You are on page 1of 117

A Thesis

entitled

Evaluation of Five GIS Based Interpolation Techniques for


Estimating the Radon Concentration for Unmeasured Zip
Codes in the State of Ohio

by

Suman Maroju

Submitted as partial fulfillment of the requirements for

the Masters of Science Degree in Civil Engineering

Advisor: Dr. Ashok Kumar

Graduate School

The University of Toledo

August 2007
An Abstract of

Evaluation of Five GIS Based Interpolation Techniques for


Estimating the Radon Concentration for Unmeasured Zip
Codes in the State of Ohio

By

Suman Maroju

Submitted as partial fulfillment of the requirement for

the degree Master of Science in

Civil Engineering

The University of Toledo

August 2007

The main objective of this thesis is to predict the geometric mean of radon

concentrations for unmeasured zip codes in the State of Ohio using the best

geostatistical analyst interpolation technique. The radon concentration data used in

this study spans from 1989-2005 (with some data sets obtained prior to 1989) and has

been collected by the University of Toledo researchers from various commercial

testing services, university researchers, and county health departments. This objective

was achieved by first dividing the data into a 80% training data set and a 20% test

iii
data set and then applying five interpolation techniques on the training data set to

evaluate the predictions for test data points. Several statistical indicators (root mean

square error (RMSE), normalized mean square error (NMSE), fractional bias (FB),

mean error (ME), mean absolute error (MAE), mean square error (MSE) and

coefficient of correlation (r)) were computed on observed and predicted radon

concentrations for evaluating the ability of predicting missing radon values using all

interpolation techniques. The confidence limits on these indicators were obtained

using the bootstrap resampling technique. The analysis of statistical indicators and

confidence limits on the indicators clearly showed that none of the techniques is

superior over the other. The spatial pattern of radon concentration generated using the

measured values for all techniques were then compared with the uranium maps for

the area. This exercise led to the exclusion of global and local polynomial

interpolation technique.

Assessment of the interpolation techniques using the soil uranium

concentrations map showed that ordinary kriging interpolation technique can be used

for predicting the radon concentrations for unmeasured zip codes. Radon

concentration values for unmeasured zip codes were obtained and upon their

examination, it was found that the number of zip codes having geometric mean of

radon concentrations exceeding 4 pCi/l, 8 pCi/l and 20pCi/l were 688, 85 and 9

respectively.

iv
ACKNOWLEDGEMENT

I wish to express my deepest gratitude to my advisor Dr. Ashok Kumar for his

continued support, guidance and encouragement, without which it would not have

been possible to complete this project. I would like to thank Dr. Brain Randolph and

Dr. Andrew Heydinger for their time, suggestions and comments during the course of

my graduate study at the University of Toledo.

Special thanks to all my friends for their enthusiastic and generous support

through my research study.

Above all, I wish to express my deep appreciation to my parents and sister

who with their love and encouragement have made all his possible.

v
TABLE OF CONTENTS

ABSTRACT..iii

ACKNOWLEDGEMENT.............................................................v

TABLE OF CONTENTS..............................................................vi

LIST OF FIGURES ......................................................................ix

LIST OF TABLES ...................................................................... xii

1. Introduction ................................................................................1

1.1 Overview....................................................................................................... 1

1.2 Problem Statement ........................................................................................ 3

2. Literature Review.......................................................................4

3. Data Collection ...........................................................................8

4. Geostatistical analyst ...............................................................10

4.1 Exploratory Spatial Data Analysis (ESDA)................................................ 11

4.1.1 Histogram............................................................................................ 11

4.1.2 Normal QQ Plot .................................................................................. 13

4.1.3 Trend Analysis .................................................................................... 15

4.1.4 Semivariogram/Covariance Cloud...................................................... 17

4.2 Deterministic Interpolation Technique ....................................................... 19

4.2.1 Inverse Distance Weighted ................................................................. 20

vi
4.2.2 Radial Basis Functions........................................................................ 22

4.2.3 Global Polynomial Interpolation......................................................... 24

4.2.4 Local Polynomial Interpolation .......................................................... 26

4.3 Geostatistical Interpolation Techniques...................................................... 26

4.3.1 Kriging ................................................................................................ 27

5. Approach and Procedure ........................................................30

5.1 Application of Interpolation Techniques .................................................... 32

5.1.1 Ordinary kriging.................................................................................. 33

5.1.2 Inverse Distance Weighting................................................................ 43

5.1.3 Radial Basis Function ......................................................................... 46

5.1.4 Global Polynomial Interpolation Technique....................................... 48

5.1.5 Local Polynomial Interpolation .......................................................... 52

5.2 Qualitative Analysis.................................................................................... 56

5.2.1 Comparison of the behavior of the prediction maps with the soil

uranium concentrations map ............................................................................... 56

5.3 Evaluation Criteria ............................................................................................ 58

5.3.1 Normalized Mean Square Error (NMSE) .................................................. 58

5.3.2 The Coefficient of Correlation (r).............................................................. 59

5.3.3 Fractional Bias (FB)................................................................................... 59

5.3.4 Root Mean Square Error (RMSE).............................................................. 59

5.3.5 Mean Error (ME) ....................................................................................... 59

5.3.6 Mean Absolute Error (MAE) ..................................................................... 60

5.3.7 Mean Square Error..................................................................................... 60

vii
5.3.8 Confidence Limits...................................................................................... 60

6. Results........................................................................................62

7. Conclusions .............................................................................100

8. References ...............................................................................101

viii
LIST OF FIGURES

Figure 1 Plot Showing the Frequency Distribution for Radon Data Set..................... 12

Figure 2 Plot Showing the Frequency Distribution of Radon Data Set after Log

Transformation............................................................................................................ 13

Figure 3 Normal QQ Plot for Radon Data Set............................................................ 14

Figure 4 Normal QQ Plot for Radon Data Set after Log Transformation .................. 15

Figure 5 Trend Analysis Dialog Box for Radon Data set........................................... 16

Figure 6 Semivariogram/Covariance Cloud Dialog Box............................................ 18

Figure 7 Radial Basis Function Surface Fitting Through a Series of Elevation Values

..................................................................................................................................... 23

Figure 8 Global Polynomial Interpolation Fitting Through the Sample Points of

Elevation on Gently Sloping Hill................................................................................ 24

Figure 9 Global Polynomial Interpolation Fitted to a Valley Using Second Order

Polynomial .................................................................................................................. 25

Figure 10 Geostatistical Method Selection Dialog Box ............................................. 33

Figure 11 Semivariogram/Covariance Modeling........................................................ 34

Figure 12 Ordinary Kriging Searching Neighborhood Dialog Box............................ 36

Figure 13 Ordinary Kriging Predicted Plot (Predicted values Vs Measured Values) 37

Figure 14 Ordinary Kriging Error Plot (Errors Vs Measured Values) ....................... 39

ix
Figure 15 Ordinary Kriging Standardized Error Plot (Standardized Error Vs Measured

Values) ........................................................................................................................ 40

Figure 16 Ordinary Kriging QQ Plot .......................................................................... 41

Figure 17 Ordinary Kriging Radon Concentration Prediction Map for the Training

Data Set....................................................................................................................... 42

Figure 18 IDW Searching Neighborhood Dialog Box ............................................... 43

Figure 19 IDW Cross-validation Tool (Measured Vs Predicted Values) ................... 44

Figure 20 IDW Radon Concentration Prediction Map for the Training Data set....... 45

Figure 21 RBF Searching Neighborhood Dialog Box................................................ 46

Figure 22 RBF Cross Validation Tool (Measured Vs Predicted Values)................... 47

Figure 23 RBF Radon Concentration Prediction Map for the Training Data Set....... 48

Figure 24 Setting the Power Parameter in Global Interpolation Technique............... 49

Figure 25 Global Polynomial Interpolation Cross validation Tool (Measured Vs

Predicted Values) ........................................................................................................ 50

Figure 26 Global Polynomial Interpolation Radon Concentration Prediction Map for

Training Data Set ........................................................................................................ 51

Figure 27 Local Polynomial Interpolation Searching Neighborhood Dialog Box ..... 53

Figure 28 Local Polynomial Interpolation Cross Validation Tool (Measured Vs

Predicted Values) ........................................................................................................ 54

Figure 29 Local Polynomial Interpolation Radon Concentration Prediction Map for

the Training Data Set .................................................................................................. 55

Figure 30 Aerial Radiometric Map of Ohio Showing the Concentration of Uranium in

Surficial Sediments and Soils ..................................................................................... 57

x
Figure 31 Ordinary Kriging (Measured Vs Predicted Values) ................................... 62

Figure 32 IDW (Measured Vs Predicted Values) ....................................................... 63

Figure 33 RBF (Measured Vs Predicted Values).64

Figure 34 Global Polynomial Interpolation (Measured Vs Predicted Values).64

Figure 35 Local Polynomial Interpolation (Measured Vs Predicted Values)............. 64

Figure 36 Ordinary Kriging Radon Concentration Prediction Map Using the Whole

Radon Data Set ........................................................................................................... 72

xi
LIST OF TABLES

Table 1 Sensitivity Analysis for Division of Data Set................................................ 31

Table 2 ME, MAE, MSE and RMSE Values for Different Interpolation Techniques

for Geometric Mean of Radon Concentrations Test Predictions ................................ 64

Table 3 NMSE, FB and Coefficient of Correlation (r) Values From Bootstrap Method

..................................................................................................................................... 65

Table 4 Robust and Seductive 95% Confidence Limits for Each Interpolation

Technique.................................................................................................................... 67

Table 5 Robust and Seductive 95% Confidence Limits Among the Techniques ....... 68

Table 6 Summary of Robust and Seductive 95% Confidence Limits Analyses on Each

Technique.................................................................................................................... 69

Table 7 Summary of Robust and Seductive 95% Confidence Limits Analyses Among

Each Technique........................................................................................................... 70

Table 8 Predictions of Geometric Mean of Radon Concentrations Zip Codes in Ohio

using Ordinary Kriging ............................................................................................... 73

xii
Table 9 Zip Codes with Geometric Mean of Radon Concentration greater than 4 pCi/l

(without Predicted Zip Codes) .................................................................................... 84

Table 10 Zip Codes with Geometric Mean of Radon Concentration greater than 4

pCi/l (with Predicted Zip Codes) ................................................................................ 89

Table 11 Zip Codes having Geometric Mean of Radon Concentrations greater than

8pCi/l........................................................................................................................... 98

Table 12 Zip Codes having Geometric Mean of Radon Concentrations greater than

20pCi/l......................................................................................................................... 99

xiii
1. Introduction

1.1 Overview

A geographical information system (GIS) provides a powerful collection of tools

for the management, visualization and analysis of spatial data. In this thesis; the ArcGIS

geostatistical analyst tool has been employed, which provides a powerful suite of tools

for spatial data exploration and surface generation using sophisticated statistical methods

and which allows creating a surface from data measurements. In addition, the

geostatistical analyst tool gives the user the power to fully understand the qualitative and

quantitative aspects of the data. By providing the freedom to predict and model spatial

phenomena based on statistics and incorporating powerful exploration tools, ArcGIS

geostatistical analyst effectively bridges the gap between geostatistics and GIS analysis.

Many environmental applications, like analysis of severity of ozone pollution and

identifying soil contamination, have been done using the geostatistical analyst

(Krivoruchko). The radon concentrations data set used for this thesis has been created

from the results submitted to the Ohio Department of Health (ODH) obtained from

various county health departments, commercial testing services and university

researchers performing radon tests for different houses. The data has been collected for

1
each zip code mostly for the years 1989 to 2005 and some, collected prior to 1989. As the

data is heavily skewed, the geometric mean of radon concentrations is used to input as a

point data source.

Radon is a naturally occurring radioactive gas produced by the breakdown of

uranium in soil, rock and water. The concentration of radon is most commonly expressed

in terms of the number of alpha particles it generates. The units of concentration are

pico-curies per liter of air (pCi/l). Radon is the second most common cause of lung

cancer after cigarette smoking, accounting for 15,000 to 22,000 cancer deaths per year in

the US alone according to the National Cancer Institute (USA). Radon gas is believed to

cause about 14% of lung cancer deaths (1000+ deaths) in Ohio annually

There are a variety of devices (alpha track, charcoal canisters, scintillation counters,

ionization chambers, positive barrier, two filter method, continuous radon monitor, sun

nuclear, e-perm) available for measuring the radon concentrations. Most of the radon

concentration tests are performed using either charcoal canisters or alpha-track detectors.

Using the geostatistical analyst, all five interpolation techniques (ordinary kriging,

inverse distance weighting (IDW), radial basis function (RBF), global polynomial

interpolation, and local polynomial interpolation techniques) are compared and the ideal

technique is used to predict the radon concentrations for unmeasured zip codes in the

State of Ohio.

2
1.2 Problem Statement

Geostatistical concepts were originally developed by Matheron (1965) and Krige

(1951) for the purpose of ore reserve estimation but are increasingly being implemented

in a variety of environmental fields. The purpose of this thesis is to estimate the radon

concentrations in zip codes where no data has been collected.

The first objective is to evaluate the best interpolation technique depending on

how the radon data is distributed. This is done by splitting the radon data points into a

80% training data set and a 20 % test data set, and the interpolation techniques are

performed using the training data set to evaluate the predictions for test data points. The

seven statistical parameters RMSE, NMSE, FB, ME, MAE, MSE and coefficient of

correlation (r) are used to evaluate the interpolation techniques.

The second objective is to perform the interpolation with the whole radon data set

using the best interpolation technique, and the geometric means of radon concentration

predictions are evaluated for zip codes where no data is collected.

The third objective is to present the impact of the results obtained from this

study. The earlier studies reported that the number of zip codes having geometric mean

of radon concentrations in three cases (over 4 pCi/l, 8 pCi/l and 20 pCi/l) is found to be

390, 85 and 9 respectively for each case. This study will attempt to calculate the number

of zip codes for the three cases mentioned above, taking into account the projected values

for the unmeasured zip codes.

3
2. Literature Review

The geostatistical analyst interpolation techniques are used to create a continuous

surface either from the measured sample points stored in a point feature layer or by using

the polygon centroids; and then predict the values at unmeasured locations from the

surface created. Many studies have been done using the geostatistical analyst and some of

the fields that benefit by virtue of these interpolation techniques include agricultural

production, temperature data, soil contamination, mining, health care and meteorology.

Some of these studies that used the geostatistical analyst interpolation techniques are

summarized below.

In the study by Jie et al., methyl-tert-butyl-ether (MTBE) concentrations were

assessed in the groundwater of the city of Temecula, California using the geostatistical

analyst. MTBE is an oxygenated fuel additive to reformulated gasoline and is considered

to be the leading groundwater contaminant that causes threat to public health and the

living environment. With the MTBE concentration data collected from the California

Regional Water Quality Control Board, a spatial surface for the study area was generated

and the MTBE concentration values were predicted using the simple kriging interpolation

4
technique and the probability maps of MTBE concentrations exceeding the critical

threshold (20 ppb) were also created.

According to a study by Kevin et al. (2001), ozone concentration was measured at

the monitoring stations throughout the state of California. The concentration levels of

ozone were known for all the stations, but to know the same for every location in

California is not quite possible as monitoring stations cannot be present everywhere for

reasons of cost and practicality. The geostatistical analyst interpolation techniques were

used to measure the level of ozone concentration for the location of interest. In this study,

ordinary kriging interpolation technique was used to create a surface from the measured

ozone concentrations and ozone concentrations at locations without any monitoring

stations were calculated.

In a study by Zheng et al. (2003), geographic variation of arsenic concentrations in

the county of Cochise, Arizona and its relationship with cancer risk was examined.

Exposure to arsenic may be a prime cause of bladder, lung, kidney and skin cancer.

People living in this county use ground water for intake, which led to the study of arsenic

concentrations in this county. Arsenic concentrations were interpolated using the ordinary

kriging, inverse distance weighting, local and global polynomial interpolation techniques.

Mathematically, of all the four ordinary kriging performed the best with lowest the

RMSE.

5
As per the study by Saby et al. (2005), an estimation of the anthropogenic stock of

Pb in the topsoil of an area around Paris was performed using the soil samplings of a grid

soil monitoring network. Topsoil enhancement technique was used to separate Pb content

due to diffused pollution from geochemical background. The lognormal ordinary kriging

was then performed to produce a smoothed kriged map of anthropogenic stock of Pb. The

total anthropogenic stock of Pb in topsoil was estimated to be 143,000 metric tons. This

amount is close to an independent estimation of total anthropogenic Pb deposition during

the last 50 years.

In the study by Liu et al. (2005), the soil heavy metal concentrations (Cu, Zn, Pb,

Cr and Cd) in paddy fields were estimated at unsampled sites from the four hundred and

fifty soil samples collected in topsoil in Hangzhoue Jiaxinge Huzhou (HJH) plain

(China). The risk assessment is also quantified, since the soil heavy metal concentration

is directly related to the crop quality and ultimately, the health of people. Ordinary

kriging and lognormal kriging were carried out to produce the spatial patterns of heavy

metals and disjunctive kriging was used to quantify the probability of heavy metal

concentrations higher than their guide values. Cokriging method was also used to

minimize the sampling density for Cu, Zn and Cr. The results showed that among the five

metals; Cu, Zn and Cd had high potential to cause environmental and human health

hazards and also, overall estimation quality of Cu, Zn and Cr using the cokriging

technique was better compared to the kriging technique.

In the study by Pardo-Iguzquiza et al. (1997), an estimation of the areal average

climatological rainfall mean in Guadalhorce river basin in southern Spain was performed

6
using the Thiessen, ordinary kriging, cokriging and kriging techniques with an external

drift; using the climatological mean and altitude known for 51 rainguage stations within

and around the river basin. Mean error, mean squared error and mean standardized square

error were calculated to compare the methods and cross-validation results showed that the

kriging technique with external drift produced the most coherent results.

In another study by Chegini et al., three interpolation methods (thin plate

smoothing splines (TPSS) with and without co-variable (TPSS-CO), weighted moving

average (WMA), and kriging (ordinary, cokriging and logkriging) were used to estimate

the monthly and annual rainfall data using the data from 167 climatic stations in

southwest Iran. The performance of the interpolation method was evaluated by mean bias

error and mean absolute error. Results showed that TPSS is the most precise method to

estimate annual rainfall and TPSS with a power of two was the most precise method to

estimate monthly and annual rainfall.

In the study by Mardikis et al. (2004), the prediction of long-term mean daily

reference evapotranspiration for each month in Greece was performed using four

interpolation methods namely ordinary kriging, inverse distance squared, residual kriging

(RK) and gradient-plus-inverse distance squared (GIDS) techniques. The accuracy of

these methods was assessed by examining ME, MAE and RMSE. The results showed that

RK and GIDS produced better accuracy of prediction when compared to the other two

methods.

7
3. Data Collection

Data for radon concentrations and other related statistics for homes have been

collected from various county health departments, commercial testing services and

university researchers. The original database for about 50,000 observations was created

by Kumar et al. (1990) and formed the basis of the Ohio Radon Information System

(Heydinger et al. (1990)). The database was extended in 1996 and 1997 to about 82,000

observations. The work continued and Sud (1998) reported total observations of 80,436

in her thesis while analyzing radon data. New data are being constantly added to the

database. Kumar and Varadarajan presented the analysis of 12,959 data points. This

thesis uses 130,826 indoor radon observations.

The database has been compiled over the last 18 years. Copies of databases were

requested from laboratories, university researchers, and others to compile a unified

database consisting of 130,826 measurements for 1590 zip code areas in Ohio. Though

this appears to be a large number, there are possibly still thousands of readings that must

have been taken but not yet reported. The majority of data received were from radon

testing organizations and have been reported in a series of technical papers, conference

proceedings and reports published by Kumar et al. (1989-2003) (References [10-22]).

8
The data sets used were diverse and had to be arranged accordingly. All included

the zip code area for the tested buildings and most provided the information on the type

of room tested, type of radon detection device used and the season of the year the test was

conducted. Only a few sources provided information on the building characteristics and

this is unfortunate, because some of these parameters may have a significant influence on

indoor radon levels.

A zip code-county listing was prepared and the following statistics have been

calculated for radon measurements in all zip code areas and counties: arithmetic mean,

geometric mean, standard deviation and variance using the Ohio Radon Information

System (ORIS). In computing these statistics, every radon measurement in a zip code was

used without regard for building, room, detection type or season. As the radon data is

heavily skewed towards the higher concentrations, the geometric mean of radon data is

used to input as a point source data.

Of the 1589 zip codes represented in the data, 253 zip codes are provided with

unknown county names and so these zip codes are not considered. Of the remaining 1336

zip codes, 270 zip codes were not shown in the Ohio zip codes shape file collected from

the ESRI website. From the radon data set of 1589 zip codes, 1066 zip codes are used for

inputting as point source data in the interpolation techniques.

9
4. Geostatistical analyst

The geostatistical analyst uses sample points taken at different locations and

interpolates a continuous surface. The sample points are measurements of some

phenomena such as radiation leaking from a nuclear power plant, an oil spill or elevation

heights. It derives a surface using the values from the measured locations to predict

values for locations where no data is collected. This tool provides two groups of

interpolation techniques:

Deterministic interpolation techniques.

Geostatistical interpolation techniques.

All methods rely on the similarity of nearby sample points to create the surface.

Deterministic techniques use mathematical functions for interpolation. Geostatistical

techniques rely on both statistical and mathematical methods that can be used to create

surfaces and assess the uncertainty of the predictions.

In addition to providing various interpolation techniques, geostatistical analyst

also provides many supporting tools. For example, prior to mapping, exploratory spatial

data analysis (ESDA) tools can be used to assess the statistical properties of data. After

exploring the data, one can create a variety of output map types (prediction, error of

10
prediction, probability and quantile maps) using many variants of kriging and cokriging

algorithms (ordinary, simple, universal, indicator, probability and disjunctive) and

associated tools (data transformation, declustering and detrending).

4.1 Exploratory Spatial Data Analysis (ESDA)

Exploratory spatial data analysis (ESDA) allows one to examine the data in

different ways. The ESDA environment is composed of a series of tools, each allowing a

view into the data. Each view can be manipulated and explored allowing different

insights into the data. Each view is interconnected with all other views as well as with

ArcMap i.e., if a bar is selected in the histogram, the points comprising the bar are

selected on any other open ESDA view and also on the ArcMap map. Certain tasks are

useful in most explorations, for defining the data distribution, looking for global and local

outliers and trends, and examining spatial autocorrelation and covariation among multiple

data sets. The different tools to view the data are:

Histogram

Normal QQ plot

Trend analysis

Semivariogram/covariance cloud

4.1.1 Histogram

The interpolation methods that are used to generate a surface give the best results

if the data is normally distributed. The histogram tool in ESDA provides a univariate

11
(one-variable) description of the data. The tool displays the frequency distribution for the

data set of interest and calculates summary statistics. The frequency distribution is a bar

graph that displays how often observed values fall within certain intervals or classes. The

relative proportion of data that falls in each class is represented by the height of each bar.

The following plots show the frequency distribution for the radon data set. The first plot

(Figure1) shows that the data is not normally distributed and so it has to be transformed

to a normal distribution to produce good results. This can be done by log transformation

and the second plot (Figure 2) shows the normally distributed data after the log

transformation.

Figure 1 Plot Showing the Frequency Distribution for Radon Data Set

12
Figure 2 Plot Showing the Frequency Distribution of Radon Data Set after
Log Transformation

4.1.2 Normal QQ Plot

QQ plot is another measure of the normality of the data. A general QQ plot is a

graph on which the quantiles from two distributions are plotted versus each other. For

two identical distributions, the QQ plot will be a straight line. Therefore, it is possible to

check the normality of the radon data by plotting the quantiles of that data versus the

quantiles of a standard normal distribution. The closer the points are to creating a straight

line, the closer the distribution is to being normally distributed. The first plot (Figure 3)

shows the data is not normally distributed, the main departure from standard normal

13
distribution line occurring at low and high radon concentration values. The second plot

(Figure 4) shows the data is close to a normal distribution after the log transformation.

Figure 3 Normal QQ Plot for Radon Data Set

14
Figure 4 Normal QQ Plot for Radon Data Set after Log Transformation

4.1.3 Trend Analysis

The Trend Analysis tool helps to identify global trends in the input data set. This

tool provides a three-dimensional perspective of the data (Figure 5).

15
Figure 5 Trend Analysis Dialog Box for Radon Data set

The radon concentration values are given by the height of each stick in the Z

dimension with the input data points on the top of the stick. The values are projected

onto the perpendicular planes, an East-West and a North-South plane. The X-axis is the

East- West axis and the Y axis is the North-South axis. A best-fit line (a polynomial) is

16
drawn through the projected points, which shows model trends in specific directions. The

green line is an east-west trend line and the blue line is a north-south trend line.

If the lines were flat, this would indicate that there is no trend. However, the blue

line in the image above starts out with low values and increases as it moves north and

then decreases. This demonstrates that the data seems to exhibit a strong trend in the

north south direction and a weaker one in the east west direction. The trend observed is

that the radon concentration values are higher in central Ohio than the values obtained as

you move away from center to north and south directions. By removing the trend, the

semivariogram will model the spatial autocorrelation among data points without having

to consider the trend in the data. The trend will automatically be added back to the

calculations before the final surface is produced.

4.1.4 Semivariogram/Covariance Cloud

The semivariogram/covariance cloud allows you to examine the spatial

autocorrelation between the measured sample points. In spatial autocorrelation, it is

assumed that things close to one another are more alike. The semivariogram/covariance

cloud lets you examine this relationship. To do so, a semivariogram value (the square of

the difference between the values of each pair of locations) is plotted on the y-axis

relative to the distance separating each pair on the x-axis (Figure 6).

17
Figure 6 Semivariogram/Covariance Cloud Dialog Box

Each red dot in the semivariogram/covariance cloud (Figure 6) represents a pair

of locations. Since closer locations should be more alike, the semivariogram will show

close locations (far left on the x-axis) having small semivariogram values (low on the y-

axis). As the distance between the pairs of locations increases (move right on the x-axis),

the semivariogram values should also increase (move up on the y-axis). However; a

certain distance is reached where the cloud flattens out, indicating that the relationship

between the pairs of locations beyond this distance is no longer correlated. In addition,

the values in the semivariogram cloud are put into bins based on the direction and

18
distance between pair of locations. These bin values are then averaged and smoothed to

produce the surface of the semivariogram. The semivariogram surface shows

semivariogram values in polar coordinates. The center of the semivariogram surface

corresponds to the origin of the semivariogram graph.

As you can see in Figure 6, the points that are close together have a high

semivariogram value. By selecting these points, the linked pairs can be seen on the arc

map. The north east data point (seen on the arc map, Figure 6) which is linked to the

surrounding points has a radon concentration of 39 pCi/L which is much higher than the

surrounding points and the south west data point which is linked to the surrounding data

points has a radon concentration of 35.4 pCi/L. So, this pair of locations shows the high

semivariogram values at low distances. These high radon concentration data points are

the possible global data outliers. Spatial autocorrelation for radon data displayed in

Figure 6 is not high due to the global outliers.

4.2 Deterministic Interpolation Technique

Deterministic interpolation techniques create surfaces from measured points,

based on either the extent of similarity (IDW) or the degree of smoothing (RBF). There

are four deterministic methods:

19
inverse distance weighted

global polynomial

local polynomial

radial basis functions

Deterministic interpolation techniques can be divided into two groups: global and

local. Global techniques calculate predictions using the entire data set. Local techniques

calculate predictions from the measured points within neighborhoods, which are smaller

spatial areas within the larger study area. geostatistical analyst provides the global

polynomial as a global interpolator and the IDW, local polynomial and RBF as local

interpolators.

A deterministic interpolation can either force the resulting surface to pass through

the data values or not. An interpolation technique that predicts a value identical to the

measured value at a sampled location is known as an exact interpolator. An inexact

interpolator predicts a value that is different from the measured value. The latter can be

used to avoid sharp peaks or troughs in the output surface. IDW and RBF are exact

interpolators, while global polynomial and local polynomial are inexact.

4.2.1 Inverse Distance Weighted

IDW interpolation explicitly implements the assumption that things close to one

another are more alike than those farther apart. To predict a value for any unmeasured

20
location, IDW will use the measured values surrounding the prediction location. Those

measured values closest to the prediction location will have more influence on the

predicted value than those farther away. Thus, IDW assumes that each measured point

has a local influence that diminishes with distance. It weighs the points closer to the

prediction location greater than those farther away, hence the name inverse distance

weighted.

A simple IDW weighting function, as defined by Shepard, is:

(1)

where w (d) is the weighting factor applied to a known value, d is the distance

from the known value to the unknown value, and p is a positive real number, called the

power parameter. Here weight decreases as distance increases from the interpolated

points. Greater values of p assign greater influence to values closest to the interpolated

point. The most common value of p is 2. A general form of interpolating a value using

IDW is:

(2)

Z is the value at an unknown location

21
4.2.2 Radial Basis Functions

RBF methods are a series of exact interpolation techniques i.e., the surface must

go through each measured sample value. There are five different basis functions:

thin-plate spline

spline with tension

completely regularized spline

multiquadric function

inverse multiquadric function

Each basis function has a different shape and results in a slightly different

interpolation surface. RBF methods are a form of artificial neural networks. RBF are

conceptually similar to fitting a rubber membrane through the measured sample values

while minimizing the total curvature of the surface. The selected basis function

determines how the rubber membrane will fit between the values. Figure 7 below

demonstrates conceptually how a RBF surface fits through a series of elevation sample

values.

22
Figure 7 Radial Basis Function Surface Fitting Through a Series of Elevation
Values

Notice in the cross section that the surface passes through the data values. Being

exact interpolators, the RBF methods differ from the global and local polynomial

interpolators, which are both inexact interpolators that do not require the surface to pass

through the measured points. When comparing the RBF to the IDW method (another

exact interpolator), IDW will never predict values above the maximum measured value or

below the minimum measured, however, the RBF can predict values above the maximum

and below the minimum measured values. The optimal parameters are determined using

cross validation in a similar manner as shown for IDW and local polynomial

interpolation.

23
4.2.3 Global Polynomial Interpolation

Global polynomial interpolation fits a smooth surface that is defined by a

mathematical function (a polynomial) to the input sample points. The global polynomial

surface changes gradually and captures coarse-scale patterns in the data. Conceptually,

global polynomial interpolation is like taking a piece of paper and fitting it between the

raised points (raised to the height of value). This is demonstrated in Figure 8 below for a

set of sample points of elevation taken on a gently sloping hill (the piece of paper is

magenta).

Figure 8 Global Polynomial Interpolation Fitting Through the Sample Points


of Elevation on Gently Sloping Hill

But a flat piece of paper will not accurately capture a landscape containing a valley.

However, if one is allowed to bend the piece of paper once you will get a much better fit.

Adding a term to the mathematical formula produces a similar result, a bend in the plane.

24
A flat plane (no bend in the piece of paper) is a first-order polynomial (linear). Allowing

for one bend is a second-order polynomial (quadratic), two bends a third-order (cubic),

and so forth, up to 10 are allowed in geostatistical analyst. Figure 9 conceptually

demonstrates a second-order polynomial fitted to a valley.

Figure 9 Global Polynomial Interpolation Fitted to a Valley Using Second


Order Polynomial

Rarely will the piece of paper pass through the actual measured points, thus

making global polynomial interpolation an inexact interpolator. Some points will be

above the piece of paper and others will be below. However, if the net differences (sum

of deviations from mean) of heights for points above and below the piece of paper are

calculated separately, the two sums should be similar. The surface in magenta is obtained

by using least-squares regression fit. The resulting surface minimizes the squared

differences among the raised values and the sheet of paper. Global polynomial

25
interpolators determine a single function which is mapped across the whole region.

Change in one input value affects the entire map.

4.2.4 Local Polynomial Interpolation

While global polynomial interpolation fits a polynomial to the entire surface,

local polynomial interpolation fits many polynomials, each within specified overlapping

neighborhoods. The search neighborhood can be defined using the search neighborhood

dialog box. The shape, maximum and minimum number of points to use, and the sector

configuration can be specified. Alternatively, a slider can be used to define the width of

the neighborhood in conjunction with a power parameter that will decrease the weights of

the sample points within the neighborhood based on distance. Thus, local polynomial

interpolation produces surfaces that account for more local variation. Local interpolators

apply algorithms repeatedly to a small portion of total set of points. Change in an input

value affects the results within the window.

4.3 Geostatistical Interpolation Techniques

The geostatistical interpolation technique is based on statistics and is used for

more advanced surface modeling that also includes errors or uncertainty of predictions.

Kriging comes under the geostatistical interpolation technique.

26
4.3.1 Kriging

Geostatistical methods create surfaces incorporating the statistical properties of

the measured data. Because geostatistics is based on statistics, these methods produce not

only prediction surfaces but also error or uncertainty surfaces, giving the user an

indication of how good the predictions are.

Kriging is divided into two distinct tasks: quantifying the spatial structure of the

data and producing a prediction. Quantifying the spatial data structure, known as

variography, is fitting a spatial dependence model to the data. To make a prediction for

an unknown value for a specific location, Kriging will use the fitted model from the

variography, the spatial data configuration and the values of the measured sample points

around the prediction location. Geostatistical analyst provides many tools to help

determine the parameters to be used and the defaults are also provided so that a surface

can be created quickly. Kriging is a moderately quick interpolator that can be exact or

smoothed depending on the measurement error model. It is very flexible and allows the

user to investigate graphs of spatial autocorrelation. Kriging uses statistical models that

allow a variety of map outputs including predictions, prediction standard error, standard

error of indicators, and probability. The flexibility of Kriging can require a lot of decision

making. Kriging assumes that the data comes from a stationary stochastic process. A

stochastic process is a collection of random variables that are ordered in space and/or

time such as elevation measurements.

27
The selection of a kriging method is based on the autocorrelation of radon

concentrations between two points, which is formulated as follows:

Z(s) = (s) + (s) (3)

Z(s) consists of two parts: a deterministic trend (s) (i.e. flow direction) and a

random auto correlated error (s). The symbol s simply indicates the location of a point.

Because (s) is a deterministic trend, the selection of a kriging method is based on

whether a directional trend exists or not. If a directional trend is unknown, then an

ordinary kriging method is appropriate. If a directional trend is known, then a simple

kriging method should be selected. Ordinary kriging was selected to map the radon

concentration distributions.

4.3.1.1 Ordinary kriging

The ordinary kriging formula is generally given by:

n (u )


n(u)

Z*(u)= (u)Z(u + 1 ( u ) m (4)


1 1

Where Z* (u) is the ordinary kriging estimate at spatial location u,

n (u) is the number of the data used at the known locations given a neighborhood,

Z (u) are the n measured data at locations u located close to u

m= mean of distribution

(u)= weights for location u computed from the spatial covariance matrix based on the

spatial continuity (variogram) model, which is given by:

28
1 n
(h)=
2 n i 1
( z (u i ) z (u i h)) 2 (5)

Where n is the number of data pairs separated by distance h

z(ui) and z(ui+h) are the data values at locations separated by distance h.

Ordinary kriging assumes a constant but unknown mean and estimates the mean

value as a constant in the searching neighborhood.

29
5. Approach and Procedure

The goal of geostatistical analysis is to predict the values where no data has been

collected. The State of Ohio consists of 1862 zip codes, but the data set obtained by the

University of Toledo from the radon testing organizations only consists of radon

concentration data for 1066 zip codes for the years 1989 to 2005 and some other data

collected prior to 1989. The geostatistical analyst is used to evaluate the radon

concentration data for the unmeasured 796 zip codes by different interpolation methods

(ordinary kriging, IDW, RBF, local and global polynomial interpolation method). The

bootstrap application is also used for validation purposes i.e., to determine the best

interpolation technique by examining the statistical parameters.

In the Ohio zip codes shape file, the geometric mean of radon concentration values

is inputted in the attribute table for each zip code and zero values are assigned to the zip

codes that are not measured. Then, the polygon features of the Ohio zip codes shape file

are converted into point features which are inputted as point data sources in the

interpolation techniques. The point featured shape file is then divided into two shape

files; one having 1066 zip codes with radon concentration data and the other contains 796

zip codes with no measured radon concentration data.

30
The first approach is to evaluate the best interpolation technique. For that, the

point featured shape file with 1066 zip codes is divided into 80% training data points

which is used for modeling i.e., to create the output surface; and the remaining 20% is

divided into test data points to validate the output surface by comparing the measured and

predicted values. The division of data set into relative percentages is based on the number

of available data points.

90-10(%) 80-20 (%) 70-30 (%) 60-40 (%)


Interpolation
Technique RMSE RMSE RMSE RMSE

Ordinary Kriging 3.46 2.23 3.33 2.86

IDW 3.46 2.40 3.31 2.29

RBF 3.46 2.36 3.31 2.93

Global 3.71 2.27 3.57 3.06


Polynomial
Local 3.41 2.28 3.30 2.91
Polynomial
Table 1 Sensitivity Analysis for Division of Data Set

The above table shows sensitivity analysis for possible divisions of the data into training -

testing data sets. 80-20 division is chosen since the RMSE values (calculated from

measured and predicted values) for the test data points are the lowest when compared to

70-30, 60-40 and 90-10 divisions.

A sufficient number of data points are required to create a surface and to make the

validation of that surface significant. Then the different interpolation techniques are

31
executed to evaluate the best interpolation technique from the surface created and

validating the test data points.

The second approach follows the step of evaluating the best interpolation

technique. Modeling is done for the whole radon data set, which creates a surface of

spatial variation of radon concentrations and the predictions for unmeasured zip codes

(where no data is collected) is evaluated from the surface created.

5.1 Application of Interpolation Techniques

The different interpolation techniques which are performed for training data set

and test data set are:

ordinary kriging

inverse distance weighting

radial basis function

global polynomial

local polynomial

32
5.1.1 Ordinary kriging

As the radon concentration data is skewed which is shown in the ESDA tool, the

log transformation is applied and as it was showing trend in the north south direction,

first order of trend removal is selected, shown in Figure 10.

Figure 10 Geostatistical Method Selection Dialog Box

The spatial autocorrelation of the transformed data was then modeled using

semivariogram/covariance modeling. The semivariogram displays the statistical

correlation of nearby data points. As the distance increases, the likelihood of these data

33
points being related becomes smaller. An empirical variogram is created by calculating

the squared difference between all pairs of points and grouping the vectors into similar

distance and direction classes (called binning) and then the average of the squared

differences for each bin is plotted as the red dot (displayed in Figure 11).

After calculating the empirical variogram for the measured data points, a model is

fitted to fit through the points. The spherical model is fitted (which is the best fit model)

and shown by the yellow line in Figure 11.

Figure 11 Semivariogram/Covariance Modeling

34
Ideally, the value of the semivariogram should be zero when the separation vector

h is zero. In practice, this is usually not true because of measurement errors. In this case,

a so-called nugget effect (nugget variance) exists having a value of 0.20487. As the

semivariogram increases, separation distance increases and reaches its maximum at a

distance called range parameter which is 46.55. The maximum semivariogram value is

the sill parameter (structured variance) which is 0.2869. The lag size is the size of a

distance class into which pairs of locations are grouped. The automated lag size and the

number of lags generated are 104950 and 12.

The color scale as shown in Figure 11 represents the calculated semivariogram

value. It provides a direct link between the semivariogram values on the graph and those

on the semivariogram surface. The value of each cell in the semivariogram surface is

color coded. The lower values are in blue and green. The higher values are in orange and

red.

The next step in kriging is searching neighborhood. As the data locations become

farther away from a location where the value is unknown, they may not be useful when

predicting the value at an unmeasured location. At some distance, the points will have no

correlation with the prediction location. Therefore, in predicting the radon concentration

of a specific point; the number of adjacent points, the searching radius and the number of

sectors of the circle (or ellipse) should be specified. As shown in Figure 12, five

neighboring points are considered and a circle with four sectors is considered. The points

highlighted in the data view give an indicator of the weights associated with each point.

35
The weights are used to estimate the value at the unknown location, which is at the center

of the crosshair.

Figure 12 Ordinary Kriging Searching Neighborhood Dialog Box

Before the final surface is produced for practical use, the cross-validation tool

from the ArcGIS Geostatistical analyst should be employed to examine how well the

surface model predicts radon concentration at a specific location. The cross-validation

tool compares measured radon values with predicted ones derived from the surface

model, and uses statistical measures to assess the surface models performance. The

statistical measures serve as guidelines for the accuracy of the surface model and its

36
prediction map. Figure 13 provides a graphical comparison between measured and

predicted values (predicted plot). The cross validation tool sequentially omits a point,

Figure 13 Ordinary Kriging Predicted Plot (Predicted values Vs Measured


Values)

predicts the associated value using the rest of the radon concentration data, and then

compares it with the measured value.

Ideally, the predicted values should be the same as the measured ones, and all data

points would form a 1:1 ratio line (the gray line in Figure 13). In reality, data points

would scatter along this line due to natural variations and uncertainties. The term

37
prediction error is used to describe the difference between the prediction and the actual

measured value. For a surface model that provides accurate predictions, the mean error

and mean standardized error should be close to zero, the root mean square error and the

average standard error should be as small as possible, and the root mean square

standardized error should be close to unity.

For the radon concentration training data set, the prediction errors for ordinary

kriging are:

mean: -0.05196

root mean square: 2.573

average standard error: 2.44

mean standardized: -0.03232

root mean square standardized: 1.111

Figure 14 shows the graphical comparison between errors and the measured

values. In general the error was small enough, since most of the points are around the

zero scale.

38
Figure 14 Ordinary Kriging Error Plot (Errors Vs Measured Values)

Figure 15 shows the graphical comparison between standardized error and

measured values. The standardized error is the measured value subtracted from the

predicted value and then divided by the estimated standard error.

39
Figure 15 Ordinary Kriging Standardized Error Plot (Standardized Error Vs
Measured Values)

The QQ plot shown in Figure 16 illustrates the quantiles of the difference between

the standardized errors and the corresponding quantiles from a standard normal

distribution. The points should roughly lie along the dashed line if the errors of

predictions from their true values are normally distributed. The plot shows that errors of

prediction are nearly normally distributed.

40
Figure 16 Ordinary Kriging QQ Plot

All of these plots show how well the ordinary kriging is predicting.

With the surface map determined from the cross validation process, the radon

concentration prediction map that shows the radon concentration distribution can be

generated. Figure 17 shows a radon concentration prediction map in which the dark color

represents high concentrations and the light color represents low concentrations.

41
Figure 17 Ordinary Kriging Radon Concentration Prediction Map for the
Training Data Set

42
5.1.2 Inverse Distance Weighting

As stated earlier, IDW interpolation follows the assumption that things close to

one another are more alike than those farther apart.

The first step in IDW is the searching neighborhood method which is conducted

on the radon concentration training data set, as shown in Figure 18. The maximum

number of neighboring points included is 15 and minimum of 10 neighboring points are

included. The shape selected is shown in the figure below with four sectors and the

maximum and minimum number of constraints available to each sector. The most

commonly used power parameter of 2 is selected.

Figure 18 IDW Searching Neighborhood Dialog Box

43
The final step is the cross validation tool (Figure 19) which shows a predicted plot

(measured values vs predicted values) and an error plot (errors vs measured values). For

IDW technique, radon concentration training data set the prediction errors are:

mean: -0.04331 and

root mean square 2.639

Unlike ordinary kriging, in the IDW, radial basis function, global polynomial and

local polynomial interpolation techniques, only two plots are generated: predicted plot

and error plot.

Figure 19 IDW Cross-validation Tool (Measured Vs Predicted Values)

44
Figure 20 shows the prediction map created by the IDW technique for radon

concentration training data sets.

Figure 20 IDW Radon Concentration Prediction Map for the Training Data
set

45
5.1.3 Radial Basis Function

Radial basis function is similar to IDW (which assumes that things close to one

another are more alike than those farther apart) except that it predicts values above the

maximum and below the minimum measured values.

Similar to IDW, the first step in RBF is the searching neighborhood method

which is conducted on radon concentration training data sets using the completely

regularized spline function with the parameter value of 0.0040628. Maximum and

minimum number of data points included is 15 and 10, respectively and the shape type is

a circle with four sectors shown in Figure 21.

Figure 21 RBF Searching Neighborhood Dialog Box

46
Figure 22 shows the final step, the cross validation tool, which illustrates the

predicted plot and the error plot. The prediction errors for RBF are:

mean: -0.00406 and

root mean square 2.625

Figure 22 RBF Cross Validation Tool (Measured Vs Predicted Values)

Figure 23 shows the prediction map created by the RBF technique for radon

concentration training data sets.

47
Figure 23 RBF Radon Concentration Prediction Map for the Training Data
Set

5.1.4 Global Polynomial Interpolation Technique

Global polynomial interpolation technique fits a plane through the measured data

points. A plane is typically a polynomial. Unlike the other interpolation techniques,

global polynomial interpolation is the only method in geostatistical analyst that does not

use a search neighborhood. The order of polynomial ranges from 1-10. The second order

48
polynomial is used for the radon concentration training data set shown in Figure 24,

because higher order polynomials showed greater root mean square error values.

Figure 24 Setting the Power Parameter in Global Interpolation Technique

The prediction errors for the radon concentration training data set which is shown

in cross validation tool (Figure 25) are:

mean -0.0008207 and

root mean square 2.752

49
Figure 25 Global Polynomial Interpolation Cross validation Tool (Measured
Vs Predicted Values)

Figure 26 shows the prediction map created by the global polynomial

interpolation technique for radon concentration training data sets.

50
Figure 26 Global Polynomial Interpolation Radon Concentration Prediction
Map for Training Data Set

51
5.1.5 Local Polynomial Interpolation

While global polynomial interpolation fits a polynomial to the entire surface, local

polynomial interpolation fits many polynomials, each within specified overlapping

neighborhoods. The conceptual basis for local polynomial interpolation is to fit many

smaller overlapping planes, and then use the center of each plane as the prediction for

each location in the study area. The resulting surface will be more flexible and perhaps

more accurate. This interpolation fits many polynomials, each within specified

overlapping neighborhoods. Local polynomial interpolation is sensitive to the

neighborhood distance.

The maximum and minimum number of data points included within the search

neighborhood is 852 and 10, respectively. The power parameter selected is 2, shown in

Figure 27.

52
Figure 27 Local Polynomial Interpolation Searching Neighborhood Dialog
Box

The prediction errors for the radon concentration training data set which is shown

in cross validation tool (Figure 28) are:

mean 0.01026 and

root mean square 2.614

53
Figure 28 Local Polynomial Interpolation Cross Validation Tool (Measured
Vs Predicted Values)

Figure 29 shows the prediction map created by the local polynomial interpolation

technique for radon concentration training data sets.

54
Figure 29 Local Polynomial Interpolation Radon Concentration Prediction
Map for the Training Data Set

55
5.2 Qualitative Analysis

5.2.1 Comparison of the behavior of the prediction maps with the soil
uranium concentrations map

The prediction maps created by the interpolations techniques are compared with

the distribution of uranium in soils map. Uranium in soils is considered as the primary

source of radon gas. The gas is derived from the radioactive decay of uranium which

occurs in some rocks, sediments and soils. The map (Figure 30) shows that the uranium

concentrations are above 3 ppm in central and western Ohio. The prediction maps

(Figures 17, 20, 23, 29) from ordinary kriging, inverse distance weighting and radial

basis function interpolations also show higher level of radon concentrations in central and

western Ohio. Moreover, central- eastern Ohio shows the elevated radon levels that result

from geologic factors not well understood presently. Many of these areas are probably

associated with relatively permeable and dry sediments (sands and gravels) or soil types

commonly found on hillsides and valley floors. Although such materials may not be

enriched in uranium, they can still cause high indoor radon levels by making it easier for

radon to get to the surface by virtue of their permeability and dryness. This same effect

may also contribute to some of the high radon levels in the central and western parts of

the state.

56
Figure 30 Aerial Radiometric Map of Ohio Showing the Concentration of
Uranium in Surficial Sediments and Soils

57
5.3 Evaluation Criteria

Certain statistical measures are built in the GIS software. These statistics are used during

the analysis while implanting interpolation techniques. Once all the interpolation

techniques are chosen, additional statistics are used based on the literature on air quality

model evaluations [23]. Confidence limits on the statistics are calculated using bootstrap

to evaluate the superiority of one technique over another. The final choice of the better

technique was based on the accuracy of the predictions, degree of confidence that could

be placed in the predictions, and the matching of overall spatial distribution of predictions

with the physical parameters that are the primary cause of indoor radon in Ohio.

The following performance measures were applied.

5.3.1 Normalized Mean Square Error (NMSE)

This performance measure emphasizes the scatter in the entire data set. The equation for

the mean square error was modified by Hanna and Heinold (1985) to obtain this

dimensionless statistic. The normalization product Cp (predicted value) and Co (observed

value) assure that the NMSE will not be baised towards the techniques that over predict

or under predict. Smaller values of NMSE denote better model performance. The

expression for the NMSE is given by:

NMSE
C 0 Cp
2

(6)
C0C p

58
5.3.2 The Coefficient of Correlation (r)

Correlation analysis involves parameters calculated from linear least squares and

associated graphical analysis. A value of r close to 1.0 implies a good correlation between

the observed and predicted values, i.e. good model performance. The numerical results

quantify the association while graphical analyses supplement by providing qualitative

results measured for the same. The correlation is estimated by the parameter r which is

given by:

r
C 0
C0 Cp C p (7)
C C
p 0

5.3.3 Fractional Bias (FB)

The bias is normalized to make it dimensionless. This FB varies between +2 and -2 and

has an ideal value of zero for an ideal model. The FB is expressed as

C0 C p
Fractional Bias (8)

0.5 C 0 C p

5.3.4 Root Mean Square Error (RMSE)

RMSE is an indicator of sensitivity of outliers. It indicates the magnitude of extreme

errors. The expression for RMSE is given by:

RMSE =
Co Cp 2 (9)
N

5.3.5 Mean Error (ME)

59
Mean error is an indicator of bias. It is expressed as:

Co Cp
ME = (10)
N

5.3.6 Mean Absolute Error (MAE)

Mean absolute error is expressed as:

Co Cp
MAE = (11)
N

5.3.7 Mean Square Error

Mean square error is expressed as:

MSE =
Co Cp 2 (12)
N

5.3.8 Confidence Limits

The various statistics are calculated using the above parameters containing a finite

number of observation predictions. Since these observations constitute a part of an

infinite distribution of samples, we must ascertain the confidence in our estimates of the

above mentioned statistics.

The bootstrapping technique is an effective tool to determine the confidence

limits at a desired percentage. Essentially, the bootstrap procedure involves random

sampling with replacement from the original sample set. Thus, any number of new

sample sets of the same size as the original data can be generated. The values of the

observed and predicted data points are picked independently of each other and not in

60
pairs. From these new n sample sets, n values of the statistic can be calculated. These

then can be used to form a cumulative distribution function of the statistic. Confidence

limits based on any required degree of certainty can be calculated from this distribution

function. The confidence limits help compare a statistic between two models.

In this study comparisons between the interpolation techniques and observed

values were attempted at a chosen value of 95 % confidence interval which should give

good agreement for any statistic measured (Kumar et al. 1993; Hanna et al. 1991). The

confidence limits are estimated on NMSE, FB, coefficient of correlation (r) for each and

among techniques in this study, using seductive and robust bootstrap resampling

techniques. The Seductive 95% Confidence Limits are based on the 2.5% and 97.5%

points on the Cumulative Distribution Function. The Robust 95% Confidence Limits are

based on the Usual Student t Approach Using Calculated Mean and Standard Deviation.

61
6. Results

After creating the surfaces from each interpolation technique using the training

data set, predictions for the test data set are evaluated from the surface created using the

validation tool. This tool also shows the graphical comparison between predicted and

measured values.

The plots given in Figures 31, 32, 33, 34 and 35 show the measured versus

predicted radon concentration values for the test data set for different interpolation

techniques.

Figure 31 Ordinary Kriging (Measured Vs Predicted Values)

62
Figure 32 IDW (Measured Vs Predicted Values)

Figure 33 RBF (Measured Vs Predicted Values)

Figure 34 Global Polynomial Interpolation (Measured Vs Predicted Values)

63
Figure 35 Local Polynomial Interpolation (Measured Vs Predicted Values)

In the above plots blue line is the fitted line drawn through the scattered points, and the

dotted line is the 1:1 ratio line. Ideally, the predicted values should be the same as the

measured ones, and the blue line should be close to the 1:1 line. The above plots illustrate

that the regression (blue) line for the ordinary kriging is close to 1:1 line, when compared

to the other interpolation techniques. Table 2 shows ME, MAE, MSE, and RMSE values,

for test data points for different interpolation techniques.

Radial Global Local


Ordinary
IDW Basis Polynomial Polynomial
Kriging
Function Interpolation Interpolation
ME 0.09 0.17 0.19 0.10 0.14

MAE 1.33 1.45 1.44 1.46 1.40

MSE 4.99 5.77 5.57 5.15 5.21

RMSE 2.23 2.40 2.36 2.27 2.28


Value

Table 2 ME, MAE, MSE and RMSE Values for Different Interpolation
Techniques for Geometric Mean of Radon Concentrations Test Predictions

64
From Table 2, it can be inferred that there is no significant difference among the error

values for all interpolation techniques. This implies that any of the interpolation

technique can be chosen.

The evaluation parameters compared for this study in order to choose the ideal

interpolation technique were also tested using additional statistical indicators. Absolute

and predicted values for the test data points were taken as the input and various error

values were generated, of which only NMSE, FB and coefficient of correlation (r) are

discussed in detail, based on air quality model evaluation studies. Table 3 below shows

the comparison for NMSE, FB and r values.

Radial Global Local


Ordinary
IDW Basis Polynomial Polynomial
Kriging
Function Interpolation Interpolation
NMSE 0.41 0.46 0.44 0.42 0.42

FB -0.026 -0.047 -0.055 -0.027 -0.041

Corr. (r) 0.50 0.42 0.45 0.48 0.47

Table 3 NMSE, FB and Coefficient of Correlation (r) Values From Bootstrap


Method

The statistical significance of evaluation parameters were further tested to

estimate whether the calculated performance measures FB, NMSE, coefficient of

correlation (r) are significantly different from zero for each technique and whether the

differences in their measures between pairs of techniques are significantly different from

65
zero (to generate 95% confidence limits). The confidence limits evaluated are shown in

Tables 4 and 5. The summaries of these results are tabulated in Tables 6 and 7.

The results indicate that when each technique is considered individually, NMSE and

coefficient of correlation (r) values are significantly different from zero while FB values

are not. This implies that the FB values could not be relied on for further evaluation.

When all five interpolation techniques are compared in their ability to produce acceptable

predictions, NMSE and correlation values are not significantly different while FB differs

significantly for ordinary kriging and radial basis function techniques. Thus, the use of

the confidence limits further strengthens the argument that any of the interpolation

techniques may be employed for the purpose of radon concentration estimation for

unmeasured zip codes because the results from a technique are not significantly different

from any other technique being considered.

66
Table 4 Robust and Seductive 95% Confidence Limits for Each Interpolation Technique

Robust 95% Confidence Limits Seductive 95% Confidence Limits

Interpolation Technique NMSE FB Corr.(r) NMSE FB Corr. (r)

Low High Low High Low High Low High Low High Low High

ordinary kriging 0.245 0.592 -0.121 0.048 0.385 0.594 0.256 0.600 -0.119 0.046 0.374 0.587

IDW 0.303 0.650 -0.152 0.037 0.274 0.545 0.314 0.650 -0.149 0.036 0.270 0.534

RBF 0.284 0.632 -0.158 0.026 0.288 0.588 0.294 0.634 -0.157 0.026 0.287 0.562

Global Polynomial 0.260 0.601 -0.123 0.055 0.393 0.547 0.272 0.608 -0.123 0.057 0.389 0.545

Local Polynomial 0.271 0.588 -0.138 0.034 0.354 0.575 0.281 0.590 -0.132 0.035 0.350 0.561

Note:

(1) The Seductive 95% Confidence Limits are based on the 2.5% and 97.5% points on the Cumulative Distribution Function.

(2) The Robust 95% Confidence Limits are based on the Usual Student t Approach Using Calculated Mean and Standard Deviation.

67
Table 5 Robust and Seductive 95% Confidence Limits Among the Techniques

Robust 95% Confidence Limits Seductive 95% Confidence Limits

Interpolation Technique NMSE FB Corr.(r) NMSE FB Corr. (r)

Low High Low High Low High Low High Low High Low High

ordinary kriging- IDW


-0.142 0.026 -0.01 0.051 -0.029 0.189 -0.153 0.015 -0.007 0.052 -0.023 0.193
ordinary kriging RBF
-0.121 0.042 0.004 0.055 -0.049 0.161 -0.127 0.033 0.006 0.057 -0.044 0.156
ordinary kriging GPI
-0.059 0.035 -0.04 0.035 -0.073 0.112 -0.06 0.038 -0.04 0.033 -0.077 0.107
ordinary kriging LPI
-0.059 0.037 -0.005 0.036 -0.05 0.1 -0.059 0.033 -0.004 0.035 -0.039 0.107
IDW- RBF
-0.012 0.049 -0.002 0.019 -0.062 0.013 -0.008 0.049 -0.002 0.018 -0.067 0.009
IDW- GPI
-0.062 0.154 -0.076 0.028 -0.218 0.097 -0.056 0.167 -0.074 0.024 -0.225 0.084
IDW- LPI
-0.006 0.101 -0.036 0.026 -0.121 0.01 -0.001 0.105 -0.038 0.021 -0.122 0.008
RBF- GPI
-0.085 0.14 -0.084 0.019 -0.203 0.131 -0.084 0.146 -0.083 0.015 -0.205 0.113
RBF- LPI
-0.022 0.079 -0.04 0.012 -0.091 0.03 -0.018 0.079 -0.042 0.007 -0.094 0.031
GPI LPI
-0.071 0.074 -0.024 0.061 -0.118 0.129 -0.068 0.071 -0.018 0.065 -0.108 0.127

68
Table 6 Summary of Robust and Seductive 95% Confidence Limits
Analyses on Each Technique

Ordinary Global Local


IDW RBF
Kriging Polynomial Polynomial
NMSE X X X X X

FB

Corr. (r) X X X X X

Note:

X indicates significantly different from zero.

Blank indicates not significantly different from zero.

69
Table 7 Summary of Robust and Seductive 95% Confidence Limits
Analyses Among Each Technique

Among Techniques
Interpolation Technique
NMSE FB Corr.(r)
Yes No Yes No Yes No
ordinary kriging- IDW

ordinary kriging RBF X

ordinary kriging - GPI

ordinary kriging - LPI

IDW- RBF

IDW- GPI

IDW- LPI

RBF- GPI

RBF- LPI

GPI LPI

Note:

Yes- Indicates significantly different from zero.

No- Indicates not significantly different from zero.

70
The prediction maps generated (Figures 17, 20, 23, and 29) from ordinary

kriging, IDW and RBF interpolations show more or less similar spatial patterns for

radon distribution as depicted by the soil uranium concentrations map in Figure 30

(Duval et al., 1985), but the global and local polynomial interpolation prediction

maps varies drastically. Hence, the global and local polynomial interpolation

techniques are unsuitable for the purpose of the study and thereby; any of the three

interpolation techniques can be employed for predicting the radon concentrations in

unmeasured zip codes. However, more work is needed to find a technique that could

produce results that are statistically different from these four techniques.

Of the three interpolation techniques, ordinary kriging interpolation has been

chosen for this study because, the prediction map generated from ordinary kriging

show much similar spatial patterns for radon distribution in central western Ohio

when compared to the soil uranium concentration map (Figure 30), than to the other

three interpolation techniques.

The decision protocol, like the choice of semivariogram model and the

searching neighborhood selected, is the same as the interpolation done with ordinary

kriging for the radon training data set. Figure 36 shows the prediction map created by

the ordinary kriging using the whole data set.

71
Figure 36 Ordinary Kriging Radon Concentration Prediction Map Using
the Whole Radon Data Set

72
Table 8 shows the predictions for geometric mean of radon concentration zip

codes with ordinary kriging interpolation technique.

Table 8 Predictions of Geometric Mean of Radon Concentrations Zip


Codes in Ohio using Ordinary Kriging
PREDICTED PREDICTED
COUNTY GM RADON COUNTY GM RADON
ZIP CODE NAME CONC. ZIP CODE NAME CONC.
45133 ADAMS 2.76 43014 ASHLAND 7.10
45167 ADAMS 2.77 44235 ASHLAND 2.87
45168 ADAMS 2.61 44287 ASHLAND 2.94
45646 ADAMS 2.77 44628 ASHLAND 6.14
45657 ADAMS 3.52 44638 ASHLAND 4.45
45661 ADAMS 2.69 44676 ASHLAND 4.21
45671 ADAMS 2.72 44691 ASHLAND 3.73
45684 ADAMS 5.35 44805 ASHLAND 3.90
45810 ALLEN 4.78 44822 ASHLAND 7.05
45812 ALLEN 4.44 44837 ASHLAND 4.21
45830 ALLEN 4.66 44851 ASHLAND 3.45
45844 ALLEN 4.12 44878 ASHLAND 4.07
45868 ALLEN 4.96 44903 ASHLAND 4.38
45877 ALLEN 4.77 44046 ASHTABULA 1.48
45887 ALLEN 4.46 44062 ASHTABULA 1.48
45896 ALLEN 3.79 44064 ASHTABULA 1.46
44086 ASHTABULA 1.50 43310 AUGLAIZE 3.40
44428 ASHTABULA 1.49 43331 AUGLAIZE 3.22
44450 ASHTABULA 1.47 43333 AUGLAIZE 3.71
43728 ATHENS 2.75 45306 AUGLAIZE 4.41
43730 ATHENS 2.48 45805 AUGLAIZE 4.44
43766 ATHENS 2.48 45845 AUGLAIZE 5.20
45651 ATHENS 2.19 45850 AUGLAIZE 3.76
45724 ATHENS 2.05 45106 BROWN 1.62
45772 ATHENS 1.44 45107 BROWN 2.21
45776 ATHENS 1.66 45120 BROWN 1.55
43716 BELMONT 2.98 45122 BROWN 2.14
43747 BELMONT 2.45 45133 BROWN 2.03
43773 BELMONT 2.34 45142 BROWN 2.27
43788 BELMONT 2.12 45144 BROWN 3.32
43793 BELMONT 2.13 45148 BROWN 2.39
43901 BELMONT 4.04 45176 BROWN 1.85
43907 BELMONT 3.93 45697 BROWN 2.30
43917 BELMONT 4.14 45003 BUTLER 4.12
43943 BELMONT 3.92 45005 BUTLER 3.54
43971 BELMONT 3.73 45030 BUTLER 2.63
43973 BELMONT 2.21 45036 BUTLER 3.02
45040 BUTLER 2.97 43044 CLARK 4.33
45231 BUTLER 2.42 43072 CLARK 5.73

73
PREDICTED PREDICTED
ZIP COUNTY GM RADON ZIP COUNTY GM RADON
CODE NAME CONC. CODE NAME CONC.
45240 BUTLER 2.45 43078 CLARK 5.03
45251 BUTLER 2.43 43153 CLARK 3.80
45252 BUTLER 2.53 45314 CLARK 3.79
45311 BUTLER 3.98 45324 CLARK 5.38
45327 BUTLER 3.93 45387 CLARK 4.07
43903 CARROLL 3.85 45424 CLARK 5.39
43945 CARROLL 4.52 45107 CLERMONT 2.22
43986 CARROLL 3.74 45111 CLERMONT 2.95
43988 CARROLL 4.23 45118 CLERMONT 1.97
44423 CARROLL 4.21 45130 CLERMONT 1.41
44621 CARROLL 4.19 45154 CLERMONT 1.88
44625 CARROLL 4.35 45162 CLERMONT 2.29
44656 CARROLL 5.02 45174 CLERMONT 2.89
44688 CARROLL 5.02 45242 CLERMONT 2.93
44695 CARROLL 4.25 45243 CLERMONT 2.88
44730 CARROLL 4.98 45244 CLERMONT 2.57
43029 CHAMPAIGN 4.42 45255 CLERMONT 2.39
43040 CHAMPAIGN 4.03 43160 CLINTON 3.60
43045 CHAMPAIGN 4.00 45054 CLINTON 3.02
43140 CHAMPAIGN 4.54 45068 CLINTON 3.20
43311 CHAMPAIGN 5.59 45122 CLINTON 2.25
43318 CHAMPAIGN 5.68 45135 CLINTON 3.22
43319 CHAMPAIGN 4.35 45142 CLINTON 2.50
43343 CHAMPAIGN 5.38 45152 CLINTON 2.65
43360 CHAMPAIGN 4.58 45162 CLINTON 2.32
45317 CHAMPAIGN 5.27 45335 CLINTON 3.48
45365 CHAMPAIGN 5.30 45385 CLINTON 3.88
43930 COLUMBIANA 5.65 43314 CRAWFORD 3.93
43932 COLUMBIANA 6.61 43335 CRAWFORD 4.10
44657 COLUMBIANA 3.66 44807 CRAWFORD 3.13
43006 COSHOCTON 6.59 44818 CRAWFORD 2.90
43014 COSHOCTON 6.90 44849 CRAWFORD 3.14
43749 COSHOCTON 4.47 44865 CRAWFORD 3.60
43762 COSHOCTON 4.87 44875 CRAWFORD 3.47
43802 COSHOCTON 5.04 44882 CRAWFORD 2.91
43804 COSHOCTON 5.92 44011 CUYAHOGA 2.53
43822 COSHOCTON 6.06 44012 CUYAHOGA 2.46
43832 COSHOCTON 5.27 44023 CUYAHOGA 2.04
43840 COSHOCTON 5.70 44028 CUYAHOGA 2.13

74
PREDICTED PREDICTED
ZIP COUNTY GM RADON ZIP COUNTY GM RADON
CODE NAME CONC. CODE NAME CONC.
44637 COSHOCTON 6.55 44039 CUYAHOGA 2.44
44654 COSHOCTON 6.02 44056 CUYAHOGA 2.15
44067 CUYAHOGA 1.98 43502 DEFIANCE 2.82
44087 CUYAHOGA 2.52 43506 DEFIANCE 2.88
44212 CUYAHOGA 2.32 43517 DEFIANCE 2.88
44233 CUYAHOGA 2.09 43527 DEFIANCE 2.69
44286 CUYAHOGA 2.06 43536 DEFIANCE 2.69
45318 DARKE 5.85 43545 DEFIANCE 2.57
45321 DARKE 5.31 43548 DEFIANCE 2.75
45337 DARKE 5.85 43557 DEFIANCE 2.95
45338 DARKE 5.35 43013 DELAWARE 4.56
45347 DARKE 5.31 43031 DELAWARE 5.03
45363 DARKE 5.74 43040 DELAWARE 4.20
45382 DARKE 5.39 43054 DELAWARE 5.11
45845 DARKE 5.21 43235 DELAWARE 5.27
45846 DARKE 4.40 43334 DELAWARE 4.05
45860 DARKE 5.02 43344 DELAWARE 4.66
45883 DARKE 4.47 43356 DELAWARE 4.54
43464 ERIE 2.74 45123 FAYETTE 4.19
44811 ERIE 4.23 45135 FAYETTE 3.60
44826 ERIE 4.08 45169 FAYETTE 3.44
44857 ERIE 4.80 45335 FAYETTE 3.40
44889 ERIE 3.19 45628 FAYETTE 4.50
43025 FAIRFIELD 5.41 43031 FRANKLIN 5.14
43062 FAIRFIELD 5.39 43064 FRANKLIN 5.94
43068 FAIRFIELD 6.63 43082 FRANKLIN 4.87
43103 FAIRFIELD 6.76 43103 FRANKLIN 7.51
43113 FAIRFIELD 6.00 43146 FRANKLIN 6.28
43135 FAIRFIELD 5.30 43162 FRANKLIN 5.90
43138 FAIRFIELD 4.60 43240 FRANKLIN 5.28
43149 FAIRFIELD 5.30 43504 FULTON 2.06
43115 FAYETTE 4.80 43522 FULTON 1.78
43143 FAYETTE 4.88 43532 FULTON 1.87
43145 FAYETTE 4.84 43545 FULTON 2.22
43570 FULTON 2.87 43153 GREENE 3.57
45656 GALLIA 1.68 43160 GREENE 3.39
45659 GALLIA 1.07 45068 GREENE 3.95
45678 GALLIA 0.99 45169 GREENE 3.41
45688 GALLIA 1.32 45177 GREENE 3.54

75
PREDICTED PREDICTED
COUNTY GM RADON COUNTY GM RADON
ZIP CODE NAME CONC. ZIP CODE NAME CONC.
45695 GALLIA 1.95 45323 GREENE 5.26
45741 GALLIA 1.98 45341 GREENE 5.32
45760 GALLIA 1.79 45368 GREENE 3.96
44040 GEAUGA 1.65 45420 GREENE 4.34
44060 GEAUGA 1.86 45424 GREENE 4.56
44077 GEAUGA 1.82 45440 GREENE 4.38
44094 GEAUGA 1.92 45458 GREENE 3.80
44099 GEAUGA 1.51 45502 GREENE 4.18
44139 GEAUGA 2.40 43713 GUERNSEY 2.27
44202 GEAUGA 2.32 43724 GUERNSEY 1.85
44231 GEAUGA 1.66 43762 GUERNSEY 2.80
44234 GEAUGA 1.80 43832 GUERNSEY 3.54
44255 GEAUGA 1.82 43837 GUERNSEY 2.84
44491 GEAUGA 1.54 43983 GUERNSEY 2.18
44699 GUERNSEY 2.74 43908 HARRISON 3.60
45013 HAMILTON 2.70 43910 HARRISON 3.95
45014 HAMILTON 2.40 43917 HARRISON 4.11
45040 HAMILTON 2.86 43950 HARRISON 4.01
45053 HAMILTON 2.86 43973 HARRISON 2.64
45069 HAMILTON 2.60 43977 HARRISON 3.00
45150 HAMILTON 2.83 44621 HARRISON 4.26
45210 HAMILTON 1.35 44683 HARRISON 4.22
43316 HANCOCK 3.90 43502 HENRY 2.81
43359 HANCOCK 3.81 43511 HENRY 2.10
43516 HANCOCK 2.51 43515 HENRY 2.02
44802 HANCOCK 3.51 43522 HENRY 1.89
44817 HANCOCK 2.73 43557 HENRY 2.95
44830 HANCOCK 2.95 43567 HENRY 2.58
45817 HANCOCK 4.90 43569 HENRY 1.91
45843 HANCOCK 3.90 45856 HENRY 2.79
45872 HANCOCK 2.39 45118 HIGHLAND 2.12
43310 HARDIN 3.67 45146 HIGHLAND 2.65
43332 HARDIN 3.94 45154 HIGHLAND 2.06
43345 HARDIN 3.58 45159 HIGHLAND 3.03
43346 HARDIN 3.83 45169 HIGHLAND 3.32
43347 HARDIN 3.54 45171 HIGHLAND 2.34
45814 HARDIN 4.65 45612 HIGHLAND 3.40
45817 HARDIN 4.99 45646 HIGHLAND 2.92
45841 HARDIN 4.95 45660 HIGHLAND 2.81

76
PREDICTED PREDICTED
COUNTY GM RADON COUNTY GM RADON
ZIP CODE NAME CONC. ZIP CODE NAME CONC.
45850 HARDIN 3.90 45679 HIGHLAND 2.41
45697 HIGHLAND 2.42 43844 HOLMES 6.29
43102 HOCKING 5.34 44608 HOLMES 3.99
43107 HOCKING 4.05 44624 HOLMES 4.23
43130 HOCKING 5.31 44627 HOLMES 3.85
43155 HOCKING 4.93 44662 HOLMES 3.74
43748 HOCKING 3.70 44676 HOLMES 3.96
43766 HOCKING 2.64 44689 HOLMES 3.88
45601 HOCKING 3.80 44842 HOLMES 5.18
45622 HOCKING 2.56 44090 HURON 3.21
45651 HOCKING 2.27 44807 HURON 3.41
45654 HOCKING 2.45 44859 HURON 3.96
45732 HOCKING 2.58 44865 HURON 4.37
45764 HOCKING 2.58 44875 HURON 3.86
43006 HOLMES 6.25 44878 HURON 4.45
43014 HOLMES 6.66 43013 KNOX 4.51
43804 HOLMES 5.78 43055 KNOX 6.37
43812 HOLMES 5.94 43071 KNOX 6.17
43824 HOLMES 6.22 43074 KNOX 4.37
43843 HOLMES 6.77 43080 KNOX 5.86
45601 JACKSON 2.30 43822 KNOX 6.35
45613 JACKSON 1.89 43843 KNOX 6.58
45634 JACKSON 1.83 44628 KNOX 6.13
45651 JACKSON 2.00 44813 KNOX 6.15
45653 JACKSON 1.77 44822 KNOX 6.93
45658 JACKSON 1.69 44842 KNOX 6.70
45672 JACKSON 2.28 44026 LAKE 1.87
45682 JACKSON 1.80 44040 LAKE 1.48
45685 JACKSON 1.88 44086 LAKE 1.82
45686 JACKSON 1.89 44132 LAKE 1.10
45695 JACKSON 1.88 44143 LAKE 1.21
43901 JEFFERSON 4.71 45623 LAWRENCE 0.97
43907 JEFFERSON 4.46 45629 LAWRENCE 1.49
43935 JEFFERSON 3.89 45656 LAWRENCE 1.65
43945 JEFFERSON 5.49 45658 LAWRENCE 1.14
43968 JEFFERSON 6.39 45659 LAWRENCE 1.28
43971 JEFFERSON 3.75 45675 LAWRENCE 1.21
43976 JEFFERSON 3.94 45688 LAWRENCE 1.31
43986 JEFFERSON 3.74 45694 LAWRENCE 1.42

77
PREDICTED PREDICTED
COUNTY GM RADON COUNTY GM RADON
ZIP CODE NAME CONC. ZIP CODE NAME CONC.
44615 JEFFERSON 4.01 43002 MADISON 6.24
43004 LICKING 6.09 43026 MADISON 6.22
43046 LICKING 5.26 43029 MADISON 4.46
43050 LICKING 5.71 43044 MADISON 4.57
43054 LICKING 5.47 43045 MADISON 4.46
43105 LICKING 5.11 43106 MADISON 4.20
43147 LICKING 6.13 43119 MADISON 6.00
43739 LICKING 5.27 43123 MADISON 5.68
43760 LICKING 4.85 43128 MADISON 3.73
43830 LICKING 5.76 45314 MADISON 3.71
43843 LICKING 6.46 45335 MADISON 3.56
43844 LICKING 6.40 45368 MADISON 4.15
43040 LOGAN 3.79 45369 MADISON 4.42
43067 LOGAN 3.63 44408 MAHONING 3.52
43326 LOGAN 4.02 44411 MAHONING 1.80
43340 LOGAN 3.51 44412 MAHONING 1.76
45340 LOGAN 4.64 44420 MAHONING 1.84
45365 LOGAN 5.15 44425 MAHONING 2.08
45895 LOGAN 3.54 44437 MAHONING 1.76
44089 LORAIN 2.75 44440 MAHONING 1.76
44256 LORAIN 2.75 44444 MAHONING 1.70
44275 LORAIN 2.63 44460 MAHONING 2.23
44280 LORAIN 2.61 44481 MAHONING 1.72
44851 LORAIN 3.22 44601 MAHONING 1.96
44859 LORAIN 3.48 43003 MARION 4.17
44880 LORAIN 2.81 43315 MARION 4.18
44889 LORAIN 2.93 43320 MARION 4.27
43402 LUCAS 1.88 43344 MARION 4.66
43445 LUCAS 2.96 43351 MARION 4.04
43449 LUCAS 2.89 44820 MARION 4.07
43460 LUCAS 2.35 44833 MARION 4.30
43522 LUCAS 1.79 45843 MARION 4.08
43551 LUCAS 2.28 45845 MERCER 5.15
43558 LUCAS 1.92 45865 MERCER 5.00
44028 MEDINA 2.44 45869 MERCER 4.69
44044 MEDINA 2.52 45874 MERCER 3.62
44090 MEDINA 2.53 45885 MERCER 4.52
44214 MEDINA 3.29 45887 MERCER 4.28
44217 MEDINA 3.57 45894 MERCER 4.16

78
PREDICTED
ZIP COUNTY PREDICTED GM COUNTY GM RADON
CODE NAME RADON CONC. ZIP CODE NAME CONC.
44233 MEDINA 2.32 45898 MERCER 3.38
44235 MEDINA 2.83 43072 MIAMI 5.25
44270 MEDINA 3.78 45304 MIAMI 5.83
44276 MEDINA 3.76 45333 MIAMI 6.00
44287 MEDINA 2.93 45344 MIAMI 5.51
44880 MEDINA 2.83 45363 MIAMI 5.87
45614 MEIGS 2.00 45365 MIAMI 5.38
45620 MEIGS 1.77 45377 MIAMI 5.33
45686 MEIGS 2.00 45380 MIAMI 5.91
45695 MEIGS 1.87 45424 MIAMI 5.33
45701 MEIGS 1.90 43724 MONROE 1.86
45710 MEIGS 1.92 43773 MONROE 2.02
45723 MEIGS 1.47 43788 MONROE 1.89
45735 MEIGS 1.54 43942 MONROE 3.15
45745 MONROE 1.89 43003 MORROW 3.96
45767 MONROE 1.85 43011 MORROW 4.12
45789 MONROE 1.96 43019 MORROW 4.74
45005 MONTGOMERY 3.60 43074 MORROW 3.89
45042 MONTGOMERY 4.08 43314 MORROW 4.12
45066 MONTGOMERY 3.36 43335 MORROW 4.25
45068 MONTGOMERY 3.48 44813 MORROW 5.15
45304 MONTGOMERY 5.24 44833 MORROW 4.19
45305 MONTGOMERY 3.98 44903 MORROW 4.34
45311 MONTGOMERY 4.18 44904 MORROW 4.39
45324 MONTGOMERY 5.17 43055 MUSKINGUM 5.92
45337 MONTGOMERY 5.06 43056 MUSKINGUM 4.97
45338 MONTGOMERY 5.04 43076 MUSKINGUM 5.11
45341 MONTGOMERY 5.26 43731 MUSKINGUM 2.95
45344 MONTGOMERY 5.44 43732 MUSKINGUM 2.34
45370 MONTGOMERY 3.48 43739 MUSKINGUM 5.04
45371 MONTGOMERY 4.82 43749 MUSKINGUM 4.21
45381 MONTGOMERY 4.65 43756 MUSKINGUM 2.99
45383 MONTGOMERY 4.87 43760 MUSKINGUM 3.99
45407 MONTGOMERY 3.69 43811 MUSKINGUM 5.26
43724 MORGAN 2.61 43812 MUSKINGUM 5.37
43730 MORGAN 2.59 43732 NOBLE 2.08
43731 MORGAN 2.91 43754 NOBLE 1.78
43732 MORGAN 2.51 43756 NOBLE 2.42
43777 MORGAN 3.01 43772 NOBLE 1.84

79
PREDICTED PREDICTED
COUNTY GM RADON COUNTY GM RADON
ZIP CODE NAME CONC. ZIP CODE NAME CONC.
45711 MORGAN 2.46 43773 NOBLE 2.08
45724 MORGAN 3.02 43778 NOBLE 1.98
45732 MORGAN 2.51 43780 NOBLE 1.89
45786 MORGAN 3.19 45715 NOBLE 2.69
45744 NOBLE 2.62 43102 PICKAWAY 5.34
45745 NOBLE 1.86 43110 PICKAWAY 7.25
45746 NOBLE 2.06 43115 PICKAWAY 5.19
43420 OTTAWA 2.53 43125 PICKAWAY 7.42
43442 OTTAWA 2.82 43135 PICKAWAY 5.00
43447 OTTAWA 3.02 43140 PICKAWAY 5.72
43469 OTTAWA 3.17 43143 PICKAWAY 5.67
43512 PAULDING 2.33 43154 PICKAWAY 6.10
43526 PAULDING 2.68 43160 PICKAWAY 5.33
43536 PAULDING 2.55 45601 PICKAWAY 5.27
43556 PAULDING 2.64 45644 PICKAWAY 4.88
45827 PAULDING 2.79 45133 PIKE 2.90
45831 PAULDING 2.39 45601 PIKE 2.49
45886 PAULDING 2.36 45612 PIKE 3.24
43025 PERRY 5.47 45648 PIKE 2.34
43046 PERRY 5.02 45657 PIKE 2.68
43107 PERRY 4.02 45660 PIKE 2.88
43138 PERRY 3.48 45671 PIKE 2.75
43148 PERRY 4.95 44023 PORTAGE 2.12
43150 PERRY 4.30 44087 PORTAGE 2.67
43758 PERRY 2.71 44224 PORTAGE 2.21
43777 PERRY 3.16 44312 PORTAGE 2.64
43782 PERRY 2.72 44444 PORTAGE 1.69
45732 PERRY 2.53 44491 PORTAGE 1.52
44632 PORTAGE 2.50 43019 RICHLAND 5.30
44685 PORTAGE 2.89 44805 RICHLAND 4.00
45042 PREBLE 4.01 44827 RICHLAND 3.65
45056 PREBLE 3.96 44833 RICHLAND 3.92
45064 PREBLE 3.99 44837 RICHLAND 4.08
45304 PREBLE 5.33 44859 RICHLAND 4.09
45309 PREBLE 5.26 44864 RICHLAND 6.38
45325 PREBLE 4.61 44887 RICHLAND 3.58
45327 PREBLE 4.33 43113 ROSS 5.19
45332 PREBLE 5.29 43135 ROSS 4.28
45345 PREBLE 4.56 43145 ROSS 4.67

80
PREDICTED PREDICTED
COUNTY GM RADON COUNTY GM RADON
ZIP CODE NAME CONC. ZIP CODE NAME CONC.
45346 PREBLE 5.32 43152 ROSS 3.69
43512 PUTNAM 2.63 43160 ROSS 4.62
43516 PUTNAM 2.68 43164 ROSS 5.13
43524 PUTNAM 2.82 45123 ROSS 4.43
43548 PUTNAM 2.96 45672 ROSS 3.05
45807 PUTNAM 4.41 45690 ROSS 3.43
45840 PUTNAM 4.10 43406 SANDUSKY 2.49
45858 PUTNAM 3.13 43416 SANDUSKY 2.81
45868 PUTNAM 4.78 44811 SANDUSKY 3.59
44824 SANDUSKY 3.39 43333 SHELBY 3.86
44830 SANDUSKY 2.60 43343 SHELBY 4.62
44836 SANDUSKY 3.10 45317 SHELBY 5.22
44841 SANDUSKY 2.68 45318 SHELBY 5.96
44870 SANDUSKY 2.87 45326 SHELBY 5.37
44883 SANDUSKY 2.79 45356 SHELBY 5.84
45613 SCIOTO 1.81 45380 SHELBY 5.55
45616 SCIOTO 4.29 45388 SHELBY 5.46
45638 SCIOTO 1.57 45860 SHELBY 5.20
45640 SCIOTO 1.81 45865 SHELBY 5.24
45656 SCIOTO 1.75 45869 SHELBY 5.09
45657 SCIOTO 3.15 45871 SHELBY 5.05
45659 SCIOTO 1.45 45895 SHELBY 3.99
45660 SCIOTO 2.91 44201 STARK 2.25
45661 SCIOTO 2.57 44411 STARK 1.89
45671 SCIOTO 2.78 44612 STARK 5.03
45682 SCIOTO 1.81 44618 STARK 3.84
43316 SENECA 3.46 44634 STARK 3.15
43407 SENECA 2.70 44645 STARK 4.08
43410 SENECA 3.17 44680 STARK 4.76
43420 SENECA 2.83 44689 STARK 3.89
43457 SENECA 2.61 44125 SUMMIT 1.71
44811 SENECA 3.83 44141 SUMMIT 2.21
44844 SENECA 2.96 44146 SUMMIT 1.90
44854 SENECA 3.11 44230 SUMMIT 3.69
44882 SENECA 2.91 44233 SUMMIT 2.27
44890 SENECA 3.83 44240 SUMMIT 2.21
44256 SUMMIT 2.66 43002 UNION 5.56
44325 SUMMIT 1.99 43015 UNION 4.72
44614 SUMMIT 4.02 43017 UNION 5.24

81
PREDICTED PREDICTED
GM RADON GM RADON
ZIP CODE COUNTY NAME CONC. ZIP CODE COUNTY NAME CONC.
44645 SUMMIT 4.09 43044 UNION 4.26
44720 SUMMIT 3.80 43060 UNION 3.87
44062 TRUMBULL 1.50 43061 UNION 4.38
44076 TRUMBULL 1.46 43065 UNION 5.09
44099 TRUMBULL 1.44 43066 UNION 4.29
44231 TRUMBULL 1.55 43302 UNION 4.48
44288 TRUMBULL 1.56 43319 UNION 3.98
43824 TUSCARAWAS 5.84 43332 UNION 4.00
44608 TUSCARAWAS 4.30 43340 UNION 3.69
44620 TUSCARAWAS 4.95 43342 UNION 4.60
44626 TUSCARAWAS 4.89 43358 UNION 3.53
44643 TUSCARAWAS 5.26 45042 WARREN 3.49
44654 TUSCARAWAS 4.87 45044 WARREN 3.12
44675 TUSCARAWAS 4.41 45050 WARREN 3.05
44699 TUSCARAWAS 3.23 45069 WARREN 2.90
45827 VAN WERT 2.85 45122 WARREN 2.27
45833 VAN WERT 3.80 45177 WARREN 3.58
45844 VAN WERT 3.47 45241 WARREN 2.84
45849 VAN WERT 2.38 45249 WARREN 2.91
45851 VAN WERT 2.46 45327 WARREN 3.69
45880 VAN WERT 2.53 45342 WARREN 3.41
45887 VAN WERT 4.38 45370 WARREN 3.71
43135 VINTON 3.59 45458 WARREN 3.30
43138 VINTON 2.63 43724 WASHINGTON 2.39
43152 VINTON 3.02 43728 WASHINGTON 2.79
45601 VINTON 2.89 43787 WASHINGTON 3.21
45647 VINTON 3.12 45711 WASHINGTON 2.49
45686 VINTON 1.90 45723 WASHINGTON 1.84
45692 VINTON 1.93 45727 WASHINGTON 2.06
45710 VINTON 1.87 45734 WASHINGTON 1.74
45741 VINTON 2.03 45778 WASHINGTON 2.15
45764 VINTON 2.30 43412 WOOD 2.88
45766 VINTON 2.01 43430 WOOD 3.00
44611 WAYNE 4.15 43431 WOOD 2.49
44614 WAYNE 4.17 43469 WOOD 2.79
44624 WAYNE 3.69 43516 WOOD 2.35
44633 WAYNE 3.80 43537 WOOD 2.31
44662 WAYNE 3.63 43566 WOOD 1.85
44666 WAYNE 4.10 43605 WOOD 2.35

82
PREDICTED PREDICTED
COUNTY GM RADON COUNTY GM RADON
ZIP CODE NAME CONC. ZIP CODE NAME CONC.
44840 WAYNE 4.08 43609 WOOD 2.40
44842 WAYNE 4.24 44830 WOOD 2.69
44866 WAYNE 3.08 45889 WOOD 2.43
43502 WILLIAMS 3.06 43302 WYANDOT 3.88
43521 WILLIAMS 2.57 43326 WYANDOT 3.97
43332 WYANDOT 4.07
43337 WYANDOT 3.80
44802 WYANDOT 3.88
44820 WYANDOT 3.74
45843 WYANDOT 3.95
45867 WYANDOT 3.72

From the data sets available prior to this study, the number of zip codes having

geometric mean of radon concentration over 4.0 pCi/l is 390; of which 94 zip codes

are not shown in the Ohio zip codes shape file. After using the ordinary kriging

interpolation technique to calculate the predictions for unmeasured zip codes, the

number of zip codes having radon concentration over 4.0 pCi/l is 688, inclusive of the

ones not shown in the shape file. The predicted geometric mean of radon

concentrations are below 8 pCi/l.

Therefore, for the cases where the geometric mean of radon concentration exceeds 8

pCi/l and 20 pCi/l, the number of zip codes from existing data is found to be equal to

that obtained by using the above mentioned interpolation technique for unmeasured

zip codes (85 and 9 for the respective cases).

And using the IDW and RBF interpolation techniques, the number of zip codes

having radon concentration over 4, 8, 20 pCi/l are found to be equal to that obtained

from ordinary kriging.

83
Table 9 Zip Codes with Geometric Mean of Radon Concentration greater
than 4 pCi/l (without Predicted Zip Codes)
ZIP GM RADON ZIP GM RADON
CODE COUNTY CONC. CODE COUNTY CONC.
45616 ADAMS 5.55 43357 CHAMPAIGN 7.8
45650 ADAMS 35.4 45344 CHAMPAIGN 6.22
45806 ALLEN 4.21 45389 CHAMPAIGN 9.72
45817 ALLEN 4.21 45502 CHAMPAIGN 4.92
45833 ALLEN 4.79 45323 CLARK 5.69
45850 ALLEN 10.42 45341 CLARK 12.43
44840 ASHLAND 6.78 45344 CLARK 6.4
44843 ASHLAND 4.3 45372 CLARK 5.18
44848 ASHLAND 8.61 45501 CLARK 8.4
44859 ASHLAND 4.34 45503 CLARK 4.7
44864 ASHLAND 8.59 45505 CLARK 4.53
44068 ASHTABULA 4.14 45506 CLARK 4.17
44088 ASHTABULA 20.87 45138 CLINTON 24.82
45806 AUGLAIZE 4.86 45164 CLINTON 4.8
45819 AUGLAIZE 21.5 43920 COLUMBIANA 5.56
45865 AUGLAIZE 5.15 43945 COLUMBIANA 4.97
45871 AUGLAIZE 6.32 44432 COLUMBIANA 4.1
45884 AUGLAIZE 6 44441 COLUMBIANA 8.24
43719 BELMONT 7.64 44443 COLUMBIANA 11.2
43906 BELMONT 4.45 44455 COLUMBIANA 11.04
43933 BELMONT 7.24 43811 COSHOCTON 9.24
43937 BELMONT 4.62 43812 COSHOCTON 4.26
43940 BELMONT 6.2 43824 COSHOCTON 17.49
43977 BELMONT 5.5 43828 COSHOCTON 8.8
45171 BROWN 5.42 43843 COSHOCTON 6.47
45018 BUTLER 12.24 43844 COSHOCTON 6.52
45043 BUTLER 5.39 43845 COSHOCTON 7.77
45062 BUTLER 6.29 44833 CRAWFORD 4.6
45067 BUTLER 4.62 44860 CRAWFORD 7.3
45241 BUTLER 4.05 44887 CRAWFORD 4.22
44427 CARROLL 9.33 44195 CUYAHOGA 5
44639 CARROLL 7.95 45303 DARKE 7.58
44643 CARROLL 6.63 45304 DARKE 5.61
44644 CARROLL 4.54 45308 DARKE 5.34
44657 CARROLL 6.4 45331 DARKE 4.93
43009 CHAMPAIGN 5.6 45346 DARKE 4.68
43047 CHAMPAIGN 15.3 45351 DARKE 5.05
43060 CHAMPAIGN 4.58 45352 DARKE 10.8
43078 CHAMPAIGN 4.9 45380 DARKE 4.35

84
ZIP GM RADON ZIP GM RADON
CODE COUNTY CONC. CODE COUNTY CONC.
45388 DARKE 4.3 43125 FRANKLIN 8
45390 DARKE 5.06 43126 FRANKLIN 9.61
43011 DELAWARE 5.34 43137 FRANKLIN 11.37
43016 DELAWARE 8.6 43140 FRANKLIN 5.87
43017 DELAWARE 6.32 43147 FRANKLIN 6.1
43021 DELAWARE 4.01 43202 FRANKLIN 4.34
43064 DELAWARE 5.43 43204 FRANKLIN 4.24
43065 DELAWARE 4.24 43205 FRANKLIN 4.74
43081 DELAWARE 11.07 43206 FRANKLIN 4.67
43342 DELAWARE 17.3 43207 FRANKLIN 6.87
44814 ERIE 4.87 43209 FRANKLIN 4.02
44824 ERIE 7.9 43213 FRANKLIN 6.15
44847 ERIE 8.03 43214 FRANKLIN 5.52
43102 FAIRFIELD 4.4 43216 FRANKLIN 17.9
43107 FAIRFIELD 4.47 43218 FRANKLIN 5.1
43110 FAIRFIELD 9.71 43219 FRANKLIN 6.46
43112 FAIRFIELD 6.63 43220 FRANKLIN 5.14
43130 FAIRFIELD 6.06 43221 FRANKLIN 6
43136 FAIRFIELD 10.19 43222 FRANKLIN 5.01
43147 FAIRFIELD 4.8 43224 FRANKLIN 5.15
43148 FAIRFIELD 4.39 43226 FRANKLIN 4.86
43150 FAIRFIELD 5.9 43227 FRANKLIN 7.48
43154 FAIRFIELD 4.04 43228 FRANKLIN 4.3
43155 FAIRFIELD 4.4 43229 FRANKLIN 5.29
43106 FAYETTE 4.61 43230 FRANKLIN 5.94
43142 FAYETTE 9.6 43231 FRANKLIN 4.44
43002 FRANKLIN 9.9 43232 FRANKLIN 9.89
43004 FRANKLIN 5.95 43234 FRANKLIN 7.4
43016 FRANKLIN 5.5 43235 FRANKLIN 5.61
43017 FRANKLIN 6.01 43279 FRANKLIN 10.4
43026 FRANKLIN 5.64 43553 FULTON 4.77
43054 FRANKLIN 4.21 45301 GREENE 10.69
43068 FRANKLIN 4.96 45305 GREENE 4.21
43080 FRANKLIN 9.81 45370 GREENE 4.07
43081 FRANKLIN 4.75 45385 GREENE 4.73
43085 FRANKLIN 7.68 45387 GREENE 4.21
43109 FRANKLIN 7.48 45430 GREENE 10.97
43110 FRANKLIN 7.01 45434 GREENE 4.66
43119 FRANKLIN 5.18 45435 GREENE 5.76

85
ZIP GM RADON ZIP GM RADON
CODE COUNTY CONC. CODE COUNTY CONC.
43123 FRANKLIN 5.12 43750 GUERNSEY 5.88
43778 GUERNSEY 4.54 43917 JEFFERSON 8.36
44230 HAMILTON 4.1 43925 JEFFERSON 10.5
45111 HAMILTON 4.36 43930 JEFFERSON 39
45140 HAMILTON 4.1 43932 JEFFERSON 15.99
45174 HAMILTON 4.27 43943 JEFFERSON 6.06
45228 HAMILTON 4.2 43005 KNOX 7.83
45263 HAMILTON 20.5 43006 KNOX 12.49
45280 HAMILTON 4.99 43011 KNOX 4.99
45841 HANCOCK 5.9 43014 KNOX 6
45868 HANCOCK 5.32 43019 KNOX 6.42
45881 HANCOCK 7.31 43022 KNOX 7.01
45890 HANCOCK 9.97 43028 KNOX 12.05
45810 HARDIN 4.47 43050 KNOX 5.26
45835 HARDIN 14 43001 LICKING 6.23
45843 HARDIN 4.37 43011 LICKING 4.69
43901 HARRISON 6.69 43023 LICKING 8.85
43907 HARRISON 5.02 43025 LICKING 5.9
43988 HARRISON 16.99 43027 LICKING 4.9
44695 HARRISON 4.53 43030 LICKING 10.19
43548 HENRY 5.86 43031 LICKING 4.16
45123 HIGHLAND 4.54 43033 LICKING 9.72
45132 HIGHLAND 4.55 43055 LICKING 10.46
45155 HIGHLAND 6.5 43056 LICKING 9.15
43111 HOCKING 21.2 43058 LICKING 8.17
43127 HOCKING 4.93 43062 LICKING 5.09
43135 HOCKING 7.06 43068 LICKING 9.9
43138 HOCKING 4.44 43071 LICKING 8.06
43144 HOCKING 23.3 43076 LICKING 4.26
43149 HOCKING 8.24 43080 LICKING 6.52
43158 HOCKING 7.5 43740 LICKING 12.25
44611 HOLMES 5.22 43822 LICKING 8.39
44637 HOLMES 7.58 43310 LOGAN 6.98
44681 HOLMES 15.9 43311 LOGAN 6.09
44811 HURON 6.5 43318 LOGAN 4.97
44826 HURON 4.08 43333 LOGAN 5.62
44837 HURON 6.8 43343 LOGAN 6.68
44846 HURON 6.17 43357 LOGAN 6.36
44847 HURON 9.59 43360 LOGAN 5.17

86
ZIP GM RADON ZIP GM RADON
CODE COUNTY CONC. CODE COUNTY CONC.
44851 HURON 5.27 43412 LUCAS 4.21
44855 HURON 6.02 43504 LUCAS 4.79
43601 LUCAS 11.2 43716 MONROE 4.64
43603 LUCAS 5.3 43915 MONROE 5.31
43618 LUCAS 4.86 45242 MONTGOMERY 4.3
43697 LUCAS 6.28 45325 MONTGOMERY 5.01
43064 MADISON 4.74 45327 MONTGOMERY 5.26
43064 MADISON 8 45401 MONTGOMERY 5.06
43140 MADISON 4.43 45404 MONTGOMERY 6.61
43151 MADISON 5.43 45409 MONTGOMERY 4.36
43153 MADISON 5.44 45413 MONTGOMERY 5.62
43162 MADISON 6.68 45415 MONTGOMERY 4.34
44443 MAHONING 10.1 45422 MONTGOMERY 7.67
44503 MAHONING 7.02 45430 MONTGOMERY 4.22
44619 MAHONING 6.59 45431 MONTGOMERY 4.31
43301 MARION 4.74 45433 MONTGOMERY 11.91
43322 MARION 4.42 45434 MONTGOMERY 4.22
43332 MARION 8.1 45439 MONTGOMERY 4.99
43335 MARION 6.1 45454 MONTGOMERY 4.46
43341 MARION 4.82 45463 MONTGOMERY 5.1
43342 MARION 5.69 45475 MONTGOMERY 7.07
44230 MEDINA 9.3 45715 MORGAN 6.66
44258 MEDINA 6.05 43317 MORROW 7.35
44280 MEDINA 6.61 43325 MORROW 5.36
45310 MERCER 6.07 43350 MORROW 6.23
45822 MERCER 4.08 43735 MUSKINGUM 7.47
45826 MERCER 4.63 43746 MUSKINGUM 5.21
45860 MERCER 5.95 43802 MUSKINGUM 4.15
45862 MERCER 5.24 43822 MUSKINGUM 5.54
45308 MIAMI 6.36 43830 MUSKINGUM 4.4
45312 Miami 6.81 43412 OTTAWA 4.01
45317 MIAMI 5.65 43439 OTTAWA 4.5
45318 Miami 5.2 45851 PAULDING 5.16
45337 MIAMI 4.98 43076 PERRY 5.04
45339 MIAMI 6.42 43739 PERRY 6.3
45356 Miami 5.24 43103 PICKAWAY 7.96
45359 Miami 8.25 43113 PICKAWAY 7.3
45361 MIAMI 6.3 43117 PICKAWAY 5.14
45371 MIAMI 4.41 43137 PICKAWAY 12.9

87
ZIP GM RADON ZIP GM RADON
CODE COUNTY CONC. CODE COUNTY CONC.
45373 MIAMI 5.46 43145 PICKAWAY 5.9
45374 MIAMI 9.2 43146 PICKAWAY 6.21
45383 MIAMI 5.04 43164 PICKAWAY 5.41
45003 PREBLE 5.68 44707 STARK 8.34
45321 PREBLE 5.93 44708 STARK 4.16
45338 PREBLE 5.24 44709 STARK 4.51
45347 PREBLE 5.25 44710 STARK 5.03
45378 PREBLE 4.2 44711 STARK 6.19
45381 PREBLE 4.89 44718 STARK 4.12
45827 PUTNAM 4.83 44721 STARK 4.71
45844 PUTNAM 4.2 44735 STARK 13.6
45853 PUTNAM 6.1 44087 SUMMIT 5.18
45877 PUTNAM 5.76 44202 SUMMIT 17.4
45893 PUTNAM 7.8 44232 SUMMIT 4.9
44813 RICHLAND 7.8 44317 SUMMIT 21.9
44822 RICHLAND 8.76 44333 SUMMIT 4.17
44843 RICHLAND 11.25 43832 TUSCARAWAS 5.45
44865 RICHLAND 4.03 43837 TUSCARAWAS 15.78
44875 RICHLAND 4.42 44612 TUSCARAWAS 4.21
44903 RICHLAND 4.1 44656 TUSCARAWAS 5.78
44904 RICHLAND 5.05 44663 TUSCARAWAS 6.06
45601 ROSS 5.06 44671 TUSCARAWAS 6.9
45633 ROSS 15.9 44680 TUSCARAWAS 18.68
45647 ROSS 5.85 44681 TUSCARAWAS 4.13
45681 ROSS 15.43 44683 TUSCARAWAS 4.85
43469 SANDUSKY 4.24 43007 UNION 8.3
45684 SCIOTO 8.27 43045 UNION 4.32
44802 SENECA 4.47 43077 UNION 4.6
44809 SENECA 10.21 45894 VANWERT 5.9
44815 SENECA 4.85 44217 WAYNE 4.22
44853 SENECA 4.76 44667 WAYNE 4.49
44883 SENECA 4.73 44677 WAYNE 8.69
45302 SHELBY 5.16 44691 WAYNE 5.28
45306 SHELBY 5.18 43080 WOOD 7.2
45333 SHELBY 5.17 43565 WOOD 4.69
45363 SHELBY 4.5
45365 SHELBY 5
45845 SHELBY 6.03
44216 STARK 12.6
44626 STARK 7.86
44657 STARK 8.65
44662 STARK 5.94
44669 STARK 5.09

88
Table 10 Zip Codes with Geometric Mean of Radon Concentration greater
than 4 pCi/l (with Predicted Zip Codes)
COUNTY GM RADON COUNTY GM RADON
ZIPCODE NAME CONC. ZIPCODE NAME CONC.
45616 ADAMS 5.55 43901 BELMONT 4.04
45650 ADAMS 35.40 43906 BELMONT 4.45
45684 ADAMS 5.35 43917 BELMONT 4.14
45806 ALLEN 4.21 43933 BELMONT 7.24
45810 ALLEN 4.78 43937 BELMONT 4.62
45812 ALLEN 4.44 43940 BELMONT 6.2
45817 ALLEN 4.21 43977 BELMONT 5.50
45830 ALLEN 4.66 45171 BROWN 5.42
45833 ALLEN 4.79 45003 BUTLER 4.12
45844 ALLEN 4.12 45018 BUTLER 12.24
45850 ALLEN 10.42 45043 BUTLER 5.39
45868 ALLEN 4.96 45062 BUTLER 6.29
45877 ALLEN 4.77 45067 BUTLER 4.62
45887 ALLEN 4.46 45241 BUTLER 4.05
43014 ASHLAND 7.10 43945 CARROLL 4.52
44628 ASHLAND 6.14 43988 CARROLL 4.23
44638 ASHLAND 4.45 44423 CARROLL 4.21
44676 ASHLAND 4.21 44427 CARROLL 9.33
44822 ASHLAND 7.05 44621 CARROLL 4.19
44837 ASHLAND 4.21 44625 CARROLL 4.35
44840 ASHLAND 6.78 44639 CARROLL 7.95
44840 ASHLAND 6.78 44643 CARROLL 6.63
44843 ASHLAND 4.30 44644 CARROLL 4.54
44848 ASHLAND 8.61 44656 CARROLL 5.02
44859 ASHLAND 4.34 44657 CARROLL 6.40
44864 ASHLAND 8.59 44688 CARROLL 5.02
44878 ASHLAND 4.07 44695 CARROLL 4.25
44903 ASHLAND 4.38 44730 CARROLL 4.98
44068 ASHTABULA 4.14 43009 CHAMPAIGN 5.60
44088 ASHTABULA 20.87 43029 CHAMPAIGN 4.42
45306 AUGLAIZE 4.41 43040 CHAMPAIGN 4.03
45805 AUGLAIZE 4.44 43047 CHAMPAIGN 15.3
45806 AUGLAIZE 4.86 43060 CHAMPAIGN 4.58
45819 AUGLAIZE 21.5 43078 CHAMPAIGN 4.90
45845 AUGLAIZE 5.20 43140 CHAMPAIGN 4.54
45865 AUGLAIZE 5.15 43311 CHAMPAIGN 5.59
45871 AUGLAIZE 6.32 43318 CHAMPAIGN 5.68
45884 AUGLAIZE 6 43319 CHAMPAIGN 4.35
43719 BELMONT 7.64 43343 CHAMPAIGN 5.38

89
COUNTY GM RADON COUNTY GM RADON
ZIPCODE NAME CONC. ZIPCODE NAME CONC.
43357 CHAMPAIGN 7.80 43824 COSHOCTON 17.49
43360 CHAMPAIGN 4.58 43828 COSHOCTON 8.8
45317 CHAMPAIGN 5.27 43832 COSHOCTON 5.27
45344 CHAMPAIGN 6.22 43840 COSHOCTON 5.70
45365 CHAMPAIGN 5.30 43843 COSHOCTON 6.47
45389 CHAMPAIGN 9.72 43844 COSHOCTON 6.52
45502 CHAMPAIGN 4.92 43845 COSHOCTON 7.77
43044 CLARK 4.33 44637 COSHOCTON 6.55
43072 CLARK 5.73 44654 COSHOCTON 6.02
43078 CLARK 5.03 43335 CRAWFORD 4.10
45323 CLARK 5.69 44833 CRAWFORD 4.60
45324 CLARK 5.38 44860 CRAWFORD 7.3
45341 CLARK 12.43 44887 CRAWFORD 4.22
45344 CLARK 6.40 44195 CUYAHOGA 5
45372 CLARK 5.18 45303 DARKE 7.58
45387 CLARK 4.07 45304 DARKE 5.61
45424 CLARK 5.39 45308 DARKE 5.34
45501 CLARK 8.4 45318 DARKE 5.85
45503 CLARK 4.70 45321 DARKE 5.31
45505 CLARK 4.53 45331 DARKE 4.93
45506 CLARK 4.17 45337 DARKE 5.85
45138 CLINTON 24.82 45338 DARKE 5.35
45164 CLINTON 4.8 45346 DARKE 4.68
43920 COLUMBIANA 5.56 45347 DARKE 5.31
43930 COLUMBIANA 5.65 45351 DARKE 5.05
43932 COLUMBIANA 6.61 45352 DARKE 10.8
43945 COLUMBIANA 4.97 45363 DARKE 5.74
44432 COLUMBIANA 4.10 45380 DARKE 4.35
44441 COLUMBIANA 8.24 45382 DARKE 5.39
44443 COLUMBIANA 11.20 45388 DARKE 4.30
44455 COLUMBIANA 11.04 45390 DARKE 5.06
43006 COSHOCTON 6.59 45845 DARKE 5.21
43014 COSHOCTON 6.90 45846 DARKE 4.40
43749 COSHOCTON 4.47 45860 DARKE 5.02
43762 COSHOCTON 4.87 45883 DARKE 4.47
43802 COSHOCTON 5.04 43011 DELAWARE 5.34
43804 COSHOCTON 5.92 43013 DELAWARE 4.56
43811 COSHOCTON 9.24 43016 DELAWARE 8.60
43812 COSHOCTON 4.26 43017 DELAWARE 6.32

90
COUNTY GM RADON COUNTY GM RADON
ZIPCODE NAME CONC. ZIPCODE NAME CONC.
43822 COSHOCTON 6.06 43021 DELAWARE 4.01
43031 DELAWARE 5.03 43145 FAYETTE 4.84
43040 DELAWARE 4.20 45123 FAYETTE 4.19
43054 DELAWARE 5.11 45628 FAYETTE 4.50
43064 DELAWARE 5.43 43002 FRANKLIN 9.90
43065 DELAWARE 4.24 43004 FRANKLIN 5.95
43081 DELAWARE 11.07 43016 FRANKLIN 5.50
43235 DELAWARE 5.27 43017 FRANKLIN 6.01
43334 DELAWARE 4.05 43026 FRANKLIN 5.64
43342 DELAWARE 17.30 43031 FRANKLIN 5.14
43344 DELAWARE 4.66 43054 FRANKLIN 4.21
43356 DELAWARE 4.54 43064 FRANKLIN 5.94
44811 ERIE 4.23 43068 FRANKLIN 4.96
44814 ERIE 4.87 43080 FRANKLIN 9.81
44824 ERIE 7.90 43081 FRANKLIN 4.75
44826 ERIE 4.08 43082 FRANKLIN 4.87
44847 ERIE 8.03 43085 FRANKLIN 7.68
44857 ERIE 4.80 43103 FRANKLIN 7.51
43025 FAIRFIELD 5.41 43109 FRANKLIN 7.48
43062 FAIRFIELD 5.39 43110 FRANKLIN 7.01
43068 FAIRFIELD 6.63 43119 FRANKLIN 5.18
43102 FAIRFIELD 4.40 43123 FRANKLIN 5.12
43103 FAIRFIELD 6.76 43125 FRANKLIN 8.00
43107 FAIRFIELD 4.47 43126 FRANKLIN 9.61
43110 FAIRFIELD 9.71 43137 FRANKLIN 11.37
43112 FAIRFIELD 6.63 43140 FRANKLIN 5.87
43113 FAIRFIELD 6.00 43146 FRANKLIN 6.28
43130 FAIRFIELD 6.06 43147 FRANKLIN 6.10
43135 FAIRFIELD 5.30 43162 FRANKLIN 5.90
43136 FAIRFIELD 10.19 43202 FRANKLIN 4.34
43138 FAIRFIELD 4.60 43204 FRANKLIN 4.24
43147 FAIRFIELD 4.80 43205 FRANKLIN 4.74
43148 FAIRFIELD 4.39 43206 FRANKLIN 4.67
43149 FAIRFIELD 5.30 43207 FRANKLIN 6.87
43150 FAIRFIELD 5.90 43209 FRANKLIN 4.02
43154 FAIRFIELD 4.04 43213 FRANKLIN 6.15
43155 FAIRFIELD 4.40 43214 FRANKLIN 5.52
43106 FAYETTE 4.61 43216 FRANKLIN 17.9
43115 FAYETTE 4.80 43218 FRANKLIN 5.1

91
COUNTY GM RADON COUNTY GM RADON
ZIPCODE NAME CONC. ZIPCODE NAME CONC.
43142 FAYETTE 9.6 43219 FRANKLIN 6.46
43143 FAYETTE 4.88 43220 FRANKLIN 5.14
43221 FRANKLIN 6.00 45868 HANCOCK 5.32
43222 FRANKLIN 5.01 45881 HANCOCK 7.31
43224 FRANKLIN 5.15 45890 HANCOCK 9.97
43226 FRANKLIN 4.86 45810 HARDIN 4.47
43227 FRANKLIN 7.48 45814 HARDIN 4.65
43228 FRANKLIN 4.30 45817 HARDIN 4.99
43229 FRANKLIN 5.29 45835 HARDIN 14.00
43230 FRANKLIN 5.94 45841 HARDIN 4.95
43231 FRANKLIN 4.44 45843 HARDIN 4.37
43232 FRANKLIN 9.89 43901 HARRISON 6.69
43234 FRANKLIN 7.4 43907 HARRISON 5.02
43235 FRANKLIN 5.61 43917 HARRISON 4.11
43240 FRANKLIN 5.28 43950 HARRISON 4.01
43279 FRANKLIN 10.4 43988 HARRISON 16.99
43553 FULTON 4.77 44621 HARRISON 4.26
45301 GREENE 10.69 44683 HARRISON 4.22
45305 GREENE 4.21 44695 HARRISON 4.53
45323 GREENE 5.26 43548 HENRY 5.86
45341 GREENE 5.32 45123 HIGHLAND 4.54
45370 GREENE 4.07 45132 HIGHLAND 4.55
45385 GREENE 4.73 45155 HIGHLAND 6.5
45387 GREENE 4.21 43102 HOCKING 5.34
45420 GREENE 4.34 43107 HOCKING 4.05
45424 GREENE 4.56 43111 HOCKING 21.2
45430 GREENE 10.97 43127 HOCKING 4.93
45434 GREENE 4.66 43130 HOCKING 5.31
45435 GREENE 5.76 43135 HOCKING 7.06
45440 GREENE 4.38 43138 HOCKING 4.44
45502 GREENE 4.18 43144 HOCKING 23.3
43750 GUERNSEY 5.88 43149 HOCKING 8.24
43778 GUERNSEY 4.54 43155 HOCKING 4.93
44230 HAMILTON 4.1 43158 HOCKING 7.5
45111 HAMILTON 4.36 43006 HOLMES 6.25
45140 HAMILTON 4.10 43014 HOLMES 6.66
45174 HAMILTON 4.27 43804 HOLMES 5.78
45228 HAMILTON 4.20 43812 HOLMES 5.94
45263 HAMILTON 20.5 43824 HOLMES 6.22

92
COUNTY GM RADON COUNTY GM RADON
ZIPCODE NAME CONC. ZIPCODE NAME CONC.
45280 HAMILTON 4.99 43843 HOLMES 6.77
45817 HANCOCK 4.90 43844 HOLMES 6.29
45841 HANCOCK 5.90 44611 HOLMES 5.22
44624 HOLMES 4.23 44822 KNOX 6.93
44637 HOLMES 7.58 44842 KNOX 6.70
44681 HOLMES 15.90 43001 LICKING 6.23
44842 HOLMES 5.18 43004 LICKING 6.09
44811 HURON 6.50 43011 LICKING 4.69
44826 HURON 4.08 43023 LICKING 8.85
44837 HURON 6.80 43025 LICKING 5.90
44846 HURON 6.17 43027 LICKING 4.9
44847 HURON 9.59 43030 LICKING 10.19
44851 HURON 5.27 43031 LICKING 4.16
44855 HURON 6.02 43033 LICKING 9.72
44865 HURON 4.37 43046 LICKING 5.26
44878 HURON 4.45 43050 LICKING 5.71
43901 JEFFERSON 4.71 43054 LICKING 5.47
43907 JEFFERSON 4.46 43055 LICKING 10.46
43917 JEFFERSON 8.36 43056 LICKING 9.15
43925 JEFFERSON 10.5 43058 LICKING 8.17
43930 JEFFERSON 39.00 43062 LICKING 5.09
43932 JEFFERSON 15.99 43068 LICKING 9.90
43943 JEFFERSON 6.06 43071 LICKING 8.06
43945 JEFFERSON 5.49 43076 LICKING 4.26
43968 JEFFERSON 6.39 43080 LICKING 6.52
44615 JEFFERSON 4.01 43105 LICKING 5.11
43005 KNOX 7.83 43147 LICKING 6.13
43006 KNOX 12.49 43739 LICKING 5.27
43011 KNOX 4.99 43740 LICKING 12.25
43013 KNOX 4.51 43760 LICKING 4.85
43014 KNOX 6.00 43822 LICKING 8.39
43019 KNOX 6.42 43830 LICKING 5.76
43022 KNOX 7.01 43843 LICKING 6.46
43028 KNOX 12.05 43844 LICKING 6.40
43050 KNOX 5.26 43310 LOGAN 6.98
43055 KNOX 6.37 43311 LOGAN 6.09
43071 KNOX 6.17 43318 LOGAN 4.97
43074 KNOX 4.37 43326 LOGAN 4.02
43080 KNOX 5.86 43333 LOGAN 5.62

93
COUNTY GM RADON COUNTY GM RADON
ZIPCODE NAME CONC. ZIPCODE NAME CONC.
43822 KNOX 6.35 43343 LOGAN 6.68
43843 KNOX 6.58 43357 LOGAN 6.36
44628 KNOX 6.13 43360 LOGAN 5.17
44813 KNOX 6.15 45340 LOGAN 4.64
45365 LOGAN 5.15 44230 MEDINA 9.30
43412 LUCAS 4.21 44258 MEDINA 6.05
43504 LUCAS 4.79 44280 MEDINA 6.61
43601 LUCAS 11.2 45310 MERCER 6.07
43603 LUCAS 5.3 45822 MERCER 4.08
43618 LUCAS 4.86 45826 MERCER 4.63
43697 LUCAS 6.28 45845 MERCER 5.15
43002 MADISON 6.24 45860 MERCER 5.95
43026 MADISON 6.22 45862 MERCER 5.24
43029 MADISON 4.46 45865 MERCER 5.00
43044 MADISON 4.57 45869 MERCER 4.69
43045 MADISON 4.46 45885 MERCER 4.52
43064 MADISON 8.00 45887 MERCER 4.28
43064 MADISON 4.74 45894 MERCER 4.16
43106 MADISON 4.20 43072 MIAMI 5.25
43119 MADISON 6.00 45304 MIAMI 5.83
43123 MADISON 5.68 45308 MIAMI 6.36
43140 MADISON 4.43 45312 MIAMI 6.81
43151 MADISON 5.43 45317 MIAMI 5.65
43153 MADISON 5.44 45318 MIAMI 5.20
43162 MADISON 6.68 45333 MIAMI 6.00
45368 MADISON 4.15 45337 MIAMI 4.98
45369 MADISON 4.42 45339 MIAMI 6.42
44443 MAHONING 10.10 45344 MIAMI 5.51
44503 MAHONING 7.02 45356 MIAMI 5.24
44619 MAHONING 6.59 45359 MIAMI 8.25
43003 MARION 4.17 45361 MIAMI 6.3
43301 MARION 4.74 45363 MIAMI 5.87
43315 MARION 4.18 45365 MIAMI 5.38
43320 MARION 4.27 45371 MIAMI 4.41
43322 MARION 4.42 45373 MIAMI 5.46
43332 MARION 8.10 45374 MIAMI 9.2
43335 MARION 6.10 45377 MIAMI 5.33
43341 MARION 4.82 45380 MIAMI 5.91
43342 MARION 5.69 45383 MIAMI 5.04

94
COUNTY GM RADON COUNTY GM RADON
ZIPCODE NAME CONC. ZIPCODE NAME CONC.
43344 MARION 4.66 45424 MIAMI 5.33
43351 MARION 4.04 43716 MONROE 4.64
44820 MARION 4.07 43915 MONROE 5.31
44833 MARION 4.30 45042 MONTGOMERY 4.08
45843 MARION 4.08 45242 MONTGOMERY 4.3
45304 MONTGOMERY 5.24 43076 MUSKINGUM 5.11
45311 MONTGOMERY 4.18 43735 MUSKINGUM 7.47
45324 MONTGOMERY 5.17 43739 MUSKINGUM 5.04
45325 MONTGOMERY 5.01 43746 MUSKINGUM 5.21
45327 MONTGOMERY 5.26 43749 MUSKINGUM 4.21
45337 MONTGOMERY 5.06 43802 MUSKINGUM 4.15
45338 MONTGOMERY 5.04 43811 MUSKINGUM 5.26
45341 MONTGOMERY 5.26 43812 MUSKINGUM 5.37
45344 MONTGOMERY 5.44 43822 MUSKINGUM 5.54
45371 MONTGOMERY 4.82 43830 MUSKINGUM 4.40
45381 MONTGOMERY 4.65 43412 OTTAWA 4.01
45383 MONTGOMERY 4.87 43439 OTTAWA 4.5
45401 MONTGOMERY 5.06 45851 PAULDING 5.16
45404 MONTGOMERY 6.61 43025 PERRY 5.47
45409 MONTGOMERY 4.36 43046 PERRY 5.02
45413 MONTGOMERY 5.62 43076 PERRY 5.04
45415 MONTGOMERY 4.34 43107 PERRY 4.02
45422 MONTGOMERY 7.67 43148 PERRY 4.95
45430 MONTGOMERY 4.22 43150 PERRY 4.30
45431 MONTGOMERY 4.31 43739 PERRY 6.30
45433 MONTGOMERY 11.91 43102 PICKAWAY 5.34
45434 MONTGOMERY 4.22 43103 PICKAWAY 7.96
45439 MONTGOMERY 4.99 43110 PICKAWAY 7.25
45454 MONTGOMERY 4.46 43113 PICKAWAY 7.30
45463 MONTGOMERY 5.1 43115 PICKAWAY 5.19
45475 MONTGOMERY 7.07 43117 PICKAWAY 5.14
45715 MORGAN 6.66 43125 PICKAWAY 7.42
43011 MORROW 4.12 43135 PICKAWAY 5.00
43019 MORROW 4.74 43137 PICKAWAY 12.90
43314 MORROW 4.12 43140 PICKAWAY 5.72
43317 MORROW 7.35 43143 PICKAWAY 5.67
43325 MORROW 5.36 43145 PICKAWAY 5.90
43335 MORROW 4.25 43146 PICKAWAY 6.21
43350 MORROW 6.23 43154 PICKAWAY 6.10

95
COUNTY GM RADON COUNTY GM RADON
ZIPCODE NAME CONC. ZIPCODE NAME CONC.
44813 MORROW 5.15 43160 PICKAWAY 5.33
44833 MORROW 4.19 43164 PICKAWAY 5.41
44903 MORROW 4.34 45601 PICKAWAY 5.27
44904 MORROW 4.39 45644 PICKAWAY 4.88
43055 MUSKINGUM 5.92 45003 PREBLE 5.68
43056 MUSKINGUM 4.97 45042 PREBLE 4.01
45304 PREBLE 5.33 45681 ROSS 15.43
45309 PREBLE 5.26 43469 SANDUSKY 4.24
45321 PREBLE 5.93 45616 SCIOTO 4.29
45325 PREBLE 4.61 45684 SCIOTO 8.27
45327 PREBLE 4.33 44802 SENECA 4.47
45332 PREBLE 5.29 44809 SENECA 10.21
45338 PREBLE 5.24 44815 SENECA 4.85
45345 PREBLE 4.56 44853 SENECA 4.76
45346 PREBLE 5.32 44883 SENECA 4.73
45347 PREBLE 5.25 43343 SHELBY 4.62
45378 PREBLE 4.2 45302 SHELBY 5.16
45381 PREBLE 4.89 45306 SHELBY 5.18
45807 PUTNAM 4.41 45317 SHELBY 5.22
45827 PUTNAM 4.83 45318 SHELBY 5.96
45840 PUTNAM 4.10 45326 SHELBY 5.37
45844 PUTNAM 4.20 45333 SHELBY 5.17
45853 Putnam 6.1 45356 SHELBY 5.84
45868 PUTNAM 4.78 45363 SHELBY 4.50
45877 PUTNAM 5.76 45363 SHELBY 5.00
45893 PUTNAM 7.80 45365 SHELBY 5.00
43019 RICHLAND 5.30 45380 SHELBY 5.55
44813 RICHLAND 7.80 45388 SHELBY 5.46
44822 RICHLAND 8.76 45845 SHELBY 6.03
44837 RICHLAND 4.08 45860 SHELBY 5.20
44843 RICHLAND 11.25 45865 SHELBY 5.24
44859 RICHLAND 4.09 45869 SHELBY 5.09
44864 RICHLAND 6.38 45871 SHELBY 5.05
44865 RICHLAND 4.03 44216 STARK 12.60
44875 RICHLAND 4.42 44612 STARK 5.03
44903 RICHLAND 4.10 44626 STARK 7.86
44904 RICHLAND 5.05 44645 STARK 4.08
43113 ROSS 5.19 44657 STARK 8.65
43135 ROSS 4.28 44662 STARK 5.94

96
COUNTY GM RADON COUNTY GM RADON
ZIPCODE NAME CONC. ZIPCODE NAME CONC.
43145 ROSS 4.67 44669 STARK 5.09
43160 ROSS 4.62 44680 STARK 4.76
43164 ROSS 5.13 44707 STARK 8.34
45123 ROSS 4.43 44708 STARK 4.16
45601 ROSS 5.06 44709 STARK 4.51
45633 ROSS 15.9 44710 STARK 5.03
45647 ROSS 5.85 44711 STARK 6.19
44718 STARK 4.12 44611 WAYNE 4.15
44721 STARK 4.71 44614 WAYNE 4.17
44735 STARK 13.6 44666 WAYNE 4.10
44087 SUMMIT 5.18 44667 WAYNE 4.49
44202 SUMMIT 17.40 44677 WAYNE 8.69
44232 Summit 4.9 44691 WAYNE 5.28
44333 SUMMIT 4.17 44840 WAYNE 4.08
44614 SUMMIT 4.02 44842 WAYNE 4.24
44645 SUMMIT 4.09 43080 WOOD 7.2
43824 TUSCARAWAS 5.84 43565 WOOD 4.69
43832 TUSCARAWAS 5.45 43332 WYANDOT 4.07
43837 TUSCARAWAS 15.78
44608 TUSCARAWAS 4.30
44612 TUSCARAWAS 4.21
44620 TUSCARAWAS 4.95
44626 TUSCARAWAS 4.89
44643 TUSCARAWAS 5.26
44654 TUSCARAWAS 4.87
44656 TUSCARAWAS 5.78
44663 TUSCARAWAS 6.06
44671 TUSCARAWAS 6.9
44675 TUSCARAWAS 4.41
44680 TUSCARAWAS 18.68
44681 TUSCARAWAS 4.13
44683 TUSCARAWAS 4.85
43002 UNION 5.56
43007 UNION 8.3
43015 UNION 4.72
43017 UNION 5.24
43044 UNION 4.26
43045 UNION 4.32
43061 UNION 4.38
43065 UNION 5.09
43066 UNION 4.29
43077 UNION 4.6
43302 UNION 4.48
43342 UNION 4.60
45887 VAN WERT 4.38
45894 VAN WERT 5.90
44217 WAYNE 4.22

97
Table 11 Zip Codes having Geometric Mean of Radon Concentrations
greater than 8pCi/l
ZIP GM RADON ZIP GM RADON
CODE COUNTY CONC. CODE COUNTY CONC.
45650 ADAMS 35.4 44681 HOLMES 15.9
45850 ALLEN 10.42 44847 HURON 9.59
44864 ASHLAND 8.59 43917 JEFFERSON 8.36
44848 ASHLAND 8.61 43925 JEFFERSON 10.5
44088 ASHTABULA 20.87 43932 JEFFERSON 15.99
45819 AUGLAIZE 21.5 43930 JEFFERSON 39
45018 BUTLER 12.24 43028 KNOX 12.05
44427 CARROLL 9.33 43006 KNOX 12.49
45389 CHAMPAIGN 9.72 43071 LICKING 8.06
43047 CHAMPAIGN 15.3 43058 LICKING 8.17
45501 CLARK 8.4 43822 LICKING 8.39
45341 CLARK 12.43 43023 LICKING 8.85
45138 CLINTON 24.82 43056 LICKING 9.15
44441 COLUMBIANA 8.24 43033 LICKING 9.72
44455 COLUMBIANA 11.04 43068 LICKING 9.9
44443 COLUMBIANA 11.2 43030 LICKING 10.19
43828 COSHOCTON 8.8 43055 LICKING 10.46
43811 COSHOCTON 9.24 43740 LICKING 12.25
43824 COSHOCTON 17.49 43601 Lucas 11.2
45352 DARKE 10.8 44443 MAHONING 10.1
43016 DELAWARE 8.6 43332 MARION 8.1
43081 DELAWARE 11.07 44230 MEDINA 9.3
43342 DELAWARE 17.3 45359 MIAMI 8.25
44847 ERIE 8.03 45374 MIAMI 9.2
43110 FAIRFIELD 9.71 45433 MONTGOMERY 11.91
43136 FAIRFIELD 10.19 43137 PICKAWAY 12.9
43142 FAYETTE 9.6 44822 RICHLAND 8.76
43126 FRANKLIN 9.61 44843 RICHLAND 11.25
43080 FRANKLIN 9.81 45681 ROSS 15.43
43232 FRANKLIN 9.89 45633 ROSS 15.9
43002 FRANKLIN 9.9 45684 SCIOTO 8.27
43279 FRANKLIN 10.4 44809 SENECA 10.21
43137 FRANKLIN 11.37 44707 STARK 8.34
43216 FRANKLIN 17.9 44657 STARK 8.65
45301 GREENE 10.69 44216 STARK 12.6
45430 GREENE 10.97 44735 STARK 13.6
45263 HAMILTON 20.5 44202 SUMMIT 17.4
45890 HANCOCK 9.97 44317 SUMMIT 21.9
45835 HARDIN 14 43837 TUSCARAWAS 15.78
43988 HARRISON 16.99 44680 TUSCARAWAS 18.68
43149 HOCKING 8.24 43007 UNION 8.3
43111 HOCKING 21.2 44677 WAYNE 8.69
43144 HOCKING 23.3

98
Table 12 Zip Codes having Geometric Mean of Radon Concentrations
greater than 20pCi/l

ZIP GM RADON
CODE COUNTY CONC.
45650 ADAMS 35.4
44088 ASHTABULA 20.87
45819 AUGLAIZE 21.5
45138 CLINTON 24.82
45263 HAMILTON 20.5
43111 HOCKING 21.2
43144 HOCKING 23.3
43930 JEFFERSON 39
44317 SUMMIT 21.9

99
7. Conclusions

Prediction maps were created using the training data set for each of the five

interpolation techniques discussed in this thesis and projected values were estimated

for the test data set. Statistical parameters (error values) for all the interpolation

techniques were evaluated and the prediction maps generated from these techniques

were compared to the soil uranium concentration map. It was shown that the ordinary

kriging interpolation technique can be used tentatively for predicting the radon

concentrations for unmeasured zip codes.

The ordinary kriging technique was chosen for this study and the geometric

means of radon concentrations were evaluated for unmeasured zip codes and a new

data set was thus created, which may be appended to the existing radon concentration

data set. Use of this technique also led to the inference that inclusive of the zip codes

not shown in the Ohio zip codes shape file, a total of 688 zip codes had a radon

concentration higher than 4.0 pCi/l, 85 zip codes with radon concentrations higher

than 8 pCi/l and 9 zip codes with radon concentrations higher than 20pCi/l. These zip

codes are given in Tables 10, 11, 12.

100
8. References

1. Krivoruchko Konstantin, Introduction to Modeling Spatial Processes Using

Geostatistical analyst.

2. Johnston Kevin, Jay M. Ver Hoef, Krivoruchko Konstantin, and Lucas Neil,

Using ArcGIS Geostatistical analyst.

3. Slotnick Melissa, Spatial and Temporal Variation of Arsenic Concentrations

in Oakland County, Michigan Groundwater, SNRE 501, Section 039,

December 13, 2002.

4. Jie Y. He, and Jia Xudong, ArcGIS Geostatistical analyst Application in

Assessment of MTBE Contamination,

5. N. Saby, D. Arrouays, L. Boulonne, C. Jolivet, A. Pochot, Geosatistical

assessment of PB in soil around Paris,France, Science of the Total

Environment 367 (2006) 212221.

6. Liu Xingmei, Wu Jianjun, Xu Jianming, Characterizing the risk assessment

of heavy metals and sampling uncertainty analysis in paddy field by

geostatistics and GIS Environmental Pollution 141 (2006) 257e264.

7. Pardo-Iguzquiza Eulogio, Comparison of GeostatisticalMethods for

Estimating the Areal Average Climatological Rainfall Mean using Data on

Precipitation and Topography, Int. J. Climatol. 18: (1998) 10311047.

101
8. M.G. Mardikis, D.P. Kalivas and V.J. Kollias,Comparison of Interpolation

Methods for the Prediction of Reference EvapotranspirationAn Application

in Greece, Water Resources Management (2005) 19: 251278.

9. Ebrahim Hosseini Chegini, Mohammad Hossein Mahdian, Sima Rahimy

Bandarabadi Mohammad Mahdavi, Survey on Application of

GeostatisticalMethods for Estimation of Rainfall in Arid and Semiarid

Regions in South West Of Iran, Soil Conservation Watershed Management

Research Center, Tehran, Iran.

10. J. A. Harrell and A. Kumar, "Multivariate Stepwise Regression Analysis of

Indoor Radon Data from Ohio, U.S.A.", Vol. 5, pp. 409-420, Journal of

Official Statistics, 1989.

11. J. A. Harrell, M. E. Belsito, and A. Kumar, "Radon Hazards Associated with

the Ohio Shale", Vol. 18, No. 1, pp. 17-26, Environmental Geology and Water

Sciences. 1991.

12. A. Heydinger, A. Kumar, and J. Harrell, "An Indoor Radon Information

System", 6 (4), pp. 194-201, Env. Software, 1991.

13. J. Harrell, J. P. McKenna, and A. Kumar, "Geological Controls on Indoor

Radon in Ohio", Investigation No. 144, Division of Geological Survey, Dept.

of Natural Resources, State of Ohio, 36 pp, 1993.

14. A. Kumar, A. Sud, and A. Heydinger, "Application of ORACLE 7.3 Database

Management System for the development of an Environmental database",

Environmental Progress, 17(3), pp. F11-F14, 1998.

102
15. A. Kumar, J. Harrell, and A. Heydinger, Ohio Goes Online to Combat Indoor

Radon, EM (by A&WMA), pp. 10-12, Feb. 2001.

16. S. Ojha, S.J. Thomas, and A. Kumar, Experience in Integrating Geographical

Information Systems (GIS) to an Indoor Radon Database, Environmental

Progress, 20(3), O7-O10, 2001.

17. A. Joshi, G. K. Manne, and A. Kumar, Management of Ohios Radon Data

with MS Access/SQL Server 7.0 , Environmental Progress, 21(4), D8-D12,

2002.

18. A. Kumar, A. Tandale, R.S. Kalapati, and S. Ghose, Management of Radon

Mitigation Data in the State of Ohio, Environmental Progress, 22(3), O5-

O10, 2003.

19. Ashok Kumar, Andrew Heydinger, James A. Harrell, Development of an

Indoor Radon Information System for Ohio and Its Application in the Study of

the Geology of Radon in Ohio, Toledo, OH, November, 1990.

20. Sud Anupam, Update and Analysis of a Residential Radon Database for the

State of Ohio, 1998.

21. Ashok Kumar, Charanya Vardarajan, Management of Radon Data in the

State of Ohio, U.S.A, Toledo, OH.

22. A. Kumar and C. Varadarajan, "Development of Ohio Radon Information

System ", Proceedings of Environmental Data Analysis - Assessing Health

and Environmental Impacts, Developing Policy, and Achieving Regulatory

Compliance Conference, Air & Waste Management Association, October 4-5,

2005, Oak Brook, IL.

103
23. Duval, J.S., 1985, Aerial Radiometric Contour Maps of Ohio: US Geological

Survey, Geophysical Investigations Map GP-968.

104

You might also like