Previous pageTable of ContentsNext Page

Interface --

Archaeology and Technology

Real-Time GIS Construction and Digital Data Recording of the Jiskairumoko Excavation, Peru

Nathan Craig


Increasingly, researchers are turning to GIS databases for organizing, analyzing, and sharing the products of their field research (M. Aldenderfer and H. Maschner, editors, 1996, Anthropology, Space, and Geographic Information Systems. Oxford University Press, New York). Two general patterns to the process of increasing GIS use involving the scale and timing of database construction are identifiable. As I will show, current practice in GIS implemention has some serious shortcomings that can be overcome. The project described hereexcavations at Jiskairumoko, a site in southern Peruillustrate an attempt to expand upon and develop new ways to collect archaeological digital spatial data.

Archaeological data is collected at two basic scales: survey and excavation. Construction of archaeological GIS datasets has taken place almost entirely at the scale of survey data. However, a principle advantage of GIS technology is that it is a scaleless spatial infrastructure (M. Aldenderfer, 1996, Introduction. In Anthropology, Space, and Geographic Information Systems. M. Aldenderfer and H. Maschner (editors); M. Goodchild, and S. Gopal, editors, The Accuracy of Spatial Databases. Taylor and Francis, New York). Given this technological fact, there is no theoretical reason why GIS datasets of excavation data could not be just as common as survey scale datasets have become.

Not only does the construction of archaeological GIS databases occur primarily at the scale of survey information, generally the timing of database construction is well after the field season is completed. In most archaeological cases, the computer database is created through the transformation of paper records into digital form. The paper to digital transformation takes place almost entirely through some combination of typing, scanning, and digitizing. Conceivably all these operations could take place in the field. In fact within the discipline of geography the collection of digital datasets directly in the field, so-called "real-time GIS" is becoming common.

Working from real-time data collection models in geography, Aldenderfer and I developed a project that proposed collecting GIS data from excavation directly in the field with no intervening paper to digital data transformation. We aimed to place data directly into the analytical context, reducing data entry time, and eliminating transcription errors that occur in the paper to digital transformation.

Implementing Infield Digital Data Collection of Excavation

The project began with the simple desire of streamlining GIS data collection. We selected the Environmental Systems Research Institute's (ESRI) ArcView 3.1 GIS software package for its ease of use, analytical power, and flexibility (ESRI, 1998, ArcView 3.1 for Windows). Through an equipment grant from the National Science Foundation, Aldenderfer obtained five Fujitsu pentop computers and two Nikon digital cameras to be used for GIS data entry during excavation of selected Late Period Archaic sites in the Rio Ilave region, which is found in the southwestern Lake Titicaca basin in southern Peru.

The key to this project is the current availability of computing technology capable of the tasks demanded of it. This project could not have been done in its present form even last year. The most critical piece of the technological puzzle is the pen computer. Although pen computers have been around for a number of years, all, until very recently, lacked a crucial feature: A sensitive screen upon which one could draw with accuracy and that could be seen under typical field conditionsi.e., in bright, indirect sunlight. So-called passive screens on pen computers, which currently dominate the market, can record images, drawings, and doodles, but they cannot render images accurately. Further, none of them have display properties that allow them to be viewed outdoors. In fact, most pen computers have been developed for the factory floor, office, and interior of vehicles.

All of this changed with the introduction of the Fujitsu Stylistic 2300. The pentop, which runs Windows 9x and NT, has a color transreflective screen that provides good, but not perfect, outdoor visibility, but more importantly, has an electromagnetic pen digitizer that moves the cursor across the screen by floating (or touching) the digitizer over it. It has extremely high resolution, at least as good and most likely better than pen and paper. In fact, it is just like drawing on paper, and depending on the software used, one can thin or thicken line width and change a whole series of other visual attributes of the drawing process.

It also is important to stress that the 2300 has a processor (in this case an Intel Pentium MMX running at 233MHz) capable of running ArcView

efficiently, a large-capacity hard drive, and a number of PC card slots. Keyboards can be hot swapped, and it comes loaded with very high-quality handwriting recognition software (which ArcView, alas, does not presently support). In short, it is a pen computer capable of performing all of the tasks demanded of it.

The second piece of the puzzle is the digital camera. For this project, we used the Nikon Coolpix 950. John Rick, in a previous issue of the SAA Bulletin, [1999, 17(3): 37] described his field use of an earlier Nikon model, the 700; it performed splendidly. The 950, in our experience, performed even more capably. Although we could have done on screen digitizing without the digital camera, it would have been very time consuming, and most likely, the process would have defeated the essential goal of the project, which was to produce a real-time archaeological GIS at least as easy to use as paper procedures.

Fitting Units into the Larger Spatial Context of an Excavation

Prior to initiating fieldwork we realized that, at a very fundamental level, spatial aspects of the excavation database needed to be designed around a matrix. Aldenderfer generally excavates sites in blocks of units removing natural layers using the decapage technique to expose large living surfaces (M. Aldenderfer, 1998, Montane Foragers: Asana and the South-Central Andean Archaic. University of Iowa Press, Iowa City). Units are coded according to an XY-coordinate system forming a matrix that covers the extent of the site.

Four minimum provenience cells of 50 cm are maintained within each unit throughout excavation. These sub-unit provenience cells are maintained so that changes in artifact density can be examined at finer resolutions, permitting more detailed interpretations of prehistoric excavated surface functions. The minimum provenience cells were used for the basis for constructing a matrix in the GIS covering the entire spatial extent of the area to be excavated (Figure 1).

The excavation matrix was constructed in the GIS with an ArcView extension, GridMaker.avx (available at All excavation units are constructed from the matrix of minimum provenience units. Provenience cells making up a unit are selected and copied to a new thematic layer. The matrix is a useful division and organization of space to replicate within the GIS. Not only does the matrix replicate a method common to many excavations, but it also is extremely useful for reconstruction of the entire site within the computer when the data are collected in the field on a unit by unit basis. Since, all of the site's units are generated by sub-setting cells of a matrix covering the entire extent of potential excavation, when reconstructing whole excavated surfaces within the GIS from data collected at the level of the unit, excavation units can match each other perfectly sharing nodes and chains (K. Clarke, 1995, Analytical and Computer Cartography. Prentice Hall, New Jersey; See Figure 2).

Figure 1 - Showing selected cells from the site excavation matrix. The table cells shown above correspond to labeled squares in the spatial data layer shown below the table. Note that every square has a corresponding record in the table and that each square is 50 cm on a side. Thus four 50-cm squares are required to construct a 1-m unit.

Adjacent units N10 and N11 of cultural level ii not sharing nodes and chains. Note overlapping polygon.

Adjacent units N10 and N11 of cultural level ii sharing nodes and chains. Note there is no overlapping polygon creating a perfect fit between adjacent units.

Figure 2 - Shows the importance of sharing nodes and chains between adjacent vector shapes in the GIS and illustrates the utility of constructing units from template files. In the example above adjacent unit polygons do not share nodes and chains and a third "in between" polygon that exists in the area of overlap. In the GIS this area may be extremely small, but could have a negative effect when reconstructing larger excavated surfaces. In the example below, unit polygons are constructed based on a template file and both polygons share connected nodes and chains. This produces a perfect fit between the two polygons and simplifies the reconstruction of larger excavated surfaces.

Digitizing each of the unit boundaries would be time consuming and is not something necessary to do in the field. Field digitizing is best directed toward recording archaeological information rather than toward database construction. By constructing each of the individual excavation units from sub-set copies of the site matrix, all units match perfectly with their eight adjacent neighbors. Using selected portions of the matrix as a template for creation of excavation units automates the creation of unit boundaries and ensures that each unit fits perfectly into the larger context of the excavation.

Automating the Creation of Attribute Fields

Specific kinds of information must be collected for each level of a unit and for each feature in a particular level of a unit. Level and feature forms generally fill at least an entire photocopied paper page. For collection of real-time GIS data of excavations, many attribute fields are necessary for sufficient description of the deposit. Creating fields from scratch for each new unit and feature theme would be a time-consuming process resulting in the wasting of valuable field time. Fieldwork is best devoted to recording information about the deposit rather than database creation. Since all units are created by sub-setting from a generic matrix covering the site, attaching all necessary fields to this unit matrix solves the problem of needing to reconstruct fields for each new theme. The generic matrix, like a paper form, contains all of the necessary fields prepackaged and ready to go as needed.

Keeping Level Data Flexible when Excavating in Natural Levels

Excavation in natural levels permits many kinds of spatial analysis difficult to achieve with only the use of arbitrary levels. When digging in natural levels, uneven living surfaces can be exposed and examined as a whole rather than restricted within some arbitrary vertical division of space. However, natural levels often are discovered during the process of excavation. Previous work in the region showed that hunter-gatherer Archaic archaeology in the region often consists of thin difficult-to-detect palimpsest layers (Aldenderfer 1998). As we excavated the site, we quickly found that not all levels are of the same depth. Given the fact that some levels did not extend across a block but were only restricted to portions of a block, nomenclature needed to remain flexible. Keeping level information consistent, but maintaining flexibility in the number and depth of levels, also was possible due to the use of template files. Even if new levels are discovered within in a unit, a new level can easily be constructed from the template file.

How to Rapidly Collect Unit Feature Data into GIS

To be useful, real-time digital data collection must not be slower than conventional paper recording. If the technique is going to be useful, there cannot be significant bottlenecking during data entry. Furthermore, data collection must be at least as accurate as it is in traditional paper methods. Increases in speed and accuracy are desirable.

Using paper records, accurately recording the extent of unit features can be quite a time-consuming task during fieldwork. Features are generally identified by the context of artifacts or often from differences in soil color and texture. Once identified, feature recording onto a paper form begins by measuring a set of bounding points on the perimeter of the object in the unit. Once taken, these measurements are transformed into the scale of the paper form, the bounding points are plotted onto graph paper, and curves are fit to these bounding points to draw the feature.

Nearly all excavators record feature and unit objects in this way, and it is possible to replicate this method of data collection in real time using nearly any GIS software. Measurements can be taken in the unit and a GIS measurement tool can be used to replicate the plotting of points in the computer. Once the set of bounding points has been created within the computer, the feature can be digitized into its correct spatial context.

Since data collection for this project is built around a GIS structure, new methods of data collection are available to the field-worker. Given the tools provided by a GIS, we found that taking vertical photographs over excavated units permits extremely rapid and accurate digital reproduction of unit contents. ArcView 3.1 supports the display and analysis of raster as well as vector spatial data. Photographs taken with the Nikon digital camera are saved as JPEG image data. With the proper extension, ArcView3.1 can display and convert JPEG image data into the native ESRI Grid format. Thus photographs taken with the Nikon digital camera could be displayed, georeferenced, and used as a background image for on-screen and in-field digitizing within the GIS.

The digital photographic data from the Nikon camera are in the image's coordinate system based on pixel addresses. This coordinate system does not relate to a coordinate system that a researcher would want to maintain for a GIS of excavation. However, once a JPEG is converted to ESRI's native Grid data format, image data can be georeferenced within the GIS. Georeferencing of JPEG image data is common to digital orthophotogrammetry and GIS-based remote sensing studies. From the scaleless perspective of the GIS, georeferencing digital photographs taken over an excavated unit is no different than rubbersheeting air photos taken over some identifiable parcel of land. When known points are observable in image data and also are plotted within the computer, a set of "from" and "to" links can be constructed in order to resample the raster data into the coordinate system of the spatial database.

All images taken of levels in units were shot at 1600 x 1200 pixels with the unit taking the largest extent possible in the field of view. We took two sets of digital photographs for each level of each unit. Unit corners were clearly marked by tape measures in all the photographs. The first photograph is taken of the unit without any etching of features in the soil. Once units were recorded without marks, then feature boundaries were etched into the top of the unit. Color and texture were carefully examined, and when features were thought to be present, feature edges were etched in the soil. Feature etching is an important aspect of Aldenderfer's excavation methods because features are excavated and pro-venienced separately. Once etchings were made to the top of a level, the second set of photographs was taken.

Photographs were then transferred from the digital camera to the computer and added the GIS database. Etched unit photographs were georeferenced to their proper unit cells in the computer. Corners marked in the photograph could be used for constructing a set of links, as could the corners of the unit in the site matrix. The presence of four links permits a first-order warp of Grid data in ArcView. Once the image was rubbersheeted to the unit, features etched in the dirt were visible on screen, in the co-ordinate system of the GIS, and could therefore be easily digitized on screen (Figure 3). Additional objects resting on top of excavated surfaces (things like projectile points, bones, fire cracked rock, manos, and other objects important to record) also were easily visible in the photographs and could be digitized on screen. Once a photograph was rubbersheeted to the correct unit, recording of complex associations of objects on the surface was quick and easy.

The digital photographs are a consistent spatial sample of the space within the camera's field of view. The photographs were taken with an eight bit byte at 1600 x 1200. Each pixel represents a measurement quantitized to 0­255, meaning that 1,920,000 measurements were made with each individual photograph taken of a unit. Once photographs are georeferenced to their corresponding unit corners in the GIS all of those nearly two million eight bit measurements are spatially accurate. Using this method many more measurements were taken for an individual unit than would have been possible using conventional strategies. Every space within the unit is consistently sampled according to its variation in the reflectance of visible light. Those measurements that define an object of interest are converted to a vector format through the process of on screen digitizing.

1 Meter
A. Unit photograph rubbersheeted to its correct spatial context



Groundstone Fragment

Late Archaic Projectile Point



Bone 1: Articulates with Bone 2

Bone 2: Articulates with Bone 1


Firecracked Rock

1 Meter
B. Detail of lower right quad showing artifacts resting on the surface of cultural level iiib.

1 Meter
C. Unit photograph showing surface artifacts digitized into the GIS database.

Figure 3 -- Showing the multi-scale nature of the GIS database and illustrating the utility of high-resolution digital cameras for recording excavated surfaces. The digitized entities were drawn to their correct scale on the screen of the Fijitsu pentop computers into ESRI's ArcView 3.1 GIS database using 1600 x 1200 pixel photographs taken with a Nikon digital camera.

All geographic features were digitized into a unit and level specific copy of the archaeological feature template theme which is quite similar to the unit template theme. The feature template theme is similar in that it is a vector layer containing all of the attribute fields necessary for recording an archaeological feature, just as the unit template is a vector theme with all of the fields necessary for recording unit level information. The feature template is different in that it contains no shapes and the attribute table contains no records, whereas the unit template consists of all possible provenience squares and a record for each in the theme attribute table. For each new unit, archaeological features and artifacts resting on the top of the level are digitized into the units feature theme. Digitizing shapes produces a corresponding record in the theme's attribute table. That record can then be given the necessary attribute information by adding it to the table.

Rubbersheeting photographs is time consuming, but the benefits gained in terms of accuracy and speed of feature recording are quite significant. Photographing, downloading, rubbersheeting, and digitizing of a given level of a unit could be completed in about 45 minutes. Since many steps could be automated, done in batches, and completed in the lab before going into the field, field recording efficiency was quite high. The most efficient routine we developed involved excavating what could be done during a day's work. Then at the end of the day, photographs of the newly exposed surfaces were taken, both etched and not etched. In the lab, either at the end of the day or in the morning before returning to the field, photographs of newly excavated units were downloaded into the computer, converted to grids, and then rubbersheeted. Some digitizing, particularly of large and obvious entities, could be done in the lab before going into the field. Thus we returned to the field each day prepared to make only those observations (drawings and attributes) that needed to be made while at the site.

Objects that need some record associated with them have to be digitized on screen. Some observations necessary for digitizing can only be made in the field. Telling the difference between a mano and a river rock, or telling whether a rock is fire cracked or naturally angular, cannot be done from unit photographs alone. Decisions about digitizing these objects are best made while at the site. Associations and precise relationships of superposition also are best digitized in the field when standing next to the unit look-ing at the photograph zoomed in close. During our first season, prep work for unit creation and photo georeferencing could be done while not in the field. Field data collection could then focus on taking elevation measurements, Munsell colors, and recording fea-tures that were difficult to discriminate from the photographs.

Tracking the Completeness of Digital Unit Data during Excavation

Keeping track of paper excavation records is difficult. Keeping track of digital records is no easier. Careful attention to developing an organized file structure and a consistently applied naming convention are essential to the success of any GIS project. Even with paper records excavators may find that some unit information has not been included in the paperwork. That is no less likely to be the case with digital records. Filling out electronic records in the field does not solve problems of ensuring that all of the attribute fields are properly entered during the excavation process. Pulling up and examining unit and feature records to check if they are complete can consume time in the field.

During excavation many units may be open within a block and keeping track of each unit's status can be quite difficult. It is no less possible to excavate away objects that do not receive their proper attributes when recording digitally than it is when recording excavations on paper. To help solve basic record-keeping problems, we developed a unit metadata form. This form is one of the few paper products used systematically during data collection. It was extremely useful to keep this record on paper because it is metadata necessary to refer to while looking at the database. The quickness and flexibility of the paper record used in conjunction with the computers made it a valuable quality control tool. For each unit-level theme, we constructed in the computer database we recorded when the theme was created, when starting and ending attributes had been entered, and when the unit was finished. At the end of the day, this record was updated into the computer. As the form grew, we would periodically print out updated versions for use in the field.


GIS is a scaleless technology permitting analytical power difficult to achieve with analog data. However, archaeological geographic databases have been primarily constructed at the regional level. Modeling of excavated archaeological data is just as possible as the modeling of surface survey data. Once the spatial context of the excavation was constructed through the creation of a site matrix within the GIS, high-resolution digital photographs could be converted into spatially referenced records of an excavated surface. The georeferenced photographs were empirical observation useful for accurate and rapid digitizing of features and artifacts resting on exposed surfaces. Lastly, our paper excavation meta-data form worked to ensure that unit and feature attribute tables were properly completed during the process of excavation.

Thus far in archaeology, GIS databases are generally constructed after fieldwork is completed. There are multiple drawbacks to this approach to GIS construction. Efficiency is lost in the post-fieldwork paper-to-digital transformation, and there is a potential for the introduction of error. Real-time digital data collection avoids these two problems. Additionally, and perhaps more importantly, real-time digital data collection permits fundamentally new approaches to recording field data. Use of the digital camera as a high-resolution measurement tool would not have been possible with post-field data entry alone. This development is only one of many possible new data-collection techniques available when recording digital data in the field.

This project is still in its infancy. Although we had great success in implementing a real-time archaeological GIS under typical field conditions, there remains much to be done to make the system easier to use. We need to find economies of scale in image warping and data entry, and we also need to be more careful and creative in the ways in which data are downloaded and archived for both immediate field needs and longer term storage. We also want to explore the interface between total stations, the pen computers, and ArcView to automate even more effectively data collection.

Over the next few months, we will be developing a Web site that will discuss these issues, as well as many others, in greater detail. Look for the announcement of its address in the News and Notes section of the Bulletin in the not-too-distant future. ·

Nathan Craig is a graduate student in archaeology at the University of California-Santa Barbara.

back to top

Previous pageTable of ContentsNext Page