AMBUSH VULNERABILITY MODEL DEVELOPMENT JOHN WILLIAM SHINSKY i CONTENTS Page 1. 2. 3. 4. 5. 6. 7. 8. 9. ACKNOWLEDGEMENTS ............................................................................................... iii LIST OF FIGURES ........................................................................................................... iv LIST OF TABLES .............................................................................................................. v LIST OF ACRONYMS ..................................................................................................... vi LIST OF TERMS .............................................................................................................. vii ABSTRACT ........................................................................................................................ 1 INTRODUCTION .............................................................................................................. 2 2.1 Background ............................................................................................................. 2 2.2 Problem Statement .................................................................................................. 2 LITERATURE REVIEW ................................................................................................... 4 3.1 Relevant Studies and Papers ................................................................................... 4 3.1.1 MICRODEM and the Weapons Fan Algorithm .......................................... 4 3.1.2 Uncertainty in Viewshed Analysis of IED Ambush Sites .......................... 4 3.2 Leveraging Previous Works .................................................................................... 5 OVERHEAD ANGLE OF ATTACK MODEL ................................................................. 6 4.1 Overhead Angle of Attack Model Inputs ................................................................ 6 4.2 Overhead Angle of Attack Model Methodology .................................................... 6 4.3 Overhead Angle of Attack Model Assumptions ..................................................... 7 4.4 Overhead Angle of Attack Model Limitations ....................................................... 7 AMBUSH VULNERABILITY MODEL ........................................................................... 8 5.1 Ambush Vulnerability Model Inputs and Variables ............................................... 8 5.1 Ambush Vulnerability Model Assumptions ........................................................... 9 5.2 Ambush Vulnerability Model Methodology........................................................... 9 5.3 Ambush Vulnerability Model Outputs.................................................................. 10 5.4 Ambush Vulnerability Model Limitations ............................................................ 11 CUMULATIVE VIEWSHED MODEL ........................................................................... 12 RESULTS ......................................................................................................................... 13 7.1 Common Inputs ..................................................................................................... 13 7.2 Overhead Angle of Attack Model Results ............................................................ 14 7.2.1 Inputs ......................................................................................................... 14 7.2.2 Outputs ...................................................................................................... 14 7.3 Ambush Vulnerability Model Results .................................................................. 15 7.3.1 Inputs ......................................................................................................... 15 7.3.2 Outputs ...................................................................................................... 16 DISCUSSION OF RESULTS .......................................................................................... 26 8.1 Comparison of Results .......................................................................................... 26 8.2 Comparison of Methodologies .............................................................................. 26 8.3 Advantages of the Ambush Vulnerability Model ................................................. 27 8.4 Disadvantages of the Ambush Vulnerability Model ............................................. 28 8.5 Summary of Results .............................................................................................. 28 CONCLUSION ................................................................................................................. 29 APPENDIX A: BASIC INSTRUCTIONS ....................................................................... 30 REFERENCES CITED ..................................................................................................... 31 ii ACKNOWLEDGEMENTS The US Army Materiel Systems Analysis Activity (AMSAA) recognizes the following individuals for their contributions to this report. The author is: John William Shinsky, Combat Systems Analysis Division, CSAD The author wishes to acknowledge the contributions of the following individuals for their assistance in the creation of this report: Dr. Peter L. Guth, Department of Oceanography, United States Naval Academy iii LIST OF FIGURES Figure 1: Flowchart depicting the OAAM methodology.................................................................6 Figure 2: Firing position point shapefile created for each route point .............................................7 Figure 3: Route centerline and the vehicle length are used to create the route points .....................9 Figure 4: Overview of the iterative viewshed analysis process for each point ..............................10 Figure 5: All viable threat rasters are summed to build the final threat map.................................10 Figure 6: CVM: Number of route points visible at each terrain position ......................................12 Figure 7: DSM for a portion of Baltimore, MD with route centerline ..........................................13 Figure 8: Shapefile containing the evenly spaced route points required by the Model .................14 Figure 9: Translation of speed and reaction time into required visible distance ...........................16 Figure 10: Effect of vehicle speed and reaction time on number of threat points .........................17 Figure 11: 32 KPH Viable threats map using 5s, 10s, and 15s reaction times ..............................18 Figure 12: 48 KPH Viable threats map using 5s, 10s, and 15s reaction times ..............................19 Figure 13: 64 KPH Viable threats map using 5s, 10s, and 15s reaction times ..............................20 Figure 14: 5s reaction time viable threats map using 32, 48, and 64 KPH vehicle speeds ...........21 Figure 15: 10s reaction time viable threats map using 32, 48, and 64 KPH vehicle speeds .........22 Figure 16: 15s reaction time viable threats map using 32, 48, and 64 KPH vehicle speeds .........23 Figure 17: AVM viable threats for each route point using a 133m reaction distance ...................24 Figure 18: AVM firing positions that can see each route point .....................................................25 Figure 19: Viable attack positions from the AVM and CVM methodologies ...............................27 iv LIST OF TABLES Table 1: Ambush Vulnerability Model user selected inputs ............................................................8 Table 2: Ambush Vulnerability Model intermediate computations ................................................8 Table 3: Ambush Vulnerability Model output files .......................................................................11 Table 4: Final statistical output table for the OAAM ....................................................................15 Table 5: User defined parameters used for the analysis ................................................................15 Table 6: Typical Model Run Times ...............................................................................................27 Table 7: Visualization Recommendations .....................................................................................30 v LIST OF ACRONYMS AoA AMSAA AVM CDD CSAD CVM DA DEM DOD DSM DTM GIA GIS OAAM MPE SF TR USGS - Analysis of Alternatives - US Army Materiel Systems Analysis Activity - Ambush Vulnerability Model - Capability Development Document - Combat Systems Analysis Division - Cumulative Vulnerability Model - Department of the Army - Digital Elevation Model - Department of Defense - Digital Surface Model - Digital Terrain Model - Geospatial Information Analysis - Geographic Information System - Overhead Angle of Attack Model - Mobility, Power, and Energy - Standard Form - Technical Report - United States Geological Survey vi LIST OF TERMS ArcGIS - Commercial software application primarily used for Geospatial analysis Feature Class - Dataset containing vector data stored within a geodatabase Geodatabase - Spatial data repository used by ArcGIS to store geospatial data in a proprietary format Polyline - A geospatial feature that is a line created by linking several points Shapefile - A dataset containing vector data in an open format Vector Data - Any geospatial data that is made up of points, lines, or polygons vii 1. ABSTRACT The Ambush Vulnerability Model (AVM) was developed to better analyze a moving vehicle’s vulnerability to an attack. The model uses several user defined parameters to provide a custom analysis that can determine the location of every viable threat position the vehicle could encounter while traversing the route. The model uses a LIDAR DSM dataset, route centerline, vehicle speed, vehicle length, targeting ranges, and a reaction time to accurately determine the positions of every point on the terrain that poses a threat to the moving vehicle based on the vehicle speed and how long they need to maintain line of sight in order to attack. 1 2. INTRODUCTION 2.1 Background In 2012, a question was asked regarding the leveraging of viewshed analysis in order to determine the vulnerability of military vehicles to an overhead attack in operational environments. Although AMSAA is one of the Army’s leading analysis organizations, covering everything from logistics analysis to combat systems analysis, no AMSAA models incorporated line of sight. Since AMSAA had little experience in geospatial analysis, a new model had to be developed. The resulting Overhead Angle of Attack Model (OAAM) was designed for the sole purpose of providing a statistical representation of the percentage of the area that is suitable for attacking the route. This refers to the number of suitable locations out of the total possible firing positions across the entire DEM. The model provides statistical output that could be used for survivability and lethality analysis from an overhead attack. The model also had to be developed within two weeks. The prototype model was completed on time and the statistical results were provided for the survivability and lethality analyses. The methodology utilizes geospatial elevation data to locate and characterize all overhead firing opportunities within a user defined radius circle of a vehicle traveling along a route. The model used a new engagement geometry methodology to capture the engagement range, attack angle, and other critical factors such as line of sight that are needed to assess vehicle vulnerability to overhead attacks. The AMSAA teams responsible for survivability and lethality analysis requested that the model be improved to incorporate a moving target, as well as to easily locate all viable threats instead of just the percentage of the area suitable for an attack. The goal of identifying viable threats would require the model to be completely revamped in order to incorporate both vehicle speed and reaction time, defined as the amount of time it would take for an attacker to acquire, aim, fire, and hit the target. The Ambush Vulnerability Model (AVM) was developed as a result of this overhaul of the OAAM. In addition to the AVM, the Cumulative Viewshed Model (CVM) was also developed to provide a quicker result than the AVM by ignoring whether or not the route points are consecutively visible to the attacker. The CVM determines how many route points a firing position can see, but it does not determine if these visible points are consecutive or split up along the route. 2.2 Problem Statement The new methodology will analyze a vehicle’s vulnerability to an attack by looking not only at line of sight, but also how long there is continuous line of sight from a shooter to a target. It builds upon the multiple viewshed analysis methods that exist in the OAAM methodology to consider the speed the vehicle will be traveling and how long the weapon 2 system will need to have line of sight in order to be considered a viable threat. The final product will be a toolbox that contains two models that can be used to answer different questions regarding a vehicle’s vulnerability to an ambush. The AVM and the CVM models provide the user with various analytical options to address ambush vulnerability. 3 3. LITERATURE REVIEW A review of past studies was done to determine if any previously used principals could be applied to this new methodology. The focus of the literature review was on viewshed and line of sight analysis methods, as well as any information on utilizing a multiple viewshed analysis. Several studies have focused on these principals and one was very similar to the Overhead Angle of Attack methodology, but nothing had previously been done with respect to analyzing the potential of an attack on a moving ground vehicle. 3.1 Relevant Studies and Papers The literature review found four papers that detail the use of multiple or cumulative viewshed analysis that is similar to the analysis methodology in the development of the AVM. Two papers by Guth (2003, 2004) are directly related to the work that was done in the OAAM, and can be directly applied to the AVM. 3.1.1 MICRODEM and the Weapons Fan Algorithm Guth (2003) described the MICRODEM program, which allows the user to define a route and the program determines where it can be seen by ambushing forces. This is done by computing a weapons fan at each point along the route which is then stored so that it can be used in the final ambush movie. The ambush movie uses a base layer and overlays the results from the weapons fan algorithm at each point to create a movie that shows the amount of the terrain that can see each point along the route. The model also creates a final image that displays the percentage of the route that can be seen from each observation point. The AVM project builds on Guth (2003). The MICRODEM model takes a route and computes a viewshed at each point along the route to determine which points on the terrain have line-of-sight to the route. When the analysis is completed there is a final output that includes a movie of the viewsheds strung together, as well as a visualization of what portions of the route are visible. In addition, Guth showed outputs identical to what is produced by the CVM. By expanding on this technology, vehicle speed and reaction times can be incorporated, bringing a higher level of complexity and accuracy to the model. Guth (2004) details how the line of sight, viewsheds, and accurate 3D perspective view models are computed from intervisibility algorithms and DEMs. The author discusses how the geometric model has the largest impact on these results. The author also discusses what methods and algorithms produce the highest quality weapons fans while minimizing error. (Guth, 2004) 3.1.2 Uncertainty in Viewshed Analysis of IED Ambush Sites Not all of the applicable studies pertain to the use of a standard multiple viewshed analysis approach. (Raehtz, 2011) covers in detail the uncertainty in viewshed 4 analysis, which can come from both the differences in algorithms and the error in elevation models. The paper also discusses a viewshed analysis approach that uses the highest resolution DEMs available in Afghanistan to analyze IED explosion sites. By considering uncertainty Raehtz, (2011) was able to extrapolate the error models to ultimately develop a more informative viewshed using a Monte-Carlo simulation. His detailed documentation on error in viewshed analysis and how to properly implement viewsheds contributed to properly implemented code in AVM. 3.2 Leveraging Previous Works The previous works above all deal with various types of cumulative and multiple viewshed analysis methodologies. The AVM relies heavily on this methodology to compute a viewshed at every point along a specified route. These time synchronized viewsheds are then used for a complex analysis to determine what points on the terrain have continuous line of sight to the vehicle. 5 4. OVERHEAD ANGLE OF ATTACK MODEL The OAAM was developed to show the locations of suitable attack positon in an urban environment. The model provides an overall statistical representation of the threat in the study area, while also providing statistical information on the distribution of attack angles and attack distances. The model uses python scripting and runs in Esri ArcGIS10.x for Desktop Advanced. 4.1 Overhead Angle of Attack Model Inputs The only two inputs required by the OAAM are a DSM and a route point shapefile to represent the vehicle waypoints to be analyzed. Depending on the type of analysis that is being conducted, the required resolution of the DSM varies. The other input ingested into the model is the route points for the vehicle. While any spacing can be used for these points, the vehicle length was used as the point spacing under the assumption that it provided for a continuous analysis is conducted and that no vehicle positions are unaccounted for. By using a larger route point spacing, the runtime of the model can be greatly reduced at the at the expense of introducing some uncertainty in the results. 4.2 Overhead Angle of Attack Model Methodology The current OAAM locates and characterize all overhead firing opportunities within a user defined radius circle of a vehicle traveling along a route (Figure 1). The model computes a viewshed for each point on the route, and characterizes all possible firing opportunities in terms of distance and attack angle. The OAAM produces a statistical output table called tablePercentages showing the percentage of the area that is suitable for an attack originating from certain distances and attack angles. The model also creates a shapefile for each route point that contains the locations of every possible firing position along with angle and distance (Figure 2). Figure 1: Flowchart depicting the OAAM methodology 6 Figure 2: Firing position point shapefile created for each route point 4.3 Overhead Angle of Attack Model Assumptions The main assumption, which has been hard coded into the model, is that the target and the firer are 2 meters tall. The model also assumes the vehicle is not in motion, and that line-of-sight provides enough time for a firing opportunity. 4.4 Overhead Angle of Attack Model Limitations The major limitations of the OAAM include static inputs and the assumption of a stationary target. The model requires the route centerline to be split into route points prior to running the model, but this functionality could easily be added to the model for convenience and is a simple procedure for an experienced GIS analyst. The OAAM produces a statistical result that only provides the user with an overall percentage of the area that is suitable for attacking the route. The model produces shapefiles that allow the user to visualize all firing positions at each route point, but there is no easy way to visualize the results for the entire route. 7 5. AMBUSH VULNERABILITY MODEL The AVM was developed to analyze a moving vehicle’s vulnerability to any attack while also addressing the limitations of the OAAM. The model allows the users to customize the most of the inputs so that the model will better fit their analysis. The model also allows the user to visualize the actual locations of the firing positions that are classified as viable threats. A viable threat exists if the firing position has continuous visibility of the target for the amount of time that was specified by the user. The end result is a highly versatile model that can simulate a moving vehicle and analyze the vulnerability to an attack from any angle or distance. 5.1 Ambush Vulnerability Model Inputs and Variables To consider a moving target, the AVM requires a few additional inputs (Table 1) beyond those used in the OAAM. These parameters allow the model the versatility to handle a wide range of applications. Table 2 shows the three variables that are computed within the model based on the user defined parameters. These variables are used throughout the model for viable threat classification, as well as output messages that the user can see as the model runs in order to keep track of the model’s progress. Table 1: Ambush Vulnerability Model user selected inputs INPUTS LIDAR DSM Route Centerline Vehicle Length (meters) Vehicle Speed (kph) Minimum Range (meters) Maximum Range (meters) Target Height (meters) Firer Height (meters) Reaction Time (sec) DESCRIPTION Elevation raster used for the viewshed analysis Polyline shapefile that will be split into evenly spaced points Used for route point spacing and the calculation of the number of route points required for a viable threat to exist Used to calculate the number of route points required for a viable threat Minimum targeting range for an attack Maximum targeting range for an attack Height of the target above the ground in meters Height of the threat above the ground in meters Used to calculate the number of route points required for a viable threat Table 2: Ambush Vulnerability Model intermediate computations VARIABLES pointsForKill totalPoints Reaction Distance (meters) DESCRIPTION Number of consecutive points that must be continuously visible for a viable threat Total number of route points analyzed Distance that must be under continuous observation for a viable threat 8 5.1 Ambush Vulnerability Model Assumptions The model assumes that one raster cell centroid equals one potential firing position. Another assumption is that analyzing the vehicle positions bumper to bumper is sufficient to assess every possible vehicle location since there would be no position on the route that the attacker could not see some portion of the vehicle. The last assumption is that a viable threat firing position requires line of sight to the vehicle for at least the amount of time specified by the user defined reaction time. 5.2 Ambush Vulnerability Model Methodology After the user sets all of the parameters and starts the model, the first process involves splitting the route into evenly spaced route points (Figure 3). These points correspond to every possible vehicle location along the route, assuming that using the vehicle length as the spacing results in a line of vehicles bumper to bumper. If a firing position can see two points in a row, it is not possible to miss anything since some portion of the vehicle would still be visible when its centroid is between the two visible locations. Figure 3: Route centerline and the vehicle length are used to create the route points Once the route points shapefile has been created, the model goes calculates and stores a variable called pointsForKill, which refers to the number of route points that must be continuously visible by a potential firing position in order to classify it as a viable threat. This is the distance traveled by the vehicle during the user defined reaction time, divided by the point spacing. The outputs of these two steps are fed into an iterative multiple viewshed analysis that loops through one point at a time for the entire route. The multiple viewshed analysis loops through every point in the route points shapefile and determines whether or not continuous line of sight exists at a firing position. This iterative analysis process (Figure 4) completes three steps for each route point. 9 Figure 4: Overview of the iterative viewshed analysis process for each point The model computes a viewshed at each route point. Once the route point number is equal to pointsForKill, the model sums the previous pointsForKill viewsheds into a single raster and then reclassifies that raster to retain only those attack points that had intervisibility for all the points. The goal of process is to compute a raster for every route point that shows the locations of firing position that had continuous line of sight to the target for the correct number of route points. These viable threats rasters will then be used to determine the location of every threat along the entire route. After looping through every point in the routepoints shapefile, the model sums every viable threat raster to form one final viable threat output raster showing the location of every viable threat that was encountered along the entire route. This final process is visualized in by the flowchart in Figure 5. Once this final output raster is created and saved, the analysis is complete. Figure 5: All viable threat rasters are summed to build the final threat map 5.3 Ambush Vulnerability Model Outputs The AVM produces several output files that can be used by the user in many ways (Table 3). The final raster layer showing the location of every threat along the route provides the final result of the model. This raster dataset that shows the location of every point on the map that has the capability of attacking the vehicle based on the vehicle’s speed and the user defined reaction time, as well as how many route points each viable threat could attack. This output map could be used to make the driver aware of the possibility of an attack from those locations, as well as to set up observation points in areas with line-ofsight to the locations that have the most visibility to the route. This would ultimately provide protection to the driver and occupants of the vehicle. This could also be used to 10 reroute a convoy by avoiding the sections of the route determined to be the most dangerous. Table 3: Ambush Vulnerability Model output files OUTPUTS DESCRIPTION allThreats raster Final raster layer showing the viable threats for the entire route routepoints shapefile viableThreats# raster viewshed# raster 5.4 Point Shapefile containing the evenly spaced route points and information on the number of viable threats and firing positions at each point Intermediate raster layer showing the viable threats at each point Intermediate output raster layer showing the viewshed computed at each point USES Main output of the AVM that allows the user to visualize every viable threat along the route Allows the user to visualize the number of viable threats and visible firing positions at each route point Used by the model to compute the final all threats raster Used by the model to compute the viable threat rasters Ambush Vulnerability Model Limitations Although the AVM addresses most of the limitations in the OAAM, it still only assesses possible threats on the rooftop level of buildings or on the ground. Since the model uses a DSM for the elevation values, it does not consider a threat positioned on a lower level balcony or shooting through a window. The model also assumes that the vehicle remains at a constant speed throughout the entire route. In doing this, every viable threat position is based on the assumption that the vehicle has not encountered a previous attack which would have caused them to drive faster and use evasive maneuvers. This is also an issue if the vehicle is forced to slow down or stop because a vehicle is attacked in front of it. The model is not capable of assessing how a drastic change in speed would affect the outcome of the remaining portion of the route. Vehicle speed has a direct effect on the distance that a vehicle must be visible in order to be attacked. If the target speeds up, the vehicle would be able to travel a larger distance within the attacker’s reaction time, which would make it more difficult for a viable threat to exist. If the vehicle slows down the opposite would occur, making the existence of a viable threat more probable. The most significant limitation of the model is runtime. The AVM can take up to 10 seconds per route point to run depending on the user’s input parameters and computer specifications. This limitation can be mitigated by either using fewer route points to improve runtime or using the CVM which was developed in response to the AVM’s runtime issue. Both of these mitigation options require the user to sacrifice precision for performance. 11 6. CUMULATIVE VIEWSHED MODEL The CVM was developed as a quicker alternative to the AVM. The model uses the same inputs as the AVM and many of the same processes. The only difference between the two models is that the CVM ignores continuous line of sight and does not determine whether or not route points are consecutively visible to a firing position. The model allows the user to quickly visualize the number of route points that are visible at each location on the terrain (Figure 6). If maintaining continuous line of sight is not important to the user and can be neglected, the CVM out performs the AVM and provides a similar result that still meets the user’s analytical requirements. Figure 6: CVM: Number of route points visible at each terrain position 12 7. RESULTS The goal of this analysis was to compare the results using the existing OAAM model and the new AVM and CVM models. 7.1 Common Inputs To conduct the analysis using both models, a LAS dataset for a portion of downtown Baltimore was downloaded from the USGS’ Earth Explorer website (http://earthexplorer.usgs.gov) and converted to a DSM elevation raster with a resolution of three meters that could be easily ingested into the model. This conversion was done within ArcGIS using the Create LAS Dataset and LAS Dataset to Raster conversion tools. A 2 kilometer route was digitized through the study area in order to ensure enough opportunities for a viable threat to exist. Figure 7 shows this DSM raster layer with the route overlaid. Figure 7: DSM for a portion of Baltimore, MD with route centerline 13 7.2 Overhead Angle of Attack Model Results 7.2.1 Inputs The OAAM requires the route centerline to already be split into evenly spaced points prior to running the model. To ensure the same data was being used for both models, the routepoints shapefile that is generated automatically by the AVM (Figure 8), using only the vehicle length for the point spacing, was used as the input for the OAAM. Figure 8: Shapefile containing the evenly spaced route points required by the Model 7.2.2 Outputs The final output from the OAAM is a statistical table showing the summarized results of the analysis broken down by angle ranges and distance bins. The percentage of the area suitable for an attack is averaged for all 406 route points that make up the vehicle route that was analyzed (Table 4). The table shows that there are very few firing positions from an angle greater than 30 degrees. At an angle of 30 degrees the firer’s height would need to be half the distance from the target. As the angle increases from this point, the firer’s height must drastically rise as well. Unless the firer is almost directly above the target firing straight 14 down, these higher angle ranges will never be achieved and should not be included in most analyses. Table 4: Final statistical output table for the OAAM 7.3 Ambush Vulnerability Model Results 7.3.1 Inputs The AVM has a great deal of versatility, allowing the user to fully customize the parameters of the model to best represent the desired analysis. Table 5 shows the additional parameters for this analysis. Table 5: User defined parameters used for the analysis PARAMETER Vehicle Speed (kph) Vehicle Length (meters) Minimum Targeting Range (meters) Maximum Targeting Range (meters) Target Height (meters) Firer Height (meters) Reaction Time (seconds) VALUES 32, 48, 64 5 30 600 2 2 5 , 10 , 15 The vehicle length of 5 meters corresponds with the approximate dimensions of a standard SUV, which is often used by military, law enforcement, and many 15 civilians. Varying reaction times were used to show how the results of the model vary depending on the parameters that have been set by the user, and to illustrate how the analysis could easily be changed to reflect a specific weapon or targeting system. The analysis was conducted nine times using three different vehicle speeds and three different reaction times. Figure 9 shows how the varying vehicle speeds and the reaction times correspond to the distance over which a firer needs continuous visibility to the target. Figure 9: Translation of speed and reaction time into required visible distance 7.3.2 Outputs The AVM creates a raster map that shows the location of every viable attack position along the route. For this 2 kilometer route through downtown Baltimore, the majority of the viable threats are close to the route at ground level and on top of nearby buildings. The results also clearly show how changing the vehicle speed and reaction time impact the final results of the model. Figure 10 shows the total number of viable threats for the nine runs. It shows an inverse correlation between vehicle speed and viable threats, so if the vehicle is traveling faster, there will be fewer viable threats with the capability of hitting the target. This also goes for the reaction times, the longer the attacker needs to hit 16 the target, the less likely they are to being classified as a viable threat. The graph shows how critical the reaction time and vehicle speed parameters are to the model’s results. It also shows that the relationship between vehicle speed, reaction time, and viable threats is exponential. This is visualized by the bend in the graph. If each additional 5 seconds reduces the number of viable threats by 60%, the reductions will not be linear since the first 60% reduction could be thousands of threats and the second 60% reduction could be hundreds of threats. In addition, an increase in reaction time creates more short stretches of road that cannot be attacked. Figure 10: Effect of vehicle speed and reaction time on number of threat points A final viable threats map overlay was created for each of the nine runs. Figure 11 shows the results for the three runs conducted at 32 kph, Figure 12 shows the results at 48 kph, and Figure 13 shows the results at 64 kph. The maps show that as the reaction time is increased the number of threat locations decreases. These maps were also created for each reaction time to show how changing the vehicle speed affects the results. Figure 14-16 show the results for 5, 10, and 15 second reaction times. To help visualize the most dangerous areas of the routepoints output file provides the capability to visualize the number of viable threats and the number of visible firing positions at each route point. Figure 17 shows the number of viable threat locations that can engage each route point and Figure 18 shows the number of firing positons that can see each route point. The map showing the number of viable threats at each route point clearly shows the most dangerous places along the route. 17 Figure 11: 32 KPH Viable threats map using 5s, 10s, and 15s reaction times 18 Figure 12: 48 KPH Viable threats map using 5s, 10s, and 15s reaction times 19 Figure 13: 64 KPH Viable threats map using 5s, 10s, and 15s reaction times 20 Figure 14: 5s reaction time viable threats map using 32, 48, and 64 KPH vehicle speeds 21 Figure 15: 10s reaction time viable threats map using 32, 48, and 64 KPH vehicle speeds 22 Figure 16: 15s reaction time viable threats map using 32, 48, and 64 KPH vehicle speeds 23 Figure 17: AVM viable threats for each route point using a 133m reaction distance 24 Figure 18: AVM firing positions that can see each route point 25 8. DISCUSSION OF RESULTS 8.1 Comparison of Results The results of the OAAM show that the majority of the suitable attack locations exist in the 150 meter to 450 meter distance ranges and come from an attack angles between 1 degree and 15 degrees (Table 4). The AVM provided results that range from 616 viable threats (0.22% of total possible firing positions) to 15,099 viable threats (5.42% of total possible firing positions) over the entire route depending on the parameters that were used. Both models have the capability of providing a number of firing positions and a total percentage, but the OAAM is limited to a stationary target where the AVM allows the user to determine vehicle speed and the firer’s reaction time. The OAAM determines the percentage of the area suitable for an attack on a stationary vehicle from various attack angles and ranges. Most of the attack positons will occur at angles below 30 degrees, since it becomes much harder to find an elevated attack position with a large angle as you get farther away from the target. At 30 degrees the firer’s elevation must be half the distance to the target. This is only possible at very close ranges. The AVM uses the vehicle speed and a reaction time to find all possible attack positions for a moving target. A faster moving target is more difficult to hit than a slower one. Also, a weapon system that requires a longer target acquisition time will also have more difficulty hitting the target. The bends and curves in the route prove to be the safest portions of the vehicle’s route. The majority of the firing positions that can see the vehicle on one street cannot see the vehicle on another. This suggests that a route that constantly turns and changes streets may be safer than a direct route that stays on the same street for long distances. Frequent turns would cause the vehicle to slow down, which would increase the risk of being attacked. 8.2 Comparison of Methodologies The comparison of the results using AVM and CVM (Figure 19) shows red areas corresponding to viable threats using both models. The yellow areas correspond to the viable threats classified only by the CVM. The comparison was done using a vehicle speed of 48 kilometers per hour and a reaction time of 10 seconds. The AVM provides a final result of 5,409 viable threats, while the CVM results in 7,264 viable threats. This equates to almost a 35% increase in the number of viable threat locations by considering continuous line of sight. The AVM is twice as time consuming as the CVM (Table 6), but also provides an increase in accuracy. While the AVM does occasionally omit a viable threat because there is a small break in visibility, it also rules out the possible threat locations that can see half of the required number of route points on one road and the other half on another. This firer would not be able to attack the target based on the user defined parameters, but it would be classified as a viable threat using the CVM methodology. This also applies to unmanned targeting systems that must have continuous visibility to the target to lock on, fire, and hit the target. A small gap in visibility may be 26 enough to restart the acquisition process. The AVM takes longer to run, but provides a higher level of confidence in the results. This needs to be taken into consideration when the user determines which methodology they plan to use. If faster run times and better model performance is more important than absolute accuracy, then the CVM is sufficient. If the user is comparing routes and is only looking to determine which route is safer, the differences between the AVM and CVM do not matter. If the user is interested in the exact locations of viable threats on a specific route that have continuous line of sight to the target, the differences in results between the two models is more significant. Table 6: Typical Model Run Times Model OAAM AVM CVM Run Time 15 Seconds Per Route Point 10 Seconds Per Route Point 5 Seconds Per Route Point Figure 19: Viable attack positions from the AVM and CVM methodologies 8.3 Advantages of the Ambush Vulnerability Model The new model has many advantages over the OAAM. These advantages range from the ability to model a moving target to the ability for the user to quickly generate a map of 27 the locations of every viable threat. The OAAM provides individual shapefiles that can be visualized for every route point, but no simple synthesis. The AVM systematically generates a summary visualization. In addition, the AVM also allows the user the capability to tune the inputs to meet the specific requirements of the analysis. This is also possible using the CVM, but the model does not consider continuous line of sight which can lead to gaps in visibility that may not be negligible. 8.4 Disadvantages of the Ambush Vulnerability Model The AVM’s runtime ranges from a few seconds per point to over 10 seconds per point depending on the input parameters and the user’s computer specifications. The excessive runtime can be mitigated depending on the user’s analysis requirements by increasing the route point spacing so the model has to process fewer points. Future ArcGIS capabilities, such as multiple core processing, can be built into new versions of the model to allow for much faster runtime by taking advantage of multiple core processing. This could allow the model to run different tasks simultaneously resulting in significant improvements in runtime. 8.5 Summary of Results The results of the three models clearly represent the limitations of the OAAM, as well as the advantages of the AVM and the CVM. The results of the OAAM provide detailed statistical probabilities, but do not provide useful information to the user other than the ability to compare different routes to see which is safe. The CVM enables the user to see how many route points can be seen by each firing positon on the terrain, which can be used to determine which firing positions pose the greatest threat to the chosen route. The AVM, which adds to the methodology of the CVM, computes the number of threat points and shows their locations. The model also provides additional outputs that make it possible for the user to determine the most dangerous portions of a route and ultimately what roads should be avoided completely. Both the CVM and the AVM provide the user with the flexibility to customize the model to meet the requirements of their analysis. Both models provide useful results that can be used easily by the user. The major difference between these two methodologies is how the viable threats are classified. The CVM simply determines the total number of route points that each position can see and ignores whether or not they are consecutively visible. This assumes that either all the visible points are consecutive or that small gaps in visibility do not matter enough to justify a longer run time. The AVM considers continuous line of sight and ensures that there are no gaps in visibility. This leave the user with a higher level of confidence that the viable threats have continuous line of sight to the target for the specified amount of time with no gaps or breaks in visibility at the expense of a much longer runtime. The user must determine whether faster runtimes is as important as a high level of confidence in the results. 28 9. CONCLUSION To improve upon the OAAM, the AVM and CVM methodologies analyze a vehicle’s vulnerability to an attack by looking at both line of sight and target acquisition timing, which depends on vehicle speed and the required weapon system engagement time. The AVM and CVM models also allow the user to specify a minimum and maximum targeting distance in order to analyze the viable threats associated with specific weapons systems with individual arming distances and range limitations. The addition of these additional parameters allow the models to more accurately simulate a moving vehicle ambush scenario. The models enable the user to conduct a much more thorough and complete analysis than the OAAM. In addition to these improvements, the new models run faster and more efficiently. The AVM runs 33% faster and the CVM runs 66% faster. The development of these new models provides the user with the capability to effectively model the ambush vulnerability of a moving vehicle, as well as determine the locations of all possible threats along a specified route. 29 APPENDIX A: BASIC INSTRUCTIONS Installation Instructions: Before the models can be used, they must be downloaded and installed. This process is not difficult and can be easily completed within ArcGIS. The installation process is as follows: Download the zipped tools folder Unzip folder Open ArcMap Open ArcToolbox Right click on Toolbox icon at the top of the ArcToolbox window Pick “Add toolbox” Navigate to the directory, and Select the “AMSAA tools.tbx” file Open the toolbox, and pick one of the two models Sample data for Baltimore is included in the “Input” directory Output Visualization Recommendations The AVM produces two output files that are useful to visualize. The most important is the allThreats.tif raster that can be found in the Threats folder within the selected output folder. The other useful output is the routepoints.shp shapefile that can be found directly within output folder. These two files can be used to produce three visualizations that make it very easy to understand the final results. (Table 7) The CVM produces one main output file that is useful to visualize. This output is the viewshedSum.tif raster that can be found directly in the output folder that was used to run the model. Using this output, it is possible to produce two visualizations that provide a detailed look at the model’s results. (Table 7) Table 7: Visualization Recommendations MODEL OUTPUTS allThreats.tif AVM routepoints.shp routepoints.shp viewshedSum.tif CVM viewshedSum.tif VISUALIZATION Location of all viable threats that are capable of attacking the vehicle Number of viable threats that exist at each route point by changing the symbology to visualize the numthreat field Number of visible firing positions at each route point by changing the symbology to visualize the numvis field Firing positions that can see a specific number of route points Number of route points that can be seen by each firing position 30 EXAMPLE Figure 19 Figure 17 Figure 18 Figure 23 Figure 19 REFERENCES CITED Guth, P.L. (2003). Ambush Movies and the Weapons Fan Algorithm: Military GIS Operations and Theory, in Proceedings of the International Conference on Military Geology and Geography, June 15-18, 2003, West Point NY. Guth, P.L. (2004). The Geometry of Line-of-Sight and Weapons Fan Algorithms: in Caldwell, D.R., Ehlen, J., and Harmon, R.S., eds., Studies in Military Geography and Geology, Dordrecht, The Netherlands, Kluwer Academic Publishers, chapter 21, p.271-285. Raehtz, S.M. (2011). Accounting for Uncertainty in Viewshed Analysis of IED Ambush Sites in Afghanistan. Michigan State University. Retrieved August 20, 2014, from http://etd.lib.msu.edu/islandora/object/etd%3A1141 31