4004.p65 ALCES VOL. 40, 2004 DETTKI ET AL. - REAL-TIME MOOSE TRACKING 13 REAL-TIME MOOSE TRACKING: AN INTERNET BASED MAPPING APPLICATION USING GPS/GSM-COLLARS IN SWEDEN Holger Dettki1, Göran Ericsson2, and Lars Edenius2 1Department of Forest Resource Management and Geomatics, Swedish University of Agricultural Sciences (SLU), SE – 901 83 Umeå, Sweden; 2Department of Animal Ecology, Swedish University of Agricultural Sciences (SLU), SE – 901 83 Umeå, Sweden ABSTRACT: To date, moose (Alces alces) tracking has relied on techniques either based on ‘Very High Frequency’ (VHF) / ‘Ultra High Frequency’ (UHF) radio collars, or Global Positioning System (GPS) collars, often requiring significant effort in the field to collect data. Here we present a technique that automatically tracks and reports moose in almost real time, and presents moose positions and movement paths with an interactive web-based map service. We equipped 25 female moose with GPS/GSM collars in Västerbotten county, northern Sweden. The GPS receivers acquired a position every 30 minutes and transmitted them after 3.5 hours as a standard Short Messaging Service (SMS) message using the Global System for Mobile communications (GSM) cell phone network. The positions were automatically extracted from the receiving local GSM-modem and stored in a database. During 18 days in March 2003, 18,638 GPS positions were transmitted by 2,719 SMS messages. Of all positioning attempts 98.1% resulted in a valid position, whereof 99.7% were 3-dimensional positions. The real-time approach allows for many new research studies; e.g., small- scale migrational studies with adapted GPS schedules for different phases of migration. Further, public access to the moose data by a web-based map can be of fundamental importance for public acceptance when dealing with local concerns. ALCES VOL. 40: 13-21 (2004) Key words: Alces alces, boreal forest, collar, GPS, GSM, moose, SMS, Sweden, tracking, ungulate A large fraction of the moose (Alces alces) population in Northern Sweden mi- grates between summer and winter home ranges (Ball et al. 2001), creating locally concentrated problems for forestry. Areas prone to extensive winter browsing by moose are often associated with extensive areas of economically valuable sapling stands of Scots pine (Pinus sylvestris). Access to accurate and up-to-date information is an important requisite for effective co-man- agement of moose and forest resources. In order to mitigate and reduce conflicts aris- ing because of strong browsing pressure in such winter concentration areas, improved knowledge about moose movement and ranging behaviour at the regional and land- scape scale in relation to available winter browse is important. Such insights may greatly facilitate co-operation among hunt- ers, foresters, and managers over manage- ment issues. To date, ungulate tracking used to, for example, study migration and movements of large ungulates has relied either on tech- niques based on traditional radio-telemetry using ‘Very High Frequency’ (VHF)- or ‘Ultra High Frequency’ (UHF) radio col- l a r s ( e . g . , H e e z e n a n d T e s t e r 1 9 6 7 , Hundertmark 1998, Ericsson et al. 2001) or in recent years more and more on Global Positioning System (GPS) telemetry (e.g., Rempel et al. 1995, Moen et al. 1996, Edenius 1997, Hundertmark 1998, Turner et al. 2000). GPS is a US based satellite-aided naviga- tion system that allows the calculation of REAL-TIME MOOSE TRACKING – DETTKI ET AL. ALCES VOL. 40, 2004 14 positions worldwide with about ± 10 m precision. Using a small handheld receiver a position can be calculated by triangulation when signals from at least 3 different of 21 available GPS-satellites are received. Ex- tensive studies have been conducted to evaluate the performance of GPS collars under different environmental conditions (e.g., Moen et al. 1996, Edenius 1997, Moen et al. 1997, Struch et al. 2001). However, existing systems have the disadvantage that significant labour is required in the field for acquiring locations (VHF/UHF) or to ex- tract on-board stored positional data from GPS collars via a local radio-link. The alternative, to download stored data after retrieval of a collar from an animal, always faces the risk of total data loss due to both mechanical collar failures or loss of the animal. Further, visualisation and spatial analysis of the data often require consider- able investments in Geographic Information Systems (GIS) software and education. The aim of this paper is to describe a framework for tracking and displaying moose positions and movement paths in almost real-time with a web-based map service, which pro- vides simple statistics on positions and move- ments. STUDY AREA The study area was located near Åsele in the county of Västerbotten, Sweden (64 °06’N 17°18’E; Fig. 1). The 2,200 km2 area has a winter density of 1.3 moose/km2 with local concentrations up to >12.5 moose/ km2 (Andersson 2002). The climate is predominantly continental with relatively short summers (1 June - 10 Sept.) and the length of the growing season is 150-160 days with an onset between 10 May and 20 May. Average annual temperature is 2- 3°C. The onset of winter is normally the first week of November and lasts to mid- April. Annual precipitation ranges between 600 and 700 mm. The ground is usually snow-covered from the first week of No- vember to the last week of April. Maximum snow depth peaks at 70-80 cm in late Feb- ruary. Climatic data were averaged from the period 1961-1990 (Raab and Vedin 1995). Fig. 1. The county of Västerbotten in Northern Sweden shown with the study area (rectan- gle), the city of Umeå, and the Arctic Circle indicated. ALCES VOL. 40, 2004 DETTKI ET AL. - REAL-TIME MOOSE TRACKING 15 METHODS GPS/GSM Collars We equipped 25 female moose with GPS collars (GPS/GSM Plus 4D) from Vectronic Aerospace GmbH (Fielitz 2003) and uniquely numbered ear-tags during 1 - 5 March 2003. The collars weighed ca 1.1 kg and were designed for a battery lifetime of 1.5 years. We immobilised the moose from a helicopter using a dart gun injecting a mixture of anaesthetic and tranquillizer (ethorphine and xylazine; Sandegren et al. 1987). Moose were aged according to tooth wear (Ericsson and Wallin 2001), which is in agreement with the cementum annuli method <5 years of age (K. Wallin and G. Ericsson, unpublished data). Twelve-channel GPS receivers mounted on the collars acquired a position every 30 minutes and stored them internally for later download. Each collar was equipped with a cell phone unit using the widely available Global System for Mobile communications (GSM) network in Europe, the second gen- eration digital cell phone technology. The unit consisted of a ‘dual band’ module, i.e., it could use the common GSM frequencies in Europe of 900 MHz and 1,800 MHz. After 7 positioning attempts with a time interval of 0.5 hours, the GSM unit was programmed to send these 7 positions each 3.5 hours as a standard Short Messaging Service (SMS) text message to a GSM- modem located at the Swedish University of Agricultural Sciences (SLU) in Umeå, Sweden (63°49’N 20°16’E; Fig. 1). De- spite the fact that the GSM technology does not require a free line of sight towards a GSM relay transmitter, the study area con- tains gaps with low or no coverage due to weak signals or partial signal blackouts in very remote or mountainous areas. Thus, not every SMS message could be transmit- ted immediately after the seventh position was acquired. Therefore the collars were programmed to always send a SMS mes- sage with the latest 7 positions first and then check for unsent positions stored in the collar when within an area with sufficient GSM coverage. If unsent positions were found, up to 10 SMS messages with a total of 70 positions were sent, starting with the latest unsent position. After sending, the GSM module went into a ‘sleep’ mode to preserve battery capacity and tried to send the next 7 positions and eventually the re- maining unsent positions during the next sending opportunity (i.e., 3.5 hours later). After the GSM-modem received a SMS message (Fig. 2), the positions were auto- matically extracted and the co-ordinates converted to the National Grid of Sweden and stored in a SQL-server database. To- gether with the positions, the exact date and time, altitude, dilution of position (DOP), type of position (2-dimensional or 3-dimen- sional), battery voltage, and the internal temperature of the collar were transmitted and stored in the database. After one year, the collars will be re- trieved manually by darting the moose. To locate a collar for retrieval in areas without GSM coverage, collars were additionally equipped with a permanent VHF transmit- ter. After retrieval, all data can be downloaded from the collars. Due to SMS message size limitations not all registered data (e.g., data from the built-in activity sensors for the animals X- and Y-axes) for each position are transmitted by SMS, but stored in the collars for later analysis. Data Retrieval and Presentation Digital topographical landcover maps (original scale 1:100,000) obtained from the Swedish National Land Survey were used as background maps. They were stored on a web server and accessed by the ArcIMS 4.0 engine (ESRI 2002) and the Internet Information server 5.0 (IIS 5.0; Microsoft Corporation 2000). A web application built with the Active Server Page technology REAL-TIME MOOSE TRACKING – DETTKI ET AL. ALCES VOL. 40, 2004 16 Fig. 2. Schematic description of the information flow: Positional information from the GPS-satellites is stored in the collar, and transmitted by the GSM network using SMS messages to a database server to be accessed as a map on the internet. (ASP) and Visual Basic, both in the .NET environment (Richter 2002), extracted the positions for one or more moose and over- laid them on the background maps (Fig. 3a) using the ActiveX connector and ArcIMS (ESRI 2002) as positions or moving paths. Further, position co-ordinates for each re- corded position (Fig. 3b) or simple statistics on path length or numbers of positions were shown as tables. RESULTS AND DISCUSSION GPS Positions and Data Transfer Between 1 - 18 March 2003, 18,638 GPS positions were registered for the 25 moose in the database. Of all positioning attempts in the field, 98.1% were success- ful (Table 1), resulting in 99.7% 3-dimen- sional positions. The dilution of precision (DOP) value as a measure of positional precision (Moen et al. 1997) was < 2.0 for 75.1% of all positions. Based on these figures, the estimated number of valid posi- tions will be ca 15,120 positions per moose and year. All stored position data were transmit- ted by 2,719 SMS messages. Of all SMS messages 95.3% were successfully sent immediately after 7 positions were regis- tered, while 4.7% were delayed because of the moose being outside an area with GSM coverage during the time of transmission (Table 2). Maximum delay time between two consecutive SMS transmissions for a 2-D 3-D Total Fix-attempts — — 19,004 Valid positions 64 18,574 18,638 DOP < 2.0 21 13,974 13,995 Table 1. Number of valid GPS-positions in the database during the first 18 days (1 - 18 March 2003). Data are given for 2-dimensional (2-D) and 3-dimensional (3-D) positions as well as positions with a dilution of position (DOP) <2.0. The number of attempts to obtain a position (fix) is also given for the same period. ALCES VOL. 40, 2004 DETTKI ET AL. - REAL-TIME MOOSE TRACKING 17 Fig. 3a. Example of a movement path for moose ‘432’ on 18 March 2003 near Åsele in the county of Västerbotten, Sweden on the interactive web-site represented (a) as a map and (b) as a table of positions within the movement path. On the map, the circle at one end of the movement path (dashed line) indicated the last transmitted position. In the table the moose name or ID, the local date and time for each position, the northing, easting, and height above sea level (‘RN North’, ‘RN East’, and ‘RN Height’; coordinates in the National Grid of Sweden, in meters), and the dilution of position (DOP) value were given. Further, it was noted in field ‘Nav’ whether the position was 3-dimensional (3D) or 2-dimensional (2D), and a temperature measure (‘Temp’, in degrees Celsius) was given. Name Date Time RN North RH East RN Height DOP Nav Temp 432 2003-03-18 23:30:56 7115322 1580400 311.6 1.2 3D 6 432 2003-03-18 23:00:24 7115313 1580406 323.3 1.4 3D 6 432 2003-03-18 22:30:06 7115320 1580406 335.9 2 3D 6 432 2003-03-18 22:00:26 7115320 1580401 322.5 3.6 3D 6 432 2003-03-18 21:30:49 7115315 1580404 322.5 1.4 3D 5 432 2003-03-18 21:00:20 7115307 1580405 319.6 1.4 3D 5 432 2003-03-18 20:30:07 7115319 1580411 318.5 2 3D 5 432 2003-03-18 20:00:49 7115310 1580411 308.6 3.8 3D 5 432 2003-03-18 19:30:20 7115348 1580418 323.5 1.6 3D 2 Fig 3b. REAL-TIME MOOSE TRACKING – DETTKI ET AL. ALCES VOL. 40, 2004 18 Time (t) #SMS received Fraction at (t) 2,719 100.0% Directly (after 3.5 h) 2,590 95.3% after 7.0 h 74 2.7% after 10.5 h 13 0.5% after 14.0 h 18 0.7% after 17.5 h 16 0.6% after 21.0 h 3 0.1% > 24 h 5 0.2% Table 2. Number and percentage of SMS mes- sages received during the first 18 days (1 - 18 March 2003) for different time intervals be- tween SMS receptions. single moose lasted 137.5 hours. We ex- pect the percentage of delayed messages to increase during the year, as many of the collared moose may eventually migrate to- wards more remote, mountainous areas with a more fragmented GSM coverage. On the other hand, the GPS positioning success rate should remain quite constant, as the vegetation coverage, which is the main ob- stacle for the reception of the satellite sig- nals for position calculation, is similar throughout the study area. Data Access and Presentation All moose data can be accessed by using a web page either as a map with positions and movement paths (Fig. 3a), or as a table of positions (Fig. 3b). Further, as the data are stored on a SQL-database server, the data are available on virtually any computer connected to the Internet. Hence, more sophisticated data analysis can be conducted locally using standard GIS packages. The data were also made accessible to the public on an interactive web page show- ing moose locations and movement paths on a map (http://www-moosetrack.slu.se). Data access, however, was restricted to positions older than 2 weeks and no tables with positional data were given to avoid disturbance of the animals. The application is already heavily used by local residents, media, hunter associations, and forest com- panies for different reasons. Interest of local residents has so far been focused on comparing local knowledge on moose move- ment to actual movement as displayed data on maps on the Internet. Furthermore, several schools in Sweden have started to use the real-time information on the web site in education on wildlife ecology and management. We also anticipate an in- creased interest from the tourist industry, for example, by using the Internet interface to explore the charismatic value of moose to attract visitors, as a wild moose in many countries is recognized as a symbol for undisturbed wilderness. Moreover, we fore- see that forest companies and hunter asso- ciations increasingly will use real-time in- formation on moose activity to facilitate communication to help resolve conflict situ- ations. Lastly, we expect to see an in- creased public interest channelled into projects such as ‘adopting’ a moose. Hence, public interest in real-time moose move- ment may deepen acceptance of moose populations in managed forest areas. Pros and Cons Monitoring of moose movement in almost real-time enables a quick reaction to possible problems with either animals or collars. For example, one of the collared moose stopped moving a few hours after darting and outfitting with a collar, and the collar temperature started to decrease over a period of 12 hours. It was possible to ALCES VOL. 40, 2004 DETTKI ET AL. - REAL-TIME MOOSE TRACKING 19 immediately revisit the animal to check on the health status of the moose. Though the moose was fine, technical problems with the collar were detected and corrected quickly with remote re-configuration to mini- mize data loss. The major advantage of the technology used in this study is the ability to collect a large amount of data in almost real-time with minimal labour required. In a previous p r o j e c t c o n d u c t e d i n t h e c o u n t y o f Västerbotten (Dettki et al. 2003), 15 moose were equipped with GPS collars over a period of 4 years, resulting in 17,036 valid positions. A varying GPS schedule was applied, resulting in 1 position per moose each hour, and up to 1 position per moose a day. The data were available first after the collars had been retrieved manually. In the current study, after only 18 days, 18,638 positions were recorded for 25 moose, ap- plying a GPS schedule with 1 position per moose and 0.5 hours. While data handling in the previous study was done manually, in the current project it is necessary to use automated routines to handle the relatively large amount of data. Only basic supervi- sion of the servers is required to handle the incoming data from the download of a SMS message from the GSM-modem, conver- sion of positions into local grid format, error check of both SMS and positional data, storage in the SQL-database and finally presentation on a map. This results in high efficiency in data retrieval and manipulation and very low costs per position, as the number of received positions is high and labour required is low. However, some disadvantages with the technology exist. First, the amount of data transmitted by SMS is mainly constrained by the allowed size of each SMS, which is internationally standardized to a maximum of 160 characters per message. This pro- hibits transmission of additional data by SMS, such as activity and mortality data. Furthermore, SMS transmission drains bat- tery power and is costly, so there are trade- offs to be made. Even though the costs per position transmitted are small (approximately US $0.02 per position with 7 positions in each SMS message), during 1 year the costs for data transmission add up to ap- proximately US $350.00 per collar with a position taken every 0.5 hours. Battery capacity continues to be a bottleneck of GPS/GSM-systems, especially for long-du- ration studies or small-sized animals. The same problems arise for transmitting data for differential post-processing, which would increase the amount of necessary SMS messages to be sent at least 3-fold. How- ever, with a precision of ± 10 m with uncor- rected positions, most applications likely do not have the need of differential real-time corrections. The far greatest disadvantage is the limitation due to the fragmented or non- existing coverage of the GSM network in remote areas in many parts of the world. The Scandinavian countries differ in this respect from, for example, North America or Russia. In Sweden, about 75% of the land area is covered by at least one GSM network (TeliaSonera 2003). However, even though all bigger roads and settle- ments in northern Sweden have at least some degree of GSM coverage, coverage is poor in more remote areas. We tried to reduce this problem by programming the collars to deliver all stored positions at least once by SMS when the animal is within or re-enters into an area with sufficient GSM coverage. During the first 18 days, all positions stored on-board were also retrieved by SMS, even though some moose collars were out of contact with the GSM- network for up to 137.5 hours. It is impor- tant to note that SMS transmissions require less net coverage than is required for oral communication by cell phones or GSM data transfer. This means that the tech- REAL-TIME MOOSE TRACKING – DETTKI ET AL. ALCES VOL. 40, 2004 20 nique is applicable also in areas where it is not possible to use a cell phone for talking, due to fragmented GSM coverage. Implications The GPS/GSM-technique gives a new possibility to ‘supervise’ animals in almost real-time. For research this means one can see when, e.g., the migration starts, and then via SMS, change the GPS schedule scheme to acquire positions more frequently. It further opens up the possibility to deter- mine in real-time when migration is over, i.e., when moose have reached their final destination. Wildlife managers then can use lethal or non-lethal methods to immedi- ately break up concentrations of moose if they are in conflict with other forestry man- agement goals in the area, for example, sapling stands of regenerating Scots pine. For researchers, the improved GPS/GSM- technique will mean less ad-hoc testing and less guess work in terms of what periods are interesting for data collection to test behavioral hypotheses. There is a great potential for time (habitat budget), distur- bance studies, detailed movement, and habi- tat studies. The notion that the GPM/GSM-tech- nique is only possible to use in Europe or other densely populated areas is not neces- sarily true. The GSM-system is expanding in Canada and the US, and moose populations near some urban settlements are growing. We suggest that a few moose equipped with GPS/GSM collars which report to an inter- active web-page will have fundamental im- portance; e.g., public acceptance when re-introducing moose. It will furthermore be of great help to make research more acceptable and stimulate public interest in wildlife management. ACKNOWLEDGEMENTS We want to thank Dipl.-Ing. Robert Schulte from Vectronic Aerospace GmbH for hard- and software development and Mats Högström for assistance on ArcIMS issues. K.G. Poole and an anonymous reviewer improved the manuscript with con- structive comments. The project is financed by the County Board administration of Västerbotten, the forestry owners associa- tion, and the Swedish Association for Hunt- ing and Wildlife Management. REFERENCES ANDERSSON, E. 2002. Älginventering från flyg i Borgsjö 2002. Länsstyrelsen i Västerbottens län. http://www.ac.lst.se/ jakt/algjakt/alginventering/. BALL, J. P., C. NORDENGREN, and K. WALLIN. 2001. Partial migration by large ungu- lates: characteristics of seasonal moose Alces alces ranges in northern Swe- den. Wildlife Biology 7:39-47. DETTKI, H., R. LÖFSTRAND, and L. EDENIUS. 2003. Modeling habitat suitability for moose in coastal northern Sweden: empirical vs. process-oriented ap- proaches. AMBIO 32:549-556. EDENIUS, L. 1997. Field test of a GPS location system for moose Alces alces under Scandinavian boreal conditions. Wildlife Biology 3:39-43. ERICSSON, G., K. WALLIN, J. P. BALL, and M. BROBERG. 2001. Age-related repro- ductive effort and senescence in free- ranging moose Alces alces. Ecology 82:1613-1620. _____, and K. WALLIN. 2001. Age-specific moose Alces alces mortality in a preda- tor free environment: evidence for se- nescence in females. Ecoscience 8:157- 163. (ESRI) ENVIRONMENTAL SYSTEMS RESEARCH INSTITUTE. 2002. Using ArcIMS. En- vironmental Systems Research Insti- tute Incorporated, Redlands, California, USA. FIELITZ, U. 2003. Remote GPS-data trans- mission via mobile phone. http:// ALCES VOL. 40, 2004 DETTKI ET AL. - REAL-TIME MOOSE TRACKING 21 www.environmental-studies.de/prod- ucts/02/gps-gsm_collars.html. HEEZEN, K. L., and J. R. TESTER. 1967. Evaluation of radio-tracking by triangu- lation with special reference to deer movements. Journal of Wildlife Man- agement 31:124-141. HUNDERTMARK, K. J. 1998. Home range, dispersal and migration. Pages 303-336 in A. W. Franzmann and C. C. Schwartz, editors. Ecology and Management of t h e N o r t h A m e r i c a n M o o s e . Smithsonian Institution Press, Wash- ington, D.C., USA. MICROSOFT CORPORATION. 2000. Microsoft Internet Information Services 5.0 Docu- mentation. Microsoft Press, Redmond, Washington, USA. MOEN R., J. PASTOR, and Y. COHEN. 1997. Accuracy of GPS telemetry collar loca- tions with differential correction. Jour- nal of Wildlife Management 61:530- 539. _____, _____, _____, and C. C. SCHWARTZ. 1996. Effects of moose movement and habitat use on GPS collar performance. Journal of Wildlife Management 60:659- 668. RAAB, B., and H. VEDIN, editors. 1995. Sveriges nationalatlas. Klimat, sjöar och vattendrag. SNA, Stockholm, Swe- den. REMPEL R. S., A. R. RODGERS, and K. F. ABRAHAM. 1995. Performance of a GPS animal location system under boreal forest canopy. Journal of Wildlife Man- agement 59:543-551. RICHTER, J. 2002. Applied Microsoft .NET Framework Programming. Microsoft Press, Redmond, Washington, USA. SANDEGREN, F., L. PETTERSSON, P. AHLQVIST, and B. O. RÖKEN. 1987. Immobilization of moose in Sweden. Swedish Wildlife Research Supplement 1:785-791. STRUCH, M., C. ANGST, and R. EYHOLZER. 2001. Möglichkeiten und Grenzen des Globalen Positionierungssystems. Ein Methodenvergleich der GPS- und VHF- Telemetrie im Gebirgswald. WILDARK & K O R A . V e r e i n e f ü r w i l d t i e r - biologische Forschung. Bern, Switzer- land. TELIASONERA. 2003. Annual Report 2003. h t t p : / / v p 0 3 4 . a l e r t i r . c o m / teliasonera_annualreport_2003en/. TURNER, L. W., M. C. UDAL, B. T. LARSON, and S. A. SHEARER. 2000. Monitoring cattle behavior and pasture use with GPS and GIS. Canadian Journal of Animal Science 80:405-413.