Welcome to your training disk! This disk enables the reader training and practical application of the knowledge that the student will get from reading the main text in the book. The exercises are in EXCEL and are easily accessible by clicking the links here. When in EXCEL, the reader gets back to this text by clicking the corresponding link that appears in the sheet at the bottom right. The EXCEL files are optimized for the display resolution 1280 by 1024 pixels. Using a display with a lower resolution may sometimes result in not the whole area of interest being on screen. In such case we suggest the reader to zoom the EXCEL files to some smaller value, e.g., 75%. To fit all the explanatory frames into one visible part of the EXCEL sheet, parts of the simulations are sometimes hidden behind the explanatory frames. The EXCEL files usually contain the following: The simulation itself in columns A and B, sometimes C (when the dynamics or more populations is simulated). A yellow-highlighted region containing the numbers of individuals at the beginning – at time zero, and all the parameters of the model. We find it convenient to separate these model inputs from the main simulation, as it allows the user to easily change them without changing the formulas in the main simulation. A figure with the result, which uses the data in columns A and B (sometimes C) as inputs. An explanation frame to the right of this figure, where it is explained, what the program simulates, what is the meaning of the resulting figure in the active sheet, if this is not quite straightforward and what are the exercises the reader is expected to perform here. Another explanation frame below this figure, where it is explained, how the program was created – what is the meaning of the formulas in the sheet. This helps the reader to create his/her own programs in EXCEL. Of course, there exist special programs for simulations of population systems. However, programming in EXCEL allows adapting the simulation for any particular situation and – last but not least – it is not necessary to buy the special programs, which are usually not free. We believe that after having performed all the exercises presented here, the reader will be able to produce own EXCEL simulations of population dynamics. Files contained in this disk This disk contains the following files, each of which can also be run independently: Introduction.doc This file PG1.EXE “Game of Life” – a population game using cellular automata Discrete exponential.xls Simulation of the discrete exponential model Discrete logistic.xls Simulation of the discrete logistic model Time delay.xls Simulation of the discrete logistic model with time delay Continuous models - 1 species.xls Simulation of several continuous models for one species: exponential, logistic, logistic with delay, Allee effect, deterioration of environment, harvesting, biocontrol Continuous models - 2 species.xls Simulation of several continuous models for two species: mutualism, competition, predator-prey Chapter 1: Game of life In Chapter 1, we introduce the reader into the “Game of Life”. You can access the program that simulates this game here. In some systems, the following message can appear: “Some files can contain viruses or otherwise be harmful to your computer. It is important to be certain that this file is from a trustworthy source. Would you like to open this file?” This is because you are trying to open an .exe file. Do not worry and click “OK”, this file is innocent. For getting back to this text, you have to stop the program by pressing Ctrl-Break. Start with 5 individuals and run simulations in different starting configurations. Explain the dynamic behavior you observe and the causes of that behavior. Repeat with different numbers of starters. What is the single most important conclusion from this exercise? Chapter 2: The elementary population model 2.1. Exponential model In Chapter 2, we examine the dynamic behavior of single-species populations and develop an elementary feedback model of the population system. The elementary population model, that of exponential growth of the population, is defined in the main text in equation (2.4): Nt Nt 1 RNt 1 . (2.4) Here Nt denotes the number of individuals in the population at time t, and R represents the individual, or per capita, rate of increase over the time period t – 1 to t. From this equation we can calculate the density of the population at the end of a particular time increment from its density at the beginning of the interval and the individual rate of increase. Use the corresponding EXCEL file in this disk to perform the simulations required in Exercise 2.1. Run simulations for different initial numbers of individuals in the population, N0, and for different values of the per capita rate of increase, R. How does the outcome depend on the choice of N0 and R? Where do you see the value of N0 in the graph? How does increase/decline in the value of R perform in the resulting growth of the population? What can you conclude about the growth of the population, if R > 0, R < 0, R = 0? You can access the EXCEL file that simulates the exponential growth here. 2.2. Logistic model Later in the chapter, we allow the individual rate of increase, R, to be dependent on the density of the population, as described in equation (2.5) in the main text: R Rm sN . (2.5) This results in the logistic model, described in equations (2.6) and (2.7): Nt Nt 1 ( Rm sNt 1 ) Nt 1 . (2.6) Nt Nt 1 Rm (1 Nt 1 / K ) Nt 1 . (2.7) Note that the equations (2.6) and (2.7) are equivalent, as we have set s Rm / K . Due to this substitution, both parameters in the equation (2.7) have a clear biological meaning: Rm, is the maximum individual rate of increase (maximum value of R under the assumption that there is a very low number of individuals in the population, so that competition for a scarce resource does not yet occur, and K is the carrying capacity of the environment - the population density where all living space is fully utilized and there is no more room for additional growth. Using the simulation model in EXCEL in this disk, solve Exercise 2.2. Then confirm that the population will show: asymptotic approach to equilibrium when Rm ≤ 1, damped stable oscillations when 1 < Rm < 2, neutrally stable oscillations with amplitude determined by the initial displacement when Rm = 2, unstable, often non-periodic behavior when Rm > 2. You can access the EXCEL file that simulates the logistic growth here. You will find out that in some cases the logistic growth described by equation (2.7) does not give biologically meaningful predictions – for some large values of Rm is the predicted population size, Nt, negative. To get rid of this problem, the following modification of the logistic equation is often used: N t N t 1e Rm (1 Nt 1 / K ) . (2.7a) Because the “regulatory term” Rm (1 N t 1 / K ) is now in exponent, the value e Rm (1 Nt 1 / K ) is never negative and so is the resulting prediction on Nt. The simulation of equation (2.7a) can be found here, in sheet “Modified logistic”, and the comparison of the numbers of individuals, Nt, predicted by equations (2.7) and (2.7a) are given in the same file (accessible here) in sheet “Comparison”. We suggest you to run the simulations of equation (2.7a) for various values of the initial numbers of individuals, N0, and for various values of the parameters Rm and K and see, how the systems behaves in response to their changes. You will find out that: the predictions of the number of individuals are always biologically reasonable, i.e., positive, independently of the choice of the initial numbers of the individuals and of the parameters; with increasing Rm, the system predicts first asymptotic approach to equilibrium, then damped stable oscillations, followed by neutrally stable oscillations. For which value of Rm does this occur? Try to calculate this critical value first, before you perform the simulations! When Rm increases beyond this critical value, we can observe unstable oscillations and finally a non-periodic solution, which is known as “chaos” in biological literature. Those interested in the mathematics of this phenomenon are referred to the following papers: R. M. May in Science, vol. 186, p. 645, 1974. R. M. May in Nature, vol. 261, p. 459, 1976. R. M. May and G. F. Oster in the American Naturalist, vol. 110, p. 573, 1976. C. Sparrow in J. Theor. Biol., vol. 83, p. 93, 1980. J.G.P. Gamarra and R.V. Solé in Ecology Letters, vol. 3, p. 114, 2001. (Here one of the most popular data sets in ecology, that of lynx returns, is analyzed in order to look for evidence for a bifurcation process and chaos.) 2.3. Logistic model with delay Finally the concept of time delay is introduced at the end of chapter 2, in equation (2.9b): Nt Nt 1 Rm (1 Nt 2 / K ) Nt 1 , (2.9b) which was then generalized to an arbitrary delay in equation (2.11b): Nt Nt 1 Rm (1 Nt T / K ) Nt 1 . (2.11b) A simulation program in EXCEL that shows the predictions of the numbers of individuals based on equation (2.11b) can be found here. Use this program to test the stability criterion for equation (2.11b): y / x RmT 1 , (2.10) where T is the length of the time delay, and stability is a quality of both Rm and T, which claims that the system is stable, if y/x < 1. We suggest you to run the simulations of equation (2.11b) for various values of the initial numbers of individuals, N0, N1, …, NT-1, and for various values of the parameters Rm and K and see, how the system behaves in response to their changes. 2.4. Continuous models 2.4.1. Introduction Note 2.4 introduced the reader to continuous models. Why to bother with these? We should, because in nature we can also find two clearly distinct groups of organisms. Butterflies can serve as an example of the first one. Their adults can usually be found on wings only during a certain period of the year – say, for instance, in July. These adults mate, reproduce and die. Their offspring undergo egg, larval and pupal stages – and appear next July as new adults. However, because their parents have died the previous year, next July we find a completely different set of adults – none of those from the previous year. We say that such species have non-overlapping generations. In contrast, aphids have overlapping generations. At any instant, we can find all age groups in their colonies: newborn larvae, large larvae, and young as well as senescent adults. Aphids – unlike butterflies – have neither a distinct “reproduction period”, nor a period of “mass death”. We can therefore approximate this situation by assumption that their populations change continuously. Thus while discrete models, which are used throughout the main text of this book, are more appropriate for modeling population dynamics of the former group of organisms, those with non-overlapping generations, represented by butterflies in our simplified example. Continuous models, on the other hand, are more appropriate for modeling population dynamics of organisms with overlapping generations, represented here by aphids. Nothing is absolute here, however. For example, if we take into consideration only the population numbers of aphids in – say – July, and look for a rule that predicts their numbers the next July, discrete models should be used. In any particular situation, therefore, we should carefully consider, which type of model is appropriate in the situation we are interested in. As it was shown in Note 2.4, in continuous models we assume that the numbers of individuals change continuously, not in discrete time instants, as in discrete models. To make it easier for the reader to distinguish, when we speak about discrete and when we speak about continuous models, we will denote the number of individuals in the latter ones by x(t) and will assume that x(t) changes continuously – mathematically, x(t) is a continuous function of time, t. For our convenience, we usually omit “t” in the bracket (mathematically – the argument of the function x) if this does not lead to confusion, and write only x instead of x(t). As a consequence, population growth can be described by a smooth curve. Thus, in continuous models, population growth is described as the change in population size (dx) that occurs during a very small interval of time (dt), and the population growth rate is then dx/dt in this case. The value dx/dt is a convenient mathematical tool, called the first derivative, sometimes also written as x’(t), or only x’, if this omission of the argument t does not lead to confusion. It represents the rate of change in population density within an infinitesimally small interval of time. Population growth is then described by a differential equation, in which the first derivative is always on the left-hand side of the equation and its right-hand side (called RHS below) describes a rule, how the number of individuals changes in dependence on the present number of individuals. Thus, for example, the equation of exponential growth, derived in Note 2.4, is mathematically expressed as dx rx , dt (note that we have replaced N by x and R by r here, to denote we are speaking about a continuous model; as explained in the main text, r log e ( R 1) ). This equation can be understood as: “the rate of change of the number of individuals at time t is proportional to the number of individuals at time t”, or – using corresponding symbols – dx/dt is proportional to x with the constant of proportionality equal to r. 2.4.2. Numerical simulation - Euler’s method The simplest method, how to simulate population growth in continuous models, is known as the Euler’s method (a simulation illustrating the use of this method can be accessed here). The way how the continuous models were derived (see Note 2.4) is used to derive the corresponding formula – the change in population size dx = x(t+dt) – x(t), during an infinitesimally small interval of time, dt, is described by the right-hand side (RHS) of the differential equation: (x(t+dt) – x(t))/dt = RHS. Now, if we make the time increment (the step), dt, very small instead of infinitesimally small, one can expect that the resulting population growth curve will not change dramatically. In other words, if we calculate the population growth as (x(t+h) – x(t))/h = RHS, where the step, h, is small, or – which is the same: x(t+h) = x(t) + h.RHS, we obtain a good approximation of the differential equation model of population growth. Starting from x(0), we can calculate iteratively – by repeated usage of this formula – the values x(h), x(2h), x(3h) ... Note that h = 1 leads us back to discrete models! In this disk, we will use the method just described with the step, h, equal to 0.1, for simulations. You will see on examples that this is a satisfactory approximation of population growth described by continuous models. If the right-hand side of the model predicts large changes of the population during a short interval of time, our simple method may not be accurate enough to make a good prediction of the population growth. In such cases more sophisticated – and more precise – formulas for simulation of continuous models, like Runge-Kutta, predictor-corrector, or a smaller step, h, of the iteration should be used. You will see, however, that this is not necessary in our simple models. For those interested in more sophisticated simulation methods for differential equation models, we recommend the book A First Course in Numerical Analysis: Second Edition by A. Ralston and P. Rabinowicz, published by Dover Publ. Inc., 2003. How does the accuracy of the forecast obtained by the Euler’s method, compared with the exact solution, depend on the step of the simulation, h? To find this out, change the size of the step, h, and watch, how this affects the accuracy of the simulation. The exact analytical solution of the equation in the grey box is in column B and shown as the blue line in the figure, while the prediction obtained from the simulation by the Euler’s method is in column C and shown as the red line in the figure. 2.4.3. Exponential growth In the file dealing with continuous models, which can be accessed here, we present several models of continuous growth. The first one is the continuous analogue of exponential population growth: dx rx . dt It assumes that the population growth rate, dx/dt, is proportional to the population size, x, with the constant of proportionality equal to the population growth rate, r. In other words, in this model, the per capita growth rate, (dx/dt)/x, is constant. This is the case, when for example the per capita birth rate, b (number of newborn individuals per one existing individual per unit time) and the per capita death rate, d (proportion of individuals that die per one unit of time) are constant and independent of the population numbers and age structure and when there is no migration into and/or out of the population studied. In this case, similarly to the discrete analogue explained in detail in the main text, r = b – d. The assumptions of the exponential growth seem to be too restrictive for this to have any practical use, but the contrary is true for three reasons: 1. In real populations, these values change with age of the individual, as older individuals are more likely to die during a unit time interval and they are less fecund. However, it has been shown that the age structure of the population very often tends to stationary values, which means that the proportion of individuals of a certain age in the population approaches an asymptotic value and does not change substantially later (see the book A Primer of Ecology by N. J. Gotelli, published by Sinauer Associates, Inc., Sunderland, Massachusetts, 1998 for a detailed analysis of this phenomenon). In such situations the mean per capita birth and death rates are constant, when averaged over the whole population. 2. The birth and death rates also depend on the population numbers, as the intraspecific competition for food, space etc. between individuals within larger populations is more severe, which usually results in their larger death rates and lower fecundity. However, if the number of individuals in the population is far below the carrying capacity of the environment, K (see the main text for the explanation of this constant), ignorance of intraspecific competition does not have a dramatic effect on the prediction of population numbers in the future. 3. When the number of individuals grows rapidly, then its numbers of births and deaths may be much larger than the number of emigrants and/or emigrants per unit time. In such case, ignoring migration does not affect the predictions of population numbers dramatically. To summarize: In rapidly growing populations, which are far below their carrying capacity and which consist of a mixture of individuals of different ages, the model of exponential growth may give a good prediction of population numbers in the future. This is the case of many insect pests during their outbreaks, spread of infectious diseases etc. Exponential growth has a great advantage in its simplicity. If this equation is integrated following simple rules of calculus (see the book A Primer of Ecology by N. J. Gotelli, published by Sinauer Associates, Inc., Sunderland, Massachusetts, 1998 for a detailed analysis), the result can be used to project, or predict, the population size: x(t ) x(0).e rt where x(0) is the initial population size, x(t) is the population size at time t, r is the instantaneous growth rate of the population, and e is the base of the natural logarithm (e ~ 2.717). Knowing the starting population size and the intrinsic rate of increase we can use this equation to forecast population size at some later time. Alternatively, we can use the Euler’s simulation procedure for the same purpose. How the accuracy of this method depends on the step of simulation, h, and the instantaneous growth rate, r, is shown here. Using the corresponding EXCEL file in this disk, answer the following questions: How does the forecast of the population numbers depend on the values of the instantaneous growth rate, r, and on the initial number of individuals, x(0)? 2.4.4. Logistic growth When the number of individuals is close to the carrying capacity of the environment, the intraspecific competition cannot be ignored any more and the logistic model is appropriate. The continuous version of the discrete logistic equation from the main text is dx x rx1 , dt K The term in the bracket, 1- x/K, represents intraspecific competition. When the population consists of only few individuals, i.e., when x is small, then this term is close to one and the growth is approximately exponential: dx/dt ~ rx. The larger the number of individuals, the smaller is this term, and therefore the smaller is the growth rate of the population. When the number of individuals is equal to the carrying capacity, K, the number of individuals does not change in the model, as dx/dt is zero in this case. We say that the carrying capacity, K, is the equilibrium, or steady state, of the system. In the continuous, differential equation systems, which we will use in this book and CD, the equilibrium can be easily calculated by setting the right-hand side of the model equation equal to zero and solving it for x – the population size. When the number of individuals is larger than the carrying capacity, K, population growth rate is negative and vice versa. Thus, in the language of the main text, its state variable, x, tends to return to or towards the steady state following an environmental disturbance, and this steady state is stable (see the section 1.6. of the main text). Simulation of this equation using the Euler’s method can be found here. If this equation is integrated following simple rules of calculus, we obtain Kx0 e rt . x(t ) K x0 (e rt 1) where x(0) is the initial population size, x(t) is the population size at time t, r is the instantaneous growth rate of the population, K is the carrying capacity of the environment, and e is the base of the natural logarithm. Try to change model parameters in the EXCEL file here. Note that the number of individuals approaches the carrying capacity, K, as time proceeds, that this approach is quicker, if r is larger, and that population grows at the maximum speed when x(t) =K/2. 2.4.5. Logistic growth with delay The feedback term in the bracket in the logistic equation, 1- x/K, represents intraspecific competition. In some systems, this feedback loop may be delayed. For example, large density of aphids on a plant may result in deterioration of plant food quality and consequently lower food (phloem sap) intake of the adults, which results in lowering the rate of development of the embryos inside the mothers and finally in lower fecundity of adult aphids some time later. The net result is that the feedback term in the bracket does not depend on the present density of aphids, x(t), but on their density some time before, x(t - T). The equation is then modified to: dx(t ) x(t T ) rx (t )1 . dt K In order to avoid confusion, we explicitly write the arguments here – i.e., we write x(t) instead of x, as before. Simulation of this equation can be found here. Similarly to the analogous discrete logistic equation with delay, increasing delay and/or increasing instantaneous population growth rate, r, results first in damped oscillations, limit cycles, and then oscillations with increasing amplitude, which finally lead to extinction of the population. Test this claim by manipulating the model parameters in the yellow-highlighted region. 2.4.6. Allee effect In the models of exponential and logistic growth, the growth rate is the largest when the population is small. This is not always true. For example, animals hunting in groups need a certain minimum size of the population (equal to the minimum size of the group), for the population not to go extinct, as too small group size might not be sufficient to overdue and kill a large prey (wolves and deer may serve as an example). In other situations, too small population density might lead to problems in finding the mating partner. Then the population growth rate might become even negative, when the population size is small. Note that the plot of the population growth rate, x’(t), against its size, x(t), is a parabola, and the right-hand side of the corresponding equation is a x rx 2 quadratic function, rx1 rx . Quite logically then (at least for those who know, how the K K plots of polynomial functions – quadratic, cubic and higher-order polynomials – look like), an appropriate function to model Allee effect is a cubic function: dx x x rx1 1 , dt K C where 0 < C < K. As said in section 2.4.3., the equilibrium can be calculated by setting the righthand side of this equation equal to zero and solving it for x – the population size. We obtain three equilibria: x = 0, x = C, and x = K. Check the EXCEL model of this system here to check that the first and last of these, 0 and K, are stable, while that in between these, C, is unstable. In other words, the number of individuals declines, if the population size is smaller C or larger than K, while it increases when the population size is between C and K. This is exactly, what the Allee effect predicts. 2.4.7. Deterioration of environment The feedback term in the bracket in the logistic equation, 1- x/K, is dependent on the instantaneous number of individuals in the population. In certain systems, however, fecundity may be reduced and mortality enhanced by the amount of waste material or excrements produced by the population during its whole existence (human beings might possibly be one example), by its cumulative effect on the host populations (for example, aphids affect negatively their host plants, and this effect seems to be proportional to their cumulative density), or simply may be proportional to the cumulative density of the population, rather than to its instantaneous size. In such situations, the regulatory, or feedback term should be dependent on the cumulative density of the population, the number of “individual-days”, or – mathematically – on the integral of the population numbers t from time zero to t, i.e.: x( )d . The resulting equation is then 0 t dx rx.1 a. x( )d , dt 0 where a is a scaling constant. Simulation of this equation can be found here. Try to change model parameters in the EXCEL file here. Note that the number of individuals initially increases, then declines and eventually approaches zero, independently of the parameter values. How does the size and timing of the peak depend on the parameters r, a, and on the initial size of the population? How does the asymptote (the maximum value, to which the curve is approaching, when time increases infinitely) of the integral (biologically: “total deterioration”, or “cumulative density”) in the lower figure depend on the parameters r, a, and on the initial size of the population? Which biological conclusions can you draw from the answers to the points above? Assume some organisms competing for a limiting resource (nitrogen in soil, phloem sap, host organisms etc.). Assume that each individual takes a units of the limiting resource per unit time. The total amount of the resource increases exponentially due to extrinsic effects - for example, the host organisms increase exponentially. Modify the model of deterioration of environment accordingly and see, whether the population would also go extinct under these assumptions. 2.4.8. Harvesting Harvesting of the population (fishing, selective logging of trees in a forest, shooting or poaching of wild animals for meat, fur or other parts of the body etc.) may dramatically affect its numbers. There is a whole array of possible models dealing with this situation. Here we will present only two simplest cases – constant harvesting of a logistically growing population and the “biocontrol” case, which will be explained in the next section. The reader is invited to modify this basic model for other situations (exponential growth, logistic growth with delay, non-constant harvesting etc.). The model for constant harvesting of a logistically growing population is – as the reader, now familiar with several other models and their construction probably expects: dx x rx.1 H , dt K where H is the number of harvested individuals per unit time. The easiest way, how to analyze the behavior of this model, is to look at the plot of dx/dt against x. When H is zero, this plot is a curveddown parabola. As H increases, this parabola moves by the value H downwards, as the reader can check and see in the EXCEL file here. The parabola then has two intercepts with the x-axis, defined by setting the right-hand side of this equation equal to zero and solving for x. The solutions are: H . x1, 2 K2 1 1 4 Kr As we have seen above, these points are equilibria of the system, i.e., the points for which dx/dt = 0, or – in other words – the population size does not change. The reader can verify in the EXCEL file that these equilibria approach each other, as H increases. The plot of the number of individuals against time shows that between these two equilibria, the population numbers increase (as the values of dx/dt are positive – see also the plot of dx/dt against x), while when the number of individuals is larger than the larger equilibrium, or smaller than the smaller equilibrium, the numbers of individuals decrease with time. Note one important biological consequence. If the population is initially, before the harvesting begins, at its carrying capacity, and the level of harvesting increases slowly, then the population attains a smaller and smaller equilibrium (the larger of the two), that one defined by H . x1 K2 1 1 4 Kr H becomes negative, there is no more Kr equilibrium of the system, as the square root in the corresponding equation does not have a real value, the whole parabola in the plot of dx/dt against x is below the x-axis, and therefore the numbers of individuals decrease continuously – the population goes extinct eventually. Mathematically, this phenomenon is called a bifurcation, and – more importantly for us – biologically this means that with increasing harvesting the population numbers will initially not respond too dramatically, but when a certain level of harvesting is exceeded, the whole population suddenly collapses and goes extinct – with obvious conservation consequences. However, when the value of H exceeds Kr/4, then 1 4 2.4.9. Biocontrol Instead of constant harvesting, one can imagine a situation, when the amount of harvested individuals is proportional to the present number of individuals – the more individuals are present in the population, the easier it is to catch them and the more of them can be harvested. Besides human harvesting, this may be the situation of a biocontrol agent (predator or parasite) released with the aim to control a pest species. The rate of catch of its victim is likely to be proportional to the density of the victim – to the likelihood that the predator will find it. In this case, the harvesting equation can be modified in this way: dx x rx.1 Hx dt K The equilibria of this system can again be calculated by equating the right-hand side of this equation to zero (thus assuming that the rate of change of the population, dx/dt is equal to zero) and solving for x; the solution is then: x1 0 H x2 K 1 r and the corresponding simulation can be found here. Chapter 4: Interactions between two species In the main text, the graphical methods developed in Chapter 3 are applied to the analysis of interactions between two species, including mutualistic, competitive, and predator-prey systems, in Chapter 4. We present discrete models in the main text, while in this disk we show simulations of interactions between two species for continuous systems. Note, however, that because of the Euler’s method used in the simulations, setting the step equal to 1 takes you back to their discrete analogs. Explore carefully all three sheets in the EXCEL file with simulations of two species systems (accessible here), and see, how changes of initial values (numbers of individuals at time zero) and changes of model parameters affect model predictions. Note that in some cases the model predictions are quite wild, or the numbers of individuals vanish quickly. This is because the two species models are more complicated, such is also their behavior and they do not always make a biological sense. One has to employ own intuition to see, where is the problem with the model in such cases. 4.1. Continuous models Generally speaking, almost all two species models we shall deal with can be expressed in the following form: dx1 f1 ( x1 ) x 2 12 ( x1 , x 2 ) dt dx 2 f 2 ( x 2 ) x1 21 ( x1 , x 2 ) dt where xi is the number of individuals (abundance) of species i (i = 1, 2), dxi/dt are the corresponding derivatives (see 2.4.1), fi are functions describing the population growth of species i if it is alone and does not interact with the other species (see chapter 2 for various models of these), and ij are interaction terms, describing the interactions between the two species: they tell you, how the abundance of one species affects changes in the numbers of individuals of the other species. Each of these interaction terms is a function of x1 and x2 – formally: ij = ij(x1, x2), which means that the strength of interaction always depends somehow on the abundances of both species. Note that each of these terms is multiplied by the number of individuals of the interacting species (x2, describing population dynamics of species 2, in the first equation, and x1, describing population dynamics of species 1, in the second equation). Thus, ij describes the interaction strength per one individual of the interacting species. This is quite convenient for the analysis of such systems, as absence of the interacting species causes vanishing of the whole interaction term and the dynamics of the species in question is then governed only by the function fi. E.g., if species 2 is absent, i.e., if x2 = 0, then the second term at the right-hand side of the first equation vanishes and the dynamics of species 1 can be described by dx1 f1 ( x1 ) . dt The three types of models are distinguished by the signs in the equations: in mutualistic systems, both signs are positive (as the abundance of each species positively influences changes in abundance of the other one); in competitive systems, both signs are negative (as the abundance of each species negatively influences changes in abundance of the other one), and in predator-prey systems one sign is positive and one is negative, as the abundance of predators negatively influences the number of individuals of prey (predator eats prey), while prey is eaten, provides food for the predator, and therefore its abundance positively affects changes in predator numbers. Any forms of the fi functions described in the section 2.4. of this text can be used; only in the predator-prey systems it is reasonable to restrict the within-species growth to the negative exponential, as – in the absence of prey – the number of predator individuals necessarily declines due to its starvation, and the most reasonable model for this is exponential decline. In the simulations, we assume logistic growth of each species in the absence of the other one (except of the predator-prey systems, where negative exponential growth is always assumed for the predator and an alternative with exponential growth of prey is also considered). In each of these models, the system dynamic trajectories and the equilibrium lines (often called isoclines) are plotted in the figures, where each axis denotes the number of individuals of one species. The point with coordinates equal to the initial conditions (numbers of individuals of each species at the beginning of the time interval considered) are indicated by a blue point – starting point of the system dynamic trajectory, which then connects all points, representing numbers of individuals of each species at all consecutive instants following the initial one – see also, e.g., Fig. 4.14. of the main text for an example of the system dynamic trajectory. As described in the main text (p. 75), the equilibrium line (isocline) of species i connects all points, in which the numbers of individuals of the two species are such that species i neither grows, nor declines in numbers. You have seen in the main text, how to find equilibrium lines in discrete models. In continuous models, this is even easier: the species corresponding to this isocline neither grows, nor declines in numbers, therefore the derivative of the function describing dynamics of this species is zero anywhere in this isocline – and such must be also the right-hand side of the corresponding model equation, as it is equal to this derivative. Thus the equilibrium lines are obtained as follows: the right-hand side of each of the model equations is set equal to zero, and then the number of individuals of the species 2 is expressed as a function of the number of individuals of species 1. An example of this calculation is given below in the section dealing with mutualistic systems. 4.1.1. Mutualistic (cooperative) systems In cooperative systems, the abundance of each of the species positively affects the changes in numbers of individuals of the other species. We assume the simplest possibility: the strength of this interaction (and therefore the influence of the abundance of one species on the rate of change of individuals of the other species) is proportional to the numbers of individuals of each of the species (so-called Lotka-Volterra interactions, named after the people who first introduced them into models of population dynamics). We assume that the interspecific interactions affect the population carrying capacity. This results in the following system of equations: x1 x1' r1 x1 1 K 1 12 x 2 x2 x 2' r2 x 2 1 K x 2 21 1 Here, x1 and x2 represent the numbers of individuals of species 1 and 2. Each of them is assumed to be growing logistically in the absence of the other species: x x1' r1 x1 1 1 K1 x x 2' r2 x 2 1 2 K2 Presence of species 2 causes an increase of the carrying capacity of species 1, which is proportional to the number of individuals of species 2: 12 x2 (see the term in the denominator of the first model equation), and vice versa. If we set x1' 0 , or – in other words – the right-hand side of the first equation equal to zero, x1 0 , we obtain an expression, which describes all possible pairs of population r1 x1 1 K1 12 x2 sizes (numbers of individuals), for which species 1 does not change in numbers (which is the meaning of x1' 0 ). We can now express x2 as a function of x1: either x1=0 (and then necessarily x K1 x1 0 , or x 2 1 r1 x1 1 , which results in the same. See the reproduction plane 12 K1 12 x2 approach in the main text for a detailed explanation. A similar calculation may be performed for the other species. Thus the isoclines of the systems can be expressed as: x2 x1 K 1 12 x 2 K 2 21 x1 The corresponding simulation can be found here. 4.1.2. Competitive systems In competitive systems, the abundance of each species negatively affects the changes in numbers of individuals of the other species. We assume that – contrary to the mutualistic systems – this performs in the numerator of the model equations. I.e., interspecific competition is expressed at the same place as the intraspecific one – in the numerator of the fraction in the bracket on the right-hand side (see below). This leads to the model x 12 x 2 x1' r1 x1 1 1 K1 x 21 x1 x 2' r2 x 2 1 2 K2 and the model isoclines are x2 K1 x1 12 x2 K 2 21x1 The corresponding simulation can be found here. 4.1.2. Predator-prey systems The general form of the predator-prey model is dx1 f1 ( x1 ) x 2 g ( x1 , x 2 ) dt dx 2 f1 ( x 2 ) x 2 h( x1 , x 2 ) dt The simulation program, which can be accessed here, allows you to choose between two types of population growth of the prey (function f), exponential and logistic, four types of functional response (function g) – Lotka-Volterra and the three Holling-type responses depicted in Fig. 4.19 of the main text, and two types of numerical response (function h) – constant conversion efficiency of prey to predator biomass and logistic predator growth. Thus the possibilities are: Prey growth curve: 1. Exponential f1(x1) = rx1 2. Logistic f1(x1) = r x1.(1- x1/K) We use exponential growth in situations, when intraspecific competition due to large density of individuals can be ignored. Predator growth curve: 1. Negative exponential f2(x2) = -dx2 We always assume that predator will exponentially decline in numbers due to starvation, if there is no prey available. Functional response: 0. Lotka-Volterra g(x1,x2) = ax1 1. Holling type I g(x1,x2) = ax1, if x1 < S/a g(x1,x2) = S, if x1 > S/a 2. Holling type II g(x1,x2) = aSx1 ax1 1 3. Holling type III g(x1,x2) = aSx12 ax12 1 The Lotka-Volterra type of functional response assumes that there is no restriction on the number of prey individuals that the predator is able to eat per unit time. This is of course unrealistic, as the predator needs some time for handling the prey: catching, killing and eating it. This type can be a reasonable approximation of reality in cases, when prey never becomes abundant and when the predator spends majority of time by searching for the prey. In such case, the assumption of linear dependence between the number of prey available and the number of prey eaten (i.e., that doubling of the number of prey available results in doubling of the number of prey eaten) may be realistic and for the simplicity of the corresponding equation even preferable, as it leads to simpler and therefore more easily tractable models. In the Holling type I functional response it is assumed that the number of prey attacked increases linearly with prey density and then suddenly stops when the predators are satiated. This type of response seems to be rather rare in nature, but may be characteristic of some filter feeders, which spend little or no time pausing after each prey is captured; that is, they do not need to stop hunting in order to kill and devour their prey. On the other hand, the type II response is typical of predators that pause after each prey is captured and, therefore, their rate of attack declines as the density of their prey increases. This type of response seems to be typical of many invertebrate predators, but it should be noted, however, that most of the data come from laboratory experiments. Type III functional responses are characteristic of predators that attack their prey at an increasing rate as prey density rises, but then the rate of attack declines as handling time becomes a factor in determining how fast prey can be caught. It is generally thought that type III responses are typical of general predators, particularly vertebrates, which switch their attack to a particular prey species when it becomes more abundant; that is, they learn to look for, or develop a “searching image” of, the more abundant species in their repertoire of prey. However, these responses have also been found in some insect parasitoids, and they may be more common in nature than was previously supposed. Numerical response: 1. Constant conversion efficiency h(x1,x2) = q. g(x1,x2) 2. Logistic predator growth x h(x1,x2) = Zx 2 1 2 x1 The assumption of constant conversion efficiency means that the predator converts a certain, constant proportion q of the prey biomass eaten into its own biomass. The assumption of logistic predator growth means that the predator grows logistically and that its carrying capacity is proportional to the number of prey available (and therefore x1 is in the denominator of the corresponding equation). Epilogue After having been through all the simulation exercises above, we are sure you now are able to develop your own models of population dynamics.