Winter 2012 Edition
By Mike Price, Entrada/San Juan, Inc.
This article as a PDF.
Sample dataset
This tutorial shows how to use the Closest Facility (CF) solver in ArcGIS Network Analyst to model travel from a single facility to the closest incidents. The model will be modified to reflect proposed changes in station locations.
In the current economic climate, public safety administrators and planners are often faced with complex decisions regarding moving or closing essential facilities. Fire station closure, consolidation, or relocation studies occur throughout the United States. Public safety officials try to make best use of existing resources and provide comparable or improved levels of service. To analyze existing and future conditions, comprehensive data is required to represent current conditions, reflect past performance, and plan for the future.
In Esri's home town of Redlands, California, the Redlands Fire Department is facing considerable commercial and residential growth in the northern part of the city. Redlands is currently protected by four fire stations in Redlands and a station in Loma Linda, located to the west.
In 2010, Redlands firefighters and emergency medical technicians (EMTs) responded to more than 8,000 incidents in and around the city. By mapping the likelihood of where, when, and how incidents have occurred, historic incidents often provide the best estimate of community risk. Historic risk is used to demonstrate an adequate Standard of Coverage; identify underserved, problem, or frequent call locations; and project future demand for services.
The CF solver in ArcGIS Network Analyst is an excellent tool for modeling optimal travel from fixed facilities to historic incidents. By adding or substituting new or relocated stations in CF templates, it is easy to perform multiple analyses.
In the Fire Data geodatabase, expand the Response_Model feature dataset and add only the prebuilt Response_Model_ND network dataset.
To begin this exercise, download the zipped training dataset. Unzip it in a project area on your computer. Open ArcCatalog and navigate to the \Redlands_Fire folder and explore its contents. This project contains two file geodatabases and several small utility files. The training data is projected in California State Plane North American Datum 1983 (NAD83) Zone 5 US Feet. To support a US National Grid spatial reference, the data frame projection is set to universal transverse Mercator (UTM) NAD83 Zone 11 Meters.
Inspect the geodatabase feature classes and layer files in the geodatabases. In the Fire Data geodatabase, open the Response_Model feature dataset and inspect Response_Model_ND. To quickly start the analysis portion of this tutorial, a very functional network dataset was prebuilt.
Notice that two layer files, First Due Station Group and Travel Time Improvement Group, cannot be viewed. Do not load them until instructed. These layer templates will be added to the map later to display results.
To load incidents, right-click Incidents and click Load From 2010 Incidents. Set the Sort Field to OBJECT_ID and select Inc_Number for Name. Click OK and wait patiently as all 8,173 incidents load.
The CF solver is traditionally used to locate one or more destination facilities within reasonable travel time of a specific incident. Turned upside down, the CF solver maps travel from one or more fixed facilities to multiple incidents. In this exercise, the CF solver will be used to model travel away from single facilities toward the closest incidents, but facilities will be modified to reflect station changes.
Using actual 2010 incidents and a section clipped from the street network provided by the regional dispatch center, this tutorial will use the CF solver to compare optimal existing coverage to coverage provided by adding one or two proposed stations in northern areas of Redlands. For another innovative use of the CF solver, see another article in ArcUser, "Run Orders: Modeling and mapping public safety arrival orders," in the Fall 2009 issue.
Right-click Facilities and select Load Locations. Set Load From to Fire Stations, Sort Field to Index, Name to Label, and Search Tolerance to 500 Feet and click OK.
In the TOC, right-click 2010 Incidents and choose Joins and Relates > Join. For Item 1, select OBJECTID, and for Item 2, specify Redlands CF Base Case > Routes. Be sure to specify Base Case!
For Item 3, select IncidentID. Click OK to join and bypass indexing. Since you are joining to an active network dataset, indexing does not work.
Now comes the tricky part. If you tackled the Fall 2009 exercise in "Run Orders: Modeling and mapping public safety arrival orders," you might remember performing a series of tabular joins coupled to field calculations. You will perform similar joins and calculations with this data, but this time, you will create two consecutive joins. Follow along carefully because this workflow is very specific.
To calculate the time change between Base Case and all incidents modified by response from Proposed A, select or type [Time_A] – [Time_Base] in the calculation window and click OK.
Now that the closest station and travel time have been calculated for Base Case, the same procedure will be used for each scenario. The only difference will be that in the first join, Item 2 will change (e.g., Redlands CF Prop A > Routes, Redlands CF Prop B > Routes, and Redlands CF Prop B > Routes). Although the workflow becomes a bit tedious now, it is critical to proceed carefully. Make the two joins for each layer, successively calculate the Station_A, Time_A, Station_B, Time_B, and Station_AB and Time_AB fields, making sure to remove all joins after calculating the stations and time for each scenario. Carefully check the calculations and save the project. If one or more joins/calculations did not work, clear and rebuild the necessary join and recalculate.
With Base Case and scenario times for all analyses, you can use the Field Calculator to calculate the Change_A, Change_B, and Change_AB fields that are currently filled with zeros in the 2010 Incidents attribute table.
[Time_A] – [Time_Base]
[Time_B] – [Time_Base]
[Time_AB] – [Time_Base]
Individually open the Travel Time Improvement Group series. This shows effects of Proposed A plus Proposed B.
Look at the results on a map. There are many ways that you could map and chart these time-based analyses. A rather basic map would plot closest stations to all incidents for each scenario. Because you have assigned a station number to all incidents, you can readily map this relationship. Station_ field is an integer. Proposed A is designated as 901 and Proposed B is designated as 902. A second map series can show the time change between the Base Case and the three scenarios. You can use a graduated size and color scheme to show how much time is gained or lost as stations open, close, or move. The fields holding differential calculation times contain necessary double precision values.
Inspect the analysis results. Notice the average times for fire responses, EMS calls, and service calls.
Return to the 2010 Incidents table and locate the IncTypeSum1 field. Right-click this field and select Summarize. Save the summary table in Fire_Data.gdb and calculate average times for Time_Base, Time_A, Time_B, and Time_AB.
Finally, summarize the Base Case and Scenario times for different incident types. Open the 2010 Incidents table, right-click the IncTypeSum1 field, and choose Summarize. Under Item 2, expand Time_Base, Time_A, Time_B, and Time_AB and check the Average box under each. Under Item 3, save the summary table as IncType_Time_Sum1 in Fire_Data.gdb. Add the resultant summary table to the map.
Inspect the results and notice the average times for fire responses, emergency medical service (EMS) calls, and service calls. The proportions and time trends in this table are not unusual. More than 75 percent of all 2010 responses were EMS or rescue calls and average times for these calls are typically the shortest. Nearly 18 percent of all calls are classified as service calls or other calls. Fire and related calls comprise less than 6 percent of all responses. Check out the average times for your scenarios. The average modeled travel time for Proposed A plus B is closing in on two minutes.
You have only scratched the surface of emergency services modeling. There are many more analyses, maps, charts, and determinations that could be prepared with this data. For a challenge, experiment with actual station relocation or removal. For example, try removing Station 263, located between Proposed A and B, and see how times increase and decrease. Also, experiment with charting in ArcGIS or your favorite spreadsheet program to graphically display the changing data. This is a training dataset, so use your imagination and have fun.
Thanks to the fine staff of Redlands Fire Department and City of Redlands GIS for providing this very interesting and complex dataset. Thanks also go to the Confire Communications Center dispatch center for the opportunity to use its excellent time-based streets network in this exercise. The Redlands Fire Department will use many of the same methods and procedures presented here to model and analyze historic, current, and future protection capabilities. Redlands and Esri are very fortunate to have the high level of emergency fire and medical response provided by the Redlands Fire Department.