Organizations must also have the means by which workload trends
can be used for long term planning of future facilities. A new
Fire station as an example requires several years of planning
and construction at a cost of multiple hundreds of thousands of
dollars. Analysis of demographic and workload patterns provide
the basis for the planning of such facilities.
This paper focuses on the techniques used to develop a workload
based automated redistricting engine currently in use by the Montgomery
County Police Department in Maryland.
The Problem:
Montgomery County, Maryland, a Northern suburb of Washington
DC has experienced a tremendous population growth in the past
25 years. As would be expected this growth has not occurred uniformly
through the County. Rather, some areas have experienced a high
development rate while other more established neighborhoods have
changed only moderately in the past two decades.
The Montgomery County Police Department (MCPD) provides law enforcement
through five primary districts. Each district has a central district
office and a district commander. Districts are divided into police
beats where individual police officers from each district are
assigned. For the purpose of data collection, beats are further
sub-divided into Police Reporting Areas (PRAs). Each call for
service placed to the Police Department is assigned to a PRA based
on its geographic location. Additionally, depending on the type
of call received, a workload value is also assigned to each incident.
Traditionally, the Police Department on an annual basis reviewed
workloads by district and beat. Where changes in demographics
had resulted in disparities in workload among districts and/or
beats, changes in assigned resources and boundaries would be made
to correct the problem. This task required the efforts of three
senior level personnel over one weeks time. Additionally, since
the process was performed manually, desired community and political
input into the process was nearly impossible. Further, the manual
process provided no mechanism for use of this data in the long
term planning of the Police Department's resources and facilities.
The Police Department outlined the following criteria in its requirement
for the development of an automated Redistricting/Beat Realignment
System (R/BRS).
The Solution:
The Redistricting and Beat Realignment System was developed
in a partnership between the Montgomery County Police Department,
Digital Engineering Corporation and the Denver and Ephrata Telephone
and Telegraph Company. The R/BRS was developed using ArcInfo
6.1.1 on a SUN Sparc 20 running SUNOS 4.1.3 using ArcInfo Amls
and ANSI C routines. The software is cross platform transferable
to all Platforms supported by ArcInfo.
The following describes the process used by the R/BRS to create
new district and beat boundaries:
Determine the workload for each PRA.
As previously discussed, The Police Department for the purpose
of data collection has divided the county into approximately 700
PRAs. The Police Department computes the workload for each PRA
on a monthly basis. The R/BRS MIS tool imports this information
from the County's Mainframe system into an INFO (and optionally
an ORACLE) database.
Produce district boundaries.
The process is started with the production of new Police District
Boundaries. The R/BRS Redistricting tool allows the user to create
multiple models depicting different district configurations. For
each model, an evaluation period is selected. The selection of
an evaluation period provides the Police Department the ability
to create models based on long term past trends or recently collected
data and perform long term forecasting functions. The total workload
for each PRA over the selected period is then calculated.
Prior to generating a new district model, two basic modeling parameters
are established:
After boundary conditions have been established, the user may
then select the number of districts into which the county should
be divided and specify the center points for each of the districts
on the graphical display of the county. The user is also permitted
to alter the amount of the total county workload each district
will be assigned.
The system now performs an automated redistricting process where
beginning with each district center point, the model annexes adjacent
PRAs to create the districts. Where hard boundary lines are encountered,
the growth of the district in the direction of the boundary is
halted. Hard boundary polygons are treated as one large PRA. Once
all PRAs have been assigned, the districts which need additional
PRAs to reach the target workload annex PRAs from districts which
have a surplus of workload. Safeguards are built to insure that
the model does not enter an endless cycle of transactions.
Upon completion of the redistricting process, the new district
boundary lines along with a bar graph, depicting the resulting
workload for each district is displayed. Manual editing tools
are available to allow the user to move PRAs from one district
to another. The manual tools also allow the adding, deleting,
and renaming of districts. Once the user is satisfied with the
district boundaries, the model may be accepted for use in beat
realignment.
Produce beat boundaries
Once balanced districts have been created, the individual beats
within each district can be redrawn to create equal workloads
between beats. The beat boundaries are constructed in the same
manner as the district boundaries. Boundary conditions are established
and the Center points for each beat is selected. The evaluation
period for the model is also selected. The results from the automated
beat realignment can be manually edited as required.
Create plots
One of the primary goals of this project was to provide the Police
Department with the ability to seek community and political participation.
To support this requirement, the R/BRS provides a menu of different
standard maps available to the user. The standard maps include:
full county maps, city maps, district maps, and beat maps. For
each map, the user is provided the option to change the elements
in the map composition. The various maps are available in standard
or E-size and color or black/white.
Additionally, the system provides the ability to create ad-hoc
maps. Ad-hoc maps are created by selecting individual components
for incorporation into ARCTOOLs compatible Views. The ARCTOOL
views can be used to generate Layouts and hard copy maps.
Label DIME file
As a part of the project, the Police Department required the system
to update the County's road center line file. The Center line
coverage which was originally derived from a USGS Dime file, contained
information on all Police Districts, Beats and PRAs. As new district
and beat configurations were adopted, the Dime file was required
to be updated with the new information.
Since, Police Districts, Beats and PRAs run along street centerlines,
the update process included information on both the Left and the
Right side of each street segment. The update process was performed
by a series of polygon overlays between the Dime file and the
various polygon Coverages. A manual edit tool was also provided
for further refining of the automated process.
Provide CAD interface
The Interface to the County's Computer Aided Dispatch system was
composed of two parts. First, the R/BRS was required to identify
the closest five beats for each beat in the County. This information
is used by the CAD operators to assign calls to the nearest beats
when officers from a particular beat are unavailable.
The system automatically calculates the nearest five beats for
each beat in the County. The system also displays and provides
a manual editing capability for modification of the nearest beat
tables.
Next, the system was required to output the nearest beat and the
district, beat and PRA configurations to a format compatible with
the County's CAD system. This requirement was satisfied by providing
text format tables compatible to the County's PSSI CAD System.
Recommend assignments.
Finally, the Police Department required the ability to generate
a report recommending the number of officers to be assigned to
each of the new districts based on workload and district configuration.
The system accepts as input the number of officers assigned to
each district and the number on temporary disability. Additionally,
the system accepts the number of new personnel available. The
R/BRS software then recommends the number of officers per district
using the weighted workload for each district as a factor.
Conclusions:
The assignment of resources to adequately provide services
in a large geographic area is often a complex and time consuming
process. Although in some cases it is possible to deploy additional
resources to areas of higher demand, such as assignment of more
sales persons in a high growth area. In many cases, the simple
assignment of more personnel will not solve the problem as in
the example of a Fire department where a combination of personnel
and facilities must be available in order to accommodate the demand
for services. In such conditions the application of the aforementioned
methodology can aid in the redrawing of geographic boundaries
and a more balanced workload demand between individual facilities.
The use of this system also provides a practical tool for long
term planning of future required facilities.
The R/BRS is currently in use at the Montgomery County Police
Department. The system can accept as its input both polygons (such
as PRAs) and points (such as location of individual calls). There
are current plans for porting a more simplified version of the
software to a PC based ARCView system.