1: \section{Analysis}
2: Our goals in this section are to find the conditions under which two prey
3: types can coexist, to determine when coexistence is steady-state versus oscillatory, and
4: to characterize the period of cycles and the phase relations during
5: oscillatory coexistence and during transients when one type is decreasing to extinction.
6: Throughout this section we consider the reduced model \eqref{eqs6}.
7: For local stability analysis it is useful to note that the model has
8: the form
9: \be
10: \dot x_i = x_i r_i(x_1,x_2,x_3) \quad i=1,2,3
11: \label{GeneralForm}
12: \ee
13: with $x_3=y$. It follows that at any equilibrium where the $x_i$ are all positive (and
14: hence the $r_i$ are all 0) the Jacobian matrix $J$ has entries
15: \be
16: J(i,j)=\tilde x_i \frac{\partial \tilde r_i}{\partial x_j}
17: \label{GeneralJacobian}
18: \ee
19: with the tilde indicating evaluation at the equilibrium with all $x_i$ present. It
20: is also useful for local stability analysis that the determinant of \eqref{GeneralJacobian}
21: is always negative unless $p_1=p_2$ (Appendix \ref{JacobianAppendix}).
22:
23: \subsection{Dynamics of a one-prey system}
24: We need first some properties of the one-prey model
25: \be
26: \begin{aligned}
27: \dot x & = x\left[\frac{m (1 - x - y)}{k + (1-x-y)}-\frac{g p y}{(k_b + px)}-1 \right]\\
28: \dot y & = y \left[\frac{gpx}{k_b + px} - 1 \right].
29: \end{aligned}
30: \label{eq7}
31: \ee
32: This is a standard predator-prey chemostat model and its behavior is well-known,
33: so we summarize here only the results that we will need later;
34: see e.g. \cite{SmithWaltman1995} for derivations and details.
35:
36: In the absence of predators,
37: the steady state for this system is
38: $
39: E_0= (1-\Lambda,0),
40: $
41: where
42: \be
43: \Lambda= \frac{k}{m-1}.
44: \label{Lambda}
45: \ee
46: $\Lambda$ is the scaled concentration of limiting nutrient at
47: which prey growth balances washout rate, so that $\dot x = 0.$
48: Similarly, steady state densities for each prey type in a predator-free two clone system are
49: $$
50: E_i= (1-\Lambda_i,0) \quad \textrm{where}\quad \Lambda_i=\frac{k_i}{m-1}.
51: $$
52: The steady state for the prey in the presence of the predator is
53: \be
54: \label{xc}
55: \bar x_c = \frac{k_b}{p(g-1)};
56: \ee
57: $\bar x_c$ is the prey density at which the predator birth and death rates are equal.
58: The model \eqref{eq7} has an interior equilibrium point $E_c=(\bar x_c, \bar y_c)$ representing
59: predator-prey coexistence if
60: \be
61: \Lambda + x_c < 1
62: \label{coexistence2}
63: \ee
64: \cite{SmithWaltman1995}, and otherwise the predator cannot persist.
65: The system then collapses to the prey by itself
66: and converges to $E_0$. Condition \eqref{coexistence2} says that
67: there is an interior equilibrium if the prey by themselves reach a steady state
68: ($1-\Lambda$) that provides enough food so that the predator birth rate exceeds
69: the predator death rate.
70:
71: The expression for the steady state of the predator, $y_c$, is easily obtained from \eqref{eq7}:
72: \be
73: \label{predatorSS}
74: \bar y_c=\bar \sigma -\bar x_c,\quad \mbox{where} \quad \bar\sigma =
75: (\bar x_c + \bar y_c) = \frac{1}{2}\left[\gamma - \sqrt{(\gamma^2 - 4m\bar x_c)}\right],
76: \ee
77: with $\gamma=k + 1 + m\bar x_c$. Similarly, the steady-state densities for the predator in a single-prey system with either prey type, $\bar y_i$, are found by substituting the steady state for the prey, $\bar x_i$,
78: in place of $\bar x_c$ and the appropriate half-saturation
79: $k_i$ in place of $k$ in \eqref{predatorSS}.
80:
81: %\be
82: %\label{predatorSS}
83: %y_i = \bar \sigma_i -\bar x_i,\quad \mbox{where} \quad \bar\sigma_i =
84: %(\bar x_i + \bar y_i) = \frac{1}{2}\left[\gamma_i - \sqrt{(\gamma_i^2 - 4m\bar x_i)}\right]
85: %\ee
86: %with $\gamma_i=k_i + 1 + m\bar x_i$, and $x_i = \frac{k_b}{p_i(g-1)}$.
87: We can use (\ref{coexistence2}) to derive the condition for predator-prey coexistence
88: in terms of the prey defense trait $p$ and the dilution rate $\delta$, recalling that
89: $\Lambda$ and $x_c$ are both implicit functions of $\delta$. Using
90: \eqref{Lambda} and \eqref{xc} we obtain from \eqref{coexistence2}
91: \be
92: \frac{k}{m(\delta)-1} + \frac{k_b}{pg(\delta)-1} < 1, \quad \textrm{or} \quad p >
93: \frac{1}{1-\Lambda}\left(\frac{k_b}{g(\delta)-1}\right).
94: \label{coexistence3}
95: \ee
96: \noindent
97: The quantity within parenthesis above is the amount of substrate present in perfect food (undefended prey with $p=1$).
98: Solving \eqref{coexistence3} for $\delta$ in terms of $p$ yields the boundary between predator extinction and stable coexistence in Figure \ref{SingleCloneBifurcations}. To the left of this line, the predator goes extinct and the equilibrium is $E_0$. As the left-hand side of the second
99: expression in \eqref{coexistence3} is an increasing function of $p$,
100: and the right-hand side is an increasing function of $\delta$, the range of $p$ values yielding coexistence
101: narrows as $\delta$ increases (see Figure \ref{SingleCloneBifurcations}).
102:
103: As in the standard Rosenzweig-MacArthur predator-prey model, the stability condition has
104: a graphical interpretation in terms of the nullclines.
105: The prey nullcline is a parabola which peaks at
106: $$ x^* = \frac{1}{m}\left[ 1-k + \sqrt{\Lambda}(m-2)\right].$$
107: The coexistence equilibrium is locally unstable if the peak of the prey nullcline is to the
108: right of the predator nullcline (i.e., if $x^* > \bar x_c $). Note that a system with defended prey ($p < 1$)
109: is always more stable than a system with fully vulnerable prey ($p=1$)
110: as reductions in $p$ shift the predator nullcline to the right.
111:
112: From \eqref{GeneralJacobian} the Jacobian of \eqref{eq7} at $E_c$ has the form
113: \be
114: J_c = \begin{bmatrix}
115: \bar x_c \frac{\partial \bar r_1}{\partial x} & - \\
116: + & 0 \\
117: \end{bmatrix}
118: \label{Jc}
119: \ee
120: so $E_c$ is locally stable if the trace
121: {\bf Tr}$(J_c)= \bar x_c \frac{\partial \bar r_1}{\partial x}$ is negative. Cycles emerge through a Hopf bifurcation
122: when the trace becomes positive. The condition {\bf Tr}$(J_c) \ge 0$ is equivalent to the following
123: expression for model \eqref{eq7}:
124: \be
125: \frac{m k}{(k + 1 - \bar x_c- \bar y_c)^2} \ge \frac{g p^2 \bar y_c}{(k_b + p \bar x_c)^2}
126: \label{OneCloneStability}
127: \ee
128: \cite{SmithWaltman1995}. Cycles begin when the rates of change
129: in prey substrate uptake (LHS) and in predator consumption (RHS)
130: with respect to the amount of substrate present as prey ($x$) are exactly equal.
131: Numerically solving \eqref{OneCloneStability} for $\delta$ in terms of $p$ yields the boundary
132: between stable coexistence and predator-prey cycles in Figure \ref{SingleCloneBifurcations}.
133: It is known that these cycles are stable and unique near the Hopf bifurcation, and numerical evidence uniformly
134: indicates that they are always unique and attract all interior initial conditions except $E_c$ \cite{SmithWaltman1995}.
135:
136: \begin{figure}[t]
137: \centerline{\includegraphics[width=3.5in]{SingleClonalBifurcations.eps}}
138: \caption{Bifurcation diagram for the rescaled, reduced clonal model with one prey type.}.
139: \label{SingleCloneBifurcations}
140: \end{figure}
141:
142: \subsection{Stability and dynamics of a two-prey system}
143: System \eqref{eqs6} has two prey types ordered so that
144: $0 < p_1 < p_2 = 1$. We refer to prey 1 as the {\it defended} type and prey 2 as the {\it vulnerable} type.
145: The cost for defense comes in the form of reduced growth rate, $1/k_1 \le 1/k_2$.
146:
147: Following Abrams \cite{Abrams1999}, we begin by finding the conditions for existence of an
148: equilibrium $(\tilde{x}_1, \tilde{x}_2, \tilde y)$ at which all three population densities are positive;
149: we refer to this as a \textit{coexistence equilibrium}.
150: Setting \eqref{eqs6} to zero and solving gives expressions for $\tilde X, \tilde Q$ and $\tilde y$ in
151: terms of model parameters (see Appendix \ref{SSAppendix}; as above $Q = p_1 x_1 + p_2 x_2$ is
152: the total prey quality, and $X=x_1 + x_2$ is the total prey density).
153: The prey steady states $\tilde x_1, \tilde x_2$ are then
154: \be
155: \left[ {\begin{array}{*{20}c}
156: \tilde x_1 \\
157: \tilde x_2 \\
158: \end{array} } \right] = \frac{1}{p_2 - p_1 }
159: \left[ {\begin{array}{*{20}c}
160: p_2 \tilde X - \tilde Q \\
161: \tilde Q - p_1 \tilde X \\
162: \end{array} } \right].
163: \label{xbars}
164: \ee
165: where
166: \be
167: \tilde Q = \frac{k_b}{g-1}.
168: \label{Qbar}
169: \ee
170: A coexistence equilibrium thus exists provided $\tilde X >0 $ and $p_1 \tilde X < \tilde Q < p_2 \tilde X$,
171: or
172: \be
173: \label{InvasionCondition}
174: p_1 < \frac{\tilde Q}{\tilde X} < p_2.
175: \ee
176:
177: Beyond the above, system \eqref{eqs6} does not yield tidy
178: analytical solutions for the steady states at coexistence. To study how parameter
179: variation affects coexistence, we start by graphically mapping the region where a
180: coexistence equilibrium exists as a function of the defended clone's parameters, $p_1$ and $k_1$
181: (Figure \ref{CCE}), without regard to whether or not the equilibrium
182: is locally stable. The coexistence region also varies with
183: $\delta$, but selecting several $\delta$ values of interest gives
184: a general sense of how the coexistence region varies as a function of dilution rate.
185:
186: The lower boundary of the coexistence region occurs when the cost of
187: defense is so high that the equilibrium density of the defended prey $x_1$ drops
188: to zero while $x_2$ and $y$ remain positive. Recalling the general form
189: \eqref{GeneralForm}, the lower boundary is thus defined by the conditions
190: $$
191: r_1(0,x_2,y)=r_2(0,x_2,y)=r_3(0,x_2,y)=0 \mbox{ with } x_2>0, y>0.$$
192: The conditions on $r_2$ and $r_3$ are solved by the steady state $E_2 = (0, \bar x_2, \bar y_2)$
193: for a one-prey system with only the vulnerable prey.
194: The lower boundary of the coexistence region is thus defined by the condition
195: $r_1(0,\bar{x}_2,\bar{y}_2)=0$, which can be written as
196: \be
197: \label{Lowercondition}
198: \frac{1}{k_1}=\frac{1}{1 -\bar x_2 - \bar y_2}\left[\frac{\bar y_2p_1 + \bar x_2}{\bar x_2(m-1)- \bar y_2 p_1}\right].
199: \ee
200:
201: \begin{figure}[t]
202: \centerline{\includegraphics[width=5.5in]{Figure3AB.eps}}
203: \caption{\small{Prey coexistence equilibria. The shaded gray regions indicate parameters $(p_1,1/k_1)$ for prey
204: type 1 giving a coexistence equilibrium (stable or unstable) with the vulnerable prey type $p_2=1$. At $\delta=1.5$ (panel A) prey type 2 cycles, and at $\delta$ = 2.0 (panel B), prey type 2 is stable. The dashed lines
205: show a representative tradeoff curve \eqref{tradeoff}, assuming roughly 50\% reduction in growth rate as the cost of being 100\% defended. Here $k_c=0.017$, $a_1=1.0$, and $a_2=0.0165$.
206: In panel A the dash-dotted line indicates where the defended type can invade the limit cycle of the predator and vulnerable prey type.}}
207: \label{CCE}
208: \end{figure}
209:
210: The upper boundary of the coexistence region occurs when the cost of defense is so low
211: that the defended prey (at the equilibrium density) drives one of the other populations to extinction.
212: In section \ref{GeneralEvoCycles} we show that for $p_1 < p^*=\tilde Q/(1-\Lambda_2)$,
213: the predator goes extinct first ($\tilde y \to 0$) as $k_1$ decreases, because the defended prey (at steady state)
214: drives the vulnerable prey to low abundance and the defended prey is very poor food. This occurs
215: at $k_1 = k_2$ (zero cost of defense). For $p_1 > p^*$, the vulnerable prey type
216: is outcompeted by the defended type before $k_1$ has reached $k_2$.
217: This boundary is therefore defined by the conditions
218: $$
219: r_1(x_1,0,y)=r_2(x_1,0,y)=r_3(x_1,0,y)=0 \mbox{ with } x_1>0,y>0.
220: $$
221: \noindent
222: The conditions on $r_1$ and $r_3$ are solved by the one-prey steady state $E_1 = (\bar x_1, 0, \bar y_1)$, so
223: the condition $r_2(\bar{x_1},0,\bar{y_1})=0$ defines the upper boundary of the coexistence region
224: for $p>p^*$. The upper boundary of the coexistence region is thus the curve
225: \be
226: \frac{1}{k_1}= \textrm{min}\left[\frac{1}{k_2}, \frac{1}{\varphi(\bar x_1,\bar y_1)} \right]
227: \label{UL}
228: \ee
229: \noindent
230: where $\varphi$ is value of $k_1$ that solves
231: \be
232: \label{UpperCondition}
233: \frac{m (1 - \bar x_1 - \bar y_1)}{k_2 + (1-\bar x_1-\bar y_1)}-\frac{\bar y_1 + p_1 \bar x_1}{(p_1 \bar x_1)} = 0,
234: \ee
235: noting that $\bar x_1$ and $\bar y_1$ are functions of $k_1$ and $p_1$.
236: The two segments of the upper boundary defined by \eqref{UL} meet at the point
237: $$ p_1=p^*= \frac{\bar Q}{1-\Lambda_2}, \quad k_1=k_2.$$ As $\delta \to 0$ (with the
238: parameter scalings in Table 2), $\bar Q \downarrow 0$ and $\Lambda_2 \uparrow 1$, so $p^* \uparrow 0$
239: so there is typically an increasingly narrow band of $p_1$ values for which a $p_1 -- k_1$ tradeoff curve
240: lies in the coexistence equilibrium region (unless the tradeoff curve happens to lie exactly inside
241: the cusp of the coexistence equilibrium region).
242:
243: As $p_1 \rightarrow 1$, the upper and lower boundaries of the coexistence region
244: meet at $p_1=p_2=1, k_1=k_2$ (Figure \ref{CCE}). That is, if the two prey are almost equally vulnerable
245: to predation, they can only coexist at equilibrium if a tiny bit of defense has a tiny cost.
246: To prove that this occurs, we show that the point $p_1=1,k_1=k_2$ lies on both boundaries.
247: At this point the two prey are identical so $\bar x_1 = \bar{x_2}$ and $\bar y_1 = \bar y_2$.
248: The upper boundary is defined by $r_2(\bar x_1,0,\bar y_1)=0$.
249: At $p_1=1, k_1=k_2$,
250: $$
251: r_2(\bar x_1,0,\bar y_1)=r_2(\bar x_2,0,\bar y_2)=r_2(0,\bar x_2,\bar y_2)=0,
252: $$
253: thus $p_1=1, k_1=k_2$ lies on the upper boundary.
254: The lower boundary is defined by $r_1(0,\bar x_2,\bar y_2)=0$.
255: At $p_1=1, k_1=k_2)$,
256: $$
257: r_1(0,\bar x_2,\bar y_2)=r_1(0,\bar x_1 \bar y_1)=r_2(\bar x_1,0,\bar y_1)=0,
258: $$
259: which shows that $p_1=1, k_1=k_2$ lies on the lower boundary.
260: Thus both boundaries converge to $k_1=k_2$ as $p_1 \to 1$.
261:
262: \subsection{Local stability of coexistence equilibria}
263: To characterize two-prey evolutionary cycles we need to find the bifurcation
264: curves in parameter space where these cycles arise.
265: The ``empirical facts'' are summarized in Figure \ref{CCEStability}, based on
266: numerical evaluations of the Jacobian and its eigenvalues within the coexistence equilibrium region.
267: In Figure \ref{CCEStability} we change the stability of the {(predator + vulnerable prey)} system by varying
268: the value of $\delta$, but the results are qualitatively the same if other parameters are varied instead
269: (e.g., varying the prey maximum growth rates).
270:
271: \begin{figure}[t]
272: \centerline{\includegraphics[width=6in,height=6in]{ClonalContourCoexist.eps}}
273: \caption{\small{Stability of coexistence equilibria for the reduced two-prey model.
274: In each panel, the horizontal axis is the palatability $p_1$ of the defended prey with the model
275: scaled so that $p_2=1$. To remain consistent with Abrams ({\it cf.} \cite{Abrams1999}, Figures 1 - 3) the vertical axis is $1/k_1$, scaled so that 0 and 1 correspond to the lower and upper limits of the coexistence equilibrium region (Figure \ref{CCE}).
276: The Jacobian matrix and its eigenvalues were
277: evaluated at an even $50 \times 100$ grid of values. Lighter gray indicates that the equilibrium is stable, darker gray that it is unstable; in all cases the computed eigenvalues with largest real part are a complex conjugate pair.
278: The vertical black line is at $p_1=p^*$, the value where the straight and curved segments of the upper limit of the
279: coexistence equilibrium region meet. The dashed curve in panels A and B is the tradeoff curve
280: $k_1=k_c + \alpha_2(1-p_1)^{\alpha_1}$, with $k_c=k_2=0.054,
281: \alpha_1=1; \alpha_2=0.05$ at $p_1$ values lying within the coexistence equilibrium region.
282: The dash-dotted line represents the
283: minimum $1/k_1$ values at which the defended prey can invade the (predator + vulnerable prey) limit cycle
284: (see Appendix \ref{InvadeEdibleCycle}). Parameter values for these plots are as follows: panel A., $\delta=1.5$; B., $\delta=1.75$; C., $\delta=2.0$ and D., $\delta=2.1$. All other parameters are unchanged and are as shown in Table 2.
285: }}
286: \label{CCEStability}
287: \end{figure}
288:
289: The stability properties in Figure \ref{CCEStability} explain the major qualitative
290: features of the two-prey model's bifurcation diagram (Figure \ref{DynamicsFigs}D). To see the
291: connection, recall that a horizontal (constant $\delta$) slice through Fig. \ref{DynamicsFigs}D
292: corresponds to a tradeoff curve between $p_1$ and $k_1$ in the panel of Figure \ref{CCEStability}
293: with the same value of $\delta$. Panel A of Figure \ref{CCEStability} has $\delta=1.5$.
294: When $p_1$ is near 1, the tradeoff curve lies above the coexistence equilibrium region,
295: and the defended prey type eventually outcompetes the vulnerable type. For $p_1$ very close to 1
296: the prey types are very similar, and the vulnerable type persists for a long time.
297: The system exhibits ``classical'' predator-prey cycles as if a single prey-type were present, even though two types are transiently coexisting.
298: For $p_1$ somewhat smaller, the vulnerable type is quickly eliminated and there are either
299: classical cycles with only the defended type (open circles in Fig. \ref{DynamicsFigs}D),
300: or (for lower values of $p_1$) the defended prey type goes to a stable equilibrium with the
301: predator (open triangles in Fig. \ref{DynamicsFigs}D).
302: As $p_1$ decreases further, Figure \ref{CCEStability}A shows that the tradeoff curve
303: enters the coexistence region in the area where the coexistence equilibrium is stable, so
304: the system then exhibits stable coexistence (cross-hatching in Fig. \ref{DynamicsFigs}D). Finally,
305: as $p_1$ decreases towards 0, the tradeoff curve enters the area where the coexistence equilibrium
306: is unstable, and it lies above the dash-dot curve marking the $k_1$ value required for the defended prey type to invade the
307: vulnerable prey's limit cycle with the predator. The system
308: exhibits evolutionary cycles with both prey types persisting (filled circles at $p \approx 0$ in Fig. \ref{DynamicsFigs}D).
309:
310: Figure \ref{CCEStability}A also shows that there is a region of parameters (below the dash-dot curve) where the coexistence equilibrium is stable and the system therefore has coexisting attractors: a locally
311: stable coexistence equilibrium, and a locally stable limit cycle with the predator and the vulnerable prey.
312:
313: Figure \ref{CCEStability}B, which has $\delta=1.75$, shows the same sequence of transitions
314: as Figure \ref{CCEStability}A, but each occurs at higher values of $p_1$, reflecting the stabilizing effect of increased
315: washout. This is reflected in Figure \ref{DynamicsFigs}D: increasing $\delta$ above 1.0 shifts
316: all the bifurcation boundaries to higher $p$ values, but the sequence of bifurcations as
317: $p$ decreases is unchanged. However for $\delta$ sufficiently high (panels C and D in Figure \ref{CCEStability}),
318: the tradeoff curve lies either below the coexistence equilibrium region or within the region where
319: the coexistence equilibrium is stable, so evolutionary cycles never occur. Instead, there is either
320: stable coexistence of the two prey with the predator, or classical predator-prey cycles with only
321: the vulnerable prey type.
322:
323: Evolutionary cycles are also eliminated -- but for a different reason --
324: as $\delta \downarrow 0$ in Figure \ref{DynamicsFigs}D. As noted above, as $\delta \downarrow 0$ we also
325: have $p^* \downarrow 0$, so unless $p_1 \approx 0$ the tradeoff curve lies above
326: the coexistence equilibrium region and only the defended prey persists with the predator,
327: cycling at higher $p_1$ and stable at lower $p_1$. Only very near $p_1 =0$,
328: a region tiny enough to be missed by our simulation grid in Figure \ref{DynamicsFigs}, can
329: there be coexistence of both prey with the predator.
330:
331: \paragraph{Stability on the edges.} We can gain some understanding of the patterns in Figure \ref{CCEStability}, and
332: see that they are not specific to the parameter values used to draw the Figure, by examining the limiting cases
333: that occur along the edges of the coexistence equilibrium region. One general
334: conclusion (explained below) is that the bottom and right edges, and the right-hand portion
335: of the top edge, all must have the same stability as the reduced system with the predator
336: and only the vulnerable prey (prey type 2). However even if this system is unstable, there
337: must be a region along the top edge where the coexistence equilibrium is stable.
338:
339: The Jacobian matrix that determines equilibrium stability is derived in
340: Appendix \ref{JacobianAppendix}. We also show there that the determinant of this Jacobian is always negative
341: at a coexistence equilibrium unless $p_1=p_2$, so the coefficient $c_0 = -\det(J)$ in the Routh-Hurwitz stability
342: criterion for 3-dimensional systems is always positive.
343:
344: \paragraph{Bottom and right edges:} Near the bottom and right edges, the coexistence equilibrium
345: has the same local stability as the {(predator + vulnerable prey)} subsystem (panels A and B versus
346: C and D in Fig \ref{CCEStability}). The bottom edge is the lower limit of the
347: coexistence equilibrium region, where $\tilde x_1 \to 0.$ The coefficients for the Routh-Hurwitz stability criterion (see Appendix \ref{HopfConditions}) are then
348: \be
349: c_0 =-\det(J) > 0, \quad c_1 = T_2(J) \to \delta_2, \quad c_2 = -T(J) \to -\tau_2
350: \ee
351: where $\delta_2$ and $\tau_2$ are the determinant and trace, respectively, of the
352: $2 \times 2$ Jacobian for the (predator + vulnerable prey) system. If this
353: one-prey system is stable then $\delta_2 > 0, \tau_2 <0$ so $c_0, c_1$ and $c_2$
354: are all positive. Moreover $c_0=\emph{O}(\tilde x_1)$ (see Appendix \ref{JacobianAppendix}),
355: so when $\tilde x_1$ is small we have $c_1 c_2 > c_0$ and the equilibrium is stable. Conversely if the steady state for the {(predator + vulnerable prey)} system is
356: unstable, $c_2$ is negative so the full system is also unstable.
357:
358: The right edge corresponds to the cusp in the coexistence region as $p_1 \to 1$.
359: Near the cusp the two prey become increasingly similar ($p_1 \approx p_2=1, k_1 \approx k_2$).
360: Using \eqref{GeneralJacobian}, the functional forms of the $r_i$ and the fact that
361: $p_1 \approx p_2$ then imply that the form of $J$ is approximately
362: \be
363: J_0 = \left[ {\begin{array}{*{20}c}
364: {a q} & {a q} & { - qb} \\
365: {a (1 - q)} & {
366: a (1 - q)} & { - (1 - q)b} \\
367: c & c & 0 \\
368: \end{array} } \right]
369: \label{Jclose}
370: \ee
371: where $q=\tilde x_1/(\tilde x_1 + \tilde x_2)$; even if $p_1$ is near $p_2$, it is not necessarily the
372: case that $\tilde x_1$ is close to $\tilde x_2$. In \eqref{Jclose} $b$ and $c$ are positive while
373: $a$ has the sign of $\partial r_1/\partial x_1$ which may be positive or negative.
374: One eigenvalue of $J_0$ is 0, corresponding to the dynamics of $x_1-x_2$. The others are
375: $\frac{1}{2}(a \pm \sqrt{a^2 - 4bc}),$
376: which are also the eigenvalues of a single-prey system at the coexistence steady state. Thus, the two-prey system
377: with $p_1 \approx p_2=1$ ``inherits'' two eigenvalues from the one-prey system with $p = 1$.
378:
379: When the one-prey system with p=1 is cyclic, the inherited eigenvalues are a
380: complex conjugate pair. In the corresponding eigenvectors, the components for the two clones are identical when
381: $p_1 = p_2$. This implies that when $p_1 \approx p_2$ the eigenvector
382: components will be similar, so the two prey types cycle almost exactly in phase.
383: The period of these oscillations is determined by the inherited eigenvalues,
384: so it is close to the period of the corresponding one-prey system.
385:
386: When the one-prey system is stable, the Routh-Hurwitz criterion
387: (Appendix \ref{HopfConditions}), using $J_0$ to approximate trace$(J)$ and $T_2(J)$ and the
388: fact that $\textrm{det}(J) < 0$ for $p_1 \not= p_2$ , implies that the full system will also be stable.
389: Therefore, a coexistence equilibrium with two nearly identical prey
390: has the same stability as the equilibrium for the corresponding one-prey systems. During damped
391: oscillations onto a stable coexistence equilibrium, or diverging oscillations away from
392: an unstable one, the clones will oscillate nearly in phase with each other and inherit
393: the cycle period of the one-prey system.
394:
395: \paragraph{Top edge:} The rightmost portion of the top edge also corresponds to the cusp in the coexistence
396: equilibrium region, so the stability here is also the same as that of the
397: {(predator + vulnerable prey)} system. In general, as $1/k_1$ approaches the upper limit of
398: the coexistence equilibrium region when $p_1> p^*$ (the curved portion), the stability of the
399: two-prey system approaches that of the {(predator + defended prey)} system with $1/k_1$ approaching $1/k_2$.
400: This must be stable if the {(predator + vulnerable prey)} system is stable, because the defended
401: prey is always more stable, as noted above. If the {(predator + vulnerable prey)} system cycles,
402: then there will be instability as $p_1 \to 1$ along the top edge.
403:
404: However, there is always stability near the top edge
405: for $p_1 \to p^*$, as follows. Along the straight portion of the top edge ($p_1 < p^*$),
406: as $1/k_1$ approaches the edge, the coexistence equilibrium converges to a limit with $\tilde y=0$,
407: while along the curved portion the limiting coexistence equilibrium has $\tilde x_2=0$. So near their intersection at
408: $p_1=p^*$, both $\tilde x_2$ and $\tilde y$ approach 0. Condition \eqref{OneCloneStability}
409: then implies that the {(predator + defended prey)} system is stable, so the coexistence equilibrium is
410: stable near the top edge for $p_1$ just above $p^*$. By continuity, there is an open region of
411: $(p_1,k_1)$ values near $p_1=p^*,k_1=k_2$ where the coexistence equilibrium is locally stable.
412: If the {(predator + vulnerable prey)} system is only weakly unstable then this
413: stability region may be quite large (Fig \ref{CCEStability} panel B), but it cannot
414: reach either the bottom or right edges.
415:
416: \paragraph{Left edge:} Finally, consider the edge $p_1 = 0$. The steady states simplify to
417: \be
418: \tilde x_1=1- \tilde Z- \tilde x_2 - \tilde y, \quad \tilde x_2 = \tilde Q = \frac{k_b}{g-1},
419: \quad \tilde y= \tilde Q(m-1)\frac{(k_1 - k_2)}{k_1 + k_2(m-1)}
420: \label{leftedgeSS}
421: \ee
422: where $\tilde Z = \frac{k_1}{m-1}$. The coexistence equilibrium exists for $\vartheta < \frac{1}{k_1} < \frac{1}{k_2}$
423: where $\vartheta$ is the value of $1/k_1$ that solves $\tilde x_2 + \tilde y + \tilde Z=1$,
424: noting that $\tilde y$ depends on $k_1$. The Jacobian matrix at \eqref{leftedgeSS} is
425: \be
426: \begin{gathered}
427: J = \left[ {\begin{array}{*{20}c}
428: { -a_1\tilde x_1 } & { -a_1 \tilde x_1 } & {-a_1\tilde x_1 } \\
429: { -a_2\tilde x_2 } & { (-a_2 + g\tilde y F^2)\tilde x_2 } & {-(a_2 + gF)\tilde x_2 } \\
430: {0} & {gk_b\tilde y F^2} & 0 \\
431: \end{array} } \right] \\
432: \end{gathered}
433: \ee
434: where setting $p_1 = 0$ and $p_2 = 1$ gives $F=\frac{1}{k_b + \tilde x_2}$ and
435: \be
436: a_i= \frac{m k_i}{(k_i + \tilde{Z})^2}.
437: \label{defnAi}
438: \ee
439: Near the lower limit of the left edge, we know that the system inherits the stability of the
440: {(predator + vulnerable prey)} system. Above the lower limit we can use the
441: Routh-Hurwitz criterion (Appendix \ref{HopfConditions}) to determine stability.
442: The coefficients $c_0$ and $c_1$ have common factor $\tilde{x}_2gF^2 > 0$. Dividing this out
443: gives modified coefficients
444: \be
445: \tilde c_0 = a_1\tilde{x}_1gk_bF > 0, \quad \tilde c_1 = k_b(a_2 + gF) - a_1\tilde x_1,
446: \quad \tilde c_2 = a_1\tilde x_1 + \tilde{x}_2(a_2 - g\tilde{y}F^2)
447: \ee
448: and the stability conditions remain the same:
449: $
450: \tilde c_0, \mbox{ }\tilde c_1, \mbox{ }\tilde c_2 > 0, \mbox{ }\tilde c_1\tilde c_2 > \tilde c_0.
451: $
452: Extensive numerical evaluations of the coefficients as $\delta$ is varied indicate that
453: loss of stability occurs when the condition $\tilde c_1\tilde c_2 - \tilde c_0 > 0$ is violated --
454: the equilibrium is stable if this condition holds and unstable if it fails. Assuming this is true,
455: loss of stability along the left edge occurs via a Hopf bifurcation (Appendix \ref{HopfConditions}).
456:
457: \begin{figure}[t]
458: \centerline{\includegraphics[width=4.25in]{RatioDefended.eps}}
459: \caption{\small{Contour plot of the long-term average fraction of defended prey. The horizontal axis is the
460: palatability $p_1$ of the defended prey, with the model scaled so that $p_2=1$. The vertical axis represents $1/k_1$, with 0 and 1 corresponding to the lower and upper limits
461: of the coexistence equilibrium region (Figure \ref{CCE}). Numerical solutions of the
462: model were used to compute the long-term average value of $x_1/(x_1+x_2)$ for
463: parameter values such that the (predator + vulnerable prey) system (same parameter values
464: as panel A of Fig. \ref{CCEStability}). In the lighter-gray region
465: the coexistence equilibrium is stable. In the darker-gray region the equilibrium is unstable. The vertical black
466: line is at $p_1=p^*$, the value where the straight and curved segments of the upper limit of the
467: coexistence equilibrium region meet.
468: %The dash-dot line is the minimum $1/k_1$ value at
469: %which the defended prey can invade the {(predator + vulnerable prey)} limit cycle.
470: Parameter values are as in Table 2 with $\delta=1.5$.
471: }}
472: \label{ContourFractionDefended}
473: \end{figure}
474:
475: \subsection{The structure of evolutionary cycles}
476: The stability analysis above delimits the situations in which evolutionary cycles occur.
477: As illustrated in Figure \ref{CCEStability}A, they arise when the $p_1$ versus $k_1$ tradeoff
478: curve passes (with decreasing $p_1$) from the region of stable coexistence equilibria
479: near $p_1=p^*, k_1=k_2$ to the region of unstable coexistence equilibria with $p_1\approx 0$.
480: For $1/k_1$ below the dash-dot curve in Figure \ref{CCEStability}A the defended prey cannot invade
481: the vulnerable prey-predator limit cycle (see Figure \ref{ContourFractionDefended}). As $1/k_1$ increases,
482: the defended prey becomes persistent and then increases in average abundance. As $1/k_1 \to 1/k_2$ the
483: characteristic features of evolutionary cycles emerge: longer cycle period and out-of-phase oscillations
484: in predator and total prey abundance.
485:
486: The phase relations on evolutionary cycles can be seen in the dominant eigenvector of the Jacobian matrix for the unstable fixed point (Figure \ref{CoCycles}). There is a codominant pair
487: of complex conjugate eigenvalues, and ( because $\det(J)<0$ ) the third eigenvalue is real and
488: negative. As the defended prey has very low palatability, the predator and
489: the vulnerable prey have the classical quarter-phase lag. Here the phase angle is $90^{o}$; because eigenvectors are only defined up to arbitrary scalar multiples, including arbitrary
490: rotations in the complex plane from multiplication by $e^{i \theta}$, only the relative phases
491: of eigenvector components are meaningful. As $1/k_1$ increases, the eigenvector components for the two prey types become out of phase with each other ($\approx 180^{o}$ phase angle, right column of Figure \ref{CoCycles}). As a result, the predator and total prey densities are out of phase with each other.
492:
493: \begin{figure}[t]
494: \centerline{\includegraphics[width=5.5in]{CoexistenceCycles.eps}}
495: \caption{\small{Coexistence of edible and defended prey on a limit cycle.
496: Parameter values for all plots were $\delta=0.9, m=3.3/\delta, g=2.3/\delta,
497: k_2=0.05, k_b=0.2, p_2=1, p_1=0.08$. Values of $k_1$ were 0.4 (top row),
498: 0.1 (center row) and 0.055 (bottom row). In each row the leftmost panel
499: shows the dynamics of total prey and predator densities, the center panel
500: shows the dynamics of the two prey types, and the rightmost panel shows the phases
501: of the Jacobian matrix components: 1=defended prey, 2=edible prey, 3=predator,
502: 4=total prey. }}
503: \label{CoCycles}
504: \end{figure}
505:
506: In the next section we show that these phase relations become
507: exact as the limit $k_1 \to k_2$ is approached, for a general version of the model
508: in which we do not specify the functional forms of the predator and prey functional responses.
509: