You are on page 1of 12

ARTICLE IN PRESS

Journal of Theoretical Biology ] (]]]]) ]]]]]]

Contents lists available at ScienceDirect

Journal of Theoretical Biology


journal homepage: www.elsevier.com/locate/yjtbi

Unifying constructal theory of tree roots, canopies and forests


A. Bejan a,, S. Lorente b, J. Lee a
a b

Department of Mechanical Engineering and Materials Science, Duke University, Durham, NC 27708-0300, USA Laboratoire Materiaux et Durabilite des Constructions (LMDC), Universite de Toulouse, UPS, INSA, 135 Avenue de Rangueil, F-31077 Toulouse Cedex 04, France

a r t i c l e in fo
Article history: Received 30 October 2007 Received in revised form 14 June 2008 Accepted 27 June 2008 Keywords: Constructal theory Design in nature Roots Trees Forests Leonardos rule Fibonacci sequence Zipf distribution Eiffel Tower

abstract
Here, we show that the most basic features of tree and forest architecture can be put on a unifying theoretical basis, which is provided by the constructal law. Key is the integrative approach to understanding the emergence of designedness in nature. Trees and forests are viewed as integral components (along with dendritic river basins, aerodynamic raindrops, and atmospheric and oceanic circulation) of the much greater global architecture that facilitates the cyclical ow of water in nature (Fig. 1) and the ow of stresses between wind and ground. Theoretical features derived in this paper are: the tapered shape of the root and longitudinally uniform diameter and density of internal ow tubes, the near-conical shape of tree trunks and branches, the proportionality between tree length and wood mass raised to 1/3, the proportionality between total water mass ow rate and tree length, the proportionality between the tree ow conductance and the tree length scale raised to a power between 1 and 2, the existence of forest oor plans that maximize ground-air ow access, the proportionality between the length scale of the tree and its rank raised to a power between 1 and 1/2, and the inverse proportionality between the tree size and number of trees of the same size. This paper further shows that there exists an optimal ratio of leaf volume divided by total tree volume, trees of the same size must have a larger wood volume fraction in windy climates, and larger trees must pack more wood per unit of tree volume than smaller trees. Comparisons with empirical correlations and formulas based on ad hoc models are provided. This theory predicts classical notions such as Leonardos rule, Hubers rule, Zipfs distribution, and the Fibonacci sequence. The difference between modeling (description) and theory (prediction) is brought into evidence. & 2008 Elsevier Ltd. All rights reserved.

1. Introduction Trees are ow architectures that emerge during a complex evolutionary process. The generation of the tree architecture is driven by many competing demands. The tree must catch sunlight, absorb CO2 and put water into the atmosphere, while competing for all these ows with its neighbors. The tree must survive droughts and resist pests. It must adapt, morph and grow toward the open space. The tree must be self-healing, to survive strong winds, ice accumulation on branches and animal damage. It must have the ability to bulk up in places where stresses are higher. It must be able to distribute its stresses as uniformly as possible, so that all its bers work hard toward the continued survival of the mechanical structure. On the background of this complexity in demands and functionality, two demands stand out. The tree must facilitate the ow of water, and must be strong mechanically. The demand to pass water is made abundantly clear by the strong geographical
Corresponding author. Tel.: +1 919 660 5314; fax: +1 919 660 8963.

correlation between the density (and sizes) of trees and the rate of rainfall (Fig. 1). It is also made clear by the dendritic architecture,

E-mail address: abejan@duke.edu (A. Bejan). 0022-5193/$ - see front matter & 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.jtbi.2008.06.026

Fig. 1. The physics phenomenon of generation of ow conguration facilitates the circuit executed by water on the globe. Examples of such ow congurations are aerodynamic droplets, tree-shaped river basins and deltas, vegetation, and all forms of animal mass ow (running, ying, swimming).

Please cite this article as: Bejan, A., et al., Unifying constructal theory of tree roots, canopies and forests. J. Theor. Biol. (2008), doi:10.1016/j.jtbi.2008.06.026

ARTICLE IN PRESS
2 A. Bejan et al. / Journal of Theoretical Biology ] (]]]]) ]]]]]]

Nomenclature a, b a0 A AB AL At AW c1, c2 C CD D Dc Dc,B DL Dt Dt,B F0 h HV It kr ks Kr Kx, Kz L factors, Eqs. (8), (9), (23), (28) and (29) factor, Eq. (25) area (m2) branch cross-section at the trunk (m2) leaf area distal to stem (m2) tree cross-section at x (m2) sapwood cross-section (m2) factors, Eqs. (48) and (49) global ow conductance, Eq. (50) drag coefcient diameter (m) canopy diameter (m) diameter of branch canopy (m) diameter at z L (m) trunk diameter (m) diameter of branch (m) drag force per unit length (N/m) frustum height (m) Huber value area moment of inertia (m4) radial specic conductivity m/(s Pa) stem specic conductivity m/(s Pa) radial permeability (m2/s) longitudinal permeability (m2/s) length (m)

LB LSC m, n m _ m _ mB p P Pg PL Pv P0 Ri sm u uB v V V VT w x, z Xs Xt

m n r

branch length (m) leaf specic conductivity exponents, Eqs. (8), (9) and (23) bending moment (N m) mass ow rate (kg/s) branch mass ow rate (kg/s) exponent pressure (Pa) ground pressure (Pa) pressure at z L (Pa) vapor pressure (Pa) branch tip pressure (Pa) rank of trees of size Di maximum bending stress (N/m2) Darcy (volume averaged) longitudinal velocity (m/s) branch Darcy longitudinal velocity (m/s) Darcy radial velocity (m/s) wind speed (m/s) volume (m3) total volume (m3) wood volume fraction longitudinal coordinates (m) side of square (m), Fig. 7b side of equilateral triangle (m), Fig. 6b viscosity (kg/s m) kinematic viscosity (m2/s) density (kg/m3)

which is the best way to provide ow access between one point and a nite-size volume (Bejan, 1997). The demand to be strong mechanically is made clear by features such as the tapered trunks and limbs with round cross-section, and other design-like features identied in this article. These features of designedness in solid structures facilitate the ow of stresses, which is synonymous with mechanical strength. According to constructal theory, plants (vegetation) occur and survive in order to facilitate ground-air mass transfer (Bejan, 2006, p. 770). Recently, constructal theory (Bejan, 1997, 2000) has shown that dendritic crystals such as snowakes are the most effective heat-ow congurations for achieving rapid solidication (Bejan, 1997; Ciobanas et al., 2006). The same mental viewing was used to explain the variations in the morphology of stony corals and bacterial colonies and the design of plant roots (Miguel, 2006; Biondini, 2008). The 23-level architecture of the lung (Reis et al., 2004), the scaling laws of all river basins (Reis, 2006; Bejan, 2006), and the macroscopic features (speeds, frequencies, forces) of all modes of animal locomotion (ying, running, swimming) (Bejan and Marden, 2006) were attributed to the same evolutionary principle of conguration generation for greater ow access in time (the constructal law). In summary, there is a renewed interest in explaining the designedness of nature based on universal theoretical principles (Turner, 2007), and constructal theory is showing how to predict the generation of natural conguration across the board, from biology to geophysics and social dynamics (for reviews, see Bejan and Lorente, 2006; Bejan, 2006; Bejan and Merkx, 2007). In this paper, we rely on constructal theory in order to construct based on a single principle the main features of plants, from root and canopy to forest. We take an integrative approach to trees as live ow systems that evolve as components of the larger whole (the environment). We regard the plant as a physical ow architecture that evolves to meet two objectives: maximum mechanical strength against the wind, and maximum access for the water owing through the plant, from the ground to the atmosphere.

Ours is a physics paper rooted in engineering. The purpose of our work is to demonstrate that the existence of tree-like architecture can be anticipated as a mental viewing based on the constructal law. The work is purely theoretical. Although comparisons with natural forms are made, the work is not intended to describe and correlate empirically the diversity of plant measurements found in nature. Although we are not nearly as familiar as our biology colleagues with the sequence of theoretical and empirical advances made on vegetation morphology, in constructal theory we have a physics method with which we have predicted natural ow design across the board (Bejan and Lorente, 2006). We bring to this table of discussion the tools of strength of materials, uid mechanics, and, above all, the engineering thinking of multi-objective design. We believe that our physics work will be of interest because of its engineering origins and purely theoretical character and message.

2. Root shape The plant root is a conduit shaped in such a way that it provides maximum access for the ground water to escape above ground, into the trunk of the plant. The ground water enters the root through all the points of its surface. In the simplest possible description, the root is a porous solid structure shaped as a body of revolution (Fig. 2). The shape of the body [L, D(z)] is not known, but the volume is xed: Z L p 2 V D dz (1) 0 4 The ow of water through the root body is in the Darcy regime. The permeability of the porous structure in the longitudinal direction (Kz) is greater than the permeability in the transversal direction (Kr). Anisotropy is due to the fact that the woody vascular tissue (the xylem) is characterized by vessels and bers that are oriented longitudinally.

Please cite this article as: Bejan, A., et al., Unifying constructal theory of tree roots, canopies and forests. J. Theor. Biol. (2008), doi:10.1016/j.jtbi.2008.06.026

ARTICLE IN PRESS
A. Bejan et al. / Journal of Theoretical Biology ] (]]]]) ]]]]]] 3

and r is the density of water. Eqs. (4) and (5) yield d 2 D u 4vD dz (6)

Summing up, the three Eqs. (2), (3) and (6) should be sufcient for determining u(z), v(z) and D(z) when the length L is specied. Here, the challenge is of a different sort (much greater). We must determine the shape [L, D(z)] that allows the global pressure difference (PgPL) to pump the largest ow rate of water to the ground level: _ mL r

p
4

D2 LuL

(7)

subject to the volume constraint (1). Instead of trying a numerical approach or one based on variational calculus, here we use a much simpler method. We assume that the unknown function D(z) belongs to the family of power-law functions: D bzm (8)

where b and m are two constants. We also make the assumption that the function P(z) belongs to the family represented by Pg Pz azn m=K z (9)

Fig. 2. (a) Root shape with power-law diameter; (b) constructal root design: conical shape and longitudinal tubes with constant (z-independent) diameters, density, u and v.

We assume that the (L, D) body is sufciently slender, so that the pressure inside the body depends mainly on longitudinal position, Pr; z Pz. This slenderness assumption is analogous to the slender boundary layer assumption in boundary layer theory. For Darcy ow, the z volume averaged longitudinal velocity is given by K z dP u m dz (2)

where a and n are two additional constants. When we substitute assumptions (8) and (9) into Eqs. (2) and (3), and then substitute the resulting u and v expressions into Eq. (6), we obtain two compatibility conditions for the assumptions made in Eqs. (8) and (9): m1 b nn 1 8
2

(10) Kr Kz (11)

The volume constraint (1) yields a third condition: b L3


2

12

(12)

where m is the uid viscosity. Because of the Pr; z Pz assumption, for the transversal volume averaged velocity v (oriented toward negative r) we write approximately: v K r P g Pz m D=2 (3)

A fourth condition follows from the statement that the overall pressure difference is xed, which in view of Eq. (9) means that Pg PL aLn ; m=K z constant (13)

The denition of the radial permeability (Kr) of the root body as a Darcy porous medium is Eq. (3). This denition is consistent with Eq. (2), which is the denition of the longitudinal permeability of the root as a nonisotropic Darcy porous medium (e.g., Nield and Bejan, 2006). The directional permeabilities Kz and Kr are two constants. The radial permeability Kr should not be confused with the concept of radial water conductivity kr, which is dened as the ratio between the radial ux of water and the radial pressure difference [e.g., Eq. (3.3) in Roose and Fowler, 2004]. The ground-water pressure (Pg) outside the body is assumed constant. This means that in this model the hydrostatic pressure variation with depth Pg(z) is assumed to be negligible, and that the root sketched in Fig. 2 can have any orientation relative to gravity. Ground level is indicated by z L: here the pressure is PL, and is lower than Pg. Throughout the body, P(z) is lower than Pg, and the radial velocity v is oriented toward the body centerline. The conservation of water ow in the body requires _ dm rpDv dz _ where m is the longitudinal mass ow rate at level z: _ mr (4)

Finally, the mass ow rate through the z L end of the body is, cf. Eq. (7):   p K z dP g P p 2 _ mL r bL2 r b anLn1 (14) dx 4 m 4 zL for which b(n) and L(n) are furnished by Eqs. (11) and (12). The resulting ground-level ow rate is 1=3    p K r 2=3 12 n1=3 _ mL r aLn 8 V (15) 4 Kz p n 12=3 with the observation that (aLn) is a constant, cf. Eq. (13). _ In conclusion, mL depends on root shape (n) according to the function n1/3/(n+1)2/3. This function is maximum when n1 (16)

Working back, we nd that the constructal root design must have this length and aspect ratio:   3VK z 1=3 (17) L pK r   L 1 K z 1=2 DL 2 K r (18)

p
4

D2 u

(5)

The constructal root shape is conical. The slenderness of this cone is dictated by the anisotropy of the porous structure (Kz/Kr)1/2.

Please cite this article as: Bejan, A., et al., Unifying constructal theory of tree roots, canopies and forests. J. Theor. Biol. (2008), doi:10.1016/j.jtbi.2008.06.026

ARTICLE IN PRESS
4 A. Bejan et al. / Journal of Theoretical Biology ] (]]]]) ]]]]]]

The root is more slender when the vascular structure is more permeable longitudinally. Another important feature of the discovered root geometry is that the longitudinal volume averaged uid velocity (u) is independent of longitudinal position (z), because n 1 means that dP/dz constant, and u K z Pg PL m L 0ozoL Kz   K r 2=3 Pg PL 3V1=3 p n Kz (19)

_ mL r

p
4

bL2 u

(20)

The morphological implications of this theoretical feature are important. If the porous structure is a bundle of tiny capillary channels, then the uid velocity through each tube must be constant, and must not depend on the size of the root crosssection (p/4)D2(z) that the channel pierces. On the other hand, earlier work on constructal design (Bejan and Lorente, 2004) has shown the following: ow strangulation is not good for ow performance, the constructal conguration of a long capillary with specied ow rate and volume is the one where the crosssection does not vary with longitudinal position, and the crosssection is round. Combining this with the new conclusion that u must not depend on z, we discover the internal structure of the constructal root body. The longitudinal tubes must be round, with diameters that do not vary with z, even though some tubes are longer than other tubes. The external shape and internal structure of the root body discovered in this section are sketched on the right side of Fig. 2. Another feature of the constructal root design is visible in Eq. (3). Because both (PgP) and D are proportional to z, we conclude that v must also be z-independent. One can show that  1=2 v Kr (21) u Kz The anisotropy of the vascular porous structure dictates the ratio between constant-v and constant-u, in the same way that it dictates the root slenderness ratio DL/L, cf. Eq. (18). There is a considerable body of literature on the modeling of water ow through roots, and a common assumption is that the root is a porous conduit with constant diameter. The analysis then yields a pressure that varies nonlinearly (exponentially) along the conduit (Landsbert and Fowkes, 1978; Frensch and Stendle, 1989; Roose and Fowler, 2004). This is consistent with the analysis shown in this section, for if this analysis is repeated by postulating that D is constant, then in place of the present conclusion (dP/ dz constant) we nd that P(z) must vary exponentially. However, measurements made on root segments show that the pressure does not vary linearly (Zwieniecki et al., 2003; McElrone et al., 2004). This is a good opportunity for next-generation analytical and experimental work. For example, the Landsbert and Fowkes (1978) modeling should be combined with the optimal tapering of the conduits, i.e., with the constructal law of facilitating ow access by allowing the ow geometry to morph.

Fig. 3. Slender tree canopy and trunk exposed to a horizontal wind with uniform velocity.

by the wind on the canopy as a problem of two-dimensional ow, in a horizontal plane that cuts the trunk and the canopy. The trunk and the canopy are modeled as two bodies of revolution, with unknown diameters Dt(x) and Dc(x), where x is measured downward from the top of the tree. The drag force per unit length (x) experienced by the tree canopy is F 0 C D Dc 1rV 2 2 (22)

where V is the horizontal wind speed and Dc(x) is the radius of the canopy at the distance x from the tree top. We assume that the Reynolds number VDc/n is greater than 103, so that the drag coefcient CD is a constant approximately equal to 1. To give our search for geometry sufcient generality, assume that the canopy has a shape that belongs in the family of powerlaw functions: Dc axn (23)

where a is a constant and the shape exponent n is not known. The bending moment experienced by the trunk at the distance x from the tree top is Z x a0 xn2 Mx (24) F 0 x x dx n 1n 2 0 where a0 is another constant: a0 a rV 2 C D 2 (25)

We now turn our attention to the maximum bending stresses in the cross-section of the trunk of diameter Dt(x): sm Mx Dt x It x 2 (26)

where It pD4 =64. The stress sm occurs in the dorsal and ventral t bers of the trunk, as the trunk bends in the wind that pushes the canopy. Optimal distribution of imperfection (Bejan, 2000) means that sm must be the same over the entire height of the tree. According to Eq. (24), the trunk diameter must vary as Dt x  32a0 =p sm n 1n 2 1=3 xn2=3 (27)

3. Trunk and canopy shape The water stream guided by the root from underground to ground level continues to ow upward through the trunk and canopy of the plant. To continue with the same analytical ease as in the analysis of root geometry, for the trunk and canopy of the plant we make the simplifying assumptions hinted at in Fig. 3, which is based on a problem proposed in Bejan (2006, pp. 831832). We assume that both the canopy and the trunk are sufciently slender. This allows us to analyze the forces exerted

This is an important result, but it is not the end of the story. It says that if we know the canopy shape (n), then we can predict the trunk shape, and vice versa (Fig. 4). To determine the trunk and canopy shapes uniquely, we need an additional idea (Section 4). If the canopy is shaped as a cone (n 1), then the trunk is also shaped as a cone, Dt/x constant. Fig. 4 shows that if the canopy has a round top (e.g., n 1/2) then the trunk diameter must vary as Dt/x5/6 constant, which is not much different than Dt/ x constant. If the canopy has a very sharp tip (e.g., n 3/2), then Dt(x) must vary as Dt/x7/6 constant, which again is not far

Please cite this article as: Bejan, A., et al., Unifying constructal theory of tree roots, canopies and forests. J. Theor. Biol. (2008), doi:10.1016/j.jtbi.2008.06.026

ARTICLE IN PRESS
A. Bejan et al. / Journal of Theoretical Biology ] (]]]]) ]]]]]] 5

Fig. 4. Three canopy shapes showing that the optimal trunk shape is near-conical in all cases.

off the conical trunk shape. In sum, we have discovered that the shape of the trunk that is uniformly stressed is relatively insensitive to how the canopy is shaped. A conical trunk is essentially a uniform-stress body in bending for a wide variety of canopy shapes that deviate (concave vs. convex) from the conical canopy shape sketched in Fig. 3. A simpler version of the problem solved in this section is to search for the optimal shape of the trunk Dt(x) when there is no canopy. The trunk alone is the obstacle in the wind, and its bending is due to the distributed drag force F0 of Eq. (22), in which Dc is replaced by Dt. The analysis leads to Eq. (27) where M(x) varies as xn+2, and sm (constant) is proportional to Mx=D3 . The conclusion is that the trunk (or t solitary pole) is the strongest to bending when it is conical, n 1. The same result follows from the subsequent discussion of Eq. (27), if we assume Dc Dt. A famous structure that only now reveals its bendingresistance design is the Eiffel Tower (Science et Vie, 2005). The shape of the structure is not conical (Fig. 4) because in addition to bending in the wind, the structure must be strong in compression. The optimization of tower shape for uniform distribution of compressive stress leads to a tower prole that becomes exponentially narrower with altitude. The shape of a tower that is uniformly resistant to lateral bending and axial compression is between the conical and the exponential. This apparent imperfection (deviation from the exponential) of the Eiffel Tower has been a puzzle until now (see the end of Section 4). This discussion of the Eiffel Tower also sheds light on a major mechanical difference between the present theory and the model of West et al. (1999). In the present work, the mechanical function is to resist bending due to horizontal wind drag, as in the upper section of the Eiffel Tower. In the model of West et al., the mechanical function is to resist buckling under its own weight, on the vertical. Of course, all modes of resisting fracture are important, but, which is the more important? Buckling is not, because the weight of the tree is static, totally independent of the notoriously random and damaging behavior of the owing environment. The wind is much more dangerous. Record breaking wind speeds make news all over the globe, and their combined effect can only be one: the cutting of the trunks, branches and leaves to size. What is too long or sticks out too much is shaved off. The tree architecture that strikes us as pattern today (i.e., the

emergence of scaling laws) is the result of this never-ending assault.

4. Conical trunks, branches and canopies The preceding section unveiled the architecture of a tree that has evolved, so that its stresses ow best and its maximum allowable stress is distributed uniformly. This tree supports the largest load (i.e., it resists the strongest wind) when the tree volume is specied. Conversely, the same architecture withstands a specied load (wind) by using minimum tree volume. In summary, the multitude of near-conical designs discovered in Eq. (27) and Fig. 4 refer to the mechanical design of the structure, i.e., to the ow of stresses, not to the ow of uid that seeps from thick to thin, along the trunk and its branches. There is no question that the maximization of access for uid ow plays a major role in the conguring of the tree. This is why the tree is tree-shaped, dendritic, one trunk with branches, and branches with many more smaller branches. How do the designs of Eq. (27) facilitate the maximization of access for uid ow? The answer is provided by the constructal root discovered in Section 2 and Eqs. (17)(21). The constructal shape for a body permeated by Darcy ow with two permeabilities (Kz, Kr) is conical. The longitudinal and lateral seepage velocities (u, v) are uniform, independent of the longitudinal position z. For a root, the lateral seepage is provided by direct (contact) diffusion from the soil, and indirect seepage from root branches, rootlets and root hairs. For the tree trunk above the ground, the lateral ow that accounts for v is facilitated (ducted) almost entirely by lateral branches. Above the ground, the lateral v is concentrated discretely in branches that are distributed appropriately along and around the trunk (see the discussion of the Fibonacci sequence at the end of this section). The theoretical step that we make here is this: the constructal ow design of the root is the same as the ow design of the trunk and canopy. From this we deduce that out of the multitude of near-conical trunk shapes for wind resistance, Eq. (27), the constructal law selects the conical shape, n 1. The conical shape is also the constructal choice for the large and progressively smaller lateral branches, provided that their mechanical design is dominated by wind resistance considerations, not by the

Please cite this article as: Bejan, A., et al., Unifying constructal theory of tree roots, canopies and forests. J. Theor. Biol. (2008), doi:10.1016/j.jtbi.2008.06.026

ARTICLE IN PRESS
6 A. Bejan et al. / Journal of Theoretical Biology ] (]]]]) ]]]]]]

resistance to their own body weight. We return to this observation in the last paragraph of this section. Recognition of the conical trunk and canopy shapes means that the analysis in this section begins with Eqs. (18) and (23), which for the tree trunk and canopy reduce to  1=2 Dt x Kr 2 b (28) x Kx Dc x a x (29)

tip of the trunk (x 0). The pressure at the tip of the branch LB is also P0. In accordance with Eq. (19), we write u K x Px P 0 m x K x;B Px P 0 m LB (38)

uB

(39)

which yield u K x LB uB K x;B x (40)

Here, it should be noted that for the tree trunk the axial coordinate (x) is measured downward (from the tree top, Fig. 3), whereas the axial coordinate of the root (z) is measured upward (from the root tip, Fig. 2). The proportionality between Dt(x) and Dc(x) is provided by Eq. (27) with n 1, in combination with Eqs. (25), (28) and (29): Dc x a 3psm K r Dt x b 2C D rV 2 K x (30)

It is reasonable to assume that the longitudinal permeability of the wood to be the same in the trunk and the branch, K x K x;B , such that Eq. (37) reduces to h 1ax 1LB 4 4 (41)

Eq. (30) recommends a large Dc/Dt ratio for trees with hard wood in moderate winds, and a small Dc/Dt ratio for trees with soft wood in windy climates. A hard-wood example is the walnut tree (Juglans regia) with sm 1:2 108 N/m2, in a mild climate represented by V$50 km/h (14 m/s). Eq. (30) with CD$1 yields Dc/Dt$2.42 106(Kr/Kx)walnut and, after additional algebra, 3=2 Dc =Ltrunk $4:8 106 K r =K x walknut . The corresponding estimates for a pine tree (Pinus silvestris) with sm 6:6 107 N/m2 in a windy climate with V$100 km/h (28 m/s) are Dc/Dt$3.4 105(Kr/ 3=2 Kx)pine and Dc =Ltrunk $6:8 105 K r =K x pine . How many branches should be placed in the canopy, and at what level x? We answer this question with reference to Fig. 5, where the aspect ratios of the trunk (Dt/x b) and canopy (Dc/ x a) also hold for the branch LB(x) located at level x: Dt;B b; LB Dc;B a LB (31)

In conclusion, the vertical segment of trunk (h) that is responsible for the ow rate into one lateral branch is proportional to the length of the branch. Another dimension that is proportional to LB(x) is the diameter of the conical branch canopy circumscribed to the horizontal LB, namely Dc,B aLB, cf. Eq. (34). Comparing h with Dc,B, we nd that hx 1 Dc;B x 2a (42)

Furthermore, in accordance with Eq. (29) for the canopy, Dc(x) is the same as 2LB(x), which means that LB x 1ax 2 Dt;B x 1abx 2 Dc;B x 1a2 x 2 (32) (33) (34)

A single branch LB(x) resides in a frustum of the conical canopy: the frustum height is h(x) and the base radius is LB(x). In the center of this frustum, there is a trunk segment (another conical frustum) of height h(x) and diameter Dt(x). The trunk frustum can be approximated as a cylinder of diameter Dt(x). The total ow rate of uid that ows laterally from this trunk segment is _ mB rvpDt h (35)

which is a constant of order 1. In other words, there is room in the global canopy (L, Dc) to install one LB-long branch on every h-tall segment of tree trunk. The geometrical features discovered in this section have been sketched in Fig. 5. One of the reviewers of the original manuscript asked us to compare this tree architecture with that of the model of West et al. (1999). This was a great suggestion because it leads to an important theoretical discovery that is hidden in the massconservation analysis that led to Eq. (41). The discovery is that Leonardos rule (e.g., Horn, 2000; Shinozaki et al., 1964) is deducible from Eq. (41), in these steps. The trunk cross-sectional 2 area at the distance x from the tip is At x p=4b x2 . At the top 2 2 of the h frustum, it is At x h p=4b x h . The reduction in trunk cross-sectional area from x to xh is DAt At x At x h. The cross-sectional area of the thick end of the single branch 2 allocated to h is AB p=4D2 p=4b L2 . The ratio between the t;B B decrease in trunk cross-sectional area and the branch crosssectional area allocated to that decrease is, after some algebra, DAt =AB 2=a1 a=8. In view of Eq. (42), where 1=2a$1, according to constructal theory the ratio DAt =AB must be a constant of order 1.

If uB is the longitudinal uid velocity along the branch LB, then the same uid mass ow rate can be written as _ mB ruB

p
4

D2 t;B

(36)

where Dt,B is the diameter of branch LB at the junction with the _ trunk. Eliminating mB between Eqs. (35) and (36), and using Eqs. (28), (32) and (33), we nd that h is proportional to x: h uB a2 x u 8 (37)

The ratio uB/u is a constant determined as follows. Let P(x) be the pressure at level x inside the trunk, and P0 the pressure at the

Fig. 5. Conical canopy with conical branches and branch-canopies.

Please cite this article as: Bejan, A., et al., Unifying constructal theory of tree roots, canopies and forests. J. Theor. Biol. (2008), doi:10.1016/j.jtbi.2008.06.026

ARTICLE IN PRESS
A. Bejan et al. / Journal of Theoretical Biology ] (]]]]) ]]]]]] 7

The area ratio would have been exactly 1 according to Leonardos rule, which was based on visual study and drawings of trees. This rule is predicted here based on the constructal law and other rst principles such as the conservation of water mass ow rate. In West et al.s (1999) model, this rule was assumed, not predicted. It was assumed along with several other assumptions (e.g., the tree-shaped structure), so that the model could become compact and useful as a facsimileas a descriptionof the real tree, just like Leonardos observations. It is because of such assumptions that the allometric relations derived algebraically from West et al.s (1999) model are description, not prediction. This remark is necessary because it contradicts West et al.s use of the words predicted values in the reporting of their derivations (e.g., Table 1, p. 667). Additional comments on West et al.s model are provided by Kozlowski and Konarzewski (2004) and Makela and Valentine (2006). In the present paper, the tree architecture and the tapering of its limbs are deduced from a single postulate which is the constructal law. Furthermore, because there is one lateral branch per trunk segment h(x), and because h decreases in proportion with x, the best way to ll the tree canopy with the canopies of the lateral branches is by arranging the branches radially, so that they ll the alveoli created in the canopy cone by two counterrotating spirals that spin around toward the top of the tree canopy. When one counts the sequence in which these alveoli arrange themselves up the trunk, one discovers the Fibonacci sequence (e.g., Livio, 2002). Like Leonardos rule, the Fibonacci sequence is the result of Eq. (42), the predicted conical canopy shape, and the geometric requirement that the next branch and canopy should shoot laterally into the space that is impeded the least by the branch canopies situated immediately above and below. The need of minimum interference between branches is a restatement of the constructal law, i.e., the tendency to morph to have greater ow access for water from ground to wind. Each branch reaches for the pocket of volume that contains the least humid air ow. This principle is universal, and is fundamentally different than ad hoc statements such as stems grow in positions that would optimize their exposure to sun, rain, and air (Livio, 2002), and phyllotaxis simply represents a state of minimal energy for a system of mutually repelling buds (Livio, 2002; after Douady and Couder, 1992). The tree structure discovered step by step up to this point consists of cones inside cones. The large conical trunk and canopy hosts a close packing of smaller conical branches and conical branch canopies. One can take this construction further to smaller scales, and see the architecture of each branch as a conical canopy packed with smaller conical branches and their smaller canopies. In such a construction, the wood volume is a fraction of its total volume, i.e., a fraction of the volume of the large canopy, which scales as L3. From this follows the prediction that the trunk length L must be proportional to the total wood mass raised to the power 1/3. This prediction agrees very well with measurements of ve orders of magnitude of tree mass scales (e.g., Table 2 in Bertram, 1989). In closing, we return to the Eiffel Tower discussed at the end of the preceding section, where we noted that strength in compression (under the weight) near the base was combined with strength in bending (subject to lateral wind) in the upper body of the tower. This discussion is relevant in the modeling of the horizontal branch, which in this section was based on the assumption that the loading is due to lateral wind. The branch is also loaded in the vertical direction, under its own weight. If we assume that the distributed weight of the branch is the only load, then the branch shape of constant strength (i.e., with xindependent sm) has the form D ax2, where a is constant. Such a branch has zero thickness in the vicinity of the tip (dD/dx 0 at x 0+), and is not a shape found in nature. This result alone

indicates that the tips of branches are not shaped by weight loading alone, and that wind loading (which prescribes D ax and nite D at small x) is the more appropriate model there. For the thick end of the branch, it can be argued that D ax2 is a realistic shape, and that near the trunk the weight loading of the beam is the dominant shaping mechanism, just like in the Eiffel Tower near the ground.

5. Forest Forests are highly complex systems, and their study has generated a signicant body of literature (for reviews, see Keitt et al., 1997; Urban et al., 1987). Multi-scale models of landscape pattern and process are being applied, for example, models with spatially embedded patch-scale processes (Weishampel and Urban, 1996). To review this activity is beyond our ability, and is not our objective. Here we continue on the constructal path traced up to this point (Fig. 1): if the root, trunk and canopy architecture is driven by the tendency to generate ow access for water, from ground to air, then, according to the same mental viewing (i.e., according to the same theory), the forest too should have an architecture that promotes ow access. The uid ow rate ducted by the entire tree from the ground to the tips of the trunk and branches is: _ m ru

p
4

D2 x L t

p b2 K x Px L P0 Dc x L 4 an

(43)

where x L indicates ground level and Dc(x L) is the diameter of the canopy projected as a disc on the ground. The important feature of the tree design discovered so far is the proportionality _ between m and Dc(x L). This also means that the total mass ow rate is proportional to the tree height L. This proportionality will be modied somewhat when we take into account the additional _ ow resistance encountered by m as it ows from the smallest branches (P0) through the leaves and into the atmosphere (Pa). See Section 6. Seen from above, an area covered with trees of many sizes (Dc,i) _ _ is an area covered with uid mass sources (mi ), where each mi is proportional to the diameter of the circular area allocated to it. From the constructal law of generating ground-to-air uid ow access follows the design of the forest. The principle is to morph the area into a conguration with mass sources (or disc-shaped canopy projections) such that the total uid ow rate lifted from the area is the largest. From this invocation of the constructal law follows, rst, the prediction that the forest must have trees of many sizes, few large trees interspaced with more and more numerous smaller trees. This is illustrated in Fig. 6a with a triangular area covered by canopy projections arranged according to the algorithm that a single disc is inserted in the curvilinear triangle that emerges where three discs touch. If the side of the large triangle is Xt, then the diameter of the largest canopy disc is D0 Xt, and the number of D0-size canopies present on one Xt triangle is n0 1/2. For the next smaller canopy, the parameters are D1 (31/21/2)Xt and n1 1. At the next smaller size, the number of canopies is n2 3, and the disc size is D2 0.0613Xt. The construction continues in an innite number of steps (n3 3, n4 6, y) until the Xt triangle is covered completely. The image that would result from this innite compounding of detail would be a fractal. The total uid ow rate vehicled by the design from the triangular area of Fig. 6a is proportional to ma
1 X i0

ni Di

1 D0 D1 3D2 0:761 X t 2

(44)

Because a canopy disc D contributes more to the global production (m) when D is large and when the number of D-size discs is large,

Please cite this article as: Bejan, A., et al., Unifying constructal theory of tree roots, canopies and forests. J. Theor. Biol. (2008), doi:10.1016/j.jtbi.2008.06.026

ARTICLE IN PRESS
8 A. Bejan et al. / Journal of Theoretical Biology ] (]]]]) ]]]]]]

Fig. 6. Multi-scale canopies projected on the forest oor: (a) triangular pattern with algorithm-based generation of smaller scales and (b) triangular pattern with more large-size canopies.

Fig. 7. Square pattern of canopy assemblies: (a) algorithm-based generation of smaller scales and (b) more numerous large-scale canopies for greater ground-air ow conductance.

a better forest architecture is the one where the larger discs are more numerous. This observation leads to Fig. 6b, where the Xt triangle is covered more uniformly by larger discs, in this sequence: D0 [(31/2+1)/2]Xt and n0 1/2, D1 [(31/21)/2]Xt and n1 1, D2 [(131/2)/2]Xt and n2 3/2, etc. The total mass ow rate is mb
n X i0

ni Di

1 3 D0 D1 D2 1:077 X t 2 2

(45)

This ow rate is signicantly greater than that of the fractal-like design of Fig. 6a. The numbers of canopies of smaller scales that would complete the construction of Fig. 6b are n3 6, n4 6, n5 6, n6 6, y, but their contributions to the global ow rate (mb) are minor. The important aspect of the comparison between Fig. 6a and b is that there is a choice [(b) is better than (a)], because each tree contributes to the global ow rate in proportion to its length scale. Had the construction been based simply on the ability to ll the area by repeating an assumed algorithm, as in fractal (space lling) practice (e.g., West et al., 1999), there would have been no difference between (a) and (b), because the triangular area is the same in both cases, and both designs cover the area. Furthermore, the fractal-like design (a) is simpler and more regular, while the better design (b) is strange, and seemingly random. One may ask, why should (b) look different than (a), and why should (b) have three large scales (D0, D1, D2) instead of just one? There is nothing strange about the evolution of the drawing (in time) from (a) to (b). This is the time arrow of the constructal law. It may be possible to nd triangular designs that are (marginally)

better than (b), but that should not be necessary in view of the global picture that will be discussed in relation to Figs. 810. Discs arranged in a square pattern also cover an area completely. One can draw and evaluate the square equivalent of Fig. 6a and by replacing the Xt triangle with a square of side Xs. The result is Fig. 7a. The numbers of discs of decreasing scales (D0 bD1 ; D2 ; . . . present on this square will be n0 1, n1 1, n2 4, etc. The performance of this regular (fractal-like) design will be signicantly inferior to that of the square pattern shown in Fig. 7b, which is the square equivalent of Fig. 6b. The canopy sizes and numbers in the square design are D0 21/2Xs and n0 2, D1 (121/2)Xs and n1 2, etc. The total mass ow rate extracted from the Xs-square is ms
1 X i0

ni Di 2D0 2D1 8D2 2:608X s

(46)

Coincidentally, one can show that the m values of Fig. 7a and b form the same ratio (namely 0.71) as the m values of Fig. 6a and b. Finally, we compare Eq. (46) with Eq. (45) to decide whether the square design (Fig. 7b) is better than the triangular design of Fig. 6b. The area is the same in both designs, therefore Xt/Xs 2/31/4 and Eqs. (45) and (46) yield mb 0:826 ms (47)

The square design is better, but not by much. Random effects (geology, climate) will make the distribution of multi-scale trees switch back and forth between triangle and square and maybe hexagon, creating in this way multi-scale patterns that appear

Please cite this article as: Bejan, A., et al., Unifying constructal theory of tree roots, canopies and forests. J. Theor. Biol. (2008), doi:10.1016/j.jtbi.2008.06.026

ARTICLE IN PRESS
A. Bejan et al. / Journal of Theoretical Biology ] (]]]]) ]]]]]] 9

Fig. 8. The hierarchical distribution of canopy sizes versus rank in the triangular forest oor designs of Fig. 6.

Fig. 9. The hierarchical distribution of canopy sizes versus rank in the square forest oor designs of Fig. 7.

Table 1 Sizes, numbers and ranks for the multi-scale canopies populating the forest designs of Fig. 6 i Size, Di/Xt (a) 0 1 2 3 4 5 6 1 0.155 0.0613 0.0325 0.0206 0.02 0.0106 (b) 0.789 0.366 0.211 0.054 0.024 0.021 0.019 2ni (a) 1 2 6 6 12 6 12 (b) 1 2 3 12 12 12 12 Rank (a) 1 2, 3 49 1015 1627 2833 3445 (b) 1 2, 3 46 724 2536 3748 4960

Table 2 Sizes, numbers and ranks for the multi-scale canopies populating the square forest design of Fig. 7 i Size, Di/Xs (a) 0 1 2 3 4 1 0.414 0.107 0.048 0.040 (b) 0.707 0.3 0.076 0.036 0.029 ni (a) 1 1 4 4 8 (b) 2 2 8 8 16 Rank (a) 1 2 36 710 1118 (b) 1, 2 3, 4 512 1320 2136

even more random than the triangle alone, the square alone, and the hexagon alone. The key feature, however, is that the design is with multiple scales arranged hierarchically, and that this sort of design is demanded by the constructal law of generating groundair ow access. The hierarchical character of the large and small trees of the forest is revealed in Fig. 8, where we plotted the size (Di) and rank of the canopies shown in Fig. 6a and b. The calculation of the rank is explained in Table 1. The largest canopy has the rank 1, and after that the canopies are ordered according to size, and counted sequentially. For example, the canopies of size D2 in Fig. 6b are tied for places 46. The sizes were estimated graphically by inscribing a circle in the respective curvilinear triangle in which the projected canopy would t. The data collected for designs (a) and (b) in Table 1 are displayed as canopy size versus the canopy rank in Fig. 8. To one very large canopy belongs an entire organization, namely two canopies of next (smaller) size, followed by increasingly larger numbers of progressively smaller scales. This conclusion is reinforced by Fig. 9, which in combination with Table 2 summarizes the ranking of scales visible in the square arrangements of canopies drawn in Fig. 7a and b. There are no signicant differences between Figs. 8 and 9. The noteworthy feature is the alignment of these data as approximately straight lines on the loglog eld of Figs. 8 and 9.

Fig. 10. The Zipan distribution of canopy sizes versus rank, as a summary of Figs. 8 and 9.

A birds eye view of this hierarchy is presented in Fig. 10. This type of alignment is associated empirically with the Zipf distribution, and it was discovered theoretically in the constructal theory of the

Please cite this article as: Bejan, A., et al., Unifying constructal theory of tree roots, canopies and forests. J. Theor. Biol. (2008), doi:10.1016/j.jtbi.2008.06.026

ARTICLE IN PRESS
10 A. Bejan et al. / Journal of Theoretical Biology ] (]]]]) ]]]]]]

distribution of multi-scale human settlements on a large territory (Bejan, 2006, pp. 774779).

6. Discussion More theoretical progress can be made along this route if we ask additional questions about the ow of water through the tree and into the atmosphere. The ow path constructed thus far consists of channels (root, trunk, branches). This construction can be continued toward smaller branches, in the same way as in Fig. 5, where we used the trunk and canopy design to deduce the design of the branch and canopy design. This step can be repeated a few times, toward smaller scales. _ The water stream m ows through this structure from the base of the trunk, P(L), to the smallest branches, P0. From the inside of the smallest branches to the atmosphere (where the water vapor _ pressure is Pv), the stream m must diffuse across a large surface that is wrinkled and packed into the interstices formed between branches (this is a model for the main path of water loss, through the variable-aperture stomata on leaf surfaces, which provide low resistance for water loss by diffusion when fully open). This, diffusion at the smallest scales, optimally balanced with hierarchical channels at larger scales, is the tree architecture of constructal theory (Bejan, 1997, 2000). It was recognized earlier in hill slope seepage and river channels, alveolar diffusion and bronchial airways, diffusion across capillaries and blood ow through arteries and veins, walking and riding on a vehicle in urban trafc, etc. This balance between diffusion and channeling, which lls the volume completely, is why the constructal trees are not fractal: if one magnies a subvolume, one sees an image that is not a repeat of the original image. Inside the tree canopy, the large surface through which _ channeled m makes contact with the owing atmosphere is provided by leaves that ride on the smallest branches. If their total surface area is A, then the global ow rate crossing A is _ m c2 AP 0 P v (48)

conrmed by a review of published measurements (Tyree, 2003) of global transpiration in sugar maple (Acer saccharum) of trunk base diameters in the range 1.3 mm10 cm, which showed a 1=3 proportionality between C and V T 1:42 . Further support for this conclusion is provided by measurement reported by Ryan et al. (2000) for ponderosa pine (Pinus ponderosa) of two sizes, 12 and 36 m high. The measurements show that under various timedependent conditions (diurnal and seasonal) the length-specic water ux [i.e., C/(length)2] for 12 m trees is approximately twice as large as the water ux for 36 m trees. This means that the ratio C(36 m)/C(12 m) is essentially constant in time and equal to 2. This also means that the exponent in the proportionality between C 1=3 and V T p is approximately p 1.37, which is in good agreement with Tyree (2003) and the discussion of Eq. (51). The balance between diffusion at the smallest (interstitial) scale and channeling at larger scales, which was demonstrated for several classes of tree-shaped ows (e.g., Reis et al., 2004; Miguel, 2006), means that there must be an optimal allocation of leaf volume to wood (xylem) volume, so that C is maximum (the xylem volumethe specialized layer of tissue through which water owsis proportionally a fraction of the total wood volume). Indeed, if we replace l with (1w) in Eq. (51), we nd C$ c1 c2 V T w1=3 1 w2=3
1=3 c1 w1=3 V T 2=3

c2 1 w2=3 V T

(52)

The conductance is zero when there are no branches and trunk (w 0), and when there are no leaves (w 1). The conductance is maximum in between. The optimal wood volume fraction is obtained by solving qC/qw 0, or, in view of the order of magnitude character of this analysis, by simply intersecting the two asymptotes of C, cf. Eq. (51). This method yields  3 w c2 $ VT (53) 2 c1 1 w The conclusion is that there is an optimal way to allocate wood volume to leaf and air volume, and the volume fraction w increases almost in proportion with (c2/c1)3VT. Larger trees must have more wood per unit volume than smaller trees. Trees of the same size (VT) must have a larger wood volume fraction in windy climates, because c2 increases with the wind speed V. The relationship between c2 and V is monotonic and can be predicted based on the analogy between mass transfer and momentum transfer (Bejan, 2004, pp. 534536). If V is small enough, so that the Reynolds number based on leaf length scale y is small, Re Vy/no104, the boundary layers on the leaves are laminar, and the mass transfer coefcient (or c2) is proportional to Re1=2 . This means that c2 is proportional to V1/2. In the opposite extreme, the entire assembly of leaves is a rough surface with turbulent ow in the fully turbulent and fully rough regime, like the ow of water in a rocky river bed. The skin friction coefcient Cf is constant (independent of Re), and the corresponding mass transfer coefcient hm is provided by the Colburn analogy for mass transfer, hm =V 1=2C f Pr 2=3 ; constant. This shows that in the high-V limit the mass transfer coefcient (or c2) increases as V. The analysis that brought us to these conclusions is consistent with analytical denitions and results used in forestry research (e.g., Tyree and Ewers, 1991; Horn, 2000). A well-established principle is the Huber rule, which relates the leaf specic conductivity (LSC) to the specic conductivity of the stem (ks): LSC HV ks (54)

where c2 is proportional to the leaf-air mass transfer coefcient, assumed known. In a stronger wind, c2 is larger and can be calculated based on boundary layer mass transfer theory. _ The corresponding shorthand expression for m traveling along the trunk and branches is, cf. Eq. (43): _ m c1 LPx L P 0 (49)

Here, we wrote L instead of Dc(x L), because Dc(x L) is proportional to L, cf. Eq. (29). Eliminating P0 between Eqs. (48) and (49) we determine the global ow conductance C, from the base of the trunk to the atmosphere:   _ m 1 1 1 C (50) Px L P v c1 L c2 A Let VT represent the total volume in which the tree resides. The volume fractions occupied by wood (trunk and branches), and leaves and air are, respectively, w and l such that w+l 1. In an order of magnitude sense, the length scales of the wood and leaf volumes are (wVT)1/3 and (lVT)1/3. Because the leaves are at, their area scales as (lVT)2/3. Together, these scales mean that Eq. (50) becomes " #1 1 1 C$ (51) c1 wV T 1=3 c2 lV T 2=3 where VT is the tree size and V T its length scale (e.g., trunk base diameter, or height). In conclusion, the global conductance C is proportional to the 1=3 tree length scale V T raised to a power between 1 and 2. This is
1=3

where HV is the Huber value, dened as the sapwood crosssection (AW) divided by the leaf area distal to the stem (AL): HV AW AL (55)

Please cite this article as: Bejan, A., et al., Unifying constructal theory of tree roots, canopies and forests. J. Theor. Biol. (2008), doi:10.1016/j.jtbi.2008.06.026

ARTICLE IN PRESS
A. Bejan et al. / Journal of Theoretical Biology ] (]]]]) ]]]]]] 11

In terms of the variables used in this paper, the specic conductivity of the stem and the leaf specic conductivity are ks _ m A2 dP=dz W _ m AW AL dP=dz (56)

LSC

(57)

Combined, Eqs. (55)(57) reproduce the Huber rule. The present analysis goes one step further, because with the optimization that led to Eq. (53) it provides an additional equation with which to estimate an optimal value for HV. In summary, it is possible to put the emergence of tree-like architectures on a purely theoretical basis, from root to forest. Key is the integrative approach to understanding the emergence of ow design in nature, in line with constructal theory and Turners (2007) view that the living ow system is everything, the ow and its environment. In the present case, trees and forests are viewed as integral components (along with river basins, atmospheric circulation and aerodynamic raindrops) in the global design that facilitates the cyclical ow of water in nature. This approach led to the most basic macroscopic characteristics of tree and forest design, and to the discovery, from theory alone, of the principle that underlies some of the best known empirical correlations of tree water ow performance, e.g., Tyree (2003) and Ryan et al. (2000). To illustrate the reach of the method that we have used, we end with another connection between this work and known and accepted empirical correlations. One example is the well-known self-thinning law of plant spatial packing, where the mean biomass of the plant increases as a power law as the number of plants of the same size decreases (Adler, 1996). A recent review (West and Brown, 2004) showed that the number of trees (Ni) that have the same linear size (e.g., Di) has been found empirically to obey the proportionality Ni $D1. The same proportionality is i found for multi-scale patches (fragments) of forests, e.g., Fig. 2 in Keitt et al. (1997). This proportionality is sketched with circles in Fig. 11. The corresponding rank (Ri) of the trees correlated as Ni $D1 is calculated by arranging all the trees in the order of i decreasing sizes, from the largest (k 1) to the trees of size i: Ri
i X k1

coincidence suggests that the success of empirical correlations between numbers and sizes of trees is another indication that the theoretical distribution of tree rankings (e.g., Fig. 10) is correct, and that the single principle on which this entire paper is based is valid. We are very grateful for the extremely insightful comments provided by the anonymous reviewers, which expanded the range of predictions made based on the constructal law in this paper. Their comments deserve serious discussion and future theoretical work, however, we use this opportunity to begin the discussion right here:

N k Dk

(58)

The resulting ordering of the empirically correlated trees is indicated with black squares in Fig. 11. The DiRi data occupy a narrow strip that has a slope between 1 and 1/2, just like the strips deduced from the constructal law in Figs. 810. This

Fig. 11. Empirical numbers of canopies of the same size (Ni), and the ranks (Ri) of such canopies.

(i) One comment was that it is not surprising that trees and forests exhibit morphologies that provide access for water ow, but generalizing this to a holistic architecture involving trees and atmospheric circulation seems much less obvious. In reality, our work proceeded the other way around. Several authors had the general principle (the constructal law) in mind, and with it they predicted with pencil and paper the morphologies of global water ow as river basins (e.g., Reis, 2006), corals and plant roots (Miguel, 2006), atmospheric circulation and climate (Reis and Bejan, 2006), animal body mass ow as locomotion (Bejan and Marden, 2006), etc. There is great diversity in this list of design predictions, ranging from the biosphere to the atmosphere and the hydrosphere, and covering all the known length and mass scales. Early on in the emerging eld of constructal theory (e.g., Bejan, 2000) it was considered obvious that the river delta too is a ow-access design for point-area ow, predictable based on the same principle, as a river basin turned inside out. Put together, the designs of river basins, deltas and ow of animal mass are facilitating the ow of water all over the globe. The same is happening in the atmosphere and the oceans, because of the patterned circulation known succinctly as climate. The summarizing question came last: what design facilitates the water-ow connection between the land based designs and the atmosphere? Vegetation is one design, for ground-air ow access. Aerodynamic droplets are another, for air-ground water access (see Fig. 1). This is a new and rich direction of theoretical inquiry in which to use the constructal law. There may be other morphological features of the biosphere that can be predicted and brought in line with the holistic architecture of the water circuit in nature. (ii) Another comment was to speculate on how the ow architecture would change if the facilitating of the water cycle is not true. First, all we have is the well-known circuit that water executes in nature, and now this paper in which we linked in very simple terms the tree-like architecture to the water-access function coupled with the wind resistance function. The generation of vegetation architecture is driven by more than two objectives (see the rst paragraph of Section 1), but the two drivers are enough for speculating as suggested by the reviewer. If vegetation is not demanded and shaped by the rest of nature (the environmental ows) to put the ground water back in the air, then, based on our analysis, xed-mass structures that must withstand the winds will all resemble the Eiffel Tower, not the botanical tree (cf., Fig. 4). In reality, vegetation is tree-shaped above and below ground, shaped like all the other point-area and point-volume ows that facilitate ow access. It is the tree shape that argues most loudly in favor of water ow access as the raison detre of vegetation everywhere. This mission comes wrapped in the strength of materials question

Please cite this article as: Bejan, A., et al., Unifying constructal theory of tree roots, canopies and forests. J. Theor. Biol. (2008), doi:10.1016/j.jtbi.2008.06.026

ARTICLE IN PRESS
12 A. Bejan et al. / Journal of Theoretical Biology ] (]]]]) ]]]]]]

of how to protect mechanically (and with xed biomass) the tree-shaped conduits between ground and air. The design solution is to endow the tree with round, tapered and, above everything, long trunks and branches. Tree size ultimately means rate of rainfall, because the tree length scale is proportional to the rate of water mass ow facilitated by the tree. The xed-mass structure must stretch into the air as high and as wide as possible, and not snap in the wind. This is how the design arrives at illustrating for us the universal tendency of trees to bulk up in stressed subvolumes, and to distribute stresses uniformly through their entire volume. To be able to put the axiom of uniform stresses (a solid mechanics design principle) under the same theoretical roof as the minimization of global ow resistance (a uid mechanics design principle) is a fundamental development in the theory of design in nature. (iii) Would this be much different if raindrops were spherical and not aerodynamically shaped? No, in fact drops start out spherical, and all sorts of random effects conspire to prevent them from falling in the way (aerodynamically) in which they would otherwise tend to fall. Things would be marginally different if all the raindrops would be spherical, however, the same random effects will prevent this uniformity of shape to occur. The global ow performance (i.e., the rate of rainfall) is extremely robust to changes and variations in the morphologies of the individuals. We have seen this in several domains investigated based on constructal theory, from the crosssectional shapes of river channels to the movement of people in urban design. Global features of ow design and ow performance go hand in hand with the overwhelming diversity exhibited by the individuals that make up the whole. Determinism and randomness nd a home under the same theoretical tent. In fact, the tree architecture is an illustration (an icon) of this duality. Pattern is discernible from a distance, so that it appears simple enough to be grasped by the mind. Diversity (chance) is discernible close up. There is no contradiction between the two, just harmony in how the individuals contribute to and benet from the global ow. Along this holistic line, we rediscover the tree as an individual shaped by the forest, and the forest as an individual shaped by the rest of the global owing environment (Fig. 1).

Acknowledgments This research was supported by the Air Force Ofce of Scientic Research based on a grant for Constructal Technology for Thermal Management of Aircraft. Jaedal Lees work at Duke University was sponsored by the Korea Research Foundation Grant MOEHRD, KRF-2006-612-D00011. References
Adler, F.R., 1996. A model of self-thinning through local competition. Proc. Natl. Acad. Sci. USA 93, 99809984. Bejan, A., 1997. Advanced Engineering Thermodynamics, second ed. Wiley, New York, pp. 798804. Bejan, A., 2000. Shape and Structure from Engineering to Nature. Cambridge University Press, Cambridge, UK.

Bejan, A., 2004. Convection Heat Transfer, third ed. Wiley, New York (Chapter 2). Bejan, A., 2006. Advanced Engineering Thermodynamics, third ed. Wiley, Hoboken, pp. 770, 779782. Bejan, A., Lorente, S., 2004. The constructal law and the thermodynamics of ow systems with conguration. Int. J. Heat Mass Transfer 47, 32033214. Bejan, A., Lorente, S., 2006. Constructal theory of generation of conguration in nature and engineering. J. Appl. Phys. 100, 041301. Bejan, A., Marden, J.H., 2006. Unifying constructal theory for scale effects in running, swimming and ying. J. Exp. Biol. 209, 238248. Bejan, A., Merkx, G.W. (Eds.), 2007. Constructal Theory of Social Dynamics. Springer, New York. Bertram, J.E.A., 1989. Size-dependent differential scaling in branches: the mechanical design of trees revisited. Trees 4, 241253. Biondini, M., 2008. Allometric scaling laws for water uptake by plant roots. J. Theor. Biol. 251, 3559. Ciobanas, A., Bejan, A., Fautrelle, Y., 2006. Dendritic solidication morphology viewed from the perspective of constructal theory. J. Phys. D: Appl. Phys. 39, 52525266. Douady, S., Couder, Y., 1992. Phyllotaxis as a physical self-organized process. Phys. Rev. Lett. 68, 20982101. Frensch, J., Stendle, E., 1989. Axial and radial hydraulic resistance to roots of maize (Zea mays L.). Plant Physiol. 91, 719726. Horn, H.S., 2000. Twigs, trees, and the dynamics of carbon on the landscape. In: Brown, J.H., West, G.B. (Eds.), Scaling in Biology. Oxford University Press, Oxford, UK. Keitt, T.H., Urban, D.L., Milne, B.T., 1997. Detecting critical scales in fragmented landscapes. Conserv. Ecol. 1 (1), 4 (online). Kozlowski, J., Konarzewski, M., 2004. Is West, Brown and Equists model of allometric scaling mathematically correct and biologically relevant? Funct. Ecol. 8, 283289. Landsbert, J.J., Fowkes, N.D., 1978. Water movement through plant roots. Ann. Bot. 42, 493508. Livio, M., 2002. The Golden Ratio. Broadway Books, New York. Makela, A., Valentine, H.T., 2006. The quarter-power scaling model does not imply size-invariant hydraulic resistance in plants. J. Theor. Biol. 243, 283285. McElrone, A.J., Pockman, W.T., Martinez-Vilaltra, J., Jackson, R.B., 2004. Variation in xylem structure and function in stems and roots of trees to 20 m depth. New Phytol. 163, 507517. Miguel, A.F., 2006. Constructal pattern formation in stony corals, bacterial colonies and plant roots under different hydrodynamics conditions. J. Theor. Biol. 242, 954961. Nield, D.A., Bejan, A., 2006. Convection in Porous Media, third ed. Springer, New York. Reis, A.H., 2006. Constructal view of scaling laws of river basins. Geomorphology 78, 201206. Reis, AH., Bejan, A., 2006. Constructal theory of global circulation and climate. Int. J. Heat Mass Transfer 49, 18571875. Reis, A.H., Miguel, A.F., Aydin, M., 2004. Constructal theory of ow architecture of the lungs. Med. Phys. 31, 11351140. Roose, T., Fowler, A.C., 2004. A model for water uptake by plant roots. J. Theor. Biol. 228, 155171. Ryan, M.G., Bond, B.J., Law, B.E., Hubbard, R.M., Woodruff, D., Cienciala, E., Kucera, J., 2000. Transpiration and whole-tree conductance in ponderosa pine trees of different heights. Oecologia 124, 553560. Science et Vie, 2005. La tour Eiffel a livre son equation (The Eiffel Tower delivered its equation), No. 1050, 1819 March. Shinozaki, K., Yoda, K., Hozumi, K., Kira, T., 1964. A quantitative analysis of plant form: the pipe model theory. I. Basic analyses. Jpn. J. Ecol. 14, 97105. Turner, J.A., 2007. The Tinkerers Accomplice. Harvard University Press, Harvard, MA. Tyree, M.T., 2003. Hydraulic limits on tree performance: transpiration, carbon gain and growth of trees. Trees 17, 95100. Tyree, M.T., Ewers, F.W., 1991. The hydraulic architecture of trees and other woody plants. New Phytol. 119, 345360. Urban, D.L., ONeill, V.O., Shugart Jr., H., 1987. Landscape ecology. BioScience 37 (2), 119127. Weishampel, J.F., Urban, D.L., 1996. Coupling a spatially-explicit forest gap model with a 3-D solar routine to simulate latitudinal effects. Ecol. Modell. 86, 101111. West, G.B., Brown, J.H., 2004. Lifes universal scaling laws. Phys. Today 57 (9), 3642. West, G.B., Brown, J.H., Enquist, B.J., 1999. A general model of the structure of allometry of plant vascular systems. Nature 400 (12), 664667. Zwieniecki, M.A., Thompson, M.V., Holbrook, N.M., 2003. Understanding the hydraulics of porous pipes: tradeoffs between water uptake and root length utilization. J. Plant Growth Regul. 21, 315323.

Please cite this article as: Bejan, A., et al., Unifying constructal theory of tree roots, canopies and forests. J. Theor. Biol. (2008), doi:10.1016/j.jtbi.2008.06.026

You might also like