Refinement of operational health physics methodology using geographical information systems


Material Information

Refinement of operational health physics methodology using geographical information systems
Physical Description:
vi, 198 leaves : ill. ; 29 cm.
Birky, Brian Kent
Publication Date:


Subjects / Keywords:
Environmental Engineering Sciences thesis, Ph. D   ( lcsh )
Dissertations, Academic -- Environmental Engineering Sciences -- UF   ( lcsh )
bibliography   ( marcgt )
non-fiction   ( marcgt )


Thesis (Ph. D.)--University of Florida, 1997.
Includes bibliographical references (leaves 188-197).
Statement of Responsibility:
by Brian Kent Birky.
General Note:
General Note:

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 027588434
oclc - 37146640
System ID:

This item is only available as the following downloads:

Full Text








I would like to thank my mentors, Dr. W. Emmett Bolch and Dr.

Charles Roessler for guiding me into the rewarding profession of health

physics. The patience of the other members of my committee, Dr. Paul

Chadik, Dr. David Hintenlang and Dr. William Properzio, throughout this

excruciatingly long process is humbly appreciated. Also, I especially

thank Mr. Ryan Richards. His knowledge of computer systems and

technical assistance with software were crucial to the success of this

research. In addition, I would like to acknowledge the contributions of

my fellow students over the years: specifically, Kurt Geber, Rich

Utrera, Debi Turner, Dave Harrison, Mike Domal, Diane Sinkowski, Susan

Stanford, Thabet Tolaymat, Peter LaPuma, George Harder and Tom Huston.

The steadfast friendships of Paul Kaiser and Art Schneider are

gratefully valued; the former for making me attend virtually every Gator

sporting event of the last decade and the latter for his unique ability

to stimulate scientific and philosophical curiosities. Finally, I would

like to thank my parents, Mr. Edgar Birky and Mrs. Belva Birky, for

their love and understanding.


ACIJOWLEDGMEIETS . ................. .. ii

Abstract .. .

INTRODUCTION . .. ....... 1
Purpose . ............... 1
Significance . . . 2
Literature Review . . 5
An Overview of GIS .... * 8
Summary . . . 14

Review of Traditional Method ... . 15
Data Acquisition and Translation for GIS . 19
Population Allocation Procedure Using GIS . 31
Case Study: DOE Pinellas Plant . .. 33
Detailed Case Study: St. Lucie .... .. . 37
Summary . . . 59

Estimate and Projection Concepts .... 60
Methods of Estimation ........ . 62
Elementary Postcensal Methods of Population Estimation 70
Population Projections . . 73
Case Study: St. Lucie Estimates and Projections 76

Overview ... . ....... * 88
Methods of Intercensal Population Estimation .. 93
Key Issues with Interpolation Techniques . .. 103
Case Study: Pinellas Plant Population Reconstruction .. 104
Summary .... . . 108

Scope of the Problem . . 110
Dealing with Error . . 120
Error in Digital Representations of Lines . .. 127
Polygon Overlay Processing . .. 130
Error Estimation for the Sector-Segment Grid .. 130

Overview ... . 136

Geographic Criteria for the Siting of Low-Level Waste Disposal
Sites . .... .. . 137
Environmental Modeling .... .. . 138
Census of Agriculture Data . .. 142
Case Study: The Cotter Mill Problem ... .144

Overview .... .. . . 155
St. Lucie Nuclear Power Plant Evacuation Estimates .157
Methodology: Evacuation Times Estimates . .. 174

Summary Statement ... . 182
Recommended Procedures ... . 183

LIST OF REFERENCES .... .. . 188


Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy



Brian Kent Birky

May, 1997

Chairperson: Dr. W. Emmett Bolch
Major Department: Environmental Engineering Sciences

The field of health physics draws upon other specialties for

information and technologies relevant to emergency planning and

radiation dose assessment. GIS has enjoyed only limited use in health

physics to date, but as its applications are demonstrated and the

computer hardware and software necessary for its implementation become

less expensive, the widespread use by many health physicists is

inevitable. This document presents a recommended protocol for the

determination of population distributions around nuclear facilities for

use as input data for estimating offsite radiation doses due to routine

and accidental releases of radioactive materials. Techniques were

developed to establish baseline population levels using U.S. Bureau of

the Census data and GIS technology, construct current population

estimates and future projections using the baseline data coupled with

predictive equations and reconstruct past population distributions using

current demographic databases. Concurrent to the fulfillment of these

objectives is an accounting of uncertainties associated with procedures

and parameters. These uncertainties propagate to yield a total

uncertainty for the population which in turn is an input parameter for

other models. The significance of this research is that techniques for

spatial data analysis are described from a practical perspective. The

reader is given the necessary information to choose a GIS platform,

obtain population and street mapping data, transform those data into

useful formats and interpret results. Proper interpretation of GIS

analyses is also important for those professionals who contract with

outside agencies for such services. Populations are reported as point

estimates with no regard for error in GIS manipulation or input data.

The problem stems from the resistance of software vendors to incorporate

uncertainty accounting algorithms within their programs and reluctance

to release proprietary codes to outside researchers. It is important

for health physicists to realize that they can effectively perform many

of their own GIS analyses, that the use of census levels that are too

aggregated may drastically affect decisions based upon GIS output and

that errors and uncertainties are generally ignored in the GIS industry

and must be semi-quantitatively estimated.



A Geographic Information System (GIS) is defined as:

a special case of information systems where the database consists
of observations on spatially distributed features, activities, or
events, which are definable in space as points, lines or areas. A
GIS manipulates data about these points, lines, and areas to
retreive data for ad hoc queries and analyses. (Deuker, 1975, p.

Many other definitions have been proposed and the ability of a GIS

to analyze spatial data is often seen as a key element in its

definition. This ability has been frequently used as the characteristic

which distinguishes the GIS from systems whose primary objective is map

production. The definition chosen to introduce this study exemplifies

the analytical approach that made GIS such an attractive tool for this

research. GIS technology is quickly breaking the bounds of cartography

and geography to become an important analysis tool for many other

disciplines in which spatially located data are accumulated. By

necessity, the field of health physics draws upon other specialties such

as biology, medicine, geology, electronics, computer science and others

for information and technologies relevant to emergency planning and

radiation dose assessment. GIS has enjoyed only limited use in health

physics to date, but as its applications are demonstrated and the

computer hardware and software necessary for its implementation become

less expensive, the widespread use by many health physicists is

inevitable. GIS is a powerful technology that can reasonably be placed

on each professional's desktop. This document presents a recommended

protocol for the determination of population distributions around

nuclear facilities for use as input data for estimating offsite

radiation doses due to routine and accidental releases of radioactive

materials. Techniques were developed to accomplish three requirements:

1. establish decennial census population levels using U.S. Bureau of
the Census data and GIS technology as a baseline,

2. construct current population estimates and future projections
using the baseline data and

3. reconstruct past population distributions using current
demographic databases.

Estimation of current and projection of future populations

necessarily involves strategies to redirect population growth in regions

where saturation, i.e., the attainment of a maximum population density

in a user-defined area which is characteristic of the region as a whole,

is a problem to adjacent areas where growth can be tolerated.

Concurrent to the fulfillment of these objectives is an accounting of

uncertainties associated with procedures and parameters. These

uncertainties propagate to yield a total uncertainty for the population

which in turn is an input parameter for other models.


The protocol for determination of population distributions

according to a complicated sector-segment grid using a GIS as set forth

herein has not been adequately described in previously published

literature. The same is true for the projection of future populations

using that data structure and the reconstruction of past populations in

areas of nuclear weapon production. In addition, a GIS can be used to

analyze agricultural production in the area surrounding a nuclear

facility so that consumption pathway parameters can be generated that

are site-specific rather than conservative default values. All of these

analyses yield point estimates using commercial software. This research

addresses this problem in a semi-empirical approach so that default

confidence bounds may be applied to results using the same or similar

GIS software and similar levels of geographic and census detail. All of

the preceding assessments must be performed or contracted for by health

physics staff for such vastly important purposes as calculation of

collective radiation exposures and radioactivity intakes to the general

public, reconstruction of past radiation doses to members of the public

at times when dose assessment methodology was not as sophisticated as

the present and emergency planning including evacuation modeling in the

event of a significant unplanned release of radioactive materials. In

short, collective doses based on population distributions assess the

public risk and this research offers new techniques to improve those

distributions. The collective dose to a population residing in a

particular sector-segment is calculated using the series of equations

that follow. The first equation calculates the dose to critical organs

or tissues of an individual from a particular type and energy of



where HTR is the equivalent dose to the organ or tissue
(sieverts) from a particular type and energy of

W, is a dimensionless "radiation weighting factor" which
accounts for differences in biological damage caused
by different types and energies of radiation

DTR is the absorbed radiation dose to that tissue or organ
in joules per kilogram (gray)

Summing over all types and energies of radiation to which the organ or

tissue is exposed gives a total equivalent dose. The next step is to

weight each organ for its own sensitivity to biological endpoints (fatal

cancers and other deleterious effects) and sum over all organs and

tissues to give a whole body effective dose.

E = E w H

Where E is the effective dose (sievert) for the whole body

W, is the tissue weighting factor that accounts for
tissue sensitivity

When the radioactivity is inhaled or ingested physical and metabolic

factors determine the retention time in the body and the dose

calculations are integrated over the subsequent 70 years after intake

for a member of the public. This is known as a dose commitment. The

collective equivalent and effective dose commitments are the integrals

over infinite time of the average individual respective rate quantities

due to a specified event, for the critical population group.

H = f H *dt
HT,coll to rate,T,avgdt

Where HT,coli is the collective committed dose equivalent to a
particular organ or tissue

Hrate,T,av is the average equivalent dose rate to the
tissue for the individuals in the critical
population group to be integrated from the time
of intake to infinity

The collective effective dose commitment is calculated by summing over

all tissues as shown before.

Radiation absorbed dose is commonly measured using integrating

thermoluminescent dosimeters (TLDs) in the environs of a facility. The

placement and numbers of such monitoring devices are critical for proper

dose assignment and this author recommends that at least one device be

placed as close as possible to the centroid of each sector-segment in

the population grid as well as known population centers. Since the

derived population number directly affects collective dose and these

doses are used to assess the operational quality of a plant (and

subsequently affect costly insurance premiums), the population should be

calculated as accurately as is reasonably achievable. In addition,

error bounds should be placed on those populations so that a calculation

employing propagation of errors can be performed to provide an

uncertainty for equivalent doses produced using them.

Literature Review

Some of the potential uses of GIS in health physics have been

addressed by others in limited detail as follows: data mapping and

conjunctive of a global positioning system (GPS) (Runyon, 1994; Sejkora

and Most, 1993), visualization (Chen, 1995), emergency response

(Mueller, 1995) and waste transportation routing (Sathisan and Chagari,

1994). Runyon states:

one of the most important factors in environmental assessment and
characterization is the ability to locate, evaluate, document and
relocate sample or data collection sites. Also fundamental to the
process is the ability to spatially relate analytical data to
those locations. Geographic Information System (GIS) tools for
building a spatial data set, for relating data sets and for
displaying and querying data are key to the integration process.
(Runyon, 1994, p. 9-4)

That paper goes on to describe the power of using both GIS and GPS

in tandem and the integration of portable gamma radiation survey

equipment. The study site is the area surrounding a non-operational

thorium milling facility. Results are presented as radiological

isopleths intended to be used for "a multitude of purposes from

direction of excavation activities to calculation of dose/risk to

exposed populations" (p. 9-4). This brief treatment does not offer

appreciable detail but may be of interest to persons who wish to use GPS

with the procedures of Chapter 7 of this dissertation which addresses a

more complex environmental assessment. Another recent publication

(Sejkora and Most, 1993) also discusses the combination of GIS and GPS

with the emphasis on mapping survey locations at environmental

thermoluminescent dosimeter placements. The authors refer to vendor GIS

data in this statement:

many of the database maps contain overlays of population and
demographic information based on 1990 Census Bureau information,
which can be useful in performing dose calculations and emergency
response and evacuation planning. (p. 51-52)

What is not stated is that many of these vendor packages of

demographic data are restricted to statistical metropolitan areas. They

are almost always constructed from the larger census regions, i.e.,

equal to or larger than the census block group. The census block is the

finest level of detail and should always be the starting point for

radiological demographic assessments when possible. Later in the paper

the authors again mention the use of GIS and population overlays as aids

in performing population dose estimates, planning evacuation routes and

estimating evacuation times for emergency planning. They further state:

geographical regions can be defined and delineated within the
database, and populations can be estimated in each region. Such
regions could be defined using compass sectors, distances, town or
city boundaries, major roadways, or combinations of such
artificial or existing delineations. (p. 52)

No references that specifically address those applications are

given. In contrast, this dissertation provides the detail missing in

the existing literature. In a Conference of Radiation Control Program

Directors (CRCPD) newsletter Peter Mueller of the Department of Energy's

Nevada Field Office reports:

to support FRMAC (Federal Radiological Monitoring and Assessment
Center) deployments and field exercises, geographic databases

consisting of co-registered "layers" of cultural, radiological,
aerial photographs, satellite imagery, and environmental data, are
being compiled by EG&G for the areas around all commercial nuclear
power plants and Department of Energy facilities. (Mueller, 1995,
p. 3)

This is a system in the development phase and appears to use

commercial off-the-shelf hardware and software. No publications in

support of this system's application regarding procedures have been

found. It appears that the system is an aggregation of hardware,

software and data; however, actual analysis protocols and data

considerations are not described. In a short (4 pages of text) paper,

population estimation techniques for routing analysis in support of

radioactive waste transport to the proposed repository at Yucca

Mountain, Nevada, are investigated (Sathisan and Chagari, 1994). The

discussion of the significance of that study points out:

population characteristics (total population and density) are
critical factors in the risk assessment, emergency preparedness
and response planning, and ultimately in route designation. (p.

The study compares alternate routes and a series of corridor

widths at block, tract and county levels of population aggregation.

This extremely limited study purports to examine the impacts of the

accuracy or level of disaggregation required for population data. They

correctly conclude that aggregation of data over extended lengths of

routes results in inaccurate estimation of critical population densities

along the routes but erroneously recommend use of tract level data

(rather than using block level data) in urban areas. An analysis so

important, i.e., the determination of the least hazardous route for

radioactive waste transport, must be based on the highest level of

detail (the block level). Potential savings in data cost, processing

time and storage requirements are trivial in this regard. Therefore,

the aggregation level sensitivity analysis is not prudent for their

overall objective. The authors did not describe the data acquisition

and translation process. The GIS analysis consisted of buffering

various widths around line segments--a simple, standard process for

which all such systems are designed thus requiring no particular

expertise or explanation.

The preceding publications relating health physics and GIS are few

and sparse in detail. There are many texts and journals devoted to GIS.

Reviews of that subject and population methodology are broader topics

that would not serve this study well. The reader should simply consult

the references found in the specific technical discussions presented


An Overview of GIS

Throughout the last 35 years there has been a very rapid rate of

theoretical, technological and organizational development in the GIS

field, culminating in a period of intense activity in the last 10 years

or so. Many decisions a health physicist must make are dictated or

influenced by some aspect of geography. A GIS allows spatial queries

that are not possible with other programs such as spreadsheets,

statistics packages, or drafting packages. A GIS also links disparate

data sets using common fields so that more complex queries are allowed.

Linking data sets by identical fields, or exact matching, is a

relatively simple process. However, not all data are collected with the

same frequency, or the same recording area. For example, environmental

data collected weekly can be grouped into monthly categories and linked

to a separate database of monthly information. The grouping could also

involve small areas aggregated into an exactly matching larger area,

e.g., census blocks to counties. Such temporal and areal groupings are

known as hierarchical matching. Note that the result of a hierarchical

matching process is an exact match. An exact match is not always a

possibility, especially when environmental data and artificial

constructs are involved. A common GIS architecture uses layering to

view and combine non-matching data sets. The concept is similar to

laying transparent maps one over another. With the spatial reference

system as the only commonality, layers can be combined to form a new

layer with all of their properties. Only then is it possible to perform

spatial queries. Combined databases have increased value because they

can be used to solve more sophisticated problems. The combination of

non-matching layers is known as fuzzy matching.

Once a spatial reference system is in place and spatially

referenced data are supplied, some questions that are uniquely answered

by a GIS are readily resolved. Inquiries about location, condition,

trends, patterns and more complex modeling are particularly suited to

this system. To begin an analysis of the area surrounding the Pinellas

Plant, for example, the GIS user may enter known latitude and longitude

coordinates, or supply the street address so that the site location is

found and marked. Places of interest in the vicinity can be located by

place name or zip code. Suppose that the user wants to find a new site

for a similar facility. The constraints are that the site must be on

the coast of Florida with a low population density in a radius of 50

miles. A spatial analysis must be done to find a location where these

conditions are satisfied, Once such a location is found, the user might

be interested in the rate of growth from 1980 to 1990, i.e., the local

trend in population growth. Other factors to consider are background

radioactivity and radiation levels and local cancer incidence. When

spatially referenced, such data reveal anomalies or patterns that may be

of concern. Finally, the GIS can assist in modeling effects of altering

local geographic features for facility construction and modeling

theoretical effluent releases and deposition in the environment. These

problems are extremely complex and deal with even more complex natural

and artificial systems. The use of systems models coupled to GISs

creates powerful problem-solving platforms that may represent the most

cost-effective way of analyzing and then making decisions concerning

environmental issues. However, there exists a danger that the use of

GIS technology will grow so much and spread to so many fields that it

will forever escape the grasp of the specialists who helped create it.

Data Management

It is not a trivial distinction that the user is part of the GIS.

Therefore users need to know the consequences of their selective use of

GIS as an analytical tool--not a map-making computer. A database

management system (DBMS) is an integral part of all modern geographic

information systems. The user's most direct interaction with the basic

components of the system is through manipulation of attribute or spatial

data through the DBMS. Present DBMS technology is very good for managing

tabular data (adding, deleting or modifying records), but even simple

mapping involves much more than just the storage and retrieval of data.

DBMS technology is not very effective for updating and managing

cartographic data, except the attribute data associated with spatial

features. This is so because of the topological relationships between

map features. When, for example, a land parcel's polygon is adjusted,

the relations of surrounding parcels to the first parcel and to each

other must also be adjusted; the geometric attributes of all these

parcels change.

The Base Map

The Topologically Integrated Geocoding and Referencing (TIGER)

File developed by the United States Bureau of the Census can serve as a

major component in a geographic information system. The principal

cartographic components of the TIGER File are map features (streets,

rivers, railroads and so forth) captured in computer-readable form. The

Census Bureau accomplished this task by establishing a cooperative

agreement with the U.S. Geological Survey to prepare the initial digital

base file. Appended to each of these map features (lines) is various

geographic information that also serves as a cartographic enhancement;

for example, feature names, address ranges, ZIP codes and so forth.

The network of lines formed by these features encloses spaces or

areas. The TIGER File contains, appended to these area records, a set

of geographic identifiers that also serve as cartographic enhancements;

for example, 1990 census block numbers, census tract numbers, city

codes, metropolitan statistical area codes and more. The TIGER File, in

fact, contains two sets of such codes: one representing the geography at

the time of the 1980 census and another representing the geography at

the time of the 1990 census. Thus, the first comparison that can be

made in a GIS environment is one that determines areal differences from

one census to the next.

From a classic GIS standpoint, the geographic area codes provide

the mechanism to link the TIGER File's cartographic base to other Census

Bureau statistical data sets such as the data tabulations from the 1980

decennial census (recorded on the several 1980 census summary tape

files), the 1990 decennial census and the numerous data sets from the

several economic, agriculture and governments censuses. The definition

of the geographic areas represented in the latter censuses may be

slightly different from what existed in either the 1980 or the 1990

decennial censuses.

The TIGER Files carry latitude and longitude coordinate values for

the intersections and end points of the lines comprising the map. These

latitude and longitude coordinates permit a GIS to relate the underlying

cartographic framework of the TIGER File to data sets from other

agencies that are defined by coordinate values; for example, the land

use/land cover data sets available from the U.S. Geological Survey, the

water resources data and geological data sets also available from the

U.S. Geological Survey, the soil data sets from the Department of

Agriculture's Soil Conservation Service, the Landsat imagery available

from EOSAT Corporation and numerous data sets available from state,

county and local units of government that have developed digital files

to represent information that interests them.

Attribute Data

The U.S. Bureau of the Census Geography Division provides similar

geographic support services for all of the Census Bureau's programs:

the decennial census, the economic and agriculture censuses, the

intercensal demographic estimates and the current surveys. The scope

and complexity of this geographic support task has changed over the

decades and the Census Bureau has changed its methodology for conducting

censuses and surveys in response.

To meet the geographic support needs of the 1990 Decennial Census,

the Census Bureau built a national automated geographic (areal) and

cartographic (map) data base called the TIGER file described previously.

This file uses a variety of new approaches that deal with the geographic

problems associated with the production of over 300,000 individual

enumerator assignment maps, appropriate maps for the field supervisory

and office staff, maps to accompany the tabulated results, automated

matching of addresses on the mailing lists the Census Bureau uses in

many of its data collection activities and preparation of computer files

that document the relationship between and among all the geographic

tabulation units for the entire United States, its territories and its

possessions. The latter is the same set of geographic units that would

be seen by looking at a full set of Census Bureau maps spliced together

into a single sheet.

Hardware Considerations

GIS applications must access increasingly large data sets,

typically several hundred megabytes for a 50-mile sector-segment grid to

several gigabytes for regional environmental modeling. The

computational activity tends to be I/O dominated, with comparatively

little time devoted to mathematical manipulation (Hendley, 1990). Rhind

(1988) goes further in explaining the tendency for I/O domination of

computational activity by stating that while the type of questions asked

of geographical data are conceptually simple, they may involve

exhaustive searches of a large database or require multiple prior

indexing. This requirement can be illustrated by examining typical GIS

operations, such as polygon overlay used in sector-segment population

determination, where a single piece of data has to be retrieved and

manipulated a number to times in the course of the operation before

finally being output (Harding and Hopkins, 1993). GIS algorithms are

usually written to take account of the fact that data volume is much

larger than system memory. Therefore, these algorithms use local disks

for temporary storage, thus compounding the I/O overhead of a GIS

application. This tendency towards I/O domination makes disk

performance a critical factor in optimizing GIS performance. The

interaction of processor time and I/O is crucial for computational

performance. Considering these factors, the GIS application is

optimized by increasing the utilization of the processor by minimizing

any I/O and increasing the processing capacity of the computer. Since

I/O is usually the bottleneck, it must be addressed first, before any

benefit from increased processing capacity can be achieved. The bottom

line on hardware is to buy a system with the fastest processor available

and a large hard drive with fast access and data transfer times.

Multiple processor systems are only necessary for extremely large

geographical areas, large attribute files and frequent analyses. For

most purposes, an off-the-shelf personal computer of high quality will

suffice. Select a brand name model since some vendors cut costs by

using lower grade components in models that appear to be equivalent, but

will perform poorly. Get the fastest processor available coupled with a

large (gigabytes in capacity) hard drive with a fast access time and

rapid transfer rate. Another necessity is a high speed CD-ROM drive to

access Bureau of the Census data.


Clearly, the application of GIS techniques to health physics

problems is beginning to happen, although the literature discussions to

the present have been superficial. The task of performing GIS analyses

is falling upon health physicists and their staffs with little guidance.

The protocols of this document fill that void. An overview of the GIS

concept was presented that serves as a foundation for the more rigorous

treatments that follow. The next chapter explains how to acquire

geographical and population data, prepare those data for use and analyze

those data in a GIS.


Review of Traditional Method

Nuclear facilities routinely record and report all emissions of

radioactivity to the environment. Although these emissions are easily

quantified via devices such as stack-mounted detectors, the eventual

radiation doses to members of the public residing near the facilities

are too low to be measured. Modeling is the only methodology to insure

compliance with Appendix I, 10CFR50. Mathematical models and the

computer programs developed from them require input of the numbers of

persons (population group) at a specific location. The population group

may be the total number of persons, a particular age grouping, one sex,

or a combination of age and sex. These populations are aggregated by

polygonal locations defined by directional radii and annular distance

from the release site. The population-location array in conjunction

with the release data radionuclidess, chemical forms, quantities, rates,

dates and times) and pathway-specific data yield collective, individual

maximum and average radiation doses to the public. Pathway specific

data for aereal emissions include release height, wind speed and

stability class. For releases directly to tributaries or potentially to

groundwater, the data demands can become more stringent involving

parameters such as river flow rate, or depth from ground surface to

groundwater, underlying geology, transverse and longitudinal

dispersivities, transmissivities, mixing volumes, etc. The resultant

doses carry a degree of uncertainty generated by all the parameters used

in their calculation. This will be addressed in Chapter 5. The focus

of this chapter is the determination of populations by location for use

as a model parameter.

The map overlay methodology is essentially a technique, in

environmental planning, to summarize many spatially distributed

variables. In "Town and Country Planning Textbook" where Tyrwhitt

delineated the scope of a planning survey in the chapter "Surveys for

Planning," she also described, almost as an adjunct, how such surveys

should be recorded; her discussion may have been the first explicit

explanation of the map overlay methodology.

As far as possible maps should be drawn on transparent paper, so
that when completed the maps to the same scale can be "sieved" -
i.e. placed one on top of another in turn so that correlations or
their absence can be noted. Where relevant "Sieve maps" should be
made bringing out the degrees of correlations noted on certain
matters. Only when this work is complete can the plan be formed,
for, until then, many important factors in determining the lines
of the plan will not be brought into focus. (Tyrwhitt, 1950, p.

In the 1960s, Ian McHarg further popularized the overlay

methodology by many planning studies based on an overlay process. These

included the Staten Island Study, published in "Design with Nature."

Later, a study in 1962 by Christopher Alexander and Marvin Mannheim at

the Civil Engineering Systems Laboratory of the Massachusetts Institute

of Technology entitled "The Use of Diagrams in Highway Route Location:

An Experiment" displayed the analytical strength of the methodology in

dealing with a complex locational problem. Steinitz et al. (1976, p.

447-448) described the three particularly important elements of this

study in relation to development in the map overlay methodology:

twenty-six different diagrams, each depicting the relative
"utility" or desirability of the highway location for one specific
requirement of the highway, are considered for analysis. The
study recognizes that not all factors can be geographically
overlaid at one time, and proposed a formal procedure "tree" for

the combination and weighting over of overlays.... This study
surpasses traditional overlay applications in several ways: in its
use of an explicit weighing system for the combination of
different components, in the idea of superimposing the different
diagrams or maps photographically as separately timed exposures on
one print, and in the procedure used to structure the synthesis of
the 26 different diagrams.

In their summary remarks, Steinitz et al. reiterated the

traditional and fundamental aspects of the map overlay methodology in

environmental planning. Although they found, no doubt, the applications

at the time of their study had become more complex, diverse and

technically sophisticated, Steinitz et al. stated that the use of the

overlay approaches differed little between their early development and

the applications then.

... It is clear that the basic methodology and the underlying
logic have changed little. We combine data maps of soil, slope,
and other elements in the same manner that Warren Manning probably
did in 1912. We commonly hand draw each data element as a
complete separate maps through overlays visually, by redrawing
or by photographic reproduction into some sort of composite
representing the analysis. (Steinitz et al., 1976, p. 449)

However, the concept and the approach in map overlay have changed

in the 1980s indeed. Ian McHarg had applied the methodology in a manner

quite unlike the earlier applications; he had computed various summary

statistics, including some cost-benefit measures, for his plans in

entirety and for individual components of those plans. These exercises

by McHarg offered the obvious opportunity of extending the map overlay

methodology beyond its original, purely visually oriented descriptive

functions into the truly analytic, prescriptive realms. The drawing and

redrawing of zones on the overlay, not just the overlaying of

transparent maps; the generation of design alternatives, not just the

evaluation of alternatives can be done with a combination of design

intuition and analytic thoughts. The preceding description of the

evolution of geographical analysis parallels the homologous process for

determination of populations around nuclear facilities.

Population and other demographic data should reflect the most

accurate methodologies available and should be presented in the most

professional manner possible. The importance of accurate estimations of

collective radiation doses to non-occupational populations demands the

modernization of standard techniques. The traditional method for

determination of population distributions can be summarized in the steps


1. Transfer census tract and block group boundaries by hand from
library maps to USGS quadrangle maps,

2. check county highway maps for more recent revisions,

3. overlay sector-segment lines and arcs by hand,

4. assign percentages of block groups to sector-segments by using
ratios of domiciles counted in the block groups,

5. multiply fractions by populations for the block groups making up a
sector-segment and sum in a spreadsheet,

6. use published economic and demographic forecasts to construct
growth curves for counties,

7. apply growth factors to sector-segments to predict future
distributions and

8. plot current and predicted distributions on a radial graph.

For the 1990 census, the Census Bureau developed the TIGER base

map as previously described. The TIGER File contains vertices (lines

and points) that define items such as roads, voting districts, census

regions and their names and address ranges. The average file size for a

state is 500 megabytes (7.5 Mb per county). The population data are on

compact disks (P.L. 94-171) containing 28 variables such as population

by race and age. The computer hardware and software necessary to take

advantage of these databases are now commonly available.

Data Acquisition and Translation for GIS

TIGER files are available on CD-ROM directly from the U.S. Bureau

of the Census or from federal repositories. The compact disks are

labeled with the State(s) they contain. If the user plans to frequently

access many locations within the U.S., it may be advantageous to

purchase a complete set from the Bureau. The TIGER data supply a

geographical framework (albeit planar) useful for visual reference,

i.e., the sector-segment grid can be overlain like the traditional hand

method over a printed map. The files that must be used for population

analyses are the census boundary files which contain no information on

roads, waterways, etc.

The GIS software of choice for this investigation is "Atlas GIS"

developed by Strategic Mapping, Inc. Geographic census boundary data

from the census bureau must be first translated to an acceptable form

(BNA file format) and then imported for use in this platform. The first

step may be accomplished using "BdryBNA Translator" available from Micro

Map & CAD. Use of the BdryBNA Translator is straightforward. Execute

the BdryBNA.EXE program and answer a series of prompts. The first

prompt asks for the name of the file(s) to be translated including the

full path and file name of the TIGER file(s). Use of wildcard characters

(* or ?) is permitted for multiple file translation at a single command.

Pressing the key will exit the program. If standard TIGER file

naming conventions are used, the shape coordinate file will

automatically be assigned by substitution of the "F42" extension for the

"F41" extension of the basic data TIGER file. If this file is not

present, the translator will now prompt the user for the shape

coordinate file name. The next prompt asks for a choice of one of there

projection options that are listed below.

1. None This outputs the data points in geographic
coordinates, i.e., longitude and latitude.

2. Mercator This outputs data points in the Mercator Projection
using a spherical model and Central Meridian of 75.00
West Longitude.

3. UTM The Universal Transverse Mercator Projection is the
projection used by the U.S. Geological Survey for many
of their maps and data sets.

The last prompt asks for a choice of boundary options. There are 24

options of which the following listed by code in Table 2-1 are relevant

to this study.

Table 2-1 Boundary Options
1 = County 13 = AIR

2 = Zip 14 = ANRC

3 = Census Tract 17 = County 1980

4 = Block Group 22 = Census Tract 1980

5 = Blocks 23 = Block Group 1980

12 = AIR FIPS 24 = Block 1980

An alternate command line mode allows multiple file processing

utilizing the DOS batch file capability. Run the BdryBNA program with

the following parameters:


1. Basic File Name: Valid DOS path filename and extension for
Basic Data File Type 1.

2. Shape File Name: Valid DOS path filename and extension for
Shape Data File Type 2.

3. Additional Codes: Valid DOS path filename and extension for
Additional Codes File Type 3.

4. Projection Choice: 1 = None, 2 = Mercator, 3 = UTM

5. Boundary Choice Codes in Table 2-1.

6. Output File Name: Optional destination filename for BNA

For example, a command line for processing the Waynesboro, VA

TIGER file to produce the "Waynes.BNA" file of census tract boundaries

in UTM projection for Atlas GIS might look like the following line.

BdryBNA TGR51820.F41 TGR51820.F42 TGR51820.F43 2 3 Waynes

The BdryBNA Translator converts basic data and shape coordinates

into BNA regions and curves. The BNA feature naming utilizes the codes

found in the TIGER Basic Record File. The codes listed in Table 2-2 are

a selection of those deemed useful for this study.

Table 2-2 TIGER codes
Code and Region Description

1. County: State FIPS Code + County FIPS Code 51820

2. ZIP: Five digit postal ZIP Code 80226

3. Tract: State + County + Tract Number + Tract Suffix

4. Block Group: State + County + Tract Number + Tract Suffix +
Block Group 518200106.002

5. Block: State + County + Tract Number + Tract
Suffix + Block Number 518200106.00239A

12. AIR FIPS: State + FIPS American Indian Reservation

13. AIR: State + 1990 Census American Indian
Reservation AIR,TJSA,TDSA,ANVSA 511234

17. County 80: State 80 + County 80 51820

22. Tract 80: State 80 + County 80 + Census Tract Number 80
+ Tract Suffix 518201234.12

23. Block Group 80: State 80 + County 80 + Tract Number + Tract
Suffix + Block Group 80 518201234121

24. Block 80: State 80 + County 80 + Tract 80 + Block 80

The TIGER data sets are compilations from two existing sources, USGS

1:100,000 scale DLG files and 1980 GBF/Dime-files. TIGER files contain

both graphic representation information and address ranging. The

address ranging is only available for street networks within the

GBF/DIME-File coverage of the 1980 Census. Areas outside of these

GBF/DIME areas have been completed with USGS DLG 1:100,000 graphics but

do not have address rangirin data fields at this time. Distribution

coverage is by county or county equivalent. Each county will typically

be made up of 12 files shown in Table 2-3.

Table 2-3 TIGER files
1. Basic Data Records Nodes and Geocoding information

2. Shape Coordinates Additional points between nodes

3. Additional Decennial Census Geographic Area Codes

4. Index to alternate Feature Names

5. Feature name list

6. Additional address Range and Zip Code data

7. Landmark Features

8. Area Landmarks

A. Additional Polygon Geographic Area Codes

I. Area Boundaries

P. Polygon Location

R. Record Number Range

BNA is a simple ASCII file format for importing and exporting

boundary line data for use in ATLAS GIS (AGIS). Text data in a BNA file

describes region, curve, circle and point entities. Once a data resource

is translated to this BNA format it can easily be imported for use with

ATLAS mapping products using the Import/Export module available from

Strategic Mapping. A simple DOS command line to perform the AGIS Import

might look like the following line.

AGISIE TGR51820.BNA Waynes.AGF /Names 1.

The BNA format does not limit the number of boundaries in a single

file. The BNA format does, however, limit regions and curves to 4,000

vertices. If any line or polygon in the DLG data exceeds this limit it

is automatically broken into two or more individual curves. BdryBNA

Translator writes the ID number of any boundary which is not closed into

a text file called "LOG.RPT". This file will include the IDs of any

boundaries broken into smaller chains as well as boundaries which are

incomplete in the original TIGER data. This file is appended with each

successive execution of the BdryBNA Translator.

Map makers have long struggled with the problem of portraying the

rounded surface of our Earth on flat media such as paper or computer

monitors. Many methods of projecting the Earth's curved surface into

X-Y coordinate systems have been devised over the years. Each has its

advantages but all are compromises which distort the true three

dimensional aspect of the Earth to fit the two dimensions of paper or


TIGER graphic data are distributed by the Census Bureau in a

standard geographic coordinate system meaning that each point is defined

in space by a longitude and latitude. The geographic coordinate system

provides the most flexibility for use within Atlas GIS. BdryBNA

Translator adds the option of converting this geographic coordinate

system to the widely recognized Mercator projection based on a spherical

assumption and a central meridian at 75 degrees west longitude. The

BdryBNA UTM projection option adds the capability of overlaying common

data sets distributed by the U.S. Geological Survey such as Digital Line

Graph and 1:24000 scale Digital Elevation Model data. The UTM coordinate

system has been used for many published large scale maps. This

coordinate system divides the Earth into 60 North to South segments each

of 6 degrees longitudinal width. By dividing the earth into smaller

segments the distortions of the Transverse Mercator projection are kept

to manageable proportions. Translation to other coordinate systems such

as a local State plane system are made possible by the use of a

coordinate translation program like Tralaine from Mentor Software, Inc.

Population Data

A census is defined as a complete canvass and enumeration of the

population and information is collected pertaining to all responding

inhabitants of the geographic area under investigation. A census is

conducted in this country every 10 years in years ending in zero.

Detailed census data are crucial in the production of population

estimates and projections. The Bureau of the Census initiated the State

Data Center program in 1976 to establish an information network at the

state level to help store and disseminate census data. Each state data

center in turn establishes its own network of affiliate agencies

throughout the state. The State Data Center program has been a

successful one that provides local outlets for census data. A modest

fee is usually charged for services involving the processing of data

tapes, providing printed reports, or providing printed copies of census

reports from media such as Microfiche or electronic media.

Population projection methodologies often require data from two

successive census periods to establish certain parameters or to

calculate population growth rates from one time period to the next. It

is therefore useful to note data products for 1990 that are comparable

to or incompatible with 1980 census data products.

Apportionment Data

Data for the latest decennial census were collected on census day,

April 1, 1990 and then released on gradually with most data products

being released between 1991 and 1993. The first major official data

product from the 1990 census must be submitted to the president by

December 31 of the census year and fulfills the constitutional mandate

to conduct a census of population every 10 years for the purpose of

apportioning the House of Representatives. A more important mandate for

this study is the delivery by April 1 of the year following the census

to each state legislature and the governor of each state the population

counts mandated by the provisions of Public Law (P.L.)94-171. The P.L.

94-171 data consist of total population; population by race for white,

black, Asian and Pacific Islander, American Indian, Eskimo and Aleut and

other, population of total Hispanic origin; and cross-tabulations of

data for persons not of Hispanic origin by race. The P.L. 94-171 data

are cross-tabulated for all ages and persons 18 years old and over for

many geographic areas such as state, county, place, census tract/block

numbering area, block group and block. Data for P.L. 94-171 are

released on CD-ROM. These data must be matched to census regions as

spatial attributes. With the release of the P.L. 94-171 data, the U.S.

Bureau of the Census completes its legal obligation with respect to the

release of data, but valuable summary data sets are prepared and

released as well.

Computer-Accessible Data

One advantage of computer tape files is that the data are directly

available in machine-readable form which may expedite the production of

population estimates and projections in some cases. Summary tape files

(STF) are extremely useful in the production of estimates and

projections. Summary tape files contain much more detail than is

usually available in conventional printed reports. Four STF products

have been designed for the 1990 census, with STF series 1 and 2

containing 100% data and STF series 3 and 4 containing sample data.

Each STF contains more than one data file and is labeled STF 1A, STF 1B

and so on. STF 3B contains data at the zip code level and may be of more

use than other filtered sets.


The four data files of STF 1 include population and housing counts

and characteristics. STF 1A contains data for states and state subareas

in a hierarchical file structure down to the block group level. The

hierarchical file structure allows users to select data at any level of

geography included on the tape (now on CD); for example, one might

select population counts at the county level, or for the census tracts

defining the county, or for the block groups comprising the census

tract. STF 1B provides similar data down to the block level. STF 1C

and STF 1D provide similar data at levels that are not very useful for

the purposes of this study.

The STF 2 series consists of three files that exhibit different

file structures, and contains population and housing data similar to the

STF 1 series but in greater detail including data by race. STF 2A is

structured around the metropolitan statistical area (MSA) level of geog-

raphy. Data are presented for each state at the census tract/block

numbering area (BNA) for MSAs and the non-MSA portion of states. Data

are presented also at the census tract/BNA level by county and places of

10,000 or more inhabitants and for whole census tracts/BNAs.

STF 2B contains similar data in an inventory-type file structure

with each geographic area presented without linkage to those above or

below it, unlike the linkage found in hierarchical file structures.

Data are presented for the following geographic areas in STF 2B: states

(including summaries for urban and rural), counties, places of 1,000 or

more inhabitants and MCDs of 1,000 or more inhabitants in selected

states. Data also will be presented for the state portion of American

Indian and Alaska Native areas. STF 3C presents data in a hierarchical

file structure for the following geographic areas: United States, census

regions, census divisions, states (including summaries for urban and

rural), counties, places of 10,000 or more inhabitants, MCDs of 10,000

or more inhabitants for selected states, MCDs of fewer than 10,000

inhabitants in New England MSAS, American Indian and Alaska Native

areas, metropolitan statistical areas and urbanized areas.

Data on STF 1 and STF 2 represent 100% items asked on all census

forms and these data are more likely to be used in the production of

population estimates and projections. STF 3 and STF 4 contain data

based an a sample of respondents from the 1990 census, but the content

is usually of little use in the preparation of estimates and projections

with the exception of STF 3B which provides population and housing data

at the 5-digit zip code level for each state. A second exception is a

special computer tape file pertaining to migration data containing all

intrastate county-to-county migration streams and significant interstate

county-to-county migration streams.

In some applications, particularly those involving population

projections, it is necessary to have data for an individual level of

geography at two points in time. Most often one requires data from the

two most recent censuses, but population reconstructions for nuclear

facilities may extend to the early days of research. The summary tape

files for the 1990 census are similar in content to those released for

the 1980 census, but are unavailable for previous censuses. There is no

guarantee of a one-to-one correspondence between geographic areas in

1980 and 1990 and in some cases one may count on significant changes.

For example, only a portion of the nation was defined in terms of census

blocks in 1980, but the entire nation was blocked for the 1990 census.

In addition, smaller areas of geography are subject to changing

boundaries over time. Incorporated places such as cities or towns may

annex additional territory between the two census periods and census

tracts or block numbering areas may be redefined from one census to

another. Caution should be exercised when comparing data from two

census periods for what appears to be the same area of geography. For a

1980 to 1990 evaluation the first task is to compare census boundaries

as supplied in the TIGER files. Not all data needs, however, may be met

through census data and other sources may be used as illustrated in the

detailed examples of the St. Lucie Power Plant population distribution,

population projections and evacuation times estimates.

Sample Survey Data

The Current Population Survey (CPS) is conducted by the U.S.

Bureau of the Census on a monthly basis. Initiated in 1942, it was

designed in response to the need for current information on unemployment

during the years of the Depression. The current version of the CPS

focuses on employment and unemployment, other characteristics of the

general labor force and the population of the United States as a whole

based on a large national sample size. Of particular interest in the

CPS are data from several of the monthly supplements such as quarterly

data on housing vacancy rates, information on migration and household

composition, data on fertility, and data on school enrollment.

Vital Registration Data

The federal government does not actually collect vital event data

directly, but rather sets standards for data quality and then collates

and reports data for births and deaths from all states. Vital

registration data for the states are collected and reported by the

National Center for Health Statistics, with annual data reported in a

multi-volume publication, "Vital Statistics of the United States." Of

particular interest are Volume I, on natality and Volume II, on

mortality. Monthly data are also published in a "Monthly Vital

Statistics Report" series. Annual reports include data at the county

level, and monthly data are presented at the national and state levels.

Vital registration data on births and deaths are frequently of use in

the production of population estimates and projections.

Each state operates its own vital registration system and more

detailed data are usually reported by the state than are available in

the U.S. summaries. One is also likely to have access to vital event

data more quickly when obtaining it directly from the state. Local

governmental units may serve as an additional source of vital event data

and may be preferred when projecting at small geographic levels (sub-


Additional Data Sources

U.S. Bureau of the Census population projections are published as

part of "Current Population Reports," Series P-25, on population

estimates and projections for the United States and individual states

and on special populations. Frequently included in the publication are

population projections produced at the state level by members of the

Federal State Cooperative for Population Projections (FSCPP). FSCPP

members are a good source for locally produced population projections,

especially when one requires detail below the state level as in this


The P-25 series of the "Current Population Reports" also publishes

the results of the Bureau of the Census's population estimation program.

The population estimation program is somewhat more ambitious than the

population projection program. Annual estimates for the United States,

all states and all counties are produced and published by the Bureau of

the Census. Subcounty estimates have been published on a biennial basis

as well. County-level estimates are produced by the Bureau of the

Census as part of its work with the Federal State Cooperative for

Population Estimates (FSCPE), which may represent an alternative source

of data. Many FSCPE members produce their own population estimates at

the state and substate level. More information about the FSCPE reports

and their usefulness is presented in Chapter 3.

Additional publications of the U.S. Bureau of the Census that may

provide helpful information in the population estimation and projection

process include the "Statistical Abstract of the United States," and

the "County and City Data Book." The "Statistical Abstract of the

United States" is a good general resource for a variety of data in

addition to the general sections on population and vital statistics at

the state level. The "County and City Data Book" provides general

population and economic data at the county level and for selected cities

in the United States. Migration data are the most difficult to obtain

in the process of producing population estimates and projections.

Estimates of migration are provided by the Immigration and

Naturalization Service (INS). Raymondo (1988) has provided estimates of

the undocumented population at the state level using the Bureau of the

Census estimate of 200,000 annual undocumented immigrants and the

estimated distribution of the undocumented population from the 1980


The migration file produced by the Bureau of the Census from

Internal Revenue Service data on matched tax returns is a useful source

of migration data. Symptomatic indicators of migration may be necessary

when working at levels below the state. Some of the more frequently

chosen symptomatic variables include driver's license applications,

motor vehicle registrations, voter registrations, utility hookups and

school enrollments. School enrollment data may be obtained from local

school districts or from state-level departments of education. Utility

data may be more difficult to obtain, especially for privately owned or

incorporated utility companies.


Individuals included in group quarters are those in institutional

settings of some type, such as college dormitories, prisons, long-term

health care facilities, military bases and related types of group

housing. These categories are important to assess in activities like

the evacuation planning demonstrated in Chapter 7. College dormitory

counts are best obtained directly from the institution as is the

population for long-term health care facilities. Prison populations are

usually available at the state level from a state department of

corrections or related agency. The Directorate of Information of the

U.S. Department of Defense is a good source for data on military base

populations,and each branch of the military has an information office

that might provide information for its particular bases.

The U.S. Bureau of the Census provides a referral service

identifying the names of both private and public data sources in its

"National Clearinghouse for Census Data Services" address list. There

are many private vendors that specialize in providing demographic

products including current population estimates and projections.

Population Allocation Procedure Using GIS

The starting point for each estimate is the 1990 population

enumerated by the U.S. Bureau of the Census (these numbers have not been

adjusted for possible undercount or overcount). Consequently, this is

also the starting point for sector-segment projections. TIGER files

must be translated to regions to match the P.L. 94-171 database. The

translated boundary files are then imported into the mapping software.

The P.L. 94-171 data are subsequently matched to the census block

regions as map attributes.

A much more accurate estimate of 1990, present and future

population distributions can be provided than possible with data and

technologies of previous censuses. The mapping software chosen for this

study has an imbedded dBASE III+ format and will calculate populations

for any polygon constructed by automatically assigning fractions, called

"area weighting," of census blocks to sector-segments. The algorithm is

less likely to fractionalize a large populace for a county with a finely

divided population, i.e., one containing a large number of blocks. In

the 1980 Census, Alachua County, Florida contained seven census tracts.

It is clear that uniquely defined boundaries that intersect portions of

those seven tracts would present difficulties in allocations within the

intersection. For the 1990 Census, Alachua County, Florida contains

over 4,500 census blocks. Since Alachua County contains a total

population of 181,596 persons, this focuses the location of about every

group of 41 people by a quantum leap in resolution. The computer

hardware and software system must be designed to handle this level of


Grid Construction and Data Assionment

A sector-segment grid is constructed originating at the latitude-

longitude coordinates of a facility with 16 sectors centered on the

compass points and radii in 1 to 10 miles, or corresponding metric

increments, to a total of typically 50 miles or 80 kilometers. This is

done by using polygon buffering algorithms imbedded in the software

program. This is not a perfect system in that no two grids are exactly

the same. An apparently uniform grid can have variation. Demographic

data associated with census blocks or block groups are assigned to

sector-segments by area-weighting as previously described.

Case Study: DOE Pinellas Plant

Eight counties are involved within the 50-mile radius of the DOE

Pinellas Plant. These are: Hardee, Hernando, Hillsborough, Manatee,

Pasco, Pinellas, Polk and Sarasota. Over 64,000 census blocks are

included in those counties. Since the total number of blocks is large,

an extensive number of blocks are contained or fractionated by even the

smallest sector-segments in the grid (See Figure 2-1). A sector-segment

grid was constructed originating at the latitude-longitude coordinates

of the DOE Pinellas Plant with 16 sectors centered on the compass points

and radii of 2, 3, 4, 5, 10, 20, 30, 40 and 50 miles. Figure 2-2 shows

the site with the top of the page as north and the 50-mile grid

providing the scale.

Figure 2-1 Census Blocks Within the One-Mile Grid for the
Pinellas Plant

Figure 2-2 Pinellas Plant and the 50-Mile Grid

The allocation of populations to the sector segment grid is accomplished

using the following steps.

1. Acquire TIGER CDs for the area of interest and translate the
census boundaries according to GIS file format (BNA for Atlas) and
map projection (or latitude/longitude pseudoprojection).

2. Acquire P.L. 94-171 data for the same area and extract population
data at the summary level in Step 1, e.g., block boundaries and
block summary level. The P.L. CDs contain an extraction program
to facilitate this procedure.

3. Import translated boundary files into the GIS using an import
program such as Import/Export for Atlas.

4. Construct common identity fields between the translated geographic
database and the extracted population database so that each census
block polygon is linked to its specific population count. It is

sometimes easier (and faster) to use an independent database
program to complete this step rather than the DBMS imbedded in the

5. Use the GIS ring buffering commands to create annular rings at
user-specified distances, e.g., 1, 5, 10 miles from a
latitude/longitude site coordinate.

6. Use line buffering commands to construct lines passing through the
origin point that dissect the rings at user-specified radii, e.g.,
so that sectors are centered on the sixteen major compass points.

7. Use the "select" command to select each sector-segment to verify
individually that each is now an independent buffered region and
that no artifact polygons have been created.

8. Perform the GIS analysis "operation" for each sector-segment to
provide summary statistics for population (POP100) and other
fields of interest, e.g., land area and total area. This step can
be automated using a simple script language to write a subroutine.

Tables 2-4 and 2-5 show the 1990 populations by sector-segments. In

addition, it might be useful to have cumulative population figures; i.e.

0-2, 0-3, 0-4, 0-5, 0-10, 0-20, 0-30, 0-40 and 0-50 miles for each of

the 16 compass directions. The cumulative totals are listed in Tables

2-6 and 2-7. These total are displayed graphically as Figure 2-3.

Table 2-4 Pinellas

Plant Populations to 5 Miles

0-2 2-3 3-4 4-5
N 2,421 5,829 5,564 6,763
NNE 3,806 4,228 4,394 4,121
NE 2,594 1,787 4,283 1,527
ENE 1,790 599 896 1,220
E 2,558 1,145 831 491
ESE 1,900 5,464 6,940 4,451
SE 1,818 3,523 5,711 5,581
SSE 1,597 4,457 6,592 8,947
S 2,324 1,518 5,828 5,879
SSW 4,317 3,170 2,477 4,714
SW 2,288 1,591 5,751 5,237
WSW 1,575 3,417 6,663 6,661
W 2,371 5,154 3,037 6,989
WNW 2,140 4,997 5,377 8,746
NW 1,553 3,515 5,692 8,266
NNW 1,013 3,446 5,945 7,894

Table 2-5 Pinellas Plant Populations from 5 to 50 Miles
5-10 10-20 20-30 30-40 40-50
N 56,502 107,728 111,008 46,903 5,324
NNE 25,146 34,356 15,540 24,095 69,288
NE 5 113,351 113,027 20,064 52,096
ENE 1,417 140,450 185,621 50,789 69,815
E 12,557 52,905 83,414 35,339 50,579
ESE 48,943 2,532 33,301 2,848 540
SE 70,541 19,709 13,577 3,355 2,307
SSE 65,311 33,030 99,304 137,954 125,700
S 30,788 7,472 12,630 2,703 0
SSW 6,530 0 0 0 0
SW 4,692 0 0 0 0
WSW 6,174 0 0 0 0
W 9,052 0 0 0 0
WNW 7,040 0 0 0 0
NW 8,451 0 0 0 0
NNW 41,734 2,959 0 0 0

Table 2-6 Pinellas Plant Cumulative Populations to 5 Miles
0-2 0-3 0-4 0-5
N 2,421 8,250 13,814 20,577
NNE 3,806 8,034 12,428 16,549
NE 2,594 4,381 8,664 10,191
ENE 1,790 2,389 3,285 4,505
E 2,558 3,703 4,534 5,025
ESE 1,900 7,364 14,304 18,755
SE 1,818 5,341 11,052 16,633
SSE 1,597 6,054 12,646 21,593
S 2,324 3,842 9,670 15,549
SSW 4,317 7,487 9,964 14,678
SW 2,288 3,879 9,630 14,867
WSW 1,575 4,992 11,655 18,316
W 2,371 7,525 10,562 17,551
WNW 2,140 7,137 12,514 21,260
NW 1,553 5,068 10,760 19,026
NNW 1,013 4,459 10,404 18,298

Table 2-7 Pinellas Plant Cumulative Populations From 5 to 50 Miles
0-10 0-20 0-30 0-40 0-50
N 77,079 184,807 295,815 342,718 348,042
NNE 41,695 76,051 91,591 115,686 184,974
NE 10,196 123,547 236,574 256,638 308,734
ENE 5,922 146,372 331,993 382,782 452,597
E 17,582 70,487 153,901 189,240 239,819
ESE 67,698 70,230 103,531 106,379 106,919
SE 87,174 106,883 120,460 123,815 126,122
SSE 86,904 119,934 219,238 357,192 482,892
S 46,337 53,809 66,439 69,142 69,142
SSW 21,208 21,208 21,208 21,208 21,208
SW 19,559 19,559 19,559 19,559 19,559
WSW 24,490 24,490 24,490 24,490 24,490
W 26,603 26,603 26,603 26,603 26,603
WNW 28,300 28,300 28,300 28,300 28,300
NW 27,477 27,477 27,477 27,477 27,477
NNW 60,032 62,991 62,991 62,991 62,991

Detailed Case StudV:

St. Lucie

This detailed example presents population estimates for the area

surrounding the St. Lucie Nuclear Power Plant. It conforms to Nuclear

Regulatory Commission (NRC) Regulatory Guide 1.70, Revision 3

requirements. It is provide to demonstrate the complexity demanded in a

typical population determination exercise, to illustrate the variety of

data that can be spatially analyzed and the display capabilities of the

system. The population estimates presented herein represent 1990

resident population levels; and estimated 1992 resident, transient and

special populations for the area within 10 miles of the St. Lucie plant.

Specification of Location

Florida Power & Light Company's (FPL) St. Lucie site is located on

Hutchinson Island, St. Lucie County, Florida. The coordinates for St.

Detailed Case Studv: St. Lucie~--

Figure 2-3 Pinellas Plant Cumulative Population Distribution

Lucie Unit 1 are latitude 27 20' 58" north and longitude 80 14' 48"

west. Approximately 300 feet to the south of St. Lucie Unit 1 is FPL's

St. Lucie Unit 2. St. Lucie Unit 2 is located at latitude 27 20' 55"

north and longitude 80 14' 47" west. The Universal Transverse Mercator

(UTM) coordinates for the midpoint (supplied by FPL) are 3,025,173

meters north and 574,326 meters east.

The eastern boundary of the site is the Atlantic Ocean and the

western boundary is the Indian River, a tidal lagoon. Other prominent

natural features within 50 miles of the site include Lake Okeechobee, 30

miles to the west-southwest of the site and a portion of the Everglades

approximately 24 miles to the south of the site. Figure 2-4 shows the

NNW (2.49%)
NW (1.09%)
WNW (1.12%)
w (1.05%)
WSW (0.97%)
SW (0.77%)
ssw (0.84%) N- (13.76%)
S (2.73%) "

-NNE (7.31%)
SSE (19.09%)

NE (12.20%)

SE (4.99%)

ESE (4.23%)-

--ENE (17.89%)

E (9.48%) -

Figure 2-4 Sector-Segment Grid for St. Lucie Positioned Over County


0 2 4 6 8

Figure 2-5 St. Lucie County with Sector-Segments

site in relation to the region within 50 miles. Figure 2-5 also shows

the site with the 50-mile grid positioned over county boundaries.

Figure 2-6 displays the level of census detail with a close-up view of

the first 10 radial miles of the grid over census block boundaries.

Prominent cities within 10 miles of the site include Fort Pierce,

approximately 7 miles to the northwest of the site on the mainland; Port

St. Lucie, 4.5 miles to the west-southwest; and Stuart, 8 miles to the

south. The largest urbanized area within 50 miles of the site is West

Palm Beach, located 36 miles to the south-southeast.

Transportation corridors within 5 miles of the site include U.S.

Highway 1 (US 1); State Roads (SR) A1A, 712 and 707: the Florida East

Coast Railroad; the Atlantic Ocean and the Intracoastal Waterway which

is located in the Indian River. SR A1A, the major north-south route on

Hutchinson Island, traverses FPL's property to the east of St. Lucie

Units 1 and 2.

Site Description

FPL owns approximately 1,132 acres of land on Hutchinson Island.

The site is generally flat and has dense vegetation characteristic of

Florida coastal mangrove swamps. At the ocean shore, the land rises

slightly to a dune or ridge approximately 19 feet above mean sea level.

The area pre-empted by the plant is about 300 acres, or approximately 27

percent of the total land owned by FPL. There are no industrial,

commercial, institutional, or residential structures within the plant

area. SR A1A traverses FPL's property approximately 1,000 feet east of

the St. Lucie Plant. The radius of the exclusion area is 0.97 miles

from the center of the St. Lucie Plant. The low population zone

includes that area within 1 mile of the center of the St. Lucie Plant.


Figure 2-6 Sector-Segment Grid Over Census Blocks Within the 10-Mile Radius

Figure 2-6 Sector-Segment Grid Over Census Blocks Within the 10-Mile Radius

Population Within 10 Miles

There are an estimated 132,958 people (1990 data) who reside

within 10 miles of the St. Lucie Nuclear Power Plant. The population is

concentrated in the cities of Fort Pierce and Port St. Lucie, both in

St. Lucie County. Most of the area within 10 miles of the plant is in

St. Lucie County. A small portion of the 10-mile area south of the

plant is in Martin County. Hutchinson Island extends along the entire

eastern boundary of the St. Lucie Plant 10-mile area, as the remaining

area east of the island is water. There are 2 portions of Hutchinson

Island, the north island and the south island. A small portion of the

north island is within 10 miles of the plant, in sector NNW 5-10. The

majority of the south island is within the plant's 10-mile radius in

both St. Lucie and Martin Counties. A number of large developments have

been built on the island, most of these being south of the plant.

Cities. Towns and Settlements

The largest population centers within 10 miles are the cities of

Fort Pierce and Port St. Lucie. The city of Fort Pierce lies northwest

and north-northwest of the St. Lucie Plant. Most of its area is located

between 5 and 10 miles of the plant to the northwest, except for the

community of Collins Park Estates which extends to approximately 3 miles

from the plant in the west-northwest and western sectors. Sector NNW

includes primarily that area of Fort Pierce which is on Hutchinson

Island. This portion of Hutchinson Island has been extensively

developed and includes a mixture of single-family homes, mobile homes,

multi-family condominium developments and a number of tourist

accommodations. The 1990 resident population in Fort Pierce was 36,830.

The city has experienced a population growth of 9% over its 1980

population of 33,802. This represents an average annual increase of

about 0.9%.

The city of Port St. Lucie lies west, southwest and south of the

plant. Its city boundaries range from about 3 miles to 13 miles from

the plant. Port St. Lucie encompasses a relatively large portion of the

area within 10 miles of the plant. In 1980, this city's population was

only 14,690. Since that time, the number of persons residing in Port

St. Lucie increased to 55,866 in 1990, an increase of 280.3%. Much of

this growth has occurred within 10 miles of the St. Lucie Plant.

Development of the area between 5 and 10 miles has been primarily in the

form of single-family homes. Most multi-family developments have

remained closer to the plant, off U.S. Highway 1.

The following is a breakdown of development according to 1990

census data in terms of housing units, housing unit density per square

mile and significant communities or residential features by sector-

segment. Tables 2-8, 2-9 and 2-10 give housing unit densities per

square mile of land. Housing unit density per area of land is used to

eliminate dilution effects of large percentages of water in some

sectors. Areas of less than 1 square mile are not considered, because

densities could be artificially inflated by using a fractional

denominator in the density quotient. First, consider the southern

sectors SE through SSW.

The SE sector approaching the 2-mile radius contains the first of

the condominium developments south of the plant on Hutchinson Island.

The rest of the island south of the plant is situated within the SSE

sector. The most intense development lies from 4 to 6 miles from the

plant where 4,563 housing units have been constructed, however numerous

multi-story condominiums and other enterprises span the length of the

Table 2-8 St. Lucie Housing Unit Densities to 5 Miles
0-1 1-2 2-3 3-4 4-5
N 0.00 **** **** 0.00 0.00
NNE 0.00 0.00 0.00 0.00 0.00
NE 0.00 0.00 0.00 0.00 0.00
ENE 0.00 0.00 0.00 0.00 0.00
E 0.00 0.00 0.00 0.00 0.00
ESE 0.00 0.00 0.00 0.00 0.00
SE 0.00 **** **** 0.00 0.00
SSE 0.00 *** **** **** ***
S 0.00 0.00 0.00 **** ***
SSW 0.00 0.00 **** 251.88 514.57
SW 0.00 **** **** 174.08 728.44
WSW 0.00 **** **** 132.25 1,337.24
W 0.00 **** **** 361.04 234.17
WNW 0.00 0.00 **** 444.57 688.64
NW 0.00 **** 0.00 *** ***
NNW 0.00 **** **** **** ***
by Annulus 0.00 0.00 0.00 272.76 700.61
**** Land area of less than 1 square mile would yield inflated densities.

southern island. These include Sand Dollar Villas, Dune Walk, Admiral,

Island Dunes I-II, Ocean Towers I-II, Islandia I-II, Empress, Princess,

Island Beach Resort, Oceana II North, Oceana I North, Miramar "Royal",

Miramar I-II, Sea Winds, Atlantis "B", Atlantis "A", Atlantis III, Ocean

Rise, Outdoor Resorts (Nettles Island), Island Club, Oceana South, Ocean

South II, Turtle Reef Club, Turtle Reef, Beach Club Colony, Ocean Dunes,

Holiday Out, Windmill Village and Villa Del Sol. Development continues

in a similar pattern in Martin County with The Dunes, Jensen Beach Club,

Indian River Point, Fairwinds Cove, Green Turtle Cove, Seaside, Joe's

Point, Sandpebble (Ocean), Sandpebble (River), Ocean View, Ocean View

(River), Beachwood Villas, Buttonwood, Shore Villas, Maritimes East,

Maritimes West, Hutchinson House East, Hutchinson House West, Angler's

Cove and Little Ocean Place. Many of the aforementioned condominium

Table 2-9 St. Lucie Housing Unit Densities to 10 Miles
5-6 6-7 7-8 8-9 9-10
N 0.00 0.00 0.00 0.00 0.00
NNE 0.00 0.00 0.00 0.00 0.00
NE 0.00 0.00 0.00 0.00 0.00
ENE 0.00 0.00 0.00 0.00 0.00
E 0.00 0.00 0.00 0.00 0.00
ESE 0.00 0.00 0.00 0.00 0.00
SE 0.00 0.00 0.00 0.00 0.00
SSE **** **** **** *** ****
S 293.44 378.39 780.24 1,158.02 ****
SSW 590.46 396.30 352.08 557.20 164.30
SW 727.41 813.95 700.98 310.83 363.63
WSW 1,036.28 1,167.46 815.26 246.65 206.08
W 221.97 166.90 159.21 7.61 0.55
WNW 525.80 201.31 104.86 98.23 65.03
NW 659.60 911.93 1,151.46 1,444.39 1,254.41
NNW *** 822.55 **** **** 300.82
by Annulus 579.28 607.35 580.58 546.13 336.40
**** Land area of less than 1 square mile would yield inflated densities.

communities are used on a seasonal basis. The island continues beyond

the 10-mile radius. This sector also contains the entirety of the

peninsular community of Sewall's Point, but only the northern, less-

developed portion lies within the 10-mile radius.

Development of interest in the S sector begins in the 4-5 segment

west of the Green River Parkway continuing into segment 5-6 north of the

county line in St. Lucie County and into Martin County along the St.

Lucie River shore involving the northern part of Pineapple Bluff. That

development is the most prominent in segment 6-7 around Savanna Road.

Proceeding into segment 7-8 Savanna Road and Jensen Beach Boulevard (SR

732) service Sugar Hill, Holly Creek, Timberwick, Pinechest Lakes and

Table 2-10 St. Lucie Housing Unit Densities to 50 Miles
10-20 20-30 30-40 40-50
N 0.00 0.00 0.00 0.00
NNE 0.00 0.00 0.00 0.00
NE 0.00 0.00 0.00 0.00
ENE 0.00 0.00 0.00 0.00
E 0.00 0.00 0.00 0.00
ESE 0.00 0.00 0.00 0.00
SE 0.00 0.00 0.00 0.00
SSE 548.48 394.14 888.79 1,653.38
S 361.21 20.21 49.67 422.43
SSW 58.54 19.17 0.55 2.82
SW 22.02 3.87 2.48 221.62
WSW 17.88 2.74 145.13 15.08
W 5.68 1.36 27.31 7.25
WNW 12.31 0.51 2.45 1.41
NW 117.63 53.21 8.09 0.83
NNW 516.40 368.45 138.04 118.28
by Annulus 184.46 95.96 140.28 271.45
**** Land area of less than 1 square mile would yield
inflated densities.

Ocean Breeze Park (mobile home community) in Jensen Beach. SR 732 also

provides the focal point for the most numerous (3,873 units) and dense

(1,158 units/mile2) housing unit construction in this sector within the

10-mile limit. This occurs in the 8-9 segment which includes Vista Del

Sol, Pine Lake Village MHP, Jensen Park Estates, Savannas Club, Tropical

Acres MHP and Beacon 21. The last segment (9-10) is comprised of 800

Place Garden Villas, Sunset Cove, Harpers Landing, as well as

development around SR 707 in the Rio area.

The SW sector incorporates part of the Harris Subdivision and

Riverview Heights plus residential areas south of Walton Road and west

of the Green River Parkway in the 3-4 segment. Growth in segment 4-5

has arisen along Melaleuca Boulevard. Community growth west of the

Florida Turnpike and east of Lennard Road supplies the greatest number

and highest density of housing units in the 10-mile radius of this

sector from 5 to 6 miles from the plant with 1,295 units and 590 units

per square mile.

Development continues in segment 7-8 along US 1 and Port St. Lucie

Boulevard in St. Lucie County, however sparse growth in the Martin

County constituent fuels a decrease in both total housing units and

their density. This dual decreasing trend continues to the 10-mile

limit. Housing is concentrated along Westmoreland Boulevard in segment

7-8 and is sporadic in Martin County. Segment 8-9 features development

along the northern shore of the North Fork of the St. Lucie River in

both St. Lucie and Martin Counties. The residential area of the

terminal segment lies south and west of Harbour Ridge Country Club

adjacent to Southbend Boulevard.

There is limited development in the SW sector until segment 3-4

where Savanna Club, an adult manufactured community, is located. In

general, development in this sector is concentrated around major

roadways. Specifically, US 1 and Walton Road in segment 4-5; US 1,

Lyngate Drive and Midport Road in segment 5-6; Midport Road, Port St.

Lucie Boulevard and north of the North Fork of the St. Lucie River in

segment 6-7; Whitmore Drive and Port St. Lucie Boulevard north of the

North Fork of the St. Lucie River and east of Westmoreland Boulevard in

the southern portion of segment 7-8; Port St. Lucie Boulevard, the

Florida Turnpike and the development of Vikings Lookout in segment 8-9;

and Port St. Lucie Boulevard, the Florida Turnpike, Savage Boulevard,

Tulip Boulevard and Darwin Boulevard in segment 9-10. The number of

housing units and their density (per square mile of land with water

subtracted) peak in segment 6-7 at 2,023 units and 814 units per square

mile of land.

Sector WSW contains parts of Spanish Lakes MHP and St. Lucie

Gardens in segment 3-4 and quickly reaches a maximum density of 1,337

units per square mile of land in the next 1-mile segment with the

inclusion of more of Spanish Lakes MHP and St. Lucie Gardens. Hidden

River Estates, River Park and River Park Unit 2 comprise segment 5-6.

There is widespread residential concentration around Floresta Drive and

Prima Vista Boulevard in segment 6-7 where the sector high of 2,951

housing units are situated. Prima Vista Boulevard and Whitmore Drive

are the hubs of development in segment 7-8. The last 2 segments contain

parts of St. Lucie West. Most of the development in segment 8-9 is east

of the Florida Turnpike and south of Juliet Avenue. Approximately equal

development exists in the last segment, primarily south of Juliet


Development in sector W is not noteworthy until segment 3-4 where

both the number of housing units and their density abruptly reach maxima

for the sector within 10 miles of the plant. These are 483 units and

361 units per square mile, respectively. Together with segment 4-5,

These segments make up the heart of the White City area and include

parts of Indian River Estates, Lexington Square, River's Edge and The

Woodlands. Segment 5-6 includes part of Driftwood Manor in the north

and Rivers Edge in the south. Continuing to the next segment (6-7),

some of Lucy Acres in the north and extensive development west of St.

James Drive as well as north and south of Bayshore Boulevard make the

main contributions, yet a decreasing trend in units and densities

continues throughout this and the remainder of the sector. The

residential pattern established in the previous segment prevails in

segment 7-8 predominantly east of the Florida Turnpike.

Development drops off precipitously in the remainder of the radius

of interest, with much of it located within the loop of the Torino


In the NNW sector, significant development is encountered in the

3-4 segment where portions of Indian River Estates in White City are

located. Gator Trace and St. James Park, both of Ft. Pierce, join

Indian River Estates to comprise the residential tracts of segment 4-5.

This is the most densely developed segment of the sector containing 689

housing units per square mile and 1,172 total units. More of St. James

Park along with parts of Driftwood Manor and Timber Ridge Estates make

up segment 5-6. Midway Road and Oleander Boulevard provide development

loci. Segment 6-7 includes pieces of Canoe Creek, River Hammock and

Lucy Acres with development focused on Sunrise Boulevard. Sweetwater

and Raintree Forest are the main communities in segment 7-8. In

addition, there is some development around Midway Road. About the same

level of development continues into segment 8-9 with Lost Tree Estates,

River Oaks Estates, Thousand Pines Estates and more all around Edwards

Road. The last mile increment of the 10-mile radius is less developed,

but features Westglen and Briargate.

The NW sector contains almost all of incorporated Ft. Pierce. As

before, segment 4-5 includes part of St. James Park. Development in

segment 5-6 is primarily east of US 1. Most of the development in

segment 6-7 is in Oleander Gardens east of US 1. In segment 7-8,

development is predominantly north of Edwards Road. It is extensive

west of US 1 to Oleander Boulevard (SR 605) and continues west of SR 605

to Sunrise Boulevard and beyond except for the area occupied by the

Lawnwood Recreation Complex. The next segment, 8-9, exhibits the peaks

of 4,749 housing units and 1,444 units per square mile for this sector.

This is due to concentrated development in Ft. Pierce around Okeechobee

Road (SR 70); the east-west thoroughfares of Orange Avenue, Delaware

Avenue and Virginia Avenue; and 25th Street (SR 615) running north-

south. Tucker Terrace and Lawnwood are the largest communities. The

final segment decreases in housing unit density with the majority of

units built along Orange Avenue and 25th Street.

The final sector under consideration is NNW. This sector

encompasses the majority of Hutchinson Island north of the plant.

Segment 6-7 has development east of AlA on the island including Surfside

Harbor. Development in segment 7-8 stretches across the island from

Jennings Cove on the lagoon side to Surfside and Tropical Beach on the

Atlantic side. Segment 8-9 contains the maximum of 1,637 housing units

which can be ascribed to Thumb Point, Bayshore Estates and Causeway

Mobile Home Park on the island with little development on the small

section of mainland in the segment. All of the South Bridge is within

the segment. Most of the North Bridge is in segment 9-10. Residential

development lies on the southern tip of the north island between Coral

Cove and Ft. Pierce Inlet Recreation Area. There is little development

on the mainland inclusion.

Population by Annular Sectors

The most heavily populated annular sectors are those which cover the

towns and developments mentioned above. The most heavily populated

sector in the 10-mile radius is annular sector NW 5-10 with a total of

35,265 residents, which includes much of the City of Fort Pierce, with a

1990 population of 36,830 residents.

Population by Annuli

The annulus between 5 and 10 miles of the St. Lucie Plant is more

densely populated than the area within 5 miles. A total of 116,230 live

between 5 and 10 miles of the plant (1990 data). Inside 5 miles, there

are about 16,728 residents. Within 2 miles of the St. Lucie Plant,

there is an estimated total of 154 residents. The entire area within 1

mile of the plant is owned by FPL and is included in the exclusion area

and low population zone. Much of the area in the one- to two-mile

annulus is water.

Population by Sectors

The most populous sector within 10 miles of the St. Lucie Plant is

the NW sector which, because of the large concentration of resident

population in the city of Fort Pierce, contains 35,636 persons (1990).

The second most heavily populated sector is WSW, which has 23,778

persons and includes part of Port St. Lucie. The adjacent sector, SW,

is the third highest with 21,019 residents in 1990.

Population Between 10 and 50 Miles

Tables 2-11, 2-12 and 2-13 show the distribution of the calculated

1990 population between 10 and 50 miles of the St. Lucie Plant; and

Figure 2-5 presents that information figuratively. The calculated 1990

population is 659,411 persons and represents 83.2 percent of the total

population within 50 miles of the plant. This population is confined to

sectors SSE through NNW since sectors N through SE, beyond the 10 mile

radius, include only the Atlantic Ocean. The major concentration of

population occurs in annular sector SSE 40-50, which includes portions

of the city of West Palm Beach, Palm Springs, Haverhill, Greenacres

City, Royal Palm Beach and Wellington. West Palm Beach is the northern

limit of the Florida Gold Coast development extending north from Miami

through Dade and Broward Counties into Palm Beach County. The 151,310

residents in annular sector S 40-50 live on approximately 174.5 square

Table 2-11 St.

Lucie Resident Populations to 5 Miles

0-1 1-2 2-3 3-4 4-5
N 0 0 0 0 0
NNE 0 0 0 0 0
NE 0 0 0 0 0
ENE 0 0 0 0 0
E 0 0 0 0 0
ESE 0 0 0 0 0
SE 0 118 0 0 0
SSE 0 0 0 197 976
S 0 0 0 196 443
SSW 0 0 38 710 2,100
SW 0 10 124 339 1,998
WSW 0 20 134 245 2,808
W 0 6 90 1,292 756
WNW 0 0 45 1,218 2,490
NW 0 0 0 23 348
NNW 0 0 0 0 0
by Annulus 0 154 431 4,220 11,919

miles of land (there are only 2.1 square miles of land in this sector-

segment). Annular sectors SSE 40-50 and SSE 30-40 have the second and

third highest populations respectively and reflect that Palm Beach

County is more highly developed than any other part of the region. Of

the total 863,518 residents of Palm Beach County in 1990, about 44

percent lived within 50 miles of the St. Lucie Plant.

Cities and Towns Within 50 Miles

Table 2-14 lists towns, cities and communities within 50 miles of

St. Lucie Plant with a 1990 population of more than 5,000, persons.

There are 12 towns with a population of more than 10,000, the largest of

which is West Palm Beach, with a 1990 population of 67,643. The second

largest is the city of Port St. Lucie, with 55,866 persons in 1990; the

Table 2-12 St. Lucie Resident Populations to 10 Miles

5-6 6-7 7-8 8-9 9-10
N 0 0 0 0 0
NNE 0 0 0 0 0
NE 0 0 0 0 0
ENE 0 0 0 0 0
E 0 0 0 0 0
ESE 0 0 0 0 0
SE 0 0 0 0 0
SSE 2,288 516 421 769 443
S 936 1,531 3,931 6,874 1,855
SSW 2,662 2,223 1,790 1,525 695
SW 3,154 5,104 4,554 2,402 3,334
WSW 3,966 6,819 6,054 1,796 1,936
W 1,247 1,120 1,161 37 5
WNW 2,310 1,378 801 819 612
NW 1,055 3,560 6,528 10,920 13,202
NNW 0 624 724 1,988 565
by Annulus 17,618 22,875 25,964 27,130 22,647

third largest is

is Riviera Beach

Fort Pierce with 36,830 persons; and the

in Palm Beach County with 27,639 persons.


urth largest

Of the 12

largest towns, 7 are in Palm Beach County. In addition to West Palm

Beach and Riviera Beach, the 7 include North Palm Beach (11,343

persons), Jupiter (24,986 persons), Greenacres City (18,683 persons),

Royal Palm Beach (14,589 persons) and Palm Beach Gardens (22,965

persons). Stuart, the largest city in Martin County, has a 1990

population of 11,936.

Of the 4 towns with populations between 5,000 and 10,000, all are

within Palm Beach County. These include the towns of Palm Beach, with

9,814 persons; Lake Park, with 9,763 persons; and Palm Springs, with

9,763 persons. Pahokee, with a 1990 population of 6,822, is also in

Palm Beach County, but is located in the northwestern quarter of the

county, on the shore of Lake Okeechobee.

Table 2-13 St. Lucie Resident Populations to 50 Miles

10-20 20-30 30-40 40-50
N 0 0 0 0
NNE 0 0 0 0
NE 0 0 0 0
ENE 0 0 0 0
E 0 0 0 0
ESE 0 0 0 0
SE 0 0 0 0
SSE 20,115 32,205 78,634 114,025
S 38,024 5,778 13,255 151,310
SSW 7,361 5,241 178 1,398
SW 4,434 1,141 246 8,810
WSW 3,012 740 15,196 1,824
W 698 381 10,176 2,478
WNW 2,078 143 811 494
NW 17,967 8,450 2,882 208
NNW 25,131 44,776 19,189 20,622
by Annulus 118,820 98,855 140,567 I 301,169

Annular Sectors

The most heavily populated annular sectors between 10 and 50 miles

from the St. Lucie Plant are those which encompass the cities and towns

with the greatest populations. The most populous annular sector, S 40-

50, includes Greenacres City (18,683 persons in 1990), Haverhill (1,058

persons in 1990), Palm Springs (9,763 persons in 1990) and Royal Palm

Beach (14,589 persons in 1990).

Po ulation b

J"" "

Population by

Figure 2-5 Resident Population from 10 to 50 Miles for the St. Lucie Plant
Figure 2-5 Resident Population from 10 to 50 Miles for the St. Lucie Plant

----I 10 ----q
0 5 10 15 20


The second most populous annular sector is SSE 40-50, which

includes portions of Palm Beach (9,814 persons in 1990), Riviera Beach

(27,639), West Palm Beach (67,643), as well as Cloud Lake (121), Glen

Ridge (207), Lake Clarke Shores (3,364) and Mangonia Park (1,453).

Table 2-14 Towns, Cities and Communities Within 50 Miles of the Plant
Communities of over 10,000 Persons

City or Town County 1970 1980 1990
Population Population Population
West Palm Beach Palm Beach 57,375 63,305 67,643
Port St. Lucie St. Lucie 330 14,690 55,866
Fort Pierce St. Lucie 29,721 33,802 36,830
Riviera Beach Palm Beach 21,401 26,489 27,639
Jupiter Paim Beach 3,316 9,868 24,986
Palm Beach Gardens alm Beach 6,102 14,407 22,965
Greenacres City Palm Beach 1,731 8,780 18,683
Vero Beach Indian River 11,908 16,176 17,350
Royal Palm Beach Paim Beach 3,423 14,589
Stuart martin 9,086 9,467 11,936
North Palm Beach aim Beach 9,035 11,344 11,343
Sebastian Indian River 2,831 10,205

Communities of between 5,000 and 10,000 Persons

City or Town County 1970 1980 1990
Population Population Population
Palm Beach Palm Beach 9,086 9,729 9,814
Palm Springs Palm Beach 4,340 8,166 9,763
Pahokee Palm Beach 5,663 6,346 6,822
Lake Park Palm Beach 6,993 6,909 6,704
The information in this table is based upon "Florida Population: Census
Summary 1990" April 1991. Bureau of Economic and Business Research,
University of Florida.

The third most populous annular sector between 10 and 50 miles of

the St. Lucie Plant lies north of West Palm Beach on the Atlantic Coast

(SSE 30-40). Although its land area is only 49.33 square miles of the

137.31 total square miles allotted to the sector-segment, i.e. 35.9

percent, it includes Lake Park (6,704 persons in 1990), North Palm Beach

(11,343), Juno Beach (2,121), as well as portions of Riviera Beach

(27,639), Palm Beach Gardens (22,965), Palm Beach Shores (1,040) and the

town of Jupiter (24,986). When the preceding three annular sectors are

combined, they comprise 52.2 percent of the total population between 10

and 50 miles of the St. Lucie Plant.

Population by Annuli

Populations of annuli between 10 and 50 miles of the St. Lucie

Plant range in number of residents from the largest, with a total of

301,169 persons (the 40-50 mile annulus), to the smallest, with 98,855

persons (the 20-30 mile annulus). The annulus between 30 and 40 miles

has the second largest population of 140,567, while the annulus between

10 and 20 miles contains 118,820 persons.

The 40-50 mile annulus has not only the largest population and the

greatest overall area (approximately 1,588 square miles of total area,

excluding the seven sectors over the Atlantic Ocean; or 1,091 square

miles of land area), but also the highest population density in the

region. The population density of the 40-50 mile annulus is 620 persons

per square mile of land (an average of the densities of the 9 sector-

segments). Eighty-eight percent of the population is located on 22

percent of the total annulus area, in sectors SSE and S, which include

West Palm Beach and environs.

Population by Sectors

The most populous sectors between 10 and 50 miles of the St. Lucie

Plant are those which cover the West Palm Beach area and the Atlantic

Coast. Sectors SSE and S have estimated 1990 populations of 244,979 and

208,367, respectively; and densities ranging from 792 to 3,188 persons

per square mile of land and from 61 to 867 persons per square mile of

land, respectively. Sector NNW has a population of 109,718 and

densities from 238 to 869 persons per square mile of land; sector NW,

the next one inland, has a total population of 29,507 and a density

range of 1.2 to 311 persons per square mile of land. The 5 remaining

sectors have densities which range from 1.3 to 623 persons per square

mile of land. The sparseness of population in the 5 interior sectors

can be attributed to extensive acreage covered by wetlands and surface

water (Lake Okeechobee), inaccessibility to population centers and the

extent of range and cropland.


This type of detailed information by annular rings, sector and

sector-segment is made possible by use of the GIS as previously

described. The sector-segment grid is constructed using the analysis

tools of the GIS so that each sector-segments a separate analytical

entity. This allows the user to query each entity for summary

statistics of the geographic attributes, e.g. land area, water area,

housing units and population. The grid is placed over the TIGER street

map features (highways, waterways, city roads, etc.) To provide

information about the specific communities. This is done by positioning

the cursor on the object of interest (line or point) to access its name.

Further enhancements include the use of more elaborate base maps to

derive that information and reference to paper maps using the GIS map as

a spatial reference.


Estimate and Projection Concepts

The concepts of population estimation and population projection

are similar in that they each involve generating a number intended to

indicate the size of the population of a given area at a specific point

in time. Both population estimates and population projections make use

of the basic demographic equation:

P = P0 + B D+ I 0

PI is the current population
P0 is the population at a previous point in time
B is the number of births since P0
D is the number of deaths since P0
I is the number of inmigrants since P0
O is the number of outmigrants since P0

This equation expresses the population at any given point in time

as a function of the population at a previous point in time and

considers the amount of natural change and net migration during the

interim. Natural change (usually an increase) is the number of births

minus the number of deaths in the study area, and net migration is the

number of inmigrants to the area minus the number of outmigrants away

from the area. The sources of vital registration and migration data

were discussed previously.

A population estimate refers to the size of the population of an

area in the recent past, while a population projection refers to the


size of the population of an area at some point in the future.

Population estimates are generally based on direct components of

population change, such as the actual number of births and deaths

occurring between the date of the previous population and the date of

the estimate; in the absence of direct data, a population estimate will

be based on symptomatic indicators of the components of population

change. Symptomatic data are those data that move in concert with the

data of direct interest. For example, a change in the school

enrollments or the number of motor vehicles being registered may serve

as symptomatic indicators of migration. Since population projections

refer to the size of the population at a point in the future, they

cannot be based on actual data comprising the components of population

change. Population projections must in one form or another be based on

the extension of either current or expected population trends into the


The only "true" population counts are available in decennial

census years, e.g. 1970, 1980, 1990, etc. Of course, the most recent

was the 1990 census which serves as the base for present and future

projections until the year 2000 census is taken and published.

Population projection methodologies at the county level have been

developed by various State agencies. In 1967, the Federal-State

Cooperative Program for Local Population Estimates (FSCPE) was formed.

Through member participation, the Bureau of the Census has an

opportunity to evaluate the level of accuracy of the population

estimates produced between censuses and to review the adequacy of

specific procedures used to prepare the estimates.

Methods of Estimation

The most widely accepted estimation methods are described in the

subsections that follow. Methodological strategies such as proration

techniques, simple ratio methods, the censal ratio method, and the vital

rates method are relatively simple to calculate and have minimal data

requirements. Alternative postcensal estimation strategies such as

composite methods, the U.S. Bureau of the Census component method II,

the ratio correlation method, the administrative records method, the

cohort component method, and the housing unit method involve somewhat

more complex calculations and have more extensive data requirements.

Vital Rates Method

The census year ratio of symptomatic data to population for a

given area is applied to a current value of that symptomatic data to

estimate the area's population. The symptomatic data are births and

deaths. First, births and deaths are converted to rates, the rates are

then adjusted to the estimate date by changes in these rates for some

broader area which can be measured over time. Next the rates are

divided into the number of births and deaths on the estimate date to

yield two population estimates. The two estimates are usually averaged

(USBC, 1990c).

Composite Method

Component, regression, and other methods are combined in an effort

to maximize the potential of each. The total for all ages results from

the sum of estimates for various age groups or age-sex groups.

Administrative Records Method

Uses births, deaths, internal migration, and immigration from

abroad. The internal migration component is derived from individual tax

return records supplied by the Internal Revenue Service. A migration

rate is derived by comparing the addresses on two years of tax returns

when the addresses are not the same. Data on Medicare enrollees

estimates the population 65 years old and over. The estimates of net

migration from the under 65 years of age tax base combined with

independent estimates of the population 65 years old and older, inmates

of institutions, college students in dormitories, military personnel

living in barracks, and other components of population change are added

to the previous year's estimate to yield an estimate of the total


Ratio-Correlation Method

Uses a multiple regression equation to relate changes in two or

more data series to estimated changes in population (the dependent

variable). All variables are expressed as the ratios of percentage

shares of the subarea to the universe in the estimate year to the

corresponding percentage shares for the base year (previous census

year). A separate regression equation is prepared and is unique for

each State. The underlying assumption of this method is that past

relationships between the independent and dependent variables will

persist into the current estimating period. Some of the independent

variables currently used are school enrollment, housing units, Federal

income tax returns and exemptions, births, and deaths. Other

possibilities include utility data, labor force data, drivers licenses,


The ratio correlation method is one of the more versatile

postcensal population estimation techniques. The ratio correlation

method produces postcensal population estimates through the use of a

statistical procedure called linear regression. Linear regression in

its simplest case is a statistical technique that allows researchers to

predict one variable based on observations of a second variable. The

variable to be predicted is referred to as the dependent variable (Y),

and the variable used to make the prediction is referred to as the

independent variable (X). Previous observations of values of X and Y

are used to determine the statistical relationship between the two

variables; given a new observation of X, one can use the statistical

model to predict the value of Y. Linear regression analysis results in

the development of an equation for a straight line that best fits the

set of previous observations. The basic equation for a linear regres-

sion model is as follows:

Y = b + mrX

Where: Y is the predicted value of the dependent variable
b is the Y intercept, the point where the prediction
line will cross the Y axis
m is the regression coefficient or slope of the
regression line
X is the known value of the independent variable

The first step in creating the linear regression model is to

calculate the value of m, the regression coefficient. The value of m

represents the ratio of the change in the value of Y, the dependent

variable, for each unit change in X, the independent variable. A

computational formula for b is as follows:

nm ( XY)- (EX) (E Y)
n- ( X2) ( X) 2

Where: n is the number of observations
(YXY) is the sum of each X value multiplied by the
corresponding Y value
(IX) is the sum of the X values
(MY) is the sum of the Y values
(CX2) is the sum of each X value squared
(CX)2 is the square of the sum of the X values

In the simplest case, a single independent variable is used to

predict the dependent variable; in more complex cases, a set of

independent variables may be used to predict the dependent variable

through a multiple regression model. The multiple regression equation

takes the form displayed below.

Y = b + mlX1 + m2X2 + + mnXn

In the ratio correlation method of population estimation, a set of

dependent variables thought to be symptomatic of population change is

used to predict the population of the estimate area. The technique is

referred to as the ratio correlation method because ratio

transformations of the independent variables, rather than the raw data,

are used in the analysis.

Component Method II

Uses birth and death statistics to measure net natural change, but

(unlike the Administrative Records method) it relies on school

enrollment figures to measure net migration. The population 65 and over

is based on the change in Medicare enrollees.

Housing Unit Method

Uses building permits, demolitions, and electric utility

connections to estimate changes in the number of housing units or

households and converts this to population change. The housing unit

method produces postcensal population estimates by equating population

to the number of occupied housing units times the average number of

persons per household. The housing unit method is one of the most

frequently applied methods of postcensal population estimation and is

the primary method of choice for subcounty population estimates. The

following formula represents the logic of the housing unit method for

estimating current population:

P = (Ho-O) + I

where P is the current population
Ho is the number of occupied housing units
O is the number of persons per household
I is the institutional population

Two major methodological strategies have evolved for calculating

the number of occupied housing units in an area. One strategy involves

an analysis of building permit and demolition data, while the other

strategy involves an analysis of utility data usually based on electric

utility meters. Both strategies begin with the number of housing units

known to exist at a previous point in time. The most recent census

provides an enumeration of occupied housing units and the average number

of persons per household which serves as the base data for the housing

unit method.

Cohort-Component Method

Requires separate assumptions for each of the components of

population change: births, deaths, internal migration, and international

migration. Each component is projected separately for each birth cohort

by sex and race.

Comparison Results

The Vital Rates Method and the Composite Method were replaced with

the Administrative Records Method after comparison of published

projections with decennial census results (USBC, 1986). For most

States, the Administrative Records Method has emerged as the single most

effective procedure for making population estimates, although accuracy

is strengthened when this method is averaged with independent estimates

based on the Ratio-Correlation Method. Other methods in widespread use

by State agencies include Component Method II and some State-specific

methods. Some States use unique methods averaged with one or more of

the above methods. One State-specific method gaining increasing favor

is the Housing Unit method. For a simple example, consider a

publication (USBC, 1988) in which the Ratio-Correlation Method was used

to determine the prediction equation for the State of Tennessee. The

regression equation for the State as a whole is given by

Y = -0.015 + 0.397 (Xl) + 0.127(X2) + 0.085(X3) + 0.412(X4)

where Xl = Federal income tax returns,
X2 = Medicare enrollees,
X3 = resident deaths, and
X4 = school enrollment in grades 1 through 8.

Note that there are key differences between State level and county

level projections.

Table 3-1 County population change

Tennessee 3.8 0.9

Bradley Co. 6.1 2.5

Hamilton Co. -1.2 -3.9

Meigs Co. 4.8 2.1

Polk Co. 0.4 -0.8

The projected total population of Hamilton County actually

decreased as driven by outmigration, while other counties and the State

as a whole experienced different rates of migration and growth. The

discussions that follow continue to contrast these levels and

demonstrate the superiority of county level predictions over larger as

well as smaller geographic and/or population coverages.

The 1990 publication "Projections of the Population of States, by

Age, Sex, and Race: 1989 to 2010" (USBC, 1990b), used a cohort-component

method that required separate assumptions for each of the components of

population change as previously described. In addition, the internal

migration component was estimated in four different ways, each with its

own alternative assumption. Although the cohort-component technique is

a good way to make projections at the state level, it is not necessarily

the best way to make projections at the county level. Many counties are

so small that the number of persons in each age-sex category is

inadequate for reliable cohort-component projections. Even more

important, county growth patterns are so volatile that a single

technique based on migration data from one time period may provide

misleading results. More useful projections can be made if several

different techniques and historical time periods are used.

It is often the objective to produce projections within a single

State rather than a consistent set of projections involving several or

all States. In such instances, examining population projections

prepared by State agencies can be useful. These State-produced

projections represent an alternative to the projections developed by the

Census Bureau. The individual State projections can be based on an

assortment of models that incorporate a wider range of variables and

data, since each State is not required to produce a set of projections

consistent with other States (USBC, 1990b). It is evident that when one

examines any particular State agency's projections that the numerical

results are meaningless without a clear statement of the methods and

assumptions that produced the numbers. FSCPE participants use those

methods that have statistically proved to generate the most accurate

projections for their respective States.

The most important and complex component of population change for

States, counties, and small geographical areas is internal migration.

With fertility and mortality rates in the U.S. becoming stable at

historically low levels, internal migration has assumed critical

importance in determining the growth or decline of counties and States

(USBC, 1990a). Trends in internal migration differ throughout the U.S.

There is also an interdependency at the State level, i.e., since net

internal migration is zero by definition at the national level, an

increase in net internal migration in one State has to be offset by

decreases elsewhere. Internal migration is sensitive to changes in

economic conditions, therefore both rapid and sizeable changes in

internal migration can be expected. This is why local bureaus that

monitor economic conditions are valuable to the purpose of population

projection over a small region, e.g., within the 50-mile radius of a

nuclear power or weapons production facility.

Population size is also a major determinant of accuracy levels.

Upon comparison of projections with decennial counts in both 1970 and

1980, smaller counties (less than 5,000 population) experienced larger

errors (7 percent) while larger counties (greater than 100,000

population) had errors of about half that amount (USBC, 1986). Clearly,

a sparsely populated county such as Polk County,TN is at risk for larger

error. However, the average error for a 10-year projection is still not

excessive. Since the counties surrounding Polk are more populous

(Bradley, McMinn, and Monroe), the inclusion of their more accurately

projected blocks in shared sector-segments will dilute the Polk error

contribution. In addition, even Polk County contains numerous census

blocks; an asset that enables the investigator to confidently place the

"true" 1990 figures in the correct sector-segments.

In a number of counties specific adjustments must be made to the

population before applying the techniques described above. This is done

to account for special populations such as university students, military

personnel, and prison inmates. Adjustments are made for counties in

which these special populations account for a large proportion of the

total population or where the special populations have moved counter to

trends for the rest of the population. The researcher draws on local

FSCPE member agencies with an intimate knowledge of these factors and

the methods best-suited to dealing with them. The application of the

predictive equations generated by these agencies is revealed in the next


Population projections for sector-segments are estimated by

applying common growth rates (predictive equations) to all census blocks

within a particular county; then the software assigns weighted fractions

of projected block populations to construct sector-segment totals.

Projections are tabulated at user-defined intervals, and can be

displayed in a variety of graphical forms (bar, rotated bar, 3-D bar,

etc.). The reliability of FSCPE member's projections coupled with the

sector-segment assignment system should result in the most accurate

delineations of projected population distributions available.

Elementary Postcensal Methods of Population Estimation

Proration Estimation

The proration method of population estimation usually is employed

as a method to allocate the population of a larger geographic area among

its smaller subareas. The method is applied by allocating a portion of

the current estimated population of a larger geographic area to a

smaller geographic area based on the ratio of the population of the

smaller area to the population of the larger area as measured in the

latest census. The formula for the proration method is as follows:

es p "el

P is the population
e is an estimated value
c is a census result
s is a geographic subarea
1 is the larger geographic area

The proration method is extremely easy to apply, has minimal data

requirements, and may be applied to any level of geography. Application

of the proration method is based on the assumption that the proportional

relationship of the smaller and larger areas has remained constant since

the time of the last census.

Simple Ratio Procedure

The simple ratio procedure is one of several strategies of

postcensal population estimation relying on symptomatic data indicators.

Symptomatic data are those data that serve as symptoms of population

change; ideally, these are data series that change in either a positive

or inverse manner in direct proportion to population changes. Logical

choices include variables that serve as direct components of population

change, such as births and deaths, along with other variables that are

more indirect correlates of population change, such as voter

registrations, school enrollments, utility hookups, and housing starts.

The logic of producing a postcensal population estimate through

the use of a simple ratio technique involves measuring the value of a

data series at two points in time along with the size of the population

at the earlier point in time. Most often the earlier point in time will

be the most recent census year. Once the data series has been measured

at the second point in time, a ratio is calculated indicating the change

in the data series using the earlier point in time as the base. The

change observed in the data series serves as an indicator of change in

population. Population change is then measured by multiplying the

population observed at the earlier point in time by the value of the

ratio for the data series. The result is a current population estimate.

One of the strengths of the simple ratio procedure is that multiple

indicators of population change (several data series) may be used.

Postcensal population estimates are obtained by averaging the various

ratios when multiple indicators are used in the simple ratio procedure.

The simple ratio procedure is a good choice when data for an area

are limited. It may be applied to a variety of levels of geography, and

it is a technique that bases a population estimate on changes in

variables that one logically can assume are related to population


Censal Ratio Method

Postcensal population estimates produced through the censal ratio

technique are based on changes in the ratio of the level of predictor

variables to the base population. In most applications some form of

adjustment is made based on national trends or some other source. When

no adjustments to the ratios are made, the censal ratio technique is

essentially the same as the simple ratio technique.

The chief advantage of the censal ratio technique over the simple

ratio technique is the ability to adjust the ratio indicating the

relationship between the predictor variable and the population size.

Any data series will undergo periodic fluctuations due to a variety of

factors, and the fluctuations are likely to be more extreme at the local

level. The ability to adjust the local ratio based on trends evident in

a larger geographic area prevents the data fluctuations in the smaller

area from being incorporated into the local population estimate.

Shyrock and Siegel (1973) suggest that it is wise to select

variables with certain characteristics. Variables for which data are

collected and reported at frequent intervals will be preferable to those

that are measured less frequently. Variables that have a sufficiently

large number of events occurring each year compared with the size of the

population are preferable to those with a smaller number of occurrences.

Predictor variables that are relatively stable over time or that change

in a predictable fashion should be chosen. Variables subject to extreme

fluctuation or variables that are especially sensitive to factors other

than population size should be avoided.

Population Prolections

Population projections are similar to intercensal estimates which

are discussed in detail in Chapter 4. Since the methodology is

essentially the same, the reader is referred to that chapter for

examples. Population projections typically are based on the assumed

continuation of the current situation, or perhaps on a high growth

scenario, or on a low-growth scenario. The general methodological

strategies for producing population projections fall into four basic

categories: (1) ratio allocation methods, (2) mathematical extrapolation

methods, (3) econometric methods, and (4) cohort component methods.

Ratio allocation and mathematical extrapolation methods are simpler

strategies for producing population projections. Ratio allocation

methods are used to allocate an existing population projection for a

larger area among the subareas that comprise it. For example, a county

population projection may be allocated among the various blocks that

comprise it, resulting not only in the original projection for the

county but in individual projections for each block. Mathematical

extrapolation methods involve the application of a selected growth rate

into the future. Econometric methods and cohort component methods are

somewhat more complex strategies for producing population projections.

These methods not only require additional technical expertise but

require a great deal more input data to complete the projection process.

The simpler methods of projection have a wider range of application and

may be used to produce population projections at almost any geographic

level. Econometric and cohort component strategies have more extensive

data demands and are less appropriate for smaller areas of geography.

For highly detailed population projections by age, race, and sex, the

clear choice is a cohort component method. Econometric methods result

in some detail by age, but not as much as cohort component methods.

Results of any methodological strategy are likely to be better the

shorter the length of the projection period. Population change is a

dynamic process with change sometimes occurring in ways that are

unanticipated. The longer the length of the projection period, the more

time will be available for the unanticipated to occur. Population

projections also tend to be more accurate for areas that have low levels

of vital rates such as fertility and mortality.

Results of population projections will tend to be more accurate

the larger the area of geography analyzed. Several factors operate to

improve the results for large areas. Population processes are volatile

at times, but subarea fluctuations tend to be averaged out as one moves

to a larger area of geography. For example, unusually high rates of

population growth in one area will often be matched by unusually low

rates of growth in another. In addition as the area of geography

increases, the problems due to migration decrease.

Ratio Allocation Methods

Ratio allocation methods of population projection provide a way to

allocate or distribute the projected population of a larger area among

its constituent subareas. Essentially, one is given a population

projection for a geographic area made up of identifiable subareas and

proceeds to allocate the population projected for a given date among the

subareas based on a set of operating assumptions. One general strategy

is to assume that the subarea's share of the total population at an

earlier date will remain constant over the length of the projection

period. The other general strategy is to assume that some pattern of

changing share is more appropriate, and the subarea will be allocated an

increasing or decreasing share of the total projected population. The

ratio allocation technique is especially well suited to small area

geographies that are not easily projected by other methods. One of the

major disadvantages is that the projections resulting from a ratio

allocation method are only as good as the population projection for the

larger area. In addition, the subarea's share of the total population

almost certainly will change over time, which among other things will

mean that the ratio allocation technique will not be as practical for

longer projection periods.

Mathematical Methods

The primary methods of mathematical population projection involve

applications of mathematical extrapolation, which is the extension of a

mathematically defined trend. There are two major types of methods of

mathematical population projection, and each type has several

variations. One major type involves mathematical extrapolation, or

simply extending a mathematical trend into the future. The mathematical

extrapolation methods are similar to the methods of intercensal

population estimation utilizing mathematical interpolation, and one may

use the same three types of assumptions. That is, given a growth rate

for a population, one may assume an arithmetic rate of growth or a

geometric rate of growth, with geometric rates of growth expressed as a

function of periodic compounding or continuous compounding. The second

major type involves fitting a growth rate to a curve. Three frequently

chosen alternatives are the Gompertz curve, the Pearl-Reed curve, and a

modified logistic curve. In each case a mathematical constant is used

to place an upper limit on the amount of growth that may take place

during the projection period. Each of the three curves results in a

growth pattern characteristic of a logistic curve which has the form of

an elongated S-shaped pattern suggestive of a period of slow gradual

growth followed by a period of rapid growth and then culminating in a

final period of no growth.

Case Study: St. Lucie Estimates and Projections

The following study is an example of an in-depth population

projection series accomplished using a GIS platform, and a new, modified

projection technique developed in this effort. The new technique is a

modified proration strategy as explained below. Instead of the typical

state to county direct procedure, this technique creates new

geographical zones (sector-segments), area-weights census blocks from

different counties to the new zones, then applies proration from county

level to block level through the DBMS. The census blocks that are

prorated from different counties are then synthesized into new composite

regions by the GIS. The result is a prorated allocated projection.

This case study demonstrates the complexity required in projections for

a nuclear power plant and the detailed data generated by the new

technique. The St. Lucie Plant projection is an expansion of the base

population presented in Chapter 2. The technique is enhanced using

multiple sources and types of information such as city and county

planning offices and housing permits.


It is recommended that the non-demographer obtain population data

professionally prepared by a State agency such as Florida's Bureau of

Economic and Business Research. These documents are inexpensive and

updated periodically. Use the county-level populations for projection

years to generate growth ratios. Create new fields in the population

database for each projection year and apply the growth ratio for each

county to its census blocks in the spreadsheet. Next, construct the

sector-segment grid as before and query the selected regions for the new

area-weighted projections. Each sector-segment projection will be a

composite of those projected blocks from the various counties that

comprise it. If a researcher has an abundance of demographic data

(especially migration data) and does not wish to use the simple

procedure designed for this study (above), then the techniques described

previously should be addressed. However, once the projections are made

for counties, the GIS procedure remains the same. The solution to the

problem of saturation lies in whether or not one accepts the following

assumptions. Assume that the highest population density for a census

block in the base census year (e.g., 1990) is a maximum that the local

population can not or does not wish to exceed. Next, assume that

adjacent or nearby high-density blocks are the more desirable to attract

excess populace. In the database write a script appropriate for the

software of choice that calculates the projected block, subtracts the

population that exceeds the density maximum, and forces that excess to

the second highest density block. Saturated blocks must be forced to


A second problem is that of null blocks. One can observe from the

case study projections that regions with zero as a baseline can never

grow using growth equations. In otherwise highly populated areas, such

blocks usually involve water, farmland, industrial areas, etc., but this

is not always the case. If the GIS user has good information about

planned developments, the null regions can be "seeded" with families to

generate future populations.

Projected Population

The population within 10 miles of the St. Lucie Plant is expected

to increase by about 19% over the 5-year period between 1990 and 1995.

The 1995 resident population is projected to be 157,625. The continued

development of Port St. Lucie and Hutchinson Island are expected to be

the largest contributors to this growth. The most significant

population increase in Port St. Lucie will be attributed to St. Lucie

West, a 4,600-acre tract of land to be developed by the J. White

Corporation. St. Lucie West is situated in sectors WSW and W. Most of

its area is between 5 and 10 miles of the plant, with some of it

extending beyond the 10-mile radius. The St. Lucie West planned

development also includes a spring training sports complex for the Mets

which was completed in 1988, new college campuses for Indian River

Community College and Barry University, public schools, an industrial

park, office park, and a regional mall.

Development along unincorporated Hutchinson Island north of the

plant has been slow and is projected to remain this way in the near

future. However, this trend will probably change as the southern part

of the island becomes more and more congested.

County planning officials have indicated that congestion of the

bridges from the mainland to Hutchinson Island could restrict

development. A bridge has been proposed which would cross the Indian

River at SR 712 and link US 1, the Florida Turnpike, and Interstate 95

to Hutchinson Island. An additional river crossing would induce

development on the island. However, it is uncertain when or where

another river crossing will be constructed, because the waters of the

Indian River in this area are part of an aquatic preserve.

Population Between 10 and 50 Miles

Projected population

Total population between 10 and 50 miles is expected to grow by

percent between 1990 and 2030, or from 659,411 to 1,266,338. The

average annual growth rate for this area would be 2.3 percent for the

year period. This rate of growth can be compared to the rate for the

State of Florida, which is expected to be 1.41 percent per year from

1990 to 2020. Florida is presently one of the most rapidly growing

states in the United States. Between 1980 and 1990, the state grew by

32.7 percent, a net addition of over three million people. Nearly 87

percent of this growth was attributed to net migration.

Areas of development

The principal area of development between 10 and 50 miles of St.

Lucie Plant occurs in Palm Beach County in the sectors including and

adjacent to the Atlantic Coast. Major development activity outside of

Palm Beach is concentrated in what can be called the "Atlantic

Corridor", the 5 to 10-mile area between the Atlantic Ocean and either

Interstate 95 or the Florida Turnpike in Martin, St. Lucie, Indian

River, and southern Brevard Counties.



Land to the west of this region is mostly used for pasture,

agricultural production (citrus, sugar cane, and truck farming), or

remains undeveloped. Access is limited and population is sparse. In a

few widely scattered sites, tracts of land have been platted and sold as

home sites or proposed for such development. No significant development

of any of these projects which lie west of the Atlantic corridor has yet

taken place.

Development is focused in the Atlantic Corridor for the following


1. proximity to existing population centers and services;

2. access to the Atlantic Ocean and Indian River, and the amenities
they provide: scenic beauty, sports and recreation, tourist and
industry potential;

3. presence of soils suitable for development of the coastal ridge;

4. zoning and planning policies developed by county and regional
agencies which permit development in these areas; and

5. availability of land suitable for development.

Only three significant clusters of development occur outside the

Atlantic Corridor between 10 and 50 miles of St. Lucie Plant. Two are

on or near the shores of Lake Okeechobee (which covers 400 square miles

in sectors SW and WSW between 30 and 50 miles of the plant). On the

southeastern shore of the lake in Palm Beach County, the community of

Pahokee serves the agricultural community of the western section as well

as the sport fishing community using the lake. A few miles north of the

lake in Okeechobee County, a regional center has developed at Okeechobee

City. The third location where significant development is occurring is

Indiantown, in south central Martin County, at the intersection of the

St. Lucie Canal and the Seaboard Coast Rail Line.

The following is a summary of development trends by county within

50 miles of St. Lucie Plant.

Palm Beach County. The principal area of growth within 50 miles

of St. Lucie Plant is in the northeastern quadrant of Palm Beach County,

which lies south of the plant, at a distance of more than 27 miles.

About 40 percent of Palm Beach County falls within 50 miles of St. Lucie

Plant; the total population of this area is expected to increase from

376,578 in 1990 to 733,136 in 2030. This increase represents a growth

of 94.7 percent over the entire period, or 2.37 percent averaged

annually. The corridor in Palm Beach County between the Atlantic Ocean

and the Florida Turnpike is intensively developed with contiguous towns

and cities such as Palm Beach, West Palm Beach, Riviera Beach, and Lake

Park. Residential development activity is expected to continue in this

area because of strong growth to date and its reputation as a desirable

place to live. Many developments include self-contained recreation

amenities. The Professional Golfer's Association (PGA) is headquartered

in Palm Beach County. Another area of growth exists in the northwestern

quadrant of Palm Beach County where Pahokee is located on the shore of

Lake Okeechobee. Pahokee is the 15h largest city within 50 miles of

St. Lucie Plant. It has an estimated population for 1990 of 6,822.

Martin County. Nearly all (approximately 84 percent) of Martin

County's 1990 population is located between 10 and 50 miles of the

plant. The remaining residents are within 10 miles of St. Lucie Plant.

The 1990 total of 84,560 persons between 10 and 50 miles for this county

is expected to grow by 119.7 percent to 101,218 by the year 2030. This

represents an average annual growth rate of 2.99 percent, the city of

Stuart is the major population center for the county; in 1990, its

population of 11,936 represented 11.8 percent of the total county

population of 100,900. Population is expected to grow in and around the

city of Stuart and on the barrier beaches in the Atlantic Corridor in

Martin County. Multi-family home construction decreased dramatically

between 1988 and 1989, when housing permits dropped by about a third.

The decline has continued through the first half of 1992. Single family

housing permits plummeted by nearly 40 percent across the Treasure Coast

the from 1989 to 1990. Martin County has seen a slight increase in

these permits in the first half of 1992 compared to 1991.

Indiantown is an incorporated area located approximately 26 miles

southwest (immediately east of the dividing line between SSW and SW) of

St. Lucie Plant at the intersection of SR 710 and SR 76. No population

number has been published for Indiantown based on the 1990 census. The

western part of Martin County is largely range and cropland, with few

permanent residents outside of Indiantown.

St. Lucie County. St. Lucie County extends from the plant site

west to the 30-mile radius. Of the county's total estimated population

of 150,171 in 1990, approximately 22.3 percent, or 33,553 persons, are

estimated to reside outside of the 10-mile radius. The number of people

outside the 10-mile radius is expected to grow by 134.2 percent (or 3.35

percent average annual rate) to a population of 78,576 in 2030. The

primary reason for this growth is the city of Port St. Lucie.

St. Lucie County's major population centers are the city of Fort

Pierce, with a 1990 population of 36,830, and Port St. Lucie with a 1990

population of 55,866. While the county as a whole grew 72.3 percent

between 1980 and 1990, the city of Fort Pierce grew 9.0 percent. Most

of Fort Pierce's growth occurred in the portion of the city on

Hutchinson Island. Inland, relatively little population growth is

taking place in Fort Pierce. As Fort Pierce is built up, development is

expected to occur within the Atlantic Corridor, outside the city limits.

Growth in Port St. Lucie has occurred at a much faster pace than Fort

Pierce. In 1980, Port St. Lucie's population was only 14,690. Since

1980, Port St. Lucie's population has increased by 280.3 percent.

Immediately to the west of the Atlantic Coastal Ridge is a

freshwater marsh system known as the Savannas. Once found along the

entire length of the Indian River Lagoon, this vanishing natural feature

has been depleted by mans development. Through the continued effort of

the State of Florida's Conservation and Recreational Land (CARL)

acquisition program, privately held properties within this area are

being acquired for perpetual public preservation. In addition to its

inland estuary and isolated wetland network, St. Lucie County has 18

miles of Atlantic Ocean shoreline, much of which is currently

undeveloped. Approximately 4.5 miles of this unincorporated oceanfront

are under public ownership. FPL owns another 2 miles of oceanfront, and

maintains the property in its natural state in conjunction with

operation of the plant. The balance of the oceanfront properties is

held in private ownership and is available for development activities,

which have historically been residential in nature. About 40 percent of

that property has already been developed.

The major use of land within the unincorporated areas of the

county is agriculture. Well over 60 percent of the county is presently

used for the production of citrus, cash crops, or ranching activities.

The western portion of St. Lucie County is especially dominated by

pasture and croplands. However, this situation should change

dramatically by 2030, as the city of Port St. Lucie expands to the

southwest. The locus of population in St. Lucie County will shift to

this area. The largest urban use of land within the unincorporated area

of the county is for detached, single family residential housing units.

Multi-family and mobile home development activities account for a little

less than a third as much acreage as the single family units. As was

observed in Martin County, multi-family home construction decreased from

1988 to 1989, but even more drastically at a reduction of about two-


thirds. The decline has continued into 1992. The Treasure Coast trend

of dropping single family housing permits is evident in St. Lucie

County, with permits leveling off over 1991 through 1992.

Indian River County. Essentially all of Indian River County falls

within the 10- to 50-mile radius. The county population of 90,208 in

1990 is expected to grow to 173,184 by 2030. This overall growth of 92

percent represents an annual average growth of 2.3 percent. The

principal community in the Atlantic Corridor is the county seat, Vero

Beach, with a 1990 population of 17,350 persons (19.2 percent of the

total county population). Other cities and towns include Sebastian,

10,205 persons in 1990; and Indian River Shores, 2,278 persons in 1990.

Only one settlement, the town of Fellsmere, with a 1990 population of

2,179, is located outside the Atlantic Corridor. Aside from the

community at Fellsmere (NW 30-40), the area west of Interstate 95 is for

the most part protected wetlands which are part of the St. Johns River

Flood Control District.

Brevard County. The portion of Brevard County which lies within

the 50-mile radius of St. Lucie Plant is included in sector-segments NNW

30-40, NNW 40-50, and NW 40-50. From the 1980 decennial census to 1990,

Brevard tied with Okeechobee County for sixth place in percent growth

(46.2 %) of the nine counties involved in the 50-mile radius. Only the

inland counties of Highlands and Glades experienced slower growth, thus

Brevard had the slowest growth rate of the coastal counties. Major

development in Brevard County over 1980 to 1990 has taken place at Cape

Canaveral, Malabar, Palm Shores, Palm Bay, and West Melbourne. It is

expected that the main growth will continue in the southern portion of

the county as vacant coastal areas become more attractive.

In southern Brevard County, development has occurred along the

Indian River and Atlantic Coast. Small communities include Barefoot

Bay, Micco, Grant, Valkaria, Melbourne Shores, and Floridana Beach. The

only incorporated town entirely within the 50-mile radius of St. Lucie

Plant is Malabar, which in 1990 had a population of 1,977. The town of

Palm Bay lies to the north of Malabar, just outside the 50-mile radius,

on the Indian River. However, part of Palm Bay's incorporated area

falls within the 50-mile radius in NNW 40-50 and NW 40-50. The

development of Palm Bay is proceeding at a rapid pace. The city grew

from 18,560 persons in 1980 to 62,632 in 1990, an increase of 237.5

percent. In southern Brevard, as in Indian River County, development

will be confined to the eastern coastal area because of restrictions

imposed in the western region by the St. Johns River Flood Control


Okeechobee County. Most of Okeechobee County's 1990 population of

29,627 persons resides within the 10- to 50-mile area. By the year

2030, this number is expected to increase by 96.5 percent to 58,219.

Okeechobee's population is concentrated in and around the county seat of

Okeechobee City. The county seat is at the convergence of US 98 and US

441 and SR 70, SR 78, and SR 710, less than 5 miles north of Lake

Okeechobee. This accessibility is expected to ensure it's continued

growth as a regional center. The city's 1990 population of 4,943

represents about 16.7 percent of the county total. The unincorporated

population of 24,684 persons accounts for the other 83.3 percent of the

county's total population. Along with Okeechobee City, the nearby

community of Cypress Quarters is split between sector-segments WSW 30-40

and W 30-40. In addition, WSW 30-40 contains Taylor Creek. W 30-40

encompasses Basswood Estates, Whispering Pines, and Country Hill


Glades, Osceola. and Highlands Counties. Three counties on the

periphery of the 50-mile study area contribute less than half of a

percent of the 1990 population between 10 and 50 miles of St. Lucie

Plant. In Glades County, in sector-segment WSW 40-50 on the northwest

shore of Lake Okeechobee, a community known as Buckhead Ridge has

developed. The only other settlement of greater size in Glades County

is the county seat of Moore Haven, which had a 1990 population of 1,432,

an increase of 14.6 percent from 1980. The unincorporated portion of

Glades County accounted for 6,159 persons in 1990, representing an

increase of 29.9 percent over 1980.

Osceola County is partially included in the 50-mile radius in

sectors NW and WNW. The only significant settlement is Yeehaw Junction,

found in sector-segment WNW 40-50.

Like Osceola, Highlands County has a small fraction of its land

area within the 50-mile radius, In this area, a small settlement has

developed on SR 70. Highlands County's predominant growth is expected

to continue outside of the 50-mile radius in the vicinity of Sebring,

Avon Park and Lake Placid, in the central part of the county. All three

interior counties reflect the lower rates of development taking place in

Florida's central regions, which are not adjacent to the Atlantic or

Gulf coasts.

Projected Growth Rates Between 10 and 50 Miles

The total population between 10 and 50 miles is expected to grow

by 92 percent from an estimated 659,411 persons in 1990 to 1,266,338 in

2030. Regarding the six main counties involved in the 50-mile radius,

the projections listed in Table 3-2 are expected. Sector-segments

likely to have the highest rates of growth are those that include

attractive areas near population centers that have reached, or are

approaching saturation. The listing above indicates a growth hierarchy

for the top three counties of St. Lucie > Martin > Palm Beach.

Table 3-2 Projections to 2030 by County
County 1990 Pop. 2030 Pop. Tot. Change % Inc.

Brevard 398,978 779,646 380,668 95.4

Indian River 90,208 173,184 82,976 92.0

Martin 100,900 221,676 120,776 119.7

Okeechobee 29,627 58,219 28,592 96.5

Palm Beach 863,518 1,750,700 887,182 102.7

St. Lucie 150,171 351,678 201,507 134.2

An analysis of 1990 resident populations per square mile of land

area available shows that sector-segment SSE 40-50 has the highest

density of 3,188 persons per square mile. This area is entirely within

Palm Beach County and includes the city of West Palm Beach. Expansion

is likely directly north into SSE 30-40; also dense, but to a lesser

degree at 1,594 persons per square mile. Growth could also continue to

the west in S 40-50 where the current density is 867 persons per square

mile. The aforementioned sector-segments of likely expansion are also

entirely within Palm Beach County. In contrast, a relatively low rate

of growth is expected for annular sector SSE 40-50 because of its high

density. Low growth rates can also be expected for sector-segments of

low density surrounded by others of low density (less than 500 persons

per square mile). Such sector-segments include SW 20-30, WSW 20-30, W

20-30, WNW 20-30, W 30-40, and WNW 30-40.

The focus of this chapter has been the estimation of populations

in the recent past but after the latest decennial census, and the

projection of populations into the distant future. Conversely, the next

chapter provides techniques to reconstruct populations bounded by past

decennial censuses, but at smaller geographic levels than originally




Data from past national decennial censuses along with county data

are plotted to yield growth curves (usually exponential) for regions of

interest. The equation describing the curve is solved for unknown

population values at some past time by assuming that subpopulations

within a region of known growth also grew at a proportional rate. It is

assumed that the uncertainty increases as the absolute value of the

difference in the present and the reconstructed time frame increases (as

with projections), however important components may be largely

subjective. Recent data with better measures of counting error are

quantifiable. Data available on CD-ROM at the block group level

facilitates a modified housing unit method where modern housing unit

densities can be correlated with their average ages to estimate

inhabitance during past years.

However, of particular interest are techniques based specifically

on population dynamics, e.g., births, deaths, migration, as opposed to

secondary indicators like housing units. Retrospective analyses have

advantages over prospective ones since limits are available for each

decennial census at the level of its smallest (or most convenient) areal

reporting. Limiting brackets consisting of successively enumerated

decennial censuses aid intercensal estimations. Methodological

strategies to estimate intercensal populations are numerous. These

strategies vary in complexity, input data requirements, and level of

detail obtained in the final computation. Some methods allow

considerable flexibility in the way they are used. Regardless of the

particular strategy, they all are based on the classic demographic

equation that relates the population at some point in time to the

population at a previous point in time by natural increase (births minus

deaths), and net migration. Annual intercensal estimations are often

made by some type of interpolative process. The beginning and ending

points are two consecutive decennial census enumerations, and an

assumption is made that population change is continuous and smooth.

There are three major techniques based on interpolation for estimation

of populations which differ in their treatment of the nature of

population growth. These are a simple arithmetic rate of change, a

compound rate of growth expressed as a geometric rate of change, or a

compound rate expressed as an exponential rate of change. The geometric

rate assumes a compound rate of growth over a specified time period,

e.g., annually, and the exponential rate assumes a continuous compound

rate of growth. In a time span of 10 years, growth rates determined by

geometric and exponential means are quite similar. Furthermore, results

obtained using arithmetic versus compound growth rates will not differ

greatly over such a short time span. There is no clear advantage

between the strategies, but a compound growth rate is usually chosen to

agree with the model for human reproduction. An advantage that all of

these interpolation techniques share is that they may be applied to any

level of geography from the entire world to the census block level. In

a detailed GIS procedure, the smallest common enumeration boundaries

between consecutive censuses, e.g., block groups or tracts, can be

analyzed and estimated independently and then aggregated into sector-

segments. In reality, the assumption of continuous and constant change

is erroneous in that a small population may sporadically increase,

decrease, or stagnate. At the levels of geography of concern in this

study, i.e., sub-state and county level, net migration is an important

factor in population change and is likely to be sporadic. Also, if a

researcher desires detailed information on race, age, or sex, the

interpolation approach is not a good choice. When such detail is

needed, the forward/reverse survival rate method is useful. For this

method, the data requirements are more rigorous and the calculations are

more difficult than the interpolation methods. Forward survival rates

are applied to the base population enumerated in the first census, and

the results are then averaged with those obtained by application of the

reverse survival rates to the subsequent census. Appropriate detailed

survival rates are less likely to be available than are detailed

population data, especially at smaller levels of geography. This is why

it is advisable to check the availability and detail of the survival

rates data before deciding the level of estimation detail to be

calculated. All things being equal, a population estimate for a given

area for a given point in time is preferable to a population projection

for the same area and time period.

The greater the level of detail in the population estimates, the

greater the level of error. Population estimates of the total

population are generally more accurate than estimates that include some

form of population detail such as age, race, or sex. Detailed

population estimates require a greater level of data and more complex

methodology than estimates for population totals only.

Estimates produced through the use of direct demographic data will

be more accurate than estimates produced through the use of indirect or

symptomatic data. Direct data are those involving any of the time

parameters of population change: fertility, mortality, and migration.

Potential indirect data sources would include a large number of

variables such as school enrollments, automobile registrations, voting

records, utility service hookups, and a variety of others. If given the

choice, no one would choose to employ indirect data over direct data,

but indirect data are often the only data available. The migration com-

ponent of population change is especially likely to be estimated only

through indirect data, particularly when population estimates are made

at subnational levels of geography.

Beyond the issue of direct versus indirect data, the higher the

quality of input data, the higher the quality of the resulting

estimates. Most estimation methodologies begin with a base population

at a previous point in time. A base population consisting of the most

recent census is of higher quality and will yield better current

estimates than a base population consisting of a previous estimate. A

similar argument can be made for other data sources. Data resulting

from official sources whose collection is mandatory will almost always

be of higher quality than data collected on an informal basis.

The longer the period of time between the enumeration of the base

population and the estimation date, the greater the degree of error in

the estimate. Estimates made for shorter periods of time--for example,

for 2 to 3 years past the base period--will generally be more accurate

than estimates made 8 or 9 years past the base year. Ironically, the

need for current estimates is more critical the farther one moves from

the base period.

In the most general case, the simpler mathematical methods of

population estimation will produce results with greater error than the

more complex methods of estimation. Simpler mathematical methods of

population estimation include interpolation, extrapolation, and

proration. Interpolation techniques are most often used for intercensal


estimates where the population is known at a beginning point and an end

point but not for the intervening years. Extrapolation methods are

essentially a simple type of population projection technique, and

proration methods involve allocating the known population of a larger

area among the several smaller subareas comprising the larger area.

More complex methods of population estimation will include cohort tech-

niques where each major age group of the population will be estimated

separately, ratio correlation techniques which involve the development

of a statistical regression equation, and several others.

The mathematical methods will usually result in estimates of

poorer quality than the more complex methods, with mathematical

extrapolation techniques resulting in the least accurate estimates. It

should be noted, however, that the use of a more complex method does not

necessarily guarantee a quality result. The more complex the method,

the more demanding the data requirements, and in some cases the required

data will simply not be available or will be of poor quality. No method

of population estimation based on poor quality data will produce an

accurate result, and complex methods of estimation are no exception. In

fact, in the absence of quality data, a simpler method of population

estimation may be the best choice.

No single method of population estimation will always be the best

choice. Over time, one will achieve better results from a program of

population estimation by employing a variety of estimation techniques.

Using multiple techniques will provide a means of checking the validity

of the estimate since similar results obtained from a variety of

different methods tend to suggest the overall accuracy of the result.

In many cases, one will be well served by averaging the results of

several estimates to obtain the final population estimate.

Methods of Intercensal Population Estimation

The methods of population estimation are employed in two broad

applications: intercensal estimation and postcensal estimation.

Intercensal estimates result when one has population totals for a

particular area from two successive censuses for the census years but

does not have population totals for the intervening years. While one

may deal with an intercensal period or a postcensal period, in each case

one is dealing with estimates in contrast to projections. Recall that

the essential difference between a population estimate and a population

projection is that population estimates are made for an area for a

specific time in the past. Most often the estimate is for a period in

the recent past, such as the previous year; population projections are

produced for a period of time in the future. In most cases the

methodologies from which one will choose when producing population

estimates or projections will differ due to the applicable time frame.

In addition, methods of population estimation are more likely to be

based on the use of symptomatic data which are likely to move in

conjunction or sympathy with population totals. Methods of population

projection are more likely to be based on some type of extension or

variation of current population trends.

A number of methodological strategies to estimate intercensal

population have been developed. The methods vary in their level of

complexity, type of data required for their computation, level of detail

in the final product, and intended purpose. In addition, many of the

methods of producing intercensal population estimates are better

described as methodological strategies; that is, there is considerable

flexibility in the way certain estimation methods are employed. One

common characteristic of population estimation methodologies is that

they all in one way or another are based on the classic demographic

equation. Not all of the estimation methodologies, however, will deal

with the classic demographic equation in as sophisticated a fashion as


Not all of the existing methodologies are described in the present

discussion. Rather, emphasis is placed on those methods that are used

more frequently by those active in population estimation. Included in

this discussion of estimation methodologies are intercensal

interpolation, assuming arithmetic and compound rates of change, and

detailed intercensal estimates through the forward/reverse survival rate

method. The discussion of intercensal estimation methodologies begins

with an analysis of the interpolation method of producing intercensal


Interpolation in Intercensal Estimates

Intercensal estimation most often involves the production of

single year population estimates for all years between two census years.

For example, we may have the 1980 and 1990 census population for an area

but would like to have a year-by-year estimate for years 1981 through

1989. Often some type of interpolation estimation process will be

chosen to derive the intercensal estimates. The two observed population

totals reported in the two most recent census results are treated as the

beginning and ending points for the population, and the interpolation

procedure then provides a series of annual estimates. One should note

that implicit in an interpolation strategy is the assumption that the

population has increased or decreased in a smooth and continuous fashion

between the two census years. In many cases the assumption of a

continuous population change will be false. Population change often

occurs in an irregular fashion with periods of rapid change followed by

Full Text
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID E3PCF2WCV_2WF5K2 INGEST_TIME 2013-10-24T18:04:50Z PACKAGE AA00017649_00001