Posted on 19 Jul 2024 — The copyright holder is the author/funder. All rights reserved. No reuse without permission. — https://doi.org/10.22541/au.172139840.

08896939/v1 — This is a preprint and has not been peer-reviewed. Data may be preliminary.

On Spatio-temporal Directional Association of Blue and Humpback

Whale’s Migratory Path Navigation with Sun, Moon, Ocean
Current & Earth’s Magnetic Field
Debashis Chatterjee1 and Prithwish Ghosh2
1
Visva-Bharati
2
North Carolina State University at Raleigh

July 19, 2024

Abstract
Since Blue and Humpback whales are classified as endangered, understanding their migration patterns is essential for devel-
oping effective conservation and management plans, thereby maintaining a balanced ecosystem across the oceans. This paper
investigates the migration patterns of blue and Humpback whales along the western coast of North America by leveraging a
comprehensive dataset on longitude, latitude, change in path direction, magnetic inclination, declination change, and date.
To account for directional changes in whale paths, a mixture of von Mises-Fisher distributions is considered for appropriate
reasons. Our investigation suggests that the nature of the directional changes exhibited by the blue and Humpback whales
can be characterized by a mixture of nine von Mises-Fisher distributions, providing evidence for the existing variation across
the whales’ directional migration. Furthermore, a novel circular-circular regression is considered to explore how the directional
changes of the whales are associated with various environmental factors, such as magnetic inclination, declination, Sun, Moon
position, and ocean currents. Almost all of our hypothesis test results reveal a statistically significant non-zero correlation
between the directional changes of the blue and Humpback whales and the considered environmental factors. Based on the
table, it’s evident that the correlation between Solar position and Geomagnetic changes has the most pronounced effect on both
types of Whales.

1
ORIGINAL ARTICLE
Journal Section

On Spatio-temporal Directional Association of

Blue and Humpback Whale’s Migratory Path
Navigation with Sun, Moon, Ocean Current &
Earth’s Magnetic Field

Debashis Chatterjee1 | Prithwish Ghosh2

1 Department of Statistics, Visva Bharati,

Santiniketan, West Bengal, India 1. Understanding the migration patterns of Blue and Humpback whales, 1

2
which are classified as endangered, is crucial for developing effective con- 2
Department of Statistics, North Carolina
servation and management plans to maintain a balanced ocean ecosys- 3
State University, 5109, SAS Hall, 2311
tem. This study investigates the migration patterns of these whales along 4
Stinson Dr, Raleigh, NC 27607, United
the western coast of North America using a comprehensive dataset that 5
States
includes longitude, latitude, changes in path direction, magnetic inclina- 6

tion, declination change, and date. 7

Correspondence
2. To accurately account for directional changes in whale paths, we applied 8
Prithwish Ghosh, Department of Statistics,
a mixture of von Mises-Fisher distributions. Our analysis indicates that 9
North Carolina State University, 5109, SAS
the directional changes of Blue and Humpback whales can be best de- 10
Hall, 2311 Stinson Dr, Raleigh, NC 27607,
scribed by a mixture of nine von Mises-Fisher distributions, highlighting 11
United States
significant variation in their migration directions. 12
Email: [email protected]
3. Additionally, we introduced a novel circular-circular regression model to 13
examine the relationship between the whales’ directional changes and 14
Funding information
various environmental factors, such as magnetic inclination, declination, 15

sun and moon positions, and ocean currents. The hypothesis tests reveal 16
a statistically significant non-zero correlation between the directional 17

changes of the whales and the environmental factors considered. The 18

results particularly emphasize that the correlation between solar posi- 19

tion and geomagnetic changes has the most pronounced effect on both 20

Blue and Humpback whales. 21

4. The findings underscore the complexity of whale migration and the in- 22
fluence of environmental variables, providing valuable insights for con- 23

servation efforts and furthering our understanding of these endangered 24

species’ behaviors. 25

KEYWORDS 26

Migration, Circular Statistics, Circular Circular Regression 27

Abbreviations: ABC, a black cat; DEF, doesn’t ever fret; GHI, goes home immediately.

1
 2 Chatterjee & Ghosh et al.

28 Introduction

29 The blue whale (Balaenoptera musculus) and Humpback Whale (Megaptera novaeangliae) is a marine mammal, is the
30 largest known animal, and is a remarkable marine species known for its incredible migration patterns. The ecological
31 theory of animal migrations remains consistent across marine and terrestrial systems [1]. Numerous whale species
32 migrate annually between high-latitude feeding grounds and low-latitude breeding grounds [2]. Although their mi-
33 grations are influenced by various factors, including prey availability, an intriguing aspect involves their sensitivity to
34 celestial and Earth-bound signals. Blue and Humpback whales exhibit a navigation ability believed to be influenced
35 by the directional change of ocean currents and the Earth’s magnetic field. These celestial and geomagnetic cues
36 significantly guide their migrations across vast oceanic distances, allowing them to find feeding and breeding grounds
37 with impressive precision [3]. This innate navigation system is a testament to the fascinating relationship between
38 Earth’s most enormous creatures and the celestial and terrestrial forces that shape their migratory journeys.
39 In this paper, we use directional statistical methodologies for examining the Spatio-temporal Directional Relationship
40 between the Migratory Paths of Blue and Humpback Whales and Changes in Sun and Moon Positions, Ocean Current
41 Direction, and Earth’s Magnetic Field. There is a wide range of literatures indirectly supporting our hypothesis. [1]
42 provides statistical evidence that the long-distance migrations of marine megafauna are influenced significantly by
43 both long-term memory and resource tracking. [4] addressed movements of blue and Humpback whales in the North
44 Pacific during the feeding season off Southern California and their southern fall migration. A comprehensive update
45 of the classic Encyclopedia of Marine Mammals is [5]. This authoritative book provides the latest information on each
46 species of blue whale, incorporating the most recent taxonomy. [6], [7] delve into intricate details about blue whale
47 species, covering various biological, ecological, and sociological aspects of marine mammals. [8] employed satellite
48 telemetry to monitor the migratory patterns and winter destinations of 23 blue whales from eastern Canada; this study
49 marks the inaugural documentation of the migratory movements of western North Atlantic blue whales. [9] contains
50 the initial documented migratory destination for eastern South Pacific blue whales. [10] accounted for differences in
51 the diving behavior of Cuvier’s beaked whales about seafloor depth, time-of-day, and lunar illumination. Research like
52 [11] has documented the austral summer residency of blue whales in this region, with indications that certain individ-
53 uals extend their stay, underscoring the significance of this area as a habitat for blue whales. Blue whale migration
54 journey is illustrated in [12]. [2] investigated the roles of geomagnetic and acoustic cues in humpback whale naviga-
55 tion and orientation. Additionally, [2] documents certain effects of human-produced sound on beaked whales [13].
56 One article explores how migrating animals navigate using geophysical cues, focusing on the role of biological clocks in
57 avian orientation studies [14]. The study investigates the decline of the endangered Antarctic blue whale population,
58 estimating their abundance and forecasting a trajectory using Bayesian models, indicating a significant reduction from
59 pre-whaling levels [15]. One study estimates the Antarctic blue whale population off the Queen Maud Land coast,
60 revealing higher densities than previously circumpolar surveys and emphasizing the importance of current abundance
61 assessments for conservation and monitoring efforts [16]. The study uses individual-based movement models to dis-
62 tinguish the roles of sea surface temperature (SST) and prey availability in driving blue whale migration, revealing
63 that both factors independently influence foraging behavior and emphasizing the importance of understanding their
64 effects on threatened species’ responses to changing ocean conditions [17].
65 In this paper Using a directional statistical approach, we analyze the spatio-temporal directional association of blue
66 whale migratory path navigation with the positions of the Sun, Moon, ocean current direction, and changes in Earth’s
67 magnetic field. The reason for using directional statistical methodologies is that almost all the parameters are angular
68 and need a special statistical paradigm. We proceed with a well-known dataset of whale migration originally published
69 in [18] and check the goodness of fit for spherical von Mises Fisher mixture distribution fit. We found that we can
 Chatterjee & Ghosh et al. 3

70 achieve that under optimal unsupervised partitioning of the datasets. After that, we create a circular-circular regres-
71 sion dataset by creating a derived dataset containing the angular spatio-temporal direction change of whale migration
72 path (as angular response variable). We calculate the change in concerned environmental factors like the Sun, Moon
73 position change, Ocean Current Direction & Earth’s Magnetic Field Change using specified mathematical equations
74 (explained later in detail). We take the angular covariate part of the circular-circular regression dataset by computing
75 the environmental factors’ directional change. We proceed in two ways: first, we validate the von Mises error suitable
76 for circular-circular regression in this scenario. Then, we propose and statistically validate a circular-circular regression
77 model using directional statistical methodologies. Second, we conduct a proper statistical hypothesis test of non-zero
78 directional correlation among covariate and response. The statistical hypothesis test’s outcome shows statistically
79 significant non-zero correlations among responses and covariates, supporting our hypothesis. The objective of this
paper can be summarized as follows.

(a) Two-dimensional plot of Blue whales path over the world (b) Two-dimensional plot of Humpback whales path over the
map, the red dot indicates the whales. world map, the red dot indicates the whales.

80

81 | Objective of this paper

82 1. Distributional check whether their location and other directional parameters follow any circular distribution.
83 2. Based on the distributions on circular paradigm, proposing a mixture distribution on Spherical paradigm(von Mises
84 Fishers, mixture distribution) with a pictorial view.
85 3. Incorporating circular-circular regression model to check the Blue whales’ directional change dependency. The
86 dependent variable is the Directional change of blue whales, and the independent variables are the Sun, Moon
87 position change, ocean current direction, and magnetic field directional change.
 4 Chatterjee & Ghosh et al.

88 Methods

89 | World Magnetic Model

90 The World Magnetic Model (WMM) provides a detailed global representation of the Earth’s magnetic field [19]. It uses
91 a spherical harmonic expansion up to the 1212 degree and order to describe the magnetic scalar potential originating
92 from the Earth’s core-generated geomagnetic main field. This model incorporates 168 spherical-harmonic "Gauss"
93 coefficients and an equal number of spherical-harmonic Secular-Variation (S.V.) coefficients. The conventions for
94 measurements in this section are as follows: angles in radians, lengths in meters, magnetic intensities in nanoteslas
95 (nT, where one tesla equals one weber per square meter or one kg.s −2 .A −1 ), and times in years [20].
96 The main magnetic field B m can be expressed as a potential field in geocentric spherical coordinates (longitude λ,
97 latitude φ ′ , radius r ) by the negative spatial gradient of a scalar potential:

B m (λ, φ ′ , r , t ) = −+V (λ, φ ′ , r , t ) (1)

98 where t represents time. This potential is expanded in terms of spherical harmonics as follows [21]:

N Õ
M   n+1
!
Õ a
V (λ, φ ′ , r , t ) = a g nm (t ) cos(mλ ) + h m ′
n (t ) sin(mλ ) p̆ n (sin φ ) (2)

m
r
n=1 m=1

99 We set N = 36 for the truncation level of the internal expansion of the World Magnetic Model. Here, a (6371200
100 m) is the geomagnetic reference radius, and (λ, φ ′ , r ) are the longitude, latitude, and radius in a spherical geocentric
101 reference frame. The coefficients g nm (t ) and h m
n (t ), which are time-dependent, describe the Earth’s main magnetic
102 field:

g nm (t ) = g nm + g¤nm (t − t 0 ) + g¥nm (t − t 0 ) 2 (3)

2
hm
n (t ) = hm
n + h¤ m ¥m
n (t − t 0 ) + h n (t − t 0 ) (4)

103 In these equations, g nm , h m

n , g¤n , h n , g¥n , h n are constants. For any real number µ, the Schmidt semi-normalized
m ¤ m m ¥m

104 associated Legendre functions p̆ nm (µ) are defined as follows:

s
(n − m )!
p̆ nm (µ) = 2 p n,m (µ ), if m > 0 (5)
(n + m )!
p̆ nm (µ) = p n,m (µ), if m=0 (6)

105 For datasets containing hourly mean observatory data, local field offsets, mainly generated in the crust, must be
106 considered. At an observatory, the magnetic field B is:

B m (λ, φ ′ , r , t ) = −+V (λ, φ ′ , r , t ) + O (λ, φ ′ , r , t ) (7)

 Chatterjee & Ghosh et al. 5

107 The offset vector O (λ, φ ′ , r , t ), known as crustal bias, is constant over time. This parameterization fits data from
108 satellite measurements and observatory hourly means. The internal part of the field is described by the following
109 equations:

N   n+2 Õ
M
∂V Õ a
d p̆ nm (sin φ ′ )
X ′ (λ, φ ′ , r , t ) = − = − g nm (t ) cos(mλ ) + h m
n (t ) sin(mλ ) (8)
r ∂φ ′ r dφ ′
n=1 m=0
N   n+2 Õ
M
∂V 1 Õ a
Y ′ (λ, φ ′ , r , t ) = − = m g nm (t ) cos(mλ ) + h m ′
n (t ) sin(mλ ) p̆ n (sin φ ) (9)

m
r cos φ ′ ∂λ r cos φ ′ r
n=1 m=0
N M
 a  n+2 Õ
∂V Õ
Z ′ (λ, φ ′ , r , t ) = =− (n + 1) g nm (t ) cos(mλ ) + h m ′
n (t ) sin(mλ ) p̆ n (sin φ ) (10)

m
∂r r
n=1 m=0

110 These equations, coupled with magnetic field vector observations, form the equations of condition. If there are
111 d data points, there are d linear equations with p parameters of the parent model:

y = Am (11)

112 Here, y is the column vector (d x 1) of observations, A is the matrix (d x p) of coefficients to the unknowns, which
113 are functions of position, and m is the column vector (p x 1) of unknowns, the Gauss coefficients of the model. Since
114 there are more observations than unknowns (d > p), the system is over-determined and does not have an exact
115 solution.
116 The final main-field coefficients for 2005.0 were derived by polynomial extrapolation of the main-field Gauss coef-
117 ficients from the parent model to this date, using equation (1). Equations (44), with the time-varying Gauss coefficients
118 g nm , h m
n replaced by their time derivatives g¤n , h n , were used to determine the final secular-variation coefficients [22].
m ¤m

119 | Haversine Formula & Calculation of Directional Changes in Blue Whale Migration Paths

120 To compute the column theta, representing directional changes in whale migration paths, we utilize the Haversine for-
121 mula (see, for example, [23]). The Haversine formula precisely calculates distances and directions between two points
122 on a sphere, making it crucial for navigation. It is a specific case of a more general formula in spherical trigonometry,
123 which relates the sides and angles of spherical triangles. The formula, which uses the latitude and longitude of the
124 two points, is particularly beneficial for small angles and distances. For further details, see [24], [25]. The formula is:

cos θ = sin φA sin φB + cos φA cos φB cos ∆λ (12)

125 where ∆λ = λ A − λ B . Point A has latitude φA and longitude λ A , while point B has latitude φB and longitude λ B .

126 | Ocean Direction and Velocity

127 Satellite altimeters not only measure surface geostrophic currents but also wave height. The geostrophic balance
128 requires that the Coriolis force balances the horizontal pressure gradient. The geostrophic balance equations are
 6 Chatterjee & Ghosh et al.

129 derived from the equations of motion, assuming no acceleration in the flow (d u/d t = dv /d t = dw /d t = 0), horizontal
130 velocities being much larger than vertical velocities (w ≤ u, v ) [26], gravity as the only external force, and minimal
131 friction. These assumptions lead to the following equations:

δp δp δp
= ρf v = ρf u = −ρg (13)
δx δx δx

132 where 2Ω sin φ is the Coriolis parameter. This can be rewritten as:

δp δp
u= u= (14)
f ρδ y f ρδx

133 At z = 0, the geostrophic approximation results in a simple relation: surface geostrophic currents are proportional
134 to surface slope. For example, at a level surface slightly below the sea surface, say two meters below, at z = −r [27],
135 we have:

u =v =w =0 (15)

136 The pressure at this level is given by:

p = ρg (ζ + r ) (16)

137 Assuming ρ and g are essentially constant in the upper few meters of the ocean, substituting this into the
138 geostrophic balance equations gives the components u s , vs of the surface geostrophic current:

g δς g δς
us = vs = (17)
f δy f δx

139 The geostrophic current velocity is computed from the eastward (U) and northward (V) components of sea water
140 velocity [28] using:

p
Velocity (ms −1 ) = U 2 + V 2 (18)

141 The angle of incidence is calculated as:

 
U
θ = tan −1 (19)
V

142 Here, V is directed from East to West or West to East, while U completes the North to South or South to North
143 direction [29].
 Chatterjee & Ghosh et al. 7

144 0.1 | Modeling the Sun and Moon’s Position in Space and Time
145 A novel approach is employed to analyze blue whale migration paths, leveraging the position of the sun. Blue whales
146 utilize the sun’s position in the sky as a navigational aid during their migratory journeys. Our results demonstrate the
147 efficacy of the proposed model in estimating the sun’s position relative to changes in longitude, latitude, and the day
148 of the year. [30]

149 | Calculating Zenith and Azimuth Angles for the Sun and Moon
150 Let the observer’s coordinates, represented by latitude and longitude, be denoted as (φ0 , λ 0 ), and the subsolar point’s
151 coordinates be (φs , λ s ). The x-, y-, and z-components of the unit vector, S, pointing from the observer to the center
152 of the Sun are computed as follows:

φs = δ,
λ s = −15(TG MT − 12 + E mi n /60),
S x = cos φs sin(λ s − λ 0 ),

S y = cos φ0 sin φs − sin φ0 cos φs cos(λ s − λ 0 ),

S z = cos φ0 sin φs − cos φ0 cos φs cos(λ s − λ 0 ).

153 The assumption is made that the influence of parallax is negligible, effectively assuming an infinite Earth-Sun
154 distance. It can be shown that this assumption holds true, leading to S x2 + S y2 + S z2 = 1. Here, TGMT represents
155 Greenwich Mean Time or UTC.

156 Remark The derivation of S x , S y , and S z becomes straightforward when working within the Earth-Centered Earth-
157 Fixed (ECEF) coordinate system. The ECEF system is a geocentric right-handed Cartesian system. The process involves
158 starting at the subsolar point in the ECEF system, constructing a unit vector pointing upward, and computing the dot
159 product of each unit vector with the vector from the observer’s coordinates to obtain S x , S y , and S z . This procedure
160 disregards the influence of parallax, assuming an infinite Earth-Sun distance for ease of calculation.

161 The solar zenith angle is then given by Z = cos −1 S z , and the solar azimuth angle, following the South-Clockwise
162 convention, is γs = atan2(−s x , −s y ).

163 Remark The azimuth angle of the sun and moon provides valuable directional information relative to an observer,
164 measured in degrees clockwise from the north. It reveals their lateral positions in the sky. The azimuth angle continu-
165 ously changes throughout the day, with the sun starting at 90 degrees (east) during sunrise and setting at 270 degrees
166 (west) during sunset. Similarly, the azimuth angle of the moon changes as it orbits the Earth. During a full moon, it rises
167 around 90 degrees, reaches its highest point at approximately 180 degrees, and sets at 270 degrees. This knowledge
168 is crucial for various applications, including astronomy, navigation, and aligning solar panels for maximum sunlight
169 exposure.

170 Remark The sun’s zenith angle is the angle between the Sun and the vertical point directly above an observer on
171 Earth. At solar noon, it is 0 degrees when the Sun is directly overhead. During sunrise and sunset, it is 90 degrees
172 when the Sun is at the horizon. The zenith angle changes throughout the day due to Earth’s rotation and varies with
 8 Chatterjee & Ghosh et al.

173 the observer’s location, time of year, and time of day. It is crucial for solar energy applications, affecting the amount
174 of solar radiation received.

175 To compute the columns Z and γs of dataset 2, we utilize the position of the Sun for each species. For blue
176 whales, the dataset is sorted concerning latitude. Then, we compute the first difference in the magnitude of the sun’s
177 position (computed using the previously discussed formulas).

178 | Testing for Directional Correlation

179 In dealing with the circular variable θ, A1 , B 1 , A2 , M 1 , and M 2 , we utilize the notation of ρ to define directional corre-
180 lation. The directional correlation coefficient ρ S A (θ, A1 ) is defined as follows:

E {sin(θi − θ̄) sin(A1i − A¯1 ) }

ρ S A (θ, A1 ) = p .
V ar (sin(θi − θ̄)V ar (sin(A1i − A¯1 ) )

181 Similar formulas apply for moon azimuth (ρM A (θ, A2 )), ocean currents (ρoc (θ, O 1 )), magnetic inclination (ρi nc (θ, M 1 )),
182 and magnetic declination (ρdec (θ, M 2 )).

183 | Properties of ρi nc and ρdec

184 1. ρi nc (θ1 , M 11 ) and ρdec (θ1 , M 21 ) do not depend on the zero direction used for both variables.
185 2. |ρi nc (θ1 , M 11 ) | ≤ 1 and |ρdec (θ1 , M 21 ) | ≤ 1.
186 3. ρi nc (θ, M 11 ) and ρdec (θ, M 21 ) equal 0 if θ, M 1 , and θ, M 2 are independent, although the converse need not be
187 true.

188 | Properties of ρ S A , ρO C , and ρM A

189 The properties of ρO C , ρ S A , and ρM A mirror those of ρi nc and ρdec and are therefore omitted.

190 | Statistical Hypothesis Framework

191 Our hypothesis can be formulated statistically as follows:

H0 : ρc = 0,

H1 : ρc , 0.

192 Consider (θ1 , M 11 ), · · · , (θn , M 1n ) and (θ1 , M 21 ), · · · , (θn , M 2n ) as some random sample of observations, both
193 measured as angles concerning the same zero direction and sense of rotation. Let (θ, M 1 ) and (θ, M 2 ) denote the
194 joint probability density function on the torus 0 ≤ θ < 2π, 0 ≤ M 1 ≤ 2π, and 0 ≤ M 2 ≤ 2π. Let M̄ 2 , M̄ 1 , and θ̄
195 represent the mean direction of three variables.
196 The correlation coefficient r i nc c,n , r dec c,n , ρO C , r S A c,n , r S Z c,n , and r M A c,n are defined similarly. The sample corre-
197 lation serves as an estimation of ρc . When the joint distributions of (θ, M 1 ) and (θ, M 2 ) are not fully specified, we
 Chatterjee & Ghosh et al. 9

198 can use the sample measures for testing the hypothesis about ρc , when n is sufficiently large. For details on the test
199 statistics and their distributional properties under H 0 , we refer to [31].

200 A spherical Mixture model based Statistical Approach

201 We commence by outlining fundamental definitions and properties of specific directional statistical distributions and
202 tests. Detailed information is provided in the Appendix section of this paper. For a more comprehensive understanding,
203 readers are directed to [32].
204 A mixture model represents the probability distribution of observations in the overall population. The Gaussian
205 mixture model is frequently extended to accommodate a vector of unknown parameters, such as the von Mises Fisher
206 Distribution [33].
207 Formally, for a directional parametric mixture distribution with density p (x |θ), where θ = (θ1 , θ2 , · · · , θK ) de-
208 notes the appropriate parameter set, we define it as follows:

K
Õ
p (θ) = φi F (x |θi ). (20)
i =1

209 Here:

210 • K represents the number of mixture components,

211 • N denotes the number of observations,
212 • θi = 1 . . . K is the parameter of the distribution of observations associated with component i ,
213 • φi = 1 . . . K signifies the mixture weight, i.e., the prior probability of a particular component i ,
214 • Φ represents a K -dimensional vector set φ1 · · · K that sums to 1,
215 • z i = 1 . . . N indicates the component of observation i ,
216 • x i =1...N denotes observation i ,
217 • F (x |θ) symbolizes the probability distribution of an observation, parametrized on θ,
218 • z i =1...N follows a Categorical(φ) distribution, and
219 • x i =1...N |z i =1...N is drawn from F (x i =1...N |θz i ).

220 In our study, the i th vector component is characterized by F (x |θi ) as either a von Mises-Fisher distribution with
221 weights φi , parameter θi as means µi , and concentration matrices κi , or as a circular uniform distribution. To fit the
222 mixture von Mises Fisher distribution to both datasets, we utilized the R package movMF [34].

223 Circular-Circular Regression for overall Dataset

224 From the above analysis, we find some Directional Parameters like the whale’s migration path Direction, Solar and Lu-
225 nar positional change, Magnetic Inclination change, Magnetic declination change, Oceanic current directional change,
226 etc. We want to show if there is any relation between the positions of the Solar, lunar change, magnetic, and oceanic
227 present directional change affecting the whale path direction. We considered the direction of the Whale Migration
228 path as the dependent variable and other angular parameters as independent. Per our theory, we believe the Whale
229 Path Direction β and the other angular parameters as α.
230 (α, β ) have joint pdf f(α, β ), 0 < α , β ≤ 2π. To predict β for a given α, consider the regression or conditional
 10 Chatterjee & Ghosh et al.

231 expectation of vector exp(i β ) given α, say

E (exp i β |α ) = ρ (α ) exp(i µ (α ) ) = g 1 (α ) + i g 2 (α ) (21)

232 Equivalently

E (cos β |α ) = g 1 (α ), E (sin β |α ) = g 2 (α ) (22)

233 From which β can be determined as

g 2 (α )
µ (α ) = βˆ = arctan ∗ (23)
g 1 (α )

234 Predicting β in this way is optional because it minimizes the below equation and is similar to the least square idea.

E | | exp i β − g (α ) | | 2

235 Without further specification on the structure of g 1 (α ) and g 2 (α ), it isn’t easy to estimate these from the data. So
236 g ( α ) and g 2 (α ) will be approximated by suitable functions. This leads to the function g ( α ) and g 2 (α ) by trigonometric
237 with m degree polynomials.

m
Õ m
Õ
g 1 (α ) ≈ (Ak cos k α + B k sin k α ), g 2 (α ) ≈ (C k cos k α + D k sin k α ) (24)
k =0 k =0

238 Thus, we have the following linear model:

m
Õ
cos β = E (cos β |α ) + ϵ1 = (Ak cos k α + B k sin k α ) + ϵ1 , (25)
k =0
Õm
sin β = E (sin β |α ) + ϵ2 = (C k cos k α + D k sin k α ) + ϵ2 (26)
k =0

239 Where (ϵ1 , ϵ2 ) is error vector [31].

240 We propose a similar regression model using the response variable a change in the direction of the whale path
241 (Variable θ calculated from our data using the Haversine formula discussed in section in supplementary, Solar and
242 Lunar positional change and the magnetic field directional changes, the Ocean currents directional changes.
243 We assume the regression model stated in (27) to detect the significance of change in magnetic inclination, the
244 regression model stated in (27) to detect the importance of Magnetic Declination change, and the regression model
245 displayed in (27) to detect the significance of Oceanic current directional change. Solar and Lunar positional change
246 are also shown in (27)

θ (ψ, λ, t ) = M S A (ψ, λ) + ρ S A A1 (ψ, λ, t ) + ϵ (27)

247 The equation for all the other physicial parameters will be similar as per the equation (27). Where M S A (ψ, λ),
248 M M A (ψ, λ), M M I (ψ, λ), M M D (ψ, λ), and M O C (ψ, λ) is the mean, ρ denotes the unknown regression coefficient, and
 Chatterjee & Ghosh et al. 11

249 the error ϵ follows von Mises Fisher with mean 0 and unknown dispersion matrix Σ = σ 2 I for all the models.( [31] )
250 We assume the following components for the model.
251 The Given components are:
252 whale path level θ (φ, λ, t ) at latitude φ, longitude λ each moment (observed time) t in (year/month/date) for-
253 mat., Solar Position change(A1 ), Lunar Position change (A1 ) , Magnetic Inclination Change (M 1 ) , Magnetic Declination
254 Change(M 2 ), Oceanic current direction angular change (O 1 ), obtained by tide Gauge, M S L (ψ, λ) is the mean sea level,
255 wi = 2πfi is the angular frequency derived from tidal component frequency.
256 The unknown components are:
257 ρ (regression coefficient signifies the directional relation between change of magnetic, ocean current change,
258 Solar and lunar positions, and the Whale migration path direction change.), Variance of error ϵ

259 Results

260 We plotted the fitted mixture model contour and density plot over the world map in the figure 2a where the pink line
261 indicates the density of the plots and the yellow color is the von Mises Fishers contour plot over the locations which
262 is done with repsect to blue whale. Same for Humbpback whale where the figure 2b where blue line indicates the
263 density of the plots and the green color is the von Mises Fishers contour plot over the locations.
264 In the figure 3a, a two-dimensional plot of Humpback whales over the world map sampling some data and showing
265 its all-over path using the green arrowed line created by some randomly selected whale location point, indicating the
266 overall direction of the humpback whale is plotted. In figure 3b, there is a two-dimensional plot of the directional
267 change of Blue whales over the world map,showing its all-over path using the blue arrowed line created by some
268 randomly selected whale location point, indicating the overall direction of the blue whale is plotted.
269 The Watson test results are given in the table S3 for blue whale and table S4 for humpback whale, where we used
270 0.01 significance level. We tested if the directional parameters follow any circular distributions or not. According to
271 the results above, the directional change of the whale path follows von Mises’ distribution for blue whale and not
272 vonMises for humpback whale. However, the other parameters do not follow any circular distributions.

r value ρ Values Regression Parameters

-0.07314078 0.115751 between the change of Whale path and Geomagnetic Inclination

0.01143854 0.1067404 between the change of Whale path and Geomagnetic declination

0.02573606 0.1071596 between the change of Whale path and Angle of Incidence(Ocean Currents)

-0.1041002 0.1108125 between the change of Whale path and Solar Positional Change

0.01417289 0.1069449 between the change of Whale path and Lunar Positional Change
TA B L E 1 This table contains information regarding the ρ values that occurred after using the circular-circular
regression model, where we have used the regression model concerning the change of direction of whale path and
change of Lunar, solar position, magnetic declination, and inclination. Also, the r-values indicate the results of the
Correlation test.

273 The table 1 of blue whale, table 2 of Humpback whale contains information regarding the ρ values that occurred
274 after using the circular-circular regression model, where we have used the regression model concerning the change
275 of direction of whale path and change of magnetic declination and inclination and Solar and lunar positional change.
276 Also, the p-values and r-values indicate the results of the Hypothesis or Correlation tests. We can see that the ocean
 12 Chatterjee & Ghosh et al.

r value ρ Values Regression Parameters

0.0128 0.25061 between the change of Whale path and Geomagnetic Inclination

0.00683 0.25233 between the change of Whale path and Geomagnetic declination

-0.01721964 0.2360995 between the change of Whale path and Angle of Incidence(Ocean Currents)

-0.01670 0.25310 between the change of Whale path and Solar Positional Change

-0.003241 0.25794 between the change of Whale path and Lunar Positional Change
TA B L E 2 This table contains information regarding the ρ values that occurred after using the circular-circular
regression model, where we have used the regression model concerning the change of direction of whale path and
change of Lunar, solar position, magnetic declination, and inclination. Also, the r-values indicate the results of the
Correlation test.

277 currents are the most effective for changing a whale’s directional path. Many other factors, like ocean temperature,
278 food habits, ecosystem, etc., are also factors for whale migration and their directional change. After the hypothesis
279 test, We got significant p-values and a clear relation between the whale directional change and the derived physical
280 earth parameters. So from the table we can clearly see that the ρ value between Solar position and Geomagnetic
281 changes effects the most for both of the Whales, given in the table 1 and Table 2.

(a) Two-dimensional plot with the density and contour over (b) Two-dimensional plot with the density and contour over
the world map for blue whales. where the pink line indicates the world map for Humpback Whales. where the pink line
the density of the plots and the yellow color is the von Mises indicates the density of the plots and the yellow color is the
Fishers contour plot over the locations vonMises Fishers contour plot over the locations

282 We use the von Mises Fisher test to get an overview of the optimal number of mixtures used to fit from the below
283 table S1. We used the von Mises Fishers test to the location parameters for k or the mixture value ten and we got nine
284 as the optimal number of mixture for fitting the spherical distribution von Mises Fishers with α Values 0.13271302,
285 0.14572542, 0.11428302, 0.03346418, 0.08099035, 0.13895723, 0.15546092, 0.14022520, 0.05818068. For
 Chatterjee & Ghosh et al. 13

(a) Two-dimensional plot of humpback whales over the world

map sampling some data and showing its all over path using (b) Two-dimensional plot of blue whales over the world map
the green arrowed line which is created by some randomly sampling some data and showing its all over path using the
selected whale location point, indicating the overall direction blue arrowed line which is created by some randomly selected
of whale whale location point, indicating the overall direction of whale

F I G U R E 3 In the following images, we employ green lines to represent the density of location parameters, while
blue lines delineate the prominent contours of the location parameter.

286 humpback whale we get that the mixture value 29 as the optimal number of mixture for fitting the spherical dis-
287 tribution von Mises Fishers with α Values 0.019483538, 0.010135764, 0.024669531, 0.010134753, 0.036215830,
288 0.022854380, 0.067264868, 0.020843837, 0.038083620, 0.037776897, 0.038258586, 0.071162208, 0.088504963,
289 0.011977560, 0.050419309, 0.017583318, 0.001367945, 0.066135326, 0.013557983, 0.021967843, 0.012944754,
290 0.003773564, 0.051776058, 0.014596476, 0.023335981, 0.092918320, 0.017010088, 0.065311910, 0.049934791

291 Discussion & Conclusion

292 In terrestrial systems, the green wave hypothesis suggests that migrating animals can optimize foraging opportuni-
293 ties by following phenological changes in high-quality forage across space, akin to tracking "resource waves." [1]. We
294 investigate the migration patterns of blue and Humpback whales along the western coast of America, utilizing a com-
295 prehensive dataset that includes parameters such as longitude, latitude, change in path direction, magnetic inclination,
296 declination change, and date. We use directional statistical tools to infer the directional distributions of whale-path
297 directional changes, indicating a distinct von Mises Fisher distribution. To further refine our understanding, we apply
298 a mixture model of von Mises Fisher under appropriate reasons, determining that a combination of nine mixtures best
299 fits the directional changes exhibited by the blue and Humpback whales based on the Bayesian Information Criterion
300 (BIC) values, suggesting the presence of multiple underlying distributions in the whale migration direction change.
301 Additionally, we conduct a novel circular-circular regression analysis suitable for the particular scenario considered in
 14 Chatterjee & Ghosh et al.

302 this paper, with directional change as the dependent variable and parameters such as magnetic inclination, declination,
303 and ocean currents as independent variables. Most of the results of our hypothesis tests demonstrated a statistically
304 significant non-zero association, implying a potential dependence of blue and Humpback whale directional changes
305 on these environmental factors. According to the table presented in both Table 1 and Table 2, it’s evident that the
306 correlation coefficient (ρ) between Solar position and Geomagnetic changes has the most significant effect on both
307 Whale species.

308 Funding information

309 No funding received.

310 Acknowledgment

311 Debashis Chatterjee thanks Visva Bharati, Santiniketan, and Prithwish Ghosh thanks North Carolina State University,
312 USA.

313 Author Contribution

314 D.C. designed, conceptualized, and developed the research and synthesized interdisciplinary statistical methodologies
315 and models. P.G. conceptualized the model, collected and prepared the datasets, wrote codes for various modified
316 datasets, and performed code-based analysis (mainly using R and Python). D.C. and P.G. covered the alignments of
317 relevant Earth science-related theories and methodologies. P.G., D.C. wrote, modified, and reviewed the manuscript.

318 Supporting Information

319 Our dataset contains the response (dependent) and covariate (independent). The data used for this article can be
320 accessed through the following links.

321 1. Response Data: The change of direction of the Whale’s path is the response data. The original data set of [18] [35],
322 which contains information about the Blue whale, can be found at the movebank repository [36]. We modified that
323 according to the definition of our response part and kept it in Zenodo https://zenodo.org/records/11205218
324 and can be accessed publicly.
325 2. Covariate Data: The independent variables we calculated consist of directional change in three components:
326 Magnetic Inclination and Magnetic Declination. Per the shared code, our compiled dataset is uploaded to Zenodo
327 https://zenodo.org/records/11205218 and can be accessed publicly.

328 references
329 [1] Abrahms B, Hazen EL, Aikens EO, Savoca MS, Goldbogen JA, Bograd SJ, et al. Memory and resource tracking drive blue
330 whale migrations. Proceedings of the National Academy of Sciences 2019;116(12):5582–5587.

331 [2] Allen AN. An investigation of the roles of geomagnetic and acoustic cues in whale navigation and orientation. PhD
332 thesis, Massachusetts Institute of Technology; 2013.
 Chatterjee & Ghosh et al. 15

333 [3] National Oceanic and Atmospheric Administration, Blue Whale 5-Year Review; Year. https://www.fisheries.noaa.gov/
334 resource/document/blue-whale- 5-year-review.

335 [4] Mate BR, Lagerquist BA, Calambokidis J. Movements of North Pacific blue whales during the feeding season off southern
336 California and their southern fall migration 1. Marine Mammal Science 1999;15(4):1246–1257.

337 [5] Wursig B, Perrin WF, et al. Encyclopedia of marine mammals. Academic Press; 2009.

338 [6] Sears R, Perrin WF. Blue whale: Balaenoptera musculus. In: Encyclopedia of marine mammals Elsevier; 2009.p. 120–
339 124.

340 [7] Ferrie C, Petrou K. ABCs of Oceanography. Sourcebooks, Inc.; 2020.

341 [8] Lesage V, Gavrilchuk K, Andrews RD, Sears R. Foraging areas, migratory movements and winter destinations of blue
342 whales from the western North Atlantic. Endangered Species Research 2017;34:27–43.

343 [9] Torres-Florez JP, Olson PA, Bedriñana-Romano L, Rosenbaum H, Ruiz J, LeDuc R, et al. First documented migratory
344 destination for eastern South Pacific blue whales. Marine Mammal Science 2015;31(4):1580–1586.

345 [10] Barlow J, Schorr GS, Falcone EA, Moretti D. Variation in dive behavior of Cuvier’s beaked whales with seafloor depth,
346 time-of-day, and lunar illumination. Marine Ecology Progress Series 2020;644:199–214.

347 [11] Redaelli L, Mangia Woods S, Landea R, Sayigh L. Seasonal Trends and Diel Patterns of Downsweep and SEP Calls in
348 Chilean Blue Whales. Journal of Marine Science and Engineering 2022;10(3):316.

349 [12] Unwin M. Migration: incredible animal journeys. Bloomsbury Publishing; 2020.

350 [13] Harris C. Whales Sing. Pan Macmillan; 2006.

351 [14] Gupta NJ, Dwivedi V, Singh BP, Bhardwaj SK. Orientation in Migrating Animals: Role of Biological Clocks. Biological
352 Timekeeping: Clocks, Rhythms and Behaviour 2017;p. 643–658.

353 [15] Hamabe K, Matsuoka K, Kitakado T. Estimation of abundance and population dynamics of the Antarctic blue whale in
354 the Antarctic Ocean south of 60° S, from 70° E to 170° W. Marine Mammal Science 2023;39(2):671–687.

355 [16] Paarman S, Vermeulen E, Seyboth E, Thornton M, Findlay K. Abundance and distribution of Antarctic blue whales
356 Balaenoptera musculus intermedia off the Queen Maud Land coast of Antarctica. African Journal of Marine Science
357 2021;.

358 [17] Dodson S, Abrahms B, Bograd SJ, Fiechter J, Hazen EL. Disentangling the biotic and abiotic drivers of emergent migratory
359 behavior using individual-based models. Ecological Modelling 2020;432:109225.

360 [18] Mate B, Palacios D, Irvine L, Follett T, Data from: Behavioural estimation of blue whale movements in the Northeast
361 Pacific from state-space model analysis of satellite tracks. Movebank data repository; 2019. http://dx.doi.org/10.
362 5441/001/1.5ph88fk2.

363 [19] Bartlett D. Essentials of positioning and location technology. Cambridge University Press; 2013.

364 [20] Chulliat A, Macmillan S, Alken P, Beggan C, Nair M, Hamilton B, et al. The US/UK world magnetic model for 2015-2020.
365 US/UK world magnetic model 2015;.

366 [21] Rebollo J, Khater S, Coupe WJ. A recursive multi-step machine learning approach for airport configuration prediction.
367 In: AIAA Aviation 2021 Forum; 2021. p. 2406.

368 [22] World Magnetic Model;. Accessed: 2023-11-30. https://www.ngdc.noaa.gov/geomag/WMM/data/WMMReports/TRWMM_

369 2005.pdf.
 16 Chatterjee & Ghosh et al.

370 [23] Todhunter I. Spherical trigonometry, for the use of colleges and schools: with numerous examples. Macmillan; 1863.

371 [24] Robusto CC. The cosine-haversine formula. The American Mathematical Monthly 1957;64(1):38–40.

372 [25] Bullock R. Great circle distances and bearings between two locations. MDT, June 2007;5.

373 [26] Access Oceanographic and Environmental Data with R;. Accessed: 2023-10-30. https://semba-blog.netlify.app/
374 10/23/2018/wind-in-the-western-indian-ocean-region/.

375 [27] Geostrophic balance;. Accessed: 2023-10-30. https://uw.pressbooks.pub/ocean285/chapter/geostrophic-balance/.

376 [28] Doglioni F, Ricker R, Rabe B, Kanzow T. Sea surface height anomaly and geostrophic velocity from altimetry measure-
377 ments over the Arctic Ocean (2011–2018). Earth System Science Data Discussions 2021;2021:1–46.

378 [29] McMillan B, Musick JA. Oceans. Simon and Schuster; 2007.

379 [30] El Mghouchi Y, El Bouardi A, Choulli Z, Ajzoul T. New model to estimate and evaluate the solar radiation. International
380 Journal of Sustainable Built Environment 2014;3(2):225–234.

381 [31] A SenGupta SRJ. Topics in Circular Statistics. Volume 5, World Scientific Publishing Co. Pte. Ltd; 2001.

382 [32] Jammalamadaka SR, SenGupta A. Topics in circular statistics, vol. 5. world scientific; 2001.

383 [33] Ghosh P, Chatterjee D, Banerjee A. On the directional nature of celestial object’s fall on the earth (Part 1: Distribution of
384 fireball shower, meteor fall &amp; crater on earth’s surface). Monthly Notices of the Royal Astronomical Society 2024
385 04;p. stae1066. https://doi.org/10.1093/mnras/stae1066.

386 [34] Hornik K, Grün B. movMF: An R package for fitting mixtures of von Mises-Fisher distributions. Journal of Statistical
387 Software 2014;58(10):1–31.

388 [35] Andrews-Goff V, Gales N, Childerhouse SJ, Laverick SM, Polanowski AM, Double MC. Australia’s east coast humpback
389 whales: Satellite tag-derived movements on breeding grounds, feeding grounds and along the northern and southern
390 migration. Biodiversity Data Journal 2023;11.

391 [36] Movebank Data;. Accessed: 2023-11-30. https://datarepository.movebank.org/entities/datapackage/b7853a1e-

392 43fc- 483d- 906a- 6991ec676413.