Air Force Institute of Technology

AFIT Scholar

Theses and Dissertations Student Graduate Works

3-19-2009

Modeling and Simulation Architecture for Studying Doppler-Based

Radar with Complex Environments
Nicholas J. Amato

Follow this and additional works at: https://scholar.afit.edu/etd

Part of the Multi-Vehicle Systems and Air Traffic Control Commons, and the Navigation, Guidance,
Control and Dynamics Commons

Recommended Citation
Amato, Nicholas J., "Modeling and Simulation Architecture for Studying Doppler-Based Radar with
Complex Environments" (2009). Theses and Dissertations. 2522.
https://scholar.afit.edu/etd/2522

This Thesis is brought to you for free and open access by the Student Graduate Works at AFIT Scholar. It has been
accepted for inclusion in Theses and Dissertations by an authorized administrator of AFIT Scholar. For more
information, please contact [email protected].
 Modeling and Simulation Architecture
for Studying Doppler-Based Radar
with Complex Environments

THESIS

Nicholas J. Amato, Captain, USAF

AFIT/GE/ENG/09-02

DEPARTMENT OF THE AIR FORCE

AIR UNIVERSITY

AIR FORCE INSTITUTE OF TECHNOLOGY

Wright-Patterson Air Force Base, Ohio

APPROVED FOR PUBLIC RELEASE; DISTRIBUTION UNLIMITED.

The views expressed in this thesis are those of the author and do not reflect the
official policy or position of the United States Air Force, Department of Defense, or
the United States Government.
AFIT/GE/ENG/09-02

Modeling and Simulation Architecture

for Studying Doppler-Based Radar
with Complex Environments

THESIS

Presented to the Faculty

Department of Electrical and Computer Engineering
Graduate School of Engineering and Management
Air Force Institute of Technology
Air University
Air Education and Training Command
In Partial Fulfillment of the Requirements for the
Degree of Master of Science in Electrical Engineering

Nicholas J. Amato, B.S.E.E.

Captain, USAF

March 2009

APPROVED FOR PUBLIC RELEASE; DISTRIBUTION UNLIMITED.

AFIT/GE/ENG/09-02

Modeling and Simulation Architecture

for Studying Doppler-Based Radar
with Complex Environments

Nicholas J. Amato, B.S.E.E.

Captain, USAF

Approved:

/signed/ 19 Mar 2009

Maj M.A. Saville, PhD (Chairman) date

/signed/ 19 Mar 2009

Maj M.J. Mendenhall, PhD (Member) date

/signed/ 19 Mar 2009

Dr. M.C. Fickus, PhD (Member) date

AFIT/GE/ENG/09-02

Abstract

This research effort develops a hybrid large-scale modeling and simulation frame-
work that defines the requirements for a program to evaluate radar-aircraft-turbine-
clutter interactions. Wind turbines and other moving structures can interfere with
a radar’s ability to detect moving aircraft because radar returns from turbines are
comparable to those from slow flying aircraft. This interference can lead to aircraft
collisions or crashes, reducing the safety for air traffic.

Two radar applications, INSSITE and IMOM, were investigated to determine

which of the subsystems, in the proposed architecture, are currently available and
which need additional development. Current radar applications either delve too deep
into details, requiring years to process, or too shallow, ignoring the Doppler effect and
assuming a static scattering value. Engineering-level radar, radiation, propagation,
and scattering models are already developed. However, engagement-level stochastic
scattering, amplitude and phase, data aren’t available. The hybrid modeling and
simulation architecture could be realized once stochastic RCS models are developed.

iv
 Acknowledgements
I’d like to thank everyone that helped me in many ways great and small.

I’d like to specifically thank several people. Maj. Saville for his guidance and
encouragement. Maj. Mendenhall and Dr. Temple for the hardware necessary for
this research. Capt. Brand for his support and for being my goto guy for everything
ARSR-4 related. Lt. Hemperly and Mr. Peltier for their assistance getting IMOM
up and running. Mr. Gloekler, Mr. Shargo, and Mr. Lindquist for their help on
operating INSSITE. Most importantly, thank you to my family and friends for all
your support, encouragement, and putting up with me.

Nicholas J. Amato

v
 Table of Contents
Page
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv

Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v

Table of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vi

List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix

List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi

List of Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xii

I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Problem Statement . . . . . . . . . . . . . . . . . . . . . 1
1.3 Proposed Solution . . . . . . . . . . . . . . . . . . . . . 3
1.4 Sponsors . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.4.1 84 RADES . . . . . . . . . . . . . . . . . . . . . 3
1.4.2 NASIC . . . . . . . . . . . . . . . . . . . . . . . 3
1.4.3 453rd EWS . . . . . . . . . . . . . . . . . . . . 3

II. Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2.1 Radar Basics . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1.1 Radar Range Equation . . . . . . . . . . . . . . 5
2.1.2 Primary vs Secondary Surveillance Radar . . . . 5
2.2 Wind Turbine Physics . . . . . . . . . . . . . . . . . . . 6
2.2.1 Line of Sight . . . . . . . . . . . . . . . . . . . . 6
2.2.2 Radar Cross-Section . . . . . . . . . . . . . . . 7
2.2.3 Doppler Effect . . . . . . . . . . . . . . . . . . . 9
2.3 Wind Turbine Effects on Radar . . . . . . . . . . . . . . 12
2.3.1 Shadow Zone . . . . . . . . . . . . . . . . . . . 13
2.3.2 Clutter . . . . . . . . . . . . . . . . . . . . . . . 14
2.3.3 Raised Threshold . . . . . . . . . . . . . . . . . 15
2.3.4 False Moving Targets . . . . . . . . . . . . . . . 15
2.3.5 Tracking . . . . . . . . . . . . . . . . . . . . . . 19
2.3.6 Receiver Saturation . . . . . . . . . . . . . . . . 19
2.4 Modeling Wind Turbine RCS . . . . . . . . . . . . . . . 20
2.5 Large Scale Modeling and Simulation . . . . . . . . . . . 21
2.5.1 How Large is Large? . . . . . . . . . . . . . . . 21

vi
 Page
2.5.2 Who Uses Large-Scale M&S? . . . . . . . . . . 22
2.5.3 M&S Hierarchy . . . . . . . . . . . . . . . . . . 22

III. Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.2 Evaluating Complexity . . . . . . . . . . . . . . . . . . . 26
3.2.1 Physics-Level . . . . . . . . . . . . . . . . . . . 26
3.2.2 Engineering-Level . . . . . . . . . . . . . . . . . 27
3.2.3 Engagement-Level . . . . . . . . . . . . . . . . . 27
3.3 Architecture Development . . . . . . . . . . . . . . . . . 28
3.4 Existing Programs . . . . . . . . . . . . . . . . . . . . . 28
3.5 Experimental Simulations . . . . . . . . . . . . . . . . . 29
3.5.1 IMOM Experiments . . . . . . . . . . . . . . . 29
3.5.2 INSSITE Experiments . . . . . . . . . . . . . . 30
3.5.3 Underlying Mathematics . . . . . . . . . . . . . 31
3.6 Analyzing the Architecture . . . . . . . . . . . . . . . . 33

IV. Radar-based Modeling and Simulation . . . . . . . . . . . . . . . 34

4.1 Connection Diagrams . . . . . . . . . . . . . . . . . . . . 34
4.1.1 Mission-Level . . . . . . . . . . . . . . . . . . . 34
4.1.2 Engagement-Level . . . . . . . . . . . . . . . . . 36
4.1.3 Engineering-Level . . . . . . . . . . . . . . . . . 37
4.1.4 Physics-Level . . . . . . . . . . . . . . . . . . . 39
4.2 Features . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.2.1 Display . . . . . . . . . . . . . . . . . . . . . . . 40
4.2.2 Radar . . . . . . . . . . . . . . . . . . . . . . . 40
4.2.3 Target . . . . . . . . . . . . . . . . . . . . . . . 42
4.2.4 EM - Propagation . . . . . . . . . . . . . . . . . 43
4.3 Understanding Computational Requirements . . . . . . . 44
4.4 Rationale behind Tip Speed - Wind States . . . . . . . . 46

V. IMOM and INSSITE Experiments . . . . . . . . . . . . . . . . . 48

5.1 IMOM . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
5.1.1 Overview . . . . . . . . . . . . . . . . . . . . . 48
5.1.2 Experimental Results . . . . . . . . . . . . . . . 48
5.1.3 Mission-Level . . . . . . . . . . . . . . . . . . . 52
5.1.4 Engagement-Level . . . . . . . . . . . . . . . . . 52
5.1.5 Engineering-Level . . . . . . . . . . . . . . . . . 52
5.1.6 Physics-Level . . . . . . . . . . . . . . . . . . . 53
5.1.7 Limitations . . . . . . . . . . . . . . . . . . . . 53

vii
 Page
5.2 INNSITE . . . . . . . . . . . . . . . . . . . . . . . . . . 53
5.2.1 Overview . . . . . . . . . . . . . . . . . . . . . 53
5.2.2 Experimental Results . . . . . . . . . . . . . . . 54
5.2.3 Mission-Level . . . . . . . . . . . . . . . . . . . 57
5.2.4 Engagement-Level . . . . . . . . . . . . . . . . . 57
5.2.5 Engineering-Level . . . . . . . . . . . . . . . . . 57
5.2.6 Physics-Level . . . . . . . . . . . . . . . . . . . 57
5.2.7 Limitations . . . . . . . . . . . . . . . . . . . . 57

VI. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
6.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . 59
6.2 Additional Research . . . . . . . . . . . . . . . . . . . . 60

Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

viii
 List of Figures
Figure Page

1.1. Typical Wind Turbine . . . . . . . . . . . . . . . . . . . . . . . 2

2.1. 2-D Radar Coverage Slice . . . . . . . . . . . . . . . . . . . . . 7
2.2. Measured vs Simulated RCS . . . . . . . . . . . . . . . . . . . 8
2.3. Turbine Power vs Wind Speed . . . . . . . . . . . . . . . . . . 10
2.4. Tip Speed vs Wind Speed . . . . . . . . . . . . . . . . . . . . . 11
2.5. Turbine Doppler . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.6. Impact of Angle on Doppler . . . . . . . . . . . . . . . . . . . . 12
2.7. Terrain Masking Shadow Zones . . . . . . . . . . . . . . . . . . 15
2.8. Clutter vs Resolution Cells . . . . . . . . . . . . . . . . . . . . 16
2.9. Clutter Maps . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.10. Turbine Blades and Aircraft . . . . . . . . . . . . . . . . . . . 18
2.11. Maximum Doppler vs Angle - Wind States . . . . . . . . . . . 21
2.12. RCS Magnitude Calculation . . . . . . . . . . . . . . . . . . . 22
2.13. Modeling and Simulation Pyramid . . . . . . . . . . . . . . . . 23
2.14. Army Military Planning Spectrum . . . . . . . . . . . . . . . . 24
3.1. Development Flowchart . . . . . . . . . . . . . . . . . . . . . . 25
3.2. RCS Complexity . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.3. Statistical RCS . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.4. Range Ring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.1. Mission-Level Simulation . . . . . . . . . . . . . . . . . . . . . 35
4.2. FAA - OE/AAA Map . . . . . . . . . . . . . . . . . . . . . . . 36
4.3. Engagement-Level Simulation . . . . . . . . . . . . . . . . . . . 37
4.4. Radar Engineering-Level Model . . . . . . . . . . . . . . . . . . 38
4.5. Physics-Level EM Wave Scattering . . . . . . . . . . . . . . . . 39
4.6. ARSR-4 Processor . . . . . . . . . . . . . . . . . . . . . . . . . 42

ix
Figure Page
4.7. Discrete set of turbine positions . . . . . . . . . . . . . . . . . 43
4.8. LOS Analysis Tool . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.9. Three Propagation Models . . . . . . . . . . . . . . . . . . . . 45
4.10. Physics-Level EM Wave Scattering . . . . . . . . . . . . . . . . 47
5.1. IMOM LOS . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
5.2. IMOM Aircraft Detection . . . . . . . . . . . . . . . . . . . . . 50
5.3. IMOM Turbine Detection . . . . . . . . . . . . . . . . . . . . . 50
5.4. Simulated Turbine - DTED Level 2 . . . . . . . . . . . . . . . 51
5.5. Simulated Turbine - DTED Level 1 . . . . . . . . . . . . . . . 52
5.6. INSSITE - Terrain . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.7. INSSITE - Turbine Convergence . . . . . . . . . . . . . . . . . 55
5.8. INSSITE - Turbine + Terrain . . . . . . . . . . . . . . . . . . . 55
5.9. INSSITE - Aircraft + Terrain . . . . . . . . . . . . . . . . . . . 56
5.10. INSSITE - Aircraft + Turbine + Terrain . . . . . . . . . . . . 56

x
 List of Tables
Table Page

2.1. Order of significance for PSR impacts, 1 is highest [13] . . . . . 13

2.2. L-Band rule-of-thumb Shadow Length . . . . . . . . . . . . . . 14
2.3. L-Band (1 GHz) rule-of-thumb RCS values [14] . . . . . . . . . 17
2.4. Proposed Wind States [8]. . . . . . . . . . . . . . . . . . . . . . 20
4.1. Sample radar and turbine parameters . . . . . . . . . . . . . . 44

xi
 List of Abbreviations
Abbreviation Page

AD Air Defense . . . . . . . . . . . . . . . . . . . . . . . . . . 1
ATC Air Traffic Control . . . . . . . . . . . . . . . . . . . . . . 1
DoD Department of Defense . . . . . . . . . . . . . . . . . . . . 1
RCS Radar Cross Section . . . . . . . . . . . . . . . . . . . . . 2
RADES Radar Evaluation Squadron . . . . . . . . . . . . . . . . . 3
NASIC National Air and Space Intelligence Center . . . . . . . . . 3
INSSITE Interactive Sensor Simulator for Terrain Environments . . 3
EWS Electronic Warfare Squadron . . . . . . . . . . . . . . . . 3
IO Information Operations . . . . . . . . . . . . . . . . . . . 3
DTED Digital Terrain Elevation Data . . . . . . . . . . . . . . . 4
IMOM Improved Many-On-Many . . . . . . . . . . . . . . . . . . 4
PSR Primary Surveillance Radar . . . . . . . . . . . . . . . . . 5
SSR Secondary Surveillance Radar . . . . . . . . . . . . . . . . 6
LOS Line-of-Sight . . . . . . . . . . . . . . . . . . . . . . . . . 6
CEM Computational Electromagnetic . . . . . . . . . . . . . . . 7
AFRL Air Force Research Laboratory . . . . . . . . . . . . . . . 8
RF Radio Frequency . . . . . . . . . . . . . . . . . . . . . . . 9
TNO Netherlands Organisation for Applied Scientific Research . 13
EM Electromagnetic . . . . . . . . . . . . . . . . . . . . . . . 14
MTI Moving Target Indicator . . . . . . . . . . . . . . . . . . . 15
PPI Plan Position Indicator . . . . . . . . . . . . . . . . . . . 17
ADT Advanced digital tracker . . . . . . . . . . . . . . . . . . . 17
NAIZ Non Auto-Initiation Zone . . . . . . . . . . . . . . . . . . 19
AGC Automatic Gain Control . . . . . . . . . . . . . . . . . . . 19
M&S Modeling and Simulation . . . . . . . . . . . . . . . . . . 21

xii
Abbreviation Page
EW Electronic Warfare . . . . . . . . . . . . . . . . . . . . . . 29
AGL Above Ground Level . . . . . . . . . . . . . . . . . . . . . 29
EOB Electronic Order of Battle . . . . . . . . . . . . . . . . . . 48
MSL Mean-Sea Level . . . . . . . . . . . . . . . . . . . . . . . . 48
TIREM Terrain Integrated Rough Earth Model . . . . . . . . . . . 48
IADS Integrated Air Defense System . . . . . . . . . . . . . . . 52
VTRPE Variable Terrain Radio Parabolic Equation . . . . . . . . . 53
APM Advanced Propagation Model . . . . . . . . . . . . . . . . 53
SBR Shooting and Bouncing Ray . . . . . . . . . . . . . . . . . 54

xiii
 Modeling and Simulation Architecture
for Studying Doppler-Based Radar
with Complex Environments

I. Introduction
1.1 Background

The United States Department of Energy set a goal for wind to produce 20% of
the US electric generating capacity by 2030 [33]. In 2007, the US cumulative turbine
capacity was 16.9 GW, which provided 1% of the total US energy. The US isn’t
the only country harnessing wind; Germany’s cumulative turbine capacity was 22.3
GW. In addition, Denmark has the highest percentage of wind power generation at
20% [36]. The primary method of transforming wind energy into electricity is a wind
operated generator, called a wind turbine.

A typical wind turbine is composed of three 40 to 60 m blades connected to a

nacelle or gear generation unit which sits on top of an 60 or 100 m tower. Fig. 1.1
shows a picture of a typical wind turbine. The blades rotate from 4 to 15 rpm to
produce tip speeds, that is, radial velocities at the tip of the turbine, of over 200
mph. Individual turbines can provide over 1.6 MW of power [21]. Turbines are often
arrayed along ridges and plateaus to capitalize on strong winds, a collection of turbines
is called a wind farm.

1.2 Problem Statement

One side effect of wind farms is that they can interfere with Air Defense (AD)
and Air Traffic Control (ATC) radars because of Doppler effects. The US and UK
have each performed several field tests in an effort to characterize the interactions
and develop policies for siting wind turbines [25]. As of 2008, the US Department
of Defense (DoD) has no formal review process [33], but it does support FAA siting

1
 Figure 1.1: The basic components of a wind turbine [23].

studies. One reason for this is a lack of analysis tools for studying wind farm effects
on AD and ATC radar performance. A comprehensive tool is needed to analyze the
interference effects to aid the U.S. Government in determining optimal locations for
new turbine construction.

Interference can be categorized by three effects: false moving targets, degraded

target tracking, and missed targets. False moving plots are created by the combination
of the turbine’s Radar Cross Section (RCS) and the high tip speed from the moving
blades. Many radars employ adaptive techniques that change the detection threshold
when a large RCS return is present during several scans in an attempt to remove
stationary objects. Signals returned from small aircraft are below the new detection
threshold, resulting in missed targets. In addition, when a radar is tracking an airplane
that flies over or near a wind farm, the radar can select the false plots and lose the
track history.

2
1.3 Proposed Solution

This thesis describes a software architecture including fidelity requirements to

model and analyze wind turbine effects on pulse Doppler radar. There are no available
programs with sufficient fidelity to evaluate current and future turbine sites. One
product of the proposed architecture is the ability to convey the affect of turbines on
a combined radar coverage map [11].

1.4 Sponsors

1.4.1 84 RADES. The mission of the 84 Radar Evaluation Squadron

(RADES) [3] is “to provide the Warfighter responsive worldwide radar-centric plan-
ning, optimization, and constant evaluation to create the most sensitive integrated
radar picture.” They are the main DOD organization charged with understanding
and evaluating the effects of wind turbines on ATC radars. The proposed modeling
and simulation architecture can be used as a baseline to develop software to analyze
the effect of wind turbines on a radar. In particular, this effort will determine which
components require additional development.

1.4.2 NASIC. The mission of the National Air and Space Intelligence Center
(NASIC) [24] is to “produce integrated, predictive air and space intelligence to enable
military operations, force modernization and policymaking.” NASIC is interested in
modeling advanced sensors and the radar principles involved with wind turbine model-
ing are very relevant to many sensors studied by NASIC. Hence, recent improvements
to a particular modeling and simulation environment, interactive sensor simulator for
terrain environments (INSSITE ), offers a means to study the wind turbine problem.

1.4.3 453rd EWS. The 453rd Electronic Warfare Squadron (EWS) [15]
“provides EW-focused Information Operations (IO) analysis, data, flagging, models
and simulations to focus combat power, increase survivability and ensure mission
success”. One effort of this thesis is to include large objects, like buildings and

3
wind turbines, within Digital Terrain Elevation Data (DTED). That capability when
paired with Improved Many-On-Many (IMOM ), a 453rd EWS product, may enhance
the accuracy in radar-based wind farm analysis.

4
 II. Literature Review

T his chapter will focus on work that others have performed on similar topics,
starting with a quick introduction of radar basics. Next is a review of physics
effects associated with wind turbines. This is followed by papers that show wind
turbines can impact a radar’s performance [21]. Finally, the proliferation of large-
scale modeling and simulation is discussed.

2.1 Radar Basics

2.1.1 Radar Range Equation. One of the first equations derived in any radar
book or paper is the received power,

σ Gr λ2
Pr = (Pt Gt ) , (2.1)
4πR2 4π(4πR2)

where Pt Gt is the transmit power directed towards a target, σ/4πR2 is the portion
of the signal reflected by the target, and Gr λ2 /(4πR)2 is the portion of the reflected
energy intercepted by the antenna [9] [30]. The 4πR2 term in the denominators
account for spherical radiation. Another important equation is the maximum radar
range,

4 Pt τ G2 λ2 σF 4
Rm = , (2.2)
(4π)3 kTs Dx (n)L
where L is the combined loss factor, F is the pattern propagation factor, and Dx (n)
is the detectability factor [9]. The detectability factor is also known as minimum
discernible signal [30].

2.1.2 Primary vs Secondary Surveillance Radar. AD and ATC radars typ-

ically employ a primary surveillance mode where a RF signal is transmitted into a
volume of space and objects in the path reflect the signal. The primary surveillance
radar (PSR) processes the returned signals to determine locations. Therefore, PSRs
are subject to two-way RF propagation losses, like (2.1).

5
 A secondary surveillance radar (SSR) is a cooperative system with aircraft.
ATCs send a coded RF signal and any aircraft equipped with a transponder or beacon
sends a coded message back to the SSR. The message from the aircraft contains
identification, heading, speed, and elevation information. SSRs are subject to one-
way RF propagation losses. SSRs are less susceptible to clutter because of the one-way
path loss.

SSRs provide more accurate data than PSRs. However, SSRs use a cooperative
system, it is ineffective if an aircraft doesn’t have it or isn’t using a beacon. For that
reason, AD radars depend on PSR to locate any and all targets both friend and foe.
Whenever the term radar is used throughout the rest of this document it is referring
to PSR.

2.2 Wind Turbine Physics

2.2.1 Line of Sight. The first question to ask when determining if a turbine
will affect a radar is: Will the turbine be within the radar’s line-of-sight (LOS)? LOS
can be estimated as

p p
D = 1.41[ ht (f t) + hr (f t)] miles, (2.3)

where, ht is the height of the turbine, and hr is the height of the radar above mean
sea level [21]. The equation assumes a 4/3 earth radius to account for beam diffrac-
tion [21]. The result is a minimum separation distance required to ensure the turbine
won’t interact with the radar.

However, (2.3) doesn’t account for terrain masking or shadowing. If there is a

tall object or terrain between the radar and the turbine, the LOS distance may be
shorter [14] [21]. Figure 2.1 illustrates how diffraction increases the radar coverage
beyond the visible region. Some models, such as IMOM, account for terrain masking
and beam diffraction around terrain to provide range detection diagrams [13].

6
Figure 2.1: Depiction of diffraction from terrain. The regions in light grey are
partially visible because of diffraction.

2.2.2 Radar Cross-Section. The σ term in (2.1) is the RCS of a target.

RCS is a measure of an object’s scattering magnitude [20]. A common misconception
is that RCS is constant for an object, when it actually depends on many factors,
including object orientation and radar wavelength.

Numerous books, like Knott [20], are dedicated to explaining methods for com-
puting RCS values. Methods range from coarse and quick which provide simple values
to detailed and time intense, like X-Patch. The quick methods rely on multiple as-
sumptions to calculate values. Decreasing the number of assumptions and increasing
the density of the observation grid will increase the accuracy, complexity, and time
of the calculations. Accuracy and timing requirements will determine what methods
can be used to model turbine RCS.

Wind turbines are very large electrically, meaning they are on the order of
several hundred wavelengths tall [19]. The following is an example of pre-processed
RCS approximation method that could be leveraged to reduce evaluation time of a
wind farm.

The RCS of a turbine, using a computational electromagnetic (CEM) program

like X-Patch with a fine computational mesh at X-Band, can take over a decade for a

7
 (a) The measured RCS of the GE windmill (170 (b) The predicted RCS of the GE windmill (170
deg yaw, 0 deg elevation, 1.5 GHz, 12.7 rpm, VV deg yaw, 0 deg elevation, 1.5 GHz, 12.7 rpm, VV
polarization) [19]. polarization) [19].

Figure 2.2: Measured vs X-Patch RCS [19]. The diagram in 2(a) shows the mea-
sured RCS of a wind turbine ranges from 40 dBsm to 20 dBsm. On average, 2(b)
shows the predicted RCS matches actual measurements. However, the spikes, in 2(b)
are artifacts of the simulation.

64-node supercomputer to process [19]. That calculation accounted for 120o of blade
rotation. For lower frequencies, such as L- or S-Band, a coarse mesh is used to reduce
computation time without sacrificing accuracy. The density and size of the mesh
directly influences the processing time. Another drawback is these calculations would
need to be performed and stored for every wind turbine model used in the US, or a
reasonable subset. But once a model has been evaluated, the data can be accessed
via look-up tables.

The Air Force Research Laboratory (AFRL) performed field tests with 10 tur-
bines in Fenner, NY. The turbines were evaluated at L-, S-, C-, and X-bands with
a variety of wind conditions to characterize RCS and Doppler characteristics [25].
The goal was to create a data set to verify RCS simulations of turbine models. Fig-
ure 2(a) shows the actual measured data and Figure 2(b) shows the simulated X-Patch
results [19]. The simulation data closely matches measured data except for the arti-
facts.

8
 2.2.3 Doppler Effect. The Doppler effect is the compression or expansion of
radio frequency (RF) signals by a moving object. Pulse Doppler radars can measure
the frequency change or Doppler shift of the transmitted signal. These radars us
the Doppler shift to calculate how fast an object was moving. This relationship is
described by
vr
fd = 2 , (2.4)
λ
where fd is the Doppler shift, vr is the radial velocity, λ is the wavelength of the
transmitted signal [30].

The wind drives the turbine blades, which are connected to a generator by
a shaft, to rotate and generate electricity. Figure 2.3 depicts a power curve for a
turbine [8]. As the wind speed increases, the turbines rotate faster producing more
power and a larger Doppler shift. Figure 2.4 shows tip speed relative to wind speed for
several common wind turbine models [8]. At maximum rotation speeds, the turbine
is comparable to a slow flying aircraft.

The Doppler returns from a turbine are well documented by QinetiQ [26] and
AFRL [4]. Figure 2.5 is an example of the Doppler returns created by a wind tur-
bine [26]. The wind is blowing perpendicular to the radar boresight creating maximum
Doppler shifts. The cyclical spikes in Doppler correspond to times when the turbine
blade is passing either straight up or straight down. A smaller spike forms just be-
fore or just after a turbine blade passes vertically downward. The smaller spike is
caused by the combined blade and tower phase center moving rapidly away from the
radar [26].

Another factor that determines the Doppler shift the radar perceives is the
angle between the turbine blades and radar boresight. The equation for calculating
the maximum Doppler shift caused by turbine blades is

wr
fd = 2 | cos(90o + φradar − φwind)|, (2.5)
λ

9
Figure 2.3: The graph depicts typical power curves for wind
turbines vs wind speed [34]. Turbines usually produce little to
no power for low wind speeds then increase linearly for normal
operating ranges before reaching a max power plateau.

10
Figure 2.4: The graph shows estimated tip speed vs wind speed for common wind
turbines [8].

−1500

−1000

−500
Doppler (Hz)

500

1000

1500
0 1 2 3
Time (s)

Figure 2.5: The graph illustrates the cyclical Doppler returns from a wind tur-
bine [26].

11
Figure 2.6: Family of curves showing apparent tip speed as wind changes direction
by 10 deg increments [8]. The apparent tip speed decreases as the angle approaches
0, when the blades are facing the radar.

where r is the length of the blades in meters, w is the rotation rate of the blades in
RPM, λ is the operating frequency of the radar, φradar is the radar azimuth (measured
from North), φwind is the wind direction, and the 90 deg factor accounts for turbine
blades spin perpendicular to wind direction [8]. The equation reaches the maximum
value when φradar − φwind is an odd multiple of 90 deg. Figure 2.6 shows a family of
curves for apparent tip speed versus wind speed for a single wind turbine based on
the angle between radar boresight and wind direction [8]. That means false targets
that move tangent to the boresight of the PSR will appear much slower than inbound
or outbound false targets which may have implications for tracking algorithms.

2.3 Wind Turbine Effects on Radar

Shadowing is the largest concern for AD radars [13]. Table 2.1 shows the signif-
icance of each impact depends on the radars use [13]. The terms “Raised Threshold”
and “Clutter” are related because a radars’ detection threshold is raised to reduce
clutter [13].

12
 Table 2.1: Order of significance for PSR impacts, 1 is highest [13]
Type of Airspace/Use of PSR
TMA/High
Density En- Low Density
Impact Airport/Approach Route En-Route Air Defence
Shadowing 3 3 2 1
Clutter 1 1 1 3
Raised Threshold 2 2 3 2
Receiver Saturation 4 4 4 4

2.3.1 Shadow Zone. In the same manner that terrain can mask wind tur-
bines, turbines can also mask targets. The region behind a turbine, from a radar’s
perspective, that is either completely or partially occluded is called a shadow zone.
The UK flight test determined that the shadow zone would affect a PSR’s ability to
detect targets within 5 km behind a wind farm [5].

A shadow zone usually forms a 4 deg or smaller wedge behind large obstructions
like turbines [14]. The two-way reduction in power [21] caused by shadowing is
s !
Drt S 2
Lsz (dB) = 40 log 1 − , (2.6)
Drw Dwt λ

where Drt is the distance between radar and target, Drw is the distance between
radar and wind turbine, Dwt is the distance between wind turbine and target, and
S is the width of the wind turbine. The greatest reduction in power happens when
the target is directly behind and in line with the wind turbine. The farther apart the
target and the wind turbine are, the smaller the shadow zone affect on probability of
detection. Diffraction is the primary reason that the partial shadowing has less effect
on radars at longer separations. Table 2.2 provides rule-of-thumb values for shadow
length, based on acceptable power loss [14].

The Netherlands Organisation for Applied Scientific Research (TNO) also in-
vestigated radar shadowing [31]. TNO focused on reducing maximum detection range
and probability of detection. They determined that placing a turbine 4 km to 8 km

13
Table 2.2: L-Band (1 GHz) rule-of-thumb shadow length values. The values are
the distance (km) the shadow extends behind the turbine. [14]
Tolerable power Distance from radar to turbine (km)
loss PL (dB) 0.5 1 2 5 10 20
0 ∞ ∞ ∞ ∞ ∞ ∞
-1 ∞ ∞ ∞ 46.0 8.2 5.8
-2 ∞ ∞ 3.4 1.7 1.4 1.3
-3 ∞ 1.7 0.9 0.7 0.7 0.6
-5 0.6 0.4 0.3 0.3 0.3 0.3
-10 0.1 0.1 0.1 0.1 0.1 0.1

away from the radar reduces the maximum detection range [31]. In addition, they
showed that for a 1 m2 target, the detection probability from a single dwell is reduced
by as much as 40% when a turbine is placed 4 km from the radar, which is quite
significant. TNO didn’t supply results for cases where turbines were 30 km or more
away from the radar, which is more common in the US.

Shadow zones are predictable which leads to a simple mitigation technique. The
technique is to site turbines at the base of plateaus or on hillsides, ensuring a portion
of the shadow zone is masked by terrain [14]. Figure 2.7 is a hypothetical scenario
that applies this method. The dashed area is the profile of the shadow zone. In effect,
this approach overlaps areas the radar can’t view.

2.3.2 Clutter. Clutter is unwanted scattered signals from trees, birds, rain,
sea, ground, buildings, and other large structures [9]. Wind turbines are large metal-
lic structures that fall under the category of clutter, the combination of size and
electromagnetic (EM) properties result in strong radar returns. To put the size in
perspective, the RCS of a 1.6 MW wind turbine can average between 20 and 40 dBsm
which is comparable to a passenger jet aircraft [25]. The rated power output of a tur-
bine increases directly with blade length and generator efficiency, but the turbine’s
RCS scales with the size of the tower and blades.

14
 Figure 2.7: The dashed line represents a shadow zone being
clipped by terrain.

2.3.3 Raised Threshold. Modern radars use clutter maps to isolate and
remove stationary clutter. Clutter maps are divided into clutter cells which can be
further divided into range resolution cells. Figure 2.8 compares 9 range resolution
cells that cover 3 km in range and 3 deg in azimuth with a clutter cell that is 3 km in
range and 3 deg in azimuth [25]. The radar keeps a history of the signal strength for
several scans and adjusts the detection threshold of each clutter cell. A large scatterer
in one range cell will cause the detection threshold of neighboring range cells, within
the clutter cell, to also increase [7]. A turbine in cell A can mask a target in cell B.

Figure 2.9 shows the affect of turbine farms on AD radar clutter maps [7]. The
thick lines are the clutter cell; cells A10, B10, and B11 contain turbines. In those
cells, the PSR returns from the aircraft are below the detection threshold and only
SSR returns were received.

2.3.4 False Moving Targets. Moving target indicator (MTI) circuits remove
stationary clutter. If a target is stationary, the successive pulses are nearly identical

15
Figure 2.8: Clutter cells are generally the size of multiple range resolution cells.

Figure 2.9: Data from a flight test showing clutter cells that contain wind turbines
(A10, B10, and B11) have a reduced probability of detection [7].

16
 Table 2.3: L-Band (1 GHz) rule-of-thumb RCS values [14]
Typical maximum Typical average Typical maximum Typical average
pre-MTI RCS pre-MTI RCS post-MTI RCS post-MTI RCS
Blade length (m) (dBsm) (dBsm) (dBsm) (dBsm)
60 54 44 53 9
55 53 43 52 8
50 51 42 51 7
45 50 41 50 6
40 49 40 48 5
35 47 39 46 4
30 45 38 44 3
25 43 36 42 1
20 40 35 39 0
15 37 33 35 -2
10 33 31 29 -4
5 29 29 18 -6

and will cancel out when subtracted. However, moving objects cause pulse-to-pulse
phase changes that are not canceled by MTI circuits. Subsequent Doppler filters are
needed to estimate the target’s velocity [25].

MTI filters can remove part of the turbines radar return, e.g. from the stationary
tower, and thereby reducing the average return energy, as shown in Table 2.3. There
are additional rule-of-thumb tables for S- and X-Band in [14]. The combination of long
turbine blades and high rotation rates results in tip speeds that are comparable to slow
flying aircraft (180 mph) [25]. The correlation of a large radar return and Doppler
shift from the turbine blades will insure radar returns pass through MTI filters as
valid targets for display on a plan position indicator (PPI). Figure 2.10 shows a PPI
display where wind farm clutter is indistinguishable from a moving aircraft [6].

There are no current techniques that can filter the returns from a moving turbine
without adversely affecting desired aircraft returns, but one is being developed by
BAE [14]. Advanced digital tracker (ADT) reduced adverse impacts caused by wind
turbines on radar [12]. However, some tracks were dropped and other tracks were

17
Figure 2.10: Plan position indicator (PPI) [6] showing how difficult it is to distin-
guish a turbine from an aircraft.

18
seduced [12]. In 2006, the UK Ministry of Defence determined the product didn’t
completely mitigate safety concerns [12].

2.3.5 Tracking. Radar target tracking is accomplished by first estimating

a targets current range and velocity and then estimating where the target will be
at the next scan and selecting a correlation gate to pick up the next return. If
there are multiple moving targets within a correlation gate, the tracker has to decide
which return belongs to which track. When a situation like that occurs, the tracker
can either select the right target/track pairing, switch the tracks, or drop the tracks
altogether and start new tracks. Tracking radars usually have a “limited” number
(several hundred to several thousand) of targets they can simultaneously track. The
hardware processing power limits the number of target tracks available.

Wind farms can severely degrade a tracking radar’s performance, based on tur-
bine layout and spacing [14]. The false moving targets can lure the tracker away from
the real target, and the raised detection threshold can mask returns causing a track
to drop. In addition, wind farms may be capable of producing enough false plots to
overwhelm the tracking processor, causing the radar to drop tracks or stop initiating
new ones.

One mitigation technique for trackers, called non auto-initiation zone (NAIZ),
allows operators to establish zones where no new tracks will be created [14]. Tracking
radars that use NAIZ will assume aircraft fly straight over a wind farm. Because of
that, the tracker will coast the track and attempt to reacquire the aircraft once it
clears a farm, thereby preserving track history. However, trackers that use NAIZ still
can’t reliably track an aircraft over a wind farm.

2.3.6 Receiver Saturation. Saturation occurs when large objects are near the
radar, resulting in large amplitude returns which saturate the receiver [13]. Receiver
saturation is one effect that is often neglected in modeling and simulation, because
many radars employ automatic gain control (AGC) to prevent saturation. When a

19
 Table 2.4:
Proposed Wind States [8].
Approximate
Tip Speed (kts) Wind State Wind Speed (m/s)
0 < vr ≤ 20 0 1
20 < vr ≤ 40 1 2
40 < vr ≤ 60 2 3
60 < vr ≤ 80 3 5
80 < vr ≤ 100 4 7.5
100 < vr ≤ 120 5 10
120 < vr ≤ 140 6 13
140 < vr ≤ 160 7 17
160 < vr ≤ 180 8 22
180 < vr 9 25+

receiver is saturated, the sensitivity and dynamic range are reduced, which reduces
the radar’s maximum detection range.

2.4 Modeling Wind Turbine RCS

As shown previously, performing CEM on wind turbines is expensive in both

time and resources. When you combine the fact that there are a vast amount of turbine
manufacturers and many have multiple models that would need to be computed, a
method using look-up tables seems unlikely in the near term.

Another method for modeling wind turbines is a concept called wind states [8]
which uses stochastic amplitude and phase data to describe wind turbine clutter. A
similar concept, the notion of sea states effecting sea clutter, is widely used in radar
receivers [8]. Table 2.4 associates wind states with ranges of tip speeds [8].

The proposed signal model

s(t) ∝ hσθ iejhφθ i s0 (t), (2.7)

where s0 (t) is the signal when modeled with the radar range equation, hσi and hφi are
the magnitude and phase of turbine clutter [8]. Aircraft returns and ground clutter

20
 Figure 2.11: Maximum Doppler versus angle between radar
boresight and wind direction based on (2.5). The graph accounts
for varying wind states.

are accounted for in s0 (t). Stochastic methods produce accurate data quickly without
requiring the use of supercomputers.

The goal of wind state modeling [8] is to create a family of curves similar to
probability of false alarm graphs for receivers [30]. Figure 2.11 displays a hypothetical
graph of maximum Doppler versus angle for various wind states. Figure 2.12 shows
how hσθ i might be calculated from a collection of turbine measurements.

2.5 Large Scale Modeling and Simulation

2.5.1 How Large is Large? The definition of large-scale modeling and sim-
ulation (M&S) varies from one source to another. One of the generic definitions is,
any system that can be separated into a number of interconnected subsystems for
either computational or practical reasons [18]. Another definition is, the dimensions
of the system are so large that conventional techniques of modeling, analysis, control
design, and optimization fail to give reasonable computational efforts [18]. That can
be simplified to “a system is large when it requires more than one controller” [22].

21
 Figure 2.12: Example using a collection of turbine measure-
ments with multiple blade positions to generate hσθ i.

2.5.2 Who Uses Large-Scale M&S? In short, almost everyone. Large-scale

models and simulations are hierarchical, decentralized information structures that ap-
proximate systems dealing with society, business and management organizations, the
economy, the environment, communication, electrical power, logistics, military com-
mand and control, information systems, aerospace, water resources, and energy [18].
Therefore, modeling the effects of turbines on a radar, in detail, would qualify as a
large-scale model.

Both the government and contractors have developed many large-scale M&S
applications to simulate radar-aircraft-clutter interactions. This report will focus on
two applications, SAIC’s INSSITE and the 453rd’s IMOM. In addition, these applica-
tions will be evaluated for their potential to model radar-turbine-aircraft interactions
in the following chapters.

2.5.3 M&S Hierarchy. One method for defining the layers of a model or
simulation is shown in Fig. 2.13 [28]. The bottom layer is physics, which controls
physical interactions and phenomena, like EM radiation and aircraft kinematics. The
physics-layer is not computationally expensive for a small scenario, and there is almost

22
 (a) (b)

Figure 2.13: Modeling and simulation pyramid. (a) Cost versus level of repre-
sentation. Physics models have the highest-level and most accurate representation.
Campaign models have the most reduced form and often depend on significant sta-
tistical analysis. (b) Data, information, knowledge versus pyramid level. Knowledge
requires the greatest amount of processed information, which means higher costs [28].

no dependence on probability or statistics. The top layer is campaigns, which are

vast scenarios encompassing many objectives or missions. An example is evaluating
the effect of turbines on the US combined radar coverage map. Campaign models
are supported by running well-over a thousand iterations of a simulation to develop
the probability and statistics information. Therefore, vast resources are needed to
generate results at the campaign level.

On the side of Fig. 2.13(b) there are three words: data, information, and knowl-
edge. Data is often viewed as the “lowest level of abstraction from which information
and knowledge are derived” [35]. Information is “giving form or shape to the mind,
as in education, instruction, or training” [29]. Information can also be thought of
as refined data, but information is not always accurate [35]. Knowledge is “what is
known in a particular field or in total; facts and information” [29]. Every layer of the
pyramid can contain data, information, and knowledge [28].

Figure 2.14 shows military planning is very similar to the M&S pyramid [32].
Along the bottom is complexity which is synonymous with computational cost. In
addition, the focus of effort scale ranges from designing (art) which seeks a systemic

23
Figure 2.14: The US Army’s spectrum of military planning [32] carries the same
message as the M&S pyramid.

and shared understanding or knowledge, to engineering (science) that could only mean
physics and engineering-level [32].

24
 III. Methodology

T his chapter covers the method used to develop the modeling and simulation
framework for analyzing the effect of complex clutter on a pulse-Doppler radar.
Two programs will be run through several set scenarios to create baseline capabilities
for future simulation software.

3.1 Overview

Figure 3.1 depicts the method that was followed to generate a M&S architec-
ture. The first step, understand the problem, was detailed in Chapters I and II.
For the second step, determine the components, the high-level systems were obvious:
radar, aircraft, turbines, terrain, and EM phenomenon. Some of the systems can be
separated into logical subcomponents. Calculating the complexity means evaluating
the computational cost of systems at several M&S levels. The evaluation includes
computing time and accuracy for every level. At this point a reasonable architecture
can be developed. The framework is presented in Chapter IV. Next, two different pro-
grams, INSSITE and IMOM, will be examined in the context of the M&S framework
to determine which components are currently developed. The results of the evaluation
are detailed in Chapter V. Finally, develop requirements will highlight components
that aren’t at the appropriate M&S level. The requirements will demand additional
data for the development of statistical models.

Figure 3.1: Represents the method used to address the prob-

lem.

25
 Figure 3.2: The left model can contain 400k or more facets
and require more than a year to produce the best representation
of one turbine’s RCS. The middle model is a reduction of the
left model, it can contain 330k scatterers but only require hours
to process a single set. The right model is as simple as possible.
It is accurate for a very limited scenario but it’s easy to process.

3.2 Evaluating Complexity

The turbine RCS was evaluated at the physics, engineering, and engagement
levels. The evaluation included a physical representation, a method, and a sample
model for acquiring RCS. Figure 3.2 illustrates the complexity of three methods used
for evaluating turbine RCS. A similar breakdown can be performed for phase data.

3.2.1 Physics-Level. Each individual turbine is represented in extreme de-

tail, thousandths of a degree blade rotations. Several methods: method-of-moments
(MoM), scattering centers, and ray tracing, are used to generate RCS values at thou-
sands of look angles. They solve very large systems of equations using matrix inversion
which has a complexity on the order of N3 . The method used to calculate the scattered
electric field is Z
s
E (t, r) = G(r, r ′ )J(t, r)dr′ , (3.1)
v

where r is the vector from the origin to the observation point, r ′ is the vector from
the origin to the source, t is time, G is the Green’s function, J is the current density,
s
and E is the scattered field which can be converted to RCS.

26
 Figure 3.3: Chi-squared function which may simulate the RCS
of a turbine. The actual model will need to be derived from
physics or engineering-level data.

3.2.2 Engineering-Level. Individual turbines are represented by a set of

fixed of blade positions, on the order of degrees. Instead of performing a full physics
estimation, look-up tables or a reduced number scattering center models are used.
These methods are advantageous because their complexity is on the order of N com-
pared with N3 for physics methods. When evaluating scattering centers, the total
RCS
N
X
σ= σi e−jk·ri , (3.2)
i

where k is the angular wavenumber and σi is the RCS from the it h range cell.

3.2.3 Engagement-Level. Whole wind farms are evaluated at this level and
blade position is account for statistically. The RCS is generated from a probability
density function. A hypothesis test is performed to determine if the radar returns
from wind farms are above detection thresholds. This method involves solving one
equation which makes the complexity on the order of 1. The model is a graph, similar
to chi-squared function shown in Figure 3.3

27
3.3 Architecture Development

The first step to developing a hybrid large-scale M&S framework is to determine

the simulation scope. The scope can be expressed in terms of the M&S pyramid based
on the number and scale of the objectives and products. The top-level objective is to
simulate the effects of turbine and clutter interference on a radar’s ability to detect an
aircraft. The next step is to recursively divide the model into logical components or
subsystems. Then, define the products and outputs for each component. In addition,
the subsystems should be separated based on the fidelity requirements. The final step
is to associate the subsystems with particular M&S levels based on its objectives,
products, and fidelity. The key to a hybrid model is the ability to access low-level,
physics-level, data when required while still being able to use higher-level, engineering
or engagement-level, information or knowledge when appropriate.

3.4 Existing Programs

This is not the first modeling architecture ever developed for radar applications.
Most radar models use either detailed RCS measurements, like CEM techniques, or a
constant amplitude for RCS, ignoring Doppler. CEM and scattering center techniques
are prohibitive because of the sheer size of the scenario. A single turbine is electrically
very large at typical radar wavelengths. Whereas using a constant amplitude for RCS
overly simplifies EM phenomenon and results in important information being lost.

There are a large variety of government and contractor radar models, each
tailored for specific purposes. This effort is based on two such programs, SAIC’s
INSSITE and the 453rd EWS’s IMOM. INSSITE operates as an engagement-level
simulation with components at the engineering and physics-levels. IMOM operates
as a campaign or mission-level simulation with subsystems at the engineering and
physics-levels.

The programs will be evaluated by reviewing program documentation, creating

experimental simulations, and corresponding with program experts. Since INSSITE

28
is an in-house design tool there is little documentation. That means discussions with
the program developers will provide a bulk of the understanding of capabilities and
limitations. IMOM is a prolific program in the electronic warfare (EW) community
and is distributed with detailed documentation.

The programs are expected to provide partial solutions for analyzing radar-
aircraft-turbines-clutter interactions. As such, the programs might verify some sub-
sections of the proposed framework. However, neither program is likely to provide a
complete solution, which led to the design of a hybrid large-scale M&S framework.

3.5 Experimental Simulations

The programs will run through a series of iterative simulation scenarios to de-
termine their strengths and weaknesses for modeling this problem. The experiments
will be slightly different for each program due to software limitations.

3.5.1 IMOM Experiments. The main limitation of IMOM is that it doesn’t

calculate phase data. However, it is very effective at calculating the magnitude of
returns from targets while also accounting for LOS.

The first experiment is to verify that LOS is taken into account. A radar, with
parameters similar to an ARSR-4, will be placed on a plateau and attempt to detect
a 1 m2 calibration target 200 ft above ground level (AGL). The resulting analysis
should show which areas were masked by terrain and which areas were visible.

The second experiment is to test the radar’s ability to detect an aircraft while
accounting for terrain clutter. The setup is similar to the first experiment except a
larger, possibly a 3D, RCS is used and the aircraft is at a higher altitude: 1500 ft
AGL. The clutter is incorporated with the various detection methods.

The third experiment is to determine if the radar can detect wind turbines with
ground clutter. Again, this is a similar analysis to the first experiment except using

29
a much larger 10,000 m2 RCS. It is anticipated that the radar will detect the turbine
at an extended range compared to the calibration target.

The fourth experiment is to combine the second and third test to determine if
the radar can detect an aircraft in the same resolution cell as a turbine. However,
IMOM evaluates each target individually, requiring other techniques to account for
the effect of wind turbine clutter on a radar. One possible method involves modifying
DTED at the turbine location to increase the terrain clutter. Another method might
be to model the turbine as a radar jammer which would raise the radar’s detection
threshold in a similar manner.

3.5.2 INSSITE Experiments. INSSITE is a SAIC in-house development

tool that provides some unique features, like the ability to simulate the movement
of objects prior to performing detailed analysis. But, the tool is not ready for main-
stream use due to some key limitations, like scenario size and sluggish response to
user manipulation.

The first experiment is to setup a radar to collect ground clutter. This task will
include creating a stationary radar with a rotating beam and incorporating DTED
into INSSITE.

The second experiment is to place a stationary turbine in freespace, then run

analysis with varying observation densities, from 1 λ up to 5 λ, to determine conver-
gence. The objective is to find the largest grid spacing that still produces an accurate
RCS magnitude for the turbine. With larger grid spacing, fewer rays are traced, and
the processing time is shorter.

The third experiment is to evaluate the RCS of a turbine when combined with
terrain. Using a higher multi-bounce setting will provide more accurate data at the
expense of simulation time.

30
 The fourth experiment is to calculate the return for an aircraft with terrain
included. The aircraft will be 1500 ft AGL like in the IMOM experiments. This may
allow some comparison between the two programs.

The fifth experiment combines the third and fourth experiments to test the
impact of turbines and clutter on a radar’s probability of detection. Because of the
small scenario size limitation, this experiment can’t be expanded to included multiple
turbines with an aircraft flying.

An optional experiment is to separate the wind turbine model into two parts;
one containing the blades and hub, the other containing the tower and nacelle. Next,
leverage INSSITE’s ability to dynamically move objects to rotate the turbine blades
in a realistic manner. This would enable the collection of phase data to generate a
complete RCS analysis.

3.5.3 Underlying Mathematics. In each of the above experiments, the im-

pact of sidelobe and backlobe interference must be included. Even though sidelobes
may be 20 to 30 dB down from the mainbeam, turbines with a RCS of 30 to 40 dB
can still interfere with the detection of small mainbeam targets. Figure 3.4 shows
a range ring and several lobes. The length of the lobes is an indication of relative
power.

For each cell the received power should be calculated using the radar range
equation (2.1). Many of the terms in the equation are the same for every cell within a
range ring and can therefore be rolled into a constant, Ci. That simplifies the equation
for power received, Pr,i = Ci G2i σi , for individual cells. The total received power at
one instant,
N
X
Pr = Pr,i , (3.3)
i=1

accounts for the contributions from every lobe. The values of σi can come from terrain,
aircraft, and turbines.

31
Figure 3.4: Range ring showing how sidelobes and backlobes
contribute to the total received strength for a single range cell.

32
3.6 Analyzing the Architecture

The goal of this effort is to determine which areas of the architecture need
further development. To accomplish the task, components of the proposed framework
were compared with capabilities inherent to IMOM and INSSITE. If neither of the
programs could perform one of the functions at the appropriate M&S level, then that
function is tagged for further development. Whereas, components that are already
at the correct M&S level should be leveraged rather than redesigned. This prioritizes
the research areas requiring additional funding.

33
 IV. Radar-based Modeling and Simulation

T his chapter breaks the simulation into components and discusses the products
and objectives at each level. First, the mission-level which provides the glue for
all the components is highlighted, followed by a brief discussion of the connections
between lower level components. Finally, the individual components are separated
and evaluated for requirements at engagement, engineering, and physics-levels.

4.1 Connection Diagrams

This section will traverse the tiers of the M&S pyramid to show how the models
will interact. When data is requested but isn’t available, the model will push-down
to a lower level to generate the data. For example, a radar requests a simple aircraft
RCS, at the engagement-level, but that data isn’t known. The aircraft RCS feature
will drill down to the engineering level to find a suitable solution or to the physics
level to generate the data. If it’s not reasonable, cost or time prohibitive, to generate
the data, an error message will notify the user.

4.1.1 Mission-Level. Figure 4.1 depicts the mission-level simulation where

statistical objectives are generated from a large number, thousands or tens of thou-
sands, of repetitions of the engagement-level M&S. At the highest level, the simu-
lation should output data on how interactions between the radar, aircraft, clutter,
and/or structures affect radar probability of detection [28]. In addition, some infor-
mation/knowledge about the effect changing structures and/or locations has on radar
performance should be generated [28]. The objective of the simulation is to determine
radar performance measures for aircraft detection despite environmental interference,
clutter and large moving structures like wind turbines [28].

Once the hypothesis tests are developed, the evaluation process can be auto-
mated to produce a radar coverage map. In addition, the analysis results must be
conveyed in a straightforward manner. Figure 4.2 shows a simplistic FAA tool for
determining if a turbine location might interfere with a radar [2]. The proposed radar

34
Figure 4.1: The mission-level simulation depends on lower-
tier, engagement, simulations. The objectives are in the form of
hypothesis tests [28].

35
 Figure 4.2: FAA tool, Obstruction Evaluation/Airport
Airspace Analysis (OE/AAA), performs rough analysis to de-
termine if a proposed wind turbine site requires additional anal-
ysis. The colored circles represent areas requiring additional
analysis [2].

coverage map should look similar to Figure 4.2 but, it should provide analysis for each
resolution range cell.

4.1.2 Engagement-Level. Figure 4.3 depicts the engagement-level simula-

tion which is dependent on both engineering and physics-level M&S to generate the
hypothesis tests [28]. The engineering-level models should be run many hundreds or
thousands of times to account for changes in the receiver, processor, target RCS, and
target Doppler. The hypothesis tests are derived by comparing the probability of
detection between four scenarios: radar with clutter, radar with aircraft and clutter,
radar with turbines and clutter, and radar with aircraft, turbines, and clutter.

36
 Figure 4.3: The engagement-level simulation depends on
interaction between engineering-level M&S and physics-level
M&S. The objectives are represented as hypothesis tests [28].

4.1.3 Engineering-Level.

4.1.3.1 Radar. The main component at the engineering-level is the

radar model which is connected to the environment as presented in Figure 4.4 [28].
The radar itself has several subsystems, that are discussed later. The radar model
follows the generic block diagram used in several text books [9] [30]. When modeling
radars, the topic of signal-level versus parametric modeling is always at the forefront.
Signal-level models represent pulse features, with nanosecond accuracy, for the time
step of radar signals. Signal-level modeling is synonymous with physics-level model-
ing. Whereas, parametric models use pulse repetition interval, several milliseconds,
for the time step and define the radar signal by a few parameters (amplitude, pulse
width, frequency, and modulation) [28]. Because of the lower level of detail, paramet-
ric modeling is equivalent to engineering-level modeling. Signal-level models are the
most detailed and accurate radar models.

37
 Figure 4.4: The main engineering-level model is the radar
model, it connects the other engineering and physics models
together through the environment component [28].

4.1.3.2 Environment. The environment block is key because it ties

together the radar, electromagnetic (EM) phenomenon, and target engineering and
physics models. The main difference between engineering-level and physics-level EM
models are the density of the observation grid and the number of assumptions used.
In addition, engineering-level EM models can use previously processed physics data,
in the form of look-up tables and scattering centers for aircraft and turbine signa-
tures [28].

The term “targets” encompasses aircraft, clutter, turbines, and other large
structures. Clutter models are often statistical functions, based on a terrain scattering
coefficient [9], that can be thought of as engineering-level models. The engineering-
level model of the aircraft and turbine are characterized by scattering center models
and/or look-up tables.

38
 Figure 4.5: The physics-level M&S for EM wave scattering
from a moving target. Maxwell’s equations form the basis of
the EM physics model, but the simulation also requires a moving
target model(aircraft or turbine) [28].

4.1.4 Physics-Level. Figure 4.5 depicts the highest fidelity physics model for
EM wave scattering. Maxwell’s equations are solved for moving aircraft and turbines
to determine the amount of scattered energy. Simulations at this level are very detailed
and require intensive processing that should be avoided for typical analysis. This
simulation level is required when pre-processed data isn’t available or when attempting
to study basic effects such as the time-evolution of the scattered EM waves.

4.2 Features

This section defines several features or components in the context of the M&S
architecture. The following list contains features that are discussed.

• Display

• Radar - Transmitter

39
 • Radar - Antenna

• Radar - Receiver

• Radar - Signal Processor

• Target - Aircraft Motion

• Target - Turbine Motion

• Target - Clutter

• EM - Propagation

4.2.1 Display. At the mission-level, the display component would show a

low resolution, much larger squares than range resolution cells, radar coverage map
based on power levels. At the engagement-level, the display consists of a red/yellow/-
green stoplight map based on LOS, probability of detection, and false alarm. At the
engineering-level, the display should operate like a PPI for one or multiple scans.

4.2.2 Radar. For simplicity, at the engagement-level many radar compo-

nents are modeled as strict power gain or loss in dB. The objective is to calculate
relative power received, Pr ≈ Pt + Gtx + Grx + σ − L. Parametric or engineering-level
models are an expanded version of the radar range equation (2.1) [30]

4 Pav GAρa σnEi (n)F 4 e−2αRmax

Rmax = . (4.1)
(4π)2 kT0 Fn (Bτ )fp (S/N)1 Lf Ls

Signal or physics-level models are the most accurate representation of a radar signal.
The equations for the signal models are discussed under each component rather than
assembling one large equation.

4.2.2.1 Transmitter. The transmitter is the component that gener-

ates the signal to interrogate the environment. The transmitter is described by many
variables: bandwidth, power, pulse width, operating frequency, pulse repetition fre-
quency, and waveform. A parametric model can incorporate all of those variables and

40
more. The definition of the transmitter output is

M −1  
X t − mT p
S(t) = Ao ao cos(2πfo t + θo ), (4.2)
m=0
τ

for a rectangular pulsed signal.

4.2.2.2 Antenna. Antennas are the interface between a radar and the
environment. Depending on the application, the transmit antenna may be different
than the receive antenna. Antenna beams are characterized by several variables:
gain, shape, azimuth and elevation beam widths, mismatch loss, sidelobe locations,
frequency range, and scan rate. Parametric models will use a subset of those variables.
The physics-level representation of the antenna
s
G(θ(t), φ(t))
αt (t) = , (4.3)
LD LC

requires a continuous equations for gain, G(θ(t), φ(t)), duplexer loss, and coupler loss.

4.2.2.3 Receiver. The receiver is responsible for down-converting the

radar returns to baseband or intermediate frequency and separating the signal into in-
phase and quadrature components. The final part of the receiver is the matched filter.
A parametric model of a receiver is characterized by its noise factor, dynamic range,
bandwidth, and minimum detectable signal. In addition, a receiver may include pulse
integration which is described by the number of pulses and integration efficiency. The
signal-level model of the matched filter is
Z ∞
yo (t) = yin (λ)s(λ − t)dλ. (4.4)
−∞

4.2.2.4 Signal Processor. The processor designs, which vary from

radar to radar, are pivotal to reducing wind farms impact on aircraft detection. Radar
processors generally have several operating modes or channels which should be eval-

41
 Figure 4.6: The processor of an ARSR-4 contains many filters
which help determine detection thresholds [10].

uated individually and in conjunction to determine the optimal method for detecting
an aircraft in the same or adjacent cell as complex clutter. Figure 4.6 presents the
processor components within an ARSR-4 radar [10]. The individual blocks require
physics-level detail to accurately model detections.

At the engagement-level the signal processor will have a detection threshold

on top of any processing gains. The engineering-level will include the gain from it’s
algorithms along with the effects of clutter maps. The physics-level will have the
actual algorithms to process signals.

4.2.3 Target. Targets include aircraft, turbines and clutter.

4.2.3.1 Aircraft Motion. Aircraft motion is important for Doppler pro-

cessing and aircraft orientation is important for RCS measurements. Aircraft motion
is simply range at the engagement-level. The engineering-level includes latitude, lon-
gitude, altitude, and velocity vector. The physics-level contains 6 degree of freedom
along with velocity and acceleration vectors.

42
 Figure 4.7: A discrete set of turbine blade positions used at
the engineering-level [26].

4.2.3.2 Turbine Motion. Turbine motion, like aircraft motion, is im-

portant for Doppler processing and turbine positioning effects RCS measurements.
Turbine motion at the engagement-level is merely the wind farm location. At the
engineering-level, there is a discrete set of blade rotations along with turbine loca-
tion, like in Figure 4.7 [26]. The physics-level model will include the continuous
motion of the blades.

4.2.3.3 Clutter. Clutter is unwanted radar returns from terrain, fo-

liage, and buildings. At an engagement-level, clutter is summarized as a negative RCS
because it reduces the ability to detect an intended target. For the engineering-level,
clutter is modeled as a distribution function [30]. The physics-level clutter model
accounts for range dependent dielectric constants.

4.2.4 EM - Propagation. Wave propagation accounts for factors like atmo-

spheric loss and terrain features to determine how far a signal can travel. Conditions
like rain, fog and foliage will reduce the maximum detection range and produce clut-
ter. At the engagement-level propagation is a straightforward attenuation factor.
The engineering-level model includes some effects like diffraction, LOS, and multi-
path due to ground or atmospheric ducting are included. Figure 4.8 illustrates a

43
 Figure 4.8: A slice of terrain evaluated for LOS [10].

Table 4.1: Sample radar and turbine parameters

Parameter Symbol Value
Wavelength λ 0.25 m
Pulse Repetition Frequency P RF 288 Hz
Azimuth Beamwidth θB 1.5 deg
Radar Scan Rate TR 12 sec
Tower Height ht 80 m
Tower Diameter dt 3m
Blade Length lb 40 m
Blade Rotation Rate ωb 2 rpm

LOS tool that accounts for 4/3rds earth radius and the first Fresnel zone [10]. A
physics-level propagation model is a full parabolic equation that handles all of the
possible engineering-level options. Figure 4.9 shows three different propagation mod-
els. APM and VTRPE are physics-level models while TIREM is an engineering-level
propagation model.

4.3 Understanding Computational Requirements

The following example illustrates how expensive it is to calculate scattering

centers for a wind turbine. The radar parameters are outlined first. Then, the turbine
parameters are generated.

44
Figure 4.9: TIREM is an engineering-level propagation model whereas APM and
VTRPE are physic-level models [27].

45
 A quick example, using the parameters from Table 4.1, shows how many pulses
hit a target in one sweep of a radar. The illumination time ti = TR θB /360 tells how
long the beam illuminates a point target. The number of pulses per illumination is
given by Np = ti P RF . For this example the number of pulses per illumination is
14.4, which is seemingly benign when considered independently. Next, the turbine
RCS measurements are calculated to put the value into perspective.

This example estimates the required number of measurements for different blade
positions. Following Nyquist, the required rotational sampling can be calculated by
φs = (360/60)(ωb/P RF )/2, or 0.21 deg. Turbine blades are usually radially symmet-
ric, which reduces the require number of measurements for different blade positions
to NRCS = 120/φs . Again, the number 576 seems benign on its own. However, that
implies 576 RCS models need to be evaluated using (3.1) to generate a look-up table
for one turbine, at one rotation rate. If physics-level RCS measurements are taken,
then each run could take multiple hours.

Another method for modeling an object’s total RCS

N
X →
− →
−
σ≈ σi e−j k · ri , (4.5)
i=1

relies on a collection of i scattering centers [20]. The turbine blades can be modeled
as a flat disc connected to a rectangle. The total surface area of the turbine A =
πlb2 + (ht − lb )dt is approximately 5146 m2 . The Nyquist sampling area As = (λ/2)2
is 1/64 m2 for the radar. The maximum number of scattering centers Ns c = A/As
is approximately 330,000. These calculations are less complex and faster to compute
than physics-level equations.

4.4 Rationale behind Tip Speed - Wind States

If two turbines have different length blades, 30 m and 50 m, but the same tip
speed, the maximum Doppler shift would be the same for both turbines. However, if
the same two turbines were rotating at the same rate, then the 50 m turbine would

46
 Figure 4.10: The physics-level M&S for EM wave scattering
from a moving target. Maxwell’s equations form the basis of the
EM physics model, but the simulation also requires a moving
target model(aircraft or turbine) [28].

have a higher maximum Doppler shift. In addition, if two turbines have the same
length blades but different gearing ratios, the blades will turn at different rates given
the same wind speed and again produce different Doppler shifts. Tip speed is the
most consistent turbine measurement that relates to maximum Doppler shift. The
one caveat is that relative wind direction has an impact on tip speed and Doppler as
shown previously. Therefore, Table 2.4 will need to be expanded to include relative
wind direction along one axis and tip speed along the other.

47
 V. IMOM and INSSITE Experiments

T his chapter discusses experiments using IMOM and two of SAIC’s in-house
evaluation tools, INSSITE and RF Scene. The first step is to introduce the
tools. Secondly, the results of the experiments are presented. Then, each program
is dissected to compare their functions to the components of the M&S framework.
Finally, their limitations for evaluating wind farms are highlighted.

5.1 IMOM

5.1.1 Overview. IMOM was developed to predict the effect of electronic

support and electronic attack systems on “radar, weapon, communication and passive
detection sites within an electronic order of battle (EOB)” [17]. At a basic level,
IMOM evaluates parametric models for EW systems to determine when radars will
detect an aircraft.

5.1.2 Experimental Results. The results of the four IMOM experiments are
detailed in the following sections.

5.1.2.1 LOS. IMOM handles LOS detection based on a target alti-

tude. There are two elevation modes: mean sea level (MSL) and above ground level
(AGL). MSL is an absolute elevation which lends itself to aircraft positions. AGL is
altitude relative to terrain which fits the analysis of turbines. Unless the terrain is
perfectly flat the two elevation modes will create slightly different LOS results.

Once elevation is determined the actual LOS calculations can be performed. The
LOS calculations are incorporated within propagation algorithms. The calculations
are based on the actual earth radius or 4/3rds radius approximation. The accuracy
of each method is directly related to the resolution of the terrain data. In addition,
some of the propagation models account for diffraction and Fresnel zones to improve
accuracy. Figure 5.1 shows the LOS blockage around King Mountain, TX of a 0.001
m2 target at 200 ft AGL using the terrain integrated rough earth model (TIREM)
over level 1 DTED. The southwest area is mostly obscured at this elevation.

48
 Figure 5.1: Depicts IMOM ’s ability to account for terrain
masking. The 0.001 m2 calibration target was placed at 200 ft
AGL.

5.1.2.2 Aircraft Detection. To explore radar-aircraft-terrain interac-

tions, a low flying, 1500 ft AGL, aircraft with 1 m2 RCS was evaluated. The low
altitude was chosen to increase the effect of terrain interference. Figure 5.2 shows the
results of the simulation. The limiting factor was still terrain masking. If the aircraft
RCS was a factor there would be a circular pattern like in Figure 5.1.

5.1.2.3 Turbine Detection. This analysis is similar to the aircraft

detection. LOS was detected for a much larger target, 10,000 m2 . The results in
Figure 5.3 show the areas that a 200 ft tall turbine would be detected. These results
can be used to tell contractors which areas are ideal, from the radar’s perspective, for
building wind turbines. However, these results don’t provide details for the extent of
impact on aircraft detection.

5.1.2.4 Aircraft-Turbine Interaction. The fourth experiment explores

the possibility of modifying DTED by stitching a wind turbine into the file. Once

49
Figure 5.2: Depicts IMOM ’s ability to detect an aircraft at
1500 ft AGL.

Figure 5.3: Depicts IMOM ’s ability to detect a wind turbine

at 200 ft AGL.

50
 Figure 5.4: Modified level 2 DTED to account for wind tur-
bine shadow zones.

the elevation data is adjusted, LOS calculations account for shadow zones. A first
order approximation adjusts the height of specific key nodes to create a hill. Fig. 5.4
shows a worst case example of the affect of adjusting a couple key elevation nodes.
Fig. 5.5 shows the best case example of using terrain to approximate shadow zones.
As expected, DTED level 2 creates a smaller cross sectional area than DTED level
1 and a much smaller area then DTED level 0. The LOS analysis of shadow zones,
using modified DTED level 2, should provide accurate results for a single turbine.
When turbines are arranged in a wind farm, the best course of action could be to
create an artificial plateau covering the whole farm. In that situation, the effect of
DTED levels would decrease as the farm size increased. However, no physical DTED
files were actually manipulated. An open-source program showed the potential to
modify DTED [1].

The second part of the fourth experiment was the idea of modeling turbines as
a radar jammer. If the turbine is modeled as a continuous noise jammer, it would
reduce the radar detection range in all directions. A self-protection jammer that acts

51
 Figure 5.5: Modified level 1 DTED to account for wind tur-
bine shadow zones.

as a repeater with large gain may be a reasonable approximation. Because of time

constraints this portion was not investigated.

5.1.3 Mission-Level. IMOM is often used as a mission-level model by run-

ning Detection Contours or Detection Radials. The EOB can account for complete
integrated air defense systems (IADS), EW strike packages, and EW support pack-
ages. The scope of some EOBs border on campaign-level.

5.1.4 Engagement-Level. When running a contour or radial analysis, area

of interest and detection grid are specified. The analysis relies on several physics and
engineering-level models: parametric radar, antenna patterns, propagation effects,
aircraft route and RCS, and terrain masking. However, it cannot test whether moving
ground structures are reported as moving aircraft.

5.1.5 Engineering-Level. Radar parametric and propagation models are

two of the strengths of IMOM. The radar, antenna, and receiver subsystems were

52
adequately modeled to be considered engineering-level. Route planning is available
which is an engineering-level version of target motion. The TIREM propagation model
performed LOS, rough atmospheric loss, and clutter effects at an engineering-level.
In addition, IMOM can process a 3D RCS pattern which is necessary for engineering-
level analysis.

5.1.6 Physics-Level. Most components of IMOM operate above the physics-

level. Propagation components are the exception. There are two very accurate models
included with IMOM, variable terrain radio parabolic equation (VTRPE) and ad-
vanced propagation model (APM). APM is a hybrid model that uses different equa-
tions based on elevation and range. VTRPE is a full parabolic equation that provides
the most accurate propagation model [27]. Both models include range dependent
dielectric constants for measuring clutter.

5.1.7 Limitations.

• Parametric model doesn’t provide I/Q data necessary for radar phase, Doppler
shifts.

• Signal processor algorithms aren’t included.

• Only one target can be evaluated at a time, meaning analysis of radar-aircraft-

turbine interaction can’t be done, unless the turbine is literally part of the
clutter.

5.2 INNSITE

5.2.1 Overview. INSSITE is a 3-D environment builder that allows users to

create or adjust terrain, incorporate features, like vegetation, and attach 3-D models
of structures. The scenes developed in INSSITE are evaluated by other tools, like
RF Scene or X Scene. One of the selling points is its simulation ability that allows
developers to view their environment in motion, before running any time consuming
evaluation tools.

53
 Figure 5.6: King Mountain, TX view from the north of the
mesa.

RF Scene is an evaluation tool that operates on scenes created in INSSITE. RF

Scene uses a shooting and bouncing ray (SBR) technique to analyze the RF spectrum
for radar signals at the signal level. That means SBR provides amplitude and phase
data for each ray. The addition of phase data enables the calculation of Doppler shift.
SBR is covered in great depth in Knott [20], including the various diffraction methods
available in RF Scene.

5.2.2 Experimental Results. Because of time constraints and technical dif-

ficulties using RF Scene, the experiments were setup but not evaluated. Figure 5.6
was a scenario with a radar illuminating hilly terrain. The experiment would have
depicted the effect of clutter on a radar. Figure 5.7 shows the turbine and radar in
freespace. Repeated analysis should be performed to determine the optimal obser-
vation grid for acceptable simulation accuracy and time. Figure 5.8 combines the
previous two experiments to include the added effect of multi-bounce. Figure 5.9
mimics the IMOM test showing radar-aircraft-clutter interactions. Figure 5.10 would
have been the capstone experiment showcasing the effect of turbines on probability
of detection.

54
Figure 5.7: Measurements of a turbine in freespace to deter-
mine the most efficient observation grid.

Figure 5.8: Scenario designed to measure the clutter environ-

ment to include a turbine.

55
Figure 5.9: Scenario involving an aircraft 1500 ft above the
mesa.

Figure 5.10: The ideal test scenario, includes radar, aircraft,

turbine and terrain.

56
 5.2.3 Mission-Level. INSSITE is capable of setting up a scenario that
includes multiple objects and sensors. However, simulating a large scene, like a full
farm, could take several days to process.

5.2.4 Engagement-Level. Engagement-level is where INSSITE is best suited,

smaller scenarios where moving objects are visible. It should be capable of analyzing
radar-aircraft-turbine-clutter interactions including Doppler effects.

5.2.5 Engineering-Level. The radar model is simplistic and adequate. The

radar is not as customizable as IMOM ’s radar models, yet it provides the essential
capabilities to produce the requested analysis. RF Scene allows the capture of phase,
RCS, and clutter data at the engineering-level. There are also hooks to call X-scene
for engineering-level RCS data. Targets in INSSITE can be manipulated to various
degrees to include routes for aircraft, turbine blade positions and turbine direction.

5.2.6 Physics-Level. Target models and antenna patterns can be extremely

well detailed in INSSITE. In addition, when RF Scene is run with a tight enough grid
it can produce physics-level phase information. X-Patch is called for physics-level RCS
analysis.

5.2.7 Limitations.

• Elevation angles are not accurate when approaching 0 deg from the horizon.
Elevation angles above the horizon are impossible.

• Requires small scene sizes, approximately one range resolution cell, to evaluate
within an hour.

• Hidden limit of 231 rays, affects maximum scene size and ray density.

• Creating a functioning ground radar is tedious.

• No native tool for creating or modifying antenna patterns.

57
• No native tool can display results of RF Scene. SAIC did provide separate
Matlabr code for this purpose.

• Program stability is an issue.

58
 VI. Conclusions

T his chapter discusses areas for further research and summarizes this research
effort.

6.1 Conclusions

The proposed large-scale modeling and simulation approach will provide a frame-
work for the development of a program to evaluate radar-aircraft-wind farm-clutter
interactions. The ability to select different levels of simulation provides a modular,
scalable analysis tool.

Many of the subcomponents are already developed and in use. Several of the
blocks: EM scattering phase and magnitude data, radar models, propagation equa-
tions, target movement and terrain data are currently generated at the necessary
engineering and physics levels. However, several components still need to be devel-
oped.

Transforming engineering and physics-level scattering data into engagement-

level stochastic scattering information will reduce the complexity of the simulation.
In addition, it will facilitate the analysis of full wind farms rather than single turbines.
Since a subsection of the scattering data is transformed into information, the amount
of physics-level data generated for each turbine is reduced.

Customizable signal processor algorithms and their ordering should be imple-

mented. The effects of clutter cell sizes on detection thresholds should be included.
Algorithms like MTI that reduce the impact of turbines must be included.

The effort to coordinate all the models and simulations across different M&S
tiers is significant. The tool will need to analyze the interference and summarize it
with a display.

Finally, there is a large amount of data that must be generated for this or any
other solution to work. There should be a method to categorize wind turbines into a
discrete number of classes to reduce the data requirements.

59
6.2 Additional Research

There are many opportunities to expand the efforts that were started for this
research. RLSTAP is another radar program that operates between the physics and
engineering-levels to provide RCS phase and magnitude measurements. It may pro-
vide additional validity to the hybrid large-scale modeling and simulation framework.

INSSITE has several components that should be studied further. Providing

analysis of a rotating turbine blade is quite powerful. However, the first step is to get
RF Scene running.

IMOM may be capable of simulating turbine interference through non-traditional

methods. Manipulating the different jammer models could simulate turbine clutter
to produce first order results quickly. The other option is to adjust the elevation of
the terrain data which will at least show the shadow zone effect and part of the RCS
interference.

Signal processor algorithms warrant further investigation. MTI techniques re-

duce the magnitude of turbine returns, there should be additional techniques to re-
duce the Doppler return. There needs to be a large set of doppler data to test the
algorithms.

Finally, the study of wind states as a method of classifying turbine clutter merits
deeper research. It may be capable of supplying the stochastic data missing from the
proposed M&S architecture. Providing a quick look analysis would be very useful.

60
 Bibliography

1. “Geospatial Data Abstraction Library”. http://www.gdal.org.

2. “Obstruction Evaluation/Airport Airspace Analysis”. https://oeaaa.faa.gov.
3. 84 RADES. http://www.rades.hill.af.mil.
4. Air Force Research Laboratory. The Air Force Research Laboratory (AFRL) Mo-
bile Diagnostics Laboratory (MDL) Wind Farm Turbine Measurements Fenner,
NY. Final report, July 20 2006.
5. AIR WARFARE CENTRE RAF LINCOLN (UNITED KINGDOM). The Effects
of Wind Turbine Farms on Air Defence Radars. Technical Report AD Number:
ADA466350, January 2005.
6. AIR WARFARE CENTRE RAF LINCOLN (UNITED KINGDOM). The Ef-
fects of Wind Turbine Farms on ATC Radar. Technical Report AD Number:
ADA467441, May 2005.
7. AIR WARFARE CENTRE RAF LINCOLN (UNITED KINGDOM). Further
Evidence of the Effects of Wind Turbine Farms on AD Radars. Technical Report
AD Number: ADA466373, August 2005.
8. Amato, N.A., P.M. Brand, M.A. Saville, and M.J. Mendenhall. “Modeling Clutter
from Windmill Energy Farms”. Air Force Institute of Technology, 2009.
9. Barton, D.K. Modern Radar Systems Analysis. Artech House, Inc., 1988.
10. Brand, P.M. Modeling of Wind Turbine Doppler Effects on Primary Search
Radar. Ph.D. thesis, Air Force Institute of Technology, 2950 Hobson Way, Wright-
Patterson AFB, OH, 45433, USA, 2009.
11. Brenner, M. and et al. Wind Farms and Radar. Technical Report JSR-08-125,
MITRE Corporation, January 2008.
12. Butler, G. “Advanced Digital Tracker - Update”, May 2007.
http://www.all-energy.co.uk/UserFiles/File/2007GeoffButler.pdf.
13. European Organisation for the safety of Air Navigation. “Assessment Methodol-
ogy to Determine the Impact of Wind Turbines on ATC Surveillance Systems”,
March 2008.
14. European Organisation for the safety of Air Navigation. “Wind Farm Impact
Assessment Techniques and Mitigation Measures”, March 2008.
15. 453rd EWS. http://www.afioc.afisr.af.smil.mil/offices/318/453/453ews/453.cfm.
16. Halliday, D. Fundamentals of Physics Extended. Wiley and Sons, 8th edition,
2008. ISBN 978-0-471-75801-3.

61
17. Idaho National Laboratory. Improved-Many-On-Many (IMOM) Engineer Help.
453rd EWS, Lackland AFB.
18. Jamshidi, M. Large-scale systems: modeling, control, and fuzzy logic. 1997.
19. Kent, B.M. and et al. “Dynamic Radar Cross Section and Radar Doppler Mea-
surements of Commercial General Electric Windmill Power Turbines Part 1: Pre-
dicted and Measured Radar Signatures”. IEEE Antennas and Propagation Mag-
azine, Vol. 50(No. 2):211–219, April 2008.
20. Knott, E.F. and et al. Radar Cross Section. SciTech Pub., 2nd edition edition,
2004.
21. Lemmon, J.J. and et al. Assessment of the Effects of Wind Turbines on Air Traffic
Control Radars. Technical Report NTIA TR: TR-08-454, National Telecommu-
nications and Information Administration, July 2008.
22. Mahmoud, M.S. Multilevel systesm control and applications: A survey. 1977.
23. Marshall, B.K. http://www.geograph.org.uk/photo/667522.
24. NASIC. http://www.naic.wrightpatterson.af.smil.mil/TableView/mission.html.
25. Office of the Director of Defense Research and Engineering. Effect of Windmill
Farms on Military Readiness 2006. Technical report, Congressional Defense Com-
mittees, 2006.
26. QinetiQ Ltd. Wind Farms Impact on Radar Aviation Interests. Technical Report
Report W/14/00614/00/REP, September 2003.
27. REMCOM. ElectroMagnetic Propagation Integrated Resource Environment (EM-
PIRE) Basic Guide to RF Propagation Models. 453rd EWS, Lackland AFB, 2007.
28. Saville, M.A. “Notes on Large Scale Modeling and Simulation”, Feb 2009.
29. Simpson, J. and E. Weiner. “Oxford English Dictionary”, 1989.
30. Skolnik, Merrill I. Introduction to Radar Systems. McGraw-Hill, Inc., 1221 Avenue
of the Americas, New York, NY 10020, third edition, 2001. ISBN: 0072909803.
31. Theil, A. and L.J. van Ewijk. “Radar Performance Degradation due to the Pres-
ence of Wind Turbines”. IEEE, 6, 2007.
32. US Army. Commander’s Appreciation and Campaign Design. Technical Report
Pamphlet 525-5-500, 2008.
33. U.S. Department of Energy. 20% Wind Energy by 2030. Technical report, July
2008.
34. Vestas Wind Systems A/S. “V82-1.65 MW Brochure”.
http://www.vestas.com/en/wind-power-solutions/wind-turbines/1.65-mw.aspx.
35. Wikipedia. http://en.wikipedia.org.

62
36. Wisner, R. and M. Bolinger. Annual Report on U.S. Wind Power Installation,
Cost, and Performance Trends: 2007. Technical Report DOE/GO-102008-2590,
U.S. Department of Energy, May 2008.

63
 Form Approved
REPORT DOCUMENTATION PAGE OMB No. 0704–0188
The public reporting burden for this collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering and
maintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this collection of information, including
suggestions for reducing this burden to Department of Defense, Washington Headquarters Services, Directorate for Information Operations and Reports (0704–0188), 1215 Jefferson Davis Highway,
Suite 1204, Arlington, VA 22202–4302. Respondents should be aware that notwithstanding any other provision of law, no person shall be subject to any penalty for failing to comply with a collection
of information if it does not display a currently valid OMB control number. PLEASE DO NOT RETURN YOUR FORM TO THE ABOVE ADDRESS.

1. REPORT DATE (DD–MM–YYYY) 2. REPORT TYPE 3. DATES COVERED (From — To)

26–03–2009 Master’s Thesis Sept 2007 — Mar 2009
4. TITLE AND SUBTITLE 5a. CONTRACT NUMBER

DACA99–99–C–9999
5b. GRANT NUMBER
Modeling and Simulation Architecture
for Studying Doppler-Based Radar
with Complex Environments 5c. PROGRAM ELEMENT NUMBER

6. AUTHOR(S) 5d. PROJECT NUMBER

N/A
5e. TASK NUMBER
Nicholas J. Amato, Capt, USAF
5f. WORK UNIT NUMBER

7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) 8. PERFORMING ORGANIZATION REPORT

NUMBER
Air Force Institute of Technology
Graduate School of Engineering and Management (AFIT/EN)
2950 Hobson Way AFIT/GE/ENG/09-02
WPAFB OH 45433-7765
9. SPONSORING / MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSOR/MONITOR’S ACRONYM(S)

84 Radar Evaluation Squadron (Shawn Jordan) 84 RADES/SC

7976 Aspen Avenue 11. SPONSOR/MONITOR’S REPORT
Hill AFB, UT 84056-5846 NUMBER(S)
((801)777-3290, [email protected])

12. DISTRIBUTION / AVAILABILITY STATEMENT

Approval for public release; distribution is unlimited.

13. SUPPLEMENTARY NOTES

14. ABSTRACT
This research effort develops a hybrid large-scale modeling and simulation framework that defines the requirements for a
program to evaluate radar-aircraft-turbine-clutter interactions. Wind turbines and other moving structures can interfere
with a radar’s ability to detect moving aircraft because radar returns from turbines are comparable to those from slow
flying aircraft. This interference can lead to aircraft collisions or crashes, reducing the safety for air traffic.
Two radar applications, INSSITE and IMOM, were investigated to determine which of the subsystems, in the
proposed architecture, are currently available and which need additional development. Current radar applications either
delve too deep into details, requiring years to process, or too shallow, ignoring the Doppler effect and assuming a static
scattering value. Engineering-level radar, radiation, propagation, and scattering models are already developed. However,
engagement-level stochastic scattering, amplitude and phase, data aren’t available. The hybrid modeling and simulation
architecture could be realized once stochastic RCS models are developed.
15. SUBJECT TERMS

Wind turbine; large-scale modeling and simulation; pulse Doppler radar; radar cross section; line-of-sight

16. SECURITY CLASSIFICATION OF: 17. LIMITATION OF 18. NUMBER 19a. NAME OF RESPONSIBLE PERSON
ABSTRACT OF Maj Michael Saville
a. REPORT b. ABSTRACT c. THIS PAGE PAGES
19b. TELEPHONE NUMBER (include area code)
U U U UU 63 (937) 255–3636, ext 4719; [email protected]
Standard Form 298 (Rev. 8–98)
Prescribed by ANSI Std. Z39.18