J. Range Manage.56: 114-126 March 2003

Development and use of state-and-transition models forrangelands

BRANDON T. BESTELMEYER1, JOEL R. BROWN2, KRIS M. HAVSTAD3, ROBERT ALEXANDER4, GEORGECHA VEZs, AND JEFFREY E. HERRICK6

IEcologist, USDA-ARS, Jomada Experimental Range, New Mexico State University, Las Cruces, N.M. 88003, 2Cooperating Research Scientist USDA-NRCS, Jomada Experimental Range, New Mexico State University, Las Cruces, N.M. 88003. 3Supervisory Scientist, USDA-ARS, Jomada Experimental Range,'New Mexico State University, Las Cruces, N.M. 88003. 4State Range Specialist, Bureau of Land Management, 1474 Rodeo Rd.. Santa Fe, N.M. 87502, 5StateRangeland Management Specialist.. USDA Natural Resources Conservation Service, 6200 Jefferson, Albuquerque, N.M. 87109. 6Soil Scientist, USDA-ARS.Jornada EXperimental Range, New Mexico State University, Las Cruces, N.M. 88003

Abstract ships can be readily understood. A knowledge of mechanisms isclosely related to the use of ecological indicators to anticipatetransitions. We conclude that models should include 1) referencevalues for quantitative indicators, 2) lists of key indicators anddescriptions of changes in them that suggest an approach to atransition, and 3) a rigorous documentation of the theory andassumptions (and their alternatives) underlying the structure ofeach model.

State-and-transition models have received a great deal of atten-tion since the introduction of the concept to range managementin 1989. Nonetheless, only recently have sets of state-and-transi-tion models been produced that can be used by agency personneland private citizens, and there is little guidance available fordeveloping and interpreting models. Based upon our experiencesdeveloping models for the state of New Mexico, we address thefollowing questions: 1) how is information assembled to createsite-specific models for entire regions, 2) what ecological issuesshould be considered in model development and classification,and 3) how should models be used? We review the general struc-ture of state-and-transition models, emphasizing the distinctionbetween changes among communities within states (pathways)that are reversible with changes in climate and "facilitating prac-tices" (e.g. grazing management), and changes among states(transitions) that are reversible only with "accelerating prac-tices" such as seeding, shrub control, or the recovery of soil sta-bility and historical hydrologic function. Both pathways andtransitions occur, so these models are complementary. Ecologicalsites and the climatically-defined regions within which they occur(land resource units) serve as a framework for developing andselecting models. We illustrate the importance of clearly delineat-ing ecological sites to produce models and describe how we havedealt with poorly-delineated sites. Producing specific modelsrequires an understanding of the multiple ecological mechanismsunderlying transitions. We show how models can represent anddistinguish alternative and complementary hypotheses for transi-tions. Although there may be several mechanisms underlyingtransitions, they tend to fall within discrete categories basedupon a few, fundamental ecological processes and their relation-

Key Words: community stability, ecological sites, ecosystemhealth, indicators, New Mexico, vegetation dynamics

Resumen

Los modelos del tipo de estado y transicion ban recibido mochaatencion desde su introduccion en 1989. Sin embargo basta harepoco tiempo se cuento con diferentes modelos de 'estado y transi-cion' que pueden ser usados por agencias gubernamentales yparticulares. Existen pocos lineamientos disponibles para sudesarrollo e interpretacion. Basados en nuestra experiencia en eldesarrrollo de estos modelos para el estado de Nuevo Mexico,hacemos las siguentes preguntas: 1) ;, como debe ensamblarse lainformacion para crear modelos para sitios especificos engrandes regiones? 2) ;, cuales aspectos ecologicos deben ser con-siderados en el desarollo del modelo y su clasificacion? y 3) ;,como deben usarse estos modelos ? Revisamos la estructura gen-eral de los modelos de 'estado y transicion' enfatizando los cam-bios entre comunidades vegetales dentro de estados que sonreversibles con los cambios en el clima y practicas facilitadoras(i.e. manejo de pastizales). Asi como cambios entre estados (tran-siciones) que son reversibles solamente mediante practicas acel-eradoras como la resiembra, control de arbustos, y recuperacionde la estabilidad y funcion hidrologica del suelo. las regionesclimaticamente definidas (unidades de recursos terrestres) y lossitios ecologicos dentro de ellos sirven como estructura para eldesarollo y seleccion de los model os. Tambien, ilustramos laimportancia de delinear claramente los sitios ecologicos paraproducir los modelos y describir como hemos resuelto el proble-ma de los sitios pobremente delineados. Para producir modelosespecificos se requiere la comprension de los mecanismos ecologi-cos que determinan las transiciones. Adicionalmente, mostramoscomo los modelos pueden representar y distinguir hipotesis alter-

We thank several agencies, institutions and individuals who contributed data,knowledge, and criticisms that are represented in this manuscript. From the BLM,Phil Smith, Lane Hauser, Rusty Stovall, Mike Ramirez, Susan Britt, and JohnSpain; from the NRCS, David Trujillo, Gene Adkins, Garth Grizzle, DarrelReasner, Tim Henry, Frank Corn, Bill Schwebke, Arlene Tugel, and Pat Shaver;from the U. S. Forest Service, Wayne Robbie and Ralph Pope, from USDA-ARS,Bob Gibbens, from New Mexico State University, Gary Donart and BarbaraNolen; and private citizens, Jim Powell and Lee Gile. Alfonso Serna-Perez andIsaac Reyes-Vera translated the abstract. We thank Sam Fuhlendorf, GeorgePeacock, Nathan Sayre, Mark Westoby, 2 anonymous reviewers, and especiallyMitch McClaran for many valuable suggestions. That should not imply that theseindividuals agree with all that we have written, however.

Manuscript accepted 21 May 02.

JOURNAL OF RANGE MANAGEMENT 5612\ March 200~

nativas y/o complementarias para explicarlas transiciones. Aun cuando puede habervarios mecanismos para explicar transi-ciones, estos se consideran dentro de cate-gorias discretas basadas en QUOS pocosprocesos ecologicos fundamentales y susrelaciones pueden ser facilmente com-prendidas. EI conocimiento de los mecan-ismos esta estrechamente ligado al uso deindicadores ecologicos para anticipar lastransiciones. Concluimos que los modelosdeben de incluir 1) valores de referenciapara indicadores cuantitativos, 2) lista deindicadores clave y descripcion de suscambios que sugieran una aproximacion auna transicion, y 3) una documentacionrigurosa de las teorias y asunciones (y susalternativas) que daD base a la estructurade cada modelo.

ecosystem change and the roles that man-agement can play in directing theseprocesses. The details of these models candraw upon a wealth of recent conceptualand technological advances in communityand landscape ecology, including the rela-tionships between positive feedbackmechanisms and threshold changes inprocesses such as erosion (Davenport et al.1998), the dependency of thresholdchanges on processes operating on differ-ent scales of space and time (Scheffer etal. 2001), and understanding the linkagesamong processes using ground-based andremotely-sensed patterns (Rango et al. inpress). Implementing these advances toimprove on-site management of range-lands, however, presents substantial chal-lenges, among them: 1) how do we drawtogether the detailed information requiredto create site-specific models that areapplicable across entire regions, 2) whatecological issues need to be considered indeveloping and classifying sets of models,and 3) given the models that can be pro-duced, what do we want to use them forand how should they be used? In address-ing these general questions about state-and-transition model development, wedraw on examples from state-and-transi-tions models developed in a range of land-scapes in New Mexico.

Ecological theory provides a basis forland management. An understanding ofthe processes inferred to cause populationor community patterns determines howmanagers should respond to patterns. Aprime example is the succession-retrogres-sion (or range condition) model ofDyksterhuis (1949) that is based on thesuccessional theory of Clements (1916)and the edaphic polyclimax concept ofTansley (1935). This model emphasizedthe return of disturbed communities to acompetitively-determined climax state andhas been a guiding principle in range man-agement (Westoby 1980). Upon recogniz-ing an undesirable trend in plant commu-nity composition, managers could respondby reducing or redistributing grazing pres-sure and effect a return to desirable condi-tions. An important reason for the successof this model is that it provided a methodto measure and compare land conditionagainst the expectations of the model (i.e.,the similarity index), thus providing a con-crete link between theoretical expectationsand management response.

Rangeland managers have long recog-nized that semiarid grasslands can trans-form into shrub-dominated states that can-not be returned to grassland through graz-ing management (Laycock 1991), contraryto applications of the succession-retrogres-sion model. Assuming that a single, com-petition-defined equilibrium plant commu-nity should exist for each site, alternativestates, and the rangelands in which theyoccur, have been referred to as "non-equi-librial" (following Wiens 1984). In fact,these alternative states may be highlyequilibrial (e.g., Muller 1940) after thetransition, so these systems are bettertermed multi-equilibrial. The increasingemphasis on processes other thancompeti-

tion in determining community patterns(Kingsland 1985), paved the way forWestoby et al. (1989, Westoby 1980) topropose a revised framework for rangemanagement. The state-and-transitionmodel formally acknowledged the multi-equilibrial nature of many rangelandecosystems and the rapid and unanticipat-ed shifts among these equilibria.Furthermore, Westoby et al. (1989)focused attention on the multiple mecha-nisms underlying alternative equilibria andemphasized an "opportunistic" style ofmanagement in which strategies varydepending upon which mechanisms areimportant. The state-and-transition con-cept provides a means for anticipatingdepartures from the monoclimax modeland incorporating this understanding intomanagement plans. Consequently, thisconcept is being widely espoused withinthe range science community of theUnited States (Society for RangeManagement 1995, USDA NRCS 1997).For agencies such as the NaturalResources Conservation Service (NRCS)and Bureau of Land Management (BLM),state-and-transition models promise toimprove assessment, monitoring, andmanagement in many semiarid rangelands.

Twelve years after the seminal publica-tion of the state-and-transition concept,however, few applications of the conceptexist that can be used by land managers.This is not to say that there has not been agreat deal of work on the concept.Researchers have provided refinements tounderlying ideas (Laycock 1991, Freidel1991, Rodriguez-Iglesias and Kothmann1997, Reitkerk and van de Koppel 1997),technological advances in the productionand quantification of models (Wiegandand Milton 1996, Allen-Diaz andBartolome 1998, Plant et al. 1999), andsome site-specific conceptual models (e.g.,Archer et al. 1988, Ash et al. 1994,George et al. 1992). There have been fewattempts, however, to develop sets of site-specific models based upon existing infor-mation that can be applied by land man-agers in a systematic way over broadareas. Here, we relay some of the insightsgained during the production of state-and-transition models for rangelands in thestate of New Mexico, USA.

Like other western states, New Mexicois dominated by semiarid rangelands andthe limitations of the monoclimax modelare very apparent in many of these ecosys-tems. State-and-transition models have thepotential to provide a framework for orga-nizing complex sets of ideas about themultiple, interactive processes driving

What is a state-and-transitionmodel?

The idea that rangeland vegetationexhibits multiple states, and transitionsamong them, has been referred to general-ly as the state-and-transition model. Theseconcepts have been adequately reviewedelsewhere (Laycock 1991, Brown 1994,Rodriguez-Iglesias and Kothmann 1997,Stringham et al. 2001). To be operational,however, specific state-and-transitionmodels must be created that describe thedetails of vegetation dynamics for particu-lar land areas. The graphical and concep-tual format provided by Westoby et al.(1989), with modifications most recentlysummarized by Stringham et al. (2001;Fig. 1), illustrate several key elements thatare communicated in these conceptualmodels. The model format presented hereis based upon that currently used and pre-sented by the NRCS.

As in vegetation mapping (Grossman etal. 1998), the most basic unit is the plantcommunity. This is the relatively homoge-neous assemblage of plants that occurs at aparticular point in space and time, and canbe defined at a scale relevant to a landmanager (e.g., 0.1-10 ha). In the modelsfor New Mexico, plant communities are

.IO1JRNAI OF RANnF MANAnFMFNT 'iRI?\ M"."h ?nn~ 11"

State C

Fig. 1. The general structure of state-and-transition models after Stringham et al. (2001).The small boxes represent individual plant communities and the dashed arrows betweenthem represent community pathways along which shifts among communities occur. Theseshifts are reversible through facilitating practices and fluctuations in climate. The largeboxes containing communities are states that are distinguished by differences in structureand the rates of ecological processes (such as erosion). The transitions among states (solidarrows) are reversible only through accelerating practices (e.g. seeding, shrub control, oraddition of soil) that can be applied at relatively great financial expense.

1997) are often expensive to apply.Generally, transitions among ecosystemstates are thought to be caused by a com-bination of external and internal, positivefeedback mechanisms that alter constraintson the presence or abundance of particularplant species (e.g., Schlesinger et al. 1990;Fig. 2). Three general classes of con-straints can be recognized: 1) the dispersalof propagules to a site and subsequentreproduction, 2) "neighbor" constraints,including the effects of competitors,predators, or parasites, as well as the ten-dency of certain life-forms to facilitate fIredisturbance, and 3) "site" constraints,including soil properties, hydrology, andclimate. Transitions occur when 1 or moreconstraints are altered by external factorsand this change catalyzes changes in posi-tive feedbacks that produce relativelyimportant shifts in vegetation structure andsoil properties. Multiple external factorscan be affected by singular processes, suchas livestock grazing (e.g., by introducingshrub propagules and decreasing competi-tion with them). Furthermore, changes inone class of constraints may reinforcechanges in other constraints, such that sev-eral positive feedback mechanisms operatetogether (c.f. Archer 1989).

For example, heavy, continuous live-stock grazing may initiate changes toshrub colonization ability by providing adispersal pathway for seeds and reducingcompetition and fire disturbance byremoving grasses. Reduced competitionand fire disturbance may permit shrubestablishment and growth. The presence ofadult shrubs 1) increases shrub seed avail-ability through reproduction and facilitatesthe dispersal of additional seeds to the siteby attracting birds, 2) increases competi-tion with grasses and limits grass reestab-lishment, and 3) increases erosion ratesand nutrient loss from shrub-interspaces byreducing grass basal cover. Alternatively,prolonged and severe drought or mechani-cal disturbance (e.g., off-road vehicles)might catalyze a similar sequence of eventsif shrub seeds were already present.

Thus, different states can be viewed asseparating positive feedbacks between dif-ferent kinds of plants and different ecosys-tem processes. For example, by retainingsoil nutrients and high infiltration capacityover relatively homogeneous areas(Ludwig et al. 1994), or by promoting fire(McPherson 1995), grasslands sustainthemselves. By promoting intershrub ero-sion and heterogeneous nutrient distribu-tions, or by outcompeting grasses forwater, shrub lands promote shrublands(Schlesinger et al. 1990). Once competi-

plant functional groups and ecosystemprocesses and, consequently, in vegetationstructure, biodiversity, and managementrequirements. For example, a grassland,whether it is dominated by one grassspecies or another, may provide sufficientcover to prevent rain-drop compaction ofthe soil surface and to intercept water andnutrients before they are lost from the sys-tem. In doing so, dominance by eithergrass species can sustain the soil condi-tions required by the other, and replace-ments along community pathways mayoccur within the state. Once grass coverhas been reduced below a critical amount(Davenport et al. 1998), or a shrub speciesinvades that leads to grass loss (Brownand Archer 1999), infiltration is reducedand erosion accelerates, a change in soilconditions occurs, and the system crossesinto a new state. This new state is charac-terized by a distinct set of plant communi-ties and a distinct range of values forecosystem attributes.

The shift between states is referred to asa "transition". Unlike community path-ways, transitions are not reversible by sim-ply altering the intensity or direction of thefactors that produced the change (c.f. the"amplitude" of Westman 1978) andinstead require the application of distinctfactors such as the addition of seeds, theremoval of shrubs, or the addition of top-soil. These "accelerating practices" asdefined by the NRCS (USDA NRCS

often defined by dominant species, orspecies that indicate the operation of par-ticular processes (e.g., an encroachingshrub species, whether or not it is nowdominant). Descriptions of plant commu-nities may also contain information on soilconditions that indicate processes (e.g., thecover of physical, chemical or cryptogam-ic crusts). It is important to identify therange of plant communities in a land areabecause these are the observable and mea-surable links to the processes embodied inthe remaining components of the model.Plant community identity at a locationmay vary in time and the arrows betweencommunities indicate changes (or "com-munity pathways") among them. Theseshifts in plant composition, as opposed tothose referred to as "transitions" (seebelow), may be caused by climate or landuse but are reversible by simply alteringthe intensity or direction of the factors thatcaused the change in composition (e.g.,practices that reduce grazing pressure orincreases in rainfall after a drought). The"facilitating practices" as defined by theNRCS (USDA NRCS 1997) would pro-duce responses along community path-ways. Thus, the succession-retrogressionmodel operates along community path-ways and is embedded within the state-and-transition model.

Communities are aggregated into states(Fig. 1) that are distinguished from otherstates by relatively large differences in

116 JOURNAL OF RANGE MANAGEMENT 56(2) March 2003

Climate change (Xchanged hydrology

II+

(e.g.reduced grassincreases erosion,

(e.g. decrease insite suitabilityfor grass)

(e.g. shrub introduction

permits competition)

(e.g. reductioo ofgrass populationspr~otes shrubestablishment)

.. +

Grazing or fire(life-form specificdisturbance)

Species introduction orlarge, general disturbance

Fig. 2. The relationships among the classes of constraints that are altered to produce anecosystem transition, and the internal and external processes that affect these constraints.External processes (dashed arrows) act as environmental triggers that set in motion posi-tive feedbacks represented by the internal processes (solid arrows). Processes affectingparticular constraints interact with one another such that multiple mechanisms produceand maintain transitions. See text.

ferences in important environmental fac-tors, including soil properties, slope, andlandscape position (e.g., in an upland orswale). These differences correspond todifferences in the structure of plant com-munities, and with respect to state-and-transition models, plant communitydynamics in the face of natural andhuman-caused disturbance (Society forRange Management 1995). Ecologicalsites are mapped by grouping soil map-ping units on which plant communities areassumed to behave similarly. Ecologicalsites occur together in a landscape as amosaic determined by patterns of geomor-phology (Fig. 3c). Like other vegetationclassification systems, ecological sites arenested within a hierarchy of climatically-defined regions (Fig 3a, 3b). The extent ofa particular ecological site is boundedwithin an area of similar geology and cli-mate, the land resource unit of the NRCSor ecoregion of the U.S. Forest Service,beyond which analogous ecological sitesmay exist. Thus, a particular state-and-transition model is intended to apply toone ecological site that can be found with-in only one land resource unit.

The definition of ecological _sites(Creque et al. 1999) and land resourceunits is often arbitrary. Given the impor-tance of soils and climate for the nature ofvegetation dynamics, the validity of anystate-and-transition model will dependupon the amount of variation in important

tive dominants are introduced, species arelost, or soil properties are significantlyaltered, transitions can be difficult andexpensive to reverse. An understanding ofthe constraints to ecosystem change, andthe relationships between the external andinternal mechanisms affecting them, sug-gest strategies for predicting and avoidingtransitions and devising restoration strate-gies (Whisenant 1999). Although thecauses of individual transitions in modelsare varied, the mechanisms involved fallinto readily-understood classes (Fig. 2)that are common to all models.

State-and-transition models, then, repre-sent postulates about the causes of bothephemeral and persistent changes in vege-tation at a site and should offer testablepredictions. Moreover, the models providea logical framework in which assumptionsand concepts about how rangelands workmust be specified. This can add clarity toour ideas about rangelands, and oftenreveals how little we really know aboutthem.

Historically, ecological sites (previouslycalled range sites) were based solely onsimilarities in the composition and produc-tivity of dominant, climax vegetation(Shiflet 1975). Ideally, ecological sites area classification of land types based on dif-

Clayeya)

lkm

Loamy.b)SDI

SD2Draw

SD3.---7

Gravelly sand

Fig. 3. The land-unit classification framework within which state-and-transition models arebeing developed by the Natural Resources Conservation Service. a) the Land ResourceRegion scale, highlighting the Western Range and Irrigated Region within which landresource units are embedded, b) the Major Land Resource Area scale: Southern DeserticBasins, Plains, and Mountains within New Mexico. Land resource units (SD-l, SD-2, SD-3)are noted with arrows. A unique set of ecological sites and state-and.transition models arecommon to each subresource area. c) a map of ecological sites for part of the JornadaExperimental Range based on groups of related soil series (draft soil map courtesy of Dr.Lee Gile and Barbara Nolen, the Desert Soils Project).

Model classification using ecologicalsites

Before a model can be created and test-ed, it is critical to define the extent overwhich a single model will apply. For aridrangelands in the United States, this extentis currently defined by the ecological sitesof the NRCS (USDA NRCS 1997).

JOURNAL OF RANGE MANAGEMENT 56(2) March 2003 117

creosotebush had encroached and dis-placed climax native grasses. Definingappropriate management standards via thestate-and-transition (or the succession-ret-rogression) approach requires that weacknowledge and accept data limitations,take into account the multi-equilibriumnature of plant communities, and developecological sites based on a detailed under-standing of plant-soil relationships.

Defining the range of alternative com-munities occurring in several ecologicalsites can be accomplished using monitor-ing data, such as those gathered by theLong-Term Ecological Research program,the BLM, the NRCS, private individuals,as well as repeated aerial or terrestrialphotography from a variety of sources(e.g., Callaway and Davis 1993, Miller1999, McClaran et al.). In many caseshowever, the number of ecological sitesmonitored or the duration of monitoring islimited. Interviews of rangeland profes-sionals, researchers, and ranchers(Bellamy and Brown 1994) in conjunctionwith structured, rapid vegetation surveysbased on soil maps and associated withsoil series determinations can add signifi-cantly to the number of communitiesencountered and provide rigorous associa-tions of communities with soil properties.

A practical limitation of using profes-sional testimony and casual field observa-tions to define communities is that it isoften unclear how communities are relatedin time. The number of communitiesincluded within an ecological site may bean artifact of the persistent environmentalheterogeneity in space included within anecological site definition, rather than tem-poral variability at points in space. Whenit is suspected that 2 communities do notoccur at the same points in space, the cre-ation of a new ecological site may becalled for. Alternatively, (or while newecological sites are developed), we candenote the absence of temporal relation-ships among communities by having them"float" within the state and not be connect-ed to other communities by communitypathways (Fig. 4). In other cases, areaswith differing aspect or slope are circum-scribed in ecological site definitions, andnorthern and southern exposures mayexhibit distinct species composition anddynamics. Nonetheless, these areas func-tion similarly enough (or are so predictablyassociated) that they are considered withinthe same ecological site and state.

soil properties and climate within ecologi-cal sites and land resource units, respec-tively. If land assigned to ecological sitesdoes not exhibit consistent properties, itmay not be clear whether the variation inplant communities observed between 2areas within the same ecological site isdue to a vegetation transition or to staticdifferences in environmental conditions(Friedel et al. 1993). For example, burro-grass (Scleropogon brevifolius Phil.) is amat-forming native perennial grass of theChihuahuan Desert, is often dominant onloamy or clayey ecological sites, and isrelatively unpalatable to livestock. Thus,dominance by burrograss is often assumedto be related to management practices anddrought. While this may be the case onStellar clay loam soils that were formerlydominated by tobosa (Pleuraphis muticaBuckl.), burro grass may have been histori-cally dominant on Reagan clay loam soils(Leland Gile, personal communication).Reagan soils are highly calcareous at shal-low depths relative to Stellar (see Gile andGrossman 1997) and are less pervious towater than are Stellar soils in a similarlandscape position (Herbel and Gibbens1989). Although Stellar, Reagan, andother soils have been grouped within thesame ecological site, these soils clearlyexhibit distinct properties and have proba-bly always harbored distinct communities.

To distinguish management-producedfrom natural patterns (especially in theabsence of historical data), it will be nec-essary to rigorously delineate ecologicalsites (Creque et al. 1999). Doing so willrequire that we distinguish eco)ogical sitesbased upon values of soil and climaticvariables that correspond to differences inthe nature of state-and-transition models.In turn, this necessitates a detailed docu-mentation of the relationships betweenplant communities and their dynamics to soilseries. Such efforts will in many cases leadto reassignments of soils to ecological sitesand the creation of new ecological sites.

lighted in the ecological site description.Furthermore, this community is oftenbelieved to be the one in which soilresources and native biodiversity is bestconserved (but see Belsky 1996). Thus,the state bearing this community (alongwith related successional stages) is theglobal management standard of the BLMand the standard is implicit in the activitiesof other agencies and non-governmentalorganizations (ci. USFS 2000, Bureau ofLand Management 2001, Strittholt andBoerner 1995).

There are, however, several limitationsto the historic climax plant communityconcept. First, in much of the westernUnited States, the historic climax plantcommunity does not now exist and cannotbe reliably estimated from historicalrecords. Thus, historic climax plant com-munities are sometimes estimated with thehidden assumption that the plants thatwere most palatable to livestock were thecompetitive dominants in the historic cli-max plant community (Westoby 1980).Second, the notion of the competitively-determined climax state is explicitlyacknowledged in the historic climax plantcommunity, and this is at odds with themulti-equilibrium concept now embracedby agencies (Svejcar and Brown 1991).That is, climaxes may shift, even withouthuman influence, such that a climax is a"moving target" over broad time scalesdue to climate change (Brown et al.1997).For example, it is possible that the domi-nance of black grama (Bouteloua eriopoda(Torr.) Torr.) on sandy soils of southernNew Mexico was a consequence of cli-matic conditions peculiar to the late nine-teenth century (Neilson 1986). A regionalchange in climate may now preclude suffi-cient sexual reproduction to reestablishblack grama as a dominant in many areas,and it is possible that overgrazing anddrought have only hastened the demise ofthis species. Given this case, it is question-able whether the restoration of a produc-tive black grama-dominated communitywould be a suitable management goal.Third, aggregating soils with distinctinherent properties into ecological sitesleads to the development of uniformexpectations where they may not be war-ranted. For example, creosotebush (Larreatridentata (DC) Cov.) has likely beendominant on erosional fan remnants at thebase of Mount Summerford in southernNew Mexico since before European settle-ment (Wondzell et al. 1996). The ecologi-cal site in which these soils are nowgrouped (Bulloch and Neher 1980), how-ever. would lead one to the conclusion that

Building state-aDd-transition modelsDefining communities and community

pathwaysOnce an ecological site has been select-

ed, the first task in creating a state-and-transition model is to define the communi-ties that can occur within that ecologicalsite. A key benchmark is represented bythe "historic climax plant community" ofthe NRCS (USDA NRCS 1997). Thiscommunity represents the composition ofplants that is known or is presumed tohave dominated an ecological site prior tothe settlement of Europeans, and is high-

Defining states and transitionsStates have been defined based on shifts

in plant community structure using multi-

JOURNAL OF RANGE MANAGEMENT 56(2) March 200311R

and conditions that lead to dominance byother species (Neilson 1986). Changed cli-mate or further soil degradation may leadto black grama extinction and dominanceby bunchgrasses that are able to reproduceby seed (transition 2). The transition to ashrub-invaded state is catalyzed by eitherthe introduction of mesquite propagulesinto grassland (Brown and Archer 1987)or by the competitive release of existingmesquite seedlings through the reductionof grass cover (Van Auken and Bush1997), fire frequency (Wright et al. 1976),or shrub seedling herbivores (Weltzin etal. 1997; transition 3a). Alternatively,propagule introduction might occur afteror concurrent with a change in climate orsoil degradation (Hennessy et al. 1983;transitions 4a, Sa). Initiated by these exter-nal triggers, the loss of black grama maybe caused by a shift in positive feedbacksinvolving competition, erosion, and physi-cal and chemical changes to soils(Schlesinger et al. 1990, Herrick et al.2002) due to the presence of shrubs (tran-sition 6). Shrubs would need to beremoved to return to the black grama-dominated (transition 3b) or limited (4b)state. Shrub expansion may need to becontrolled in order to remain in the shrub-invaded state, although coexistence

Fig. 4. A draft state-and-transition model for the gravelly loam ecological site within the SD-2 land resource unit of southern New Mexico. Transition 1 is caused by grass loss and sub-sequent shrub invasion, whereas transition 2 is caused by soil degradation. Transition 3requires shrub removal, grass seeding, and restoration of soil fertility and permeability.The communities within the eroded shrubland state are not connected by arrows, indicat-ing that there is no evidence for replacement among these community types. Instead, thesedistinct community types seem to reflect variation among soils included within the ecologi-cal site, although there is not yet enough information to reliably split soils into separate eco-logical sites. Black grama, bush muhly (Muhlenbergia porteri Scribn.), tobosa, and burro-grass are perennial grasses; burrograss is usually least palatable to cattle. Creosotebush andtarbush (Flourensia cernua D.C.) are shrubs.

38 I Black Slama II Mesquite I~

Shrub-invaded

grasslands6!IBunchgrasses I'I IMesqulte I

I Bad! grama 1'4-+ 1 aad! grama 1I. - 1 1 Dropseeds I.. ~, , !I " I" :' +

1 Blad! grams 1 1 I Dropseeds 11 Threeawns 1 i 1 Black grams I.

~ 4/;'""-~~.2!~;;;:;:ed rassla1d1 / ~~18.. 11b ,/4)

5 I

!5b i

7

variate analyses of long-term data sets(Allen-Diaz and Bartolome 1998). Whilethis approach is objective and potentiallyrepeatable, it does not consider theprocesses or mechanisms involved incommunity changes. Furthermore, mostapplications do not demonstrate that plantcommunities have entered a new domainof variability (e.g., Friedel 1991) that indi-cates a fundamental change in the func-tioning of the ecosystem (Whisenant1999). Thus, it is unclear whether changesin plant composition can be reversedthrough facilitating or accelerating prac-tices (Stringham et al. 2001).

Given these uncertainties and the pauci-ty of long-term data supporting the exis-tence of multiple domains of variability,states and transitions can be constructedbased upon postulates of vegetationchange in combination with empiricalobservations of community structure andenvironmental conditions. For example, anumber of explanations for the well-docu-mented loss of black grama and increasein honey mesquite (Prosopis glandulosaTorr.) may pertain to Sandy ecologicalsites in south-central New Mexico (Fig.5). The transition to a black grama-limitedstate (transition la) represents the shiftbetween climatic or soil fertility condi-tions conducive to black grama dominance

21

~

Fig. 5. A state-and-transition model for the sandy ecological site within the SD-2 landresource unit of southern New Mexico. Black grama, dropseeds (Sporobolus R. Br. spp.)threeawns (Aristida L. spp. ) are perennial grasses. Black grama is palatable to cattle for alonger duration than the other species and comparatively sensitive to grazing pressure.Snakeweed (Xanthocephalum Willd. spp.) is a subshrub that tends to invade with reduc-tions in grasses or with adequate winter-spring precipitation. Mesquite is a large shrub thattends to invade intact or degraded grasslands, promote the loss of grasses in intershrubspaces, and concentrate resources beneath its canopy.

JOURNAL OF RANGE MANAGEMENT 56(2) March 2003 119

lands to blue grama (Bouteloua gracilis(H.B.K.) Lag. ex Steud. Hitchcock) grass-lands characteristic of upland sites.Subsequently, agradation of soil into chan-nels will lead to a return of floodingcycles, and bottomland grasslands mayreestablish with improved grazing man-agement. Ecological sites and regions maydiffer in the degree to which hydrology,climate, or other physical features createstability over particular time scales, blur-ring the distinction between pathways andtransitions among "stable" states(Fuhlendorf et al. 2001). Nonetheless,when intensive accelerating practices(such as gully stabilization) can be used tohasten relatively slow natural processes orrecovery, the recognition of distinct statesis useful. Stability must be defined relativeto a time scale, so it is important to con-sider how the scale over which naturalrecovery is observed matches managementtimeframes when defining states.

may be sufficient to describe persistentchanges due to various causes and creatingnew ecological sites would be an unneces-sary semantic complication.

Unexpected transitions among states (orsites) may depend upon the transitionsoccurring in adjacent sites. On the loamysoils of the Jomada Experimental Range,for example, wind erosion of degradedsandy soils to the west of loamy soils hasresulted in the accumulation of sand on theloamy soils (C. Monger, personal commu-nication). These patches support blackgrama grass that is absent on the unalteredsoils. Herbel et al. (1972) speculate thatthe increased abundance of tobosa relativeto black grama and other grasses on lowerpiedmont clayey sites is due to theincreased water run-on from degradedgravelly sites occurring upslope. It isimportant to recognize that some transi-tions may have extrinsic causes thatdepend upon landscape context rather thanlocal management.

between bunchgrasses and mesquite with-out shrub control may persist for longperiods if soil degradation is not aggravat-ed by environmental stress. If grazing,drought, or shrub encroachment continueto reduce the remaining grass cover, ero-sion and/or increases in rodent and rabbitdensities (Campbell 1929) eventually leadto the formation of stable coppice duneswith the loss of most grasses (transition 7).At this point, natural reestablishment maynot be possible for most grass species. Inprinciple, the mechanical restoration andstabilization of topsoil and addition of soilnutrients and microorganisms followingshrub removal could be used to initiategrassland recovery (transitions 8, 9)

In this example, the definition of each ofthe states depend critically on our notionsof the processes driving vegetation changeand our responses to them. For example,the shrub-invaded state may be defined byeither the presence of shrub seedlings, orthe presence of shrubs seedlings in thecontext of particular stresses or changes todisturbance regimes. If the presence ofshrubs with grasses for long periods accel-erates grass loss, a shrub-invaded statemay be supported only by intensive accel-erating practices. Alternatively, this situa-tion might simply be considered as anearly stage of a shrub-dominated state ifshrub control was not implemented. Inaddition, we proposed that some commu-nities (e.g. the dropseed/black grama com-munity) occur in 2 distinct states. Thisdenotes that communities with similarvegetation structure may have very differ-ent responses to management due to dif-ferences in climate or soil propertiesbetween the states. States are human con-structs that represent our understanding ofand relationships with rangelands.

By definition, the recognition of statesand transitions also depends on temporalscale (Friedel 1997) and, thus, our ideas ofstability and equilibrium. In many cases,such as when grasslands are converted tomesquite dunes, the difference betweencommunity pathways and transitionsbetween states is clear. In other cases, it isnot. For example, bottomland/draw systemdegradation in the New Mexico Plateausand Mesas land resource unit may exhibita cyclic sequence of "states" over a periodof 50-100 years without the use of accel-erating practices (G. Adkins, NRCS, per-sonal communication). Erosion and chan-nelization due to reduced grass cover leadsto decreased soil moisture availability anda transition from giant sacaton(Sporobolus wrightii Munro) or alkalisacaton (S. airoides (Torr.) Torr.) grass-

The relationship between states, tran-sitions, and ecological sites

When a transition involves changes insoil structure, the resulting vegetation/soilchange may be so permanent that it is con-sidered a new ecological site rather than astate. The NRCS National Range andPasture Handbook states "Severe physicaldeterioration can permanently alter thepotential of an ecological site to supportthe original community" (USDA NRCS1997). This may apply when soils aretruncated to the point where an erodedphase is recognized in soil classification.For example, the loss of sandy surfacehorizons on some sandy loam soils mayexpose clay-rich strata that no longer sup-port the germination or survival of former-ly dominant species (Gile and Grossman1997). The point at which a new ecologi-cal site should be recognized, however,depends on the fuzzy distinction betweenchanges that are "persistent" (Stringham etal. 2001) without the use of acceleratingpractices (i.e., a transition) and "perma-nent" change. We should also be aware ofthe consequences of losing track of theoccurrence of historic communities at asite-do we want to maintain documenta-tion that a particular area used to be grass-land but is now shrubland? The amount offinancial resources required to apply accel-erating practices to reverse a transition maybe a suitable criterion for deciding when tocreate a new ecological site. Constraintsdue to biotic interactions, for example, areoften less expensive to overcome than abi-otic limitations (Whisenant 1999). We sug-gest, however, that the concept of "state"

Patterns in sets of state-and-transi-tion models

Given the issues and approaches dis-cussed in the preceding sections, whatgeneralizations can be drawn from themodels we have created? To date, we haveproduced about 60 draft or completedmodels spanning 4 land resource units insouthwestern-southcentral New Mexico.These land resource units intergrade withone another, and differ in subtle waysbased on moisture and temperature (Table1). Some of the patterns that we haveobserved can be compared among landresource units using models from a set ofcommon ecological site types representinga gradient of landscape position and soilproperties. In general, we see that morestates per ecological site were generatedfor thermic, aridic soils of SD-2 (i.e., thesouthern desert unit 2 land resource unit)than for the soils experiencing more usticand/or mesic regimes in higher elevationor more northerly land resource units. Thismay be due to 1 or 2 non-exclusive caus-es: 1) more ecological sites within rela-tively WanD and arid regions are subject toa variety of processes that lead to severalstates (e.g. erosion plus shrub invasion orexpansion; Table 1) and 2) land-use pro-fessionals recognize more states in SD-2because of the relatively extensiveresearch conducted there. Another featureapparent in the groups of models is thewide range in the number of states. Somemodels (e.g. SD-l Hills) have only onestate, implying that the site is resilientwith facilitating practices alone and that

JOURNAL OF RANGE MANAGEMENT 56(2) March 2003

Table 1. The number of states and key constraints defining the states for 7 common classes of ecological sites within 4 Land Resource Units (e.g., SD-2)in southern New Mexico. The Land Resource Units chosen differ in mean annual precipitation (MAP) and soil moisture regime (MR) and/or meanannual temperature (MAT) and soil temperature regime (TR) but all occur within south-central to southwestern New Mexico. The number of statesand key constraints are based on published literature and interviews with rangeland professionals. These models may be viewed at the NRCS NewMexico website: http://www.nm.nrcs.usda.gov/techserv/fotg/Section2/esd.htm

MLRA 36, WP-3*Latitude: 32° 22' - 34° 35'Longitude: 106° 58' - 109° 02'12-16" MAP;13.3°C MATMR: Ustic-AridicTR: Mesic

Ecological Site #ofstates

MLRA 42, SD.l*Latitude: 330 27' - 35° 21'Longitude: 106025' - 107° 07'

8-11"MAP; 13.3°CMATMR: Ustic-AridicTR: Thermic-mesic

#ofstates

MLRA 42, SD-4*Latitude: 310 40' - 320 30'Longitude: 105039' - 105055'12-14" MAP; 15°C MATMR: Ustic-AridicTR: Thermic

# of Key constraintsstates

Bottomland "5 GullyingBlocked run-on

GullyingBlocked run-on

GullyingBlocked run-on

Swale/Draw GullyingBlocked run-on

NA NA

Clayey/Clay upland 6 Soil sealingErosionlsoilloss

Blocked run on

Soil sealingGullying

4 Soil fertility lossErosionlsoilloss

-carny (Draw) Creosotebush invasion 2Erosion/soil loss

Soil fertility loss?Gullying

Sandv/Loarnv sand Soil fertility loss?Sand sage expansion

Sand sage3 expansion Mesquite invasion

Gravelly Soil fertility lossCreosotebushinvasion

Creosotebush expansion ErosionlsoillossJuniper4 invasion

MLRA 42, SD-2*Latitude: 31019' - 340 24'Longitude: 105050' - 109002'

8-10" MAP;15.5°C MATMR: AridicTR: Thermic

Mean number of states** 4.8 2.8

* MLRA refers to the Major Land Resource Area (USDA NRCS 1997), a climatically-defined unit within which Land Resource Units are nested. Land Resource Unit designations

include SD 1-4 (Southern Desert units 1,2, and 4) and WP 3 (Western Plains and Mesas unit 3). See also Fig. 3.** Swale/Draw and Hills Ecological Sites were excluded from the mean because they do not occur in all Land Resource Unit's (Hills) or are functionally different (Swale/Draw).

IBlocked run-on refers to the alteration of surface hydrology by features such as darns, ditches, or roads that inhibit natural water run-on patterns2lnvasion refers to the presence of a plant species that was not present in historic communities, whereas expansion refers to increased density of a plant relative to its density in historiccommunities.3 Artemisiafilifolia Torr.4 Juniperus monospenna (Engelm.) Sarg.

subject to erosion and loss of soil fertility.The importance of shrub or tree invasion,on the other hand, seems to depend on theecological site/land resource unit combi-nation in question, although in some landresource units (e.g. SD-2) it is moreimportant than in others. Overall, we seethat a subset of common processes in vari-ous combinations explain vegetationdynamics within different ecological sites.This allows us to make some generaliza-tions, while requiring that these general-izations be carefully evaluated for eachecolo~ical site and land resource unit.

the succession-retrogression model ade-quately describes its dynamics. At a conti-nental level, this may be true of manyrangeland communities that are resistant tosoil degradation, not dependent on surfacehydrologic inputs, or not subject to inva-sion by competitive species.

Can we generalize about the importanceof particular processes in different ecolog-ical site types or land resource units? Insome cases, we find that the same types ofprocesses are invoked to explain transi-tions in similar ecological sites: lowlandsites such as bottomlands, draws andclayey sites are subject to changes in sur-face hydrology and surface soil structureand chemistry with respect to infiltration.Changes among states in upland sites suchas sandy, gravelly, and hills sites are often

rounding past vegetation changes on eco-logical sites and to use this information toanticipate and interpret changes in thefuture. Beyond this general function, iden-tifying the operation of processes underly-ing transitions will be key elements inusing models to improve land manage-ment. In many cases this is difficult to do,however, because we often do not fullyunderstand the causes of observed vegeta-tion changes, particularly their interac-tions.

A conceptual approach to deal with thisis to postulate that there is usually a domi-nant proximate variable or a characteristicpattern of correlation among several vari-ables that regulates particular transitionson ecological sites. In the southwesternU.S., the duration and timing of available

Using models: proximate variables,indicators, and predicting transitions

The primary use of state-and-transitionmodels is to depict the circumstances sur-

JOURNAL OF RANGE MANAGEMENT 56(2) March 2003 121

>-QC(J.):J0-

~L1.

~.100 100~ 200+~& 50-100 100-000 200+ 20'30 -~

soil water may be a key variable involvedin many kinds of transitions (Whitford etal. 1995, Devine et al. 1998, Breshears andBarnes 1999; Table 1) and also relatesclosely to nutrient availability (StaffordSmith and Morton 1990, Reynolds et al.1999, Dodd et al. 2000). Transitions maybe due to threshold (non-linear) changes inthe availability of soil water over time dueto shifts in runoff patterns (e.g., percola-tion thresholds; Davenport et al. 1998) orto threshold responses of species perfor-mance to gradual changes in the soil water(e.g., species-specific soil moisture limita-tions). Thus, nonlinear responses may becaused by both positive feedbacks thatcreate threshold changes in variables aswell as through species tolerance limits(e.g., Austin 1985).

To anticipate transitions, we need toknow something about the changes inproximate variables underlying thresholdresponses in vegetation and soils. Despitea wealth of information, these issues haveseldom been directly addressed in NewMexico, or elsewhere. For example, is thepersistent reduction of black grama abun-dance due to a reduction in soil moistureavailability, reductions in the abundanceof mycorrhizal fungi, a change in rodentherbivory levels, some other unconsideredfactor, or a combination of factors? If soilmoisture changes are important for blackgrama germination, survival, or reproduc-tion, then, across what values of moistureavailability do threshold responses occur?Are threshold or continuous changes inthese variables observed in nature andwhat are the causes of this variation? Arethreshold responses of particular speciesdetermined by the same environmentalvariables in all instances? Addressingthese questions for key grass and shrubspecies will greatly improve our capacityto provide flexible, predictive models oftransitions with management utility thatlink retrospective data and observationswith comparative studies and experiments.

But what is a manager to do with state-and-transition models while we gather andsynthesize these critical data? Even whenmechanisms are fully understood, proxi-mate variables such as soil moisture ormycorrhizal populations are going to bedifficult to monitor directly or may changeonly after a transition. A promisingapproach is to measure factors related tokey proximate variables that indicate theoperation of processes that can be alteredto prevent (or facilitate) a transition(Herrick 2000, Ludwig et al. 2000, Kuehlet al. 2001). For example, the presence,expansion, and spatial arrangement of bare

Bare ground patch size class (cm)

Fig. 6. Histograms for gap-intercept data (see Herrick et al. 2001) representing differentstates within the loamy ecological site in the SD-2 land resource unit of southern NewMexico. Graphs are arranged to parallel the state-and-transition model. Gap (bare-patch)sizes were measured along a 50-m tape. Gaps were the distance between perennial plantbases (where stems emerge from the ground) that were intercepted by the edge of the tape.Only gaps greater than 20 cm were tallied. Data were gathered on representative sites onthe Jornada Experimental Range.

the plant composition data gathered bymanagement personnel to identify states.Together, the suite of qualitative andquantitative measurements can do forstate-and-transition models what the simi-larity index does for succession-retrogres-sion models: they provide a means to con-nect field observations with theoreticalexpectations and management responses.

It is important to recognize that transi-tions usually do not happen simultaneous-ly across entire landscapes, grazing allot-ments, or even pastures. In many cases,transitions occur 1 patch at a time, occur-ring first on areas within ecological sitesthat are most sensitive to change due toslight variations in soils or landscape posi-tion (Fig. 7). Over time, or across space,these patches may coalesce to producelandscape-scale phenomena (Gosz 1993).Furthermore, changes in ecosystem func-tioning in 1 patch may affect adjacentpatches. In many arid systems, nutrientslost from patches that have undergonetransitions may be redistributed to adja-cent patches, increasing local productionand grass cover in those patches (Ludwigand Tongway 1995). These patch-scaledynamics indicate that 1) monitoringchanges in the frequency, size, or spatialarrangement of patch-scale transitionsmay be used to indicate the consequencesof management activities, 2) monitoringtransects located in "nutrient sinks" or

ground patches or shrubs, reduction in soilstability, or the formation of rills, litterdams, and terracettes indicate erosion thatreduces soil fertility and microbial popula-tions, water infiltration, and may eventuallytruncate entire soil horizons (Unpublisheddata, Herrick et al.). Thus, a trend towards atransition involving one or several interact-ing mechanisms may be revealed by indica-tors, even when the precise mechanisms(and critical values) are not well known.

By linking specific indicators to state-and-transition models, model developerscan provide tools to help land managersdetermine the state that land is in and toevaluate the probability of a transition.Different qualitative indicators (Pellant etal. 2000) relate to different processes, andmodels can point to specific indicators thatsignal an approach to a particular transi-tion. Models can include ranges in valuesfor quantitative indicators such as perenni-al plant cover, shrub density, bare groundpatch size, frequency, and spatial arrange-ment, soil compaction, and soil surfacestability to define the range of structureand function characterized by states andthat provide benchmarks for measure-ments of the processes leading to transi-tions. In particular, indicators such asbasal and canopy gap size (Unpublisheddata, Herrick et al.; see Fig. 6) are corre-lated with other indicators of soil qualityand erosion and can easily complement

JOURNAL OF RANGE MANAGEMENT 56(2) March 2003122

~

can convey what has happened on differ-ent soils, and what will happen given theoperation of the processes embodied in amodel. Managers can use case-specificinformation, supplemented with indicatorsin many instances, to evaluate the likeli-hood that particular processes are operat-ing, evaluate management options withrespect to those processes, and make adetermination of the costs and benefits ofthose options. The refinement of transi-tion-specific indicators and reporting ofreference values with specific models willaid this process.

In using reference indicator values, it isill-advised to consider maximizing grazingreturns by pushing a rangeland as close aspossible to a point of transition without"crossing" into another state. It is clearthat temporally unpredictable and uncon-trollable factors (e.g., climate) and a highdegree of spatial variability in site proper-ties (e.g., amounts of run-on, variation insoil gravel content) interact with factorsunder management control to cause rapidchanges in key, proximate variables inspace and time. Thus, it is doubtful thatindicator values for specific transitionswill be precise (Hobbs and Morton 1999,Muradian 2001). Management withrespect to state-and-transition models callsfor conservative strategies aimed at detect-ing and reversing trajectories toward vari-able environmental thresholds. This is theessence of "adaptive management" in thestate-and-transition model context.

- ~~~.~""~, Fig. 7. An area within the clayey ecological site on the Sevilleta National Wildlife Refuge

north of Socorro, N.M. in the SD-lland resource unit. Grasses in the foreground includegalleta (Pleura phis jamesii Torr.), alkali sacaton, and burrograss. This refuge has beenexcluded from livestock grazing since 1973. Despite rest, some large bare patches remain.These patches tend to have low soil aggregate stability indicating reduced capacity forinfiltration. This suggests that a transition has occurred, albeit at a small scale.

locations will be incapable ofearly warning of a broad-scaleand 3) it is necessary to inter-

u__.~ indicators within patchesPellant et al. 2000) by considering

.-transition- agencies and individuals adopt the

model framework, andmodels are produced, it is impor-

consider how these concepts can, management and

models will be used. First, theimplies that, for

. , --~..- ---~ -~ ...~~-being restored to a historic state

. practices. What.~~ "for all practical purposes"on the processes maintaining an

agement responses. For example, byassigning land within the gravelly ecologi-cal site to a creosotebush-invaded state, itis believed that the recovery of grass pro-duction may occur through use of a shrub-specific herbicide. In this case, competi-tion for water and nutrients and propagulelimitation define the properties of thestate. On the other hand, if there had beena transition from the creosotebush-invadedstate to the creosotebush-dominated state,the application of herbicide and seedingwould be ineffective for regenerating agrassland state. State-and-transition mod-els provide a framework for recognizingdistinctions among the causes of vegeta-tion change and distinguishing amongthese alternatives in the field.

The grassland-shrubland transition ongravelly soils also illustrates a potentialhazard of the state-and-transitionapproach. Degradation-caused shrub landson gravelly soils may be presumed to bepractically unrecoverable when they may,in fact, be recoverable through manage-ment practices (J. Powell, personal com-munication). If states are misidentified,sites may be "written off' prematurely,leading to missed opportunities and con-tinued degradation. For this reason, localknowledge and open debate are necessaryingredients for applying state-and-transi-tion models. State-and-transition models

- - ~

. is detennined by the- -~ ~_._~ of a dominant speciesit may be relatively inexpensive

the other hand, if the tran---_u_~ by soil degradation, thenof reversing it would be far

Conclusions and recommendationsState-and-transition models organize the

combined understanding of scientists andmanagers to explain ecosystem dynamicsacross variable rangeland landscapes. It isimportant to recognize that this frameworkis not a blanket replacement of an outdatedsuccession-retrogression model, but acomplement to it that accounts for theexistence of multiple equilibria, as well asthe return to equilibrium following pertur-bations. Furthermore, the contrast betweencommunities and states can be used to dis-tinguish the need for facilitating and accel-erating management practices. The resultsof a broad range of studies and personalexperiences can be summarized within thisframework and the resulting views can becontinually updated as new observationsand ideas emerge (Bradshaw and Borchers2000).

In order for the understanding represent-ed in state-and-transition models to becommunicated, refined, and comparedagainst the field observations of managers,we suggest the inclusion of several com-

land to a state, we assert-- ~ processes andthat land that indicate man-

OF RANGE MANAGEMENT 56(2) March 2003123

Third, both academic ecologists andresource managers should attempt to linkinformation on other components ofrangelands valued by land owners andsociety (e.g., biodiversity) to state-and-transition models so that their responsescan be interpreted alongside those ofplants and soils. (e.g., Van der Haegan etal. 2000, Bestelmeyer and Wiens 2001).

Our efforts in New Mexico lead us toconclude that a great deal of informationhas yet to be gathered, but that a great dealof understanding already exists that hadnot yet been captured in a useful form.The models we produced have been well-received by many field managers, for bothcomplementing and organizing theirknowledge as well as highlighting uncer-tainties. Models can improve the capacityof managers to evaluate the costs and con-sequences of management decisions andrally researchers around unanswered ques-tions. For this reason alone, the productionof state-and-transition models is a worth-while endeavor. As models become firmlylinked to a mechanistic understanding ofplant pattern, dynamics and indicators ofthe processes underlying them, the modelswill become invaluable.

Literature cited

Allen-Diaz, B. and J. W. Bartolome 1998.Sagebrush-grass vegetation dynamics: com-paring classical and state-transition models.Ecol. Appl. 8:795-804.

Archer, S. 1989. Have southern Texas savan-nas been converted to woodlands in recenthistory? Amer. Natur. 134:545-561.

Archer, S., C. Scifres, C.R. Bassham, and R.Maggio. 1988. Autogenic succession in asubtropical savanna: conversion of grasslandto thorn woodland. Ecol. Monogr.52:111-127.

Ash, A. J., J. A. Bellamy, and T. G. H.Stockwell. 1994. State and transition modelsfor rangelands. 4. Application of state andtransition models in northern Australia. Trop.Grassl. 28:223-228.

Austin, M. P. 1985. Continuum concept. ordi-nation methods, and niche theory. Annu.Rev. Ecol. Syst. 16:39-61.

Bellamy, J. A. and J. R. Brown. 1994. Stateand transition models for rangelands. 7.Building a state and transition model formanagement and research on rangelands.Tropical Grassl. 28:247-255.

Belsky, A. J. 1996. Viewpoint: Westernjuniper expansion: Is it a threat to arid north-western ecosystems? J. Range Manage.49:53-59.

Bestelmeyer, B. T. and J. A. Wiens. 2001.Ant biodiversity in semiarid landscapemosaics: the consequences of grazing vs.nat-ural heterogeneity. Ecol. Appl. 11:1123~1140.

Bradshaw, G. A. and J. G. Borchers. 2000.Uncertainty as information: narrowing thescience-policy gap. Conserv. Ecol. 4: 7.[online] URL:http://www.consecol.org/voI4/issl/art7, 15May 2002

Breshears, D. D. and F. J. Barnes. 1999.Interrelationships between plant functionaltypes and soil moisture heterogeneity forsemiarid landscapes within the grassland/for-est continuum: a unified conceptual model.Landscape Ecol. 14:465-478.

Brown, J. H., T. J. Valone, and C. G. Curtin.1997. Reorganization of an arid ecosystem inresponse to recent climate change. Proc. Nat.Acad. Sci. USA 94:9729-9733.

Brown, J. R. 1994. State and transition modelsfor rangelands. 2. Ecology as a basis forrangeland management: performance criteriafor testing models. Trop. Grassl. 28:206-213.

Brown, J. R. and S. Archer. 1987. Woodyplant seed dispersal and gap formation in aNorth American subtropical savanna wood-land: the role of domestic herbivores.Vegetatio 73:73-80.

Brown, J. R. and S. Archer. 1999. Shrubinvasion of grassland: recruitment is continu-ous and not regulated by herbaceous biomassor density. Ecol. 80:2385-2396.

Bulloch, H. E. and R. E. Neher. 1980. SoilSurvey of Dona Ana County Area, NewMexico. USDA Soil Conservation Service,Washington, D.C.

Bureau of Land Management. 2001. H-4180-1- Rangeland Health Standards, BLMManual, Release 4-107, Bureau of LandManagement, Washington, D.C.

Campbell, R. S. 1929. Vegetative successionin the Prosopis sand dunes of southern NewMexico. Ecol. 10:392-398.

Callaway, R. M. and F. W. Davis. 1993.Vegetation dynamics, fire, and the physicalenvironment in coastal central California.Ecol. 741567-1578.

Clements, F. E. 1916. Plant succession: ananalysis of the development of vegetation.Carnegie Inst., Washington Pub. 242:1-512.

Comrie, A. C. and E. C. Glenn. 1998.Principal components-based regionalizationof precipitation regimes across the southwestUnited States and northern Mexico, with anapplication to monsoon precipitation vari-ability. Climate Res. 10:201-215.

Creque, J. A., S. D. Bassett, and N. E. West.1999. Viewpoint: Delineating ecologicalsites. J. Range Manage. 52:546-549.

Davenport, D. W., D. D. Breshears, B. P.Wilcox, and C. D. Allen. 1998. Viewpoint:Sustainability of pinon-juniper ecosystems-a unifying perspective of soil erosion thresh-olds. J. Range Manage. 51:231-240.

Devine, D. D., M. K. Wood, and G. B.Donart. 1998. Runoff and erosion from amosaic tobosagrass and burrograss communi-tY in the northern Chihuahuan Desert grass-land. J. Arid Environ. 39:11-19.

Dodd M. B., W. K. Lauenroth, and I. C.Burke 2000. Nitrogen availability through acoarse-textured soil profile in the shortgrasssteppe. Soil Sci. Soc. Amer. J. 64:391-398.

JOURNAL OF RANGE MANAGEMENT 56/2\ March 20031?A

Dyksterhuis, E. J. 1949. Condition and man-agement of rangeland based on quantitativeecology. J. Range Manage. 2:104-115.

Friedel, M. H. 1991. Range condition assess-ment and the concept of thresholds: a view-point. J. Range Manage. 44:422-426.

Freidel, M. H. 1997. Discontinuous change inarid woodland and grassland vegetationalong gradients of cattle grazing in centralAustralia. J. Arid Environ. 37:145-164.

Friedel, M. H., G. Pickup, and D. J. Nelson.1993. The interpretation of vegetation changein a spatially and temporally diverse aridAustralian landscape. J. Arid Environ.24:241-260.

Fuhlendorf, S. D., D. D. Briske, and F. E.Smeins. 2001. Herbaceous vegetation changein variable rangeland environments: The rel-ative contributions of grazing and climaticvariability. Appl. Veg. Sci. 4:177-188.

George, M. R., J. R. Brown, and J. Clawson.1992. Application of nonequilibrium ecologyto management of Mediterranean grasslands.J. Range Manage. 45:436-440.

Gile, L. H. and R. B. Grossman. 1997. TheDesert Project Soil Monograph. NaturalResources Conservation Service, USDA Nat.Soil Surv. Center, Lincoln, Neb.

Gosz, J. R. 1993. Ecotone hierarchies. Ecol.Appl. 3:369-376.

Grossman, D. H., D. Faber-Langendoen,A.S. Weakley, M. Anderson, P.Bourgeron, R. Crawford, K. Goodin, S.Landaal, K. Metzler, K. Patterson, M.Pyne, M. Reid, and L. Sneddon. 1998.Terrestrial Vegetation of the United States.Volume 1: The National VegetationClassification System:Development, Status,and Applications. The Nature Conservancy,Arlington, Virgo

Hennessy, J. T, R. P. Gibbens, J. M.Tromble, and M. Cardenas. 1983.Vegetation changes from 1935-1980 inmesquite dunelands and former grasslands ofsouthern New Mexico. J. Range Manage.36:370-374.

Herbel, C. H. and R. P. Gibbens. 1989.Matric potential of clay loam soils on aridrangelands in southern New Mexico. J.Range Manage. 42:38~392.

Herbel, C. H., F. N. Ares, and R. H. Wright.1972. Drought effects on a semi desert grass-land range. Ecol. 53:1084-1093.

Herrick, J. E. 2000. Soil quality: an indicatorof sustainable land management? Appl. SoilEcol. 15:75-83.

Herrick, J. E., J. R. Brown, A. Tugel, P. L.Shaver, and K. M. Havstad. 2002.Application of soil quality to monitoring andmanagement: Paradigms from rangelandecology. Agro. J. 94:3-11.

Herrick J. E., W. G. Whitford, A. G. deSoyza, J. W. Van Zee, K. M. Havstad, C.A. Seybold, and M. Walton. 2001. Fieldsoil aggregate stability kit for soil quality andrangeland health evaluations. Catena44:27-35.

Plant, R. E., M. P. Vayssieres, S. E. Greco,M. R. George, and T. E. Adams. 1999. Aqualitative spatial model or hardwood range-land state-and-transition dynamics. J. RangeManage. 52:51-59.

Rango, A., S. Goslee, J. Herrick, M.Chopping, K. Havstad, L. Huenneke, R.Gibbens, R. Beck, and R. McNeely. (Inpress). Remote sensing documentation of his-toric rangeland remediation treatments insouthern New Mexico. J. Arid Environ.

Reitkerk, M. and J. van de Koppel. 1997.Alternate stable states and threshold effectsin semiarid grazing systems. Oikos79:69-76.

Reynolds, J. F., R. A. Virginia, P. R. Kemp,A. G. De Soyza, and D. C. Tremmel. 1999.Impact of drought on desert shmbs: effects ofseasonality and degree of resource islandsdevelopment. Ecol. Monogr. 69:69-106.

Rodriguez Iglesias, R. M. and M. M.Kothman. 1997. Structure and causes ofvegetation change in state and transitionmodel applications. J. Range Manage. 50:399-408.

Scheffer, M., S. Carpenter, J. A. Foley, C.Folke, and B. Walker. 2001. Catastrophicshifts in ecosystems. Nature 413:591-596.

Schlesinger, W. H., J. F. Reynolds, G. L.Cunningham, L. F. Huenneke, W. M.Jarrell, R. A. Virginia, and W. G.Whitford. 1990. Biological feedbacks inglobal desertification. Sci. 247: 1043-1048.

Shiflet, T. N. 1975. Range sites and soils in theUnited States. p. 26-33 In: D. N. Hyder (ed.)Arid Shrublands: Proceedings of the ThirdWorkshop of the U.S./Australia RangelandPanel. Soc. for Range Manage., Denver,Colo.

Society for Range Management, Task Groupon Unity in Concepts and Terminology.1995. New concepts for assessment of range-land condition. J. Range Manage.48:271-282.

Stafford Smith, D. M. and S. R. Morton.1990. A framework for the ecology of aridAustralia. J. Arid Environ. 18:255-278.

Stringham, T. K., W. C. Krueger, and P. L.Shaver. 2001. States, transitions, and thresh-olds: Further refinements for rangelandapplications. Oregon State Univ. Agr. Exp.Sta., Special Rep. 1024.

Strittholt, J. R. and R. E. J. Boerner. 1995.Applying biodiversity GAP analysis in aregional nature-reserve design for the edge ofAppalachia, Ohio (USA). Conserv. Bioi. 9:1492-1505.

Svejcar, T. and J. R. Brown. 1991. Failures inthe assumptions of the condition and trendmodel for managing natural ecosystems.Rangelands 13:165-167.

Tansley, A. J. 1935. The use and abuse of veg-etational concepts and terms. Ecol.16:284-307.

USDA Natural Resources ConservationService. 1997. National Range and PastureHandbook. U.S. Dep. Agr., Washington, DC.

USFS. 2000. National Forest System LandResource Management Planning; Final Rule.Federal Register 65 (219.20): 67574-67575.

JOURNAL OF RANGE MANAGEMENT 56(2) March 2003 125

Van Auken, O. W. and J. K. Bush. 1997.Growth of Prosopis glandulosa in responseto changes in aboveground and belowgroundinterference. Ecol. 78:1222-1229.

Van der Haegen W. M, F. C. Dobler, and D.J. Pierce. 2000. Shrubsteppe bird response tohabitat and landscape variables in easternWashington, USA. Conserv. Bioi.14:1145-1160

Vandekerckhove, L., J. Poesen , D. O.Wijdenes, J. Nachtergaele, C. Kosmas, M.J. Roxo, and T. deFigueiredo. 2000.Thresholds for gully initiation and sedimenta-tion in Mediterranean Europe. Earth SurfaceProcesses and Landforms 25:1201-1220.

Weltzin J. F., S. Archer, and R. K.Heitschmidt. 1997. Small-mammal regula-tion of vegetation structure in a temperateSavanna. Ecol. 78:751-763.

Westman, W. E. 1978. Measuring the inertiaand resilience of ecosystems. Biosci.28:705-710.

Westoby, M. 1980. Elements of a theory ofvegetation dynamics in arid rangelands.Israel J. Bot. 28:169-194.

Westoby, M, B. Walker, and I. Noy-Meir.1989. Opportunistic management for range-lands not at equilibrium. J. Range Manage.42:266-274.

Wiegand, T. and S. G. Milton. 1996.Vegetation change in semiarid communities:Simulating probabilities and time scales.Vegetatio 125:169-183.

Wiens, J. A. 1984. On understanding a non-equilibrium world: Myth and reality in com-munity patterns and processes, p. 439-457 In:D. R. Strong, D. Simberloff, L. G. Abele, andA. B. Thistle (eds). Ecological Communities:Conceptual Issues and Evidence PrincetonUniv. Press. Princeton, N.J.

Whisenant, S. G. 1999. Repairing damagedwildlands: a process-orientated, landscape-scale approach. Cambridge Univ. Press,Cambridge, U.K.

Whitford, W. G., G. Martinez-Turanzas,and E. Martinez-Meza. 1995. Persistence ofdesertified ecosystems: explanations andimplications. Environ, Monitoring andAssess, 37:319-332.

Wondzell, S. M., G. L. Cunningham, and D.Bachelet. 1996. Relationships between land-forms, geomorphic processes, and plant com-munities on a watershed in the northernChihuahuan desert. Landscape Ecol.11:351-362.

Wright, H. A., S. C. Bunting, and L. F.Neuenschwander. 1976. Effect of fire onhoney mesquite. J. Range Manage.29:467-471-

WE CATER TO SOME

PRETTY TOUGHCfJSTOMERS

~ . . with the West'slargest selection ofpasture. range and

reclamation seed.

ROO native and. varieties for

/ --~-~.., seed blends.plus a full line of erosion

control andhydroseeding products.81IIimrTH"iB1 consulting

services and trainingavailable.

~

Call or fax for our catalog. (801) 76B4422, fax (801) 768-,3967G'anit8 Seed Co., 1697 W. 21 m Nortl1, Lehi, UT 84043

www.graniteseed.com

126 JOURNAL OF RANGE MANAGEMENT 56(2) March 2003