uruk, based on texts and archaeological surveys of the 1970s, extends as far north as the lands around Marad, northeast to Karkara (Tell Jidr) and east to Larsa and the Sealand marshes. These areas seem to coincide with actual terminus points of three central canals leaving Uruk, according to updated survey maps of ancient canals in the alluvial plain made by Jotheri (Figure 4).

Figure 4: Palaeo-canals around Uruk based on data from Jotheri, main canals entering Uruk are marked in red.

The first attempt at a visualization of Uruk’s hinterland using the gazetteer, was creating a relational network graph of the current known toponyms and hydronyms. The aim of this was to visualize the overall amount of recorded connections between known placenames in the gazetteer, removing previous doubt in scholarship about connections, or lack thereof, between certain places. We used the open-source software Gephi to visualize our network.

Beforehand, the data from the gazetteer had to be preprocessed. Two CSV files are needed, with the following mandatory columns: one for all the nodes, i.e., the placenames, and another for edges, i.e., connections between the nodes. For the primary analysis, all data is extraneous, apart from the “Site_ID” and the associated “LOC” listings, also recorded by their corresponding “Site_ID.”3 The reduced dataset was then transformed into a two-column listing of unique pairs of connected “Site_ID.” For toponyms with multiple connections, this required creating multiple lines, each between the head “Site_ID” and a specific “LOC.” For Example, Nār-Bānītu has four “LOC:” Bēlāya, Babylon, Kish, and the Euphrates. As described in the “Location Notes” field, it is also, of course, connected to Uruk. This listing, which exists in one row of the original gazetteer TSV, is reduced to these six toponyms, and then reduced to five rows, each with two entries (“Site_ID” have been replaced with full toponyms here for ease of understanding):













Table 1: Example of all the connections to the site of Nār-Bānītu

This also provides an opportunity to clean up the dataset, specifically removing so-called “leaf” listings,4 as well as duplicate listings. Duplicate, in this case, refers to a toponym with a “LOC” which, in its own entry, carries a “LOC” to that first toponym. To take the example above, if Bēlāya lists a “LOC” of Nār-Bānītu, it is not required to keep this duplicate link, since the network is not directional. Furthermore, all connections with Uruk were removed: all the entries in the gazetteer are connected to it so keeping it would create a meaningless node. The resulting visualizations can be seen in Figures 5c and 5d, and can be compared to previous visualizations of the landscape in Figures 5a and 5b.

Figure 5a: Cocquerillat’s (143 Pl. 3) reconstruction of Uruk’s hinterland.

Figure 5b: Joannès’s (116 Fig. 19) reconstruction of Uruk’s hinterland.

Figure 5c: Network visualization of the MAPA gazetteer in ForceAtlas2 layout with Lin-Log mode. The network is divided into colour groups based on modularity statistics (Jacomy et al.). The network is not georeferenced, but it was rotated so that in general the cardinal points fit the offered schematic reconstruction.

Figure 5d: Network visualization of the MAPA gazetteer in the Fruchterman layout.

The visualizations created from the gazetteer already provide a more robust and connected imagined landscape than previous studies done in the 60s and 70s on a chosen subset of the texts. This landscape was first reconstructed in the pioneering work of Denise Cocquerillat, based on the 1930s 1:100,000 survey maps of Nöldeke (Heinrich and Falkenstein, fig.18). Revised several times, her thesis was based on geographic and social information extracted mainly from two types of Neo-Babylonian documents: imittu-debt notes (see further below) and administrative cadastral texts. Later studies built upon Cocquerillat’s initial results, to produce a more detailed and updated picture, taking into account archaeological surveys (Adams and Nissen), as well as remote sensing data of aerial and satellite imagery (Figure 5b). The new visualizations, while they are not a map per se, provide another way to imagine the landscape. Although the edges of the graph are not absolute quantifiable distances, they give general regions of relatively close or distant placenames.

Within the larger network of connections, different coloured nodes reflect spatial clusters in the immediate hinterland of the city of Uruk and beyond (Figures 5c and 5d). Even though Uruk was removed, the centre of the network shows the placenames closest to the city, and the colour groups further away from the centre are respectively further away from Uruk. Most of the larger nodes reflect important canals: Nār-Šarri (RNSR, in green), Sumandar (SUM, in purple-blue), Bāb-ḫilti (BHLT, in dark grey), Ḫiltu (HLT), and Nār-Aššurītu (RNAS). These are man-made structures, unlike rivers or streams, which connect towns, villages, fields, and other gazetteer entities across the “landscape” (i.e., the network).5 The central canal, called the Nār-Šarri, or King’s Canal, was the main artery connecting Uruk to the capital city Babylon (BAB, in blue). Therefore, we know it exited to the north of the city, in the direction of Marad and then Babylon. On the banks of this canal, a large number of places were located which are visible in the network as the green group.

Another important group in the network is the purple-blue group. Although it is relatively small, one can see that it includes some of the larger, more central nodes. In general, it seems that this group is made up of central crossroads, linking the different groups together. A good example for this is the central Šaqillatu node (SQL, purple-blue), which seems to sit at the centre of this group, connecting the other nodes. The Šaqillatu is connected to the King’s Canal, which, as said, is the central artery of the network. It is also connected to Bāb-ḫilti (BHLT) and Ḫiltu (HLT), the blue group, which has a central position in the network, when one compares the previous reconstruction offered by Cocquerillat. Through Rudāya (RDJ), Šaqillatu is connected to one of the grey groups, that are further away from the centre. Rudāya is also connected to the Takkīru canal (RTKK, in grey), an important artery south of Uruk, exhibiting substantial agricultural activity. The Takkīru is connected to another important canal, the Ḫarri-kibbi (KHRK, in orange), on the banks of which several other geographical entities can be reconstructed (in orange).

Lastly, the Šaqillatu is connected to Ḫuṣṣēti (HST), which is then connected to the Sumandar or Sumundar canal (SUM), the second largest node in the network. This canal was excavated in the late second millennium BCE, and was used to lead water from the area of the Euphrates to the city of Nippur in the north-east (Abraham and Gabbay). In the network, it is connected to the pink group, which can be understood as the geographical entities related to the tribal group Bīt-Dakūri. These entities, like Bīt-Amukāni (in teal), are in fact territories where tribal groups by that name—whose livelihood depended on agriculture and livestock management (sheep and cattle)—congregated in small settlements. The other placenames connected to them are villages or even houses, or canals linked with these groups, displaying the traces of their settlement patterns in Uruk’s hinterland.

Newly published texts corroborate the findings from our visualizations regarding the location of the groups Bīt-Dakūri and Bīt-Amukāni (see also introduction above). In a recent article, Beaulieu concluded based on these texts that their tribal lands were situated between Babylon and Uruk, but did not extend further east than the city of Nippur. Given the central connection to the Sumandar canal in the network, it would seem that this canal was the eastern border of Bīt-Dakūri, whose lands stretched all the way to Babylon. Similarly, the territories of Bīt-Amukāni in the network are linked to Ḫiltu (HLT, in blue)—very likely another central canal linked north to the area of Babylon and Borsippa.

One can see, then, the power of such visualizations: the gazetteer is built from previous research, yet only when that research was visualized, could this insight be obtained. Such methods then could be a vital tool for the field of cuneiform studies, or any field which deals with missing data, in order to make the most out of the information we have in a heuristic method. This method of distant reading significantly enhances understanding and presentation of landscapes.

Spatial and Temporal Networks

A second stage of the MAPA project, which has been performed on a minimal scale thus far, is geoparsing: linking primary textual sources to placenames documented in the gazetteer. This includes tagging relevant data within the texts in addition to placenames, such as people and commodities mentioned, Babylonian dates and the type of text. This was performed using the Recogito software developed by Pelagios commons, from which one can download the annotated data in various formats (Figure 6). This process already allows for an additional visualization of Uruk’s hinterland, informed by the movements of agents and objects in space and time.

Figure 6: Tagging in Recogito of placenames based on gazetteer entries in an ancient contract from Uruk. On the left-hand side is a modern handcopy of the clay tablet in Babylonian cuneiform and on the right-hand side its corresponding transliteration into Latin characters.

 One type of text produced by temple and state personnel is particularly rich in spatial information: debt notes of agricultural yield estimation, called in Akkadian imittu. Cultivators of the land were managed by contractors, who signed leases with them, designating the share of land under control, their expected yield, and the share they owe the Eanna, the main temple of Uruk, which owns the land. The contractors moved between the fields in the hinterland of the city during the agricultural season and before the harvest, which is done on the seventh month (Akkadian Tašrītu, i.e., September/October). The irrigated fields were usually located on the banks of the canals outside of the city. Together with a permanent team they document for each place the estimated share of the produce to be paid by the cultivator to the general contractor, including any other provisions or additional payments, like tax. The imittu-debt notes issued are a written attestation of the important process of agricultural yield estimation, which allowed the temple administrators to assess the total impost.

The imittu records contain several relevant information points: the estimate of agricultural output, the name of the debtor, the name of the lease contractor, the place of the date garden, the place where the document was written, and the Babylonian date which includes the king’s regnal year, month, and day. Since we know where the rental contractors moved on what day of the month, it is possible to reconstruct the distance between different places along their path. Therefore, using these texts it was possible to try to geographically locate the reconstructed travel routes on these canals. The goal was to create an imaginary geographic network whose nodes are the names of the places the rental contractors visited and whose edges are the distance in days from place to place.

Cocquerillat created her reconstruction of Uruk’s hinterland based on forty imittu texts (Figure 5a). In our corpus, there are currently 179 tagged texts, from the eleventh year of Nebuchadnezzar II (594 BCE) to the first year of Darius I (521 BCE), during which time six different contractors are attested in the corpus. The texts were tagged and collected based on editions and research done most recently by Janković in her unpublished dissertation.

Using Python, the Recogito-created CSV was adapted to preprocess suitable nodes and edges CSV files for Gephi. The Babylonian date formulae were altered to be machine readable: days, months, years, and kings were separated into different columns and changed to integers. Then, iteratively, if two texts had the same contractor, the dates of the debt-notes were subtracted from each other, and an edge relation was created between the places, consisting of the number of days between them. If the two places appear connected by the same contractor in other texts, the resulting days are compared and if a shorter distance in days is attested, it overwrites the previous distance. The network is undirected, therefore, there was no need to differentiate between the source and the target.

This way, the final CSV includes the shortest distance in days attested in the corpus between each two placenames. Furthermore, for placenames that are attested in the same text, a distance of one day was decided, since if more than one place is mentioned, it should be within a walking-day’s distance. Lastly, the strength column had to be created. The strength column is vital to determine and correctly display the level of connectedness and closeness between nodes. The higher the integer in the strength column, the closer the two nodes. The number of days, however, are the other way around—the smaller the number, the closer the two placenames. The highest number of distance in days was taken from the corpus plus one, in our case fifty-eight, and the number of days for each edge connection was subtracted from fifty-eight, creating the strength column. Thus, a distance of one day has the strength of fifty-seven, the highest in our corpus, and two places that have a distance of fifty-seven days have the strength one, the lowest in the corpus.

It is important to note that ancient cuneiform texts are often broken or abraded, and as a result, some of the vital information is missing: such as the date, the contractor, or the field name. In case any of those were completely missing, the text could not be used. If they were partially missing, then they were used as much as possible—i.e., if a placename is partially missing, it could still be used, or if a date formula is missing the day but not the month, the maximum possible distance in days was calculated, in the hopes that it will be replaced with a better attested text. Nevertheless, there are not many big gaps of days in the corpus: out of 225 edges, only twelve have a distance of more than a month. Most records, 103, have a distance of a day, and more than half, 167, have a distance of a week or less. To conclude, our database is large enough so that most edges can be considered the real distance in days between places, despite the limitations of the sources. The resulting network can be seen in Figure 7.

Figure 7: The network of relative distance in days between placenames in Uruk’s hinterland attested in imittu-debt notes. The colours dividing the network are according to the contractors who visited those places. Some places were visited by one contractor, others by several.

Currently the disambiguation of homonyms is performed manually. At first, homonyms are separated into different entities, and in an iterative fashion we examine which homonyms are likely to be the same place based on their spatial proximity on the graph. Thus, this method is not only a result but also a better visual tool for deciding matters of disambiguation, which are a prominent issue when dealing with ancient geographies or prosopographies. Such a method or tool, which combines rich databases and visualization tools, can be a vital resource for an interactive and interpretable method for disambiguation. Although it is outside the scope of this current project, it is certainly a desideratum for historical studies and beyond.

In this network, we can see a conglomeration of the paths along the canals which these contractors used, over a significant period of time, ca. seventy years. It therefore provides a long-term view of the changes in land management policies in Uruk’s hinterland. In the present corpus, the activities of three contractors stand out: (1) Šum-ukin, who was active in the time of Nabonidus (Figure 8a); (2) Ardia, who was active in the time of Cyrus and Cambyses (Figure 8b); and (3) Gimillu, who operated in the first two years of Darius (Figure 8c). Each node in the networks of itineraries is one of two possible types of places: the date grove (red) and the place where the document was written (blue). Sometimes, the document is written in the same place as the date grove, and so it is marked with its own colour (green).

Figure 8a: The network of the contractor Šum-ukin, attested from the reigns of Neriglissar and Nabonidus (ca. 560-549 BCE).

Figure 8b: The network of the contractor Ardia, attested from the reign of Cambyses (ca. 530-523 BCE).

Figure 8c: The network of the contractor Gimillu, attested from the first regnal year of Nebuchadnezzar IV (521 BCE).

 Šum-ukin’s route shows that his work was concentrated along at least two distinct canals. Ardia’s route, however, shows not only walking along a canal, but also administrative centres that become major intersections in the network. Gimillu’s trajectory, the last in this chronological chain, shows an increase in central junctions and fewer trajectories along canals. It can be seen in these cases that the circular motion revolves around the places where the documents are written (blue and green). This can also be seen in the declining incidence of “green” places, i.e., it becomes less common to sign the imittu-document in the date grove itself.

It has been noted that the imittu-documents’ formulae and legal stipulations became standardised during the reigns of Cambyses and Nebuchadnezzar IV (Janković 408), which corresponds to the activities of Ardia and Gimillu, respectively. This evidence coincides with the centralisation efforts evident in the itineraries. It follows that the contractors may have drawn up a draft of the imittu-documents in advance at administrative centres,6 or they had to go to the nearest administrative centre to finalise the assessment.

The standardisation of the documents coupled with the centralisation of agricultural assessment in the hinterland of Uruk, would suggest a higher level of institutional involvement by the temple or state bureaucracy in this process. This can be attributed to a previously unnoticed administrative trend following the rise of Achaemenid rule in Babylonia. It should be stressed that the date industry was one of the main sources of wealth in the Babylonian economy, henceforth a vital resource to control.

Remote Sensing and the MAPA Gazetteer

To complement the detailed image of the land around Uruk gained from textual sources, the project has begun tapping into the growing online archive of satellite imagery. These are becoming standard in current archaeological exploration of landscapes (Parcak). Many studies show the serious potential in using similar interdisciplinary methods with southern Mesopotamian canals and settlement patterns. There have been several works using remote sensing to understand the physical geography, settlement patterns, and topography of this area (Al-Ḥamdani; Hritz, Landscape and Settlement; Jotheri; Pournelle). While all of them analyze satellite and aerial imagery to survey landscape features like archaeological sites, palaeo-canals, and levees, none have employed texts and the placenames therein for their reconstruction.

Earlier surveys made use mostly of aerial photography and lower resolution LANDSAT images, but all projects done from the late 90s make extensive use of declassified CORONA intelligence satellite photographs taken in the 60s and 70s. Their advantage lies in predating most modern industrial development in Iraq, which transformed the landscape, and in being of relatively high ground resolution (c. 2-5m) than most other satellite programs of the time, like LANDSAT or ASTER (Hritz, “Contributions of GIS and Satellite-Based Remote Sensing;” Ur). Surveys from the last two decades also achieve a much better pattern of the changing alluvial landscape and its micro-topography with thousands of previously unidentified and undated sites and palaeo-canals on the ground (Hritz, “Tracing Settlement Patterns and Channel Systems;” Jotheri). The most recent systematic remote sensing survey with dating of ground samples in the area of Uruk was led by a British-Japanese-Iraqi team (Jotheri et al.).

An example of the usefulness of combining textual and remote sensing data is that of the canals leaving the city of Uruk. Texts from Uruk name three locations of quays that were likely to be part of harbour areas: Kār-Eanna, Kār-Nanāya and Kār-Ninurta. In addition to scant information from two texts, suggesting Kār-Ninurta should be located between Uruk and Larsa, to the east of Uruk, additional information can be brought from a CORONA photograph of Uruk and its hinterland, processed and georeferenced by Jason Ur on the Harvard WorldMap portal (Figure 9a). There, it is possible to identify the King’s Canal and the Takkīru canal. However, their course is more of a complicated mosaic of secondary channels than one large clearly identifiable canal.

Nevertheless, immediately to the northern tip of the Urukean wall, one can identify a smaller settlement, maybe a suburb of the city (Figure 9b). It does not only have a darker texture which very likely marks large irrigated (agricultural?) land, but is also criss-crossed with what appear to be roads or secondary ditches. A main channel goes through this site and parallel to it runs a large feature which could be a wide road, since it does not have the usual accompanying levee formation. Thus, identifying this large channel as the King’s Canal makes sense, as it ran parallel to the King’s Highway. We would not hazard to identify the small suburb as Kār-Eanna, but it does suggest that one should be looking for the three quays in places where the large canals leave the city wall. Coincidently, only three places on the CORONA photograph would fit such a description: the one to the north already mentioned, one to the south-west (Takkīru?) and another one to the east heading towards Larsa (Figure 9c). This could be the lower branch of the King’s Canal mentioned in some Urukean texts, as it continues to run parallel to the King’s Highway.

Figure 9a: Uruk CORONA image 1103-1041DA059 (4 May 1968), Center for Advanced Spatial Technologies, University of Arkansas/US Geological Survey.

Figure 9b: View of Uruk’s northern city wall marked with arrow. Three palaeo-canals entering at one point inside the wall are marked with hatched white lines. Possible remains of a settlement visible immediately outside the wall.

Figure 9c: CORONA view of Uruk’s south and east wall sections, where palaeo-canals emerge from the city on their path south and east, respectively (in blue).

This small case study is a prime example to show the potential of combining textual and remote sensing information. It is now further possible to train machine learning models to identify ground features in satellite images, such as palaeo-canals, levees, sites, or hollow ways (i.e., remains of ancient tracks or roads; Altaweel et al.). Once they are identified, these can be used in combination with the relative geographical networks shown above, and thus provide possible identifications of more placenames with physical features on the ground.

Summary and Conclusions

The MAPA project has produced the first LOD driven gazetteer of placenames in and around first millennium BCE Uruk, enriching the growing family of ancient gazetteers online. This gazetteer can already be visualized in different ways, providing a robust reconstruction of Uruk’s hinterland based on the latest scholarship. Further information can be gathered to enrich the gazetteer from primary sources. This provides a more fine-tuned reconstruction of not only Uruk’s hinterland, but also the administrative and economic functions of this highly productive agricultural land, and new interpretative methodologies for distantly reading ancient cuneiform sources.


Works Cited

Abraham, Kathleen, and Uri Gabbay. “Kaštiliašu and the Sumundar Canal: A New Middle Babylonian Royal Inscription.” Zeitschrift für Assyriologie und Vorderasiatische Archäologie, vol. 103, no. 2, Jan. 2013, pp. 183–195.

Adams, Robert McC, and Hans Jörg Nissen. The Uruk Countryside: The Natural Setting of Urban Societies. University of Chicago Press, 1972.

Al-Ḥamdani, Abdulameer M. Shadow States: The Archaeology of Power in the Marshes of Southern Mesopotamia. Stony Brook University, 2015.

Altaweel, Mark, Adel Khelifi, Zehao Li, Andrea Squitieri, Tasnim Basmaji, and Mohammed Ghazal. “Automated Archaeological Feature Detection Using Deep Learning on Optical UAV Imagery: Preliminary Results.” Remote Sensing, vol. 14, no. 3, 2022, p. 553.

Bagg, Ariel M. Répertoire géographique des textes cunéiformes. VII/1 Teil 1, Die Orts-und Gewässernamen der neuassyrischen Zeit: Die Levante. Reichert, 2007.

Bagg, Ariel M. Répertoire géographique des textes cunéiformes. 7.2.1/2: Die Orts- und Gewässernamen der neuassyrischen Zeit: Zentralassyrien und benachbarte Gebiete, Ägypten und die arabische Halbinsel. Reichert, 2017.

Bagg, Ariel M. Répertoire géographique des textes cunéiformes. 7.3.1/2: Die Orts- und Gewässernamen der neuassyrischen Zeit: Babylonien, Urartu und die östlichen Gebiete. Reichert, 2020.

Bagg, Ariel M. “The Geography of the Neo-Assyrian Empire.” The Ancient Near East Today, 2021,

Beaulieu, Paul-Alain. “Arameans, Chaldeans, and Arabs in Cuneiform Sources from the Late Babylonian Period.” Arameans, Chaldeans, and Arabs in Babylonia and Palestine in the First Millennium B.C, edited by Angelika Berlejung and Michael P. Streck, Harrassowitz Verlag, 2013, pp. 31–55.

Berman, Merrick L., Ruth Mostern, and Humphrey Southall, editors. Placing Names: Enriching and Integrating Gazetteers. Indiana University Press, 2016.

Boochs, Frank, Kai-Christian Bruhn, Christophe Cruz, Ashish Karmacharya, and Tobias Kohr. “HiGeoMes: Distributed Geodatabases in an Archaeological Joint Research Project.” CAA Proceeding, 2013.

Bruhn, Kai-Christian. “TEXTELSEM mapping data package.” 10 May 2018. Zenodo,

Cancik-Kirschbaum, Eva, and Christian Hess. Toponyme der mittelassyrischen Texte: Der Westen des mittelassyrischen Reiches: Obermesopotamien im 2. Jt. v.Chr. – MTT I/2. Collège de France, 2017. OpenEdition Books,

Cocquerillat, Denise. Palmeraies et cultures de l’Eanna d’Uruk (559-520). Mann, 1968.

Ermidoro, Stefania. “New Data on the Babylonian Hydraulic Landscape: An Update to the Répertoire Géographique des Textes Cunéiformes, Vol. 8.” Kaskal, vol. 13, 2016, pp. 135–174.

Fink, Christoph. Fundorte und Karten: Obermesopotamien im 2. Jt. v.Chr. – MTT I/3. Collège de France, 2017. OpenEdition Books,

Freitas, Andre, Edward Curry, João Oliveira, and Sean O'Riain. “Querying Heterogeneous Datasets on the Linked Data Web: Challenges, Approaches, and Trends.” IEEE Internet Computing, vol. 16, 2012, pp. 24–33.

George, Andrew R. Babylonian Topographical Texts. Peeters, 1992.

Gordin, Shai, and Samuel Clark. “DHALab/MAPA: MAPA Gazetteer.” Zenodo, Feb. 2021,

Heinrich, Ernst, and Adam Falkenstein. “Forschungen in der Umgebung von Warka.” Neunter vorläufiger Bericht über die von der Deutschen Forschungsgemeinschaft in Uruk-Warka unternommenen Ausgrabungen, edited by A. Nöldeke et al., Mann, 1938.

Hill, Linda L. Georeferencing: The Geographic Associations of Information. MIT Press, 2006.

Hritz, Carrie. “Contributions of GIS and Satellite-Based Remote Sensing to Landscape Archaeology in the Middle East.” Journal of Archaeological Research, vol. 22, no. 3, Sept. 2014, pp. 229–176. Springer Link,

Hritz, Carrie. Landscape and Settlement in Southern Mesopotamia: A Geo-Archaeological Analysis. University of Chicago, 2005.

Hritz, Carrie. “Tracing Settlement Patterns and Channel Systems in Southern Mesopotamia Using Remote Sensing.” Journal of Field Archaeology, vol. 35, no. 2, June 2010, pp. 184–203. Taylor and Francis Online,

Hyvönen, Eero. Publishing and Using Cultural Heritage Linked Data on the Semantic Web. Morgan & Claypool, 2012.

Jacomy, Mathieu, Tommaso Venturini, Sebastien Heymann, and Mathieu Bastian. “ForceAtlas2, a Continuous Graph Layout Algorithm for Handy Network Visualization Designed for the Gephi Software.” PLOS ONE, vol. 9, no. 6, June 2014, p. e98679. PLoS Journals,

Janković, Bojana. Aspects of Urukean Agriculture in the First Millennium BC. University of Vienna, 2013.

Joannès, Francis. Textes économiques de la Babylonie récente. Editions Recherche sur les civilisations, 1982.

Jotheri, Jaafar. Holocene Avulsion History of the Euphrates and Tigris Rivers in the Mesopotamian Floodplain. Durham University, 2016.

Jotheri, Jaafar, Mark Altaweel, Akihiro Tuji, Ryo Anma, Benjamin Pennington, Stephanie Rost, and Chikako Watanabe. “Holocene Fluvial and Anthropogenic Processes in the Region of Uruk in Southern Mesopotamia.” Quaternary International, vol. 483, 2018, pp. 57–69.

Parcak, Sarah H. Satellite Remote Sensing for Archaeology. Routledge, 2009.

Pedersén, Olof, et al. “Cities and Urban Landscapes in the Ancient Near East and Egypt with Special Focus on the City of Babylon.” The Urban Mind. Cultural and Environmental Dynamics, edited by Paul J. J. Sinclair et al., Uppsala University, 2010, pp. 113–148.

Pedersén, Olof. “Waters at Babylon.” Water and Urbanization, edited by Terje Tvedt and Terje Oestigaard, I.B. Tauris, 2014, pp. 107–129.

Pirngruber, Reinhard. “A Diplomatics Approach to the Eanna Archive: the Livestock Dossier.” Archiv für Orientforschung, vol. 54, 2021, pp. 88–108.

Pournelle, Jennifer. R. Marshland of Cities: Deltaic Landscapes and the Evolution of Early Mesopotamian Civilization. University of California, San Diego, 2003.

Rattenborg, Rune, et al. “An Open Access Index for the Geographical Distribution of the Cuneiform Corpus.” Cuneiform Digital Library Journal, vol. 2021, no. 1, Nov. 2021.

Rattenborg, Rune, et al. “Cuneiform Inscriptions Geographical Site Index (CIGS).” 1.4, 1 Nov. 2021. Zenodo,

Röllig, Wolfgang “Aims and Organization of the ‘Répertoire Geographique Des Textes Cunéiformes’ and Historical Geography.” Sumer, vol. 42, 1986, pp. 40–43.

Sallaberger, Walther. “Zur Genese der mesopotamischen Götterwelt: Eine Auseinandersetzung mit Thorkild Jacobsens.” Mu-u an-za-še₃ kur-ur₂-še₃ ḫe₂-g̃al₂: Altorientalistische Studien zu Ehren von Konrad Volk, edited by Jessica Baldwin and Jana Matuszak, 2020, pp. 391–412.

Ur, Jason A. “CORONA Satellite Imagery and Ancient Near Eastern Landscapes.” Mapping Archaeological Landscapes from Space, edited by Douglas C. Comer and Michael J. Harrower, Springer, 2013, pp. 21–31. Springer Link,

Vitale, Valeria, et al. “Pelagios – Connecting Histories of Place. Part I: Methods and Tools.” International Journal of Humanities and Arts Computing, vol. 15, no. 1-2, Edinburgh University Press, Oct. 2021, pp. 5–32. Edinburgh University Press Journals,

Zadok, Ran. Geographical Names According to New- and Late-Babylonian Texts. Reichert, 1985.

Zadok, Ran. “New Documents about Uruk, Its Countryside and the Sealand.” Des polythéismes aux monothéismes: mélanges d’assyriologie offerts à Marcel Sigrist, edited by Uri Gabbay and Jacques Pérennès, Peeters, 2020, pp. 491–534.

Ziegler, Nele, and Anne-Isabelle Langlois. Les toponymes paléo-babyloniens de la Haute-Mésopotamie: La Haute-Mésopotamie au Iie millénaire av. J.-C. – MTT I/1. Collège de France, 2017. OpenEdition Books,

No comments here
Why not start the discussion?