(PDF) Mapping simulated scenes with skeletal remains using differential GPS in open environments: An assessment of accuracy and

G Model FSI-7120; No. of Pages 14 Forensic Science International xxx (2013) xxx–xxx Contents lists available at SciVerse ScienceDirect Forensic Science International journal homepage: www.elsevier.com/locate/forsciint Case report Mapping simulated scenes with skeletal remains using differential GPS in open environments: An assessment of accuracy and practicality Brittany S. Walter a,*, John J. Schultz b a University of South Carolina, Department of Anthropology, 1512 Pendleton Street, Columbia, SC 29208, United States b University of Central Florida, Department of Anthropology, 4000 Central Florida Boulevard, Orlando, FL 32816, United States A R T I C L E I N F O A B S T R A C T Article history: Scene mapping is an integral aspect of processing a scene with scattered human remains. By utilizing the Received 13 July 2012 appropriate mapping technique, investigators can accurately document the location of human remains Received in revised form 13 January 2013 and maintain a precise geospatial record of evidence. One option that has not received much attention Accepted 17 February 2013 for mapping forensic evidence is the differential global positioning (DGPS) unit, as this technology now Available online xxx provides decreased positional error suitable for mapping scenes. Because of the lack of knowledge concerning this utility in mapping a scene, controlled research is necessary to determine the practicality Keywords: of using newer and enhanced DGPS units in mapping scattered human remains. The purpose of this Forensic archaeology Forensic anthropology research was to quantify the accuracy of a DGPS unit for mapping skeletal dispersals and to determine Differential global positioning systems the applicability of this utility in mapping a scene with dispersed remains. First, the accuracy of the DGPS Mapping skeletal dispersals unit in open environments was determined using known survey markers in open areas. Secondly, three Geographic information systems simulated scenes exhibiting different types of dispersals were constructed and mapped in an open environment using the DGPS. Variables considered during data collection included the extent of the dispersal, data collection time, data collected on different days, and different postprocessing techniques. Data were differentially postprocessed and compared in a geographic information system (GIS) to evaluate the most efficient recordation methods. Results of this study demonstrate that the DGPS is a viable option for mapping dispersed human remains in open areas. The accuracy of collected point data was 11.52 and 9.55 cm for 50- and 100-s collection times, respectfully, and the orientation and maximum length of long bones was maintained. Also, the use of error buffers for point data of bones in maps demonstrated the error of the DGPS unit, while showing that the context of the dispersed skeleton was accurately maintained. Furthermore, the application of a DGPS for accurate scene mapping is discussed and guidelines concerning the implementation of this technology for mapping human scattered skeletal remains in open environments are provided. ß 2013 Elsevier Ireland Ltd. All rights reserved. 1. Introduction emergence of recovery methods more geared toward forensic contexts. Recoveries involving skeletal remains have traditionally The recent re-emphasis of methodological techniques in been documented with hand-drawn maps that note the distribu- forensic archaeology is a development that has provided forensic tion of evidence. More recently, technology is being continually anthropology with ‘‘a new conceptual framework, which is adapted to record locational data of evidence at a forensic scene, broader, deep, and more solidly entrenched in the natural providing a more comprehensive spatial component for analysis sciences’’ [1, p. 33]. This shift is currently changing the goals of [2]. forensic anthropologists and how they approach situations in the While standard global positioning system (GPS) units generally field. Dirkmaat et al. [1] attribute the current configuration of do not offer the appropriate degree of accuracy for mapping forensic anthropology to four developments: (1) improvement of remains at a scene [3], portable differential global positioning field archaeology methods, (2) new technology, (3) new techniques system (DGPS) units offer decimeter to centimeter error margins used in the analysis of spatial data in the field, and (4) the which may be appropriate for mapping scattered remains. These enhanced units have the potential to collect accurate positional information of objects and provide the location of the object on the earth. In situations where scattered remains are extensively * Corresponding author at: Department of Anthropology, 1512 Pendleton Street, Hamilton College, Room 317, University of South Carolina, Columbia, SC 29208, dispersed over a large area or a topographically varied area, the United States. Tel.: +1 813 716 6724; fax: +1 803 777 0259. utilization of standard mapping techniques can be a difficult task E-mail address:

(B.S. Walter). [3,4]. There are a variety of benefits for integrating DGPS for 0379-0738/$ – see front matter ß 2013 Elsevier Ireland Ltd. All rights reserved. http://dx.doi.org/10.1016/j.forsciint.2013.02.027 Please cite this article in press as: B.S. Walter, J.J. Schultz, Mapping simulated scenes with skeletal remains using differential GPS in open environments: An assessment of accuracy and practicality, Forensic Sci. Int. (2013), http://dx.doi.org/10.1016/j.forsciint.2013.02.027 G Model FSI-7120; No. of Pages 14 e2 B.S. Walter, J.J. Schultz / Forensic Science International xxx (2013) xxx–xxx mapping scenes. This technology can be easily moved to map each commonly used in the determination of complementary satellite skeletal element over a large area without introducing additional geometry and is defined as the ‘‘ratio of the volume of a error as a result of long-distance measuring. In addition, only a tetrahedron created by the four most widespread, observed single operator is required during data collection. Most impor- satellite to the volume defined by the ideal tetrahedron’’ [10, p. tantly, the integration of accurate DGPS data to scene documenta- 183]. The PDOP is expressed as a number with the ideal tion is beneficial, as it provides the ability to spatially analyze and tetrahedron being 1. The closer the satellites are to each other, display survey data in a GIS, while maintaining a geospatial record the less accurate the satellite geometry, increasing the PDOP of the scene [5–9]. number [14]; thus, a lower PDOP is more desirable. Standard DGPS receivers will automatically choose the satellite constellation with 1.1. Differential global positioning systems theory the lowest PDOP. Additionally, PDOP is predetermined and can be acquired using planning software before data is collected in the The global positioning system (GPS) is a satellite-based field to ensure accurate data collection. positioning system involving 24 satellites circling the earth [8]. A GPS receiver uses positional information from these satellites 1.2. Previous DGPS research in forensic archaeology to calculate the position of an object on the earth. A DGPS unit is a more accurate enhancement of a standard GPS unit and Limited research has been conducted concerning the use of requires two receivers; one remains stationary while the other DGPS in the mapping of skeletal remains. The seminal research records positional data via a receiver (Fig. 1). The stationary provided by Listi et al. [3] determined that the use of a mid-price receiver, a basestation, relates all of the satellite measurements GPS receiver was not as accurate as traditional mapping to a single local reference [8]. The basestation measures the techniques because of influential variables such as tree cover timing errors and provides correction information to the other density, proximity of remains to structures and trees, and the receiver during postprocessing or in the field with real-time positioning of satellites which resulted in erratic data. It must be differential correction. Differential postprocessing software noted, however, that the GPS receiver used by Listi and obtains known basestation information via the internet and colleagues has become obsolete and affordable models that then compares this information to the mapped point data for are more accurate have been developed in the last five years that increased positional accuracy [10]. Differential GPS units are offer decimeter accuracy with postprocessing. While this was an handheld units that are compact and easy to transport to and important study that first tested the utility of GPS for mapping from a scene [4]. skeletal dispersals, it was limited in scope, as a single dispersal Though these enhanced units provide increased accuracy, with eight features in a mixed environment was the only several variables can limit the accuracy of points recorded by a application of this technology in the field. Thus, it is crucial that receiver such as cloud cover, satellite position, and obstruction of the accuracy of these new and enhanced DGPS units be assessed satellites from buildings and tree cover. Atmospheric delays can by utilizing more comprehensive controlled scenes with introduce error when satellite signals travel through the iono- numerous bones and concentrations of bones of differing size. sphere [11]. Changes in charged particle density in the ionosphere In a more recent study, Spradley et al. [15] utilized a DGPS for a and fluctuations of atmospheric density from temperature change scene involving taphonomic research. While this research in the atmosphere can affect the travel speed of the satellite signals primarily focused on the analysis of scavenging patterns of [12]. Differential GPS units, however, use dual-frequency receivers vultures on a human cadaver rather than the development of a to differentially correct this information by comparing the methodology concerning DGPS, the study is an important information collected by the receiver and the basestation to take example of applying this technology to analyzing the dispersal into account these changes and create sophisticated base models of a body. This limited research underlines the necessity of that reduce error margins [10]. controlled scenarios wherein different variables can be con- The geometry of satellite positions can also affect the positional trolled and tested to assess the accuracy of using a DGPS in error of a DGPS receiver. Satellites are most accurate when they are mapping skeletal dispersals. It is because of this lack of spaced farther apart, as close-set satellites overlap, causing areas of published research, that it is imperative for forensic anthro- positional uncertainty when signals intersect [10,12]. The geome- pologists to experiment with mapping technologies to deter- try of a constellation of satellites is expressed by a number called mine their applicability in the field. Furthermore, a better the Dilution of Precision (DOP). Positional DOP (PDOP) is the most understanding of this technology will aid in the development of efficient and accurate methods for data collection, maximizing point accuracy at a scene. More specifically, this research (1) determined the accuracy of using a DGPS unit in an open environment; (2) constructed scenarios in order to simulate scenes that may be encountered in real-life forensic cases; (3) collected geospatial and attribute data of features from skeletal dispersals using DGPS; (4) processed, analyzed, and generated maps of the data in geographic informa- tion systems (GIS); and (5) determined the applicability of utilizing DGPS for mapping skeletal dispersals. Because the DGPS is a relatively new technology that has yet to be comprehensively employed in the mapping of human remains, different aspects of this utility, such as data collection time, data collected on different days, proximity of features, feature data collection, and post- processing methods were also considered. Finally, the results of this study will contribute to the formulation of guidelines and recommendations for using a DGPS unit in scene mapping, supporting the current shift of integrating innovative technology Fig. 1. Illustration of differential GPS (DGPS) during location acquisition. into forensic archaeology. Please cite this article in press as: B.S. Walter, J.J. Schultz, Mapping simulated scenes with skeletal remains using differential GPS in open environments: An assessment of accuracy and practicality, Forensic Sci. Int. (2013), http://dx.doi.org/10.1016/j.forsciint.2013.02.027 G Model FSI-7120; No. of Pages 14 B.S. Walter, J.J. Schultz / Forensic Science International xxx (2013) xxx–xxx e3 2. Materials and methods for accuracy determination often determined using high-precision Global Navigation Satellite System (GNSS) technologies, such as commercial-grade DGPS units 2.1. Differential GPS unit and are accurate to the centimeter [10]. In civil engineering studies, it is standard practice to compare The differential GPS unit used for this research was a Trimble collected points to known points in order to ascertain GPS receiver GeoExplorer 2008 Series GeoXH handheld differential GPS receiver error [14,18–20]. This study utilized two survey markers as known with Zephyr antenna (Fig. 2). The receiver uses a field computer points on the earth to determine the error of the Trimble powered by Microsoft Windows Version 6 operating system and GeoExplorer 2008 Series DGPS unit. The coordinates of the known Terrasync software. The receiver also utilizes both H-star and points used for this project were provided by the Department of EVEREST multipath technology to provide heightened accuracy Transportation of the State of Florida and consisted of two points in after postprocessing using the internal antenna. The addition of the open areas in Deland, Florida (Fig. 3). external Zephyr antenna to this DGPS unit provides better locational recordation with 10–30 cm accuracy when data are 2.3. Distance accuracy differentially postprocessed [16]. The use of an external antenna and rangepole is also beneficial, The mapping of proximate bones as separate features or as increasing the accuracy of DGPS data in three ways [17]. First, by clusters were be analyzed by mapping bones at 5 cm, 10 cm, 15 cm, placing the antenna on a rangepole, the antenna is anchored to an 20 cm, 25 cm, and 30 cm distances. The 5 cm distance interval was associated point on the ground that holds the configuration in chosen to simulate the different levels of bone clustering. plumb and the addition of the tripod keeps the unit stationary Additionally, the accuracy of the DGPS unit in mapping long during data collection. The operator is, thus, given a definitive bones was determined by comparing maximum distance mea- point on the ground that he or she intends to map. Secondly, by surements of long bones with collected point data of the bones at placing the antenna above the operator, the DGPS unit does not proximal and distal ends using both 50- and 100-s collection times. suffer from the abstraction the operator creates with his or her This was conducted in order to determine the level of accuracy of body. Finally, by using an antenna that is of a higher quality, such the DGPS in mapping long bone length and orientation, and, thus, as the Zephyr, the quality and strength of the signal is increased demonstrated the accuracy of the unit when illustrated on a map. drastically. The Zephyr also adds an additional antenna frequency The result from the analysis of these data was then applied to data to the single frequency DGPS unit. These frequencies are typically collection for the simulated scenarios in the next section. associated with survey-grade DGPS units because the combination of both can be used to calculate a far higher level of precision. 2.4. Data collection parameters 2.2. Controlled points Prior to the day of data collection, Trimble planning almanac software which provides the potential satellite positioning was Survey markers, or benchmarks, are known points on the earth consulted to determine the best time for data collection. The best maintained by various federal and state agencies such as the time of day for data collection can be determined by considering Department of Transportation [10]. The exact coordinates, loca- the greatest number of satellites in view, with at least four tion, and description of these survey markers may be obtained satellites being the most desirable, and the least PDOP value, with from these agencies. Currently, survey marker locations are most values less than two being the most desirable [14]. Data collection was conducted during this time period. Also, the orientation of the DGPS receiver was considered, as the vertical orientation of the GPS receiver has been found to significantly influence accuracy, with vertical orientation of the receiver yielding coordinates that are more accurate, rather than horizontal orientation [20]. Additionally, during data collection, the end of the rangepole was positioned at predetermined point adjacent to the skeletal element on the ground and then leveled using the dot level on the rangepole and remained stationary throughout the data collection time via the tripod. Data was collected in US State Plane 1983, Florida East, with the NAD 1983 Conus datum, as this is the coordinate system and datum used by the Florida Department of Transportation in Deland, Florida for the survey markers. Point data was collected using the batch method, which is the average of the point data in 1-s intervals. The collection times for this research were chosen in accordance with the 100-s collection used in previous research [3], with the addition of the more efficient 50-s collection time for comparison purposes. Point data were collected at each survey marker in 1-s intervals for both 50 and 100 s. Using this method, 25 points were collected consecu- tively for each survey marker at both 50- and 100-s time intervals. 2.5. Data processing After point data were collected in the field, unprocessed data were transferred to a desktop computer and imported into Trimble Pathfinder Office1 version 5.2 for differential postprocessing. Uncorrected data were differentially postprocessed against Fig. 2. DGPS unit with labeled components mapping a longbone in the field. the closest public basestation in Deland, Florida (CORS96), Please cite this article in press as: B.S. Walter, J.J. Schultz, Mapping simulated scenes with skeletal remains using differential GPS in open environments: An assessment of accuracy and practicality, Forensic Sci. Int. (2013), http://dx.doi.org/10.1016/j.forsciint.2013.02.027 G Model FSI-7120; No. of Pages 14 e4 B.S. Walter, J.J. Schultz / Forensic Science International xxx (2013) xxx–xxx Fig. 3. Labeled map of survey markers in open areas in Deland, FL. approximately 8 km from the mapped area with an integrity index the 100-s collection time and 11.55 cm for the 50-s collection time. of 94.7. The integrity index is a grading system by Trimble that Correspondingly, Survey Marker 2 displayed an accuracy mean of monitors basestations used for differential processing and rates a 9.51 cm for 100-s collection time and 11.48 cm for the 50-s basestation on its reliability, accuracy, and precision [17]. collection time. Therefore, the data collected at survey markers in Additionally, the integrity index value for a basestation is adjusted open areas produced a mean accuracy of 9.55 cm for 100-s in consideration of the basestation’s proximity to where the data collection time and 11.52 cm for the 50-s collection time. was collected by the receiver. Trimble recommends postprocessing Moreover, there was an approximate 20% accuracy increase using against a basestation with an integrity index of 80 or higher that is 100-s data collection when compared to 50-s data collection within 200 km of the surveyed area [16]. Processed data was (Table 2). exported into ESRI1 ArcGIS version 10 for analysis in ArcMap to The results of an independent samples t-test of mean accuracies determine the distance of the collected points to the survey for the collection times are shown in Table 3. The results from this markers. test demonstrate that for both survey markers, there was a significant difference in accuracy between 50- and 100-s collection 2.6. Calculating accuracy times (SM 1 p = .000, SM 2 p = .001). These results support the hypothesis that the accuracy of the DGPS unit was significantly The following formula is a distance formula commonly used to increased by 100-s collection time, compared to 50-s collection determine the distance between two points and was used to time. determine the distance of the collected points to the known points Table 4 shows the results of the independent samples t-test [10]: conducted between the different collection times for both survey qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi markers in open areas. The analysis indicates that the different Accuracy ¼ ðxi x0 Þ2 þ ðyi y0 Þ2 open areas did not yield significantly different GPS coordinates for both 50- and 100-s collection times (50-s p = .884, 100-s p = .893), where xi is the collected horizontal coordinate, x0 is the known suggesting that the DGPS unit produced consistent results during horizontal coordinate, yi is the collected vertical coordinate, and y0 data collection of the open areas for both collection times, which is the known vertical coordinate. The accuracy was calculated by were also collected on different days. exporting attribute data into Microsoft Excel version 10 using the formula tool, which makes batch calculations of data. Accuracy 3.2. Long bone accuracy data was imported into GIS as attribute information with additional statistical analyses discussed later. To ensure correct The actual maximum length of each long bone was compared to accuracy calculations, the accuracy of each point was then cross- the measurements between collected points in ArcGIS for both 50- checked using the measuring tool in ArcMap. For all collected and 100-s collection times. Fig. 4 illustrates the points collected at points, the calculated accuracy and the distance measurement in the proximal and distal aspects of the long bones in the field for GIS were equal to the nearest hundredth centimeter. both collection times in an open area. Lines were added in ArcMap to illustrate the orientation of the long bones. Overall, the 100-s 3. Results of accuracy determination collection time was consistently more accurate than the 50-s when collected points were compared to actual lengths (Table 4). Long 3.1. Survey marker accuracy bones with a maximum length greater than 25 cm demonstrated collected points that were closer to the actual maximum length The accuracy for each survey marker and collection time was (Table 5). Additionally, the orientation of long bones with a calculated using the aforementioned formula and descriptive maximum length greater than 25 cm demonstrated correct statistics were determined using IBM1 SPSS version 20 for orientation and less varied orientation between collection times comparison purposes (Table 1). The two survey markers (SM 1 when compared to shorter long bones (Fig. 4). and SM 2) demonstrated similar results for both collection times. Positional data were also collected for known distances (0 cm, Survey Marker 1 demonstrated an accuracy mean of 9.59 cm for 5 cm, 10 cm, 20 cm, 25 cm, and 30 cm), to simulate bone Please cite this article in press as: B.S. Walter, J.J. Schultz, Mapping simulated scenes with skeletal remains using differential GPS in open environments: An assessment of accuracy and practicality, Forensic Sci. Int. (2013), http://dx.doi.org/10.1016/j.forsciint.2013.02.027 G Model FSI-7120; No. of Pages 14 B.S. Walter, J.J. Schultz / Forensic Science International xxx (2013) xxx–xxx e5 Table 1 Summary of the accuracy results for SM 1 and SM 2 with both collection times. Survey marker Mean (cm) Standard Range (cm) 95% confidence deviation (cm) interval (cm) 50-s collection time SM 1 11.55 1.65 9.13–13.38 10.89–12.21 SM 2 11.48 1.75 7.98–14.79 10.80–12.16 100-s collection time SM 1 9.59 1.82 7.36–12.84 8.86–10.32 SM 2 9.51 2.25 5.30–13.87 8.63–10.39 clustering, and were compared in ArcMap using the measurement (an electronic form that can be exported to the DGPS unit that tool. The measuring tool is a simple function that calculates the queries the user to collect certain information during data actual distance between areas on a map. Like the long bone collection) with bone type, bone side, aspect, collection time, measurements, the 100-s collection time was consistently more and notes. These attributes were exported into ArcGIS with the accurate than the 50-s collection time, but not by more than 2 cm point data for analysis and map creation. (Table 6). Most notably, the error decreased as the distances increased, with the 25 cm and 30 cm distances demonstrating a 4.3. Data processing and map creation difference of less than 1 cm. After point data were collected in the field, unprocessed data 4. Materials and methods for simulated scenarios were transferred to a desktop computer and imported into Trimble1 GPS Pathfinder 1 Office for differential postprocessing. 4.1. Scenarios Uncorrected data were collected in the field and then differentially corrected against the same basestation as the previous section Scenarios were constructed to depict various examples of (CORS96). Processed data were then exported into ArcGIS 10 for skeletal dispersals that may be encountered in real-life situations. analysis in ArcMap and were analyzed to determine the Skeletal dispersals have been reported to range from relatively practicality of utilizing a DGPS unit in mapping skeletal dispersals. articulated skeletons to skeletal elements dispersed over hundreds The wide skeletal dispersal (Scenario 1) underwent additional of meters [21]. Three scenarios were chosen to represent dispersal postprocessing to ascertain the influence of different base stations scenarios of human remains in an outdoor setting and consisted of and multiple basestations for differential correction. This scenario a wide scatter (Scenario 1), tight scatter (Scenario 2), and relatively was processed using the Deland basestation (DLND), Cape articulated skeleton (Scenario 3) in an open environment. Data Canaveral basestation (CCV6), and Leesburg basestation (LEES), collection for all of the scenarios was conducted in the Deep and subsequently processed using multiple basestations (DLND, Foundations and Geotechnical Research Site located at the CCV6, and LEES) (Table 7). University of Central Florida Arboretum in Orlando, Florida The mean accuracy of the DGPS unit in open environments was (Fig. 5). Point data was collected with 50- and 100-s collection applied to each scenario using the buffer tool in ArcMap. This tool times on the same day for all of the scenarios. The skeletal material creates a buffer area around the point with the calculated radii used for all of the scenarios in this project consisted of plastic cast determined as the mean of the error from both survey markers for model skeletal elements. both collection times. The tool then exports a shapefile, a file format in GIS designed to contain geometric information on a map. 4.2. Data collection parameters By adding this shapefile, one is able to take into account the error of the DGPS for the collected points visually for the scenarios in As in the previous section, the differential GPS used for this addition to the collected points. A buffer shapefile was created research was a Trimble GeoExplorer 2008 Series GeoXH handheld using the buffer tool by inputting the point shapefile and setting differential GPS receiver with Zephyr antenna and rangepole setup. the radius to 11.52 cm and 9.55 cm for 50- and 100-s collection Data collection methods were duplicated from the previous section times, respectively. with the exception of the coordinate system and datum which was Universal Transverse Mercator (UTM), Zone 17 North and the WGS 5. Results of simulated scenarios 1984 datum. The distance analysis conducted in the previous section was Scenario 1 consisted of a wide skeletal dispersal in an open area implemented during data collection by collecting data for bones and was postprocessed using different basestations. The skeletal with a maximum length of less than 25 cm as a single point, and elements in this scenario were also flagged and collected during with a maximum length exceeding 25 cm as two points at opposite two different days using the same data collection methods. After aspects (i.e. proximal and distal ends). Skeletal elements clustered analysis in ArcMap, it was determined that the orientation of the within 25 cm of each other were recorded as a single feature. long bones was generally maintained for all postprocessed data Finally, data were collected using a predefined data dictionary collected using a 100-s time collection, while the orientation of the Table 3 Table 2 Results of independent samples t-test for both collection times and survey markers. Average error and difference between collection time error of collected points to survey markers. Survey Collection Mean (cm) s.d. (cm) t df p marker time Survey Mean 50-s Mean 100-s Mean error Percentage SM 1 50 s 11.55 1.65 3.83 48 0.000 marker error (cm) error (cm) difference (cm) increase 100 s 9.59 1.82 SM 1 11.55 9.59 1.96 20.7% SM 2 50 s 11.48 1.75 3.47 48 0.001 SM 2 11.48 9.51 1.97 20.4% 100 s 9.51 2.25 Please cite this article in press as: B.S. Walter, J.J. Schultz, Mapping simulated scenes with skeletal remains using differential GPS in open environments: An assessment of accuracy and practicality, Forensic Sci. Int. (2013), http://dx.doi.org/10.1016/j.forsciint.2013.02.027 G Model FSI-7120; No. of Pages 14 e6 B.S. Walter, J.J. Schultz / Forensic Science International xxx (2013) xxx–xxx Table 4 Table 5 Independent samples t-test of mean error (cm) for 50- and 100-s collection times Comparison of actual long bone length and lengths between collected points and mean percentage changes between collection times for open areas. measured in GIS for 50- and 100-s collection times. Collection time SM 1 SM 2 t df p Actual max GIS max length Difference length (cm) (cm) (cm) 50-s 11.55 11.48 0.146 48 0.884 100-s 9.59 9.51 0.136 48 0.893 50 s 100 s 50 s 100 s Humerus 31.0 35.5 31.8 4.5 0.8 Radius 23.0 21.4 25.2 1.6 2.2 long bones using the 50-s collection time was not as consistent. Ulna 25.0 28.5 27.8 3.5 2.8 Femur 42.0 47.5 44.4 5.5 2.4 This trend was also demonstrated when considering the actual Tibia 32.0 34.4 33.4 2.4 1.4 length of the long bones, with the data collected using the 100-s Fibula 35.0 37.1 34.3 2.1 0.7 collection time collecting a maximum length consistent with the Os coxa 22.0 21.5 22.1 0.5 0.1 actual maximum length of the long bones. Mean difference 2.9 1.5 Furthermore, when considering postprocessing using different basestations (Table 7), the basestation with the greatest distance from Scenario 1, LEES, produced the most inconsistent orientation Point data collected on different days for Scenario 1 were and the greatest difference from the actual maximum length of the postprocessed against the DLND basestation and compared in long bones (Fig. 6). This is to be expected, as an increase of distance ArcMap (Fig. 7). The general orientation of the long bones were decreases the reliability of a basestation [10]. The CCV6 base- maintained with both collection times for Day 2 but contrasted station, with a slightly lower integrity index than the DLND slightly from the orientation of the long bones for Day 1. Also, the basestation maintained the general orientation and maximum 100-s collection time recorded maximum lengths closer to the length of the long bones for both collection times (Fig. 6). actual maximum lengths of the long bones for both days when Postprocessing conducted against multiple basestations demon- compared to the 50-s collection time. However, the data collected strated similar results to the LEES basestation with inadequate on Day 2 were within the error buffers determined in the previous maintenance of long bone orientation and maximum length, which chapter for both collection times. may have been due to the low integrity index of one of the Scenario 2 consisted of a tight skeletal dispersal in an open area basestations (LEES) influencing the overall integrity index when all using 50- and 100-s collection times and data were differentially of the basestations were combined (Fig. 6). corrected against the DLND basestation. Like the wide scatter in an Fig. 4. Collected points of long bone measurements for both collection times with the maximum length of each bone in the field. Please cite this article in press as: B.S. Walter, J.J. Schultz, Mapping simulated scenes with skeletal remains using differential GPS in open environments: An assessment of accuracy and practicality, Forensic Sci. Int. (2013), http://dx.doi.org/10.1016/j.forsciint.2013.02.027 G Model FSI-7120; No. of Pages 14 B.S. Walter, J.J. Schultz / Forensic Science International xxx (2013) xxx–xxx e7 Table 6 Table 7 Comparison of known lengths and lengths of collected points measured in GIS for Basestations used during postprocessing of Scenario 1 with location, distance, and 50- and 100-s collection times. integrity index from the mapped area. GIS length (cm) Difference (cm) Basestation Location Distance (km) Integrity index 50 s 100 s 50 s 100 s DLND Deland, FL 51 90.97 5 cm 7.5 7.3 2.5 2.3 CCV6 Cape Canaveral, FL 65 84.98 10 cm 12.6 11.9 2.6 1.9 LEES Leesburg, FL 66 81.20 15 cm 17.6 16.2 2.6 1.2 20 cm 22.2 21.4 2.2 1.4 25 cm 26.4 25.9 1.4 0.9 limited because of the decreased radius of the error buffers for the 30 cm 31.1 30.6 1.1 0.6 100-s collection time. This lack of overlap was demonstrated in the tight scatter and wide scatter, but was not shown in the relatively open area (Scenario 1), increased collection time for point data articulated scatter; thus, the DGPS unit was not able to preserve collected in Scenario 2 maintained the relative orientation and the general disposition of the separate elements for this type of maximum length of the long bones (Fig. 8). Also, the decreased scatter because of the close proximity of the bones. radius of the error buffer and dispersal type caused minimal With the inevitable technological improvements of DGPS, overlapping of the features, resulting in an accurate depiction of increased accuracy of DGPS receivers is an endeavor that is the overall dispersal pattern. expected in the years to come [12]. Studies conducted seven years Scenario 3 consisted of a relatively articulated skeleton in an ago, show error levels of 1.7 m for DGPS units in open open area using 50-s and 100-s collection times. Increased environments [20], clearly demonstrating the rapid pace at which collection time for point data collected in this scenario shows DGPS technology is developing when compared to the decimeter preservation of the general orientation and maximum length of the accuracy determined from this study. It is, thus, the responsibility long bones (Fig. 9). Some single features, however, did not of the forensic anthropologist to evaluate the use of these new maintain an exact position but still fell within the error buffer. technologies in the field within different environments and Furthermore, increased overlapping of the error buffers when conditions before applying them in forensic situations. compared to the other scenarios was noted for this type of Recommended guidelines are provided in this section to dispersal, resulting in difficulty when distinguishing between maximize the efficiency and accuracy of this technology for separate features. mapping a skeletal dispersal. Before data collection can occur, it must be determined if mapping using a DGPS unit is appropriate 6. Discussion and guidelines for the scene in question. It is recommended that a DGPS unit may be utilized to map skeletal dispersals only in open environments, A number of important data collection parameters were as the accuracy of DGPS units in obstructed environments has not determined for implementation of a DGPS unit for mapping yet been explored. The use of traditional mapping methods such as dispersed skeletal remains in open environments. The accuracy a compass survey or baseline may introduce additional error from calculated from the survey markers for both collection times long distances and obstructions for widely dispersed remains [3]. produced consistently accurate positional error for both collection Furthermore, a total station may not be a viable mapping method times at two separate areas on different days, demonstrating that in these situations because of the obstruction of the line of site over the DGPS used for this research produces relatively consistent long distances which can result in the constant relocation of the results at different locations on different days when in the same transit point to gain an accurate sight line from the total station environment. Additionally, the 100-s collection time was found to unit to the stadia point [3,4,9]. A DGPS unit may be a better option be approximately 20% more accurate than the 50-s collection time. than more traditional mapping methods for wide scatters because The point data collected for the scenarios demonstrated a of its portability and ease of use over long distances. However, for consistent maintenance of long bone orientation and maximum situations with a relatively articulated skeleton or skeletal length. Furthermore, the overlapping of features in open areas was elements in close proximity, a DGPS unit may not be the best Fig. 5. Aerial image of the open environment location at the Deep Foundations and Geotechnical Research Site on the University of Central Florida campus in Orlando, FL, with inset image of the area. Please cite this article in press as: B.S. Walter, J.J. Schultz, Mapping simulated scenes with skeletal remains using differential GPS in open environments: An assessment of accuracy and practicality, Forensic Sci. Int. (2013), http://dx.doi.org/10.1016/j.forsciint.2013.02.027 G Model FSI-7120; No. of Pages 14 e8 B.S. Walter, J.J. Schultz / Forensic Science International xxx (2013) xxx–xxx Fig. 6. Composite image created using ArcMap of DLND basestation against CCV6 (A), LEES (B), and multiple basestations (C) with actual orientation of long bones in the field for Scenario 1. mapping option because of overlapping error margins, and a more each skeletal element. It was determined through this research appropriate method, such as a baseline, should be employed. that bones longer than 25 cm (i.e. crania, long bones, etc.) should be Moreover, if rain is an issue, the DGPS unit should not be used, as mapped using two points to illustrate orientation, while bones less the model used for this research is not waterproof. than 25 cm (i.e. scapulae, ribs, vertebrae, etc.) should be mapped Before a DGPS unit is used for mapping, the accuracy of the using a single point. Predetermined points should be assigned to unit must be determined by comparing collected coordinate bones less than 25 cm in length and should be consistent information to a known survey marker. This may be accom- throughout data collection. For example, in this study, the scapulae plished by following the practices provided in Section 2 of were measured at the glenoid fossa and all vertebrae were this article. Data concerning survey markers may be obtained measured at the anterior aspect of the body (Table 8). Also, it is from the Department of Transportation in the user’s area. If the recommended to map bones as individual features when skeletal error determined by the DGPS unit is within the accuracy elements are at least 25 cm apart, and to map clusters of two or range desired by the user, the DGPS unit may be employed in more bones that are less than 25 cm apart as one feature. The user similar environmental scenarios (i.e. open area error applied to may then include information in the ‘‘Notes’’ data entry describing dispersals in open areas). the skeletal elements comprising the cluster. Additionally, the type Developing a consistent methodology for data collection is of equipment should also be consistent throughout data collection. crucial for mapping any scene. By following predetermined A rangepole should be used to better pinpoint the exact location of procedures, there is less of a chance of introducing additional the element that the user is measuring. The addition of a tripod to error or not including important information. Before data the rangepole is recommended to keep the DGPS unit level and collection, determine how the point data will be collected for stationary during location acquisition. Please cite this article in press as: B.S. Walter, J.J. Schultz, Mapping simulated scenes with skeletal remains using differential GPS in open environments: An assessment of accuracy and practicality, Forensic Sci. Int. (2013), http://dx.doi.org/10.1016/j.forsciint.2013.02.027 G Model FSI-7120; No. of Pages 14 B.S. Walter, J.J. Schultz / Forensic Science International xxx (2013) xxx–xxx e9 Fig. 7. Map of 100-s and 50-s collection times with accuracy buffers of Scenario 1 using ArcMap. After determining how point data will be collected for each type Environment type of bone and what equipment will be employed, a data dictionary Date collected should be created to include any additional attribute information. Time collected It is recommended that the following information be included in Notes order to ensure a complete inventory of the skeletal elements and characteristics: It is possible to assign default selections to a field within the data dictionary, such as type of environment, to save time during Bone type (i.e. cranium, mandible, etc.) data collection. Also, time of data collection and date of data Side collection can be automatically generated by the software rather Completeness (i.e. complete, 75%, etc.) than being entered manually. Another advantage of using a data Aspect (i.e. proximal, distal, etc.) dictionary is that the software will not allow the user to save a Collection time feature unless all of the required information is entered, ensuring Please cite this article in press as: B.S. Walter, J.J. Schultz, Mapping simulated scenes with skeletal remains using differential GPS in open environments: An assessment of accuracy and practicality, Forensic Sci. Int. (2013), http://dx.doi.org/10.1016/j.forsciint.2013.02.027 G Model FSI-7120; No. of Pages 14 e10 B.S. Walter, J.J. Schultz / Forensic Science International xxx (2013) xxx–xxx Fig. 8. Map of 100-s and 50-s collection times with accuracy buffers for Scenario 2 using ArcMap. that crucial information is not missed in the field. By including a free, accessible through the DGPS unit or available through major field for notes in the data dictionary, the user is able to include any DGPS retailers, and is obtainable up to a year in advance. Planning additional information concerning the feature. Finally, it is software provides information including the satellite position data, recommended that the data dictionary be as generic as possible, DOP data, and elevation data on specific days. It is recommended so that it may be applied to several types of scenes rather than to that data collection be determined by considering the greatest create a new data dictionary each time a scene is mapped. number of satellites, with at least four satellites being the most As previously mentioned, an important aspect of successful desirable, and the least PDOP value, with values less than two data collection is good satellite geometry, which is the desired being the most desirable [14]. Time windows fitting these number and position of satellites for optimal accuracy during data characteristics should be at least 3 h long to allow for adequate collection [10]. To achieve good satellite geometry, it is recom- time to survey the scene and allow additional time for unforeseen mended that the user consult a satellite almanac to determine the circumstances. If preplanning is not possible, it is recommended best time of day for data collection. Planning almanac software is that the PDOP value and number of satellites be recorded during Please cite this article in press as: B.S. Walter, J.J. Schultz, Mapping simulated scenes with skeletal remains using differential GPS in open environments: An assessment of accuracy and practicality, Forensic Sci. Int. (2013), http://dx.doi.org/10.1016/j.forsciint.2013.02.027 G Model FSI-7120; No. of Pages 14 B.S. Walter, J.J. Schultz / Forensic Science International xxx (2013) xxx–xxx e11 Fig. 9. Map of 100-s and 50-s collection times with accuracy buffers for Scenario 3 using ArcMap. data collection so if poor satellite geometry were to occur, it could The collection time for point data should be determined in explain the increased error of point data. consideration of the number of elements to be mapped and the After a data collection method has been established and once amount of time available in the field. The collection time should be the best time of day for data collection has been determined, point as long as possible, as increased collection time adds point data data may be collected in the field. It is recommended that all that will be averaged for the final coordinate. It is recommended predetermined methods be executed consistently throughout the that at least a 100-s collection time be implemented, as this data collection process. First, skeletal elements and evidence collection time was found to be more accurate than a 50-s should be flagged, so that features are not missed. It is also collection time; however, if time is an issue, a 50-s collection time recommended that the DGPS unit be oriented vertically, so as not was found to provide sufficient accuracy in open areas. Once the to degrade satellite reception [20]. If possible, one user should collection time is determined, it should be utilized for all point conduct the entire survey, so that possible error from differences in data. data collection methods is not introduced and features are not After positional data has been collected for the objects of skipped. interest, collected features should be cross-checked with the Please cite this article in press as: B.S. Walter, J.J. Schultz, Mapping simulated scenes with skeletal remains using differential GPS in open environments: An assessment of accuracy and practicality, Forensic Sci. Int. (2013), http://dx.doi.org/10.1016/j.forsciint.2013.02.027 G Model FSI-7120; No. of Pages 14 e12 B.S. Walter, J.J. Schultz / Forensic Science International xxx (2013) xxx–xxx Fig. 10. Flow chart of guidelines for collecting and processing DGPS data for skeletal dispersals. flagged skeletal elements and evidence. This may be accomplished must be imported into differential correction software, such as GPS by looking at the list of features collected or with a map of the area Pathfinder1 Office. Generally, there is a short period of lag time on the data view screen of the unit. The user may also collect (approximately 1 to 2 h) in receiving basestation data through the features such as trail entrances, datums, and buildings to provide Internet. When choosing a basestation to differentially correct the context to the site. This may be done by creating lines for roads, DGPS data against, it is crucial to choose the basestation that is polygons for buildings using the DGPS unit, or simply with point within the closest proximity to the scene at which the data were data of features. collected. Trimble recommends postprocessing against a base- Once all data has been collected, it is necessary to transfer the station with an integrity index of 80 or higher and within 200 km of data to a computer where the data will be differentially corrected the site [16]. Once the data has been successfully corrected, data against a nearby basestation. After the data is transferred, the data must then be exported as a shapefile for analysis in a GIS, Please cite this article in press as: B.S. Walter, J.J. Schultz, Mapping simulated scenes with skeletal remains using differential GPS in open environments: An assessment of accuracy and practicality, Forensic Sci. Int. (2013), http://dx.doi.org/10.1016/j.forsciint.2013.02.027 G Model FSI-7120; No. of Pages 14 B.S. Walter, J.J. Schultz / Forensic Science International xxx (2013) xxx–xxx e13 Table 8 DGPS units, as only a single DGPS unit was considered in this Description of points and number of points collected on surveyed bones and study. number of points collected for each skeletal element. Furthermore, now that DGPS units may be utilized for mapping Skeletal element Number Description of points collected dispersals, various options that are available by integrating DGPS of points data from a scene into a GIS should be explored. Geographic Cranium 2 If oriented sideways: anterior information systems can be valuable for presentation purposes by and posterior aspects incorporating basemaps to provide context to a scene, drawing If oriented long ways: superior tools to highlight important aspects of the dispersals, and the and inferior aspects Mandible 1 Anterior aspect buffer tool to illustrate the accuracy of the DGPS unit. Moreover, Vertebrae 1 Anterior aspect of the body several analytical tools may be employed to investigate the spatial Sternum 2 Superior and inferior aspect distribution of remains and analyze scatter patterns of a skeletal Ribs 1 Medial aspect of head dispersal; however, an abundance of tools available in GIS analysis Scapulae 1 Lateral aspect (glenoid fossa) Clavicle 1 Anterior aspect of midshaft have not yet been utilized to assess their usefulness for a scene Os coxa 2 Superior and inferior aspects with human remains. Many factors can be deduced from spatial Humerus 2 Proximal and distal aspects analysis of a scene, including the quantification of commingled Radius 2 Proximal and distal aspects remains [22–24], dispersal distance [21], and scavenging patterns Ulna 2 Proximal and distal aspects [15] that may aid in understanding and reconstructing events that Carpal 1 Distal aspect Metacarpal 1 Distal aspect occurred at a site prior to its discovery. Thus, further research with Manual phalanx 1 Distal aspect these tools is necessary to determine their utility when analyzing Articulated hand 2 Proximal and distal aspects and displaying skeletal dispersals for both small-scale and large- Femur 2 Proximal and distal aspects scale situations. Patella 1 Distal aspect Tibia 2 Proximal and distal aspects Fibula 2 Proximal and distal aspects Acknowledgements Tarsal 1 Distal aspect Metatarsal 1 Distal aspect We would like to thank the Department of Transportation in Pedal phalanx 1 Distal aspect Deland, Florida, for providing the survey marker information used Articulated foot 2 Proximal and distal aspects in this research. Also thanks to the Departments of Civil, Environmental and Construction Engineering at the University of Central Florida for providing access to the Deep Foundations and Geotechnical Research Site. The DGPS unit utilized in this research maintaining the same coordinate system and datum used during was provided by the Department of Anthropology at the University data collection. Finally, it is recommended that during map of Central Florida. We would also like to thank Chelsea Nicole creation, the accuracy of the DGPS unit be illustrated by using the Stewart for providing assistance in data collection. Finally, thanks buffer tool in ArcMap and that a basemap (an aerial map of the area to anonymous reviewers and Carrie Healy for constructive provided by ArcGIS) be added to give context to the scene. Fig. 10 is comments leading to the final version of this paper. a flowchart of the recommended guidelines provided from this research. References [1] D.C. Dirkmaat, L.L. Cabo, S.D. Ousley, S.A. Symes, New perspectives in forensic 7. Conclusions anthropology, Am. J. Phys. Anthropol. 137 (2008) 33–52. [2] D.C. Dirkmaat, G.O. Olson, A.R. Klales, S. Getz, The role of forensic anthropology in This research has served as the first thorough investigation of the recovery and interpretation of the fatal-fire victim, in: D.C. Dirkmaat (Ed.), A Companion to Forensic Anthropology, Blackwell Publishing, Oxford, 2012, pp. utilizing a DGPS unit and GIS in mapping scattered human 113–135. remains. Research concerning the use of DGPS equipment and [3] G.A. Listi, M.H. Manhein, M. Leitner, Use of the global positioning system in the GIS software in scene mapping and analysis is limited [3]. field recovery of scattered human remains, J. Forensic Sci. 52 (2007) 11–15. [4] L.K. Napton, E.A. Greathouse, Archaeological mapping, site grids, and surveying, Mapping dispersed remains and associated evidence in the field in: T.R. Hester, H.J. Shafer, K.L. Feder (Eds.), Field Methods in Archaeology, 7th ed., can be challenging because of long distances and varied Leftcoast Press, Walnut Creek, CA, 2009, pp. 177–234. environments in which human remains are generally found. [5] D.W. Lowe, B.A. Burns, Using GPS and GIS to create a historic base map, Cult. Res. This study presented a quantification of the influence of different Man. 21 (1998) (1998) 38–39. [6] J. Gao, Integration of GPS with remote sensing and GIS: reality and prospect, J. variables likely to affect the accuracy of a DGPS receiver. Though Photogram. Eng. Remote Sens. 68 (2002) 447–453. several factors may influence the accuracy of the DGPS unit, the [7] J. Spencer, B.G. Frizzelle, P.H. Page, J.B. Vogler, Global Positioning System: A Field error determined for this unit in an open environment is Guide for the Social Sciences, Blackwell Publishing Ltd., Malten, MA, 2003. [8] A. El-Rabbany, Introduction to GPS: The Global Positioning System, 2nd ed., appropriate for mapping skeletal dispersals. Most importantly, Arctech House, Inc., Norwood, MA, 2006. the use of error buffers for point data of bones in maps [9] T.L. Dupras, J.J. Schultz, S.M. Wheeler, L.J. Williams, Forensic Recovery of Human demonstrated the error of the DPGS unit, while the point data Remains: Archaeological Approaches, 2nd ed., CRC Press, Boca Raton, FL, 2011. [10] P. Bolstad, GIS Fundamentals: A First Text on Geographic Information Systems, successfully maintained the entire context of the dispersed 3rd ed., Eider Press, White Bear Lake, MN, 2000. skeleton over the area. [11] P.M. Kintner, B.M. Ledvina, The ionosphere, radio navigation, and global naviga- Though this research has provided a comprehensive set of tion satellite systems, Adv. Space Res. 35 (2005) 788–811. [12] W. Lechner, S. Baumann, Global navigation satellite systems, Comput. Electron. guidelines for utilizing a DGPS unit in mapping skeletal remains, Agric. 25 (2000) 67–85. the influence of additional variables to DGPS locational acquisi- [14] C.E. Johnson, C.C. Barton, Where in the world are my field plots? Using GPS tioning must still be ascertained. Further research is necessary to effectively in environmental field studies, Front. Ecol. Environ. 2 (2004) 475–482. [15] M.K. Spradley, M.D. Hamilton, A. Giordano, Spatial patterning of vulture scav- determine the influence of additional variables to the accuracy of enged human remains, Forensic Sci. Int. 219 (2011) 57–63. DGPS units such as the type of environment (i.e. a tree-covered [16] Trimble Navigation Limited, GeoExplorer 2002 Series User Guide, Trimble area or an area near a tall structure), level of cloud cover, time of Navigation Limited, Westminster, CO, 2009. day, PDOP, and additional collection times. In addition to [17] Trimble Navigation Limited, Why Postprocess GPS Data? Mapping and GIS White Paper, Trimble Navigation Limited, Westminster, CO, 2004. postprocessing, real-time differential processing should also be [18] K. Serr, T. Windholz, K. Weber, Comparing GPS receivers, U.R.S.A. J. 18 (2006) investigated. Also, error levels should be determined for other 19–23. Please cite this article in press as: B.S. Walter, J.J. Schultz, Mapping simulated scenes with skeletal remains using differential GPS in open environments: An assessment of accuracy and practicality, Forensic Sci. Int. (2013), http://dx.doi.org/10.1016/j.forsciint.2013.02.027 G Model FSI-7120; No. of Pages 14 e14 B.S. Walter, J.J. Schultz / Forensic Science International xxx (2013) xxx–xxx [19] A.J. Graettinger, T.W. Rushing, J. McFadden, Evaluation of inexpensive global A Companion to Forensic Anthropology, Blackwell Publishing, Oxford, 2012, pp. positioning system units to improve crash location data, Transport. Res. Rec. 1746 175–196. (2001) 94–101. [23] H. Tuller, U. Hofmeister, S. Daley, Spatial analysis of mass grave mapping data to [20] T. Sando, R. Mussa, J. Sobanjo, L. Spainhour, Quantification of the accuracy of low assist in the reassociation of disarticulated and commingled human remains, in: priced GPS receivers for crash location, J. Trans. Res. For. 44 (2005) 19–32. B.J. Adams, J.E. Byrd (Eds.), Recovery, Analysis, and Identification of Commingled [21] M.H. Manhein, G.A. Listi, M. Leitner, The application of geographic information Human Remains, Humana Press, Totowa, NJ, 2008, pp. 7–29. system and spatial analysis to assess dumped and subsequently scattered human [24] N.P. Herrmann, J.B. Devlin, Assessment of commingled human remains using a remains, J. Forensic Sci. 51 (2006) 469–474. GIS-based approach, in: B.J. Adams, J.E. Byrd (Eds.), Recovery, Analysis, and [22] L.L. Cabo, D.C. Dirkmaat, J.M. Adovasio, V.C. Rozas, Archaeology, mass graves, and Identification of Commingled Human Remains, Humana Press, Totowa, NJ, resolving commingling issues through spatial analysis, in: D.C. Dirkmaat (Ed.), 2008 , pp. 257–269. Please cite this article in press as: B.S. Walter, J.J. Schultz, Mapping simulated scenes with skeletal remains using differential GPS in open environments: An assessment of accuracy and practicality, Forensic Sci. Int. (2013), http://dx.doi.org/10.1016/j.forsciint.2013.02.027