1: \documentstyle[aps,pre,manuscript]{revtex}

2: 

3: \begin{document}

4: 

5: \draft

6: 

7: \title{Correlated noise induced control of prey extinction}

8: \author{Suman Kumar Banik {\footnote {e-mail: pcskb@mahendra.iacs.res.in} }

9: }

10: \address{Department of Physical Chemistry \\

11: Indian Association for the Cultivation of Science \\

12: Jadavpur, Calcutta 700 032, India}

13: 

14: \date{\today}

15: 

16: \maketitle

17: 

18: \begin{abstract}

19: We study the steady state properties of a phenomenological two-state 

20: predator model in presence of correlated Gaussian white noise. 

21: Based on the corresponding Fokker-Planck equation for probability 

22: distribution function the steady state solution of the probability 

23: distribution function and its extrema have been investigated. We 

24: show for a typical value of noise correlation there is a 

25: giant loss of bistability which in turn prevents the prey population

26: from going into extinction.

27: \end{abstract}

28: 

29: 

30: \pacs{PACS number(s): 05.40.-a, 87.10.+e}

31: 

32: \section{Introduction}

33: The subject of noise-induced transition has got wide applications in

34: the field of physics, chemistry and biology \cite{hl}. In most of these 

35: theories the noise affects the dynamics through system variable, i.e., the

36: noise is multiplicative in nature. The focal theme of these investigations 

37: is the study of steady state properties of the system where the

38: fluctuations, in general, are applied from outside and are independent

39: of the system's characteristic dissipation. Such systems are generally

40: termed as open systems \cite{lw}, since they lack the principle of detailed

41: balance which ensures fluctuation-dissipation relation to hold for the

42: thermodynamically closed systems. However, it may also happen that the

43: external fluctuations instead of affecting only some system's parameters 

44: affect the system directly, i.e., they drive system dynamics multiplicatively 

45: as well as additively. Because the two noise processes owe a common origin 

46: they get correlated in the relevant timescale of the problem \cite{ft,cw}. 

47: Correlated noise processes have found applications in studying steady

48: state properties of a single mode laser \cite{szhu}, in

49: analyzing bistable kinetics \cite{wck}, in giant supression of activation

50: rate \cite{ajrm}, in producing

51: directed motion in spatially symmetric periodic potentials \cite{lh}, 

52: in studying stochastic resonance in linear systems \cite{bg}, 

53: in steady state entropy production \cite{bbr}, etc.

54: In this brief communication we investigate a simple noise-driven two-state

55: predator model \cite{hl} and show how noise correlation can dynamically 

56: prevent the prey population from extinction.

57: 

58: \section{The Model}

59: 

60: To start with we consider an environment of the prey which in absence

61: of predation grows logistically and at the same time its density in a

62: territory depends linearly on a constant source of migration. We also 

63: consider a population of predators in the given territory which lives by 

64: feeding on prey. The characteristic time scale over which the population

65: of prey and predator varies are very much different, so one can consider

66: the predator population to be constant within the generation time of prey.

67: The predators are engaged in two types of activities, viz, hunting or resting.

68: The time scale of predator's two activities are very short compared to the

69: generation time of prey, i.e., $\tau_R$, $\tau_H$ $\ll$ $\mu^{-1}$ where

70: $\tau_R$ and $\tau_H$ are the characteristic average time of resting and

71: hunting, respectively and $\mu$ is the birth rate of prey. 

72: The activity of the predator in the territory 

73: resembles the mode of action of enzymes or catalysts in a chemical reaction.

74: The enzymes or catalysts in a chemical reaction transform substrates in

75: a continuous manner without destroying themselves. The constant

76: predator population acts in a similar way by feeding on the prey. To put this 

77: ideas in a quantitative way we write the evolution equations for the

78: predator and prey \cite{hl},

79: \begin{eqnarray}

80: \label{eq1}

81: \dot{X} & = & A + \mu X \left ( 1 - \frac{X}{K} \right ) - 

82: \frac{1}{\tau_H} XY \; \; , \\

83: \label{eq2}

84: \dot{Y} & = & - \; \frac{1}{\tau_H} XY + \frac{1}{\tau_R} Z

85: \end{eqnarray}

86: 

87: \noindent

88: where $X$ is the density of prey in a given territory. The constant $A$

89: in Eq.(\ref{eq1}) is due to a constant source of prey through immigration. 

90: The second term in (\ref{eq1}) is the Fisher logistic growth term with birth rate 

91: $\mu$ and carrying capacity $K$. $Y$ and $Z$ are the numbers of predators in 

92: the hunting and resting state, respectively. $E$ is the total constant 

93: population of the predators, i.e., $E \equiv Y(t) + Z(t) =$ constant.

94: The last term in (\ref{eq1}) describes the decay rate of prey. The model

95: is hybrid in nature in the sense that it has virtue of taking into consideration

96: of the logistic growth model as well as of the predator-prey model.

97: 

98: Following Ref.\cite{hl}

99: we now consider that the predator population, $E$ is small compared to

100: prey population $X$. To study the overall dynamics within the timescale 

101: $\mu^{-1}$ we make the following transformation 

102: \begin{equation}

103: \label{eq3}

104: \tau_H = \varepsilon \tau_H^* \; \; , \; \; \tau_R = \varepsilon \tau_R^*

105: \; \; , \; \; Y = \varepsilon Y^*  \; \; {\rm and} \; \;

106: Z = \varepsilon Z^*

107: \end{equation}

108: 

109: \noindent

110: where $\varepsilon$ is a small quantity, $\tau_H^*$, $\tau_R^*$ are 

111: quantities of order $\mu^{-1}$ and $Y^*$, $Z^*$ are quantities of order $X$. 

112: Using (\ref{eq3}) in (\ref{eq1}) and (\ref{eq2}) we arrive at

113: \begin{eqnarray}

114: \label{eq4}

115: \dot{X} & = & A + \mu X \left ( 1 - \frac{X}{K} \right ) - 

116: \frac{1}{\tau_H^*} XY^* \; \; , \\

117: \label{eq5}

118: \varepsilon \dot{Y}^* & = & - \; \frac{1}{\tau_H^*} XY^* + 

119: \frac{1}{\tau_R^*} Z^* \; \; .

120: \end{eqnarray}

121: 

122: Now eliminating $Y^*$ from (\ref{eq4}) and using the limit 

123: $\varepsilon \rightarrow 0$ we arrive at the following dimensionless

124: evolution equation for prey

125: \begin{equation}

126: \label{eq6}

127: \dot{x} = \alpha + x (1-\theta x) - \beta \frac{x}{1+x}

128: \end{equation}

129: 

130: \noindent

131: where

132: \begin{equation}

133: \label{eq7}

134: x = \frac{\tau_R^*}{\tau_H^*} X \; \; , \; \; 

135: \alpha = \frac{A \tau_R^*}{ \mu \tau_H^*} \; \; , \; \;

136: \beta = \frac{E}{\mu \tau_H} \; \; {\rm and} \; \;

137: \theta = \frac{\tau_H}{\tau_R K} \; \; .

138: \end{equation}

139: 

140: \noindent

141: It is interesting to note that the third term in Eq.(\ref{eq6}) is the

142: predation term which essentially emerges from the two-state of predator

143: activities. The steady state solution of Eq.(\ref{eq6}) shows a cusp

144: type of catastrophe. The corresponding critical point 

145: ($\alpha_c, \beta_c, x_c$) is given by \cite{hl}

146: \begin{eqnarray*}

147: \alpha_c = \frac{(1-\theta)^2}{27\theta^2} \; \; , \; \;

148: \beta_c = \frac{(1+2\theta)^3}{27\theta^2} \; \; {\rm and} \; \;

149: x_c = \frac{1-\theta}{3\theta} \; \; .

150: \end{eqnarray*}

151: 

152: \noindent

153: The necessary condition to have a physically realizable critical point

154: i.e., for $\alpha_c$, $x_c$ to be positive, is $\theta < 1$. Thus the 

155: steady state curve of $x$ as a function of $\beta$ always shows a bistable

156: region for small values of $\theta$. The smallness condition may be

157: maintained by increasing the carrying capacity $K$ or by decreasing

158: the ratio $\tau_H/\tau_R$.

159: 

160: Eq.(\ref{eq6}) is the starting point of our further analysis. It may be noted

161: that $\alpha$ and $\beta$ are the two quantities which appear in the

162: prey evolution equation as a constant and a multiplicative factor,

163: respectively. Expressions for $\alpha$ and $\beta$ in (\ref{eq7}) suggest

164: that they are connected by a common parameter $\mu$, the birth rate

165: of the prey. Now if due to some environmental external disturbance the birth

166: rate of the prey fluctuates, it is likely to affect both $\alpha$ and

167: $\beta$ in the form of additive and multiplicative noises which are

168: connected through a correlation parameter. Or in other words the

169: external fluctuations affect the parameter

170: $\beta$ which fluctuates around a mean value, thus generating 

171: multiplicative noise and at the same time environmental fluctuations 

172: perturbs the dynamics directly

173: which gives rise to additive noise. As a result we have the stochastic

174: differential equation in Stratonovich prescription,

175: \begin{equation}

176: \label{eq8}

177: \dot{x} = \alpha + x (1-\theta x) - \beta \frac{x}{1+x} -

178: \frac{x}{1+x} \xi (t) + \eta (t)

179: \end{equation}

180: 

181: \noindent

182: where $\xi (t)$ and $\eta (t)$ are the stationary Gaussian white noises

183: with the following properties

184: %\begin{mathletters}

185: 

186: \begin{eqnarray}

187: \langle \xi (t) \rangle & = & \langle \eta (t) \rangle = 0  \; \; , \\

188: \langle \xi (t) \xi (t') \rangle & = & 2 \sigma \delta (t-t') \; \; ,

189: \\

190: \langle \eta (t) \eta (t') \rangle & = & 2 D \delta (t-t') \; \; {\rm and}

191: \\

192: \langle \xi (t) \eta (t') \rangle & = & \langle \eta (t) \xi (t') \rangle

193: = 2 \lambda (\sigma D)^{1/2} \delta (t-t')

194: \end{eqnarray}

195: 

196: %\end{mathletters}

197: 

198: \noindent

199: where $\lambda$ denotes the degree of correlation between noise processes

200: $\xi (t)$ and $\eta (t)$

201: with $ 0 \leq \lambda \leq 1$. Using the above mentioned noise properties

202: we write down the corresponding Fokker-Planck equation (in Stratonovich 

203: prescription) for the evolution of probability distribution function 

204: \cite{cw,wck},

205: \begin{equation}

206: \label{eq9}

207: \frac{\partial}{\partial t} P(x,t) = - \frac{\partial}{\partial x}

208: A(x,t) P(x,t) + \frac{\partial^2}{\partial x^2} B (x,t) P(x,t)

209: \end{equation}

210: 

211: \noindent

212: where

213: %\begin{mathletters}

214: 

215: 

216: \begin{equation}

217: A (x,t) = 

218: \alpha + x (1-\theta x) - \beta \frac{x}{1+x} +

219: \sigma \frac{x}{(1+x)^3}  - \lambda (\sigma D)^{1/2} \frac{1}{(1+x)^2} 

220: \end{equation}

221: and

222: \begin{equation}

223: B (x,t)  =  D + \sigma \frac{x^2}{(1+x)^2}  - 

224: 2 \lambda (\sigma D)^{1/2} \frac{x}{1+x} \; \; .

225: \end{equation}

226: 

227: %\end{mathletters}

228: 

229: \section{Steady state analysis and results}

230: 

231: Using the zero current condition at the stationary state we derive

232: the stationary probability distribution function (SPDF) with $0$ and 

233: $\infty$ as the natural boundaries,

234: \begin{equation}

235: P_s (x) = N \frac{1}{ B(x)} \exp \left [ 

236: \int^x \frac{ A(x') }{ B(x') } dx' \right ] 

237: \end{equation}

238: 

239: \noindent

240: where $N$ is the normalization constant. Using the explicit

241: forms of $A(x)$ and $B(x)$ we have the following explicit forms of

242: SPDF

243: \begin{equation}

244: P_s (x) = N (1+x) g^{\nu-\frac{1}{2}} (x)

245: \exp [ q_1 x^3 + q_2 x^2 + q_3 x + q_4 f(x) ]

246: \end{equation}

247: 

248: \noindent

249: where

250: %\begin{mathletters}

251: \begin{equation}

252: g(x) = a + bx + cx^2

253: \end{equation}

254: 

255: \begin{equation}

256: \label{eq10}

257: \begin{array}{ccccc}

258: f (x) & = & -2/(b+2 c x) &  {\rm for} & \lambda = 1 \\

259:       & = & (2/ \sqrt{\Delta} ) \arctan [ (b+2cx)/ \sqrt{\Delta} ]

260:       &  {\rm for} &  0 \leq \lambda < 1

261: \end{array}

262: \end{equation}

263: 

264: \noindent

265: with

266: \begin{eqnarray}

267: \label{eq11}

268: a & = &  D \; , \;  b = 2 [ D-\lambda (\sigma D)^{1/2} ] \; ,

269: \nonumber \\

270: c & = & D + \sigma - 2 \lambda (\sigma D)^{1/2} \; , 

271: \Delta = 4 \sigma D ( 1 - \lambda^2 )

272: \end{eqnarray}

273: 

274: \noindent

275: along with

276: \begin{eqnarray}

277: q_1 & = & \frac{-\theta}{3c} \; , \nonumber \\

278: q_2 & = & \frac{1-2\theta}{2c} + \frac{b \theta}{2 c^2} \; , \;

279: q_3 =  \frac{\alpha -\beta - \theta +2}{c} - 

280: \frac{b (1-2\theta) + a \theta}{c^2} - \frac{b^2 \theta }{c^3} \; , 

281: \nonumber \\

282: q_4 & = & \alpha - \frac{ b (2\alpha-\beta+1)}{2c} +

283: \frac{(b^2-2ac)(\alpha-\beta-\theta+2)}{2c^2} - 

284: \frac{ b^2(b^2-3ac) \theta}{2c^4} \nonumber \\

285: & &  + \frac{ a(b^2-2ac) \theta - b (b^2-3ac) (1-2\theta)}{2c^3}  \; \;

286: {\rm and} \nonumber \\

287: \nu & = & \frac{2\alpha-\beta+1}{2c} -

288: \frac{ b ( \alpha -\beta \theta +2)}{2c^2} +

289: \frac{ (b^2-ac)(1-2\theta) - ab\theta}{2c^3} \nonumber \\

290: & & + \frac{b (b^2-ac) \theta}{2c^4} \; \; .

291: \end{eqnarray}

292: 

293: %\end{mathletters}

294: 

295: \noindent

296: The extrema of SPDF is calculated using the condition $A(x)-B'(x)=0$,

297: \begin{equation}

298: \label{eq12}

299: \alpha + x (1-\theta x) - \frac{\beta x}{1+x} - 

300: \frac{\sigma x}{(1+x)^3} + \frac{\lambda (\sigma D)^{1/2} }{ (1+x)^2} = 0

301: \; \;  {\rm for} \; 0 \leq \lambda \leq 1 \; \; .

302: \end{equation}

303: 

304: \noindent

305: For zero noise correlation, i.e., for $\lambda=0$ the last term of 

306: Eq.(\ref{eq12}) vanishes and we have the extrema of SPDF for pure 

307: multiplicative noise processes \cite{hl}. For zero correlation the additive 

308: noise has no extra effects in the steady state dynamics. To illustrate 

309: this we have plotted extrema of SPDF as a function of $\beta$ in 

310: Fig.(1) using the parameters given in \cite{hl}. For zero noise correlation 

311: the curve shows a sharp minima which decreases on increasing $\lambda$. 

312: Similarly, in Fig.(2) we have plotted extrema of SPDF as a 

313: function of $\beta$ for different values of additive noise strength $D$ 

314: with maximum correlation ($\lambda=1$). As the additive noise strength 

315: increases the well gets flattened and almost vanishes for a large enough 

316: value of $D$.

317: 

318: In Fig.(3) we show the effect of correlation parameter $\lambda$ on SPDF. 

319: For a low value of $\lambda$ the SPDF shows the typical bistable region 

320: ( see Fig.3(a) ) which vanishes for higher values of $\lambda$ ( see 

321: Fig.3(b) ). As the value of correlation parameter 

322: $\lambda$ increases the peak on the lower values of $x$ decreases while

323: for a higher value of $\lambda$ we have a single peak at a higher

324: values of $x$. Since $x$ denotes the prey population, it is clear from

325: Fig.(3) that with the increase of $\lambda$ values the prey population 

326: recovers from going into extinction. In other words, the distribution

327: of prey which was mainly peaked about zero (for a low value of $\lambda$)

328: signifying high extinction rate, moves away from zero with the increase of 

329: correlation between noises thus favouring the prey's survival.

330: Though Gaussian white noise acting independently and

331: multiplicatively favours the extinction of prey \cite{hl}, the

332: extinction rate decreases drastically for a simultaneous perturbation

333: of additive and multiplicative white noise originating from a common source,

334: hence connected through a correlation parameter.

335: 

336: From the expressions of $f(x)$, $b$ and $c$ given in Eqs.(\ref{eq10}) and

337: (\ref{eq11}) it is clear that for $\lambda$ = 1.0 we have always a singular

338: distribution for $\sigma=D$, since it makes both the parameter $b$ and $c$

339: zero and eventually leads to the divergence of all the  $q$'s and $\nu$.

340: However this divergence can be removed for appreciable difference between

341: the $\sigma$ and $D$ values. In Fig.(4) we have plotted the typical behaviour 

342: of SPDF for maximum correlation $\lambda$ = 1 which shows monotonic

343: decreasing behaviour. In contrast to the behaviour shown in Fig.(3), Fig.(4),

344: however, shows the hastening of prey's extinction for a full correlation

345: between additive and multiplicative noises.

346: 

347: In this brief communication we have studied the effect of environmental 

348: fluctuation of the birth rate of the prey in terms of external correlated 

349: noise processes which appreciably modify the macroscopic behaviour of a 

350: two-state predator model. We have shown how the correlation between the two 

351: noise processes which owe a common origin may drastically prevent the

352: extinction of the prey.

353: 

354: I express my sincerest gratitude to Prof. D S Ray for suggesting me the 

355: problem and for his continuous inspiration during the progress of this work.

356: This work was supported by Council of Scientific and Industrial Research

357: (C.S.I.R.), Govt. of India.

358: 

359: 

360: \begin{thebibliography}{99}

361: 

362: \bibitem{hl} W. Horsthemke and R. Lefever, {\it Noise-Induced Transitions}

363: (Springer-Verlag, Berlin, 1984).

364: 

365: \bibitem{lw} K. Lindenberg and B. West, {\it The Nonequilibrium Statistical

366: Mechanics of Open and Closed Systems} (VCH, New York, 1990).

367: 

368: \bibitem{ft} A. Fulinski and T. Telejko, Phys. Lett. A {\bf 152}, 19 (1991).

369: 

370: \bibitem{cw} Li Cao and Da-jin Wu, Phys. Lett. A {\bf 185}, 59 (1994).

371: 

372: \bibitem{szhu} S. Zhu, Phys. Rev. A {\bf 47}, 2405 (1993).

373: 

374: \bibitem{wck} Wu Da-jin, Cao Li and Ke Sheng-zhi, Phys. Rev. E {\bf 50},

375: 2496 (1994); Ya Jia and Jia-rong Li, Phys. Rev. E {\bf 53}, 5786 (1996).

376: 

377: \bibitem{ajrm} A. J. R. Madureira, P. H\"anggi and H. S. Wio, Phys. Lett. A

378: {\bf 217}, 248 (1996).

379: 

380: \bibitem{lh} J. H. Li and Z. Q. Huang, Phys. Rev. E {\bf 53}, 3315 (1996);

381: {\it ibid} {\bf 57}, 3917 (1998).

382: 

383: \bibitem{bg} V. Berdichevsky and M. Gitterman, Phys. Rev. E {\bf 60},

384: 1494 (1999)

385: 

386: \bibitem{bbr} B. C. Bag, S. K. Banik and D. S. Ray, Phys. Rev. E {\bf 64},

387: 026110 (2001).

388: 

389: 

390: \end{thebibliography}

391: 

392: \begin{figure}[h]

393: %\begin{center}

394: %\includegraphics*[width=6cm]{cpe1.eps}

395: %\end{center}

396: \caption{ Plot of extrema of SPDF

397: as a function of $\beta$ for different values of noise

398: correlation $\lambda$ using $\alpha$ = 4.5, $\theta$ = 0.1, $\sigma$ = 33.0

399: and $D$ = 3.0.

400: }

401: \label{fig1}

402: \end{figure}

403: 

404: \begin{figure}[h]

405: %\begin{center}

406: %\includegraphics*[width=6cm]{cpe2.eps}

407: %\end{center}

408: \caption{ Same as in Fig.(1) but for different values of additive noise

409: strength $D$. The other parameters are same except $\lambda$ = 1.

410: }

411: \end{figure}

412: 

413: \begin{figure}[h]

414: %\begin{center}

415: %\includegraphics*[width=6cm]{cpe3.eps}

416: %\end{center}

417: \caption{ Plot of $P_s(x)$ against $x$ for different values of noise

418: correlation $\lambda$ using $\alpha$ = 4.5, $\theta$ = 0.1, $\beta$ = 7.5,

419: $\sigma$ = 3.0 and $D$ = 0.3.

420: (a) For low values of $\lambda$ and (b) for high values of $\lambda$.

421: }

422: \end{figure}

423: 

424: \begin{figure}[h]

425: %\begin{center}

426: %\includegraphics*[width=6cm]{cpe4.eps}

427: %\end{center}

428: \caption{ Same as in Fig.(3) but for $\lambda$ = 1.0 and $D$ = 2.12.

429: }

430: \end{figure}

431: 

432: 

433: \end{document}

434: