USING OF TRANSPORT NETWORK MODEL TO ESTIMATE TRAVELLING TIME AND DISTANCE FOR GROUND ACCESS A FOREST FIRE

Since regional forest protection services often have limited material resources, the emergency response to the emerging forest fires requires to choice an optimal maneuvering solution and a method to transfer available forces. One of the possible ways is to create a regional transport model for the case of forest fire based on the network of public roads and forest glades. Paper describes a method of calculation of travelling time and distance to a forest fire, research results for an experimental transport model, created by Network Analyst ArcGIS, to build the shortest routes from the fire stations to the forest fires. Spatially-distributed data on the fire trucks’ average speed for different types of roads and the elevation values were used in the model for the test area (Irkutsk region of Russia). In total 16251 routes were built and analyzed for 16 years (2002-2017). The model was validated using the data on forest fires detected by the MODIS-Aqua/Terra spectroradiometer within the ground and forest aviation zones of test region. A map showing the fire routes within one-, twoand threehours ground transport accessibility is created for the forest fire ground protection zone of the test region. The model’s work quality was validated for the forest fires detected within the ground zone. As a result, 98% (2661) of forest fires in the ground zone are accessible within three hours and less, that indirectly confirms the correctness of model. At the same time, the majority of forest fires are located within one(68%) and two(24%) hour’s availability. Finally, recommendations on using the transport model for the managerial decisions on the forest fire fighting on regional level were given.

Forest fires, annually occurring in different regions of Russia, cause significant damage to the country's economy. Besides to their timely detection, there is a task of promptly delivery of fire brigade and special equipment to the forest fire. Under the conditions of high and extreme forest fieriness, choice of a method (ground or aviation) to deliver the forces and equipment from a fire-and-chemical (PHS -Russian acronym for Fire Station) station to the forest fire is an actual problem in the forest fire control at the regional level.
Methods of mathematical modeling are used to solve the task to design a forest transportation network and access optimization for the forest areas (Khodakov, Zharikova, 2011;Runova, Kostyaev, 2012;Tarantsev, 2013;Faraonov, 2014). Modeling is conducted by the well-known office programs, as well as with the help of geo-information systems tools. We note that the important issue of defining the protection zones boundaries was discussed in scientific publications in a debatable way (Tarantsev, 2015), where Microsoft Excel tools were used for the analysis.
Integration of transport spatial models and remote sensing data (Earth remote sensing) on forest fires in modern geographic information systems can be a new level of development in this area. GIS technologies allow designing traffic routes in space, evaluating a set of parameters for decision making when organizing forest firefighting (Loupian et al., 2006) and logistics (Abousaeidi et al., 2016b;Feng et al., 2017).
As it was noted in the paper (Kotelnikov et al., 2017), insufficient funding of forest fires detection and organization of the forest firefighting in Russia practically means the necessity of resources dispatching. This article also underlines the need for retrospective data collection and updating the knowledge base on the access to forest fires. Present study is a development of this direction and is carried out by both creating a multi-year archive of access routes and its thematic analysis.
The problem to organize the operational transfer of forces and means to the forest fire sites is still actual for many "forest" countries, taking into account the national specifics of forest management, forest growing conditions, local climate, terrain and road network features. We can mention among foreign works the following articles (Kumar et al., 2005;Akay et al., 2012;Akay, Aziz, 2015) on the similar topic.
In accordance with Russian Forest Code, administrative units of Russia are engaged in the detection and suppression of forest fires. Some certain regulations, rules and terminology that have been formulated on the federal and regional levels over the past decade to describe the organization of access to the forest fire sites, for example, (On approval of the classification of natural forest fire danger, 2011).
There is one valid document (Recommendations on the use of forces and technical means, 2014), among the other existing regulatory documents, which states that access to the detected forest fire should be organized within three hours from the moment of its detection.
Namely, paragraph 48 of the guidelines recommends that in order to ensure the prompt elimination of forest fire a forest firefighting division should arrive at the forest fire place and start its extinguishing work: (a) if Forests classified as Class I of the natural fire danger -no later than one hour after the forest fire detection; (b) if Forests classified as Class II of the natural fire danger -no later than two hours after the fire detection; (c) if Forests assigned to the III -V classes of the natural fire danger -not later than three hours after the fire detection.
Important aspects of collaboration between the ground and aviation forces were investigated in the following papers (Bryukhanov, 2011;Bryukhanov et al., 2017), among the topics covered there are an urgent need for state funding, and the importance of interaction between the aviation crews and ground forces is highlighted when organizing work by aviation or ground methods.
The purpose of present study is to make an estimation of travelling time and distance from the nearest fire station to a detected forest fire while driving a special vehicle (fire truck, tank truck, patrol car, etc.) along the public and special roads, including fire roads and forest glades, within the ground protection zone by the example of the Irkutsk region.
To do so the following tasks have been formulated: -Creation of ground forest fire transport model for the test region; -Creation of access routes from fire stations to the forest fires, where the trucks can be delivered in less than three hours; -Creation of map of forest fire sites distribution within one-two-three hour's accessibility in ground and aviation zones (data of protection zones boundaries are of 2017).

MATERIALS AND METHODS
Study included the stages to create the region's transport network using ArcGIS tools and extensions, then to build a forest fire transport model for ground access using the data on public roads, forest glades, fire stations, and forest fire sites. At the final stage, shortest routes to forest fires are being created based on proposed model, then the calculation of travelling time, taking into account the elevation values and putting the routes into a database.
Thus, we have implemented the following sequence "Fire station (point) -Access route (polyline) -Forest fire (point)". A scheme which includes all steps of data processing and analysis is shown on the Fig. 1. Irkutsk region has been chosen as test area because of its high annual forest inflammability during the fire hazard season (Goldammer et al., 2003;Runova, Dolenko, 2006;Ponomarev et al., 2017). As it was shown in the paper (Kaibicheva, Kaibichev, 2013), Irkutsk region is ranked 6-2-3-8-9 on the Dow Jones forest fire index (number of forest fires) within the time interval of 5 consecutive years (2006)(2007)(2008)(2009)(2010), in which 30 Russian regions were included in the selection. These statistics indicate the relevance of choosing such region as a typical example.
For the test region we used geodatabase with settlements of point and polygon types, road network (including forest glades), hydrographic objects and administrative division of 1:200 000. Datasets of such scale are characterized by necessary details and spatial extent to perform a regional level study, which allows to organize access in the emergency situations (Oleynikov, Markov, 2014).  We have used ArcGIS extensions for the calculations: 3D Analyst and Network Analyst. Network Analyst based on the Dijkstra's algorithm (Dijkstra, 1959), was used to create a transport model. This algorithm is shown on graphs and allows finding the shortest paths from one of the graph nodes to all the others. The algorithm is widely used in the transport studies (How to create…, 1996;Ni et al., 2014), there are some modifications, for example, (Alazab et al., 2011).
According to the Dijkstra's algorithm we named created access routes the shortest ones. Field (obtaining the numerical characteristics of the polyline). Statistical estimation of the number of forest fires within the ground and aviation zones was carried out using the methods of mathematical statistics.  As we can see from the Fig. 4, combination of roads with the forest glades layer can expand the forest trucks' access to the forest. The combination of data on the two types (roads and glades) allows to simulate the options for moving on all the roads and to differentiate the movement of special transport, depending on the classes and conditions (seasonality) of their use.

THEORY AND CALCULATIONS
Following input parameters were used to create a transport model: Taking into consideration data on the altitude allows to correct (decrease) the vehicle's speed in the mountainous conditions. Each road (or glade) is a polyline consisting of elementary straight lines (or segments) between two adjacent points. Polylines were converted into points, a height value from the elevation raster was added to each of the points.
To adjust the speed on every segment of road/glade we have used the slope coefficient R: where: Erv, Srvelevation values from ETopo2 for the first (beginning) and last (ending) points of polyline, m; Llength of polyline, m. Then R-coefficient was added to the attributes of polylines, so we can calculate the speed among the full length of road/glade: (2) where: Vcspeed with 3D value included, km per hour, Vnormative speed of special vehicle, km per hour.
Calculation of R-coefficient for the roads and glades showed that 30% of their length within the studied data set in the Irkutsk region has a slope factor close to "1". The rest of the road network requires a speed adjustment depending on the terrain slope, so this confirms the need of 3D-adjustment procedure. The speed is adjusted downwards depending on the relief drop (ascent or descent) on the route.
At the final stage of transport model construction we calculate travelled time taking into account 'adjusted" speed with the formula: where: Tmmovement time время with 3D value, minutes; L3dlength with 3D value, km.
Thus, the transport model contains all the necessary information about roads, speed and travel time on each segment of the road/glade, taking into account the terrain slope.
Calculations were done in ArcGIS Desktop.
In Fig. 5 shows the model of the road network within the boundaries of ground protection and forest aviation zones with the speed distribution with the slope coefficient (3D value). Based on the collection of routes to fires in the database, a statistical analysis of the fires number within one, two, three and more than three-hour intervals is performed (Recommendations on the use of forces and technical means, 2014). According to these recommendations, one of the requirements when designing a ground protection zone is the arrival time limitation, namely, not more than three hours from the forest fire's detection moment. Thus, assuming that the ground zone is designed correctly, the correctness of setting the transport model's parameters is indirectly verified. If the majority of fires registered in the ground protection zone would have the access time of less than three hours, then we could assume that the model works close to the real situation.

DISCUSSION
As an implementation of the methodology, described above, the shortest access routes were created for the archive of forest fires, detected in 2002-2017 within the ground and aviation protection zones. Access routes were created under condition that fire stations and detected forest fires were located within these two zones. Every fire hazard season has its assessment of route's travelled length and transportation time. Fig. 6 shows distribution of 16251 access routes. As can be seen from the Fig. 6, the minimum number of routes (236) was built for the data of 2004, the maximum (3297)is for 2003, which indicates important differences in the number of detected forest fires from different years and their accessibility by ground means.
We note that the number of days during which forest fires occurred during the fire hazard season varies from 170 to 213 days, so the range is more than one month (43 days). The analysis of travelled time and length is given in the Table. 2 Distribution number of routes registered within the ground protection zone based on the time limit of three-hour interval (absolute values and percentage) is presented in the Table 3.
As can be seen from the Table. 3, most of the fires (2661 or 98.5%) registered within the ground protection zone are accessible in three hours. This indirectly confirms the correctness of transport model's settings and shows that the ground protection zone was designed with taking into account the possible arrival to the forest fire within three hours. At the same time, the majority of forest fires are located at hourly (1754 or 68.2%) and two-hour (744 or 24.3%) availability from the fire stations. If we look at the ratio of all routes (16251) created from the fire stations to the forest fires in relation to the ground protection and forest aviation zones, then most of the fires (Fig.7)    This is also confirmed by the statistics of routes distribution within a three-hour interval (Table 4). Calculations took into account the forest fires location relative to the zones to verify the possibility to change the boundaries of these two monitoring zones. According to the data from this Table, the ground zone can also be extended. If we consider the full length of the routes from fire stations to fires (Table 5) within the ground protection and forest aviation zones, the model allowed to build 1 million 218  Fig. 9. Distribution of relative length of routes (%) to the forest fires within the ground and aviation protection zones, travelling time is three hours (green color), more than three hours (blue color), and total length of routes within ground zone (red color) According to these statistics, the ground protection zone can also be extended to three hours ( Fig. 9), if it is economically profitable with the respect to aviation transport modes and methods of extinguishing forest fires.

CONCLUSIONS
In the paper we have shown a possibility to use the forest fire transportation model to make the quantitative estimations of travelled time and length for the access routes made from the fire stations to the forest fires within the ground and aviation protection zones. Based on the results obtained, the following conclusions can be drawn.
1. Created transport model allowed building routes for 98% of forest fires registered in the ground protection zone under the condition that the transfer of human forces and technical means for them is not more than three hours. This indirectly confirms the correctness of the selected parameters for the model. It is planned to compare the simulation results (as they are shown in this paper) to the real tracks of the fire trucks for the next fire hazard seasons. This will also allow to do a spatial analysis of coincidence between routes in the real conditions and the model calculations.

2.
It is necessary to update the regional attribute and spatial data on the roads and forest glades in order to add the practical value to the created transport model. Transport model can be used to choose a method of forces and equipment delivery (ground / aviation) to the forest fires, as well as their maneuvering in conditions of high and extreme forest burning with limited regional resources. It will also expand the opportunities for interaction between the different regional departments and the tenants of the forest fund lands involved into the forest fires extinguishing.
3. The transport model created can become a basis for solving more complex management tasks, namely, estimating the time spent on the forces and equipment transfer between a forest fire site and a fire station, taking into consideration transport movement parameters in the cities, maneuvering with the firefighting forces involvement between the regions and forest fire units of federal subordination.

4.
Established long-term access routes database allows to evaluate and to adjust the boundaries between ground and forest aviation protection zones, taking into account temporary accessibility of the forest fund territory by public and special roads (fire roads, forest glades, etc.).