1: \documentclass[doublespacing]{elsart}
2: %Uses the elsevier package
3:
4: \usepackage{graphicx}
5: \usepackage{natbib}
6: \usepackage{lineno}
7: \journal{Ecological Complexity}
8:
9: \bibliographystyle{elsart-harv}
10: %\slugcomment{Prepared for Ecological Complexity, \today}
11: %\shortauthors{Abraham}
12: %\shorttitle{Urchin feeding fronts}
13: \begin{document}
14:
15: \begin{frontmatter}
16: \title{Sea urchin feeding fronts}
17:
18: \author{Edward R. Abraham\thanksref{now}}
19: \address{National Institute of Water and Atmospheric Research (NIWA),
20: P.O. Box 14-901, Kilbirnie, Wellington, New Zealand}
21: \ead{edward@dragonfly.co.nz}
22:
23: \thanks[now]{Present address: Dragonfly, 10 Milne Terrace, Island Bay, Wellington, New Zealand (www.dragonfly.co.nz)}
24:
25:
26:
27:
28: \begin{abstract}
29: Sea urchin feeding fronts are a striking example of spatial pattern formation in an ecological system. If it is assumed that urchins are asocial, and that they move randomly, then the formation of these dense fronts is an apparent paradox. The key lies in observations that urchins move further in areas where their algal food is less plentiful. This naturally leads to the accumulation of urchins in areas with abundant algae. If urchin movement is represented as a random walk, with a step size that depends on algal concentration, then their movement may be described by a Fokker-Planck diffusion equation. For certain combinations of algal growth and urchin grazing, travelling wave solutions are obtained. Two dimensional simulations of urchin algal dynamics show that an initially uniformly distributed urchin population, grazing on an alga with a smoothly varying density, may form a propagating front separating two sharply delineated regions. On one side of the front algal density is uniformly low, and on the other side of the front algal density is uniformly high. Bounds on when stable fronts will form are obtained in terms of urchin density and grazing, and algal growth.
30: \end{abstract}
31: \end{frontmatter}
32:
33: \linenumbers
34: \modulolinenumbers[2]
35:
36:
37: \section{Introduction}
38: \label{sec:intro}
39:
40:
41:
42: Dense, linear aggregations of sea-urchins are sometimes seen. These
43: features, known as feeding-fronts, generally occur at the boundary
44: between grazed and ungrazed habitat \citep{Dean84, Scheibling99, Alcoverro02, Gagnon04}. The fronts propagate slowly
45: towards the ungrazed region. Because of the high urchin densities,
46: they are often destructive. A striking example was an
47: aggregation of the urchin \emph{Lytechinus variegatus},
48: observed invading sea-grass habitat in Florida Bay \citep{Macia99}.
49: The aggregation was estimated to be 2 - 3 m wide and 4 km long, with a
50: density of order 100 urchins m$^{-2}$. It is reported to have moved at a rate of up to 6 m
51: day$^{-1}$, reducing above-ground seagrass to less than 2\% of its
52: initial biomass. Although it became more diffuse with time, the front
53: remained as a coherent feature for at least 10 months. Similar features have been seen in other benthic invertebrates. Linear
54: aggregations of starfish have been recorded invading extensive mussel
55: beds \citep{Dare82}, and traveling fronts of strombid conch have also
56: been observed in the Caribbean \citep{Stoner89, Stoner94} and in
57: Australia (A. MacDiarmid, pers.~comm.). Because of the strong
58: influence of such aggregations on the benthic habitat, it is
59: interesting to question how they are formed and maintained.
60:
61:
62: Herds, flocks, schools, and swarms are all aggregations of social
63: animals. The aggregation is caused by the interaction between the
64: individuals, which attracts them together at large distances
65: \citep{Okubo80}. For animals such as sea-urchins there is little
66: evidence that they are social. In uniform habitat their clumping is
67: mild \citep{Andrew86, Hagen95}. Experiment suggests that urchins will
68: aggregate in the presence of food \citep{Vadas86}, but there is no
69: evidence for a strong social interaction. Moreover, studies of urchin
70: movement have found that while they may exhibit a chemosensory
71: response to algae, they do not show any directed movement towards
72: it \citep{Andrew86}. A recent flume tank study shows that the urchin \emph{Lytechinus variegatus} can move in a directed manner towards a food source under some flow conditions \citep{Pisut02}. This may explain how urchins locate their food at short distances. Both the flow and the chemical signals are likely to be more complex in the urchins' natural environment. In field studies the direction of urchin movement is usually found to be either random or weakly directional \citep{Duggan01, Dumont06, Lauzonguay06}. The question then is how to explain the formation
73: of intense aggregations in an asocial animal, which appears not to be
74: able to move in a directed manner.
75:
76:
77: \begin{figure}
78: \begin{center}
79: \includegraphics*[width=7cm]{f1.eps}
80: \end{center}
81: \caption{Movement of red sea urchins,
82: \emph{Strongylocentrotus franciscanus}, near the boundary of a kelp
83: forest at Santa Cruz Point \citep[redrawn from ][]{Mattison77}. The
84: figure shows the average rate of urchin movement, measured over a 24
85: hour period, at four locations. For comparison, the percentage of
86: urchins which were observed to be feeding, and the weight of algae
87: attached to the urchins' oral surface, are also shown. Within the
88: kelp forest (shaded), feeding is high and movement rates are low.}
89: \label{fig:mattison}
90: \end{figure}
91:
92: A recurrent observation is that there is an inverse relation between urchin movement and macrophyte density
93: \citep{Mattison77, Andrew86, Dance87, Dumont06}. A study by \citet{Mattison77} of
94: red sea-urchins (\emph{Strongylocentrotus franciscanus}) near Santa
95: Cruz found that urchins within a kelp forest moved by 7.5 cm
96: day$^{-1}$, whereas outside it the movement rate increased to over 50
97: cm day$^{-1}$ (Fig.~\ref{fig:mattison}). The reasons for the difference in movement rates between habitats is not clear. Some studies find that movement rate is more for starved urchins \citep{Dix70, Hart90}, whereas others find either no effect \citep{Dumont06} or the opposite relation \citep{Klinger85}. It has also been shown, by using physical models of large algae, that the movement of foliose algae by the water may restrict urchin movement \citep{Konar03}. In this paper, the consequences of differential motility in different habitats will be explored, whatever its cause. Four simple assumptions are made about sea urchin movement:
98: \begin{enumerate}
99: \item Sea urchins are asocial, with the movements of individual
100: urchins being independent
101: \item The direction of sea urchin movement is random (over a suitable
102: time period, which we take to be 24 hours)
103: \item The sea-urchin
104: movement rate decreases as the macrophyte density increases
105: \item The distance moved in a 24 hour period is related to the seaweed density at the beginning of the time-period.
106: \end{enumerate}
107:
108: The consequences of these assumptions are explored, using both
109: analytical techniques and direct simulation. It might seem to be intuitively reasonable that if the urchins are randomly moving then they will disperse, and it will be impossible for them to accumulate into an organised structure like a feeding front. In this paper it is shown that under
110: certain circumstances, and with a suitable representation of
111: macrophyte growth and urchin grazing, the assumptions about urchin movement may lead to persistent urchin feeding fronts.
112:
113: There are other features of urchin movement which are not accounted for by this model. A recent study \citep{Lauzonguay06} of sea urchin movement, which followed the movements of individual urchins using video techniques, showed that the distance moved decreased with increasing urchin density. This effect is not included in the present study. Other authors have concluded that the urchin response to predators may mediate the formation of feeding fronts \citep{Bernstein81}. The model we discuss is a minimal model. The complexities of differential feeding on multi species algal assemblages \citep{Gagnon04, Wright05}, size dependent urchin movement \citep{Dumont04, Dumont06}, seasonal variations in movement rate \citep{Konar01, Dumont04}, relation between behaviour and the supply of drift algae \citep{Dayton84}, interactions between movement and the substrate \citep{Laur86}, or between water movement and urchin movement \citep{Kawamata98} are not included. All demographic processes such as urchin growth, recruitment and mortality have also been ignored. If sufficient data were available these processes could be represented. However, while their inclusion would lead to a more realistic model of a specific system, the purpose of this paper is to explore the consequences of a single urchin behaviour.
114:
115:
116:
117:
118: \section{Urchin movement and the Fokker-Planck equation}
119: \label{sec:movement}
120:
121: \begin{figure}
122: \label{fig:errorfunc}
123: \centering
124: \includegraphics[width=7cm]{f2.eps}
125: \caption{Solution to the Fokker-Planck equation, eq.~(\ref{eq:fp}),
126: describing the dispersal of an initially uniform population of
127: sea-urchins in response to a step-change in macrophyte density. The
128: right-hand side, $x>0$, is kelp forest with the urchin movement being
129: $\lambda_+ = 0.1$ m and the left-hand side, $x<0$, is barren with
130: $\lambda_- = 1 $ m (here $\Delta t = 1$ day). These values are
131: chosen to be comparable with \citet{Mattison77}. (a) The urchin
132: distribution after 30 days, with the initial population having a
133: value $u/u_\infty = 1$. There is a net movement of urchins from the
134: barren region to the kelp-forest, with a sharp peak appearing at the
135: kelp boundary. (b) The width of the peak, $2 \sqrt{ D_+ t}$,
136: increases very slowly. Even after 100 days it is less than two meters
137: wide. The maximum urchin density is constant with time, at 10 times
138: the initial population.}
139: \end{figure}
140:
141:
142: The four assumptions above may be used to formalize sea-urchin movement as a random walk. If $x_i(t)$ is the
143: position of urchin $i$ at time $t$, then its position a time $\Delta
144: t$ later may be represented as
145: \begin{equation}
146: \label{eq:randomwalk}
147: x_i(t+\Delta t) = x_i(t) + \eta(t)\lambda(s(x_i(t))) ,
148: \end{equation}
149: where $\eta(t)$ is a dimensionless random variable with a zero mean and a unit
150: variance, and $\lambda(s(x))$ (dimensions [$x$]) is a characteristic step-size which is a function of the macrophyte density,
151: $s$.
152:
153: If the movement of individual sea urchins satisfies eq.~(\ref{eq:randomwalk}), then the dispersal of the population may be approximated by the continuous
154: Fokker-Planck equation \citep{Turchin98},
155: \begin{equation}
156: \label{eq:fp}
157: {\partial u \over \partial t} = {\partial^2 \over \partial x^2}(D u) ,
158: \end{equation}
159: where $u(x, t)$ is the urchin density and the motility $D(s)$ (dimensions [$x^2$ $t^{-1}$]) is
160: related to the random-walk parameters by
161: \begin{equation}
162: \label{eq:rwp}
163: D(s) = {\lambda(s)^2 \over 2 \Delta t}.
164: \end{equation}
165:
166: The long term behavior of the population $u$ is well-known. If the total number of sea-urchins is constant with time, then the steady
167: state solution to eq.~(\ref{eq:fp}) is
168: \begin{equation}
169: \label{eq:steadystate}
170: u(x,t) = c/D(s),
171: \end{equation}
172: where $c$ is a constant. At equilibrium, the population density will
173: be inversely related to the motility. The sea-urchins will accumulate in areas where the seaweed concentration is higher, and so the individual urchins are moving more slowly. The aggregation of randomly
174: walking foragers in regions with higher food density, is known variously as preytaxis
175: \citep{Kareiva87}, orthokinesis \citep{Okubo80}, or phagokinesis
176: \citep{Andrew86}. An experimental study of ladybugs feeding on an
177: inhomogeneous aphid population showed that, in this case,
178: eq.~(\ref{eq:steadystate}) provided a good description of the data \citep{Turchin98}. The random walk formalism is similar to (although simpler than) that used to understand the formation of traveling bands of bacteria through chemotaxis \citep{Keller71b}.
179:
180: While it has been observed that urchin movement is higher when the
181: algal density is lower, little is known about the functional form of
182: $\lambda(s)$. In the absence of any data, we will simply assume that
183: there is a threshold algal density, $s_c$, at which the rate of urchin
184: movement changes from a minimum to a maximum value,
185: \begin{equation}
186: \label{eq:Dsc}
187: D = \left\{
188: \begin{array}{cl}
189: D_-, & s<s_c \\
190: D_+, & s\geq s_c \\
191: \end{array}
192: \right.,
193: \end{equation}
194: where $D_- > D_+ >0$. Within this model, the urchins have only two behaviours. This simplifying assumption has the advantage of making
195: analytic solutions to the Fokker-Plank equation possible.
196:
197: \section{Analytical solutions}
198:
199: \subsection{Solving for a fixed boundary}
200:
201:
202: As a first step towards understanding the formation of
203: feeding-fronts, the response of an urchin population to a step-change
204: in the motility is considered. The boundary between the barren and the kelp regions is assumed to be fixed, with the macrophyte density being
205: greater than the critical density, $s_c$, for $x>0$ and less than $s_c$ for $x\leq 0$. It
206: follows from eq.~(\ref{eq:Dsc}) that the motility is $D = D_+$,
207: ($x>0$) and $D = D_-$, ($x \leq 0$), where $D_\pm =
208: \lambda^2_\pm/2\Delta t$. If it is assumed that the urchin population
209: is initially uniformly distributed, then $u(x, 0) = u_\infty$, where
210: $u_\infty$ is a constant.
211:
212: Away from the boundary between the two regions, the motility is
213: constant and eq.~(\ref{eq:fp}) reduces to a diffusion equation. If we
214: write $u(x, t) = u_+(x, t)$, ($x \geq 0$) and $u(x, t) = u_-(x, t)$,
215: ($x < 0 $) then, for the derivatives on the right hand side of
216: eq.~(\ref{eq:fp}) to be continuous, we require that
217: \begin{equation}
218: \label{eq:continuity}
219: D_+u_+(0,t) = D_-u_-(0,t).
220: \end{equation}
221: We will look for a solution which has both $u_+(0,t)$ and $u_-(0,t)$
222: constant with time, and so will require that $\partial^2 (D u)/\partial x^2 |_{x=0} =
223: 0$. Because the total urchin population is constant, any increase in the urchin density at positive $x$ must be matched by a decrease in density at negative $x$,
224: \begin{equation}
225: \label{eq:areas}
226: \int_0^\infty (u_+-u_\infty) dx = \int_{-\infty}^0
227: (u_\infty - u_-) dx .
228: \end{equation}
229: The solution to a diffusion equation with a constant boundary is given
230: by the complementary error function,
231: \begin{equation}
232: \label{eq:erfc}
233: \mathrm{erfc}(x) = 1 - {1 \over \sqrt{\pi}} \int_0^x e^{-\beta^2} d\beta,
234: \end{equation}
235: with $\beta$ being an integration constant. The solution
236: for the urchin population may be written as
237: \begin{equation}
238: \label{eq:uprime}
239: u_\pm(x, t) = u_\infty\left(1 \mp \gamma_\pm \mathrm{erfc}\left(|x|/2\sqrt{D_\pm t}\right)\right),
240: \end{equation}
241: where $\gamma_\pm$ are constants which must satisfy
242: \begin{equation}
243: \label{eq:gamma}
244: D_+\gamma_+ + D_-\gamma_- = D_+ - D_-
245: \end{equation}
246: in order to solve eq.~(\ref{eq:continuity}). For eq.~(\ref{eq:areas}) to hold,
247: \begin{equation}
248: \label{eq:area}
249: \gamma_+/\gamma_- = \sqrt{D_-/D_+}.
250: \end{equation}
251: With this ratio $\partial^2 Du/\partial x^2|_{x=0} = 0$, and the
252: Fokker-Planck equation is solved throughout the domain. From
253: eqs.~(\ref{eq:gamma}) and (\ref{eq:area}) it follows that
254: \begin{equation}
255: \label{eq:gammapm}
256: \gamma_\pm = {D_+ - D_- \over \sqrt{D_\pm}(\sqrt{D_-}+\sqrt{D_+})}.
257: \end{equation}
258:
259: A plot of the solution is given in fig.~(\ref{fig:errorfunc}). The initially uniform urchin density develops a peak at the boundary between the two regions. There is an increased urchin density just inside the kelp, and a depleted region on the barren side of the boundary. The height of the peak is constant with time, but the width grows steadily. On the barren side of the peak there is a region where the sea-urchin density is less than the initial value.
260:
261: \subsection{Solving for a moving boundary}
262:
263:
264: We now look for traveling wave solutions of the Fokker-Planck
265: equation, representing a steadily moving urchin front. At this stage, the grazing of the urchins is not considered, it is simply assumed that the boundary between the two regions moves at a constant velocity $c$. The variable
266: $z = x - ct$ is introduced. The traveling solutions are functions of $z$ only, and
267: they satisfy the equation, derived from eq.~(\ref{eq:fp}),
268: \begin{equation}
269: \label{eq:fpz}
270: -c {du \over dz} = {d^2 Du \over dz^2},
271: \end{equation}
272: where $u = u(z)$ and $D=D(z)$. If the boundary between the grazed and ungrazed regions falls at $z=0$, the motility is
273: \begin{equation}
274: \label{eq:Dz}
275: D(z) = \left\{
276: \begin{array}{cl}
277: D_-, & z\leq 0 \\
278: D_+, & z > 0 \\
279: \end{array}
280: \right. .
281: \end{equation}
282: By integrating eq.~(\ref{eq:fpz}) twice, an integral equation for the urchin density is obtained,
283: \begin{equation}
284: \label{eq:Du}
285: D(z)u(z) = - c \int_{-\infty}^z (u(x) - u_\infty) dx + D_-u_\infty ,
286: \end{equation}
287: where the constant of integration, $u_\infty$, has been chosen so that $u(\pm \infty) = u_\infty$.
288:
289: It is straightforward to verify that the solution to eq.~(\ref{eq:Du}) is given by the function
290: \begin{equation}
291: \label{eq:usolution}
292: {u(z)\over u_\infty} = \left\{
293: \begin{array}{ll}
294: 1, & z \leq 0 \\
295: {D_- - D_+ \over D_+}e^{-cz/D_+} + 1, & z>0
296: \end{array}
297: \right. .
298: \end{equation}
299: If the motility is larger in the grazed region, $D_- > D_+$, then the traveling wave solution has the form of a feeding front, with a peak at the boundary between the regions. The maximum density within the feeding front occurs on the boundary, with a density $u_\infty D_-/D_+$. The urchin density is constant throughout the barren region, and decays exponentially towards the ungrazed side of the boundary, the front having a width of $D_+/c$.
300:
301: The feeding front can only propagate continually if there is a non-zero urchin density within the ungrazed region. Otherwise, the front will lose urchins as it travels and decay away.
302:
303:
304: \section{Introducing seaweed}
305: \begin{figure}
306: \label{fig:ghplot}
307: \centering
308: \includegraphics[width=7cm]{f3.eps}
309: \caption{Variation in macrophyte growth $g(s)$ (solid line) and
310: urchin grazing $h(s)$ (dashed line) as a function of the macrophyte
311: density, $s$. The curves follow eqs.~(\ref{eq:gs}, \ref{eq:hs}),
312: with the parameters $\mu_s = 0.01$ day$^{-1}$; $s_0 =
313: 0.03$ $s_{\max}$;
314: $k_s = 0.05$ $s_{\max}$; $\alpha = 0.001$ $s_{\max}$ urchin$^{-1}$ day$^{-1}$. The dashed line is drawn for an
315: urchin density of $u_\infty = 2$ urchin m$^{-2}$. The three intersection points of
316: $g(s)$ and $h(s)$ are labeled by the macrophyte densities $s_1$,
317: $s_2$ and $s_3$. The upper and lower dotted lines show the urchin
318: grazing with the same parameters, but with urchin densities of
319: $u_\infty = 3$ urchin m$^{-2}$ and $u_\infty = 1$ urchin
320: m$^{-2}$, respectively. With these densities there is only one
321: solution of $ds/dz = 0$ (eq.~\ref{eq:dsdz}), and so there are no
322: possible traveling wave solutions that could represent an urchin feeding-front.}
323: \end{figure}
324:
325: Having identified a frontal solution to the urchin density when the boundary is moving steadily, the question is whether there are
326: traveling wave solutions to the coupled seaweed-urchin equations. The change in algal density is taken to occur through a combination
327: of growth and grazing,
328: \begin{equation}
329: \label{eq:dsdt}
330: {\partial s \over \partial t } = g(s) - h(s)u,
331: \end{equation}
332: where $g(s)$ describes the algal growth and $h(s)u$ is
333: the grazing rate of the urchins on the seaweed. There is no explicit seaweed
334: dispersal included. Recruitment from a wider
335: seaweed population is simply represented by a non-zero intercept of
336: $g(s)$.
337: For a traveling wave solution to exist, $s$ must be a function
338: of $z = x - ct$ only, so
339: \begin{equation}
340: \label{eq:dsdz}
341: {\partial s \over \partial z } = h(s) u/c - g(s)/c.
342: \end{equation}
343: At $z=\pm \infty$ the population must be in equilibrium, $h(s)u_\infty
344: = g(s)$, with $s(\infty) > s_c$ and $s(-\infty) \leq s_c$. There must
345: be at least three real, positive solutions to
346: \begin{equation}
347: \label{eq:condition}
348: \partial s/\partial z|_{u=u_\infty} =0,
349: \end{equation}
350: which we shall call $s_1$, $s_2$, and $s_3$ ($s_3>s_2>s_1$). The
351: solutions $s_1$ and $s_3$ are stable, and $s_2$ is unstable. In order
352: that $s<s_c$ for $z<0$ it is required that
353: \begin{equation}
354: \label{eq:scbound}
355: s_2>s_c>s_1.
356: \end{equation}
357: If this does not
358: hold then no traveling wave solutions can be obtained. The propagation
359: speed can be obtained by requiring that $s = s_c$ at $z=0$, where the
360: urchin density, $u$,
361: in eq.~(\ref{eq:dsdz}) is obtained from eq.~(\ref{eq:usolution}).
362:
363:
364:
365: As a plausible example, assume that macrophyte growth is logistic
366: \begin{equation}
367: \label{eq:gs}
368: g(s) = \mu_s (s + s_0)(1 - s/s_{\max}),
369: \end{equation}
370: where $\mu_s$ (dimensions $[t^{-1}]$) is the growth-rate
371: $s_{\max}$(dimensions $[s]$) is the macrophyte carrying
372: capacity and the term $\mu_s s_0$ (dimensions $[s t^{-1}]$) represents
373: a background recruitment rate. With this growth function, the
374: macrophyte will grow to a density $s_{\max}$ in the absence of
375: urchins, and this growth to a maximal density will take a time of
376: order $\mu_s^{-1}$.
377:
378: An appropriate representation of grazing is the Holling
379: type II or Michaelis-Menten equation \citep{Holling59, Begon96}
380: \begin{equation}
381: \label{eq:hs}
382: h(s) = { \alpha s \over s + k_s},
383: \end{equation}
384: where $\alpha$ (dimensions $[s u^{-1} t^{-1}]$) parameterizes the maximal grazing rate per urchin,
385: and $k_s$ (dimensions $[s]$) is the half-saturation constant for urchin grazing. At low
386: algal densities the grazing function decreases to zero,
387: representing the difficulty that urchins have in locating food when the macrophyte is sparse.
388:
389:
390: As an example, growth parameters relevant to the New Zealand
391: alga \emph{Ecklonia radiata} are used. This species grows to a mature
392: size within a year, and so an order-of-magnitude growth-rate is estimated to be
393: $\mu_s = 0.01$ day$^{-1}$. The recruitment density $s_0$ will be site
394: specific, depending on the abundance of mature alga in the surrounding
395: area. It is simply assumed that $s_0$ is a small fraction of the
396: maximum density, $s_0 = 0.03 s_{\max}$. An estimate of urchin
397: grazing rates may be obtained from the results of a small experiment carried out by Russell Cole (1993). A square
398: meter quadrat was loaded with urchins (\emph{Evechinus chloroticus}),
399: to a density of 60 m$^{-2}$, and the decrease in the abundance of the
400: alga \emph{E. radiata} was monitored. Even at this high urchin
401: density the decline in alga was slow, with a time-scale of $\sim$ 20
402: days. The maximum grazing rate is therefore $\alpha =
403: 1/(20\times 60) = 0.001$ $s_{\max}$ urchin$^{-1}$ m$^{2}$ day$^{-1}$. The
404: algal density at which the urchin grazing is half of its maximum is
405: taken to be $k_s = 0.1 s_{\max}$. In the absence of any data on the
406: variation of urchin motility with algal concentration, it will simply
407: be assumed that the critical algal density is $s_c = k_s$. The growth
408: and grazing curves that result from these parameters are shown in
409: fig.~(\ref{fig:ghplot}), for three differing urchin densities. Detailed
410: experiment would be needed to verify both the functional form and the
411: parameterization of the growth and grazing functions. The intent here
412: is to illustrate the qualitative features of the urchin-macrophyte
413: system, rather than quantitative modeling of a specific case.
414:
415: The existence of three solutions to
416: eq.~(\ref{eq:condition}) could be determined by directly solving this cubic
417: equation. While analytically tractable, the general solution will be
418: complicated. A more amenable estimate of when three real, positive solutions can
419: be found is readily obtained by graphical inspection of the growth and
420: grazing functions, $g(s)$ and $h(s)$. If the recruitment density $s_0$
421: is zero, then three solutions to eq.~(\ref{eq:condition}) will only be
422: found if the initial slope of the grazing function is larger than the
423: initial slope of the growth function. This will only hold if
424: \begin{equation}
425: \label{eq:ugreater}
426: u_\infty > {\mu_s k_s \over \alpha}.
427: \end{equation}
428: If $k_s << s_{\max}$, then the maximal grazing rate also needs to be less than
429: the maximal growth rate. This implies that
430: \begin{equation}
431: \label{eq:uless}
432: u_\infty < {\mu_s s_{\max} \over 4 \alpha}.
433: \end{equation}
434: Both of these
435: inequalities, (\ref{eq:ugreater}) and (\ref{eq:uless}), can only be
436: satisfied simultaneously if
437: \begin{equation}
438: \label{eq:kcondition}
439: k_s < s_{\max}/4.
440: \end{equation}
441: If the recruitment density $s_0$ is non-zero but small, $s_0 <<
442: s_{\max}$, then these conditions will still be relevant. For the
443: parameters used in fig.~(\ref{fig:ghplot}) the conditions given in
444: eqs.~(\ref{eq:ugreater}, \ref{eq:uless}) translate to the requirement
445: that 1 urchin m$^{-2}$ $< u_{\infty} <$ 2.5 urchin m$^{-2}$. These are not exact bounds, but they provide a useful estimate of the range
446: over which three solutions to eq.~(\ref{eq:dsdz}) can be found.
447:
448: For a feeding-front solution to exist it is also necessary that the
449: transition from high to low urchin motility occurs at a macrophyte
450: density, $s_c$, which is between $s_1$ and $s_2$
451: (eq.~\ref{eq:scbound}). In the case presented in
452: fig.~(\ref{fig:ghplot}), this would be satisfied by $s_c = 0.1
453: s_{\max}$. The range of initial urchin densities over
454: which a feeding front solution develops is small, with a factor of less than 3 between a density that leads to macrophyte beds and a density
455: that results in urchin barrens.
456:
457: \section{Numerical simulations}
458: \label{sec:numsim}
459:
460: \begin{figure}
461: \label{fig:peakcontour}
462: \centering
463: \includegraphics[width=7cm]{f4.eps}
464: \caption{Comparison between the theoretical form of a traveling wave
465: (eq.~\ref{eq:usolution}) and numerical simulations, as described in
466: section \ref{subsec:traveling_wave}. (a) The theoretical peak height, $D_-/D_+$ (b) The
467: theoretical width, $D_+/c$. (c) The maximum height
468: of the peak from the simulations (d) The width of the peak. The half-shading masks the region where the peak height is too small to allow a width to be reliably calculated.}
469: \end{figure}
470:
471: \subsection{The traveling wave}
472: \label{subsec:traveling_wave}
473: For comparison with the analytic solutions numerical simulations are carried out.
474: The first set of simulations aims to check the validity of the traveling wave
475: solution, eq.~(\ref{eq:usolution}). A one dimensional model is built, which
476: begins with uniformly distributed urchins, each urchin having a real-valued
477: position. The boundary between the low and high motility regions begins
478: at $x = 1200$ m and moves towards the right at a velocity $c = 1$ m day$^{-1}$.
479: At each timestep, for each urchin, a random number $\eta$ is generated from a normal
480: distribution with zero mean and variance. If
481: the urchin is to the right of the boundary it is moved by $\lambda_+ \eta$.
482: Otherwise, the urchin is moved by $\lambda_- \eta$. These rules capture
483: the assumptions which led to the derivation of eq.~(\ref{eq:usolution}). A
484: window 800 m wide is maintained around the boundary, with a border 150 m
485: wide beyond that. Urchins are added or removed from the simulation to hold
486: the density constant within the two border regions. Any urchins which move
487: beyond the border are removed. The simulation starts with a uniform density
488: of 50 urchins m$^{-1}$. It is run for 2000 timesteps, with data from the final
489: 200 timesteps being grouped into 1 m long bins and averaged. The whole
490: simulation is repeated for a range of $\lambda_+$ and $\lambda_-$ ($\lambda_- > \lambda_+$). A comparison
491: of the theoretical and the numerical peak widths and heights are shown in
492: fig.~(\ref{fig:peakcontour}). There is good agreement between the two approaches,
493: confirming that these simple assumptions can lead to a propagating peak in urchin
494: density.
495:
496:
497:
498: \subsection{Two dimensional simulations with macrophyte}
499:
500: Finally, a simulation is run to check the stability of the
501: feeding fronts in a two-dimensional setting, with macrophyte. A numerical domain
502: is used which represents a 500 m $\times$ 500 m square, divided into 1 m$^2$ cells.
503: Each cell has a seaweed density, $s$, with the density going from $s = 0$ on the left
504: hand side of the domain to $s = s_{\max}$ on the right hand side. The seaweed distribution
505: has some initial variability, introduced by adding a random function to the linear gradient
506: (fig.~\ref{fig:seaweed}a). The random function has greater variability at longer length
507: scales, with a Fourier transform that decays as $f^{-3/4}$, where $f$ is the wavenumber.
508: This is done to introduce noise into the model, capturing in some way the natural environmental
509: variability. Urchins are then added, uniformly distributed through the whole domain,
510: and with an average density of 1.5 urchin m$^{-2}$. At each timestep the seaweed within each cell changes
511: according to eq.~(\ref{eq:dsdt}). A simple finite-difference approximation is used, and
512: the seaweed density is always kept above zero. The urchin density is calculated from the
513: number of urchins within each 1 m$^2$ cell, and the seaweed is grazed accordingly. The
514: urchins are then moved by a random amount, with the size of the step, $\lambda$, depending
515: on whether the seaweed density exceeds the threshold. Any urchins moving outside the
516: domain are reflected back into it, so the total number of urchins within the domain is constant.
517:
518:
519: \begin{figure}
520: \label{fig:seaweed}
521: \centering
522: \includegraphics[width=7cm]{f5png.eps}
523: \caption{Simulation of seaweed (a, c, e, g) and urchins (b, d, f, h), showing the
524: formation of a feeding front. The pictures are made at 0 (a, b), 600 (c, d),
525: 3000 (e, f) and 6000 (g, h) model days.}
526:
527: \end{figure}
528:
529: The results are shown in fig.~(\ref{fig:seaweed}). In the simulations shown, the following
530: parameters have been used, $\lambda_+ = 0.05$ m day$^{-1}$; $\lambda_- = 1$ m day$^{-1}$;
531: $k_s = 0.05 s_{\max}$; $s_0 = 0.01 s_{\max}$; $\mu = 0.01$ day$^{-1}$; $\alpha = 0.001
532: s_{\max}$ urchin$^{-1}$ m$^{2}$; and $s_{c} = k_s$, similar to the parameters in fig.~(\ref{fig:ghplot}).
533: The simulations are run for 10,000 model days, with the figure showing
534: the seaweed density and the urchin density at 0, 600, 3000 and 6000 model days.
535:
536: From the start of the simulation the seaweed density becomes increasingly polarized,
537: with areas of urchin barren, and areas of close to maximum density. A feeding front
538: develops along the boundary between the regions, and the boundary slowly propagates
539: towards the ungrazed region. The front appears stable, becoming smoother with time.
540:
541: \section{Discussion}
542:
543: The simple assumptions of differential urchin movement in response to seaweed density
544: lead to the formation and propagation of an urchin feeding front, in qualitative agreement
545: with observations. No social behavior needs to be assumed to explain the persistence of
546: the front, and the motion of each urchin can be random. The system provides an excellent example of how simple individual processes can lead to spatial pattern. The development of the fronts shows the importance of correctly representing
547: movement. Diffusion approximations, based on Fickian diffusion, are often used to represent
548: animal dispersal \citep{Okubo80}. Because Fickian diffusion will always lead to the density of a population
549: decreasing (at least in the absence of any reproduction or migration) it is unable to generate sharp fronts. A simple change to the representation of dispersal, from Fickian to Fokker-Planck, leads to a model that captures the qualitative features of the system. The focus of the analysis has been on demonstrating that the system can develop a stable propagating front. This simple model may also be used to explore the dynamics of more transient phenomena, such as the effect of a localised recruitment of urchins, or how a patchy mosaic of barrens and macrophyte habitat can be maintained. With the two states that are stable to small perturbations and the transitional wave transforming one to the other, the urchin-macrophyte system has many of the features of excitable media \citep{Murray93}. There is ample scope for further exploration of this analogy.
550:
551: With the movement rates used here, the propagation speed of the front is very slow.
552: In the two dimensional simulation, the front moved at a speed of 10 m per model year. This is on a similar order to propagation speeds of 2.5 m month$^{-1}$ reported from field observations \citep{Gagnon04}.
553: In contrast the aggregation of \emph{Lytechinus variegatus} in Florida Bay was reported to
554: move at 6 m day$^{-1}$. The propagation rate will be strongly dependent on the details of
555: the urchin grazing. It is likely that the assumption of asociality, or of urchin independence, breaks down at the
556: high densities encountered in the front. Because of the very narrow spatial extent of the frontal region, the urchin behavior at high densities will effect the outcome of the model. To produce a quantitatively accurate model would
557: require more detailed observations. Studies which focus on the movements of individual urchins \citep{Lauzonguay06} are likely to generate the data required to build a better representation of the frontal dynamics. For example the inclusion of a traffic-jam effect, where the movement rate of the urchins decreases as the density increases \citep{Lauzonguay06}, will result in an increased urchin density within the front.
558:
559: As discussed in the introduction, there are other processes that are known to influence urchin behaviour which could be represented within a model of this nature. Unfortunately, the effects of many of these factors have only been measured in a few isolated experiments, and there is insufficient published data to include them. The model developed here is in many ways a null model. It is hoped that it will inspire experimentalists to collect the individual based data which is needed to understand the full detail of how urchin feeding fronts are formed and maintained.
560:
561:
562:
563:
564: \section{Acknowledgements}
565: I am grateful to Andrew Visser for discussions on random walks; to Russell Cole for discussions on urchin biology and to Alison MacDiarmid and Alistair Dunn for their support and encouragement. This project was funded by the New Zealand Foundation for Research in Science and Technology.
566:
567: \bibliography{kk}
568:
569: \end{document}
570:
571:
572:
573: