3. Model development and Model examples 3.1 Fundamental equations of the system, restriction to the essential information To develop a model we need basic equations of the system under consideration. These laws often contain processes which are not relevant and can be neglected. Example for constructing a simple ocean model The hydrodynamic equations for an incompressible fluid like the ocean could be written as follows v Velocity vector ρ Mass density T Temperature 1 ∂ 2 v p Pressure 2 = − ∇p − gk + ν h ∇ h v + ν v 2 , νh Eddy viscosity (horizontal) ∂z ρ νv Eddy viscosity (vertical) α Thermal expansion coefficient Continuity equation ∇ ⋅ v = 0 , κ h Eddy conductivity (horizontal) ρ = ρ 0 (1 + α (T − T0 )), Equation of state κ v Eddy conductivity (vertical) g Gravity acceleration 2 ∂T ∂ T 2 f Coriolis-parameter Temperatur e equation + v ⋅ ∇T = κ h ∇ h T + κ v 2 , k Vertical unit vector ∂t ∂z ∂v + ∇ ⋅ ( vv ) + fk × v Momentum equation ∂t CM-98 We can, however, discard the equation of state and the temperature equation when the density variations within the fluid are nearly negligible due to small temperature contrasts. Then, we end up with the equations ∂v 1 ∂2v 2 + ∇ ⋅ ( vv ) + fk × v = − ∇p − gk + ν h ∇ h v + ν v 2 , ∇ ⋅ v = 0 , ρ ∂t ∂z in which ρ = const. We can simplify the system further if the geometrical properties of the flow are special. These are often idealizations which may help to better understand the dynamics. Example Wind driven ocean model in a rectangular basin (SpOzMo, Julia Dellnitz 2000). The fluid is bounded by rigid planes at z=-D and z=0 (simple model for an ocean with depth D). Then the vertical integration of the continuity equation yields ( v h denotes horizontal velocity vector) D ∇ ⋅ ∫ v h dz = − w z = 0 + w z = − D = 0 , 0 since vertical velocity w vanishes at the boundaries. CM-99 Figure: Sketch of the model domain Therefore, the vertically averaged horizontal flow vh is nondivergent and we can describe the vertically averaged horizontal flow by a streamfunction ψ so that ∂ψ ∂ψ vˆ = − , ,0 ∂y ∂x 0 1 with vˆ = v h dz. ∫ D −D Vertically averaging of the horizontal momentum equation leads to (note that ∇ h = ∇ in 2D) 0 1 0 1 τ − τb ∂vˆ 1 + ∇ ⋅ ∫ v h v h dz + (wv h z = 0 − wv h z =. H ) + fk × vˆ = − ∇ ∫ pdz + ν h ∇ 2 vˆ + t , ∂t D 444 ρD − D ρD D −D 1 424444 3 =0 where τ t = ν v ∂v h / ∂z z = 0 denotes the horizontal stress at the top and τ b = −ν v ∂v h / ∂z z = − H the horizontal stress at the bottom of the ocean. The horizontal momentum flux vhvh is usually small in ocean dynamics. A compromise is to neglect only vertical anomalies so that the equation becomes 0 τt − τb ∂vˆ 1 2 + ∇ ⋅ ( vˆ vˆ ) + fk × vˆ = − ∇ ∫ pdz + ν h ∇ vˆ + , ∂t ρD − D ρD CM-100 So, the components of the momentum equation read ∂ ∂ψ I − ∂t ∂y ∂ ∂ψ II ∂t ∂x ∂ψ ˆ + ∇ ⋅ − v ∂y + ∇ ⋅ vˆ ∂ψ ∂x 0 ∂ψ ∂ 1 ∂ψ 2 ν − f pdz = − + ∇ − ρD ∫ h ∂y x x ∂ ∂ −D 0 − f ∂ψ = − ∂ 1 2 ∂ψ ν pdz + ∇ vˆ h ∫ ∂y ∂y ρ D − D ∂x τ tx − τ bx , + ρD + τ ty − τ by . ρD Applying ∂/∂x(II)- ∂/∂y(I) gives the so-called vorticity equation ∇ ⋅ [k × ( τ t − τ b )] ∂ 2 ∂ψ ∂∇ 2ψ ∂ψ ∂∇ 2ψ ∂ψ (∇ ψ ) − + +β = ν h ∇ 2 (∇ 2ψ ) − . ∂t ∂y ∂x ∂x ∂y ∂x ρD where β=∂f/∂y is the so-called Beta-parameter. Consequently, we have simplified the original system with four unknown three-dimensional fields to a system with only one two-dimensional field, namely, the streamfunction ψ = ψ ( x, y , t ). CM-101 3.2 Low order model formulation (Galerkin method) For a low order model we represent the model fields in terms of suitable orthogonal functions, e.g. ∞ ψ = ∑ ψ k (t ) f k ( x, y ). k =1 This can be used to convert the model equation, e. g. ∂ψ = Fψ (ψ ). ∂t to an infinite set of ordinary differential equations by the operation ∫∫ ..... f dxdy ≡ k f k .. A This would lead due to the orthogonality (i.e. < fj | fk >=δjk) to the following dynamical system dψ k = f k Fψ dt for k = 1,....., ∞ Of course we are not able to solve an infinite number of equations. Therefore, we must truncate the expansion at a finite k. For dynamical system analysis it is appropriate to choose rather a low value of truncation number k. CM-102 Application to the wind-driven ocean model: Nonlinear vorticity advection Planetary vorticity advection Bottom friction ∂ 2 (∇ ψ ) + ∂ψ ∂∇ ψ − ∂ψ ∂∇ ψ + β ∂ψ ∂t ∂x ∂y ∂y ∂x ∂x 2 Wind stress 2 δ 1 = ∇ ⋅ (k × τ t ) − E ∇ 2ψ + ν h ∇ 4ψ , 2D ρD Horizontal Diffusion where D is the depth of the ocean, β the planetary vorticity gradient (meridional derivative of Coriolis-parameter), τt the wind stress vector and δE the bottom friction parameter. The stream function has to fulfil free-slip boundary conditions: ψ = 0 at x = 0 , L x and y = 0 ,L y (no flow across the boundary) ∇ 2ψ = 0 at x = 0 ,L x and y = 0 ,L y (no drag at the boundary) CM-103 With the Fourier expansion πk ψ = ∑ ∑ψ k ,l (t ) sin l =1 k =1 Lx Nl N k πl x sin y , Ly the boundary conditions are automatically fulfilled. Applying the aforementioned technique leads to a dynamical system with Nk×Nl ordinary differential equations for a given wind stress. Truncation at Nl=2 and Nk=2, neglect of bottom friction and a windstress field of the form πk 1 − ∇ ⋅ (k × τ t ) ∝ sin ρD Lx πl x sin y Ly lead to the Veronis model (Veronis 1963) In this model γ denotes the γ dA 4 dB 8 =− B − εA + , = A + AC − εB amplitude of windstress and ε dt 3π 2 dt 15π a viscosity parameter. dC 8 dD 1 =− D − AB − εC , = C − εD . dt 15π dt 3π CM-104 Some results from numerical simulations at higher truncations (from Dellnitz 2000) Streamfunction of the steady state Bifurcation Diagram (Re Reynolds number) CM-105 3.3 Box model formulation A box model predicts only integrated quantities. Therefore, one shifts from local budget equations to global budget equations. The disadvantage of this method is the inaccurate description of the impact of local anomalies on the global budget. In most cases it has to be described in an empirical way. a) The zero dimensional energy balance model (EBM) The simplest models in climate physics are the so-called zero-dimensional climate models. These are indeed box-models in which the complete atmosphere-ocean system has been aggregated into one box. Such a box model is used to predict the mean temperature T of the climate system and is able to describe ice-albedo feedbacks. Absorbed radiation Ri T T Emitted radiation Ro Ro = ε (T )sT 4 Ri = Q0 [1 − A(T )] Figure: Schematic illustration of a zero-dimensional climate model (adapted from Fraedrich 2001) CM-106 The governing equation of the EBM is C [ ( )] ( ) dT = Q0 1 − A T − ε T sT 4 dt Notation: Q0 =342W/m2 s =5.67×10-8W/m2/K4 A ε T C Global mean solar radiative input (present day value) Stefan-Boltzmann constant Planetary albedo Greyness of the system Annually averaged temperature of the global earth-atmosphere system Heat capacity of the system The equation expresses the approximate radiation balance between absorbed radiation Ri = Q0 [1 − A(T )] and emitted radiation Ro = ε (T )sT 4 = R ↑ (0) CM-107 Any slight imbalance between Ri and Ro leads to a change in the temperature of the system. In the simple model the albedo A and emissivity ε have to be expressed (parameterized) as a function of the temperature T . It is plausible that the change of ice influences the albedo. This leads to the ice albedo feedback (Budyko and Sellers). A simple parameterization is given by Amax 2 Tu − T 2 A(T ) = Amin + ( Amax − Amin ) 2 2 T − T u l Amin for T < Tl for Tl ≤ T ≤ Tu for T > Tu where Amax is the maximum albedo (snow ball earth), Amin the minimum albedo (ice free planet), Tl the lower temperature threshold and Tu the upper temperature threshold. The realistic temperature range is between Tl and Tu so that one may only wish to consider the second case in the albedo formula, namely A( T )=A0-b T 2 (Fraedrich 1978). The greyness of the system depends upon the greenhouse effect. With increasing temperature the water vapour amount increases and, therefore, also the greenhouse effect. This reduces the emissivity of the system since a bigger part of emitted surface radiation heats the atmosphere that emits radiation back to the surface. A simple parameterization (Fraedrich 1978) is given by ε (T ) = ε 0 - κT 2 CM-108 The graph shows the dependency of ε and A on temperature T . We use the following parameters according to Fraedrich (1978): A0=1.2, b=10-5 1/K2, ε0=0.87 and κ =3x10-6 1/K2. For a realistic temperature range (250K< T <320K) albedo and greyness have physically valid values. Present day values (assuming T = 288.6K) are: A=0.284 and ε=0.62. CM-109 Fixed albedo and fixed greyness For fixed present day albedo and greyness the analysis is simple. Then, the dynamical system reduces to: C dT = Q0 [1 − A] − εsT 4 dt The equilibrium becomes Te = 4 Q0 [1 − A] εs Therefore, no bifurcation occurs. The linearization gives C Attractor dT ′ = −4εsTe 3T ′ dt The stability of the equilibrium is obvious since the growth rate σ = −4εsTe 3 /C is negative. CM-110 Figure: Heating as a function of temperature: Net heating (red line), absorbed radiation (green line) and emitted radiation (blue line). Fixed greyness (only ice-albedo feedback) For fixed greyness the system equation becomes C dT = Q0 [1 − A0 + bT 2 ] − εsT 4 dt Now, the equilibria result from the biquadratic equation bQ0 2 Q0 [ A0 − 1] = 0 T − T + εs εs 4 and the physically valid equilibria become Repellor Attractor 2 Te1, 2 bQ0 bQ0 Q0 = ± [ A0 − 1] − 2εs εs 2εs Since A0>1 no solution occurs for 4εs Q0 < 2 [ A0 − 1] b Figure: Heating as a function of temperature: Net heating (red line), absorbed radiation (green line) and emitted radiation (blue line). CM-111 At the critical solar input we obtain for the equilibrium temperature and the albedo Tec = bQ0 A −1 = 2 0 = 200K 2εs b , Ac = A0 − bTec = 2 − A0 = 0.8 2 The critical albedo is quite large but possibly below Amax (snowball earth). Note that with the full albedo formula we possibly have two additional equilibria: Tes = 4 Q0 [1 − Amax ] for Tes < εs A0 − Amax b (Snowball Earth) Teh = 4 Q0 [1 − Amin ] for Teh > εs A0 − Amin b (Ice free planet) and The first solution represents a snowball earth and the other a warm planet without any ice and snow. These additional equilibria are stable if they exist. CM-112 Linearizing the governing equation yields dT ′ C = (2Q0 bTe − 4εsTe 3 )T ′ dt Inserting the equilibrium temperature gives the growth rate 2 2 bQ T bQ Te 0 e 0 σ = 2Q0 b − 2bQ0 m 4 − εsQ0 [ A0 − 1] = m4 − εsQ0 [ A0 − 1] C 2 2 C Obviously, Te1 is stable and Te 2 is unstable. Therefore, the bifurcation at 200K is a saddle node bifurcation. The figure on the right displays the equilibria for the model that takes a maximum albedo (Amax=0.9) and a minimum albedo (Amin=0.1) into account. Then, a further saddle node bifurcation includes a new branch for a stable snowball earth solution. However, the snowball earth cannot exist for Q0>εs(A0-Amax)2/(1-Amax)/b2=316.4W/m2. Therefore, this Figure: Equilibria of the conceptual climate model state is impossible for present day conditions. model including the ice-albedo feedback. CM-113 Fixed albedo (only water vapor feedback) For fixed albedo the system equation becomes C dT = Q0 [1 − A] − (ε 0 − κT 2 )sT 4 dt Now, the equilibria result from the sixth-order polynomial equation ε 0 4 Q0 T − T + [1 − A] = 0 κ κs 6 Attractor Repellor Whether any two positive equilibria exist depends on the 6 4 minimum of the curve f (T ) = T − ε 0 / κT that is given by ε 2ε 2ε 1 2ε f (Tmin ) = 0 − 0 0 = − 0 at Tmin = 23 κ 3 κ κ 3 κ 3 2 3 2 ε0 3κ Two solutions occur for f (Tmin ) < −Q0 (1 − A) / κs . This yields κs 2 ε 0 = 858W/m2 Q0 < 2(1 − A) 3 κ 3 Figure: Heating as a function of temperature: Net heating (red line), absorbed radiation (green line) and emitted radiation (blue line). CM-114 Linearizing this equation gives dT ′ C = (6κsT 5 − 4ε 0 Q0 sT 3 ) T ′ dt The growth rate increases without limit at high temperatures leading to the so-called runaway greenhouse effect (the ocean evaporates). It has to be noted that the greyness cannot become negative but already zero greyness would lead to a 2 runaway greenhouse effect. Therefore, ε 0 > κT must be fulfilled for a stable climate. However, this is the case for T < 538.5K and the equilibrium temperatures are within this range. It is obvious from the figure showing the heating terms (previous slide) that the upper equilibrium is unstable and the lower one stable. Figure: Equilibria of the conceptual climate model including water vapour feedback. CM-115 Full model (ice-albedo and water vapor feedback) The system equation including temperature-dependent albedo and greyness becomes C dT = Q0 [1 − A0 + bT 2 ] − (ε 0 − κT 2 )sT 4 dt In the full model up to three equilibria having positive temperatures can arise. However, taking the physically necessary upper and lower limits of albedo into account even up to four equilibria can be found. The stability of these equilibria can be determined by calculating the derivative of the right hand side of the system equation with respect to temperature. The figure on the right shows the equilibria of the full model. In the model a transition of the present day climate to a snowball earth is more likely than the transition to a runaway greenhouse. However, the simple model may include incorrect parameter assumptions and exclude many processes that may counteract a climate changes (e.g. clouds). Therefore, the model should rather be considered as a roughand-ready theory for climate dynamics without making a statement for real changes in the future. For the latter purpose complex climate models are appropriate. CM-116 Runaway greenhouse Icefree Planet Snowball Earth Moderate Climate b) Stommels two-box-Model of the meridional overturning circulation The ocean is divided into a polar and an equatorial box. The model predicts the temperature and salinity average of each box (see figure adapted from Dijkstra 2005). heat flux CT (Tea-Te) fresh water flux CS (Sea-Se) heat flux CT (Tpa-Tp) fresh water flux CS (Spa-Sp) volume flux Ψ~αT(Te-Tp)+ αS(Sp-Se) temperature Te temperature Tp salinity Se salinity Sp volume V volume V CM-117 We obtain the following governing equations of the system V dT p = CT (T p − T p ) + Ψ (Te − T p ) , a dt dT a V e = CT (Te − Te ) + Ψ (T p − Te ), dt dS p a = CS (S p − S p ) + Ψ (Se − S p ) , V dt dS e a V = Cs (Se − Se ) + Ψ (S p − Se ) dt where CT and CS are relaxation constants and the volume flux Ψ is proportional to the density difference ∆ρ between the boxes Ψ =γ ∆ρ ρ0 [ ] = γ α T (Te − T p ) + α S ( S p − S e ) , where αT and αS are the thermal expansion and haline contraction coefficients, respectively. CM-118 Therefore, we can reduce the system to two equations for the temperature difference ∆T=Te-TP and salinity difference ∆S=Se-SP between the boxes: d∆T = C 'T ( ∆T a − ∆T ) − γ ' α T ∆T − α S ∆S ∆T , dt d∆S = C ' S ( ∆S a − ∆S ) − γ ' α T ∆T − α S ∆S ∆S , dt where C´T=C T /V, C´S=C S /V and γ'=2γ/V. Nondimensionalization (see Dijkstra 2005) leads to the following dynamical system: dT = η1 − T (1 + T − S ) ~ dt dS II ~ = η2 -S (η3 + T − S ) dt I where ~ t = C 'T t , T = (γ 'α T / C 'T )∆T , S = (γ ' α T / C 'T )∆S η1 = (γ 'α T / C 'T )∆T a , η 2 = (γ 'α T / C 'T )∆S a , η 3 = C 'S / C 'T CM-119 Equilibria From I and II we obtain 0 = η1 − Te (1 + Te − S e ) ⇔ Te = η1 1 + Te − S e 0 = η 2 − S e (η 3 + Te − S e ) ⇔ S e = η2 η 3 + Te − S e Combining both leads to Te − S e = η1 1 + Te − S e - η2 η 3 + Te − S e Setting η3=1 (relaxation time scales for heat and salinity are identical) we get the equilibrium for the nondimensional overturning mass flux − 1 / 2 + 1 / 4 + η1 − η 2 for η1 > η 2 TH state (sinking in the north) 1 / 2 − 1 / 4 + η 2 − η1 for η1 < η 2 SA state (sinking in the south) ψ~e = Te − S e = CM-120 We see that the sign of the overturning circulation changes sign when η1= η2. In this case the thermal expansion and the saline compression are compensating each other. ↑ ψ η1 -η2→ Figure: Overturning strength in Stommels box model as a function of χ-η2 (for η3= 1). CM-121 Equilibria T, S η1=0.25 η3=0.3 T, S η1=0.25 η3=3.0 η2 η2 TH+SA SA η2 TH η1 Figure: Temperature difference (T) and salinity difference (S) in Stommels box model as a function of η2 for a) η1=0.25,η3= 0.3 and b) η1=0.25,η3= 3.0. c) displays a regime diagram as a function of η1 and η2 for η3= 0.3. Note that multiple equilibria arise due to the imperfection η3≠1. (adapted from Dijkstra 2005). CM-122 3.4 Population dynamics: The Malthusian model Malthus developed a theory for population growth in 1798. It is also a theory for economic growth in the pre-industrial era. The evolution of the human population N (number of humans) is governed by the birth rate B and death rate D. Population growth requires that the birth rate is larger than the death rate. The governing equation becomes dN = N [B − D ) ] dt To obtain a simple model the rates B and D should be parameterized as a function of N itself. They could also depend upon other variables but these are prescribed as exogenous parameters in the present model. The Malthusian model assumes that the birth rate B is independent of the population. In modern societies this is possibly not realistic as, e.g., the birth rate decreases in urban areas with large populations. However, it should be acceptable in the pre-industrial era when family planning was still not established. The death rate D is on the other hand a decreasing function of wealth (consumption C per capita) so that B = const., D = D (C / N ) , dD <0 d (C / N ) CM-123 A parameterization for the death rate like in the figure below is given by cD D= C/N Population decay Population growth Figure: Relative population tendency, birth rate B and death rate D as a function of per capita consumption. CM-124 To solve the problem one has to parameterize the consumption as a function of population. In Malthus model it is assumed that a fixed area of land A exists and that N labourers cultivate the area. The yield Y per time unit of this cultivation is parameterized by the following production function N Y ( A, N ) = c A A arctan N0 where cA is a technology parameter and N0 a reference population. The yield Y increases with population but the additional yield per additional labourer decreases with N. This law of diminishing returns finds its reason in the limited amount of agricultural area. The maximum possible yield is Y=cA Aπ/2. Then, additional labourers cannot increase the yield anymore since all land areas are cultivated and optimally utilized. When the market is cleared, that is, the yield of the cultivation is immediately sold and consumed, we can equate production Y with consumption C N C = Y ( A, N ) = c A A arctan N0 CM-125 C=cA Aπ/2 Figure: Consumption as a function of population. The diminishing returns lead to a decrease of consumption per capita with increasing population. Using these parameterizations we obtain the following differential equation for population growth cD N dN = N B − dt N c A arctan A N0 CM-126 For large N>>N0 this equation approaches the logistic equation dN N = σ N N 1 − dt N m where σN=B and Nm=cAπ/(2cD). In this case the maximum population is N= Nm . However, the equilibrium is possibly already attained at a smaller value. The equilibrium is stable because the linearized equation becomes dC dN dN ' dD = B − De − N e dt d (C / N ) N e Ne 123 Ne =0 dD dC dN N e N e − C e = − Ne d (C / N ) N e Ce − 2 Ne N ' N ' Since C(0)=0 and d 2C/ dN 2<0 (diminishing returns) we have dC dN N e − Ce < 0 Ne and using the assumption dD/d(C/N) <0 yields a negative growth rate. CM-127 This becomes also clear by looking at the tendencies in a diagram Attractor The most important message of the Malthusian model is that the equilibrium is associated with the relation B = D (C / N ) ⇒ C / N = const . at the equilibrium state. Therefore, increase of land area A (e.g. by discovering new land) or technological progress (increase of cA) does not eventually lead to a higher consumption per capita C/N (wealth)! Therefore, the wealth cannot be elevated and the humans stay poor since the population growth compensates the higher food production. CM-128 3.5 Macroeconomic approach in economics In economics an approach analogous to the box model formulation exists. We may understand an economic system as a system of many households, firms and banks. A description of the financial budget for each household or firm is gained by so-called microeconomic models. On the other hand macroeconomic models only describe the budget of many households, firms and banks. A simple common approach divides the complete (national) economy into one aggregate for households, one for firms and one for banks. This gives such a picture: Notations Households H Interest iB Profit Π Wage wL Consumption C Firms F Saving S Investment I Banks B Figure: Monetary circuit in a simple macroeconomic model CM-129 H F B S I C Π K L w i Households Firms Assets (administrated by the bank) Saving Investment Consumption Profit Capital Labour Wage per work unit Interest The monetary circuit of this macroeconomic model can be summarized by three equations: dH = iB + wL + Π − C − S dt dF II = −iB − wL − Π + C + I dt dB III =S dt I d (H + F + B ) = I ≠ 0 ! ⇒ dt Note that the assets B increase monotonically although the savings are transferred in the form of investments I to the firms and, consequently, B also appears in the form of a credit. Therefore, money is not conserved in this economic system. Saving S must become zero if an equilibrium exists. However, this appears illusory as households with large incomes are not able to spend their money only on consumption. Therefore, such an economy is not sustainable in the long term. The model completely covers the growing gap between rich and poor since heterogeneous households are summarized into one aggregate. On the other hand a mathematical proof that the inequality between rich and poor rises in such an economic system has been given in the textbook by Kremer (2012)1. 1 Kremer, J.. 2012: Grundlagen der Ökonomie. Metropolis Verlag, Marburg, 267. The proof is also described in the following article: Kremer, J., 2010: Dynamic Analysis - Investigating the long-term behavior of Economies. (http://www.rheinahrcampus.de/fileadmin/prof_seiten/kremer/DynamicAnalysis.pdf) CM-130 If the firms pay out their profits completely, the tendency for F vanishes (the firm does not accumulate money). This gives the following profit Π = C + I − wL − iB The investment I will of course not directly be paid out in the form of money to the firm-holders. Instead, the investment will be paid to other firms that produce capital goods (production facilities). So, much of the investment will eventually be paid out in the form of wage (capital goods can be viewed as “coagulated work”). Therefore, a part of wL compensates much of the investment I. We already see a problem here. If the assets B increase monotonically, the debt of the firms (-B) will also increase. To avoid a negative profit due to increasing interest payment, the firm has to raise its production and sales. However, firm-holders could avoid further investments because of the increasing interest payments. In such a case saved money will be increasingly extracted from the monetary circuit. This leads to a decrease of income (stemming from the investments) and increasing unemployment that may cause an economic crisis. To circumvent this, the state can borrow saved money in the form of a bond and the borrowed money can be feeded back into the monetary circuit by public spending. This kind of politics was suggested by the economist J.M. Keynes and is called Keynesianism. However, the increasing debt and the accompanying interest payments of the government also lead to problems in the long term (debt crisis). Therefore, Keynesianism also does not make the economic system really sustainable. CM-131 So far, the model includes monetary exchanges without a rule to evaluate them. To determine these exchanges more relations have to be introduced. First, the economy produces goods and its amount per time unit can be measured by the so-called production function: Y = Y ( K , L) Y describes the production of consumer and capital goods. Furthermore, depreciating capital K (production facilities) must be recovered. Therefore, we have the relation Y= dK + δK + C dt where δ is the depreciation rate of capital goods. The households spend a part of their income on consumption and another part will be saved. The saving leads (in this model) to investments, that is, the firms use this money to extend and recover their production facilities. dK S= + δK → Y = S + C dt With the saving ratio s (the fraction of saved money) we obtain dK S = sY = + δK dt CM-132 Therefore, the economy will be governed by the following differential equation: dK = sY ( K , L ) − δK dt This is the so-called neoclassical Solow-Swan model. The so-called Cobb-Douglas production function is widely used and is given by Y = AK α L1−α , where A denotes the level of technology. α is a constant with 0<α <1. This function allows the substitution of labour L by capital K and vice versa. However, an optimal ratio K/L exists depending on the costs for capital (interest and depreciation) and labour (wage). Prescribing the labour L, the level of technology A and the saving ratio s as exogenous constants we obtain the following dynamical system equation dK = sAK α L1−α − δK dt In more elaborate models L, A and s are endogenous variables and will be predicted. CM-133 This equation has two equilibria: K e1 = 0, K e 2 = (sA ) 1 1−α L The first trivial equilibrium is unstable. This instability triggers economic growth. The second equilibrium is stable. One can see this by looking at the following graph sY δK Ke1=0 Ke2 K→ When sY>δK, economic growth takes place and this appears for Ke1<K<Ke2. Therefore, the system changes monotonically towards the state Ke2 away from state Ke1. CM-134 This nontrivial equilibrium state is, however, in disagreement with the monetary cycle. Due to equation III we have (assuming that all saved money will be invested) t t dK B (t ) − B (t 0 ) = ∫ Sdt' = ∫ + δKdt' dt t0 t0 Therefore, the assets of the households (debt of the firms) will increase further although the development of capital K has reached equilibrium. On the other hand the increasing debt leads to increasing capital costs and diminishing profits that eventually causes a collapse of the economic system. The only way to circumvent this is to abolish the interest or to introduce never-ending technological progress that will increase A. In the former case the investor will not receive a return for their investment. Then, it is likely that the investor will retain their money and this also provokes a collapse of the economy. The cost for Real Capital K depreciation could be hided in the price for the goods. Financial Capital B Then however, a steady state must be accompanied with zero savings. This appear unlikely as rich people can hardly spend their complete income on consumption. Technological progress together with inventing new products for consumption remains as the last possibility. Then, the economy can never reach Time t→ equilibrium. This so-called growth imperative clearly Figure: Divergence of real and financial capital disagrees with sustainable development. CM-135 3.6 World-2 model World-2 is a global ecological-economic model that makes a prediction of population growth, economic growth, food production, non-renewable resources and pollution. Figure: Schematic diagram of Forrester’s World-2 model (from Forrester 1971) CM-136 World-2 has five variables: N R Population non-renewable resources P X Pollution Agricultural fraction of capital K Capital These are predicted by the following equations dN = N [c B BS ( S ) BF ( F ) B N ( N ) BP ( P ) − c D DS ( S ) DF ( F ) D N ( N ) DP ( P )] dt dK = c K NK S ( S ) − δK dt dP P K = N PK − dt N τ P (P ) N0 dR =− NR S ( S ) dt R0 dX X F ( F , S ) − X = dt τX where S denotes the material standard of living and F the food ratio CM-137 The auxiliary variables material standard of living S and food ratio F are defined by S= K 1− X S R ( R) N 1− X 0 , K F = FN ( N ) FP ( P ) FX X N The material standard of living S increases with capital per person and amount of nonrenewable resources but it decreases with the agricultural fraction of capital. The food ratio F decreases with population and pollution (FN and FP are monotonically decreasing functions) but increases with the amount of agricultural capital per person. All functions occurring in this system are tabulated. The population N, the capital K, the pollution P have been normalized with dimensional values N0, K0 and P0 so that their values in 1970 are near to one. The non-renewable resources have been normalized with the value R0 occurring in the year 1900. The time is measured in years. The following constants are used in World-2 N0=3.6x109, cB=0.04, K0=3.6x109, cD=0.025, P0=3.6x109, cK=0.05, R0=900x109, δ=0.025, X0=0.3, τX=15 To understand World-2 better it is useful to consider isolated model parts. We begin with the macroeconomic core of the model. CM-138 Economic part of World-2 dK = c K NK S ( S ) − δK dt For simplicity KS is approximated by a linear function of S, namely K S ( S ) = 0.85 S + 0.15 The latter is indeed nearly true in a suitable range of S. Then, the economic equation becomes dK 1− X = c K 0.85 KS R ( R ) + 0.15 N − δK dt 1− X 0 The similarity to the Solow-Swan model is evident when we assume that labor L is proportional to the population N. In this case the equation for capital takes formally the form dK = sY ( K , L, X , R ) − δK dt CM-139 Therefore, the production function now also depends upon the agricultural fraction X of capital and the amount of natural resources R. X decrease and R increase the production. By assuming constant X and R we obtain the following form for the production function Y ( K , L ) = aK + bL This is the so-called von Neumann production function which is characterized by a linear increase of production with capital and labour. Therefore, the impact of substitution of labour by capital is independent of the actual amount of capital. This has important consequences for the solution. For as>δ no equilibrium exists and the solution exhibits unlimited exponential growth. This is indeed the case in World-2 when no resources are depleted (R=R0). δK CM-140 Effect of non-renewable resources on economic growth in World-2 A termination of economic growth takes place when the equation for the non-renewable resources is included. The function RS(S) is approximately given by RS(S)=S. In this case the dynamical system takes the form (assuming X=X0): dK = c K [0.85 KS R ( R ) + 0.15 N ] − δK dt N dR = − 0 KS R ( R ) dt R0 The function SR(R) can be written as (giving a perfect fit with the tabulated data): S R ( R ) = −0.5 cos(π R ) + 0.5 for 0 ≤ R ≤ 1 By setting tendencies to zero we obtain the equilibrium R = 0, K = 0.15c K N / δ This means that in this state humans are maintaining some capital without using nonrenewable resources. This could be a state similar to the preindustrial era. CM-141 Numerical solution Starting values: K(1900)=0.1, R(1900)=1 The economic growth ceases at around 2025 and afterwards economic decay takes place. Note that even in the year 2200 non-renewable resources are available and decay very slowly due to the low amount of capital. CM-142 Effect of population dynamics on economic growth in World-2 The population equation is rather complicated since birth and mortality rates depend upon all model variables. Here, we neglect the impact of pollution P and agricultural fraction X of capital. Then, the equation becomes dN = N [1.02cB BS ( S ) BF ( F ) BN ( N ) − 0.92cD DS ( S ) DF ( F ) DN ( N )] dt Furthermore, the food ratio is approximately given by the following expression K F ( K , N ) = 1.02 0.4 + 0.6 [2.4 − 1.5 arctan(1.35 N )] N The various functions in the population equation can be approximated as follows (4 − F ) 2 BF ( F ) = 2 − , 8 B S ( S ) = −0 .2 S + 1 .2 , B N ( N ) = − 0 . 1N + 1 . 1 , 1 F DS ( S ) = 2.5 exp( −1.5S ) + 0.5 DF ( F ) = D N ( N ) = 0 .2 N + 0 .8 CM-143 The birth rate decreases with food ratio, material standard of living and population. The mortality rate only decreases with food ratio and material standard of living while it increases with population. Neglecting the dependence on material standard of living and population yields a Malthusian population dynamics. This hints at the possibility of multiple equilibria. In the first equilibrium the population should be large with many fasting people and high birth rates. In the second equilibrium the population should be low with a high material standard of living and high food ratio. This could lead to a demographic transition dynamics. However, the model does not reveal this in a convincing way. In the equilibrium the capital K as a function of R and N becomes: Ke = 0.15c K N e 0.15c K N e = δ − 0.85c K S R ( R ) δ − 0.85c K [1 − cos(π R )] / 2 Obviously, no equilibrium is possible for R≥ 2δ arccos1 − = 0.556 π 0.85c K 1 CM-144 Figure: Population of the equilibrium states as a function of R In the model humans can survive even if all non-renewable resources are exhausted. Then, the number becomes 380 million peoples. Multiple equilibria arise only near R=0.55. For R→0.556 the equilibrium population diverges dramatically. CM-145 Simulations with resource extraction and population dynamics Simplified World-2 model (only prediction of N,K and R) Full World-2 model (including pollution and agricultural dynamics) We see that the simplified model already simulates an evolution which is quite similar to that of the full model. Note that the real world population rises much more dramatically after 1970. CM-146 A better agreement with the real world population results when we increase the economic growth parameter from cK=0.05 to cK=0.1 after 1970. Simulation with increased capital growth parameter Comparison of “poverty” index (death rate, probability to die with in 1 year) simulations. In this scenario the capital attains a much larger value and the poverty decreases below the minimum of the standard simulation. However, the poverty increases much faster in the economic decay phase. CM-147