1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33

THE EFFECT OF WOODY PLANT ENCROACHMENT ON LIVESTOCK
PRODUCTION: A COMPARISON OF NORTH AND SOUTH AMERICA

José D. Anadóna, 1, Osvaldo E. Salaa,b, B. L. Turnerb,c, Elena M. Bennettd

a

School of Life Sciences, Arizona State University, Tempe, Arizona 85287 USA
School of Sustainability, Arizona State University, Tempe, Arizona 85287 USA
c
School of Geographical Sciences & Urban Planning, Arizona State University, Tempe, Arizona
85287 USA
d
Department of Natural Resource Sciences and McGill School of Environment, McGill
University, Sainte-Anne-de-Bellevue, Quebec, H9X3V9 Canada
b

1

Corresponding author: J.D. Anadón, School of Life Sciences, Arizona State University, Tempe,
Arizona 85287 USA, jdanadon@asu.edu, 480-727-3728

Classification: BIOLOGICAL SCIENCES/Sustainability

Keywords: woody-plant encroachment, land-cover change, livestock production, global change,
food production, N and S American rangelands, net primary productivity

1

34

Abstract

35

A large fraction of the world grasslands and savannas are undergoing a rapid shift from

36

herbaceous to woody-plant dominance. This land-cover change is expected to lead to a loss in

37

livestock production, but the impacts of woody-plant encroachment on this crucial ecosystem

38

service have not been assessed. We evaluated how tree cover has affected livestock production at

39

large scales in rangelands of contrasting socio-economic characteristics in the U.S. and

40

Argentina. Our models indicated that in areas of high productivity, a 1% increase in tree cover

41

resulted in a reduction in livestock production ranging from 0.6 to 1.6 reproductive cows per

42

km2. Mean livestock production in the U.S. is 27 Rc km-2, so a 1% increase in tree cover results

43

in 2.5% decrease in mean livestock production. This effect is large considering that woody-plant

44

cover is increasing at 0.5-2% per year. On the contrary, in areas of low productivity, increased

45

tree cover had a positive effect on livestock production. Our results also show that ecological

46

factors account for a larger fraction of livestock production variability in Argentinean than in

47

U.S. rangelands. Differences in the relative importance of ecological versus non-ecological

48

drivers of livestock production in Argentina and the U.S. suggest that the valuation of ecosystem

49

services between these two rangelands might be different. Current management strategies in

50

Argentina are likely designed to maximize livestock production whereas land managers in the

51

U.S. may be optimizing multiple ecosystem services, including conservation or recreation,

52

alongside livestock production.

53

2

54

Significance Statement

55

Grasslands all over the world are undergoing a rapid shift from a regime dominated by

56

herbaceous plants to one dominated by woody plants, a phenomenon known as woody-plant

57

encroachment. The impact of this global phenomenon on livestock production, the main

58

ecosystem service provided by grasslands, remains largely unexplored. We quantified, for the

59

first time, the impact of woody-plant encroachment on livestock production at a large scale,

60

finding a reduction of between 0.6 and 1.6 reproductive cows per km2 for each 1% increase in

61

tree cover. By comparing the largest rangelands of the Americas (U.S. and Argentina), we also

62

showed how the impact of woody-plant encroachment is mediated by social-economic factors.

63

Our manuscript represents a significant advance in our understanding of grasslands as complex

64

social-ecological systems.

65
66

3

67

Introduction

68

Grasslands, shrublands, and savannas, collectively ‘rangelands’, constitute ca. 50% of the Earth’s

69

land surface (1). Although characterized by low yet highly variable annual rainfall, these areas

70

provide 30-35% of terrestrial net primary productivity (2), contain >30% of the world’s human

71

population, and support the majority of the world’s livestock production (3, 4). Besides

72

livestock production, rangelands also provide a variety of other ecosystem services, including

73

fiber production, carbon sequestration, maintenance of the genetic library (conservation), and

74

recreation (5).

75

One of the most striking land-cover changes in rangelands worldwide over the past 150

76

years has been the proliferation of trees and shrubs at the expense of perennial grasses (6). In the

77

U.S., non-forest lands undergoing woody-plant encroachment are now estimated to cover up to

78

335 million ha, i.e. 40% of the coterminous U.S., (7) and the increase in woody cover ranges

79

from 0.5 to 2% per year (8). The causes of this vegetation change are debated and the main

80

potential drivers include intensification of livestock grazing, changes in climate and fire regimes,

81

the introduction of non-native woody species, and declines (natural and human-induced) in the

82

abundance of browsing animals (9-12). Historical increases in atmospheric nitrogen deposition

83

and atmospheric carbon dioxide concentration have also been suggested to play a role (10, 11).

84

Woody-plant encroachment has long been of concern to a broad range of stakeholders,

85

from pastoralists to ranchers, because of the expected negative impact on livestock production

86

(13). In response, brush management has been widely used to reduce the cover of encroaching

87

woody-vegetation on both public and private lands. For example, the U.S. Natural Resources

88

Conservation Service spent $127M in brush management programs in the period 2005-2009,

89

implemented on more than 1 million ha of rangeland (14). Despite claims about impact of

90

woody-plant encroachment on livestock production and the large amounts of federal, state, and

91

private spending on brush management, the impact of woody-plant encroachment on livestock

92

production has seldom been quantified (15). Here, our objectives are i) to quantify how woody-

93

plant encroachment affects livestock production at large spatial scales, and ii) to assess how this

94

impact is modified under different ecological and social-economic conditions.

95

We developed a general framework in which livestock production depends on net

96

primary production, woody-plant cover, and other non-biological determinants. Net primary

97

production sets the total amount of biomass and energy that is available to herbivores (16). The
4

98

most common view on woody-plant encroachment is that encroachment diverts herbaceous

99

productivity, on which cattle feed, to unpalatable woody-plant productivity, thus reducing

100

potential energy intake (17-19). Thus overall, primary production and woody-plant

101

encroachment jointly determine the livestock carrying capacity of an ecosystem.

102

In natural ecosystems from forests to deserts, there is a tight correlation among primary

103

productivity and secondary productivity and animal biomass (16). Social and economic factors

104

determine how close current livestock stocking is to the carrying capacity of the site, which is

105

determined by NPP. Oesterheld et al. (20) assessed the relationship between net primary

106

productivity and livestock production in managed rangelands in Argentina, where management

107

focuses on food production, and found that the link between primary and secondary productivity

108

was even tighter than in natural ecosystems. Management practices such as providing water and

109

minerals, regulating animal distribution, and reducing parasitism, predation and diseases,

110

resulted in stocking rates that were closely associated with net primary productivity.

111

We expect that in advanced industrial societies, where the production of goods (e.g., food

112

by means of agriculture and ranching) plays a secondary role in the economy (21), landscapes

113

will be managed to maximize multiple ecosystem services, and thus livestock production might

114

be less driven by ecological drivers. Ecological factors, including net primary productivity and

115

woody-plant cover, determine potential stocking rate but actual stocking rate is modulated by

116

manager’s decisions (22). In some cases, land managers overstock rangelands leading to

117

degradation and desertification (23) while in other cases managers understock. The latter results

118

from pursuing optimization of multiple ecosystem services of which food production is only one.

119

Rangelands managed for multiple purposes and ecosystem services (24) seek provisioning of

120

food, fiber, firewood, carbon sequestration, conservation or recreation.

121

Our hypotheses are i) that overall livestock production decreases with woody-plant

122

encroachment, ii) the effect of woody-plant encroachment on livestock production is modulated

123

by NPP, with a larger negative impact of woody-plant encroachment in those areas with higher

124

net primary productivity, and iii) the role of ecological drivers (net primary productivity and tree

125

cover) on livestock production is larger in regions where the demand for ecosystem services is

126

concentrated exclusively on food production.

127
128

The scarcity of studies attempting to quantify the impact of woody-plant encroachment
on livestock production reflects the enormous difficulties of addressing this issue by means of
5

129

conventional field approaches. An experimental approach necessitates monitoring the change in

130

livestock production in a number of locations during the encroachment process, a process that

131

might take decades (11). Our approach has been to explore how current rangeland livestock

132

production varies at a regional scale along sites with different net primary productivity and

133

woody cover. We thus assessed the consequences of the process of woody-plant encroachment

134

by evaluating the relationship between tree-cover and livestock production at a given point in

135

time across multiple locations. This approach of swapping time for space has been used to

136

predict future trajectories of species in an ecological succession (25), and more recently, the

137

expected change of organisms ranging from microbes (26) to trees (27) under a changing

138

climate. We are aware of the limitations of the approach mostly associated with the existence of

139

lags that result in different models through space and time (28). Given the limitations of

140

alternative options and the urgency of the problem, we consider our approach to be promising.

141

To test our hypotheses, we collected information about woody-plant cover and primary

142

productivity from remote sensing sources and about livestock production from agricultural

143

census data. Woody-plant encroachment occurs when there is an increase in the abundance of

144

trees or shrubs. The type of woody component depends on mean annual precipitation with arid

145

systems being invaded by shrubs and mesic ecosystems being invaded by trees. In our study

146

areas, the transition between shrub and tree domains occurs approximately at 600 mm of annual

147

precipitation (Fig S2). In the present work, we focused on encroachment of trees (i.e., areas >600

148

mm) because current remote sensing tools assess tree cover with accuracy, but they do not

149

adequately estimate shrub cover (29) and thus our approach is not feasible in shrublands. We

150

aggregated data at the county level and combined remote sensing and census data in a model that

151

yields estimates of the impact of woody-plant cover on livestock production at large scales. To

152

account for the effects of socio-economic factors, we quantified the impact of tree cover on

153

livestock production in two regions of the world that have extraordinary environmental similarity

154

but have contrasting socio-economic characteristics (30, 31). The two regions are the U.S.

155

Central Grassland Region and the Argentinean Central Grassland. Both share similar temperature

156

and precipitation gradients, yielding vegetation types that are remarkably similar (31) (Fig. 1).

157

These environmental similarities contrast with large socio-economic differences in the rural

158

sector, and specifically regarding livestock production (Supplementary Information, Fig. S1).

159

During the last decades in the U.S., there has been a reduction of people making a living from
6

160

agriculture (40% reduction since 80s) and a negative trend in the number of cattle in the region

161

(22% reduction since the 70s). At present, a large proportion of stakeholders in the U.S. are not

162

full-time ranchers but maintain livestock production as a source of secondary income or for

163

cultural or recreational reasons (USDA economic service; www.ers.usda.gov). In Argentina,

164

although the relative importance of ranching has decreased due to the expansion of crop

165

products, especially soybean, the reduction in the number of cattle has been much smaller (4%

166

reduction since the 70s, Fig. S1); and beef is still the agricultural commodity with the largest

167

output value (28% of the total agricultural production 2005-2007) (32). As a result, we expected

168

stocking rates in Argentina to be closer to the NPP-derived carrying capacity of the system, and

169

thus more tightly driven by ecological factors, than in the U.S. (20).

170
171

Results and Discussion

172

In both the U.S. and Argentina, livestock production shows a W to E gradient of increasing

173

reproductive density. The maximum value in the U.S. is 66 reproductive cows (Rc) per km2 in

174

the eastern part of the region. In Argentina, this gradient is more apparent than in the U.S.,

175

reaching maximum values of 43 Rc km-2 (Fig. 2). This directional gradient is the same for NPP

176

and tree cover gradients in both regions, following mean annual precipitation gradients (Fig. 1).

177

In accordance with our first hypothesis, woody-plant encroachment in both rangelands

178

had a negative impact on livestock production. An increase of 1% in tree cover resulted in an

179

overall decrease in livestock production ranging from 0.6 to 1.6 Rc km-2 (Fig. 3, Table 1). In the

180

U.S., an increase in tree cover of 1% decreased livestock production by 0.57 Rc km-2. Mean

181

livestock production in the U.S. is 27 Rc km-2, so a 1% increase in tree cover results in 2.5%

182

decrease in mean livestock production of the region. In NPP units, a 1% increase in tree cover

183

had the same impact on livestock production as an NPP decrease of 41 g C m-2 y-1. The

184

magnitude of the impact can be gauged when taking into account that, in North America, the

185

increase of woody cover ranges from 0.5 to 2% per year (8).

186

As in our second hypothesis, in Argentina, a significant interaction between NPP and tree

187

cover as drivers of livestock production exists, although we did not find this interaction when

188

evaluating the U.S. data (Fig. 3, Table 1). At high productivity values (900 g C m-2 y-1), an

189

increase of 1% tree cover decreased livestock production by 1.6 Rc km-2, relative to livestock

190

production ranging between 1 and 43 Rc km-2. However, at productivity values of less than 365
7

191

g C m-2 y-1, tree cover enhanced livestock production. In low productivity (300 g C m-2 y-1) areas

192

in Argentina, an increase in tree cover of 1% increased livestock production by 0.24 Rc km-2.

193

This result contradicts current understanding of the impact of woody-plant encroachment, which

194

is thought to have a negative impact on livestock production (6, 17-19, 33). Note that the lower

195

limit of NPP in our study area in the U.S. occurs above 365 g C m-2 y-1, obscuring a possible

196

positive effect of tree cover on livestock production at low productivity values. Potential

197

explanations of this positive effect of woody-plant encroachment on livestock production at low

198

productivity values may be found in factors other than the amount of food available for livestock

199

production. For example, most of the areas of low productivity in our study area are associated

200

with low precipitation and high temperature (Fig. 1). In these areas, tree cover might provide

201

shelter and shade or overall near-ground temperatures, decreasing animal respiration costs (34) .

202

Our results showed that the effect of NPP and tree cover on productivity was larger in

203

Argentina than in the U.S. (R2= 50% and 24% respectively, Table 1), indicating a strong

204

difference between the two study areas in the importance of the drivers of livestock production.

205

This aligns with our third hypothesis, that the role of ecological drivers (net primary productivity

206

and tree cover) on livestock production would be larger in regions where the demand for

207

ecosystem services is concentrated exclusively on food production. The effect of tree cover on

208

livestock production relative to the effect of net primary productivity on livestock productivity

209

was similar in the two study regions, with the explanatory power of NPP being five times larger

210

than that of tree cover (U.S.: R2NPP = 20% and R2TC = 4%; Argentina: R2NPP = 42% and R2TC = 8%,

211

being R2NPP and R2TC the percentage of variance accounted for by net primary productivity and

212

tree cover) (Fig. 4). The similarity in the relative importance of NPP and tree cover indicates

213

that, despite the difference in socio-economic differences, the underlying ecological mechanisms

214

driving livestock production are similar.

215

Differences in the relative importance of ecological vs. non-ecological (social) drivers on

216

livestock production in Argentina and the U.S. suggest that the value of the various ecosystem

217

services provided by rangelands may be different in these two regions. Rangelands produce a

218

variety of ecosystem services including food and fiber production, carbon sequestration,

219

maintenance of the genetic library (conservation), and recreation (5). Current management

220

strategies in Argentina are likely to be designed to maximize a single ecosystem service

221

(livestock production). On the contrary, land managers in the U.S. appear to be optimizing
8

222

multiple ecosystem services, including conservation or recreation alongside livestock production.

223

Therefore, it is important to measure the effects of woody-plant encroachment on the entire

224

portfolio of ecosystem services that are provided by rangelands. Most changes in ecosystem

225

services due to woody-plant encroachment remain unclear and have been identified only in a

226

qualitative fashion (but see (33)). Future quantitative studies taking into account multiple

227

ecosystems services are needed in order to assist in decision making whether to implement or not

228

brush management actions. Livestock production is currently one of the most important

229

ecosystem service provided by rangelands but the development trajectory highlighted by the

230

differences between Argentina and the U.S. point out that other ecosystem services will likely

231

become increasingly important as economies undergo a transition from the production of goods

232

to the provision of services.

233

Our study demonstrates that livestock production is part of an integrated socio-ecological

234

system where ecological and social-economic drivers interact along gradients of climate and

235

economic development (22). In high productivity regions, woody-plant cover negatively affects

236

livestock production mainly through reductions in forage availability. The negative effect of

237

woody plants on forage availability is overwhelmed in low productivity regions by the positive

238

effects of woody cover that may be linked to the amelioration temperature, a possible linkage

239

that requires examination. As economic development increases the demand for ecosystem

240

services from rangelands becomes more diversified. In least developed regions, food and fiber

241

dominate the demand for ecosystem services. On the contrary, in developed regions there are

242

multiple demands from rangelands beyond food production that include conservation, carbon

243

sequestration, water supply and recreation. As development increases and demand diversifies the

244

importance of ecological drivers decreases while that of social-economic factors increases. The

245

future of woody-plant encroachment and its consequences on ecosystem services will be

246

modulated by changing climate and social and economic conditions.

247
248

Methods Summary

249

Study areas

250

We modeled the impact of woody plant encroachment on livestock production at a county

251

resolution for both U.S. and Argentinean rangelands (Fig. 1). Both areas share a similar

252

latitudinal temperature gradient and a longitudinal precipitation gradient, with precipitation
9

253

increasing from W to E. These similar climatic patterns yield vegetation types that are

254

remarkably similar (31). These similarities contrast with large social-economic differences (see

255

Introduction and Figure S1), which make them a perfect study system to address the impact of

256

woody-plant encroachment on livestock production at a regional scale and the variation of this

257

impact between different socio-economic regions.

258

The U.S. and Argentinean rangelands constitute, together with the Brazilian Cerrado, the

259

two main rangelands of the Americas (35). Here, we used rangelands in a very broad sense; our

260

two study areas comprise the transition between the desert and the forest biomes. We defined our

261

study areas in the U.S. and Argentina as those encompassing the prairie, savanna, and temperate

262

and subtropical desert and steppes divisions and regimen mountain divisions, according to

263

Bailey’s ecoregions (1). Within those areas, we excluded those counties with mean annual

264

precipitation values below 600 mm, thus focusing on the tree dominion (Fig S2) and excluding

265

woody cover due to shrubs. The resulting areas in the U.S. and Argentina had the same

266

precipitation lower limit (600mm) but differed in their upper limit (U.S.=1260 mm, Argentina=

267

2270 mm). In order to make the analysis of both areas fully comparable we limited the upper

268

precipitation limit of Argentina to that of the U.S. (i.e., 1260 mm). Taking into account also

269

those counties excluded due to low representation of non-crop lands (see below), the resulting

270

number of sampling units (i.e., counties in the U.S., departments in Argentina) was 242 for the

271

U.S. and 125 for Argentina.

272
273

Livestock production data

274

Data on livestock production were obtained from the USDA Census Database

275

(www.agcensus.usda.gov) and Argentinean Food and Agriculture Administration (SENASA;

276

http://www.senasa.gov.ar) (Fig. 2). In both cases, we used the last available livestock data (2007

277

for the U.S., and 2010 for Argentina). We focused on cattle, which is the main livestock type in

278

both areas. For comparability, we used the number of reproductive animals, a metric present in

279

both data bases. This metric corresponded to the class ‘Cows incl. calves’ in the USDA Census

280

data and to the class ‘Cows’ in the SENASA database (range: 1.5-66.4 and 0.5-43.2 animals per

281

km2 for the U.S. and Argentina respectively). In the U.S. we subtracted the number of cows on

282

feedlots, also available in the U.S. Census Database, from the total number of cows.

283
10

284

Environmental data

285

Net primary productivity, tree cover, and land uses per county were quantified by using

286

Moderate-resolution Imaging Spectroradiometer (MODIS) products (Fig. 2). All environmental

287

variables were characterized by the mean annual values of the year of the livestock data (2007

288

for the U.S., and 2010 for Argentina) and the four previous years. The value of the Net primary

289

productivity was assessed using the Photosynthesis and Net Primary Productivity algorithm

290

MOD17A3 (36). Here, production is determined by first computing a daily net photosynthesis

291

value which is then composited over an 8-day interval of observations over a year, to produce a

292

net primary productivity measure. Tree cover was assessed by means of MODIS Vegetation

293

Continuous Fields product MOD44B (29). This product represents Earth’s terrestrial surface as a

294

proportion of three surface cover components: percent tree cover, percent non-tree cover, and

295

percent bare ground. Land uses were assessed by the MODIS product MCD12Q1 (37). This

296

land-use remote sensing data allowed us to exclude crops and urban areas in our analysis, and

297

thus to obtain a more accurate measure of the net primary productivity available for livestock

298

consumption per county. Additionally, in order to remove those counties with low sampling size,

299

we also excluded from our analyses those counties with less than 1000 km2 or 25% of

300

rangelands.

301

Mean annual precipitation values were obtained from Earth observations and climatic

302

models. Specifically, annual precipitation values for the study periods in Argentina were

303

obtained from the Tropical Measuring Mission (TRMM; www.trmm.gsfc.nasa.gov) at a 0.25°

304

resolution. In the U.S., annual climatic data at a 2.5’ resolution were obtained from the PRISM

305

Climate Group (Oregon State University; www.prism.oregonstate.edu).

306
307

Hypotheses testing

308

Our first two hypotheses describe the impact of net primary productivity and tree cover on

309

livestock production and were tested by means of the model LP = β0 + β1*NPP + β2*TC +

310

β3*NPP*TC, where LP is livestock production, NPP is net primary productivity and TC is tree

311

cover. The sign and significance of β2 and β3 in the models fitted for the two study areas (U.S.

312

and Argentinean rangelands) tested first and second hypotheses.

313

The third hypothesis, that states that the role of ecological drivers (net primary productivity and

314

tree cover) on woody-plant encroachment on livestock production would be larger in regions
11

315

where the demand for ecosystem services is concentrated exclusively on food production, was

316

tested by means of examining model results in the U.S. and Argentinean rangelands separately.

317

In particular, we examined the explained variance of the model in each country. The relative

318

explanatory power of NPP and tree cover was assessed by means of a variance partitioning

319

analysis (38, 39), which allowed us to break down the total explained variance in four fractions:

320

pure effects of NPP, pure effects of TC (i.e., variance exclusively explained by NPP or TC), join

321

effect of NPP and TC (i.e., variance explained simultaneously by NPP and TC) and the effect of

322

the synergistic interaction between the two drivers (variance explained by NPP*TC).

323

The model LP = β0 + β1*NPP + β2*TC + β3*NPP*TC was fitted with three candidate sets

324

of variables describing NPP and TC considering one, three or five years of previous information:

325

a) variables describing NPP and TC values for the year of census (2007 for the US and 2010 for

326

Argentina), b) variables describing the average NPP and TC values of the year of the census and

327

the two previous years and c) the average NPP and TC values of the year of the census and the

328

four previous years. For both the U.S. and Argentina, the three candidate set of variables yielded

329

very similar patterns, although the models with largest values of explained variance, and thus

330

those presented here, were those with independent variables describing NPP and TC the year

331

before the livestock census data.

332
333

Acknowledgements

334

This work has been supported by the National Academies Keck Futures Initiative award 025512

335

(NAKFI), and by NSF DEB 09-17668 and DEB 1235828. K. Havstad, S. Archer and G. Ruyle,

336

E. Jobággy and C. Rueda provided valuable suggestions during the development of the project.

337

F. Maestre kindly reviewed a previous draft of the manuscript. X. Li and Y. Zhang

338

(Environmental Remote Sensing and Geoinformatics Lab, Global Institute of Sustainability,

339

ASU) assisted with the processing of MODIS data.

340
341

REFERENCES

342
343
344
345

1.
2.

Bailey RG Ropes L (1998) Ecoregions: the ecosystem geography of the oceans and
continents (Springer New York).
Field CB, Behrenfeld MJ, Randerson JT, Falkowski P (1998) Primary production of the
biosphere: integrating terrestrial and oceanic components. Science 281(5374):237-240.

12

346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391

3.

4.
5.

6.

7.
8.

9.
10.
11.
12.
13.
14.

15.

16.
17.
18.

19.
20.

Safriel U Adeel Z (2005) Dryland systems. Ecosystems and human well-being, current
state and trends, eds Hassan R, Scholes R, & Ash N (Island Press, Washington), Vol 1,
pp 625-658.
Reynolds JF, et al. (2007) Global desertification: building a science for dryland
development. Science 316(5826):847-851.
Sala O Paruelo J (1997) Ecosystem services in grasslands. Nature’s services: Societal
dependence on natural ecosystems, ed Daily GC (Island Press, Washington, D.C.), pp
237-251.
Archer SR (2010) Rangeland conservation and shrub encroachment: new perspectives on
an old problem. Wild rangelands: Conserving wildlife while maintaining livestock in
semi-arid ecosystems, eds du Toit JT, Kock R, & C DJ (John Wiley & Sons, Ltd,
Chichester, UK), pp 53-97.
Pacala SW, et al. (2001) Consistent land-and atmosphere-based US carbon sink
estimates. Science 292(5525):2316-2320.
Barger NN, et al. (2011) Woody plant proliferation in North American drylands: a
synthesis of impacts on ecosystem carbon balance. Journal of Geophysical Research
116(G00K07).
Archer S (1995) Tree-grass dynamics in a Prosopis-thornscrub savanna parkland:
reconstructing the past and predicting the future. Ecoscience 2(1):83-99.
Van Auken OW (2000) Shrub invasions of North American semiarid grasslands. Annual
Review of Ecology and Systematics 31:197-215.
Briggs JM, et al. (2005) An ecosystem in transition: causes and consequences of the
conversion of mesic grassland to shrubland. BioScience 55(3):243-254.
Naito AT Cairns DM (2011) Patterns and processes of global shrub expansion. Progress
in Physical Geography 35(4):423-442.
Scholes R Archer S (1997) Tree-grass interactions in savannas. Annual Review of
Ecology and Systematics 28:517-544.
Tanaka JA, Torell LA, Brunson MW, Briske DD (2011) A social and economic
assessment of rangeland conservation practices. Conservation benefits of rangeland
practices: assessment, recommendations, and knowledge gaps, ed Briske DE (US
Department of Agriculture, Natural Resources Conservation Service, Washington, DC),
pp 371-422.
Dalle G, Maass BL, Isselstein J (2006) Encroachment of woody plants and its impact on
pastoral livestock production in the Borana lowlands, southern Oromia, Ethiopia. African
Journal of Ecology 44(2):237-246.
McNaughton SJ, Oesterheld M, Frank DA, Williams KJ (1989) Ecosystem-level patterns
of primary productivity and herbivory in terrestrial habitats. Nature 341(6238):142-144.
Ethridge D, Dahl B, Sosebee R (1984) Economic evaluation of chemical mesquite control
using 2, 4, 5-T. Journal of Range Management 37(2):152-156.
Hedrick DW, Hyder DN, Sneva FA, Poulton CE (1966) Ecological response of
sagebrush-grass range in central Oregon to mechanical and chemical removal of
Artemisia. Ecology 47(3):432-439.
Hyder DN Sneva FA (1956) Herbage response to sagebrush spraying. Journal of Range
Management 9(1):34-38.
Oesterheld M, Sala OE, McNaughton SJ (1992) Effect of animal husbandry on herbivorecarrying capacity at a regional scale. Nature 356(6366):234-236.
13

392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436

21.
22.

23.
24.
25.
26.

27.

28.

29.

30.
31.

32.

33.
34.

35.
36.
37.
38.

Bell D (1973) The coming of post-industrial society: a venture in social forecasting.
(Basic Books, New York), p 507.
Díaz S, et al. (2011) Linking functional diversity and social actor strategies in a
framework for interdisciplinary analysis of nature's benefits to society. Proceedings of
the National Academy of Sciences 108(3):895-902.
Geist HJ Lambin EF (2004) Dynamic causal patterns of desertification. BioScience
54(9):817-829.
Pratt D Gwynne MD (1977) Rangeland management and ecology in East Africa (Hodder
and Stoughton).
Clements FE (1916) Plant succession: an analysis of the development of vegetation
(Carnegie Institution of Washington, Washington, DC) p 512.
Garcia-Pichel F, Loza V, Marusenko Y, Mateo P, Potrafka RM (2013) Temperature
Drives the Continental-Scale Distribution of Key Microbes in Topsoil Communities.
Science 340(6140):1574-1577.
Thuiller W, Lavorel S, Araújo MB, Sykes MT, Prentice IC (2005) Climate change threats
to plant diversity in Europe. Proceedings of the National Academy of Sciences of the
United States of America 102(23):8245-8250.
Sala OE, Gherardi LA, Reichmann L, Jobbágy E, Peters D (2012) Legacies of
precipitation fluctuations on primary production: theory and data synthesis. Philosophical
Transactions of the Royal Society B: Biological Sciences 367(1606):3135-3144.
Hansen MC, et al. (2003) Global percent tree cover at a spatial resolution of 500 meters:
First results of the MODIS vegetation continuous fields algorithm. Earth Interactions
7(10):1-15.
Paruelo JM, et al. (1995) Regional climatic similarities in the temperate zones of North
and South America. Journal of Biogeography 22(4/5):915-925.
Paruelo JM, Jobbágy EG, Sala OE, Lauenroth WK, Burke IC (1998) Functional and
structural convergence of temperate grassland and shrubland ecosystems. Ecological
Applications 8(1):194-206.
Lence SH (2010) The agricultural sector in Argentina: major trends and recent
developments. The shifting patterns of agricultural production and productivity
worldwide, eds Alston JM, Babcock BA, & Pardey PG (The Midwest Agribusiness Trade
Research and Information Center, Iowa State University, Ames, Iowa), pp 409-448.
Eldridge DJ, et al. (2011) Impacts of shrub encroachment on ecosystem structure and
functioning: towards a global synthesis. Ecology Letters 14(7):709-722.
Tucker CB, Rogers AR, Schüzt KE (2008) Effect of solar radiation on dairy cattle
behaviour, use of shade and body temperature in a pasture-based system. Applied Animal
Behaviour Science 109(2-4):141-154.
Wint W Robinson TP (2007) Gridded livestock of the world 2007 (Food and Agriculture
Organization of the United Nations Rome).
Heinsch FA, et al. (2003) User's Guide GPP and NPP (MOD17A2/A3) Products NASA
MODIS Land Algorithm. pp 1-57.
Friedl MA, et al. (2010) MODIS Collection 5 global land cover: Algorithm refinements
and characterization of new datasets. Remote Sensing of Environment 114(1):168-182.
Borcard D, Legendre P, Drapeau P (1992) Partialling out the spatial component of
ecological variation. Ecology 73(3):1045-1055.

14

437
438

39.

Cushman SA McGarigal K (2002) Hierarchical, multi-scale decomposition of speciesenvironment relationships. Landscape Ecology 17(7):637-646.

439
440

15

Table 1. Models assessing the effect of ecological drivers (net primary productivity and tree
cover) on livestock production in the U.S. and Argentinean rangelands. R2 = % of explained
variance. N.s. = non-significant effect (not included in the final model).

U.S.
Intercept
NPP
Tree cover
NPP*Tree cover
R2

Estimate
-40.8044
0.133
-0.5754
24.01

Argentina
p-value
0.8424
<0.0001
0.0005
n.s.

16

Estimate
-22.75
0.09796
1.1360
-0.003
50.26

p-value
0.6015
<0.0001
0.0006
0.0001

Figure 1. Main environmental gradients (mean annual precipitation and mean annual
temperature) in the U.S. and Argentinean rangelands. Rangelands are defined in this work as
those areas encompassing the prairie, savanna, and temperate and subtropical desert, steppes and
mountain divisions, according to Bailey’s ecoregions (1). Within these areas our work focused
on those counties with mean annual precipitation values between 600 and 1260 (see Methods
Section and Fig. 2). For both areas, national (bold lines) and county (thin lines) borders are
drawn. In the US state borders are also drawn (bold lines).

17

Figure 2. Livestock production, net primary productivity (NPP) and tree cover for our study
counties. Rangelands not included in the analyses (in grey) are those counties with annual
precipitation less than 600 mm or larger than 1260 mm (light gray) or those counties with less
than 1000 km2 in rangelands or less than 25% of their total area in rangelands (dark gray; see
Methods).

18

Figure 3. Response models of livestock production to net primary productivity and tree cover in
the U.S. and Argentinean rangelands. Equations for response models are shown in Table 1. The
red area indicates the NPP range where the impact of tree cover on livestock production is
negative. The green area indicates positive effect. Rc km-2= Number of reproductive cows per
km2.

19

40
10

20

30

pure NPP
pure TC
interaction NPP-TC

0

Explained variance (%)

50

Figure 4. Explanatory power of net primary productivity (NPP) and tree cover (TC) on livestock
production in the U.S. and Argentinean rangelands as assessed by a variance partitioning
analysis. This analysis breaks down the explained variance of the model into a) the pure effects
of NPP or TC (i.e. the portion of the variance explained exclusively by one this factors), b) the
join effects of NPP and TC (i.e. the portion of the variance explained jointly by NPP and TC, due
to, for example collinearity between them), and c) the interaction between NPP and TC.

U.S.

Argentina

20

SUPPLEMENTARY INFORMATION

0.95
0.85

ARG
U.S.

0.75
0.65

% of variation in total cattle

Figure S1. Evolution of the number of cattle (top) and agricultural population (bottom) by fiveyear periods in the U.S. and Argentina. Source: FAOSTAT (http://faostat.fao.org/; last accessed
Jul 13th, 2013).

1960

1970

1980

1990

2000

1990

2000

0.95
0.85

ARG
U.S.

0.75
0.65

% of variation in total cattle

Years

1960

1970

1980

Years

21

0.5
0.4
0.3
0.2
0.1

Mean tree cover per county (%)

Figure S2. Mean annual precipitation range (600-1260 mm) in relation to tree cover in the U.S.
The lower limit of our study area was set at 600 mm, thus excluding those areas where tree cover
is marginal (<10%). 1260 mm equals the maximum annual precipitation value for a county in our
study area. Annual precipitation from WorldClim database (http://www.worldclim.com/), tree
cover from MODIS (see Methods section).

400

600

800

1000

Annual precipitation (mm)

22

1200