Coverage maps are commonly associated with analyses of the availability of mobile telephony. These maps display areas according to the power of the available telecommunications signal, but this is not the only application. Various map solutions are implemented in the field of health and safety: for example, coverage maps can be created for the analysis of immunization areas  or the degree of patient choice and competition between Adult Congenital Heart Disease Centers, based on a feasible travel time of 120 minutes . A similar case of hour service areas for hospital are presented in . Another example is presented in , which describes a solution that provides the geographic information needed to support real-time routing of alerts for accidents to the nearest air medical transport service. This paper describes the authors’ development of the Atlas Database and the Air Medical Services Database. The map shows circular areas that can be reached by aircraft in 10, 20 or 30 minutes. Coverage maps can be also treated as a component of the Smart City solutions .
In the context of ambulance travel, some examples of coverage maps can be also given. In , polygon map overlays showing 8- and 15-minute arrival times for the city of Haifa are presented. Time-based Ambulance Zoning was presented in , and  presents simulation-based probability-of-coverage maps that demonstrate the probability that the nearest available ambulance can respond to a call. Probability-of-coverage maps are also mentioned in . Another strategy to reduce the average response time from the initial call to the arrival of ambulances in Singapore using geospatial-time analysis is presented in  and are presented as coverage maps. Analysis of ambulance coverage within 10 minutes in Samsun, Turkey is the scope of .  shows a solution which makes it possible to plan ambulance distribution to maximize the coverage efficiency for a given number of ambulances in any particular area in Johor Bahru (Malaysia). In , a predicted future call coverage map is discussed.  presents another solution in which maps display analyses of the fastest routes to incidents and the average response time with associated serviceability in North Dakota, USA. A computational model of demand, bases and distribution of hospitals in Melbourne is presented in . An analysis of rural area coverage by EMS in Western Cape, South Africa is presented in . The ambulance deployment and demand coverage for Odunpazari District (Turkey) is shown in  with a 5 minute response time as the goal. In , the probability of the nearest ambulance arriving within 4 minutes is presented for Toronto using the city road grid as the reference map. The general issue of ambulance redeployment has also been covered by numerous additional articles [12, 15, 19, 20, 21, 22, 23, 24, 25].
This papers outlines the construction and determination of ambulance coverage mapping that can be updated in real time. For a given area (represented by a properly prepared grid), it graphically represents the shortest driving time of a ambulance. The aforementioned solution was designed in 2015 for the Malopolskie Voivodship Office ; and includes not only static analysis of sample stationary ambulance patterns, but also real-time dynamic analysis of emergency situations.
The basic element of the project is the map. The WGS84 reference system is used for all visualizations and calculations. The project was limited to the map of the Malopolskie Voivodship. The most important issue here is the representation of the area in context of the visualization of the potential travel times of ambulances. In [27, 28], the analyzed area is represented as a grid of intersections; in , a coverage map defines travel times from general medical practitioners (a list of 11,631 postal codes) to ten ACHD (Adult Congenital Heart Disease) centers. The project described in this article assumes representation of an area as an orthogonal grid whose points are 250 m apart in urban areas, while 1000 m intervals are considered sufficient in rural areas. The grid G is constructed as a list of points Pi, for which each is assigned coordinates and a unique identifier.
Pi - point of grid,
G - grid of points (set of points).
This solution allows for flexible representation of an area, with high-density classification in urban areas (Fig. 1, e.g. Cracow and neighborhoods) and low-density classification in areas where points are meaningless (e.g. lakes). Selected points can be moved to other representative locations, and special points such as hospitals or known fixed ambulance stations can be added.
The locations of stationary ambulances A on standby are essential and are rounded to the closest grid point of G. In order to increase the accuracy of the point grid, the appropriate ambulance stationing points designated by the Voivodship Office were added. In Malopolskie Voivodeship, the number of ambulances is variable and was 119 at the time the project was created.
A - set of ambulances,
Ai - current location point of ambulance i, Ai is equal to Pj (closest point of grid G).
2.2 Determining the pre-calculated times of ambulance journeys
In order to determine the ambulance journey times for the entire coverage map in real-time, a map of pre-calculated travel times (PT) was created that shows travel times between any two points of the grid in both directions [29, 30, 31, 32].
where PT[x, y] - travel time from Px to Py.
Determining travel times for a map of pre-calculated values is time-consuming because of the Cartesian product of all possible routes, therefore limits are imposed that allow irrelevant calculations to be rejected . A custom plug-in for MapPoint was used for this project; calculations were then made on a cluster of 30 computers with Intel Core i7 processors, which took a several days to complete a map of the city of Krakow (distance 250 m) and Malopolskie Voivodship (distance 1000 m).
2.3 The model of ambulance velocities
One very important issue is to choose the general model of velocities that are characteristic for ambulances . The typical driving model used by Google Maps and MapPoint determines the characteristic speed values for five types of roads:
Limited access roads,
Other (major) roads,
Arterial (minor) roads,
The standard values for vehicles complying with standard traffic rules are determined by the vector (104, 96, 80, 56, 32) [km/h]. However, in the case of ambulances, these speeds are considerably higher, and so based on data collected from real-world systems, the value of these speeds were determined. For this, a system for calculating the inverse problem was created . A Monte Carlo algorithm was used to match the velocity values of the simulated ambulance journeys that were closest to historical data. In this way, a velocity vector (145, 121, 82, 71, 38) was compiled with a compensation error of 13%. Further consideration is needed however to determine whether this model is sufficient to describe traffic regardless of time of day, or day of the week or month.
2.4 Verification of the model based on instantaneous speeds
Creating an ambulance traffic model is not an easy issue. There are few works that theoretically or empirically estimate this form of vehicle movement model [8, 9, 18, 27]. The primary problem is determining whether periodic increases in road traffic actually affect travel times  with or without traffic signal control [37, 38]. The author used data obtained from the Malopolska Office which contain approx. 450,000,000 records describing the time, current position of the ambulance, its status and instantaneous speed. These data were collected in the period 2013-2017 (2017: I-VI) from GPS devices mounted in ambulances. Records were generated every 2 minutes when stationary and every 30 seconds when moving (technical or call-related). These data were analyzed for daily fluctuations of traffic congestion with an accuracy of one hour, and graphs of mean instantaneous velocities were constructed. Also shown were the days of the week, months, and variations in subsequent years. The research outlines are shown in the graphs (Fig. 2, 3).
Analysis of the results allowed many intriguing conclusions to be drawn. The most important conclusion is that despite peak traffic, ambulance speeds on emergency callouts do not significantly change (max 2-4 km/h, in opposite to Shenzhen, ). This can be explained by the use of special bus lanes in cities and the culture of the drivers on the roads. This explanation also confirms the result that during technical trips (i.e. non-emergency), there is a significant reduction in ambulance speeds at peak road usage hours. Moreover, ambulances are not slowed down by traffic jams at weekends (Saturday and Sunday). Annually, there is a slight reduction in ambulance speeds (4km/h) in the winter. However, by comparing the speed of emergency trips to technical trips it can be concluded that one general ambulance model can be used if small deviations in speed are accepted.
Constructing the arrival coverage map
To calculate the coverage map CM, each grid point Pi should have calculated the minimum travel time from all standby ambulances:
Ai - current location point of ambulance i,
Pi - grid point i.
As a result, the calculated values for each grid point can be applied to the base map by giving the grid coverage of ambulances with the lowest call times. The coverage can be represented with color, e.g. the example scale uses shades of blue for times of 0-10 minutes; green for 11-15 minutes; yellow for 16-20 minutes; and red or black for more than 20 minutes. This is related to statutory travel times of 15 minutes within the city and 20 outside the city.
The illustration shows an example of an arrival time coverage map for Cracow (Fig. 4, where two ambulances are on standby: one in the northwest and one in the south. Attention should be paid to the nonlinear nature of such coverage maps, as the journey time for points near motorway exits is significantly shorter than that for points in other areas. The coverage map also illustrates that the northeastern part of the city remains uncovered by emergency medical services within the 20 minute statutory time limits.
The coverage designated in the manner described above varies according to the type of urban area: densely built up with a regular and extensive road network (Figure 5(a)); adjacent to main roads with many junctions (Fig. 5(b)); and mountainous (valley) (Fig. 5(c)). Therefore, coverage maps could be of great use to local authorities when creating ambulance station mapping or in crisis situations. The system also implements additional layers related to population density and frequency of accidents based on historical data .
3.1 Real-time application
The calculation time of the coverage map in the tested model, with 21448 points covering the city of Krakow with an accuracy of 250 m and the remaining areas of the province with an accuracy of 1000 m, takes about 40 ms on a typical computer with Intel i5 Core CPU. A longer processing time may be caused by preparing a graphic view of map, but such tasks can be performed on the GPU. Therefore, the low requirement for computing power allow realtime depiction of the coverage of ambulance services. Fig. 6 presents a fragment of the map depicting a vehicle drive for two time moments (point A and point B). It is noticeable that the 10-minute access area (blue) is moving with the vehicle position. Such an application has been prepared for implementation by ambulance units of the Malopolskie Voivodeship Office.
This article describes a novel approach which supports ambulance fleet planning, management, and redeployment. The proposed solution makes it possible to present service coverage of urban and rural areas by EMS and produces real-time coverage maps that can be used by dispatch or crisis management centers.
The construction details of the proposed system are described in the article. An interesting issue related to the ambulance model based on instantaneous speed was discussed. The results of matching the characteristic velocities used by the network calculation systems are presented. Based on the collected speed data, it was determined that in the administrative area of Malopolskie Voivodeship the speed of ambulances traveling on call is not significantly affected by traffic jams. This proves that one general coverage map is a sufficient approach.
The first implementation of the system took place in 2015 and it was submitted for the Malopolskie Voivodeship Office as a tool supporting the Medical Rescue and Sanitary Unit of the Social Policy Department.
A sample application demo is available on the site:
The author would thank to Governor J. Miller; M. Lechowicz, S. Janusz and S. Choroba from Malopolskie Voivodeship Office; F. Banas from KPR and supporting team of M. Lupa, K. Szostek, I. Plokita, P. Lukasik, T. Pazdalski and R. Rumanek.
This work was financed by the AGH - University of Science and Technology, Faculty of Geology, Geophysics and Environmental Protection as a part of statutory project.
Kamadjeu R, Tolentino H (2006) Web-based public health geographic in- formation systems for resources-constrained environment using scalable vector graphics technology: a proof of concept applied to the expanded program on immunization data. International Journal of Health Geographics 5(1):24 CrossrefGoogle Scholar
Diller GP, Kempny A, Piorkowski A, Grübler M, Swan L, Baumgartner H, Dimopoulos K, Gatzoulis MA (2014) Choice and competition between adult congenital heart disease centers. Circulation: Cardiovascular Quality and Outcomes 7(2):285-291 Web of ScienceGoogle Scholar
Schuurman N, Fiedler RS, Grzybowski SC, Grund D (2006) Defining rational hospital catchments for non-urban areas based on travel-time. International Journal of Health Geographics 5(1):43 CrossrefGoogle Scholar
Flanigan M, Blatt A, Lombardo L, Mancuso D, Miller M, Wiles D, Pirson H, Hwang J, Thill JC, Majka K (2005) Assessment of air medical coverage using the atlas and database of air medical services and correlations with reduced highway fatality rates. Air medical journal 24(4):151-163 CrossrefGoogle Scholar
Chmielewski S, Samulowska M, Lupa M, Lee D, Zagajewski B (2018) Citizen science and WebGIS for outdoor advertisement visual pollution assessment. Computers, Environment and Urban Systems, 67, 97-109 CrossrefWeb of ScienceGoogle Scholar
Shuib A, Zaharudin ZA (2010) Framework of tazopt model for ambulance location and allocation problem. World Academy of Science, Engineering and Technology 70:678-683 Google Scholar
Budge S, Ingolfsson A, Zerom D (2010) Empirical analysis of ambulance travel times: the case of calgary emergency medical services. Management Science 56(4):716-723 CrossrefWeb of ScienceGoogle Scholar
Ingolfsson A (2013) Ems planning and management. In: Operations Re- search and Health Care Policy, Springer, pp 105-128 Google Scholar
Ong MEH, Chiam TF, Ng FSP, Sultana P, Lim SH, Leong BSH, Ong VYK, Tan C, Ching E, Tham LP, et al (2010) Reducing ambulance response times using geospatial-time analysis of ambulance deployment. Academic Emergency Medicine 17(9):951-957 CrossrefWeb of ScienceGoogle Scholar
Terzi O, Sisman A, Canbaz S, Dündar C, Peksen Y (2013) A geographic information system-based analysis of ambulance station coverage area in Samsun, Turkey. Singapore Med J 54(11):653-658 Google Scholar
Azizan MH, Lim CS, Hatta WALWM GT, Teoh S (2013) Simulation of emergency medical services delivery performance based on real map. International Journal of Engineering and Technology 5(3):2620-2627 Google Scholar
Mason AJ (2013) Simulation and real-time optimised relocation for improving ambulance operations. In: Handbook of Healthcare Operations Management, Springer, pp 289-317 Google Scholar
Maxwell MS, Ni EC, Tong C, Henderson SG, Topaloglu H, Hunter SR (2014) A bound on the performance of an optimal ambulance redeployment policy. Operations Research 62(5):1014-1027 CrossrefWeb of ScienceGoogle Scholar
Vanderschuren M, McKune D (2015) Emergency care facility access in rural areas within the golden hour?: Western cape case study. International Journal of Health Geographics 14(1):5 CrossrefWeb of ScienceGoogle Scholar
Swalehe M, Aktas SG (2016) Dynamic ambulance deployment to reduce ambulance response times using geographic information systems: A case study of Odunpazari District of Eskisehir Province, Turkey. Procedia Environmental Sciences 36:199-206 CrossrefGoogle Scholar
Westgate BS, Woodard DB, Matteson DS, Henderson SG (2016) Large- network travel time distribution estimation for ambulances. European Journal of Operational Research 252(1):322- 333 CrossrefWeb of ScienceGoogle Scholar
Maxwell MS, Henderson SG, Topaloglu H (2009) Ambulance redeployment: An approximate dynamic programming approach. In: Winter Simulation Conference, Winter Simulation Conference, pp 1850-1860 Google Scholar
Sasaki S, Comber AJ, Suzuki H, Brunsdon C (2010) Using genetic algorithms to optimise current and future health planning-the example of ambulance locations. International Journal of Health Geographics 9(1):4 Web of ScienceCrossrefGoogle Scholar
Bernas M, Wisniewska J (2013) Quantum road traffic model for ambulance travel time estimation. Journal of Medical Informatics & Technologies 22:257-264 Google Scholar
Simić D, Simić S (2012) Hybrid artificial intelligence approaches on vehicle routing problem in logistics distribution. Hybrid Artificial Intelligent Systems pp 208-220 Google Scholar
Simić D, Kovaćević I, Svirćević V, Simić S (2015) Hybrid firefly model in routing heterogeneous fleet of vehicles in logistics distribution. Logic Journal of the IGPL 23(3):521-532 CrossrefWeb of ScienceGoogle Scholar
Trebacz I (2015) Karetki dojada szybciej (ambulances will arrive more quickly). In: Biuletyn AGH, vol 11, AGH, p 33, URL http://www.agh.edu.pl/en/blog-naukowy/info/article/ambulances-will-arrive-more-quickly-scientific-blog/
Westgate BS, Woodard DB, Matteson DS, Henderson SG (2013) Travel time estimation for ambulances using bayesian data augmentation. The Annals of Applied Statistics pp 1139-1161 Google Scholar
Lewandowski M, P laczek B, Bernas M (2017) Decentralized control of traffic signals with priority for ambulances. Journal of Medical Informatics & Technologies 26:17-25 Google Scholar
Koziel G (2014) Algorytmy wyznaczania optymalnej trasy przejazdu [Algorithms for optimal route determine]. Logistyka 3:3206-3212 Google Scholar
Mitosz M, Ztomaniec P, Badurowicz M (2014) Modele matematyczne optymalizacji tras w tranporcie medycznym [Mathematical models of route optimization in medical transport field], Logistyka, 6(2014), 7524-7533. Google Scholar
Cichociński P, Dębińska E (2012) Baclanie dostępności komunikacyjnej wybranej lokalizacji z wykorzystaniem funkcji analiz sieciowych [Accessibility study of a selected location using network analysis functions]. Roczniki Geomatyki-Annals of Geomatics, 10(4(54)), 41-48. Google Scholar
Lewandowicz E, Flisek P (2017) Dostępność komunikacyjna w analizach sieciowych w przestrzeniach heterogenicznych [Communication availability in network analysis in heterogeneous spaces]. Roczniki Geomatyki-Annals of Geomatics, 15(4(79)), 375-389. Google Scholar
Płokita I, Piórkowski A, Lupa M, et al (2016) Comparative analysis of algorithms for calculating arrival times of emergency vehicles. Geoinformatica Polonica 15:85-91 Google Scholar
Fisher R, Lassa J (2017) Interactive, open source, travel time scenario modelling: tools to facilitate participation in health service access analysis. International Journal of Health Geographics 16(1):13 CrossrefWeb of ScienceGoogle Scholar
Lukasik P, Piórkowski A (2016) Development of ambulance speed characteristics based on actual data. Studia Informatica 37(1):113-124 Google Scholar
Amram O, Schuurman N, Hameed SM (2011) Mass casualty modelling: a spatial tool to support triage decision making. International Journal of Health Geographics 10(1):40 CrossrefWeb of ScienceGoogle Scholar
Lewandowski M, P laczek B, Bernas M (2017) Self-organizing traffic signal control with prioritization strategy aided by vehicular sensor network. In: IFIP International Conference on Computer Information Systems and Industrial Management, Springer, pp 536-547 Google Scholar
Bernaś M, P laczek B, Porwik P, Pamuła T(2014) Segmentation of vehicle detector data for improved k-nearest neighbours-based traffic ow prediction. IET Intelligent Transport Systems 9(3):264- 274 Web of ScienceGoogle Scholar
Jenelius E, Koutsopoulos HN (2017) Urban network travel time prediction based on a probabilistic principal component analysis model of probe data. IEEE Transactions on Intelligent Transportation Systems Web of ScienceGoogle Scholar
About the article
Published Online: 2018-06-11
Citation Information: Open Geosciences, Volume 10, Issue 1, Pages 167–173, ISSN (Online) 2391-5447, DOI: https://doi.org/10.1515/geo-2018-0013.
© 2018 Adam Piórkowski, published by De Gruyter. This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. BY-NC-ND 4.0