INFORMATION TECHNOLOGY AND SUPPLY CHAIN GOVERNANCE: A CONCEPTUAL MODEL

Articles

The coexistence of two very different life-formswithin savannas, namely grasses and trees, has perplexed

ecologists for decades and has been labeled the “savannaproblem” (Sarmiento 1984). Because the lack of trees inmany grasslands is equally perplexing, we suggest that thereis actually a “savanna–grassland problem”that applies to a largeproportion of the global terrestrial landscape. To date, muchof the research on the savanna problem has taken a classical,Popperian approach, whereby data (usually derived fromsmall-scale experiments) are presented to refute or falsifytheories about processes that must have operated over longtime frames and over a range of spatial scales. In our view, thisheavy reliance on falsification is impeding progress toward anunderstanding of the ultimate factors governing savanna andgrassland distribution. In this article we depart from such afalsification approach and instead plot a fundamentally newcourse for tackling this problem: a course based on adaptiveinference (Holling and Allen 2002), consideration of scale, andrecognition of ecosystems and biomes as complex adaptivesystems (Levin 1998).

Using the analogy of a horse race, the classic hypothetico-deductive approach, in which ecologists quickly select and pursue a candidate hypothesis using a process that avoids typeI errors (false positives), is analogous to shooting most ofthe horses at the starting gate before the race gets under way.When the chosen horses fail to stand the test of time, the jockeys go back to the start and either try to revive some of

the prematurely culled horses or else begin the race overagain with different horses, but with the same rules thateliminated most of them before the race. We propose a fun-damentally different approach whereby all horses (even thosethat appear lame on first appearance) are nurtured andcoaxed to their full capacity. With this approach, there is notnecessarily a single winner, but rather a recognition that eco-logical understanding will arise from the results of races runat different scales at different times. These races may take placein different contexts, such as macroecology, microbiology, orsoil science. The difficulty of advocating this new approachis that ecologists mired within the falsification paradigm will be tempted to shoot it down before it has seen the light of day.In all likelihood, it will be up to a different generation ofecologists to take up the challenge and develop this new paradigm further.

Many hypotheses have been put forward to explain the savanna–grassland problem (box 1). Only one, however—namely, fire (Bond et al. 2003, Bond and Keeley 2005, Bond

Anthony J. Mills (e-mail: [email protected]) is a soil scientist and ecologist in

the Department of Soil Science, Stellenbosch University, Matieland, 7602,

South Africa, and the South African National Biodiversity Institute, Claremont,

7735, South Africa. Kevin H. Rogers and Marc Stalmans are ecologists, and

Ed T. F. Witkowski is a plant ecologist, in the School of Animal, Plant and

Environmental Sciences, University of the Witwatersrand, PO Wits, 2050,

Johannesburg, South Africa. © 2006 American Institute of Biological Sciences.

A Framework for Exploring theDeterminants of Savanna andGrassland Distribution

ANTHONY J. MILLS, KEVIN H. ROGERS, MARC STALMANS, AND ED T. F. WITKOWSKI

An understanding of the factors governing grass–tree coexistence in savannas and exclusion of trees in grasslands remains elusive. We contend that progress in understanding these factors is impeded by a reliance on a falsification approach and an excessive concern over type I errors (false positives), which results in premature rejection of hypotheses, inadequate attention to scale, and a miring rather than galvanizing ofecological discussions. An additional hindrance to progress may be that investigations tend to focus on processes within either savannas or grasslands,while ignoring the boundary between the two. We propose a new scientific framework for identifying determinants of savanna and grassland distribution, which advocates (a) the recognition of ecosystems and biomes as complex adaptive systems, (b) a scientific methodology based on adaptive inference, and (c) explicit consideration of patch boundaries at various scales. Analysis of processes operating at dynamic savanna–grassland boundaries should permit better separation of ultimate from proximate factors controlling grass–tree interactions within the individual biomes. The proposed savanna–grassland framework has potential for application in other areas of ecology facing similar problems.

Keywords: boundary, scale, ecotone, nutrient limitation, savanna problem

www.biosciencemag.org July 2006 / Vol. 56 No. 7 • BioScience 579

Dow

nloaded from https://academ

ic.oup.com/bioscience/article-abstract/56/7/579/234338 by guest on 14 April 2019

et al. 2005)—appears to be considered seriously at present. Ourpurpose is not to dismiss this hypothesis, but rather to pointout potential contradictions and flaws in the assumptions usedto support it, and to examine why other hypotheses havefailed to gain support. Our overall aim is to restart the racewith a new set of conventions and philosophies, call the fa-vorite back to the start, resuscitate the injured, and finally en-courage the birth of some new foals.

A challenge for ecologistsConservation managers rely on ecology to provide answersto problems encountered on the ground, yet answers are of-ten conspicuously absent. When it comes to savanna andgrassland conservation, an understanding of the fundamen-tal processes governing tree abundance in a landscape wouldprobably help managers to tackle three critical questions: (1)Why are grasslands largely treeless? (2) Why are tree densi-ties increasing in many savannas? and (3) Why are trees encroaching into many grasslands? To take South Africa as a

case study, fire and frost are often cited as the causal agentsmaintaining largely treeless landscapes on the Highveld(Acocks 1953, Bond et al. 2003), yet numerous tree species,both within South Africa (e.g., Acacia karroo, Protea caffra, Pro-tea roupelliae, and Leucosidea sericea) and worldwide, are ca-pable of withstanding frequent, intense fires as well as frost(Rundel 1981). The lack of trees in South African grasslandshas also been attributed to accidents of evolutionary history,whereby trees tolerant of both frequent grass fires and coolerclimates (such as eucalypts) did not evolve in the region(William J. Bond, Department of Botany, University of CapeTown, Cape Town, South Africa, personal communication, 30March 2005). In the face of imperfect knowledge of what con-trols grass–tree interactions, it is difficult to assess the role ofchance and of environmental factors in shaping grasslands andsavannas.

Problems of scale also emerge when trying to answer thequestions posed above. Results from grass–tree plot-scaleexperiments (hundreds to thousands of square meters) within

Articles

580 BioScience • July 2006 / Vol. 56 No. 7 www.biosciencemag.org

Humans. According to one early hypothesis, grasslands were anthropogenically derived and maintained by fire over the last 1000years (Acocks 1953). This view has been discounted because of the great diversity and large degree of endemism within thesegrasslands (McKenzie 1989), and because palynological data show the presence of this biome since the late Pleistocene (>10,000years before the present; Meadows and Linder 1993), well before cultivation (Feely 1987).

Waterlogging. Another hypothesis suggests that regolith structures (e.g., shallow, impermeable layers or illuvial B horizons), whichcause waterlogging in the wet season and yet reduce water storage capacity in the dry season, exclude trees from grasslands (Tinley 1982). Although grasses are often dominant on poorly drained soils within savannas, this contention is incompletelysupported, as grasslands do occur extensively on well-drained, loose, deep, sandy soils (Werger and Coetzee 1978).

Frost. A third hypothesis proposes that frost excludes woody elements across large parts of the grassland biome (Acocks 1953).Although frost limits the pool of tree species available for colonization of grassland, it does not seem to be an adequate explanation for their exclusion, because there are many exceptions, such as Acacia karroo, Rhus lancea, Leucosidea sericea, andother woody plants that tolerate frost.

Fire. A more widely accepted view is that frequent and intense fires (promoted by specific climatic conditions) exclude trees fromthe grassland biome. High productivity (as a result of relatively high rainfall), seasonal growth, and frost curing result in highfuel loads, which, together with a high incidence of lightning, sustain frequent and intense fires. Fire frequency is likely to increase with rainfall (Van Wilgen et al. 2004), but O’Connor and Bredenkamp (1997) noted that invoking fire as the main agentexcluding trees from grasslands is “possibly inadequate given the occurrence of satellite grasslands not dependent on fire andsome savannas well adapted to high fire frequencies.” Coetzee and colleagues (1994) also noted that the fire regimes ofsavannas and grasslands north and south of Pretoria, respectively, do not differ.

Nutrient availability. The water content of topsoils (0 to 10 centimeters) during the growing season is greater in grasslands thanin adjacent, drier savannas. One hypothesis is that this promotes continuous mineralization of soil organic matter and releaseof nutrients, which renders grasses more competitive than woody species, irrespective of fire regimes (Mills 2003). Dominanceof grasses over woody species, even in the absence of fire, is not without precedent (Knoop and Walker 1985). Peters (2002) modeled grass–shrub interactions from field observations in semiarid New Mexico and predicted that increased summer precipitation would result in dominance of the grass Bouteloua eriopoda over the shrub Larrea tridentata. The grass was morecompetitive on physiological grounds, and its dominance was not related to fire. Similarly, Palmer and colleagues (1999) notethat in the South African Karoo, dominance of shrubs over grasses increases as the coefficient of variation of mean annual rainfall increases.

Box 1. Hypotheses explaining the lack of trees in South African grasslands.

Dow



a patch of savanna may contradict results of broadscale stud-ies across biomes (hundreds of square kilometers). For ex-ample, frost may be more frequent in grasslands than insavannas at a biome scale in South Africa, but in some local-ities, as a result of cold air drainage, frost is more frequent onlower wooded slopes than on the upper grassy slopes. Goszand Sharpe (1989) similarly noted that biome delineation isoften correlated with large-scale climatic features, whereas fine-scale ecotones can be determined by site-specific character-istics such as soil discontinuities.

The challenge for ecologists is to identify processes oper-ating at different scales and to differentiate the drivers fromthe modifiers of biome structure. Can factors such as fire orherbivory, for example, change a savanna into a grassland orprevent trees from encroaching into grasslands, or are thesefactors only “modifiers” (Stott 1991), with the ultimate dri-ver of biome structure being climate? Contingency may alsoplay a role, whereby the vegetation structure of a particularlandscape is dependent on a unique set of interacting factors(McNaughton 1983). Subtle abiotic differences and bioticinteractions within apparently similar landscapes couldthereby result in vastly different vegetation structures.

Levin (1998) observed that traditional approaches in ecol-ogy are inadequate for broadscale questions, such as what fac-tors determine biome distributions, because of a dividebetween population and ecosystem scientists. He suggestedthat a way forward is to recognize ecosystems as complex adap-tive systems, in which patterns at higher levels emerge fromlocalized interactions and selection processes operating atlower levels. Such systems are nonlinear, with historical de-pendency and multiple possible outcomes of dynamics. In ad-dition, the systems have been assembled from parts that haveevolved over longer timescales and broader spatial scalesthan the current system or biome. Levin (1998) proposed sev-eral pertinent questions for determining the degree to whichsystem features are determined by environmental conditionsor by self-organization. These include, among others, (a) Arepatterns of biodiversity distribution and organization uniquelydetermined by local conditions, or are they historically andspatially contingent? (b) How do ecosystems become as-sembled over time, particularly with respect to evolutionaryprocesses? and (c) What are the relationships between ecosys-tem structure and functioning?

We propose that to analyze the determinants of grasslandand savanna distributions within the context of complexadaptive systems and Levin’s (1998) questions, a new, broad-based and integrative conceptual framework is required.Frameworks serve as scientific maps for new areas of en-deavor and show how facts, hypotheses, models, and expec-tations are linked, thereby indicating the scope to which ageneralization or model applies (Pickett et al. 1999). They alsoencourage interdisciplinary interaction at appropriate scalesand help to order phenomena and material, thereby reveal-ing patterns (Rapport 1985). No such framework has been de-veloped for the savanna problem, because most ecologists havepersisted in the pursuit of a single cause for the distinction

between the two biomes. The framework we propose shouldallow ecologists to separate ultimate from proximate deter-minants at different scales. It is based on principles of land-scape ecology (Gosz and Sharpe 1989) and adaptive inference.The framework emphasizes the potential importance of (a)scale, (b) processes operating across boundaries, (c) confir-matory data, and (d) the development of multiple lines of rea-soning.

Separating ultimate from proximate factors in the grassland–savanna problemThe perplexing absence of trees in South African grasslands(where mean annual rainfall exceeds 700 millimeters [mm]in many parts) is a prime example of the savanna–grasslandproblem. According to Tainton and Walker (1993), “Theinteraction of rainfall, temperature (particularly frost), fire andsoil type determines the type of vegetation, but it’s not alwaysclear how or why some areas are pure grasslands and othersnot” (p. 271).

O’Connor and Bredenkamp (1997) concluded that thedistribution of grasslands is governed by a “subtle interplayof climate, topography, fire and grazing.”The difficulty of iso-lating the factors and subtle interplays determining vegetationstructure is not, however, restricted to grasslands and savan-nas. Orians and Solbrig (1977) noted three decades ago that“predictive theories about community structure and func-tioning are nearly absent in ecology”(p. 254), a statement thatstill largely rings true today.

A variety of factors is likely to affect the development of agrassland or savanna (figure 1). Feedback effects betweenseveral of the components are evident. Separating proximatefrom ultimate factors is difficult. A chicken and egg problemoften develops. Fire, for example, may be a proximate factor,an inevitable result of the combination of dry grass and light-ning. The presence of large herds of herbivores such as blackwildebeest (Connochaetes gnou), springbok (Antidorcas mar-supialis), and extinct large mammals such as the giant buffalo(Pelorovis antiquus) and giant hartebeest (Megalotragus priscus;Klein 1984) during the evolution of southern African grass-lands may also have been a proximate factor, given that graz-ers are likely to be attracted to flushes of grass after fire. Theclassification of biomes and vegetation types tends to per-petuate rather than alleviate the chicken and egg problem. Thisis because various abiotic and biotic characteristics (e.g., fre-quency of fire, cover of grass, incidence of frost, hydrologi-cal status of soils) are used to divide the landscape intodefined units, and the abiotic characteristics are often sub-sequently assumed to be causal.

Hypotheses of grass–tree coexistence also suffer fromchicken and egg problems. Demographic-bottleneck models,such as the storage effect hypothesis (Higgins et al. 2000), con-tend that frequent fires, competition from grass, and soilmoisture limitations usually prevent the recruitment of treeseedlings in savannas, and that adult trees store the potentialfor seedling recruitment for the few occasions when abioticconditions are suitable. Competition-based models, by

Articles


Dow



contrast, propose that trees and grasses coexist because of theirdifferential ability to acquire and partition resources (seeWalter [1971] for an explicit example of classic niche sepa-ration through separate rooting zones). Both models proposethat one main factor or several interacting factors determinea particular ecosystem state. Yet the factors are not indepen-dent of the state, and consequently there is a danger of circularreasoning. Grass competition, fire frequency, and soil mois-ture are, for example, highly dependent on the amount of grassand tree biomass.

Effects of fire exclusion on vegetation structure have recentlybeen modeled using dynamic global vegetation models(DGVMs). Results from the Sheffield DGVM suggest that “vastareas of humid C4 grasslands and savannas, especially inSouth America and Africa, have the climate potential to formforests”(Bond et al. 2005), and that if fire were excluded, mostof the eastern half of South Africa would be “dominated bytrees instead of grasses”(Bond et al. 2003). Findings from long-term fire exclusion experiments in grasslands and savannasappear to support the model results, in that woody biomassusually increases with a decrease in fire frequency. The pos-sible mechanisms by which fire excludes trees from grasslandsand the potential role of grass–tree competition require somediscussion.

The effects of fire are likely to be considerably more com-plex than the direct damage caused to living tissues. Fire in-fluences subsequent (postfire) nutrient availability, soil water

content, soil temperatures,rates of mineralization, andlight availability (Stock andLewis 1986, Blair 1997,Knapp et al. 1998), all ofwhich are likely to influencethe competitive ability ofgrasses and trees. For exam-ple, the mean sunlit photo-synthetic rates of several treesin a hardwood forest in Wis-consin were stimulated afterfire, probably because of anincrease in nitrogen avail-ability (Reich et al. 1990).There may also be contin-gency effects, whereby fireinteracts with soil moisture,nutrient availability, lightavailability, and grass–treecompetition in subtle, non-linear ways. As McNaughton(1983) noted with respect tothe Serengeti grasslands,“theproximate mechanisms reg-ulating species abundancesare many weak forces actingprobabilistically, so that thecumulative effects are large,

but the individual effects are minor, interactive, and uncer-tain”(p. 315). He highlighted grazing as one force dependenton many intersecting probability functions, such as speciescomposition, phenological stages of grasses, tree canopy den-sity, species of grazers, density of grazers, soil properties, fre-quency of burning, and soil water content. The influence ofeach one of these factors is likely to vary through time, andit may prove more fruitful to map and describe such dy-namic probability functions rather than pursuing static, lin-ear chains of cause and effect when examining factors thatgovern biome distribution.

Despite the apparently compelling evidence from modelsand long-term experiments supporting the present role of firein tree exclusion, an intriguing inconsistency remains: Whydid fire-tolerant tree floras not evolve to dominate all fire-prone ecosystems, when trees from families such as Cae-salpiniaceae, Fagaceae, Pinaceae, and Myrtaceae cover vastareas of the globe? Trees show great plasticity with respect tofire tolerance, with adaptive features such as corky bark (e.g.,Quercus suber), seedlings with protective grassy covering(e.g., Pinus palustris), lignotubers, and epicormic buds (e.g.,Acacia and Eucalyptus spp.). Furthermore, convergent evo-lution shows that plants in similar environments on differ-ent continents evolve into remarkably similar growth forms(Orians and Solbrig 1977, Cody and Mooney 1978, Cowlingand Witkowski 1994). The principle of convergent evolu-tion seems, however, to falter when it comes to fire-prone flo-

Articles


Figure 1. Hypothetical relationships between factors likely to affect the evolution of a grasslandor savanna. Note that there are feedback effects from several of the center components; even cli-mate is not immune from feedback. There are also numerous feedback effects between the cen-ter components (e.g., between fire and herbivory pressure, and between fire andmineralization) not indicated in the diagram.

Dow



ras, with some fire-prone systems being dominated by fire-tolerant trees and others by fire-tolerant grasses. This is a ma-jor inconsistency when invoking fire as a primary determinantof vegetation structure across large parts of the planet. An al-ternative view that may resolve the issue is that vegetationstructure is primarily governed by climate, and fire is proxi-mate and incidental.

If the climate in a fire-prone environment favors the dom-inance of low-growing herbaceous vegetation for physiolog-ical reasons, it is conceivable that through evolutionary timea positive feedback could develop, whereby the vegetation ben-efits from fire because nutrients are returned to topsoils ormoribund material is removed, or both. The vegetation is likelyto evolve attributes that promote fire, and the competitive abil-ity of the vegetation may become largely dependent on fire.Woody plants invading herbaceous vegetation are likely to belimited by the competitive ability of the herbaceous vegeta-tion (through competition for resources such as light, water,and nutrients), but will also probably not be tolerant of thespecific fire regime that coevolved with the herbaceous veg-etation. If, however, climate in another fire-prone environmentfavors the growth of trees, then through evolutionary time,a tree flora may emerge that is tolerant of or promotes fire,and this flora may even come to depend on nutrient-rich ashbeds for the germination and successful recruitment ofseedlings. This could explain the wide range in vegetationstructure across fire-prone ecosystems and the increase inwoody vegetation in fire-exclusion experiments. Herbaceousvegetation that coevolved with fire over millions of years is un-likely to be competitive when fire is removed. Removing firefrom fire-dependent grasslands is likely to reduce grass vigorand thereby create an artificially competitive advantage fortrees. Consequently, it may be premature to conclude fromthe results of fire-exclusion experiments (Bond et al. 2005) thatfire is more important than climate in shaping vegetationstructure.

Indeed, the wide range of fire regimes evident across south-ern African vegetation types suggests that, through evolution-ary time, plants exerted a strong influence on the fire regimerather than the other way around. Subtropical thicket in the Eastern Cape, South Africa, for example, occurs in a warm,semiarid climate (250 to 650 mm mean annual rainfall) witha large potential for growth of grass,yet can exclude fire becauseof the relative paucity of grass cover and the predominance ofsucculent shrubs such as Portulacaria afra (Vlok et al. 2003).Rainforest patches in northern Australian savannas provideanother example of vegetation that excludes fire. Bowman andcolleagues (2004) showed that rainforest patches in some regionsoccur on the most clayey soils.They suggested that the rainforestspecies thrive in these fertile, moist soils and can persist by preventing grass recruitment and excluding fire.

Much of our understanding of fire–vegetation interactionshas been derived from experimental studies at the plot scale.It is difficult to use experiments (e.g., fire exclusion, watering,nutrient addition) to separate ultimate from proximate causesbecause of mismatches in temporal scale. Savanna and

grassland biomes did not develop under the conditions im-posed in such experiments. If such conditions had been pre-sent through evolutionary time, and yet the ultimate factorswere still in place, it is conceivable that similar biomes wouldhave arisen, but with different ecological processes. If fire, forexample, had been excluded from the South African Highveldfor several million years, a grassland might have developed thatrelied on microbial decomposition and herbivores, ratherthan fire, for nutrient cycling. It is conceivable that today’sgrasslands are a response to a certain suite of abiotic condi-tions in soils (e.g., water content, degree of aeration, specificmacro- or even micronutrient content) that promote grasses,and thereby enable them to dominate and exclude trees, ir-respective of the fire regime.Although experiments can be in-structive, it is increasingly acknowledged that withoutcomplementary emphasis on large-scale phenomena throughtime and space, it is difficult to determine which results re-flect idiosyncrasies of individual treatments, species, or siteconditions, and which reflect the operation of more univer-sal processes (Brown 1995).

An alternative explanation for the lack of fire-toleranttrees in South African grasslands is an accident of evolu-tionary history. It may be that tree genera that are extremelytolerant of fire in other parts of the world (e.g., Eucalyptus,Melaleuca, Banksia, Pinus, Quercus) by chance did not migrateto or evolve in South Africa (Richardson et al. 1992). Acker-ley (2004) notes that the presence of a species within anecosystem does not necessarily reflect a process of natural se-lection and adaptation, because many species may have mi-grated into the system from surrounding areas. Nevertheless,convergent evolution (e.g., such as occurs across Cactaceae andEuphorbiaceae) suggests that plant growth forms are flexible,highly responsive to abiotic conditions, and unlikely to be lim-ited by the available gene pool. Furthermore, many tree fam-ilies present in South African grasslands (in forest patches,along drainage lines, or on rocky slopes) have fire-tolerant treegenera in other parts of the world (e.g., Proteaceae and Myrtaceae). This suggests that genetic limitations for the development of a fire-tolerant tree flora were unlikely, and we consequently propose that the principle of convergent evolution warrants an in-depth analysis in the context of thesavanna–grassland problem. Furthermore, we caution that the“accident of evolutionary history” explanation is potentiallydangerous when deployed within the falsification paradigm,because in its extreme form, no abiotic explanations of veg-etation structure are necessary, given that all observed dif-ferences between biomes can be explained away by chance.Unfortunately, this explanation starts from an inherentlydomineering position (in that it can be invoked to explain anypattern), and it inevitably stultifies or nullifies research focusingon abiotic explanations for ecological differences at largescales (e.g., across continents).

It is evident that scientists are no closer than Tainton andWalker (1993) were 13 years ago to solving the question of whysome areas are pure grassland and others are not. The prin-ciple of convergent evolution suggests that fire-tolerant wood-

Articles


Dow



lands should have developed in many pure grasslands. Small-scale experiments are unlikely to separate ultimate fromproximate factors governing vegetation structure, becauseof mismatches in temporal scale, and research within the in-dividual biomes tends to run up against problems of circu-lar reasoning. Ecologists need to regroup and develop a newstrategy to deal with these problems and to tackle this in-tractable question.

Toward a new framework that encompasses complexadaptive systems, scale, boundaries, and adaptiveinferenceFrameworks facilitate linkages between different paradigmsand prepare common ground for scientists using different con-ceptual approaches. In this light, we discuss below several dif-ferent approaches that we advocate using in tandem forhoming in on the factors ultimately determining biome dis-tributions.We suggest that new insights are likely to arise fromthe mere process of combining methodologies and ideas. Inthe words of Jacob (1977),“Novelties come from previouslyunseen association of old materials. To create is to recombine”(p. 1163).

Complex adaptive systems and scale. The identification of ul-timate and proximate factors governing vegetation structurerequires consideration of complexity, chance, evolution, andscale. Levin (1998) brings this diverse range of issues to-gether under the banner of “complex adaptive systems” andthereby provides a useful starting point for investigating the

savanna–grassland problem. Tackling the different issues willrequire research on several fronts. To date, most grassland andsavanna research has taken place within the biomes. We sug-gest that a new focus on scale and processes operating atgrassland–savanna boundaries will generate additional insightsinto factors influencing the individual biome structures anddistributions. A scaled approach was recently used by Gillson(2004) in an East African savanna. Using the concept ofhierarchical patch dynamics (Wu and Loucks 1996), Gillsonsuggested that different ecological processes—namely, soiltype, disturbance by fire or herbivory, and climate—determinetree abundance over hundreds of years at micro, local, andlandscape scales, respectively. A similarly scaled perspectiveon boundaries (Gosz and Sharpe 1989), from individualplant patches to the biome boundary (figure 2), could bevery useful in separating ultimate and proximate causation ingrassland and savanna differentiation.

Using boundaries to separate proximate from ultimate causation. Boundary research has principally aimed at doc-umenting and understanding boundaries. We suggest, how-ever, that boundaries can be used for understanding ecologicalprocesses shaping the communities or biomes on either sideof the boundary. This is a major conceptual departure bothfrom current boundary research and from “within-biome”re-search. The potential relevance of boundaries to within-biome research is highlighted by Cadenasso and colleagues(2003), who note that because the patches that the boundaryseparates are distinguished from each other by some defin-

Articles


Figure 2. A hierarchical patch-dynamic approach to the study of the grassland–savanna boundary. Factorsgoverning nitrogen mineralization and boundary patchiness highlight the effect of scale on ecologicalprocesses. Abbreviations: cm, centimeter; km, kilometer.

Dow



ing characteristic, the gradient in that characteristic is steeperin the boundary than in either of the neighboring patches. Thissuggests that ultimate factors governing grassland and savannadistribution are likely to show sharp changes over the bound-aries between the vegetation types and are likely to controlshifts in boundaries over time. By contrast, we suggest thatproximate factors within grasslands or savannas are likely tobe more loosely tied, and may or may not show changes atboundaries (figure 3).

Scrutiny of processes occurring across boundaries cantherefore potentially identify proximate factors and by aprocess of exclusion develop a shortlist of potential ultimatefactors (figures 3, 4). Boundaries that are shifting through timeare likely to be particularly powerful in this regard (figure 4).Potential ultimate factors driving a boundary shift (e.g., soilmoisture in surface soils) would, for example, be expected tobe closely tied to the present boundary line and might evenbe slightly ahead of it (figure 4). In contrast, potential prox-imate soil properties (e.g., soil organic matter, pH, base sta-tus and redox status at depth) may lag behind the boundaryshift. It is likely that processes and differences across bound-aries, as yet not envisaged, would be uncovered with a sys-tematic examination, especially if conducted at differentscales.

Although ecological properties such as soil type can changeacross boundaries (Cole 1992), it cannot be assumed thatabrupt changes in grassland–savanna boundaries will neces-sarily be reflected in abiotic variables. This is because ecological

boundaries may also be the product of nonlinear behavior,whereby gradual changes in environmental variables elicit dra-matic changes in population and community variables whenthresholds are reached (Fagan et al. 2003). Boundaries mayalso have properties that are unique to the boundary (Faganet al. 2003), thereby complicating the isolation of ultimate fac-tors. As an example, the woody species Maesa lanceolata oc-curs mostly at the boundary between Afromontane forests andgrasslands but is largely absent from either of the two for-mations. We note that there are likely to be multiple causesfor changes in vegetation structure across boundaries and con-sequently advocate that research is conducted on numerousfronts using numerous hypotheses.

Adopting an adaptive inference approach with multiple linesof reasoning. Grassland and savanna research has focused pre-dominantly on experimentation, falsification, and avoidanceof type I error, and has tended to generate inconclusive bi-variate empirical studies and factorial experiments (Weiheret al. 2004). The main drawback of this use of Popperian phi-losophy is that, because it does not deal well with contingencyand multiple causality, it results in type II errors (false neg-atives), shutting down promising avenues of research. Wesuggest that adaptive inference (Holling and Allen 2002) is amore suitable approach for trying to separate the proximateand ultimate factors shaping vegetation structure and foranalyzing systems with complex (nonlinear) dynamics.Adap-tive inference draws on both experimental and correlative data.

Articles


Figure 3. Changes in ecological processes or factors across a grassland–savanna boundary. Abbreviation: cm, centimeter.

Dow



It specifically tries to avoid type II errors in the early phasesof advancing understanding by keeping numerous hypothe-ses alive and actively seeking confirmatory data for each hy-pothesis. Research at this early stage would focus on buildingcases for each hypothesis, rather than trying to refute them.Given that processes are likely to vary considerably at differ-ent scales of analysis and experimentation, the rejection of anyhypothesis in the early stages of investigation is probably un-warranted.

The investigation of numerous lines of reasoning at differentscales cannot be achieved by an individual scientist, andtherefore requires a collective and focused effort by the ap-propriate ecological community (Pickett et al. 1994). Adap-tive inference has been shown to be effective (Levin 1998,Holling and Allen 2002) because, as understanding improvesand some lines of reasoning are slowly winnowed out, the mostparsimonious explanations are identified. Ecologists can thenset quantitatively more specific, more precise, and narrowertests of components of the lines of reasoning. This in turn en-courages a switch to more deductive tests (e.g., manipulativeor controlled experiments, null hypothesis testing) that avoidtype I errors.

New foci for grassland–savanna researchMcNaughton (1983) noted that the composition of vegeta-tion is probably governed not by a single overpowering forcebut rather by constellations of weak forces. Following this

approach, the outcome of grass–tree competition may be a function of numerous interacting factors, such as fire, soilwater content, availability of energy (sunlight), nutrients,frost, and microbial activity. Although this constellation offorces may operate in complex ways and at different scales (see,e.g., factors affecting nitrogen mineralization shown in figure2), the art of investigating such complexity may lie in the iden-tification of those processes and agents that play the greatestrole in shaping the system (i.e., the ultimate factors). To thisend, we suggest several new research foci in the paragraphsbelow.

First, the role of fire in the exclusion of trees from grass-lands warrants critical analysis. Data on fire intensity and fre-quency gradients across grassland–savanna boundaries arelikely to be instructive, as are data on the relationships betweenfire intensity and survival of trees at different life historystages at the boundary (figure 5). Interactions between fire andgrass–tree competition are largely unexplored. For example,the increased nutrient availability after fire as a result of in-corporation of ash into topsoils (Stock and Lewis 1986) andincreased rates of mineralization (Knapp et al. 1998) could in-crease grass vigor and enable the dominance of grasses overtree seedlings, whether fire damages the tree seedlings ornot. Experiments that shield tree seedlings from fire, yet al-low adjacent grass plants to burn, are a possible way of teas-ing out such interactions. Furthermore, considerably moreinformation on the effects of herbaceous vegetation on woody

Articles


Figure 4. Factors that may lag, keep abreast of, or move ahead of dynamic grassland–savannaboundaries and thereby provide clues for separating proximate from ultimate processes governing biome distributions.

Dow



plant growth, recruitment, seed production, and seedlingestablishment is required to separate the effects of com-petition on grassland and savanna structure from those ofdisturbance (House et al. 2003).

Second, determining the effects of climate on nutrientavailability in topsoils (Hooper and Johnson 1999), and theimplications of these effects for grass–tree competition,could be revealing. Savanna models have tended to focuson water rather than nutrients as the main resource lim-iting growth (Sankaran et al. 2004). The importance of nu-trient availability with respect to grass–tree interactions andvegetation structure is intuitive yet largely unexamined(Walker and Langridge 1997, House et al. 2003). Recent ex-perimental work shows that nitrogen fertilization in semi-arid savannas can result in an increase in grass vigor andreduced tree seedling survival (David Ward, School ofBiological and Conservation Sciences, University ofKwaZulu-Natal, Scottsville, South Africa, personal com-munication, 20 December 2005). This suggests that grassesmay dominate in landscapes where the supply of nutrientsfrom mineralization is sustained above a certain thresh-old during periods of plant growth. Mineralization is con-trolled by numerous factors operating at different scales(figure 2). At the macro or biome scale, the desiccation oftopsoils, which occurs frequently during the growing season in South African savannas, may reduce grass com-petitiveness not only through reduced water availability butalso through a reduced nutrient supply from mineralization.If the dominance of grasses over trees is largely dependent ona sustained supply of nutrients, this desiccation, or “switch-ing off the nutrient pump,” may be a key factor enablingtrees to get a foothold in savanna systems.

Third, the jury is still out as to whether fire and herbivoryare modifiers (Stott 1991, Sankaran et al. 2004) or primary de-terminants (Bell 1982, Higgins et al. 2000, Bond et al. 2005)of savanna structure. Even extremely intense fire regimes orbrowsing pressure have, to our knowledge, never converteda savanna into a treeless grassland (Laws 1970). This suggeststhat these factors are modifiers rather than drivers of thesystem. But is the same true at the boundary? It is conceiv-able that manipulation of fire or herbivory is more likely toexclude trees at grassland–savanna boundaries than within thesavanna system itself. The vulnerability of trees to herbivoresand fire may, for example, be a function of distance from theboundary (Palmer et al. 2003), and may correlate with spe-cific environmental factors such as the nutrient supply fromtopsoils or the distribution of water in soil profiles.

Finally, the vertical distribution of water in the soil profilehas been invoked as a key determinant of grass–tree ratios insavannas, with tree biomass increasing as the amount of sub-soil water increases (Walter 1971, Walker and Noy-Meir1982). The focus of these classic equilibrium models of sa-vannas has been on soil water content in topsoils relative tosubsoils. Effects of water content on vegetation structure ingrasslands and savannas may, however, also play out at anotherscale. The top few centimeters of mineral soil, or pedoderm

(Mills and Fey 2004), tends to have more soil organic matterthan the deeper layers within the conventionally defined Ahorizon, and consequently mineralization is often dispro-portionately greater in this surface layer than below it (Woods1989, Purnomo et al. 2000). Seasonal and daily fluctuationsin the water content of the pedoderm (such as occur underconditions of mist and dew) are likely to influence the avail-ability of nutrients, and therefore to affect grass–tree com-petition. A pedoderm that seldom dries out during thegrowing season, for example, may be necessary for grasses todominate tree seedlings. Ellery and colleagues (1991) notedthat mean temperature during the growing season is gener-ally lower in South African grasslands than in savannas. Doesthis climatic pattern influence biome structure through aconstellation of interacting factors, such as the water contentof the pedoderm, rates of soil organic matter mineraliza-tion, and rates of photosynthesis in grass leaves? Our under-standing of grass–tree interactions is likely to improve ifresearch paradigms are shifted to a much finer vertical reso-lution when examining the uppermost part of soil profiles.

We hope that housing these different research foci withinthe proposed framework of a scaled exploration at the bound-ary, with an emphasis on adaptive inference, will providenew vigor and direction for researchers tackling the vexing savanna–grassland problem.

The challenge concludedIn this article we have explored the many theories of grass–treeinteraction that have been invoked to explain the differencesbetween grasslands and savannas. We have also introducedsome new ideas, and different ways of thinking about old ideas,

Articles


Figure 5. A dynamic grassland–savanna boundary near Barber-ton, Mpumalanga, South Africa. Fire frequency is likely to be similar across the boundary, although differences in grass speciesflammability may influence fire intensity at a microscale andthereby affect tree seedling establishment. Photograph: Marc Stalmans.

Dow



into the debate. The list of ideas and theories can be sum-marized as follows: One system is favored over another as aresult of (a) the present suite of abiotic conditions; (b) abi-otic history; (c) chance-driven evolutionary history; (d) aparticular regime of fire, frost, herbivory, nutrients, or water;(e) contingent interactions between some or all of theseregimes; and (f) complex adaptive processes involving someor all of the above.

Each of these lines of reasoning has merit as a working hy-pothesis, model, or theoretical construct, but our currentunderstanding does not provide a useful basis for acceptingone over the other, or for choosing one or a combination asthe most profitable avenue of scientific exploration. There arethree main reasons for the limitations in current under-standing of the savanna–grassland problem: First, previousstudies of the problem have been conducted within savannaor grassland systems, making it almost impossible to separateultimate drivers from proximate modifiers in the explorationof causation. Second, the predominance of a falsificationparadigm behind the design of previous studies has led to pre-mature debates over which single factor is the dominantcause of the differences between grasslands and savannas.Third, the lack of attention to spatial and temporal scale,particularly evident in experimental and falsification ap-proaches to science, has lead to confusion between cause andeffect.

The challenge ahead is therefore to acknowledge the com-plexity of the practical, analytical, and paradigmatic problemsfaced when exploring ecosystem-level questions and to adopta fresh approach. We recommend that a community of sci-entists from different disciplines explore and winnow theworking models by (a) adopting an adaptive inference par-adigm of scientific exploration; (b) applying analytical ap-proaches that allow type I and type II errors at appropriatejunctures of the scientific process; and (c) focusing scientificeffort at the grassland–savanna boundary, where the conse-quences of scale, history, contingency, and adaptive feedbackare most likely to be unraveled. Such a scaled, adaptive, andinterdisciplinary approach will provide a challenge for futuregenerations of ecologists and collaborating scientists.

AcknowledgmentsThe authors extend their grateful thanks to the Mellon Foun-dation for funding the research.

References citedAckerley DD. 2004. Adaptation, niche conservatism, and convergence:

Comparative studies of leaf evolution in the Californian chaparral.American Naturalist 163: 654–671.

Acocks JPH. 1953. Veld types of South Africa. Memoirs of the Botanical Survey of South Africa 28: 1–192.

Bell RHV. 1982. The effect of soil nutrient availability on community structure in African ecosystems. Pages 193–216 in Huntley BJ,Walker BH,eds. Ecology of Tropical Savannas. Berlin: Springer-Verlag.

Blair JM. 1997. Fire, N availability, and plant response in grasslands: A testof the transient maxima hypothesis. Ecology 78: 2359–2368.

Bond WJ, Keeley JE. 2005. Fire as a global “herbivore”: The ecology and evolution of flammable ecosystems. Trends in Ecology and Evolution 20:385–394.

Bond WJ, Midgley GF, Woodward FI. 2003. What controls South African vegetation—climate or fire? South African Journal of Botany 2003:79–91.

Bond WJ, Woodward FI, Midgley GF. 2005. The global distribution ofecosystems in a world without fire. New Phytologist 165: 525–538.

Bowman DMJS, Cook GD, Zoppi U. 2004. Holocene boundary dynamics ofa northern Australian monsoon rainforest patch inferred from isotopicanalysis of carbon (14C and �13C) and nitrogen (�15N) in soil organic mat-ter. Austral Ecology 29: 605–612.

Brown JH. 1995. Macroecology. Chicago: University of Chicago Press.Cadenasso ML, Pickett STA,Weathers KC, Bell SS, Benning TL, Carreiro MM,

Dawson TE. 2003. An interdisciplinary and synthetic approach to ecological boundaries. BioScience 53: 717–722.

Cody ML, Mooney HA. 1978. Convergence versus non-convergence inMediterranean-climate ecosystems. Annual Review of Ecology and Systematics 9: 265–321.

Coetzee JP, Bredenkamp GJ, van Rooyen N. 1994. An overview of the physical environment and vegetation units of the Ba and Ib land typesof the Pretoria–Witbank–Heidelberg area. South African Journal ofBotany 60: 49–61.

Cole MM. 1992. Influence of physical factors on the nature and dynamicsof forest–savanna boundaries. Pages 63–75 in Furley PA, Proctor J,Ratter JA, eds. Nature and Dynamics of Forest–Savanna Boundaries.London: Chapman and Hall.

Cowling RM, Witkowski ETF. 1994. Convergence and non-convergence ofplant traits in climatically and edaphically matched sites in MediterraneanAustralia and South Africa. Australian Journal of Ecology 19: 220–232.

Ellery WN, Scholes RJ, Mentis MT. 1991. An initial approach to predictingthe sensitivity of the South African grassland biome to climate change.South African Journal of Science 87: 499–503.

Fagan WF, Fortin M-J, Soykan C. 2003. Integrating edge detection and dynamic modeling in quantitative analyses of ecological boundaries.BioScience 53: 730–738.

Feely JM. 1987. The Early Farmers of Transkei, Southern Africa, before A.D.1870. Oxford (United Kingdom): British Archaeological Reports.Cambridge Monographs in African Archaeology 24. British ArchaeologicalReports International Series 378.

Gillson L. 2004. Evidence of hierarchical patch dynamics in an East Africansavanna? Landscape Ecology 19: 883–894.

Gosz JR, Sharpe PJH. 1989. Broad-scale concepts for interactions ofclimate, topography, and biota at biome transitions. Landscape Ecology3: 229–243.

Higgins SI, Bond W, Trollope WSW. 2000. Fire, resprouting and variability:A recipe for grass–tree coexistence in savanna. Journal of Ecology 88:213–229.

Holling CS, Allen CR. 2002. Adaptive inference for distinguishing crediblefrom incredible patterns in nature. Ecosystems 5: 319–328.

Hooper DU, Johnson L. 1999. Nitrogen limitation in dryland ecosystems:Responses to geographical and temporal variation in precipitation.Biogeochemistry 46: 247–293.

House JI, Archer S, Breshears DD, Scholes RJ. 2003. Conundrums in mixedwoody–herbaceous plant systems. Journal of Biogeography 30: 1763–1777.

Jacob F. 1977. Evolution and tinkering. Science 196: 1161–1166.Klein RG. 1984. The large mammals of southern Africa: Late Pliocene to

recent. Pages 107–148 in Klein RG, ed. Southern African Prehistory andPaleoenvironments. Rotterdam (The Netherlands): A. A. Balkema.

Knapp AK, Conrad SL, Blair JM. 1998. Determinants of soil CO2 flux from a sub-humid grassland: Effect of fire and fire history. Ecological Applications 8: 760–770.

Knoop WT, Walker BH. 1985. Interactions of woody and herbaceous vegetation in a southern African savanna. Journal of Ecology 73: 235–253.

Laws RM. 1970. Elephants as agents of habitat and landscape change inEast Africa. Oikos 21: 1–15.

Levin SA. 1998. Ecosystems and the biosphere as complex adaptive systems.Ecosystems 1: 431–436.

Articles


Dow



McKenzie B. 1989. Medium-term changes of vegetation pattern in Transkei.South African Forestry Journal 150: 1–6.

McNaughton SJ. 1983. Serengeti grassland ecology: The role of compositeenvironmental factors and contingency in community organization.Ecological Monographs 53: 291–320.

Meadows ME, Linder HP. 1993. A palaeoecological perspective on the origin of Afromontane grasslands. Journal of Biogeography 20: 345–355.

Mills AJ. 2003. Reciprocal relationships between vegetation structure and soilproperties in selected biomes of South Africa. PhD dissertation. Universityof Stellenbosch, Bellville, South Africa.

Mills AJ, Fey MV. 2004. Frequent fires intensify soil crusting: Physico-chemical feedback in the pedoderm of long-term burn experiments inSouth Africa. Geoderma 121: 45–64.

O’Connor TG, Bredenkamp GJ. 1997. Grassland. Pages 215–257 in CowlingRM, Richardson DM, Pierce SM, eds. Vegetation of Southern Africa.Cambridge (United Kingdom): Cambridge University Press.

Orians GH, Solbrig OT. 1977. Convergent Evolution in Warm Deserts.Stroudsburg (PA): Dowden, Hutchinson and Ross.

Palmer AR, Novellie PA, Lloyd JW. 1999. Community patterns and dynam-ics. Pages 208–223 in Dean WRJ, Milton SJ, eds. The Karoo: EcologicalPatterns and Processes. Cambridge (United Kingdom): Cambridge University Press.

Palmer SCF, Hester AJ, Elston DA, Gordon IJ, Hartley SE. 2003. The perilsof having tasty neighbors: Grazing impacts of large herbivores at vegetation boundaries. Ecology 84: 2877–2890.

Peters DPC. 2002. Plant species dominance at a grassland–shrubland ecotone: An individual-based gap dynamics model of herbaceous andwoody species. Ecological Modelling 152: 5–32.

Pickett STA, Kolasa J, Jones CG. 1994. Ecological Understanding: The Na-ture of Theory and the Theory of Nature. San Diego: Academic Press.

Pickett STA, Burch WR, Grove JM. 1999. Interdisciplinary research:Maintaining the constructive impulse in a culture of criticism.Ecosystems 2: 302–307.

Purnomo E, Black AS, Smith CJ, Conyers MK. 2000. The distribution of netnitrogen mineralisation within surface soil, 1: Field studies under awheat crop. Australian Journal of Soil Research 38: 643–652.

Rapport A. 1985. Thinking about home environments: A conceptual framework. Pages 255–286 in Altman I, Werner CM, eds. Home Environments. New York: Plenum Press.

Reich RB, Abrams MD, Ellsworth DS, Kruger EL, Tabone TJ. 1990. Fire affects ecophysiology and community dynamics of central Wisconsin oak forest regeneration. Ecology 71: 2179–2190.

Richardson DM, MacDonald IAW, Holmes PM, Cowling RM. 1992. Plantand animal invasions. Pages 271–308 in Cowling RM, ed. The Ecologyof Fynbos: Nutrients, Fire and Diversity. Cape Town (South Africa):Oxford University Press.

Rundel PW. 1981. Fire as an ecological factor. Pages 501–538 in Lange LO,Nobel PS, Osmond CB, Ziegler H, eds. Physiological Plant Ecology I:Responses to the Physical Environment. Berlin: Springer-Verlag.

Sankaran M, Ratnam J, Hanan NP. 2004. Tree–grass coexistence in savannasrevisited—insights from an examination of assumptions and mechanismsinvoked in existing models. Ecology Letters 7: 1–11.

Sarmiento G. 1984. The ecology of neotropical savannas. Cambridge (MA):Harvard University Press.

Stock WD, Lewis OAM. 1986. Soil nitrogen and the role of fire as a miner-alizing agent in a South African coastal fynbos ecosystem. Journal ofEcology 74: 317–328.

Stott P. 1991. Recent trends in the ecology and management of the world’ssavanna formations. Progress in Physical Geography 15: 18–28.

Tainton NM,Walker BH. 1993. Grasslands of southern Africa. Pages 265–290in Coupland RT, ed. Natural Grasslands: Eastern Hemisphere and Résumé. Amsterdam: Elsevier.

Tinley KL. 1982. The influence of soil moisture balance of ecosystem patternsin southern Africa. Pages 175–192 in Huntley BJ, Walker BH, eds.Ecology of Tropical Savannas. Berlin: Springer-Verlag.

Van Wilgen BW, Govender N, Biggs HC, Ntsala D, Funda XN. 2004. Responseof savanna fire regimes to changing fire-management policies in a largeAfrican national park. Conservation Biology 18: 1533–1540.

Vlok JHJ, Euston-Brown DIW, Cowling RM. 2003. Acocks’Valley Bushveld50 years on: New perspective on the delimination, characterisation andorigin of subtropical thicket vegetation. South African Journal of Botany69: 27–51.

Walker BH, Langridge JL. 1997. Predicting savanna vegetation structure onthe basis of plant available moisture (PAM) and plant available nutrients(PAN): A case study from Australia. Journal of Biogeography 24: 813–825.

Walker BH, Noy-Meir I. 1982. Aspects of stability and resilience in savannaecosystems. Pages 556–590 in Huntley BJ, Walker BH, eds. Ecology ofTropical Savannas. Berlin: Springer-Verlag.

Walter H. 1971. Ecology of Tropical and Subtropical Vegetation. Edinburgh(Scotland): Oliver and Boyd.

Weiher E, Forbes S, Schauwecker T, Grace JB. 2004. Multivariate control ofplant species richness and community biomass in blackland prairie.Oikos 106: 151–157.

Werger MJA, Coetzee BJ. 1978. The Sudano–Zambezian region. Pages301–462 in Werger MJA, ed. Biogeography and Ecology of SouthernAfrica. The Hague: Junk.

Woods LE. 1989. Active organic matter distribution in the surface 15 cm ofundisturbed and cultivated soil. Biology and Fertility of Soils 8: 271–278.

Wu J, Loucks OL. 1996. From balance of nature to hierarchical patch dynamics:A paradigm shift in ecology. Quarterly Review of Biology 70: 439–466.

Articles


Dow



Back issues of BioScience are now available online for institutional andindividual subscribers through JSTOR, the not-for-profit, online digital

archive. Users can browse, search, view, download, and print the full-text PDF files of BioScience from its original inception as the AIBS Bulletin in 1951 up to the most recent five years. AIBS maintains a

five-year moving wall between BioScience content archived with JSTORand more recent BioScience content.

Institutions that participate in JSTOR’s new Biological Sciences Collection can access back issues of BioScience at www.jstor.org.

The Biological Sciences Collection will include at least 100 titles whenit is completed at the end of 2007, including the 29 journals in JSTOR’sexisting Ecology and Botany Collection. The journals in this collectionoffer greater depth in fields such as biodiversity, conservation, paleon-tology, and plant sciences, in addition to introducing new areas such as

cell biology and zoology.

Individual members of AIBS can access BioScience back issues through JSTOR at www.aibs.org/bioscience, where both individuals

and institutions can also access the most recent five years ofBioScience online.

AIBS and BioScience are proud to collaborate with JSTOR to preserveand make widely available the historic literature of our field.

BACK ISSUES OF BIOSCIENCE NOW AVAILABLE

ONLINE WITH JSTOR FOR INSTITUTIONAL

AND INDIVIDUAL SUBSCRIBERS!

Dow



Documents

INFORMATION TECHNOLOGY AND SUPPLY CHAIN GOVERNANCE: A CONCEPTUAL MODEL