Bollettino di Geofisica Teorica ed Applicata

Vol. 57, n. 3, pp. 275-285; September 2016

DOI 10.4430/bgta0176

Is the bimodal oscillating Adriatic-Ionian circulation

a stochastic resonance?
F. Crisciani1 and R. Mosetti2
1
Department of Physics, University of Trieste, Italy
2
Istituto Nazionale di Oceanografia e di Geofisica Sperimentale - OGS, Trieste, Italy

(Received: October 21, 2015; accepted: April 14, 2016)

ABSTRACT The observed pseudoperiodic reversal of the upper-layer circulation of the Ionian Sea
has been assumed to be related to some internal feedback processes (density driven)
by the so-called BiOS (Adriatic-Ionian Bimodal Oscillating System) hypothesis. The
idea behind this paper is that the Ionian circulation is driven by what happens in the
boundary of the region, and so a very simple dynamical model has been derived by
integrating in the bounded flow domain (the Ionian Basin) on the f-plane of the two-
layer quasi-geostrophic evolution equations. Furthermore, a nonlinear term which
mimics the mutual interaction between the moving layers has been introduced. By
considering a stochastic Gaussian noise acting as a forcing and a decadal periodic
input term having an amplitude much smaller than the noise variance, a stochastic
resonance appears in the solution for a certain range of the values of the parameters.
The decadal forcing at the boundary of the Ionian basin can be related to the low-
frequency atmospheric variability in the surrounding basins. The resonance has a period
consistent with the observed decadal time scale of the reversal between cyclonic and
anti-cyclonic phases.

Key words: Water circulation, dynamic model, Adriatic Sea, Ionian Sea.

1. Introduction

In recent years, the conjecture that the Adriatic-Ionian basin behaves like a bimodal
oscillating system, between cyclonic and anticyclonic upper-layer circulation, has been taken
into account to explain some observational facts (Fig. 1).
This mechanism is called BiOS. [Adriatic-Ionian Bimodal Oscillating System; Civitarese
et al. (2010), Gacic et al. (2010)]. BiOS is a new and different way of explaining the Ionian
alternance in the circulation between cyclonic and anti-cyclonic states, because it attributes its
causes to internal processes more than to external ones (such as wind stress). In fact, looking
at Fig. 1, it is clear that the wind-stress curl is unable to reverse the circulation. BiOS can
give a more general explanation for the phenomenon, based on an “intrinsic” pseudo-periodic
oscillation. In addition, the overall surface circulation of the Adriatic Sea is always cyclonic,
and salinity and density data collected in the southern Adriatic (the main source of eastern
Mediterranean deep water) show a decadal variation coherent with the northern Ionian change
in sea level height, and in general, their periodic inversion is compatible with the inversion of

© 2016 – OGS 275

Boll. Geof. Teor. Appl., 57, 275-285 Crisciani and Mosetti

Fig. 1 - Circulation time series in the Adriatic and Ionian basins and wind-stress curl. Data have been obtained by
satellite altimetry by using the geostrophic assumption (courtesy of Miroslav Gacic).

circulation. The phenomenology can be described as shown in Fig. 2. The low-salinity MAW
(modified Atlantic water) advected by the anticyclonic circulation enters the Adriatic basin: this
causes a density decrease in the Adriatic Sea (since MAW is less salty), so the ADW (Adriatic
deep water) produced in this period has a lower density than usual. This “light” water spreads
towards the Ionian flanks, so that the water level along Ionian flanks gradually increases,

Fig. 2 - Stochastic resonance found for parameters: μ = 0.11811, T = 100 months, A = 0.05 (lower). Solution with
μ = 0.13241: random motion (upper) (x-axis: time (months); y-axis: upper layer circulation;positive cyclonic).

276
Is the bimodal oscillating Adriatic-Ionian circulation a stochastic resonance? Boll. Geof. Teor. Appl., 57, 275-285

generating an upwelling of the denser water of the lower layers (and a consequent downwelling
of isobars). This leads to a weakening of the anticyclonic upper-layer circulation, which at the
end gets inverted. The cyclonic regime favors LIW (Levantine intermediate water) to enter the
Adriatic, causing deep waters to be saltier, and thus heavier. As before, denser ADW spreads
along Ionian flanks increasing their density: this causes a gradual downwelling of the upper
lighter density water (and a consequent upwelling of isobars) and a gradual weakening of the
cyclonic regime, which is also, then, inverted.
Contextualizing to the time interval, in 1994, the anticyclonic gyre extended over the
northern Ionian, advecting MAW into the Adriatic (low salinity water). In 1995, anticyclonic
gyre became less and less strong and, in 1998, a cyclonic circulation was established. The
salinity in the Adriatic basin started to increase again, since saltier LIW was now advected into
it, as is evident in Fig. 2. It is important to underline that the anticyclonic trend advects mostly
MAW, since the pathway followed by waters is towards the western border of the Ionian. On the
other hand, the cyclonic motion induces waters mostly on the eastern part of the Ionian, pushing
waters saltier than MAW into the Adriatic basin. The inversion occurred in 1997, in fact,
between 1998 and 2005 a cyclonic gyre was observed, and this caused salty Levantine or Cretan
water to enter, inverting the salt content of the ADW formed until now. Another inversion started
in 2006. Currently, the circulation seems to be in a cyclonic phase (see again Fig. 1).

2. Model equations

In this section, we derive the nondimensional, two-layer quasi-geostrophic model [see

Cavallini and Crisciani (2013), for full details of this section], which will be used in a suitable
integrated form. With reference to the f – plane and in standard notation, in both layers the
governing equation has the form:
∂ 2 ∂w 1
∇ ψ + J (ψ, ∇ 2 ψ )= + Re −1 ∇ 4 ψ (1)
∂t ∂z
Each layer extends over a bounded and simply connected fluid domain D of the (x, y) plane.
Consider first the upper layer (superscript “I”) included between the rigid lid in z = 1
and the impermeable interface below z = 1, say in zi = (Hi + h)/H. According to this notation,
z* = H is the dimensional top of the whole fluid body at rest, while z* = H + h (< H) is the
dimensional interface between the moving layers, where h (<< Hi) is the fluctuating part of the
interface. Because of mass conservation and because of the impermeability of the interface

∫Dhdxdy = 0 (2)

Thus, vertical integration of Eq. 1 over the mean depth (H – Hi)/H of upper layer yields

∂ 2 I
∇ ψ + J (ψ I , ∇ 2 ψ I )=
H
[w 1 (z = 1)− w 1 (z = z i )]+ Re −1 ∇ 4 ψ I (3)
∂t H − Hi


Indeed the final equations (see Eqs. 11 and 17) are not sensitive to a possible β – effect.

277
Boll. Geof. Teor. Appl., 57, 275-285 Crisciani and Mosetti

Because no stress is applied in z = 1, then

w1 (z = 1) = 0. (4)

On the other hand, given that h is related to both the stream functions ψ I, ψII by the
relationship

f 0 UL II (5)
h= (ψ − ψ I )
g ʹ

where g´ is reduced gravity, then the vertical velocity at the interface can be written as

⎡∂ ⎤ (6)
w (z = z i ) = Fε ⎢ (ψ II − ψ I )+ J (ψ I , ψ II )⎥ + ℑ(ψ I − ψ II )
⎣ ∂t ⎦

f 2 L2
where F = 0 is the rotational Froude number. The first term at the right hand side of
gʹH
Eq. 6 takes into account the effect of fluctuations of the interface on its vertical velocity. The
second term represents the mutual interaction between the moving layers which affects the
same velocity further on. As in the barotropic limit ψI = ψII, in general ℑ(0 ) = 0 . This term
will be suitably parameterized after the integration of the vorticity equations on D. Keeping in
mind Eqs. 3, 4, and 6 and recalling that, up to the first order in ε, w = εw1 we have after a little
algebra,
D0 2 I
[∇ ψ − FI (ψ I − ψ II )]+ H ℑ(ψ I − ψ II )= Re −1 ∇ 4ψ I (7)
Dt ε(H − H i )

D ∂ f 02 L2
where, in the upper layer, 0 = + J (ψ I , ⋅ ) and F I = . At this point, by using
Dt ∂t gʹ(H − H i )

Reynolds’ transport theorem at the left hand side and the divergence theorem at the right hand
side, Eq 7 can be integrated over D to give

(8)

Setting

CI = ∫ u I ⋅ tˆds, CII = ∫ u II ⋅ tˆds (9)

∂D ∂D

where C is the circulation of the geostrophic current uu = k̂ × ∇ψ along the unique boundary

d dC I
�D of the fluid domain, we have ∫ ∇ 2 ψ I dxdy = and, because of Eqs. 2 and 5,
dt D dt

278
Is the bimodal oscillating Adriatic-Ionian circulation a stochastic resonance? Boll. Geof. Teor. Appl., 57, 275-285

�(ψI – ψII) dxdy = 0. As anticipated, the parameterization of ∫ ℑdxdy is now assumed in the form
D

ℑ(ψ I − ψ II ) dxdy = (C I − C II) )

1
(10)
33

εD
∫
Some consequences of Eq. 10 in the present context are discussed in the Appendix. Finally,
the flux of relative vorticity ∇ 2 ψ I across the boundary of the domain, that is

∫ ∇(∇ ψ )⋅ n̂ds
2 I

∂D

is prescribed in terms of a given function of time only, say φ (t) . This term comes from the
lateral action of the surrounding flow which forces the upper layer at its boundary. On the
whole, Eq. 8 yields
dC I
+
H
(C I − C II ))33 = φϕ(t ) (11)
dt (H − H i )
Consider now the lower layer (superscript “II”) included between the interface in z = zi
and the bottom in z = η* / H = η. Vertical integration of Eq. 1 over the mean depth Hi / H of the
lower layer yields
∂ 2 II H
∇ ψ + J (ψ II , ∇ 2 ψ II )= [w 1 (z = z i )− w 1 (z = η)]+ Re −1 ∇ 4 ψ II (12)
∂t Hi
where the vertical velocity is again given by Eq. 6 and
E v 2 II
w 1 (z = η) = FJ(ψ II , η)+ ∇ψ (13)
2ε
Eq. 13 is nothing but the nondimensional version of Eq. 4.9.36 of Pedlosky (1987). By using
Eqs. 6 and 13 into Eq. 12, after a little algebra, one obtains
DD0 0 2 2 IIII IIII IIII I I
[∇[∇ψψ −−FF (ψ(ψ −−ψψ )])]−− HH ℑℑ(ψ(ψI I−−ψψIIII))++ HH FFIIIIJJ(ψ(ψIIII, ,ηη))
Dt
Dt εεHHi i HHi i
(14)
HH EEv v 2 2 IIII
++ ∇∇ψψ ==Re
Re−1−1∇∇4 ψ
4 IIII
ψ
HHi i 22εε
D ∂ f 2 L2
where, in the lower layer, 0 = + J (ψ II , ⋅ ) and FII = 0 . Eq. 14 can be integrated on the
Dt ∂t gʹH i

domain D along the same lines as Eq. 8 and noting also that, because of Stoke’s theorem,

∫ J(ψ
II
, η) dxdy = ψ II ∫ ∇η × k̂ ⋅ n̂ds = ψ II ∫ ∇η ⋅ t̂ds = 0 (15)
D ∂D ∂D

where ψ II is the constant value taken by the stream function on the boundary. Eq. 15 shows that
the model is not sensitive to bottom topography. Hence, integration of Eq. 14 over D yields
dC II H I
− (C − C II ))33 + H E v II
C = Re −1 ∫ ∇(∇ 2 ψ II )⋅ n̂ds (16)
dt Hi εH i 2 ∂D

279
Boll. Geof. Teor. Appl., 57, 275-285 Crisciani and Mosetti

The line integral at the right hand side of Eq. 16 is the flux of relative vorticity crossing the
boundary of the lower layer, which has the same meaning as that of the upper layer. However,
unlike the upper layer, this flux is prescribed to be a constant (say φ0), so the final form of Eq. 16 is

− (C − C II )) +
dC II H I H E v II
C =φ (17)
33
ϕ0
dt Hi εH i 2
Note that

H Ev HE 1 (18)
=
εH i 2 H i 2ε

where HE is the thickness of the Ekman bentic layer. According to the orders of magnitude (in
S.I. units ) consistent with this investigation and listed below

HE = O(10), Hi = O (3.5×103), U = O (5×10-2), L = O (5×105)

the quantity in Eq. 18 is O(1). To summarize, the coupled Eqs. 11 and 17 give the time evolution
of the circulation in both fluid layers once all the inputs are established and definite initial
conditions

C I (0 ) = C 0I , C II (0 ) = C 0II (19)

are fixed. In what follows, the values (in S.I. units)

H Ev
H = 4 × 10 3 , H i = 3 .5 × 10 3 , =1 (20)
εH i 2
will be adopted, while φ(t) and φ0 will be chosen within the framework of stochastic resonance,
so the constants in Eq. 19 are expected to play a minor role in the long-term behaviour of Eq. 9.

3. The stochastic resonance hypothesis

The mechanism of stochastic resonance was introduced in a seminal paper (Benzi et al.,
1981). They discovered that a stochastic system (having motion in all time scales) forced by a
periodic signal may create in the solution a bi-stable response characterized by a rapid transition
between the two states with the same period of the forcing term. Since the discovery, the
concept of stochastic resonance has been widely applied to several disciplines in the context of
dynamical systems and in particular to climate dynamics (e.g., Benzi et al., 1982). Furthermore,
stochastic resonance has been investigated in the framework of the Stommel thermohaline
model (Eyink, 2005).
The idea is to explore if the dynamical system described by Eqs. 11 and 17 could exhibit
a stochastic resonance when a periodic decadal forcing and a random noise are introduced.
Because the BiOS process is supposed to be a bi-stable system, the stochastic resonance could
be a candidate model to explain the observed pseudo-periodic reversal of the Ionian vorticity.

280
Is the bimodal oscillating Adriatic-Ionian circulation a stochastic resonance? Boll. Geof. Teor. Appl., 57, 275-285

Notice that in the framework of this model, the forcing is acting on the boundary region of
the Ionian Sea. This forcing derives from the lateral action of the surrounding flows. The
introduction of a decadal periodicity into the system can be supported by the fact that large-scale
ocean processes are influenced by atmospheric forcing and the appearance of a decadal scale
is quite common (Cessi and Louazel, 2001; Wunsch, 2011, Kondrashov and Berloff, 2015).
Following these considerations, we modify the system of Eqs. 11, and 17 as follows:
dC I
+
H
(C I − C II ))33 = φϕ(t ) = A sin(2πt / T) + μs(t )
dt (H − H i )
(21)
(C − C II))33 + H v C II = 0
dC II
H I E
−
dt Hi εH i 2

where A is the amplitude of the forcing; T is the period (decadal); μs(t) is a Gaussian noise
process physically representing the high-frequency signal (time scales from the monthly/
seasonal to the interannual) of the forcing acting on the boundary region.
The numerical integration of the stochastic system 21 has been made by means of a
MATLAB code based on the Runge-Kutta ODE solver. A large set of runs have been done
by small changes in the variance of the random noise since the stochastic resonance usually
appears within a very narrow range of variance (Fig. 3). As can be seen in Fig. 3, the solutions
are strongly sensitive to the ratio between the periodic amplitude and the noise variance. The
stochastic resonance can be found, being A= 0.05, in a range between μ = 0.115 and μ = 0.120.

Fig. 3 - The BiOS mechanism (after

Civitarese et al., 2010).

Notice that, despite the weakness of the periodic forcing with respect to the noise variance,
the solution shows stochastic transitions which are correlated to the period of 100 months
present in the forcing. Note also that the average time duration of a cyclonic or anticyclonic
phase is one-half of the period (with the transition time much shorter between the two states).
This is a typical behaviour of a stochastic resonance. In other words, the mean residence time τ
of the stable states is related to the period T by the relation T � 2 τ.

281
Boll. Geof. Teor. Appl., 57, 275-285 Crisciani and Mosetti

4. Discussion

The driving mechanisms behind the decadal reversal of the Ionian Sea upper-layer circulation
provoked a considerable debate in the Mediterranean scientific community. It is still unclear
what the driving force is. It has been suggested that the reversal could be driven by variations
in wind-stress curl over the basin, baroclinic dynamics acting within the Adriatic-Ionian
System (AIS), or baroclinic dynamics driven by thermohaline properties at the AIS eastern
boundary. The variability observed in the Ionian circulation has been often associated with a
change in the wind-stress curl over the area (Pinardi et al., 1997, 2015). Recently, Theocharis
et al. (2014) suggested that the decadal variability observed in the Ionian circulation reflects
an internal mechanism driving the alternation of the Adriatic Sea and the Aegean Sea as main
DWF (deep water formation) sites for the eastern Mediterranean (EMED). This thermohaline
pumping (Theocharis et al., 2014; Velaoras et al., 2014) involves the whole thermohaline
cell of the EMED, and not only the Adriatic-Ionian Sea, as in the BiOS hypothesis (Gacic
et al., 2010). In both cases, the role of long term atmospheric forcings has been considered
negligible. In a recent paper by Mihanović et al. (2015), it is claimed that BiOS is the dominant
generator of the Adriatic decadal variability. On the contrary, in the present paper, it is assumed
that the upper-layer Ionian circulation is driven by the forcing acting on the boundary of the
Ionian region as suggested by the scheme of circulation depicted in Fig. 2. In this hypothesis,
neither the baroclinicity nor the wind-stress curl inside the Ionian basin are responsible for the
inversion of the circulation at decadal time scale. The forcing is due to both interannual and
decadal variability of the Adriatic and the adjacent basins’ meteo-oceanographic conditions
forcing the boundary of the basin (it could also be the effect of the Aegean Sea in the case
of the East Mediterranean Transient). The presence of a decadal variability in the Adriatic
Sea oceanographic condition has been recently shown by, among others, a new analysis
of experimental data in the paper by Mihanović et al. (2015). In our opinion, this is due to
the corresponding decadal scale in the atmospheric forcing. In this sense, the already cited
paper by Pinardi et al. (2015) reinforces this conjecture, since it gives a clear picture of the
mean circulation of the Mediterranean Sea emerging from a 23-year eddy-resolving model
reanalysis. The multi-decadal mean flow emerging from this analysis is consistent with many
of the previous findings, but new circulation structures also become apparent. It is clearly
stated that “The Ionian reversal mechanism is also evident in the stream function, thus giving
further evidence of the wind-driven nature of the mechanisms underlying the northern Ionian
reversal”. In the present approach, even if the decadal atmospheric signal could be very weak
with respect to the interannual variability, it is reflected into the boundary of the upper layer of
the Ionian region driving the circulation inversion through a stochastic resonance mechanism.
The hypothesis that a purely termohaline process could explain the Ionian circulation reversal
has also been tested by using a thermohaline (with temperature and salinity state variables)
three-layer box model following the approach by Ashkenazy et al. (2012), in a Master’s degree
thesis (Capuano, 2015). By using the typical values for deep-water formation rate in the Adriatic
Sea as well as the realistic values for the fluxes between the Adriatic box, the Ionian box, and
the eastern Mediterranean box, no periodic solutions appear, and so there is no possible reversal
of the Ionian gyre in the upper layer (500 m). In this sense, the model developed here moves
to a possible conclusion that BiOS is a forced mechanism induced by the typical atmospheric

282
Is the bimodal oscillating Adriatic-Ionian circulation a stochastic resonance? Boll. Geof. Teor. Appl., 57, 275-285

decadal scales. It is still not clear if the decadal variability is connected, for example, to the
NAO index or to some other atmospheric teleconnections in the Mediterranean area.
The model developed here does not have any relation to the deep water formation processes.
As speculated in the above-cited paper by Mihanović et al. (2015), the dense water outflow
rates from the Adriatic through the Otranto Strait (-0.3 Sv) does not have the capacity to rapidly
invert the circulation in the upper Ionian Sea. On the contrary, the stochastic resonance shows,
as a typical feature, rapid switches from one regime to the other. However, due to the detailed
knowledge of the thermohaline structure of the Adriatic-Ionian basins coming from longtime
CTD data , further improvements could be expected by using a quasi-geostrophic continuously
stratified numerical model to investigate the role of baroclinicity and that of the Adriatic deep
water.

Acknowledgements. The authors gratefully thank their colleague dr. Fabio Cavallini for a useful discussion
about mathematical aspects of the investigation. Thanks also to dr. Miroslav Gacic for his comments on the
BiOS phenomenology.

REFERENCES
Ashkenazy Y., Stone P.H. and Malanotte-Rizzoli P.; 2012: Box modeling of the Eastern Mediterranean sea. Phys. A:
Statistical Mechanims and its Applications, 391, 1519-1531.
Benzi R., Sutera A. and Vulpiani A.; 1981: The mechanism of stochastic resonance. J. Phys. A: Math. Gen., 14, L453-
L457.
Benzi R., Parisi G., Sutera A. and Vulpiani A.; 1982: Stochastic resonance in climate change. Tellus, 34, 10-16.
Capuano C.; 2015: Non-linear box models for the study of the interannual variability of the thermohaline parameters
of Adriatic-Ionian basin. University of Trieste, Inter athenaeum Master’s Degree in Physics - Earth and
Environmental Physics Curriculum, Academic Year 2013-2014. Supervisor: Renzo Mosetti.
Cavallini F. and Crisciani F.; 2013: Quasi-geostrophic theory of oceans and atmosphere: topics in the dynamics and
thermodynamics of the fluid earth. Springer, Dordrecht, The Netherlands, 385 pp.
Cessi P. and Louazel S.; 2001: Decadal oceanic response to stochastic wind forcing. J. Phys. Oceanogr., 31, 3020-
3029.
Civitarese G., Gacic M., Borzelli G.L. and Lipizer M.; 2010: On the impact of the Bimodal Oscillating System (BiOS)
on the biogeochemistry and biology of the Adriatic and Ionian Seas (Eastern Mediterranean). Biogeosci., 7,
3987-3997.
Eyink G.L.; 2005: Statistical hydrodynamics of the thermohaline circulation in a two-dimensional model. Tellus, 57A,
100-115.
Gacic M., Borzelli G.L.E., Civitarese G., Cardin V. and Yari S.; 2010: Can internal processes sustain reversals of the
ocean upper circulation? The Ionian Sea example. Geophys. Res. Lett., 37, L09606.
Kondrashov D. and Berloff P.; 2015: Stochastic modeling of decadal variability in ocean gyres. Geophys. Res. Lett.,
42, 1543-1553, doi:10.1002/2014GL062871.
Mihanović H., Vilibić I., Dunić N. and Sepić J.; 2015: Mapping of decadal middle Adriatic oceanographic variability
and its relation to the BiOS regime. J. Geophys. Res., 120, 5615-5630, doi:10.1002/2015JC010725.
Pedlosky J.; 1987: Geophysical fluid dynamics. Springer, New York, 710 pp.
Pinardi N., Korres G., Lascaratos A., Roussenov V. and Stanev E.; 1997: Numerical simulation of the interannual
variability of the Mediterranean Sea upper ocean circulation. Geophys. Res. Lett., 24, 425-428.
Pinardi N., Zavatarelli M., Adani M., Coppini G., Fratianni C., Oddo P., Simoncelli S., Tonani M., Lyubartsev V.,
Dobricic S. and Bonaduce A.; 2015: Mediterranean Sea large-scale low-frequency ocean variability and water
mass formation rates from 1987 to 2007: a retrospective analysis. Prog. Oceanogr., 132, 318-332.
Theocharis A., Krokos G., Velaoras D. and Korres G.; 2014: An internal mechanism driving the alternation of the
eastern Mediterranean dense/deep water sources. In: Borzelli G.L.E., Gacic M., Lionello P. and Malanotte-
Rizzoli P. (eds), The Mediterranean Sea: temporal variability and spatial patterns, John Wiley, Oxford, U.K., pp.
113-137, doi:10.1002/9781118847572.ch8.

283
Boll. Geof. Teor. Appl., 57, 275-285 Crisciani and Mosetti

Velaoras D., Krokos G., Nittis K. and Theocharis A.; 2014: Dense intermediate water outflow from the Cretan Sea:
a salinity driven, recurrent phenomenon, connected to thermohaline circulation changes. J. Geophys. Res., 119,
4797-4820, doi:10.1002/2014JC009937.
Wunsch C.; 2011: The decadal mean ocean circulation and Sverdrup balance. J. Mar. Res., 69, 417-434.

Corresponding author: Renzo Mosetti

Istituto Nazionale di Oceanografia e di Geofisica Sperimentale - OGS
Via Beirut 4, 34134 Trieste, Italy
Phone: +39 040 2249710; fax: +39 040 2249770; e-mail: rmosetti@ogs.trieste.it

Appendix. Some consequences of parameterization in system 10

The dynamic systems 11 and 17 without bottom friction and forcing becomes

⎧ dC H
)
I
II 33
⎪⎪ dt + H − H (C − C ) = 0
I

i (A.1)
)
⎨
dC II H I
⎪ − (C − C II )33 = 0
⎩⎪ dt Hi

The question is whether system A.1 describes some kind of unforced circulation controlled
only by the interaction between the quasi-geostrophic layers. Initial conditions in Eq. 19
complete this model. Integration of system A.1 with Eq. 19 yields

CC00 ++ CC00 ++ ((CC00 −−CC0II0II)(

)()(11++KK33 tt)))−−1–1/2
HH−−HHi i I I HHi i IIII HHi i I I
CCI I((tt))== (A.2)
1/ /22

HH HH HH

) =I (tH) =− H i−CH0I i+CH0I i+CH0II −i CH0II −+ HHi i ((CC0I 0I −−CC0II0II))(

CII (tC ()(11++KK3 3 t )t−))1−/–1/2
12/ 2
(A.3)
H H H H HH

where K 3 = 2(C 0I − C 0II ))

H2
(>0). The evolution of the circulation goes as follows
22

H i (H − H i )

H − H i I H i II
C 0I → C I (t )→ C0 + C0
t >0
H
t →∞
H (A.4)
H − H i I H i II
C 0II → C II (t )→ C0 + C0
t >0 t →∞
H H
In other words, system A.1 evolves towards a homogeneous and steady circulation, given
by the average of C 0I , C 0II , each of them being weighted by the relative thickness of the related
layer. However, the rate of the asymptotic convergence shown in Eqs. A.4 is, in general,
different in each layer. In fact, solution of Eqs. A.2 and A.3 gives the ratio
dC I dC II Hi
/ =− (< 0 ) (A.5)
dt dt H − Hi
For instance, if H – H i << H i (in accordance with Eq. 20), then Eq. A.5 implies
dC I / dt >> dC II / dt . Inequality A.5 also shows that the final steady state is reached by a

284
Is the bimodal oscillating Adriatic-Ionian circulation a stochastic resonance? Boll. Geof. Teor. Appl., 57, 275-285

decrease of the circulation rate in one layer and the increase in the other.
The monotonic character of solution of Eqs. A.2 and A.3 implies that CI (t) reverts its sign at
most once in the course of time if and only if CI (0) and lim C I (t ) have discordant signs, that is
to say if and only if
t →+∞

⎛ H − H i I H i II ⎞
C 0I ⎜ C0 + C0 ⎟ < 0 (A.6)
⎝ H H ⎠
For the same reason, CII (t) reverts its sign at most once in the course of time if and only if

⎛ H − H i I H i II ⎞
C 0II ⎜ C0 + C0 ⎟ < 0 (A.7)
⎝ H H ⎠
However, the inversion of the circulation in both layers is not possible, even in deferred
times. In fact, both Eqs. A.6 and A.7 imply

C 0I C 0II < 0 (A.8)

so C0I and CII0 should have concordant signs because of Eqs. A.6 and A.7, but discordant because
of Eq. A.8.
Another question is the reason why just the third power of the difference CI – CII is introduced
in Eq. 10. Integration of system A.1 with n = 2,3,4... in place of the third power appearing in
system A.1 results in the equations

(C0 − C0II)(
)(11+ +KKn ntt))1−1– nn
H − H i I H i II H i I 11
C I (t ) =
––––––

C0 + C0 +
H H H
(A.9)
(C0 − C0II)()(11+ +KKn ntt))1−1– nn
H − H H H − Hi I 11
C II (t ) =
––––––
i
C 0I + i C 0II −
H H H
where

KKn n==(n(n−−11))()(C C0II0II0II)))

n −nn–1 HH2 2
(CC 0I 0I0I−–−CC  1−1
(A.10)
(H−−HHi )i )
HHi (i H
Solution of Eqs. A.9 coincides trivially with Eq. A.2 and A.3 for n = 3. Consider the function
of time

φ(t ) = (1 + K n t ))11– n
11

(A.11)
––––––
−n

included in the above class of solutions and, owing to the arbitrariness of initial conditions,
assume C0I – CII0 < 0. The following alternative holds:
~
1) the integer n is even. Hence Kn < 0 so Eq. A.8 has a singularity for some t > 0 ;
2) the integer n is odd. Hence Kn > 0 so Eq. A.8 is positive definite ∀ t ≥ 0 .
Because of the arbitrariness of initial conditions, an even value of n must be ruled out and the
smallest odd integer for which the solution exists is n = 3.

285