Coastal Carolina University

CCU Digital Commons

Honors Theses HTC Honors College

Spring 2024

Error Analysis of Numerical Weather Prediction Forecasts and

Remotely Sensed Air Properties
Caleb R. Sease
Coastal Carolina University, [email protected]

Follow this and additional works at: https://digitalcommons.coastal.edu/honors-theses

Part of the Atmospheric Sciences Commons, Oceanography Commons, and the Other Physical
Sciences and Mathematics Commons

Recommended Citation
Sease, Caleb R., "Error Analysis of Numerical Weather Prediction Forecasts and Remotely Sensed Air
Properties" (2024). Honors Theses. 493.
https://digitalcommons.coastal.edu/honors-theses/493

This Thesis is brought to you for free and open access by the HTC Honors College at CCU Digital Commons. It has
been accepted for inclusion in Honors Theses by an authorized administrator of CCU Digital Commons. For more
information, please contact [email protected].
Approved, DCN# 543-2217-24 8/13/2024

Error Analysis of Numerical Weather Prediction Forecasts and

Remotely Sensed Air Properties

Caleb R. Sease

Department of Marine Science

Submitted in Partial Fulfillment of the

Requirements for the Degree of Bachelor of Science
In the HTC Honors College at
Coastal Carolina University

Summer 2024

Louis E. Keiner Erin E. Hackett

Director of Honors Professor
HTC Honors College Department of Marine Science
Gupta College of Science

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

Abstract.......................................................................................................................................... 3
1.0 Introduction ............................................................................................................................. 4
2.0 Background ............................................................................................................................. 5
2.1. The Marine Atmospheric Surface Layer ......................................................................... 5
2.2 Surface Ducts and X-band Radar Wave Propagation ..................................................... 6
2.3 Modern Methods of Modeling and Measuring Air Properties within the MASL ........ 7
3.0 Research Objectives .............................................................................................................. 10
4.0 Methods.................................................................................................................................. 11
4.1 Data ..................................................................................................................................... 12
4.2 Analysis Methods............................................................................................................... 16
4.2.1 Modified Refractivity Comparisons.......................................................................... 17
4.2.1.1 Duct Height Differences ..................................................................................... 17
4.2.1.2 Root-Mean-Square Error ................................................................................... 18
4.2.1.3 Altitudes of Maximum Residuals ....................................................................... 19
4.2.1.4 Correlation Coefficients ..................................................................................... 19
4.2.1.5 Time and Range Differences .............................................................................. 20
4.2.2 COAMPS Temperature and Humidity Comparisons ............................................. 20
5.0 Results and Discussion.......................................................................................................... 21
5.1 Remotely Sensed Refractivity .......................................................................................... 21
5.2 COAMPS Refractivity ...................................................................................................... 29
5.3 Time and Range Effects on Refractivity Errors ............................................................. 33
5.4 COAMPS Temperature and Humidity ........................................................................... 36
6.0 Conclusions ............................................................................................................................ 42
References .................................................................................................................................... 45

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

Abstract

Communication and sensing technologies are influenced by atmospheric properties such

as the air’s index of refraction, humidity, and temperature in the marine atmospheric surface layer

(MASL). Recent research has demonstrated that radar can remotely sense the index of refraction

in the MASL and, sometimes, numerical weather prediction models (NWP) can be used to aid this

process. These developments are important because they allow for larger spatial coverage and finer

temporal resolution than is possible with in-situ atmospheric sensors. More recently, on-going

research is working toward extending this remote sensing capability to humidity using refraction

as an intermediary. In order to advance these remote sensing methods an assessment of the

accuracy of remotely sensed refractivity and NWP forecasts is needed. As such, this study focuses

on determining the errors of NWP forecasts of refractivity, temperature, and humidity, and

remotely sensed refractivity based-on statistical comparisons to direct meteorological

measurements during the Coupled Air–Sea Processes and Electromagnetic Ducting Research-East

(CASPER-East) field campaign. These analyses enable an overall assessment of the NWP

forecasts and refractivity that can be used to test and evaluate on-going research aimed at extending

remote-sensing technologies.

Error statistics show that a majority of NWP and remotely sensed refractivity estimates are

accurate (low root-mean-square errors and high correlation coefficients) relative to in-situ

atmospheric measurements. When large errors do occur, they often occur during highly unstable

atmospheres, which are abnormally dry relative to the other time periods during the CASPER-East

field campaign. Furthermore, NWP temperature forecasts, in general, contain less error than

humidity forecasts. Thus, because humidity has the largest effect on refraction, errors on NWP

refraction are likely driven by inaccurate humidity predictions. Moreover, spatiotemporal

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

differences between the NWP or remotely sensed refractivity and in-situ measurements had no

statistically significant linear relationship to error magnitude.

1.0 Introduction

Modern meteorological modeling and measurement techniques provide valuable information

about the surrounding environment. These techniques provide necessary data for military,

scientific, and civilian applications. However, these technologies have various pitfalls, including

low resolution, poor coverage, measurement uncertainties, and contain assumptions that can lead

to inaccuracies. These issues are especially challenging in near-surface marine regions because the

physical dynamics cause rapid changes over fine temporal and spatial scales, which makes both

meteorological measurements and modeling difficult. Therefore, there has been various research

efforts to improve meteorological modeling and measurement techniques, especially with respect

to the application of remote sensing technologies in the marine environment.

Remote sensors are desirable when measuring the atmosphere in a marine environment as they

can provide data without being physically in contact with the entity they are observing, which has

a number of logistical and economic benefits. Radar has been explored as a remote sensor to

estimate atmospheric refractivity because refractivity has one of the largest effects on radar wave

propagation. However, refractivity is only one property of the complex atmosphere, and while it

has been successfully estimated using point-to-point radar remote sensing, humidity and

temperature have not been widely explored using similar techniques in the near-surface region. As

such, understanding the accuracy of current methods that estimate refractivity, temperature, and

humidity should help inform and verify the validity of these future techniques.

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

Section 2.0 provides relevant background information for this study. The research objectives

for this study are described in section 3.0. Following the objectives, Section 4.0 describes the

methods, including both the data sources and error statistics used in the study. Section 5.0 analyzes

and discusses the results. Finally, a conclusion of the findings is provided in Section 6.0.

2.0 Background
The following sections provide background information on the marine atmospheric surface

layer (MASL), evaporation ducts and their effect on radar remote sensing, and a new approach to

remote sensing. This background provides an overview of the foundational knowledge to

understand the relevance and importance of this error analysis as it pertains to making accurate

retrievals of humidity from current radar remote sensing techniques.

2.1. The Marine Atmospheric Surface Layer

The Earth’s atmosphere consists of many layers each with their own vertical temperature

gradient, that notionally alternates in sign throughout Earth’s atmosphere. The troposphere is the

first layer of the atmosphere where temperature typically decreases with altitude. Temperature and

pressure gradients within the troposphere cause the movement of air creating atmospheric

circulation patterns including localized weather phenomena. Within the troposphere exists a

boundary layer that is influenced directly by the Earth’s surface, known as the planetary boundary

layer. This layer extends approximately 1 km above the Earth’s surface, but its height can vary

depending on surface meteorological conditions. This layer, when overlying the ocean is called the

marine atmospheric boundary layer. Within the first ~10 percent or ~100 m of the marine

atmospheric boundary layer, exists the MASL. This layer is further impacted by physical processes

occurring at the air-sea boundary, including, but not limited to, ocean surface waves, wind shear,

and surface fluxes (Liu et al., 2022). Flow within the MASL is turbulent in nature and changes on

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

a vast range of spatiotemporal scales including small spatiotemporal scales that can often be

neglected in other regions of the atmosphere. This range of scales is challenging to measure and

model and thus, study of the MASL is an open and on-going area of research. As such, this study

is focused on measurements and modeling within the MASL.

2.2 Surface Ducts and X-band Radar Wave Propagation

Naval vessels and commercial ships commonly operate within the MASL and rely on the

transmission of electromagnetic (EM) waves to understand their surroundings and uncover

potential threats. The MASL can have a large impact on the propagation of EM waves. A

phenomenon known as an evaporation duct (ED), which is caused by negative humidity gradients

between the marine surface and the first tens of meters in the atmosphere, exists at most times

within the MASL, occurring over 95 % of the ocean (Hitney et al., 1985; Babin, 1997). This

negative humidity gradient leads to a negative modified refractivity gradient (which is atmospheric

refractivity adjusted for the curvature of the Earth), which defines a duct. Although the occurrence

of EDs is common, another type of duct referred to as a surface-based duct (SBDs) can also occur

due to temperature inversions near the sea surface caused by either cool ocean waters, or the

advection of cool, often dry, air over a body of water (Paulus, 1989). Therefore, a negative

modified refractivity gradient can form in conjunction or separately from an ED. A strong enough

negative modified refractivity gradient of either ducting scenario can cause EM waves to refract

and reflect or scatter off the ocean surface, which are subsequently refracted back to the surface in

a phenomenon known as trapping. Trapping can extend radar ranges far beyond expected limits

and lead to holes in radar detection above the duct (Skolnik, 1962). Therefore, it is necessary to

accurately quantify the existence and height of EDs for accurate EM propagation predictions. And

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

conversely, due to the same sensitivity, they can be used to determine properties of the air by radar

remote sensing.

Quantification of EDs or SBDs is commonly done via calculation of modified refractivity,

which is related to temperature, pressure, and humidity. Using the equation from Bean & Dutton

(1968) we can quantify the effect of these factors on modified refractivity:

77.6 4810𝑒 𝑧
𝑀= [𝑝 + ] + ( ) × 106 (1)
𝑇 𝑇 𝑅𝑒

where T is temperature in Kelvin, p is pressure in millibars, z is altitude in meters, e is partial water

vapor pressure in millibars (which can be calculated using specific humidity (q) and atmospheric

pressure), and Re is the radius of the Earth in meters. At typical temperatures and humidities within

the MASL, e is often considered to have the largest effect on atmospheric refractivity (Skolnik,

1962). Modified refractivity is particularly useful when exploring SBDs or EDs because ducting

regions occur anytime M decreases with height.

2.3 Modern Methods of Modeling and Measuring Air Properties within the MASL

As stated previously, measuring and modeling atmospheric properties within the MASL

can be challenging due to the complex physical processes that occur throughout this region.

Current methods include use of numerical weather prediction (NWP) models and/or in-situ

atmospheric sensors. Modern NWP models, such as the Coupled Ocean/Atmosphere Mesoscale

Prediction System (COAMPS®; Hodur, 1997) provide excellent forecasting capability but suffer

from poor spatial and temporal resolution due to the computational power required to run forecasts

continually (Haack et al., 2010). The low vertical resolution of NWP models can be particularly

problematic for modeling properties in the MASL because it changes rapidly both spatially and

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

temporally. Moreover, these predictions rely on approximations for various unresolved physical

processes. Specifically, approximations regarding atmospheric and ocean mixing, shear, and

turbulent processes must be applied (Hodur, 1997). NWP simulations are initialized using initial

boundary conditions, and subsequently solve the Navier-Stokes equations to provide regional or

global forecasts on a prescribed grid. Often boundary and initial conditions are poorly known or

are inaccurate. All of these limitations can lead to inaccuracies in NWP forecasts.

In-situ atmospheric sensors are another method of obtaining atmospheric properties most

commonly using radiosondes or weather stations. These sensors are most often affixed to weather

balloons, buoys, or towers. Due to fiscal considerations, in-situ measurements are often performed

at single discrete locations with sparse sampling periods resulting in poor temporal and spatial

coverage (Pastore et al., 2021). The size of the sensor housing and mounting apparatus often

prevents measurements very close to the surface (< 1 m), where humidity gradients are largest

(Rainer, 2016). Moreover, measured data has uncertainty associated with it, which can further be

amplified by data extrapolation and fitting measurements to models. Nonetheless, in-situ

measurements are generally considered the most accurate representation of the atmosphere at the

location of the sensor at the time of measurement.

Due to the limitations of modern mereological measuring and modeling techniques,

research regarding the remote sensing of atmospheric properties has become prominent in recent

decades. There have been several methods concerning the remote sensing of humidity and

temperature. Ground based microwave radiometers have been used extensively to gather relative

humidity and temperature data. However, they cannot be used during rainy or uncertain weather

conditions, as a water film on the radiometer will greatly increase error (Che et al., 2018). In marine

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

environments, uncertain weather conditions can be prominent, meaning this technique may be less

applicable for marine system applications. Moreover, these radiometers are often used to quantify

the occurrence and size of clouds, which are at much higher altitude and are a larger scale

phenomena than that associated with the MASL. The use of wireless, passive remote sensors to

measure the atmosphere is an advantageous technique in regions not easily accessed (Ren et al.,

2015). However, these remote sensors have flaws akin to radiosondes and other in-situ methods.

They have low spatial resolution as well as relatively high measurement uncertainty. Conversely,

radar wave propagation has been used to gather humidity in the atmosphere as seen in Cooper et

al. (2021). These methods use G-band radar, which are not commonly available. These high

frequency waves estimate humidity based on reflection, typically occurring when the EM wave

comes in contact with a cloud. Similar to radiometers, this technique is not particularly useful

when attempting to gather humidity within the MASL. Although many limitations currently exist

as reported throughout literature, modified refractivity’s large dependence on humidity has

important implications regarding modern remote sensing techniques.

It has been shown in Pastore et al. (2022) that inversely determining modified refractivity

from radio frequency (RF) signals can be an effective method of remote sensing refractivity. The

RF data used in Pastore et al. (2022) was collected using bi-static X-band radar. Other authors have

also performed similar RF inversions for refractivity (Pozderac et al., 2018) and numerous authors

have estimated refractivity from sea clutter (RFC; Karimian et al., 2011; Karimian et al., 2013).

Generally, the point-to-point (bi-static radar) inversions tend to be more accurate. This better

accuracy may be because they do not require the estimation of sea surface scattering coefficients

that are needed for the RFC technique, which are typically difficult to predict (Plant & Irisov,

2017). Pastore et al. (2022) show that refractivity inversions estimate refractivity at least as well

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

as NWP forecasts, and are in relative agreement with in-situ measurements. Furthermore, the

point-to-point refractivity inversions enable high vertical resolution estimates of refractivity within

the MASL. This resolution is necessary to retrieve refractivity gradients within the first several

meters of the atmosphere. The humidity gradients are what cause EDs, which are largely the cause

of the refractivity gradients. Therefore, the ability for X-band frequency, point-to-point radar

inversions to accurately estimate refractivity gradients may allow for the retrieval of humidity

gradients using refractivity as an intermediary. This research is ongoing and aimed at obtaining

humidity using X-band RF data. To develop these techniques, it is important to compute error

statistics for the remotely sensed refractivity and NWP forecasts that may be used to verify or aid

humidity retrievals. The statistical analysis of remotely sensed atmospheric refractivity enables

better interpterion of inaccuracies in humidity retrievals. Moreover, the statistical analysis of NWP

forecasts is necessary as they may be used to either constrain humidity retrievals or verify them.

For example, if NWP forecasts are inaccurate and the humidity retrievals are constrained by them,

it may cause inaccuracies in the humidity retrievals. Similarly, if NWP humidity estimates are used

to verify humidity retrievals, it is important to understand when their accuracy is both high and

low.

3.0 Research Objectives

This study aims to evaluate the accuracy of NWP forecasts and refractivity inversions

based on measured radar data from the CASPER-East field campaign. Remotely sensed modified

refractivity from Pastore et al. (2022) is analyzed for evaluating its use in the remote sensing of

humidity using radar-derived modified refractivity as an intermediary. The evaluation of NWP

forecasts will provide insight into the model’s performance as well as aid in evaluating the use of

10

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

NWP forecasts to constrain and verify developing humidity remote sensing techniques. The

specific objectives of this research are:

i. Quantify the accuracy of twelve radar inversion datasets relative to the closest, in time

and space, direct atmospheric measurements

ii. Quantify the accuracy of thirty-six NWP forecasts relative to the closest, in time and

space, direct atmospheric measurements

Error statistics considered include root-mean-square errors, correlation coefficients,

residual magnitudes, and evaporation duct height (EDH) differences. Moreover, the error statistics

will be evaluated to gather insight into the effect of space and time differences on the accuracy of

NWP forecasts and radar refractivity inversions. Understanding the accuracy of NWP forecasts

and radar-derived refractivity will help inform test cases for use in the remote sensing of humidity

in order to properly interpret the impact of inaccuracies in M and T on humidity retrievals.

4.0 Methods

Section 4.1 explains all data utilized in this study including the numerical weather

prediction (NWP) refractivity, temperature, and humidity forecasts, remotely sensed refractivity,

as well as the in-situ data that is used as reference for the comparisons. Section 4.2 explains how

the error statistics are computed as well as how they are used to evaluate the numerical weather

prediction forecasts and remotely sensed refractivity estimates.

11

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

4.1 Data

The Coupled Air Sea Processes and Electromagnetic Ducting Research - East field

Campaign (CASPER-East) was held off the coast of Duck, North Carolina from October 10 to

November 6, 2015 (Wang et al., 2018). The goal of CASPER-East was to curate a comprehensive

dataset to support modern research regarding EM propagation in ducting environments as well as

support in the improvement of NWP model performance in the MASL. This field campaign

included collection of RF and in-situ atmospheric and oceanic measurements, as well as simulation

of NWP forecasts covering the timeframe of the field campaign.

In-situ atmospheric measurements were gathered during the CASPER-East field campaign

using the Marine Atmospheric Profiling System (MAPS). MAPS data were collected utilizing a

small workboat deployed from the larger research vessel (R/V) Sharp. This smaller vessel helped

negate any major effects the research vessel may have had on the surrounding atmosphere. A

radiosonde attached to a tethered weather balloon completed seven ascents and descents from the

small workboat. The radiosonde measured air temperature, pressure, and relative humidity, which

were subsequently converted to modified refractivity (eq. 1). These measurements were fit using

a seventh order polynomial for each respective variable and measurement set (Wang et al., 2018;

Kang and Wang, 2016). Thus, these MAPS datasets consist of 36 modified refractivity,

temperature, and relative humidity vertical profiles that were fit to data collected from October 13

to November 1, 2015 with two to four datasets taken a day sporadically. Table 1 shows the details

of the datasets used in the study. Fig. 1 shows an example MAPS refractivity dataset (MAPS 1).

12

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

Table 1: Position and time of MAPS measurements, COAMPS forecasts and radar measurements
used in the refractivity inversion estimates (UTC). Position is shown as distance offshore of Duck
Pier, in Duck, North Carolina. Maximum radar (inversion) range is also referenced to Duck Pier.

Max
MAPS MAPS COAMPS Inversion
Time Distance COAMPS Distance Inversion Time Range
Dataset (UTC) (km) Time (UTC) (km) (UTC) (km)
10/13/2015 10/13/2015 10/14/2015 00:57 -
MAPS 1 16:49 4.3 17:00 2.0 03:20 38.8
10/13/2015 10/13/2015 10/14/2015 00:57 -
MAPS 2 19:47 30.2 20:00 28.6 03:20 38.8
10/14/2015 10/14/2015 10/15/2015 00:01 -
MAPS 3 13:06 28.1 13:00 26.6 01:53 38.6
10/14/2015 10/14/2015 10/15/2015 00:01 -
MAPS 4 16:42 3.9 17:00 2.0 01:53 38.6
10/14/2015 10/14/2015 10/15/2015 00:01 -
MAPS 5 19:32 27.7 20:00 26.6 01:53 38.6
10/15/2015 10/15/2015 10/15/2015 17:48 -
MAPS 6 12:22 28.1 12:00 26.6 19:09 37.8
10/15/2015 10/15/2015 10/15/2015 17:48 -
MAPS 7 15:03 3.3 15:00 2.0 19:09 37.8
10/16/2015 10/16/2015 10/16/2015 23:23 -
MAPS 9 13:12 28.1 13:00 26.6 10/17/2015 01:19 76.2
10/16/2015 10/16/2015 10/16/2015 23:23 -
MAPS 10 16:16 3.2 16:00 2.0 10/17/2015 01:19 76.2
10/16/2015 10/16/2015 10/16/2015 23:23 -
MAPS 11 18:38 29.9 19:00 28.6 10/17/2015 01:19 76.2
10/17/2015 10/17/2015 10/17/2015 17:57 -
MAPS 12 12:07 28.3 12:00 26.6 19:09 38.8
10/17/2015 10/17/2015 10/17/2015 17:57 -
MAPS 13 15:04 3.8 15:00 2.0 19:09 38.8
10/17/2015 10/17/2015 10/17/2015 17:57 -
MAPS 14 18:06 29.6 18:00 28.6 19:09 38.8
10/17/2015 10/17/2015 10/17/2015 22:19 -
MAPS 15 22:25 28.5 22:00 26.6 10/18/2015 00:08 39.9
10/20/2015 10/20/2015 10/20/2015 19:42 -
MAPS 16 13:41 28.9 14:00 26.6 21:09 38.6
10/20/2015 10/20/2015 10/20/2015 19:42 -
MAPS 17 16:57 3.9 17:00 2.0 21:09 38.6
10/20/2015 10/20/2015 10/20/2015 19:42 -
MAPS 18 19:29 30.4 19:00 28.6 21:09 38.6
10/21/2015 10/21/2015 10/21/2015 16:34 -
MAPS 19 11:01 43.6 11:00 42.9 18:54 55.5
10/21/2015 10/21/2015 10/21/2015 16:34 -
MAPS 20 12:57 28.3 13:00 26.6 18:54 55.5

13

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

Max
MAPS MAPS COAMPS Inversion
Time Distance COAMPS Distance Inversion Time Range
Dataset (UTC) (km) Time (UTC) (km) (UTC) (km)
10/21/2015 10/21/2015 10/21/2015 16:34 -
MAPS 21 14:29 16.3 14:00 14.3 18:54 55.5
10/21/2015 10/21/2015 10/21/2015 16:34 -
MAPS 22 16:10 3.3 16:00 2.0 18:54 55.5
10/23/2015 10/23/2015
MAPS 23 12:43 63.6 13:00 61.3
-- --
10/23/2015 10/23/2015
MAPS 24 15:44 77.7 16:00 61.3
-- --
10/24/2015 10/24/2015
MAPS 25 12:05 61.7 12:00 61.3
-- --
10/25/2015 10/25/2015
MAPS 26 15:25 28.5 15:00 26.6
-- --
10/25/2015 10/25/2015
MAPS 27 18:11 16.4 18:00 14.3
-- --
10/25/2015 10/25/2015
MAPS 28 19:58 4.0 20:00 2.0
-- --
10/25/2015 10/25/2015
MAPS 29 22:28 28.5 22:00 26.6
-- --
10/31/2015 10/31/2015
MAPS 30 21:19 110.9 21:00 61.3
-- --
11/1/2015 11/1/2015
MAPS 32 12:31 139.4 13:00 61.3
-- --
11/1/2015 11/1/2015
MAPS 33 14:57 135.4 15:00 61.3
-- --
11/1/2015 11/1/2015
MAPS 34 17:30 122.2 18:00 61.3
-- --
11/1/2015 11/1/2015
MAPS 35 19:39 114.6 20:00 61.3
-- --
11/1/2015 11/1/2015
MAPS 36 21:32 107.5 22:00 61.3
-- --

14

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

(a) (b)

Figure 1: Example MAPS modified refractivity dataset 1 (a) and 13 (b), measured on 13 October
2015, 16:49 and 17 October 2015, 15:04 UTC respectively, compared to the closest in time and
range COAMPS (a) and inversion-based (b) modified refractivity.

NWP forecasts were simulated for every hour during the CASPER-East field campaign up

to ~60 km offshore Duck Pier with approximately 2 km horizontal spacing and ~32 m of vertical

resolution. These forecasts were predicted using the Coupled Ocean/Atmosphere Mesoscale

Prediction System (COAMPS®). The forecasts are initialized using atmospheric boundary

conditions from the Navy Global Environmental Model, and oceanic boundary conditions from

the Global Hybrid Coordinate Ocean model (Hogan et al., 2014; Halliwell et al., 1998). However,

COAMPS does not have the vertical resolution necessary to resolve EDs (Karimian et al., 2013).

Therefore, the COAMPS model results were blended with a surface layer model, specifically, the

Navy Atmospheric Vertical Surface Layer Model (NAVSLaM; Frederickson, 2015) to increase the

resolution of the forecasts in the lower 100 m of the atmosphere. These blended COAMPS-

NAVSLaM forecasts of temperature, specific humidity, and modified refractivity (which hereafter

will be referred to as COAMPS forecasts for simplicity) are used in this study and have decimeter

vertical sampling. An example COAMPS modified refractivity vertical profile is shown in Fig. 1a

15

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

along with the closest MAPS dataset in space and time. The details of the forecasts used in this

study are shown in Table 1.

The CASPER-East field campaign also involved data collection of multiple point-to-point

RF datasets, specifically, shore-to-ship RF signals from approximately 3 to up to 60 km offshore,

which took one to four hours to acquire, and are presented in Table 1. The signal was transmitted

from Duck Pier at 10.7 GHz frequency (X-band) from between 10.3 to 11.1 m above sea level

depending on the tide. The receivers were located at 4.0, 6.0, 9.2, and 12.3 m above sea level and

were mounted on the A-frame of the R/V Atlantic Explorer as it drove out in range away from

Duck Pier (Wang et al., 2018). Twelve RF datasets were used to inversely determine modified

refractivity via a genetic algorithm as described in Pastore et al. (2022). These twelve refractivity

inversions span over nine days and each occurred within at least twelve hours of MAPS in-situ

measurements. The surface refractivity used in the inversions in Pastore et al. (2022) was chosen

arbitrarily to be 333 M-units, therefore, each profile was shifted by the difference between 333 M-

units and the surface refractivity of each MAPS profile such that the surface refractivity matches.

These remotely sensed refractivity vertical profiles are evaluated in this study and have 0.1 m

vertical spacing. An example remotely sensed refractivity vertical profile is shown in Fig. 1b. For

more information regarding the inversion process refer to Pastore et al. (2022).

4.2 Analysis Methods

This study uses error statistics to assess the accuracy of COAMPS forecasts and inversely-

determined modified refractivity. We consider the MAPS in-situ measurements as the “true”

atmosphere for this study, but due to polynomial fitting, measurement uncertainty, and

spatiotemporal variation of the MAPS measurements (relative to the NWP forecasts and

inversions), MAPS measurements could deviate from the “true” atmosphere at the time of the RF
16

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

data acquisition or the NWP forecast. Nevertheless, it seems most logical to evaluate the accuracy

against direct in-situ measurements obtained during the CASPER-East field campaign.

The closest COAMPS forecast and RF (or remotely sensed) refractivity dataset to each

MAPS measurement time and location are used in this study (see Table 1). COAMPS forecasts all

occur within an hour of each MAPS dataset and are selected based-on the profile closest in range

compared to each MAPS dataset. Unlike COAMPS forecasts, RF data were recorded only once or

twice daily, therefore, there are occasional time differences of more than eleven hours between

inversely determined refractivity and direct MAPS measurements. Twelve inversions are

compared with 36 MAPS profiles, meaning that there are typically multiple MAPS datasets that

can be compared to one inversion result, allowing for the analysis of the relationships between

measurement time difference and the accuracy statistics discussed in this section. Conversely, no

position difference is considered for the inversely determined modified refractivity because it

represents a range of RF data extending from 3-60 km offshore; therefore, a MAPS profile was

always measured within this range.

4.2.1 Modified Refractivity Comparisons

Modified refractivity is compared using several error metrics that are described in the

subsequent subsections.

4.2.1.1 Duct Height Differences

The duct height occurs where there is a change in the sign of the M profile slope or where

the vertical gradient is zero, and also typically occurs at the minimum modified refractivity for an

ED. The ED height has one of the largest effects on EM propagation, especially at the X-band

frequency (Pastore, 2023; Skolnik, 1962). Therefore, accurate prediction of the duct height by the

17

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

inversion and COAMPS forecasts is necessary for the accurate prediction of EM propagation. In

this study, the duct height is determined as the minimum modified refractivity for each modified

refractivity profile: COAMPS, MAPS, and inversion. In the presented statistics, 𝜆 represents either

the COAMPS or inversion modified refractivity vertical profile, and 𝜏 the MAPS modified

refractivity vertical profile. z is altitude. The duct height difference (D0) is defined:

𝐷0 = |𝑧𝜆 − 𝑧𝜏 |
𝑧𝜆 = 𝑧(min(𝜆)) (2)
𝑧𝜏 = 𝑧(min(𝜏))

4.2.1.2 Root-Mean-Square Error

Root-mean-square error (RMSE) between each COAMPS or inversion modified

refractivity profile and the MAPS refractivity is computed:

1/2
(𝜆𝑖 −𝜏𝑖 )2
Γ = (∑𝑛𝑖=1 ) (3)
𝑛

where n is the number of altitudes. This metric provides an estimate of the mean difference between

the predicted values (COAMPS/inversion) and the expected values (MAPS). RMSE was

calculated for the entire profile (Γ) as well as only at altitudes below the duct height (𝛤𝐵 ), where

the duct height used is the larger duct height among the two compared profiles. This restricted

altitude range helps highlight differences within the ED as opposed to the mixed layer above it.

Due to the large effect the ED has on propagation, it is important to quantify the performance of

the NWP forecasts and inversion at representing the ED.

18

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

4.2.1.3 Altitudes of Maximum Residuals

Residuals are calculated for every altitude in the COAMPS/inversion to MAPS modified

refractivity comparison. The residual is defined as |𝜆𝑖 − 𝜏𝑖 |. The altitude of the maximum residual

is determined for each COAMPS/inversion to MAPS comparison:

Ω = 𝑧(max (|𝜆𝑖 − 𝜏𝑖 |)) (4)

The altitude of the largest residual is computed both over the entire profile (𝛺) and also

over only the altitudes below the larger duct height (𝛺𝐵 ). The largest residuals provide insights

into the performance of the inversion and COAMPS forecasts. For example, if the altitude of the

largest residual occurs at the maximum altitude (𝛺=60 m), then that would suggest the main

discrepancy is with divergent mixed layer slopes, which is not as impactful on EM propagation

predictions as differences within the ED (Lentini and Hackett, 2015; Pastore and Hackett, 2024).

Similarly, the altitude with the largest residual below the duct height being at the surface (𝛺𝐵 =0

m) would suggest the ED is well represented with surface estimates being most inaccurate.

4.2.1.4 Correlation Coefficients

Correlation coefficients provide insights into the overall effectiveness of the

inversion/COAMPS to predict the trend/shape of the MAPS modified refractivity profiles with

altitude. The Pearson correlation coefficient is computed as:

𝑛 ̅
1 𝜆𝑖 −𝜆 𝜏𝑖 −𝜏̅
𝑅= ∑ ( )( ) (5)
𝑛−1 𝑖=1 𝜎𝜆 𝜎𝜏

where an overbar indicates a mean and 𝜎 represents standard deviation for the subscript variable.

This metric can be considered as one of the most important statistics in analyzing the performance

19

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

of the inversion/COAMPS as high positive correlation means the profiles change in the same way

vertically, suggesting similar gradients, while inverse correlation indicates that the vertical trends

of the two profiles oppose each other potentially indicating very different vertical gradients. Owing

to the fact that the vertical gradient of refractivity is what impacts propagation, this statistic is

especially useful in evaluating the accuracy of the profiles in terms of EM propagation predictions.

4.2.1.5 Time and Range Differences

The time difference between COAMPS and MAPS are calculated as the difference between

the start time of the MAPS measurements and the start time of the COAMPS forecast. For the

inversion to MAPS time difference, the inversion time used for comparison is the mid-time of the

RF data collection. Therefore, to calculate the time difference, the MAPS measurement start time

is subtracted from the middle time of the RF data. This calculation of time difference ensures

MAPS datasets collected immediately following RF data collection are considered as close in time

as a MAPS dataset collected immediately preceding the RF data collection. Because the MASL is

a turbulent region where air-sea interactions and dynamic processes are constantly shifting the

measured atmospheric properties in both space and time, we hypothesize that a potential cause of

discrepancies could be the time and space discrepancy between the COAMPS/inversion and

MAPS direct measurements. Therefore, linear regressions are computed and coefficients of

determination (𝑅2 ) are calculated for the difference in time or position versus RMSE, correlation

coefficients, and D0. Recall for the inversion, only the influence of time difference can be

quantified as all MAPS profiles occurred within the range of the RF measurements.

4.2.2 COAMPS Temperature and Humidity Comparisons

Because temperature may be used to constrain humidity retrievals and COAMPS forecasts

of humidity may be used to evaluate the results of humidity retrievals, it is important to also
20

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

examine the accuracy of COAMPS temperature and humidity relative to MAPS measurements.

The RMSE, altitudes of maximum residuals, and correlation coefficients are also computed for

COAMPS and MAPS temperature (K) and specific humidity (g/kg) vertical profiles. The same

equations (eq. 2, 3, 4, and 5) apply except with 𝜆 and 𝜏 representing either temperature or specific

humidity rather than modified refractivity for COAMPS and MAPS, respectively. The assessment

of COAMPS performance in respect to temperature and specific humidity can also provide

understanding into what factors control the accuracy of the modified refractivity profiles, as

modified refractivity is calculated via humidity, temperature, and pressure (eq. 1) (Bean & Dutton,

1968).

5.0 Results and Discussion

In this section, the results of the error analysis are discussed. Sections 5.1 and 5.2 address

the results of the error analysis of remotely sensed (inversion) and COAMPS refractivity,

respectively. Section 5.3 examines the impact of spatiotemporal differences on these refractivity

errors, and Section 5.4 presents results of the error analysis of COAMPS temperature and humidity.

5.1 Remotely Sensed Refractivity

Results for RMSEs both over the entire profile as well as below the duct height are shown

in Table 2 and Fig. 2. The majority of RMSEs are below 5 M-units, which is reasonably accurate.

RMSE below the duct height can give insights into the accuracy of the inversion. A low RMSE

below the duct height (red) is notable because EM propagation is most affected by the duct region

of the profile. In contrast, large RMSE for the entire profile (black) could relate to either

discrepancies within the duct or the mixed layer. RMSE is frequently similar but slightly smaller

below the duct height indicating that some of the discrepancy is coming from the mixed layer

21

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

region of the profile, which is less important for propagation predictions (Pastore et al., 2024), but

a significant portion of the differences stem from the ducting region of the profile. These relatively

large RMSEs below the ED height (in comparison to the entire profile) are due to near-surface

differences between inversion-based refractivity and MAPS. This near-surface region, i.e.,

between the surface and ~0.5 m, is still an active area of research and the shape of this region is

still not fully understood (Stanek et al., 2023). Also, it is evident that MAPS 9 through 16 shows

larger RMSEs, which could be related to the meteorological conditions at the time of data

collection. These datasets were collected from October 16 to October 20, 2015, when the lowest

humidity throughout all the MAPS profiles was measured at 12 m, reaching levels less than 5 g/kg,

indicating that large humidity gradients are occurring very close to the surface because the

humidity must be high at the surface. Moreover, wind speed fluctuated significantly from 0 to 15

m/s during this time. The atmosphere was also very unstable as reported by Wessinger et al. (2024),

with an air sea temperature difference reaching less than -5 degrees Celsius. A cold front arrived

on October 18, explaining the higher windspeeds, atmospheric instability, and dry air, which can

be brought into the measurement region by cold air advection associated with the frontal system

(Wang et al., 2018). The lowest RMSE over the entire profile occurs for MAPS dataset six (3.35

M-units); this profile is shown in Fig. 3a. Conversely, MAPS 15 showed the highest RMSE (31.82

M-units), and is shown in Fig. 3b. MAPS 15 has multiple changes in slope, which typically does

not occur in evaporation ducts. The inversion process is not designed to model these anomalies,

leading to the large RMSE. Additionally, the multiple changes in gradient could be an artifact of

the polynomial fitting, which is not a valid representation of the true atmosphere (Fig. 3b)

22

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

Table 2: Error statistics for remotely sensed refractivity and COAMPS (see section 4.2 for a
description of the statistics shown).

Duct
Height 𝛤 𝛤𝑩
Duct 𝜴 (m) 𝜴𝑩 (m) (m) 𝑫𝟎 (M-units) (M-units) R
height COAMPS COAMPS COAMPS COAMPS COAMPS COAMPS COAMPS
(m) / / / / / / /
Dataset MAPS Inversion Inversion Inversion Inversion Inversion Inversion Inversion
3.0 3.0 9.9 0.2 8.0 8.2 0.88
MAPS / / / / / / /
1 9.7 0.4 0.4 9.1 0.6 3.5 1.3 0.93
4.0 4.0 11.3 33.4 4.5 3.5 -0.25
MAPS / / / / / / /
2 44.7 0.6 0.6 9.1 35.6 6.4 4.1 -0.72
3.1 3.1 8.1 2.0 5.8 6.7 0.79
MAPS / / / / / / /
3 10.1 60.0 0.8 12.8 2.7 4.5 3.6 0.91

1.0 1.0 5.1 3.7 10.5 9.5 0.80

MAPS / / / / / / /
4 8.8 60.0 0.4 12.8 4.0 2.7 1.9 0.92
2.0 2.0 6.7 6.3 14.0 13.3 0.76
MAPS / / / / / / /
5 13 60.0 0.7 12.8 0.2 3.4 3.0 0.92
7.0 7.0 10.2 1.8 8.2 9.3 0.66
MAPS / / / / / / /
6 12 60.0 0.3 5.3 6.7 1.0 0.6 0.91
60.0 1.9 10.0 2.0 7.1 8.2 0.69
MAPS / / / / / / /
7 8 0.1 0.1 5.3 2.7 3.6 3.5 0.82
58.0 6.8 12.7 5.4 3.5 5.4 0.44
MAPS / / / / / / /
9 18.1 60.0 2.8 6.3 11.8 10.3 9.0 0.78
3.0 3.0 8.8 11.7 1.7 2.6 0.71
MAPS / / / / / / /
10 20.5 1.1 1.1 6.3 14.2 6.9 7.1 0.72
3.4 3.4 8.4 10.6 4.3 4.3 0.66
MAPS / / / / / / /
11 19 60.0 2 6.3 12.7 7.8 8.0 0.70
24.8 10.7 10.7 4.1 8.2 8.5 0.43
MAPS / / / / / / /
12 6.6 58.5 14.9 14.9 8.3 10.3 8.6 0.71
25.1 4.1 10.5 0.3 5.7 7.5 0.66
MAPS / / / / / / /
13 10.8 51.3 1.5 14.9 4.1 8.4 7.5 0.84
60.0 9.2 9.2 1.3 14.3 16.6 0.53
MAPS / / / / / / /
14 7.9 60.0 14.9 14.9 7.0 11.0 9.1 0.66
21.3 10.5 10.5 4.9 4.1 7.6 0.42
MAPS / / / / / / /
15 5.6 25.4 19.9 19.9 14.3 31.8 30.3 0.33
15.6 6.4 6.4 1.5 14.0 16.0 0.70
MAPS / / / / / / /
16 4.9 60.0 18.5 18.5 13.6 7.7 5.7 0.49

23

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

Duct
Height 𝛤 𝛤𝑩
Duct 𝜴 (m) 𝜴𝑩 (m) (m) 𝑫𝟎 (M-units) (M-units) R
height COAMPS COAMPS COAMPS COAMPS COAMPS COAMPS COAMPS
(m) / / / / / / /
Dataset MAPS Inversion Inversion Inversion Inversion Inversion Inversion Inversion
60.0 1.5 5.8 6.0 14.1 14.0 0.78
MAPS / / / / / / /
17 11.8 60.0 0.6 18.5 6.7 4.6 2.4 0.75
2.1 2.1 7.7 2.2 13.9 15.1 0.67
MAPS / / / / / / /
18 9.9 60.0 0.5 18.5 8.6 3.4 2.4 0.81
3.2 3.2 7.9 2.5 15.3 15.5 0.67
MAPS / / / / / / /
19 10.4 60.0 0.8 6.6 3.8 4.0 2.9 0.92
2.9 2.9 5.5 3.4 12.0 12.8 0.73
MAPS / / / / / / /
20 8.9 60.0 1.3 6.6 2.3 4.9 3.7 0.92
1.4 1.4 5.5 1.4 5.0 6.4 0.83
MAPS / / / / / / /
21 6.9 60.0 0.4 6.6 0.3 3.4 2.4 0.94
1.5 1.5 4.5 1.9 4.7 5.8 0.85
MAPS / / / / / / /
22 6.4 60.0 0.7 6.6 0.2 5.0 3.2 0.92
2.9 2.9 6.2 9.0 8.6 7.2 0.81
MAPS / / / / / / /
23 15.2 -- -- -- -- -- -- --
2.1 2.1 7.3 3.7 10.8 10.2 0.90
MAPS / / / / / / /
24 11 -- -- -- -- -- -- --
3.6 3.6 11.5 7.5 3.0 3.6 0.71
MAPS / / / / / / /
25 19 -- -- -- -- -- -- --
1.4 1.4 4.4 9.6 11.4 9.9 0.79
MAPS / / / / / / /
26 14 -- -- -- -- -- -- --
1.9 1.9 6.6 8.3 12.7 10.9 0.56
MAPS / / / / / / /
27 14.9 -- -- -- -- -- -- --
3.0 3.0 15.4 2.2 12.2 12.1 0.91
MAPS / / / / / / /
28 17.6 -- -- -- -- -- -- --
0.9 0.9 6.2 19.8 8.6 7.5 0.40
MAPS / / / / / / /
29 26 -- -- -- -- -- -- --
1.9 1.9 4.8 3.6 2.7 4.8 0.77
MAPS / / / / / / /
30 8.4 -- -- -- -- -- -- --
0.8 0.8 5.7 7.3 2.0 4.2 0.48
MAPS / / / / / / /
32 13 -- -- -- -- -- -- --
0.9 0.9 5.5 9.7 2.9 3.4 0.13
MAPS / / / / / / /
33 15.2 -- -- -- -- -- -- --
0.8 0.8 38.0 25.7 8.3 7.6 0.59
MAPS / / / / / / /
34 12.3 -- -- -- -- -- -- --

24

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

Duct
Height 𝛤 𝛤𝑩
Duct 𝜴 (m) 𝜴𝑩 (m) (m) 𝑫𝟎 (M-units) (M-units) R
height COAMPS COAMPS COAMPS COAMPS COAMPS COAMPS COAMPS
(m) / / / / / / /
Dataset MAPS Inversion Inversion Inversion Inversion Inversion Inversion Inversion
0.4 0.4 4.5 11.1 6.0 3.9 0.36
MAPS / / / / / / /
35 15.6 -- -- -- -- -- -- --
60.0 1.8 7.2 0.2 13.8 13.7 0.89
MAPS / / / / / / /
36 7.4 -- -- -- -- -- -- --

MAPS Dataset

Figure 2: RMSE between remotely sensed refractivity and MAPS refractivity for the entire
profile (black) and only below the duct height (red). MAPS and inversion datasets are listed in
Tables 1 and 2.

25

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

(a) (b)

Figure 3: An example of an accurate inversion, MAPS 6 (a) and an example of an inaccurate

inversion, MAPS 15 (b). MAPS and RF (inversion) datasets are listed in Tables 1 and 2.

𝛺 and 𝛺𝐵 show the height of the largest discrepancy in refractivity with respect to both

the entire profile and only within the ED and are presented in Table 2 and Fig. 4. Most MAPS

datasets have 𝛺 located at the maximum altitude within each profile comparison (z = 60 m),

indicating the largest error is likely the result of divergent and/or shifted modified refractivity

within the mixed layer. Inaccuracy of the mixed layer slope has a lesser impact on EM propagation

than that within the ED (Pastore and Hackett, 2024). MAPS 5, 6, 21, and 22 also have an 𝛺𝐵 near

the surface, as shown in Fig. 4, indicating dominant discrepancies are between the inversion and/or

MAPS surface estimates. Conversely, MAPS 12, 14, 15, and 16 have 𝛺𝐵 occurring at the higher

duct height. This result implies that the least accurate part of the remotely sensed refractivity occurs

near the top of the duct, which is not ideal for propagation predictions and is consistent with the

relatively large difference in duct height (large D0). This result is also consistent with the large

RMSEs below the duct height for these same datasets (Fig. 2). These discrepancies appear to occur

when the inversion estimates a larger EDH, plausibly causing the largest residual to occur around

these altitudes (e.g., Fig. 3b). MAPS 1, 5, 21 and 22 have 𝐷0 less than 1 m. This indicates close

26

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

similarity in the duct heights of the two profiles, which is further supported by small RMSE and

high R for these datasets (Fig. 2 and 5). The largest 𝐷0 occurs for MAPS 2, because of a large EDH

(44.7 m) measured by MAPS and a much smaller one for the remotly sensed refractivity (11.3 m).

MAPS Dataset

Figure 4: Altitude of the maximum residual (𝛺, blue), altitude of the maximum residual below
the duct height (𝛺𝐵 , red), and the duct height differences 𝐷0 (black) between remotely sensed
refractivity and MAPS refractivity. MAPS and inversion datasets are listed in Tables 1 and 2.

Correlation coefficients (R), are an important statistic for quantifying similarity in the

vertical gradients of inversion and MAPS refractivity. Most MAPS datasets (with the exception of

2, 9-12, and 14-17) have R > 0.8, as shown in Fig. 5 and Table 2, indicating the majority of

inversion refractivity profiles are changing the same way vertically in relation to the MAPS

refractivity profiles. In other words, their refractivity gradients are similar, which is important

because refractive gradients are what impacts EM propagation predictions (Skolnik, 1962). It is

worth noting high correlation coincides with “typical” MAPS evaporation duct refractivity

profiles, meaning there is an ED with a large surface refractivity gradient present as well as a

mixed layer with a consistent slope (for an example of a typical ED refractivity profile view Fig.

27

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

1a). Inversely, MAPS 15 and 16 have relatively low correlation (R < 0.5), while MAPS 2 has a

relatively strong inverse correlation. These differences are caused by significant differences in ED

height (EDH) and/or the occurrence of a non-standard mixed layer slope (e.g., very steep gradients

and/or multiple changes in slope). The MAPS 2 refractivity profile has a large duct height (44.7

m), creating a large negative gradient up to the EDH, and the inversion profile has a large positive

mixed layer slope in the same region due to the lower EDH. Meaning, the profiles diverge for the

majority of the comparison, creating the high negative correlation. Similar to MAPS 2, MAPS 15

has poor correlation owing to differences in the gradient below the larger EDH. The MAPS

𝜕𝑀
refractivity profile measured a small gradient below the EDH ( ≲ -0.2 M-units/m) whereas the
𝜕𝑧

𝜕𝑀
inversion estimated a large gradient below the EDH ( ≳ -5.8 M-units/m). This difference may
𝜕𝑧

be associated with uncertainty in the estimates within the first ~0.5 m of the atmosphere, and the

different methods of estimating refractivity in this region by MAPS and the inversion.

MAPS Dataset

Figure 5: Correlation coefficients (R) between inversion-based refractivity and MAPS refractivity.
MAPS and inversion datasets are listed in Tables 1 and 2.

28

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

Overall, MAPS 6 has the lowest RMSE and a high correlation coefficient (Fig. 3a). The

profiles change in a similar way vertically and therefore have similar refractivity gradients. MAPS

15 has the highest RMSE and smallest positive correlation coefficient. This MAPS refractivity

profile does not have the classical evaporation duct shape and also has multiple sign changes in

slope above the duct leading to poor error statistics (Fig. 3b). Overall, the majority of the inversion-

MAPS comparisons had low RMSEs, high correlation coefficients, and small duct height

differences. Only certain inversions had inaccurate estimates of modified refractivity, as presented

in this section.

5.2 COAMPS Refractivity

Table 2 and Fig. 6 show RMSEs for COAMPS forecasts relative to MAPS measurements.

RMSE is frequently larger below the duct height, signifying the least accurate refractivity estimates

are below the ED and mixed layer slopes are highly accurate, which act to reduce the RMSE

because there are typically more points in the mixed layer than in the ED. This result is likely due

to the lack of large gradients in the MAPS data in the first few meters of altitude relative to the

COAMPS forecasts, which always show a steep gradient within the first meter or so. As previously

mentioned in section 5.1.1, the true nature of this surface region is largely unknown. Thus, it should

be noted that MAPS and/or COAMPS near surface estimate inaccuracies may be leading to the

large RMSEs within the ED. Furthermore, MAPS 16, 19, and 20 have typical MAPS and

COAMPS ED refractivity profiles indicating the large RMSE may be caused by surface point

disagreement more than differences in vertical gradients over the majority of the profile. The

lowest RMSE over the entire profile occurs for MAPS 10 (4.14 M-units). Conversely, MAPS 14

shows the highest RMSE (16.58 M-units), shown in Fig. 6, plausibly owing to the lack of a large

MAPS refractivity gradient at the surface as previously discussed. Moreover, MAPS 14

29

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

refractivity profile has a change in slope near the maximum altitude of the profile, this phenomenon

is known as an elevated duct. The COAMPS forecasts, due to limitations of the Monin-Obukhov

similarity theory used in the blended surface layer model, cannot model this elevated duct, leading

to the large RMSE. Similar to inversion results, discrepancies between COAMPS and MAPS

refractivity stem from multiple slopes in the mixed layer, mismatched EDH, and lack of strong

near-surface gradients in the MAPS data.

MAPS Dataset

Figure 6: RMSE between COAMPS refractivity and MAPS refractivity for the entire profile
(black) and only below the duct height (red). MAPS and COAMPS datasets are listed in Tables 1
and 2.

Fig. 7 and Table 2 shows results for the altitudes of the maximum residuals, where MAPS

7, 14, 17, and 36 have 𝛺 located at the maximum altitude within each profile (z = 60 m) and

(excluding MAPS 14) have an 𝛺𝐵 near the surface. All other datasets have 𝛺 and 𝛺𝐵 within or near

the ED, which is consistent with the larger RMSEs below the EDH (Fig. 6) for these datasets.

Again, this result is likely due to the differences between gradients estimated near the surface from

COAMPS and MAPS. The steep surface refractivity gradients likely occurring near the surface

30

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

are extremely difficult to measure directly and thus their existence within the surface layer at the

time of COAMPS forecasts and MAPS measurements is unknown (as previously discussed).

MAPS Dataset

Figure 7: Altitude of the maximum residual (𝛺, blue), altitude of the maximum residual below
the duct height (𝛺𝐵 , red), and duct height differences, (𝐷0 , black) between COAMPS and MAPS
refractivity. MAPS and COAMPS datasets are listed in Tables 1 and 2.

COAMPS forecasts for MAPS 2, 29, and 34 have especially large 𝐷0 (>20 m). MAPS 30-

36 occur at over 100 km offshore, while COAMPS forecasts only extend to 61 km offshore,

therefore, there is over 40 km difference between the MAPS and COAMPS profiles for these

datasets, which could at least partly explain the duct height discrepancies for these datasets. This

far offshore the Gulf Stream can impact refractivity distributions, which would not be present in

the profiles closer to shore resulting in truly different duct heights. In other words, the high

variability of atmospheric properties with respect to distance offshore may cause the discrepancies

between MAPS and COAMPS refractivity profiles for these datasets.

31

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

The Pearson correlation coefficients shown in Fig. 8 and Table 2 (R) between COAMPS

and MAPS refractivity are mostly consistent with RMSE, 𝛺, and 𝐷0 results (i.e., high correlation

occurs when RMSE and 𝐷0 are small). The majority of Pearson’s correlation coefficients are

relativity high (R > 0.6), with slight slope variations above the duct height conceivably being the

cause of weaker correlations. The highest correlation occurs for MAPS 28 (R = 0.91). The negative

correlation for MAPS 2 is due to the large discrepancy between the EDHs (large D0), where MAPS

data reflects a much higher EDH (44.7 m). This EDH difference causes the profiles to frequently

have divergent and/or shifted slopes at altitudes surrounding the higher EDH. The weakest positive

correlation occurs for MAPS 33 (R = 0.13), which is shown in Fig. 9b. Specifically, COAMPS

shows the existence of a surface-based duct above the mixed layer, which is not present in the

MAPS measurements, likely causing the weak correlation. However, unlike many of the

comparisons, the low correlation for MAPS 33 is not consistent with RMSE and residual

calculations. This observation shows the importance of all the error statistics. While the COAMPS

forecast does not change the same way vertically after 50 meters, it has low RMSE and a relatively

similar gradient below the duct height. Meaning the propagation predictions from this COAMPS

profile could be relatively accurate despite the low correlation (Figs. 6, 7, and 8) due to the

additional duct at high elevations. That being said, the majority of R values are consistent with

RMSEs and 𝛺, but MAPS 33 is an important comparison demonstrating the significance of every

statistic used in this study. An example of an accurate comparison (high correlation coefficient,

low RMSE, and small D0) is shown in Fig. 9a for MAPS 22.

32

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

MAPS Dataset

Figure 8: Correlation coefficients (R) between MAPS and COAMPS refractivity. MAPS and
COAMPS datasets are listed in Tables 1 and 2.

(a) (b)

Figure 9: A representation of an accurate COAMPS forecast, MAPS 22 (a) and a representation

of an inaccurate COAMPS forecasts, MAPS 33 (b). MAPS and COAMPS datasets are listed in
Tables 1 and 2.

5.3 Time and Range Effects on Refractivity Errors

Coefficients of determination (𝑅2 ) were calculated to quantify the linear relationship

linking time or range differences between the COAMPS or remotely sensed modified refractivity

33

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

and MAPS refractivity to the RSMEs. Recall that the time difference for the remotely sensed

refractivity can span upwards of 11 hours, while COAMPS forecasts are simulated for every hour,

meaning COAMPS never has a time difference over an hour. Moreover, the impact of range

differences on RMSE can only be evaluated for COAMPS refractivity because the remotely sensed

refractivity is based on radar data that covers variable distances offshore (up to 60 km).

There was no statistically significant linear relationship found between time or range

differences and COAMPS/inversion error statistics (p > 0.05); moreover, all linear relationships

had small coefficients of determination (𝑅2 < 0.1 for all variables). This result is demonstrated in

Fig. 10 and 11 for time differences for both inversion and COAMPS, respectively. Furthermore,

R2 values were computed for forecast hour after COAMPS initialization time and RMSE, which

also resulted in very low 𝑅2 (< 0.1) that was not statistically significant. Additionally, COAMPS

refractivity forecast range relative to the distance offshore of the MAPS refractivity measurements

showed no statistically significant relationship (p > 0.05). Therefore, we can conclude

spatiotemporal differences between COAMPS, inversion, and MAPS refractivity have a random

influence on the error statistics, plausibly due to the relatively small sample size of this study.

34

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

Figure 10: Linear regression for time difference between remotely sensed and MAPS refractivity
estimates versus inversion refractivity RMSE for the entire profile (blue) and only below the duct
height (black).

Figure 11: Linear regression for time difference between COAMPS forecast time and MAPS
measurements time versus COAMPS refractivity RMSE for the entire profile (blue) and only
below the duct height (black).

35

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

5.4 COAMPS Temperature and Humidity

All error statistics used to evaluate the accuracy of remotely sensed and COAMPS

refractivity were also used to evaluate COAMPS temperature and humidity against MAPS

measured temperature and humidity. The results are shown in Table 3 and Fig. 12-14. For the

majority of datasets, COAMPS predicts temperature more accurately than humidity. RMSEs for

temperature were typically below 2 K and RMSE for humidity was typically below 2 g/kg, as

shown in Fig. 12. Thus, COAMPS humidity nominally differs from MAPS humidity by 10% of

the mean humidity, while temperature by only about 1% of the mean temperature given typical

values of mean temperature and specific humidity in the MASL for this data. COAMPS

temperatures had higher average correlation with MAPS temperatures when compared to

correlations between COAMPS humidity and MAPS humidity. Most temperature profiles had R

> 0.8, while the majority of humidity profiles had R < 0.7, as shown in Fig. 13. The negative

temperature correlation shown in MAPS 2 and 29 are due to temperature increasing with altitude

in the MAPS data that was not predicted by COAMPS. Notably, none of the COAMPS humidity

profiles had a negative correlation with MAPS humidity.

COAMPS humidity profiles show similarity to the results for COAMPS refractivity. This

similarity is seen when comparing Fig. 7 to 11 (and Table 2 to 3) as 𝛺 and 𝛺𝐵 are located at roughly

the same altitudes for most of the datasets. This result is logical owing to the relationship between

refractivity, temperature, and humidity, where humidity has the largest influence on refractivity

(eq. 1).

36

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

Table 3: Error statistics for COAMPS and MAPS humidity and temperature (see Section 4.2 for a
description of error statistics).

𝛤 𝛤𝑩
𝜴 (m) 𝜴𝑩 (m) Humidity (g/kg) Humidity (g/kg) R
Humidity / Humidity / / Temperature / Temperature Humidity /
Dataset Temperature Temperature (℃) (℃) Temperature
MAPS 28.8 / 9.9 / 0.4 / 0.5 / 0.83 /
1 60.0 9.9 2.7 2.4 0.61
MAPS 4.7 / 4.7 / 0.3 / 0.4 / 0.79 /
2 60.0 44.7 2.0 1.9 -0.99
MAPS 3.1 / 3.1 / 1.1 / 1.3 / 0.67 /
3 13.4 10.0 0.6 0.6 0.94
MAPS 0.9 / 0.9 / 1.4 / 1.3 / 0.55 /
4 0.5 0.5 0.1 0.2 0.95
MAPS 1.6 / 1.6 / 1.7 / 1.6 / 0.57 /
5 0.5 0.5 0.4 0.7 0.86
MAPS 6.7 / 6.7 / 1.1 / 1.2 / 0.66 /
6 60.0 4.6 0.1 0.2 0.95
MAPS 60.0 / 1.6 / 1.0 / 1.2 / 0.69 /
7 0.4 0.4 0.4 0.3 0.96
MAPS 57.8 / 6.9 / 0.5 / 0.6 / 0.39 /
9 57.8 9.6 2.2 2.1 0.85
MAPS 3.3 / 3.3 / 0.7 / 0.8 / 0.65 /
10 51.5 20.5 3.0 2.9 0.80
MAPS 3.3 / 3.3 / 0.4 / 0.5 / 0.59 /
11 2.6 2.6 0.4 0.5 0.88
MAPS 24.7 / 10.7 / 1.1 / 1.2 / 0.61 /
12 54.2 10.7 0.5 0.7 0.89
MAPS 24.7 / 3.7 / 0.8 / 1.1 / 0.68 /
13 1.5 1.5 0.1 0.3 0.92
MAPS 60.0 / 9.2 / 1.8 / 2.2 / 0.44 /
14 56.4 9.2 0.5 0.5 0.84
MAPS 60.0 / 10.5 / 0.5 / 1.1 / 0.50 /
15 60.0 10.5 0.2 0.3 0.87
MAPS 16.2 / 6.4 / 1.6 / 2.0 / 0.46 /
16 60.0 6.4 1.5 1.4 0.68
MAPS 60.0 / 1.6 / 1.4 / 1.4 / 0.50 /
17 48.6 2.9 2.7 2.7 0.80
MAPS 2.1 / 2.1 / 2.5 / 2.7 / 0.55 /
18 60.0 4.0 2.8 2.8 0.79
MAPS 3.1 / 3.1 / 2.2 / 2.2 / 0.61 /
19 1.8 1.8 0.2 0.3 0.95
MAPS 2.7 / 2.7 / 2.0 / 2.1 / 0.47 /
20 60.0 1.8 1.1 1.1 0.89
MAPS 1.4 / 1.4 / 1.0 / 1.2 / 0.56 /
21 51.3 1.7 1.5 1.4 0.92
MAPS 1.4 / 1.4 / 1.0 / 1.1 / 0.52 /
22 60.0 0.9 1.5 1.5 0.95
MAPS 3.1 / 3.1 / 0.9 / 0.8 / 0.64 /
23 60.0 15.2 0.7 0.7 0.94
MAPS 2.3 / 2.3 / 1.2 / 1.1 / 0.78 /
24 59.6 11.0 1.2 1.1 0.94

37

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

𝛤 𝛤𝑩
𝜴 (m) 𝜴𝑩 (m) Humidity (g/kg) Humidity (g/kg) R
Humidity / Humidity / / Temperature / Temperature Humidity /
Dataset Temperature Temperature (℃) (℃) Temperature
MAPS 3.6 / 3.6 / 0.3 / 0.5 / 0.73 /
25 4.4 4.4 0.6 0.7 0.89
MAPS 1.6 / 1.6 / 1.4 / 1.2 / 0.76 /
26 45.2 14.0 0.6 0.5 0.94
MAPS 2.0 / 2.0 / 1.5 / 1.3 / 0.82 /
27 58.3 14.9 0.7 0.6 0.97
MAPS 3.0 / 3.0 / 1.4 / 1.4 / 0.97 /
28 54.9 0.0 1.2 1.1 0.34
MAPS 0.9 / 0.9 / 1.0 / 0.9 / 0.91 /
29 56.7 26.0 0.6 0.3 -0.94
MAPS 1.8 / 1.8 / 0.3 / 0.7 / 0.44 /
30 1.9 1.9 1.8 1.8 0.80
MAPS 0.8 / 0.8 / 1.3 / 1.7 / 0.70 /
32 1.3 1.3 5.2 5.3 0.94
MAPS 0.8 / 0.8 / 1.0 / 1.4 / 0.60 /
33 0.5 0.5 4.7 4.9 0.95
MAPS 0.7 / 0.7 / 2.3 / 2.3 / 0.85 /
34 0.2 0.2 5.0 5.1 0.90
MAPS 0.3 / 0.3 / 0.4 / 0.8 / 0.90 /
35 0.0 0.0 4.4 4.5 0.96
MAPS 60.0 / 1.7 / 3.2 / 3.2 / 0.91 /
36 60.0 7.4 4.5 4.4 1.00

38

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

(a)

(b)
.5
RMSE (g/kg)

MAPS Dataset

Figure 12: RMSE between MAPS and COAMPS temperature (a) and MAPS and COAMPS
humidity (b) for the entire profile (black) or only below the duct height (red). MAPS and
COAMPS datasets are listed in Tables 1-3.

39

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

(a)

(b)

MAPS Dataset

Figure 13: Altitude of the maximum residual (𝛺, blue), altitude of the maximum residual below
the duct height (𝛺𝐵 , red) for COAMPS temperature (a) and humidity (b). The duct height
difference (D0, black) is shown in both panels for reference and is the same as that presented in
Figure 7. MAPS and COAMPS datasets are listed in Tables 1-3.

40

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

(a)

(b)

MAPS Dataset

Figure 14: Correlation coefficients (R) between MAPS and COAMPS temperature (a) and
humidity (b). MAPS and COAMPS datasets are listed in Tables 1-3.

41

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

6.0 Conclusions

This study evaluates the accuracy of NWP forecasts and inversely determined refractivity

from RF measurements. 36 NWP forecasts and 12 radar refractivity inversions were evaluated via

comparison to 36 in-situ atmospheric measurement sets, which are considered to be representative

of the “true” atmosphere (despite spatiotemporal differences). Each inversely determined

refractivity and NWP forecast were selected for comparison with an in-situ measurement dataset

based on the closest proximity in both space and time relative to the in-situ measurements. Multiple

error statistics were computed to assess the accuracy including: RMSE, duct height differences,

altitudes of the largest residuals (both below the duct height and for the entire profile), correlation

coefficients, as well as differences in location and time between the in-situ measurements and

either NWP forecasts or radar inversion estimates.

The results for the remotely sensed refractivity generally show low RMSE, high

correlation, and relatively small duct height differences for the majority of the refractivity

comparisons; however, a few specific radar inversions had high RMSEs, low correlation, and large

duct height differences, which were attributed to irregularities within the MAPS profiles (small

near-surface refractivity gradients and/or multiple sign changes in the mixed layer slope). Thus,

these inaccuracies could be the result of extrapolation of MAPS measurements to the surface, or

limitations of the parametric model used in the refractivity inversions that cannot account for these

irregularities. Unlike inversion-based refractivity, RMSEs for COAMPS refractivity were mostly

larger below the duct height than over the entire profile indicating highly accurate mixed layer

slopes and less accurate comparisons within the ED; however, correlation coefficients were

generally similar, being only slightly lower on average, indicating reasonably accurate refractivity

gradients. Generally, the error statistics show that the NWP refractivity is representative of the

42

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

“true” atmosphere; however, inaccurate forecasts do occur. These inaccurate forecasts may be the

result of uncertainties within the in-situ measurements because they were most often related to

differences in the near-surface region, where COAMPS almost always predicts large gradients

(due to the surface layer model) that can be absent in the MAPS measurements. This result could

be due to the extrapolation of the MAPS measurements to the surface, as mentioned previously.

COAMPS forecasted humidity and temperature were also evaluated. Results show that

temperature is consistently more accurate than humidity. The humidity results are quite

comparable to the refractivity results, meaning inaccuracies in NWP humidity estimates likely lead

to inaccuracies in NWP refractivity estimates, which is logical given the large dependence of

refractivity on humidity. Spatiotemporal differences between COAMPS/inversion refractivity and

in-situ refractivity had no statistically significant relationship. This result implies that

spatiotemporal differences are not a dominant factor leading to the magnitudes of the error

statistics.

The results found in this study will be used for future work regarding humidity retrievals

from RF signals. Knowledge of the accuracy of inversely determined refractivity is necessary to

allow study of the effect of these errors on humidity retrievals. Moreover, the accuracy of NWP

humidity is necessary if humidity retrievals will be compared with COAMPS humidity forecasts

for verification of the humidity retrieval. Furthermore, NWP temperature could be used to

constrain humidity retrievals and thus, its uncertainty needs to be known in order to evaluate the

impact of these inaccuracies on humidity retrievals. Meaning, inaccurate NWP temperatures could

potentially increase uncertainty for the retrieval process. In summary, the results of this study will

help inform test cases for evaluating the humidity retrieval technique. The improvement of modern

remote sensors could increase the spatial and temporal coverage of weather measurements

43

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

including allowing for both broader spatial coverage at finer temporal resolution than current

technologies allow.

Acknowledgements: The author would like to thank Coastal Carolina University and the Office

of Naval Research (N000 I 4-23-C-2033) for funding this research, and Daniel Greenway for his

mentorship regarding this study.

44

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

References

Babin, S. M., Young, G. S., Carton, J. A. (1997). A new model of the oceanic evaporation duct.
Journal of Applied Meteorology, 36, 193–204. https://doi.org/10.1175/1520-0450

Bean, B. B., Dutton, E. J. (1968). Radio meteorology (1st ed.). College Entrance Book Company:
New York, NY.

Che, Y., Ma, S., Xing, F., Li, S., Dai, Y. (2019). An improvement of the retrieval of temperature
and relative humidity profiles from a combination of active and passive remote sensing.
Meteorological and Atmospheric Physics, 131, 681–695. https://doi.org/10.1007/s00703-
018-0588-3

Cooper, K. B., Roy, R. J., Dengler, R., Monje, R. R., Alonso-Delpino, M., Siles, J. V.,
Yurduseven, O., Parashare, C., Millan, L., Lebsock, M. (2021). G-band radar for
humidity and cloud remote sensing. IEEE Transactions on Geoscience and Remote
Sensing, 59(2), 1106–1117. https://doi.org/10.1109/TGRS.2020.2995325

Frederickson, P. A. (2015). Further improvements and validation for the Navy Atmospheric
Vertical Surface Layer Model (NAVSLaM). In Proceedings of the 2015 USNC-URSI
Radio Science Meeting (Joint with AP-S Symposium), Vancouver, BC, Canada, 19–24
July 2015 (p. 242). IEEE. doi: 10.1109/USNC-URSI.2015.7303526

Haack, T., Wang, C., Garrett, S., Glazer, A., Mailhot, J., Marshall, R. (2010). Mesoscale
modeling of boundary layer refractivity and atmospheric ducting. Journal of Applied
Meteorology and Climatology, 49(12), 2437–2457.
https://doi.org/10.1175/2010jamc2415.1

Halliwell, G., Bleck, R., Chassignet, E. (1998). Atlantic Ocean simulations performed using a
new hybrid-coordinate ocean model. In EOS, transactions AGU, Fall 1998 AGU meeting.

Hitney, H. V., Richter, J. H., Pappert, R. A., Anderson, K. D., Baumgartner, G. B. (1985).
Tropospheric radio propagation assessment. Proceedings of the IEEE, 73(2), 265–283.
https://doi.org/10.1109/proc.1985.13138

Hodur, R. M. (1997). The Naval Research Laboratory’s coupled ocean/atmosphere mesoscale

prediction system (COAMPS). Monthly Weather Review, 125(7), 1414–1430.
https://doi.org/10.1175/1520-0493(1997)125<1414>2.0.co;2

Hogan, T. F., Liu, M., Ridout, J. A., Peng, M. S., Whitcomb, T. R., Ruston, B. C., et al. (2014).
The navy global environmental model. Oceanography, 27(3), 116–125.
https://doi.org/10.5670/oceanog.2014.73

Kang, D., Wang, Q. (2016). Optimized estimation of surface layer characteristics from profiling
measurements. Atmosphere, 7(2), 14. https://doi.org/10.3390/atmos7020014

45

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

Karimian, A., Yardim, C., Gerstoft, P., Hodgkiss W.S., Barrios, A.E. (2011). Refractivity
estimation from sea clutter: An invited review. Radio Science, 45, 1-16.

Karimian, A., Yardim, C., Haack, T., Gerstoft, P., Hodgkiss, W. S., Rogers, T. (2013). Toward
the assimilation of the atmospheric surface layer using numerical weather prediction and
radar clutter observations. Journal of Applied Meteorology and Climatology, 52(10),
2345–2355.

Lentini, N. E., Hackett, E. E. (2015). Global sensitivity of parabolic equation radar wave
propagation simulation to sea state and atmospheric refractivity structure. Radio Science,
50(10), 1027–1049. https://doi.org/10.1002/2015rs005742

Liu, B., Ma, X., Ma, Y., Li, H., Jin, S., Fan, R., Gong, W. (2022). The relationship between
atmospheric boundary layer and temperature inversion layer and their aerosol capture
capabilities. Atmospheric Research, 271, 106121.
https://doi.org/10.1016/j.atmosres.2022.106121

Pastore, D. M. (2023). Characteristics of refractivity and sea state in the marine atmospheric
surface layer and their influence on X-band propagation. (Dissertation). Coastal Carolina
University.

Pastore, D. M., Greenway, D. P., Stanek, M. J., Wessinger, S. E., Haack, T., Wang, Q., Hackett,
E. E. (2021). Comparison of atmospheric refractivity estimation methods and their
influence on radar propagation predictions. Radio Science, 56(9), 1–17.
https://doi.org/10.1029/2020rs007244

Pastore, D. M., Hackett, E. E. (2024). Relative influence of sea state and mean, turbulent, and
heterogeneous refractivity on X-band propagation. IEEE Transactions on Antennas and
Propagation, 72(6), 5223-5234. https://doi.org/10.1109/TAP.2024.3396644.

Pastore, D. M., Wessinger, S. E., Greenway, D. P., Stanek, M. J., Burkholder, J. R., Haack, T.,
Wang, Q., Hackett, E. E. (2022). Refractivity inversion from point-to-point X-band radar
propagation measurements. Radio Science, 57(2), 1-16.
https://doi.org/10.1029/2021RS007345

Plant, W. J., Irisov, V. (2017). A joint active/passive physical model of the sea surface
microwave signatures. Journal of Geophysical Research: Oceans, 122, 3219–3239.
https://doi.org/10.1002/2017jc012749

Paulus, R. A. (1989). Specification for environmental measurements to assess radar sensors.

Naval Ocean Systems Center. DTIC. Technical Document 1685.

Pozderac, J., Johnson, J., Yardim, C., Merrill, C., de Paolo, T., Terrill, E., et al. (2018). X-band
beacon-receiver array evaporation duct height estimation. IEEE Transactions on
Antennas and Propagation, 66(5), 2545–2556. https://doi.org/10.1109/tap.2018.2814060

46

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.

Approved, DCN# 543-2217-24 8/13/2024

Rainer, R. B. (2016). In-situ observation of undisturbed surface layer scaler profiles for
characterizing evaporative duct properties. (M.S. Thesis). Naval Postgraduate School.

Ren, Q. Y., Wang, L. F., Huang, J. Q., Zhang, C., Huang, Q. A. (2015). Simultaneous remote
sensing of temperature and humidity by LC-Type passive wireless sensors. Journal of
Microelectromechanical Systems, 24(4), 1117-1123.
https://doi.org/10.1109/JMEMS.2014.2384591

Skolnik, M. I. (1962). Radar handbook (1st ed.). McGraw-Hill: New York, NY.

Stanek, M. J., Pastore, D. M., Hackett, E. E. (2023). A novel near-surface wave-coherent

instantaneous profiling system for atmospheric measurements. Sensors, 23(8), 4099.
https://doi.org/10.3390/s23084099

Wang, Q., Alappattu, D. P., Billingsley, S., Blomquist, B., Burkholder, R. J., Christman, A. J., et
al. (2018). CASPER: Coupled Air-Sea Processes and Electromagnetic Ducting Research.
Bulletin of the American Meteorological Society, 99(7), 1449–1471.
https://doi.org/10.1175/bams-d-16-0046.1

Wessinger, S. E., Greenway, D. P., Haack, T., Hackett, E. E. (2024). Near-surface thermodynamic
influences on evaporation duct shape. Atmosphere, 15(6), 718.
https://doi.org/10.3390/atmos15060718

47

DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.