cond-mat0307567/mep.tex
1: \documentclass[aps,preprint,groupedaddress,floatfix]{revtex4}
2: \usepackage{graphics}
3: \usepackage{epsfig}
4: \bibliographystyle{apsrev}
5: %\bibliographystyle{unsrt}
6: \begin{document}
7: \title{Maximisation Principles and Daisyworld}
8: \author{G.J. Ackland}
9: \affiliation{School of Physics, The University of Edinburgh, Mayfield Road, 
10: Edinburgh EH9 3JZ, UK 
11:  E-mail: gjackland@ed.ac.uk}
12: 
13: 
14: \begin{abstract}
15: {\bf We investigate whether the equilibrium time averaged state of 
16: a self-organising system with many internal
17: degrees of freedom, 2D- daisyworld, can be described by 
18: optimising a single quantity.
19: Unlike physical systems where a principle of maximum energy production  
20: has been observed, Daisyworld follows evolutionary dynamics rather
21: than Hamiltonian dynamics.  We find that this is sufficient to
22: invalidate the MEP principle, finding instead a different principle, 
23: that the system self-organises to a state which maximises the amount of life.}
24: 
25: Keywords: Daisyworld, entropy, feedback, gaia, logistic
26: \end{abstract}
27: 
28: \date{\today}
29: \maketitle
30: 
31: 
32: 
33: \section{Introduction} 
34: 
35: 
36: Thermodynamics provides an excellent macroscopic description of the 
37: physics of matter in terms of time-independent 
38: averaged quantities such as energy, entropy, free energy.
39: These arise from the underlying microscopic, time-dependent motions of 
40: atoms. The two descriptions are unified through statistical 
41: mechanics, and therefore contain the same information, but in 
42: in the absence of atomic-level detail 
43: the thermodynamic picture is far more useful for everyday application.
44: 
45: It is interesting to consider whether ecosystems can similarly be described
46: by macroscopic averages rather than details of individual species.
47: 
48: Recent work has suggested that the principle of
49: maximum entropy production (MEP) may be a good way to investigate
50: complex, self organising systems.  Lorenz et al (2001,2002) 
51: suggest that the atmosphere of Earth and Titan may be in such a state.  
52: Dewar (2003) has shown
53: that other known distributions of open, dynamical systems, such as
54: self-organised criticality (Bak et al, 1988), the fluctuation theorem (Crooks, 1999) and Zipf's Law (Zipf 1932).
55: The basis of Dewar's argument is that if one takes an appropriate
56: average over all its internal possibilities\footnote{This is implicit 
57: in using the sum over all paths to build a partition function}, 
58: consistent with the known
59: fluxes and internal energies, MEP will emerge.  Of course, real
60: systems do not actually sample all possibilities, so this only
61: provides a sufficient conditions for MEP, the necessary condition is
62: that the average over possibilities actually observed converges
63: rapidly. This parallels an argument in equilibrium statistical
64: mechanics, where thermodynamic averages are calculated from an ergodic
65: hypothesis (that all microstates contribute according to  Boltzmann
66: statistics) to which the chaotic trajectories of real systems
67: rapidly converge.  
68: 
69: Boltzmann statistics follow from Hamiltonian
70: dynamics. In the case of ecological systems, 
71: the dynamics are driven by evolution, and no proof of 
72: convergence of an evolving
73: system to a Boltzmann-type distribution yet exists. It is interesting
74: to ask whether more general principles, such as MEP, can be applied in
75: an ecological context.  In this paper we explore a particular
76: ecological model, 2D-Daisyworld, in which the MEP state can be
77: rigorously defined, to see whether the system does, in fact,
78: self-organise thereto.  
79: 
80: The 
81: 2D-Daisyworld, is based on
82: previous work (Von Bloh et al., 1997,1999, Ackland et al 2003).  Life is
83: reduced to a single species 'daisies' with a single property
84: (albedo) upon which selection can act. The local environment is
85: reduced to a single variable temperature. The planet is warmed by
86: a Sun whose heat production (insolation) increases slowly. Growth is
87: only possible across a narrow range of temperatures (5-40$^o$C) with
88: the mid point 22.5$^o$C being optimal and the growth rate dropping
89: parabolically to zero at the extremes. Local temperature is partly
90: determined by insolation and albedo and partly by diffusive heat flow
91: across the planet. Growth occurs by seeding of bare ground adjacent to
92: existing daisy populations, eliminating the need in the original,
93: zero-dimensional daisyworld of Watson and Lovelock, (1983) to predefine
94: different local environments for different albedo types. Each
95: (asexual) offspring mutates in albedo at random. 
96: 
97: The original daisyworld  is deterministic, cast in
98: the form of differential equations, and has spawned many variants
99: to incorporate specific evolutionary dynamics or natural selection
100: (Lenton, 1998, Lenton and Lovelock 2000, Lovelock 1998, Robertson 
101: and Robinson, 1998, Staley 2002, Sugimoto, 2002).  
102: The 2D version is stochastic, with
103: many internal degrees of freedom, allowing an entropy production to
104: be defined by the range of albedo (called biodiversity), as well as 
105: through heat flow. 
106: 
107: Historically, daisyworlds were designed to illustrate the 
108: self-regulating effects of coupling between life and the 
109: environment, and their impact on evolution by group selection.
110: This is not the issue of interest here - daisyworld is used as 
111: a test case for MEP. 
112: 
113: 
114: \section{Methods} 
115: The temperature field changes as: 
116: 
117: \begin{eqnarray} C\dot{T}(x,y,t) &=& D_T\nabla^2{T}(x,y,t) \nonumber \\ 
118:                                  & & -\sigma_B{T}^4(x,y,t) \nonumber \\
119:                                  & & + S_L(1-A(x,y,t)) \end{eqnarray}
120: 
121: 
122:  Where $C=2500$ is the heat
123: capacity, T the temperature field, $D_T=D/C$ the thermal diffusion 
124: constant $\sigma_B=5.67\times 10^{-8}$
125: the Stefan-Boltzmann constant, $S_L=S/432.3$, the insolation 
126: normalised to that which gives T=295.5K on a bare planet
127: and $A$ the albedo field (daisy/bare ground). Each lattice point has an
128: associated temperature and albedo. $A$ ranges from 0 to 1, with 0.5
129: being the value for bare ground.  The daisy field is discrete and
130: evolves stochastically. We use a square lattice with eight
131: neighbouring sites. At each time step each site is examined and: 
132: 
133: (1) If occupied by a daisy, it changes to bare ground with probability
134: $\gamma(T)$
135: (the "death rate"). 
136: 
137: (2) If
138: occupied by bare ground, the site is populated with probability $\beta(T)$ 
139: (the "growth rate"). 
140: 
141: The new daisy 
142: has the albedo of its parent, from a randomly chosen neighbouring site, 
143: with a random fluctuation
144: (mutation) drawn from a uniform distribution between r and -r.  If the
145: neighbouring site is bare, no growth occurs. 
146: 
147: 
148: 
149: \section{MEP or Gaia?} 
150: 
151: Two global principles will be advanced, these are not new, but formal 
152: definitions for the present system are made.
153: 
154: 1/  Maximal Entropy Production (Dewar, 2003) - the system self organises 
155: to produce entropy as rapidly as possible
156: 
157: 2/  The Gaia principle (Lovelock, 1998) - the system self organises 
158: to maximise the amount of life
159: 
160: In previous work on the 2D case (Ackland et al 2003), it was shown that 
161: multiple solutions satisfied the Gaia principle.  A secondary
162: criterion was found, maximisation of biodiversity 
163: entropy\footnote{Finite diffusivity leads to clump formation, which confounds this
164: result due to neglect the entropy associated with variation in clump
165: size. In principle this could be incorporated, but for the present
166: purposes the maximum entropy principle observed in the limit of
167: infinite diffusivity is sufficient.}
168:   $N_A \ln N_A$
169: where $N_A$ is the number of daisies with albedo A, leading to a mean
170: distribution 
171: 
172: \begin{equation} \langle N_A \rangle = e^{\beta A} \end{equation}
173: 
174: with $\beta$  given by the solution of:
175: 
176: \begin{equation} \langle  A \rangle \int^1_0 \exp(\beta A)dA =  
177: \int^1_0 A \exp(\beta A)dA\end{equation}
178: 
179: 
180:  Thus the global state of 2D daisyworld was classified 
181: in terms of two quantities, maximised
182: growth (i.e. temperature regulation at the favoured growth
183: temperature) and maximised biodiversity.  These are equivalent 
184: to satisfying both MEP and Gaia principles simultaneously.
185: 
186: It is possible to make a small change to the 2D daisyworld so that 
187: Gaia and MEP cannot be satisfied simultaneously.
188:  To do this simply requires making the
189: death rate temperature dependent, and the temperature for minimum
190: death rate ($T_d$) different from that for maximum birthrate ($T_g$).  In
191: this case the MEP will predict that the temperature will adjust close
192: to $T_g$, while the Gaia hypothesis predicts it will adjust close to $T_d$.
193: \footnote{We say "close to": these are the limiting cases for low death rates
194: and high birth rates, assuming uncorrelated dead patches.}
195:   The
196: explanation for this is straightforward: MEP predicts that entropy
197: (biodiversity) be produced at the fastest possible rate: maximising
198: the birthrate.  By contrast, the Gaia hypothesis
199: is that the total number of daisies be maximised: minimising deathrate.  
200: Thus investigating whether the mean
201: temperature tends to $T_g$ or $T_d$ determines which (if either) 
202: of the two hypotheses
203: applies.  Figure.\ref{fig:2d2}  shows the variation of
204: mean temperature and population as a function of insolation: the mean
205: temperature obtained is $T_d$ and the population is $1-\gamma(T_d)$.  
206: Thus we conclude that the dynamics satisfy Gaia rather than MEP.
207: 
208: 
209: 
210: \section{Thermal Entropy Production}
211: 
212: The arguments above pertain to biodiversity entropy.  With a finite
213: diffusivity, one can also investigate whether the MEP can be applied
214: to entropy production by heatflow in daisyworld.  There are conceptual
215: difficulties with defining the system here - the models for MEP on
216: Earth (Ozawa et al, 2003), and Dewar's (2002)
217: formulation assume constant flows of
218: energy into and out of the system; the self-organising albedo of
219: daisyworld is able to regulate these flows.  Similarly, the diffusion
220: constant is fixed in 2D-daisyworld, whereas in atmospheric models it
221: is assumed to self adjust.  Notwithstanding this, there is a range of
222: temperatures in 2D daisyworld, and heat is transported from hotter to
223: colder regions, so we can define thermal entropy production\footnote{factor of two arises from double counting} as
224: 
225: \begin{equation}
226: dS/dT = \int \frac{dQ}{T} = \sum_{i} \sum^8_{k=1} \frac{D_T (T_i - T_k)}{2T_i}
227: \end{equation}
228:  
229: Given the constraints on flow, it is not straightforward to determine
230: the maximum possible entropy production.  However, after
231: an abrupt change in the external forcing (insolation) the system will
232: be out of equilibrium, and will then evolve towards a steady state. If
233: MEP holds, the state immediately after the change should have lower
234: entropy production than the steady state.  Figure \ref{fig:entropy}
235:  shows no sign of
236: such behaviour.  Monitoring entropy production also enables us to
237: monitor another property, the net heat flow in and out of the system.
238: In steady state, the mean values of these heat flows should be equal 
239: - in 0D daisyworlds solutions initially assumed 
240: input and output were in balance at all times -
241: this is not required for a nonlinear system, but proof of the
242: stability of these solutions came later (Saunders, 1994).  In the
243: 2D-daisyworld Figure \ref{fig:ebalance}
244: shows the time variation of flow in and flow out.  An
245: interesting feature here is that flow in (basically albedo) 
246: fluctuates much faster than flow out
247: (basically temperature), which also has a delayed response 
248: of many daisy lifetimes. 
249: 
250:  Observations of planetary MEP are
251: presumed to indicate that large-scale atmospheric heat transfer mechanisms are
252: self-organising.  However, even in the curved-planet
253: version of 2D-daisyworld (Ackland et al, 2003), 
254: very strong temperature regulation
255: eliminates large scale temperature variation and hence the driving
256: force for the atmosphere (except in the case of desert formation).
257: Figure \ref{fig:entropy}
258: shows that the mean heat entropy produced per cell does not tend 
259: to a maximum during equilibration after a sharp change in insolation.
260: We this conclude that no MEP principle applies to heatflow in 2D daisyworld.
261: 
262: We examined whether there was an optimum value for
263: the diffusion constant D: at given S, heat entropy
264: is largely independent of D.  This is non-trivial, since it
265: implies that for high D temperature variations are suppressed
266: precisely enough to compensate for the increased diffusion constant.
267: However, it indicates that even if D could be optimised by 
268: a hypothetical daisyworld atmosphere, in the presence of strong temperature
269: regulation no detectable maximum entropy producing value
270: of D exists. 
271: 
272: \section{ MEP in Original daisyworld }
273: 
274: In Watson and Lovelock's (1983) 0D
275: daisyworld two daisy albedos exist, denoted by suffices b and
276: w, with albedo $A_b=0.25$ and $A_w=0.75$ 
277: respectively. The relevant equations are:
278: 
279: 
280: Change of fractional areas occupied by each daisy type (set to zero at
281: equilibrium) 
282: 
283: \begin{equation} da_i/dt = a_i(x\beta_i-\gamma);
284:  \label{D1} \end{equation}
285: 
286:  Fraction of free surface for growth 
287: 
288: \begin{equation}x=1-\sum_i a_i; \label{D2} \end{equation} 
289: 
290: Growth rates for daisies.
291: 
292: \begin{equation}\beta_i={\rm max}\{0,1-((22.5-T_i)/17.5)^2 \};
293:  \label{D3}\end{equation} 
294: 
295: Mean planetary albedo (reflectivity) 
296: 
297: \begin{equation} A_p = x/2 + \sum_i A_ia_i; \label{D4} \end{equation} 
298: 
299: Local temperatures over daisyfields
300: 
301: \begin{equation} \label{D5}
302: (T_i+273)^4 = q(A_p-A_i)+(T_e+273)^4 \end{equation}
303: 
304: One further equation closes the set - this represents the equality of
305: absorbed and emitted heat, 
306: As we
307: saw in the 2D case, this equality strictly should be considered as a
308: time average, however if we assume that no variable large heat
309: reservoirs (e.g. icecaps) exist the timescales are such that the
310: following equality holds: 
311: 
312:   \begin{equation}(T_e +273)^4 = LS/\sigma_{B}(1-A_p) \label{D6} \end{equation} 
313: 
314: S being an insolation, L a variable of order unity, and
315: emitted by Stefan-Boltzmann radiation.  
316: 
317: The usual conclusion from this model is that the daisies
318: moderate the temperature, however 
319: the graph of T $vs$ L (figure \ref{fig:0d2}a): shows that 
320: $T_e$ is not optimised to the optimum for growth.   
321: In contrast, a plot of x $vs$ L  (figure \ref{fig:0d2}b) 
322: shows that x is a minimum for all L.  
323: 
324: The daisyworld equations can be solved analytically.  The solutions 
325: fall into four
326: independent classes: no daisies, black-only, white-only, both types.
327: At a given
328: insolation 
329: only one ``living'' solution is stable if the simulation is 
330: started with non-zero concentration of both daisy types.  Dead and living 
331: solutions coexist in regions where the dead planet temperature lies 
332: outside the limits for growth (5-40$^o$C).
333: The stable solutions (Fig.\ref{fig:0d2}) have been found dynamically, 
334: (iterating equations \ref{D1}-\ref{D6} to self consistency) by
335: numerous authors.  The important point here is that
336: the {\it stable} solution 
337: can  be determined by minimisation of $x$ among allowed 
338: solutions.  Adopting the lowest $x$ is the  
339: 0D equivalent to the Gaian principle of maximum life, 
340: as found in the 2D case.
341:  
342: As a further test, we introduce a third daisy species (grey) with 
343: albedo 0.5, the same as bare ground and therefore with no regulating effect.  
344: Equations \ref{D1} show that coexistence of three species is impossible, 
345: and in fact the stable solutions (Fig.\ref{fig:0d3}) 
346: traverse regions of black-only, 
347: black and grey, grey only, grey and white, white only.  In each region 
348: unstable equilibria exist for other combinations, and in each case the 
349: unstable equilibria have higher $x$ than the total one.  This result is 
350: robust against changes to  $q$, $A_i$ and $\gamma$ (including species 
351: dependent $\gamma$).
352: 
353: The equilibrium solution to daisyworld can also be obtained by 
354: either integrating equation \ref{D1} or with a minimisation of 
355: $x$ with respect to $a_i$ subject to constraints.
356: Each gives the same solution (Figs.\ref{fig:0d2},\ref{fig:0d3}), 
357: independent of the history of the system.  Thus {\it the equilibrium state 
358: in 2D and 0D daisyworld can be determined assuming the Gaia principle, 
359: without knowing 
360: detailed dynamics, just as the equilibrium states of thermodynamic systems 
361: can be determined by minimising free energy.}
362: 
363: 
364: \section{Gaia and MEP in the logistic Map}
365: 
366: The logistic map is another widely used model for population growth
367: (Hassell, Lawton and May, 1976).  
368: The resulting populations famously follow the period doubling route 
369: to chaos.  There is no natural selection and again it is interesting 
370: to ask whether this model displays any kind of maximum population/maximum 
371: entropy behaviour.   
372: 
373: The classical logistic map gives an nth population size
374: of
375: 
376: \begin{equation} 
377:   y(n)=ay(n-1)[1-y(n-1)]
378: \end{equation}
379: 
380: The equilibrium distribution for $y$  depends only on the parameter $a$. 
381: The mean population, averaged over 40000 iterations, is shown 
382: as a function of $a$ in fig\ref{log1}. It exhibits a lot of 
383: structure which can be connected with the bifurcation diagram of the 
384: logistic map.  The mean population of the attractor is higher than would be 
385: expected from maximum entropy (all levels equally occupied). 
386: 
387: Turning now to entropy, there are many definitions related to  the 
388: population distribution by analogy with the 2D daisyworld.  For example, 
389: the Tsallis entropy (Latora et al 2000)
390: 
391: \begin{equation}
392: S = \sum_{i=1}^N log(P(y_i)) / Nlog(M/N)
393: \end{equation} 
394: 
395: where $P(y_i)$ is the number of occurences of a population 
396: between $\frac{i-1}{N}$ and $\frac{i}{N}$ during M iterations 
397: of the logistic map (taken after a 1000 iteration equilibration period).
398: 
399: This quantity has a maximum possible value of 1, and 
400: (neglecting numerous distinct special cases of 
401: limit cycles) its value tends relatively smoothly to this value 
402: with increasing $a$. However, its actual 
403: value depends on N.  This only 
404: tells us that the area lying within the attractor increases with $a$.
405:   
406: The attractors in the logistic map provide some constraints 
407: on the mean value of $y$ and the Tsallis entropy: it is 
408: difficult to define what the 
409: ``maximum life'' solution would be.  Thus while it is notable that 
410: perturbed logistic map dynamics gives mean population in its 
411: stable region above 0.5 for all $a$, no quantitative principle emerges.
412: 
413: \section{Conclusions} The 2D-daisyworld model has been used as a
414: testbed to determine whether any global maximisation is applicable to a
415: system governed by the dynamics of natural selection.  
416: We find that Maximum Entropy Production is not valid here, either 
417: for biodiversity or heat transfer entropy.  This casts doubt on some
418: recent work in which the MEP is assumed as a constraint and used to
419: obtain a solution (e.g. Pujol, 2003).  Analysis of 2D and 0D cases 
420: suggests that daisyworlds appear to obey a separate 
421: optimisation principle, in the simple 2D case indistinguishable from 
422: MEP, that of maximising the amount of life present
423: in the system.  The logistic map attractors have above average population, 
424: but for this system it is difficult to define what the maximal population 
425: value should be.
426: 
427: The more fundamental implication is that stability of a
428: dynamic equilibrium can be determined as an optimisation problem on
429: the properties of the system as observed, and that detailed
430: integration of system dynamics through time may be unnecessary in
431: determining the equilibrium state.
432: 
433: 
434: \section{References}
435: 
436: 
437: P. Bak, C. Tang, and K. Wiesenfeld, "Self-organized
438: criticality", Physical Review A 38 (1988) 364
439: 
440:  G. E. Crooks, 
441:  "The Entropy Production Fluctuation Theorem and the Nonequilibrium Work Relation for Free Energy Differences",
442:  Phys. Rev. E 60, 2721 (1999),
443: 
444:  Dewar, R.C. 
445: Information theoretic explanation of maximum entropy production, the fluctuation theorem and self-organized criticality in non-equilibrium stationary states.
446: J.Phys.A 36 631 (2003)
447: 
448: 
449: %Harding, S.P. and Lovelock, J.E. (1996). Exploiter-mediated Coexistence and Frequency
450: %Dependent Selection in a Numerical Model of Biodiversity. Journal of Theoretical Biology 182, 109-116.
451: 
452: Hassell, M. P., Lawton, J. H. and R. M. May. 1976. Patterns of dynamical behavior in single species populations. J. Animal Ecology. 45: 471-476.
453: 
454: 
455: 
456: Kirchner, J.W. (2002). The Gaia hypothesis: Fact, theory, and wishful thinking. Climatic 
457: Change 52, 391-408.
458: 
459: Lenton, T.M. (1998). Gaia and natural selection. Nature 394, 439-447.
460: 
461: Lenton, T.M. and Lovelock, J.E. (2000). Daisyworld is Darwinian: Constraints on 
462: Adaptation are Important for Planetary Self-Regulation. Journal of Theoretical Biology 
463: 206, 109-114.
464: 
465: %Lenton, T.M. and Von Bloh, W. (2001). Biotic feedback extends the life span of the biosphere. Geophysical Research Letters 28, 1715-1718.
466: 
467:  Lorenz, R.D. (2002)
468: Planets, life and the production of entropy.
469: International Journal of Astrobiology 1, 3-13.
470: 
471: Lorenz, R.D., Lunine, J.I., Withers, P.G. and McKay, C.P. (2001)
472: Titan, Mars and Earth: Entropy Production by Latitudinal Heat Transport.
473: Geophys. Res. Lett. 28, 415-418. 
474: 
475: Lovelock, J.E. (1988). The Ages of Gaia - A Biography of Our Living Earth. New 
476: York: W. W. Norton and Co.
477: 
478: % Ya.B. Pesin, Russian Math. Surveys 32 (1977) 55.
479: 
480: Latora,V.  Baranger,M.  Rapisarda,A. and C. Tsallis,C. "Generalization to nonextensive systems of the rate of entropy increase: the case of the logistic map", Physics Letters A 273 (2000) 97. 
481: 
482: May, R. M. 1976. Simple mathematical models with very complicated dynamics. Nature. 261: 459-467.
483: 
484: Ozawa,H. Ohmura,A. Lorenz,R.D. Pujol,T.
485: The second law of thermodynamics and the global climate system - A review of maximum entropy production principle. Reviews of Geophysics (accepted)
486: 
487: Prasad, A, Mehra, V and Ramaswamy, R.  Strange nonchaotic attractors 
488: in the quasiperiodically forced logistic map
489: Physical Review E, 57, 1576-84 (1998)
490: 
491:  Pujol, T. (2002)
492: The consequence of maximum thermodynamic efficiency in Daisyworld.
493: J. theor. Biol. 217, 53-60. 
494: 
495: % R.Rai, R. S. Johal 
496: %Generalized bit cumulants for chaotic systems: Numerical results.
497: %CondMat/9909288
498: 
499: 
500: Robertson, D. and Robinson, J. (1998). Darwinian Daisyworld. Journal of Theoretical 
501: Biology 195, 129-134.
502: 
503: Saunders, P.T. (1994). Evolution without Natural Selection: Further Implications of the 
504: Daisyworld Parable. Journal of Theoretical Biology 166, 365-373.
505: 
506: Staley M
507:     Darwinian selection leads to Gaia in the long run
508: J Theor Biology 218 35 2002 
509: 
510: Sugimoto, T. (2002). Darwinian evolution does not rule out Gaia hypothesis. Journal of 
511: Theoretical Biology.218, 447-455.
512: 
513: Von Bloh, W., Block, A. and Schellnhuber, H.J. (1997) Self-stabilization of the biosphere 
514: under global change: a tutorial geophysiological approach. Tellus 49B, 249-262.
515: 
516: Von Bloh, W., Block, A. and Schellnhuber, H.J. (1999) Tutorial Modelling of geosphere-
517: biosphere interactions: the effect of percolation-type habitat fragmentation. Physica A 
518: 266, 186-196.
519: 
520: Watson, A.J. and Lovelock, J.E. (1983). Biological homeostasis of the global 
521: environment: the parable of Daisyworld. Tellus 35B, 284-289.
522: 
523: Zipf, G.K  Selective Studies and the Principle of Relative Frequency in Language  (Harvard University Press, Cambridge, 1932) 
524: 
525: 
526: \begin{figure}[htb]
527: \protect{\includegraphics[width=\columnwidth]{daisydie.eps}}
528:         \caption{\label{fig:2d2} 
529: Fractional daisy coverage and temperature 
530: (in units of optimal growth temperature,  $T_g=295.5$ ) from a calculation
531: with a 200x200 lattice 
532: ($\gamma=0.02(1 - (\delta+T_d-T)(T+\delta-T_d)/2\delta^2$,
533: $\beta(T) =(\delta+T_g-T)(T+\delta-T_g)/\delta^2$. 
534:   $T_d=305.5$, $\delta=17.5) D=0.1, r=0.02$) 
535: with scaled insolation S varied from 0.7 to 1.7 
536: over $10^6$ updates of daisy and temperature fields with a temperature 
537: dependent death rate, minimal at $T_d=305.5$.   The graph shows that 
538: in the non-desert regime population tends to its maximal value 
539: ($1-x=\gamma/(\gamma+\beta)=0.99$ 
540: in the mean field approximation, where regrowth is always possible) 
541: while temperature is regulated at $T_d$, rather than $T_g$ which would maximise(bio-)entropy production.  Thus the governing global principal appears to be Gaian 
542: (maximal life) rather then MEP. At higher S, fluctuations lead to long 
543: lived deserts, which cause the mean field approximation to break down 
544: as daisies cannot grow in the interior. 
545: }
546: \end{figure}
547: 
548: 
549: 
550: \begin{figure}[htb]
551: \protect{\includegraphics[width=\columnwidth]{Flow_inset2.eps}}
552:         \caption{\label{fig:ebalance} Graph showing net total energy flows into (solid line) and out of 
553: (dotted line) a 200x200 cell 
554: 2D daisyworld system (diffusion = 0.1, death = 0.02, mutation = 0.02). 
555: The simulation starts with S=1.0 and increases by discrete jumps of 0.1
556: every 10000 timesteps.  Up to S=1.7 this is sufficient to allow the 
557: system to return to equilibrium: the excess 
558: incoming flow of energy is reduced by evolution of the albedo.  
559: Beyond 1.7 (see inset) the sudden change generates sufficient die back 
560: to create deserts in the system, which initially cause positive feedback 
561: and increase the incoming flow still further.  Ultimately, evolution of 
562: paler daisies in the non-desert regions allow reinvasion of the deserts, 
563: but the simulation had to be run for 20000 timesteps (S=1.7) and 50000 
564: (S=1.8) to achieve this.  Even in equilibrium, transient desert regions 
565: occur.  At S=1.9 the simulation was run for 120000 
566: timesteps, throughout which desert and populated regions coexisted and 
567: migrated around the planet (there being no curvature to favour one 
568: region over another).
569: Throughout, note that the energy emitted is smoother than that absorbed
570: and lags the energy absorbed by about 500 timesteps (10 mean daisy lifetimes)
571: $\gamma = 0.02, D = 0.1, r=0.02$.
572: }
573: \end{figure}
574: 
575: 
576: \begin{figure}[htb]
577: %  Look for minimum
578: \protect{\includegraphics[width=\columnwidth]{ent.eps}}
579:         \caption{\label{fig:entropy} Graphs showing variation of 
580: configuration (biodiversity) entropy 
581:  $ \langle \sum_{i=1}^{20} p(A_i) \ln p(A_i) \rangle $ from partitioning
582: the albedos into 20 groups, and (heat) entropy 
583: production $ \langle \int dQ/T \rangle  $
584: averaged over the 200x200 lattice and over a period of 50 timesteps
585: (one mean daisy lifetime) for the same calculation as Fig.\protect{\ref{fig:ebalance}}.
586: Each changes monotonically as the temperature is increased through the
587: region of full coverage.  Beyond S=1.7, where fluctuating deserts appear,
588: each measure becomes less well defined.  There is significantly increased 
589: heat flow from the hot deserts to the covered regions, so the heat flow 
590: becomes dominated by the area of desert.  Conversely, the biodiversity 
591: is reduced when much of the planet is unpopulated.
592: Reversing the change in temperature shows that the entropy change 
593: is reversible.
594: }
595: \end{figure}
596: 
597: \begin{figure}[htb]
598: %  Energy volume
599: \protect{\includegraphics[width=90mm]{Temp0D}}
600: \newline
601: \protect{\includegraphics[width=90mm]{Cover0D}}
602:         \caption{\label{fig:0d2} 
603: Results from 0D daisyworld with parameter values from Lenton and Lovelock:
604:       $S/\sigma_B = 1.68\times 10^{10}K^4$,
605:       $q = 2.06425\times10^9$ and 
606:       $\gamma = 0.3$.
607: a) Temperature variation with increasing solar insolation.
608: Thick line show planetary average, dotted lines show $T_b$ and $T_w$.  
609: The four regimes are shown in their region of stability: Simulations
610: commencing with temperatures outside the growing range can stabilise
611: the bare planet solution, but single-daisy and two-daisy solutions do
612: not coexist. Only stable solutions are shown, including the bistable 
613: living/dead regime. 
614: b) coverage: Thick line denotes bare ground, thin lines denote individual species.
615: %The full analytic expression for $T_w$ in terms of $q$ is: 
616: %\protect{$ T_w=\frac{
617: %(-1397124q+1397124\sqrt{2204936436478415714892+q^2})^{2/3}-
618: %162662955948+5698562(-1397124q+1397124\sqrt{2204936436478415714892+q^2})^{1/3})}{2364
619: %((-1397124*q+1397124\sqrt{2204936436478415714892+q^2})^{1/3})} - 273
620: %$
621: %
622: %with a1=2204936436478415714892
623: %}
624: For non-zero $a_b$ and $a_w$,
625: $da_i/dt=0$  means that $\beta_b=\beta_w$ and 
626: from \ref{D3} and \ref{D5}, $T_w$ and $T_b$ are also independent of
627: $L$: $T_w = 22.5$ for $q=0$ falling almost linearly to zero for
628: $q=9.34\times10^9$. For standard parameter values from Watson and Lovelock,
629: $\beta_b=\beta_w=0.918$ $T_w$=17.5 and $T_b$=27.5.  
630: With $T_w$ and $\beta_w$ fixed by the choice of $q$,
631: independent of $L$, as is the amount of bare ground at equilibrium:
632: $x=\gamma/\beta_w$; (from 1).  Equations \ref{D2} and \ref{D4} 
633: now give us simple
634: linear expressions for $a_b$ and $a_w$ in terms of $x$ and the single
635: variable $A_p$.  Thus, once the assumption of species coexistence is
636: made, the only $L$-dependent variables are $T_e$, $A_p$ which can be
637: solved for analytically.  There are three unstable equilibrium solutions, 
638: either without daisies or with a single species Saunders (1994).
639: }
640: \end{figure}
641: 
642: \begin{figure}[htb]
643: %  Three species daisyworld
644: \protect{\includegraphics[width=100mm]{Temp3}}
645: \newline
646: \protect{\includegraphics[width=100mm]{Cover3.eps}}
647:         \caption{\label{fig:0d3}
648: Daisyworld with additional grey daisy species with albedo 0.5. 
649: Other parameters as for Fig 4. 
650: a) Temperature - thick line shows planetary average, thin lines denote 
651: local temperatures over daisies.  Note that the grey-only solution 
652: is stable because it maximises life, despite having no temperature 
653: regulatory effect.
654: b) Coverage: Thick line denotes bare ground, thin lines denote 
655: individual species.  Note that graph is continuous, with bare ground 
656: (x) falling below 0.30625 only in the single-species regions.
657: }\end{figure}
658:  
659: \begin{figure}[htb]
660: % Logistic map
661: \includegraphics[width=0.9\columnwidth]{logx.eps}
662:  \caption{\label{log1}
663: Bold line shows the mean amount of life from the logistic map, 
664: plotted against recursion parameter $a$.  Background dots show the 
665: bifurcation diagram.  
666: A small amount of random noise added to $y(n)$ (drawn from a flat 
667: distribution $\pm 0.001$) eliminates most of the  
668: ordered behaviour, and the  gives mean population above 0.5 throughout.
669: Without noise, the mean population is still above 0.5 for all the 
670: chaotic cases, and for all the ordered windows except for the tiny
671: region between 3.9602 and 3.9616, 0.04\% of the total.  
672: Populations above 0.5 are found for
673: stable attractors across the family of logistic maps of the form 
674: $y_{n+1}=a[y_n(1-y_n)]^\nu$
675: }
676: \end{figure}
677:  
678: 
679:  
680: \end{document}
681: 
682: 
683: 
684: 
685: 
686: