1

THE SOCIAL FACTORS THAT FACILITATE OR CONSTRAIN RESTORATION 1

MANAGEMENT: WATERSHED REHABILITATION AND WET MEADOW (BOFEDAL) 2

RESTORATION IN THE BOLIVIAN ANDES 3

Brett D. Hartmana, David A. Clevelanda,b 4 aDepartment of Geography, University of California Santa Barbara, Santa Barbara, CA 93106 5

bDepartment of Environmental Science, University of California Santa Barbara, Santa Barbara, CA 93106 6

Abstract 7

Restoration ecology holds promise for addressing land degradation in impoverished rural 8

environments, provided the approach is adapted as it is applied to rural development settings. 9 There is a need for increased integration of social dynamics in land restoration, however few 10

systematic studies exist. We explored the social factors that influence restoration management, 11 including local motives and perceived benefits, incentives, land tenancy, institutional factors, 12 conflict resolution, accessibility, off-farm labor, and outmigration. The study area is a successful 13 watershed rehabilitation and wet meadow restoration project ongoing in the Bolivian Andes since 14

1992. We used household survey methods (n=237) to compare the four highest restoration 15 management intensity communities (HighRMI) with the four lowest restoration management 16

intensity communities (LowRMI). The results suggest several factors that facilitate investments in 17 land restoration, including aligning restoration objectives with local motives and perceived 18 benefits, ensuring incentives are in place to stimulate long term investments, conflict resolution, 19

land tenancy, and accessibility. Institutional factors such as community organization and 20 leadership capacity play a supporting or mediating role. In addition to social benefits reported by 21

local communities, land restoration helped slow the rate of rural to urban migration, with 29.1% 22 outmigration in HighRMI communities compared to 70.3% in LowRMI communities. Results 23

suggest that land restoration projects that integrate community development into project planning 24 and implementation will achieve greater success. 25

Keywords: Aymara, bofedal, check dams, erosion control, restoration ecology, community development, 26 socio-ecological system 27

1. Introduction 28

Reversing land degradation in impoverished rural environments is a global priority 29

(Gisladottir and Stocking 2005). Restoration ecology provides a promising framework (Brown 30

and Lugo 1994; Hobbs and Norton 1996; Lamb et al. 2005), but it will need to be modified when 31

applied to a rural development setting. The major challenge for restoring heavily managed rural 32

environments lies in increased integration of social dynamics into restoration efforts (Burke and 33

Mitchell 2007, Temperton 2007, Aronson et al. 2010, Shackelford et al. 2013). There is an 34

emerging body of literature which suggests that local and indigenous people can be effective at 35

land restoration, provided there are sufficient levels of social coordination and mobilization 36

2

(Walters 2000, Long et al. 2003, Amede et al. 2007, Stringer et al. 2007, Weston et al. 2015, 37

Hartman et al. 2016). Restoring ecosystem services also creates a series of social benefits that 38

help maintain rural livelihoods (Badola and Hussain 2005, Walton et al. 2006, Blay et al. 2008, 39

Nielson-Pincus et al. 2013, Barral et al. 2015, Reed et al. 2015). While these studies provide 40

preliminary insight into the social dynamics influencing land restoration, the social conditions 41

required for restoration success remain poorly studied. There is a need to better understand the 42

factors that support the social coordination and mobilization required for impoverished rural 43

communities to invest in land restoration. 44

People make land use decisions in response to a complex set of individual, social, 45

economic, and institutional factors that ‘pressure’ or ‘nudge’ change (Zimmerer 1993, Lambin et 46

al. 2003). Land degradation can stress societies by reducing the flow of ecosystem services 47

(Lamb et al. 2005, Scherr 2000, Carter et al. 2007), leaving populations vulnerable to food 48

scarcity, infectious disease, civil conflict, and involuntary migration (Myers 1997, Collins 2001, 49

Patz et al. 2004, Theisen 2008, Reuveny and Moore 2009). These stressors can, in turn, trigger 50

fundamental social changes that reduce the local ability to sustain the investments in time, 51

energy, and resources required to return degraded ecosystems to full functionality (Brown and 52

Lugo 1994). Robust social systems are needed for local communities to be able to initiate and 53

sustain land restoration, or at the very least, such systems need to be developed in the course of 54

land restoration (Leach et. al. 1999, Anderies et al. 2004, Amede et. al. 2007). The imperative is 55

to identify key social factors that either facilitate or constrain restoration trends, and which can 56

be managed through community development or policy initiatives. 57

Previous studies have identified several factors that influence land management 58

decisions, but their role in facilitating or constraining land restoration have been poorly studied. 59

These include aligning restoration objectives with local motives and perceived benefits (Blay et 60

al. 2008, Bullock et al. 2011, Reed et al. 2015); credits and incentives to stimulate long-term 61

investment (Mekuria et al. 2011, Montagnini and Finney 2011, Schiappacasse et al. 2012); road 62

networks and accessibility (Jungerius 2002, Nyssen et. al. 2002, Pender 2004, Valentin et. al. 63

2005); land tenancy and institutional factors (Agrawal and Gibson 1999, Leach et. al. 1999, 64

Hodge and McNally 2000, Gebremedhin et. al. 2004); conflict resolution (Amede et. al. 2007, 65

Theisen 2008); and pursuit of alternative economic opportunities through off-farm labor and 66

outmigration (Gray 2009, Baptista and Rudel 2006, Kull et. al. 2007, Izquierdo et. al. 2008). We 67

3

conducted an exploratory analysis to investigate the social factors that facilitate or constrain land 68

restoration at a long-term and large-scale watershed rehabilitation and wet meadow restoration 69

program in the highland Andes. In this paper, we first describe the social and environmental 70

characteristics of the study area, and the history of land degradation and subsequent restoration. 71

We then discuss the study variable selection process, and describe the household survey and 72

other methods used to collect data. Finally, we present results, contextualize them in the 73

literature, and draw preliminary conclusions about the implications for land restoration in a rural 74

development setting. 75

2. Study Area 76

The study area is a watershed rehabilitation and wetland restoration program in the 77

Bolivian Andes. It is located the Ayllu Majasaya-Aransaya-Urunsaya, an indigenous Aymara 78

territory situated along the Cochabamba-Oruro Highway in the Tapacarí Province, Department 79

of Cochabamba, Bolivia (Figure 1). Ayllus are communal territories, with agropastoral activities 80

conducted through institutions such as the ayanoka (communal planting areas), the mink'a 81

(communal planting and work days), the ayni (reciprocal work exchanges between families), and 82

traditional authorities such the hilakata (regulates communal planting and grazing activities). 83

Population densities are low (14.7 people/km2), with families living in dispersed and isolated 84

ranchos. The climate is semi-arid, with the majority of the rain falling between November and 85

March. Elevation ranges from 3,800 to 4,600 m and vegetation is dominated by Puna grasslands 86

with seeps and wet meadows (bofedales) embedded within the grassland matrix. Local 87

livelihoods are based on llama and sheep grazing in high elevation grasslands, with agriculture 88

concentrated in the more mesic valleys. 89

The highland Andes have been subjected to human influence for 7,000-8,000 years 90

(Ellenberg 1979; Baied & Wheeler 1993; Chepstow-Lusty et al. 1998). Land degradation in the 91

study area is a result of modern population growth, cultivation on steep slopes, and overgrazing, 92

which led to severe gully erosion, reduced agropastoral production, and bofedal degradation 93

(Siebert 1983; Harden 2001; Brandt & Townsend 2006). Bofedales provide important dry season 94

grazing, and impacts to these systems can negatively impact local communities (Squeo et al. 95

2006, Washington-Allen et al. 2008). According to oral histories, land degradation intensified 96

after the land reforms of 1952, which superimposed private land tenancy over the ayllu system 97

4

and created a complex of individual, communal, and intra-communal land use rights (La Fuente 98

1997, Delgado 2001, Rist et al. 2003). The complex of land use rights weakened the traditional 99

ayllu system of agropastoral management, and worsened resource disputes. Land degradation 100

exacerbated the levels of poverty and contributed to a reliance on off-farm labor and 101

outmigration to urban centers (e.g. to Cochabamba and Oruro) and to the coca (Erythroxylem 102

coca) regions in the Chapare (La Fuente 1997). People began working in the city for 3-4 weeks 103

out of the year to earn money as a supplement to their production in order to buy such things as 104

sugar and kerosene. The trips to the city became more prolonged over the years, and some 105

families established households in the city. Some individuals moved abroad, primarily to 106

Argentina and Spain, typically returning after a number of years to invest earnings in cargo 107

trucks or taxis that they would use to earn money in the region. 108

Land restoration began in 1992 through a partnership between the Ayllu Majasaya-109

Aransaya-Urunsaya and the Dorothy Baker Environmental Studies Center (Centro de Estudios 110

Ambientales Dorothy Baker, CEADB). CEADB documented the restoration process in project 111

records that included quarterly monitoring, annual evaluations, unpublished reports, and thesis 112

studies. Land restoration was planned through participatory community workshops in 1992 and 113

1995, consultation with community work groups, and feedback from community leaders and 114

promotores (community volunteers that received special training in land restoration techniques, 115

and that helped coordinate restoration activities and provide training to the rest of the 116

community). The partnership between CEADB and the Ayllu Majasaya-Aransaya-Urunsaya 117

eventually grew into to a program that included over 30 communities and multiple governmental 118

and non-governmental organizations. One of these organizations (Food for the Hungry, FHI) 119

used a food-for-work program to encourage terrace construction in exchange for foodstuffs. 120

When land restoration efforts began, conflict was identified as a major impediment to building 121

erosion controls. The sources of conflict identified included 1) disputes over land tenancy and 122

resource use rights; 2) intergenerational conflict from the younger generation leaving for military 123

service, and upon their return, not respecting the old ways, and 3) interpersonal disputes 124

(Hartman 1996, La Fuente 1997). Effective land restoration requires a high degree of 125

coordination, and conflict resolution was perceived to be a critical step before land restoration 126

could begin. To help address resource use disputes, several communities elected to privatize 127

communal lands. Communities that privatized lands allocated 42 ha to each family, with a very 128

5

small minority of families currently living on parcels from 5 - 21 ha (CEADB project records). 129

Land privatization was facilitated by legal reforms enacted in 1996 that granted a greater degree 130

of local control over economic and natural resources. 131

Land restoration was conducted in community work groups (aines). Land restoration 132

started in the upper reaches of the watershed in the early 1990s. Community work groups began 133

by building check dams in the headwaters of gullies along the paved Cbba-Oruro Highway, 134

working their way down slope as gullies were stabilized. Other erosion control structures such as 135

terraces, inlfiltration ditches, gabions, tree planting, grazing exlcosures, and stock ponds were 136

introduced in later stages of project development. The project was highly succesful, and over 137

30,000 erosion control structures (ECSs) were built. ECSs led to increased bofedal vegetation 138

and standing water in gullies, and increased greenness on approximately 50 km2 (Hartman et al. 139

2016a). 140

3. Methods 141

3.1 Sample frame 142

The sample frame for this study is based on a comparison between communities that had 143

conducted the most restoration management and those that had conducted the least. There is a 144

high degree of variability in the density of erosion control structures in project communities 145

(from 14.7 - 225.3 ECSs/km2, according to CEADB project records). Therefore, restoration 146

management intensity (RMI) was defined as RMI = No ECSs/km

2. We explored alternate 147

measures of RMI (e.g. normalizing RMI by population density). However, such measures did not 148

improve the differentiation of HighRMI and LowRMI communities. Using RMI, four high 149

restoration management intensity communities (HighRMI) and four low restoration management 150

intensity communities (LowRMI) were identified as study communities (Figure 1, Table 1). The 151

study communities are from the same ayllu system, and social structure, cultural norms and 152

practices, production methods, population density, and income are all similar. The study 153

communities are from a similar elevation belt, and soils and vegetation types are similar as well. 154

However, several communities in the lower reaches of the watershed (below 3,850 m) began 155

land restoration after 2003, and these were not considered in the selection of HighRMI and 156

LowRMI communities. Control communities were not selected for the household survey, as 157

6

many of the questions asked are not relevant outside the context of a land restoration program. 158

Control communities were included in two related studies to measure the biophysical effect of 159

ECSs (Hartman et al. 2016a) and evaluate changes in local knowledge and attitudes with 160

increased restoration management intensity (Hartman et al. 2016b). Two additional communities 161

with low to medium levels of RMI were surveyed (Table 1), and data from these communities 162

were used in selected data analyses as described below. 163

3.2 Study variable selection 164

This research was exploratory, and hypotheses about potential social factors influencing 165

land restoration in impoverished rural environments were developed from a review of the rural 166

development and natural resource management literature. The study variables (Table 2) were 167

selected based on relevance to a watershed rehabilitation and wet meadow restoration program 168

conducted by an Aymara indigenous group in the highland Andes. Study variable selection was 169

informed by prior research conducted in the study area in 1996, semi-structured interviews with 170

five key informants and nine community members in a reconnaissance survey in 2008, and 171

CEADB project records. Social factors that are not relevant to the study area but may be relevant 172

to land restoration in other social and environmental contexts were excluded from the analyses, 173

e.g. ecotourism (Blangy and Mehta 2006). The final list of study variables included: participation 174

rates, ECSs built, motives for and perceived benefits of land restoration, incentives (food-for-175

work program), institutional factors (land tenancy, active leadership, community development), 176

conflict resolution, accessibility (road networks), off-farm labor, and outmigration. The 177

information was solicited through a combination of open questions that allowed a variety of 178

responses, closed questions to elicit specific information, and scaled response questions to 179

evaluate perceptions about the factors that facilitate or constrain ECS construction (Table 2). 180

Data were evaluated by triangulating responses about motivations for and perceived benefits of 181

land restoration, perceptions about the factors that facilitate or constrain ECS construction, and 182

available study variable data. 183

3.3 Household survey 184

A questionnaire was developed for the household survey. Information solicited included: 185

1) participation rates, including whether or not they participated, year the respondent began 186

project participation, number of erosion controls constructed, whether or not the respondent 187

7

received any benefits from land restoration, and whether or not the respondent experienced any 188

negative side effects from land restoration efforts; 2) motivational factors for participating in 189

land restoration efforts and the perceived benefits of land restoration; 3) scaled response 190

questions about the levels of soil erosion when the land restoration began, and the current levels 191

of land restoration; 4) a facilitations and constraints analysis, which consisted of scaled responses 192

about the factors that either facilitate or constrain erosion control construction; 5) information 193

about levels of conflict, conflict resolution, and cooperation via comparisons between the present 194

level of conflict with the level of conflict at the time of project initiation; 6) days spent in off-195

farm labor, locally and in the city; 7) number of family members that live in the city or abroad; 196

and 8) general demographic information (age, sex, plot size, family size, household location). 197

The questionnaire was designed to begin with relatively simple information and then cover 198

potentially sensitive topics once a rapport was established, ending with easily answered 199

quantitative data, in case respondents became fatigued. The interview was designed to be 200

completed in 30 minutes or less. The questionnaire was tested with 11 respondents from 201

communities that were not part of the HighRMI and LowRMI sample population. Following 202

each test interview, the questionnaire was reviewed and revised by the survey team. Additional 203

information on the survey team, training, and interview protocols in included in the 204

supplementary information. 205

The facilitation and constraints analysis was a set of scaled response questions designed 206

to elicit perceptions of whether selected social factors facilitate or constrain the construction of 207

erosion control structures (ECSs). The facilitating and constraining factors were: improved 208

accessibility through road construction, incentives such as food-for-work, off-farm work in the 209

city, off-farm work in the ayllu, land privatization, living in ayanokas (collective lands), if 210

leaders are active or not, if community organization is strong, if there is conflict in the 211

community, and if conflict resolution occurs. The scaled response (-2, -1, 0, +1, +2) was 212

translated to verbal categories. Respondents were asked if the factor under consideration helped 213

(ha ayudado), hindered (ha afectado), or did not affect (no ha afectado) their ability to build 214

erosion control structures. If the response was neutral, no further questions were asked. If the 215

response was that the factor helped or hindered, then a follow-on question about magnitude was 216

asked, i.e. did it help a lot, or help a little? 217

8

The household survey was conducted between October 2010 and February 2011, during 218

the growing season. Interviews were conducted in one of three languages - Aymara, Quechua, or 219

Spanish. The questions were asked the same way each time without leading information or 220

gestures, and probing questions (who, what, when, where, and why) were asked when necessary. 221

Answers were recorded in Spanish directly into the spaces in the questionnaire, and later 222

translated to English. Interviews were conducted with the male or female head of household, 223

with a household defined as a residence unit comprising parents, children, and any additional 224

extended family members. Only one family member was interviewed per household. There were 225

a total of 210 respondents in the HighRMI and LowRMI groups (Table S1). An additional 27 226

respondents were interviewed in the two communities with low to medium levels of restoration 227

management intensity (n=237 total). Data from these communities were not included in the 228

comparison of HighRMI and LowRMI respondents, but were used to compare communities with 229

communal and private land tenancy and to calculate outmigration rates. Respondents were 76.7% 230

male and 23.3% female. Family size ranged from 2 - 14 people, and respondents ranged from 20 231

- 84 years of age. 232

Permission to conduct the survey was granted by the regional indigenous authorities. 233

With the support of the indigenous authorities, the non-response rate was not significant, and 234

only one household refused to be interviewed. Given the low population densities, a 100% 235

sample was conducted in most study communities. In the larger communities (Lakolakoni, Yawri 236

Totora, Japo, and Tallija), sampling was conducted by selecting routes that covered all 237

geographic regions of the community. Routes were selected in consultation with the promotores 238

and by reviewing aerial photographs. Routes were selected that traversed town centers, 239

secondary roads, and remote areas accessible by footpaths. Once a route was selected, a 100% 240

sample of all households along the route was conducted, including any households located in 241

side valleys. There was a potential sample bias created by people who were absent at the time of 242

survey as they were working in the city, planting or tending herds. Therefore, a list of households 243

that were not at home during the 1st visit was kept. The survey team returned for a 2

nd or 3

rd visit 244

to secure an interview, or the resident was located in other areas, such as at regional fairs. This 245

measure increased the sample from 52% to 67.5% of the households. 246

9

3.4 Additional data sources 247

A detailed census of the project area was conducted by CEADB in 2006. The census 248

included data on the number of males, females, and families resident in project communities in 249

2006. The 2006 census also included data on the number of families that had emigrated as of 250

2006. During the 2012 household survey, the number of resident families resident in each 251

community was documented in consultation with community leaders. 252

Town center points were identified from Instituto Geografico Militar 1965 1:50,000 253

topographic maps (6240 I, 6240 IV, 6241 II, 6241 III) and adjusted based on ground-truthing and 254

Googe Earth® imagery. Community boundaries were delineated in ArcGIS. Data sources 255

included the 1:50,000 topographic maps, detailed sketch maps created by local communities, 256

GPS points of key landmarks, and watershed boundaries from a 30m resolution ASTER Global 257

DEM. In general, the community boundaries followed ridgelines and river center lines, and it 258

was possible to delineate community boundaries by merging and clipping watershed polygons. 259

Road networks were delineated in Google Earth® and exported to ArcGIS. 260

3.5 Data analysis 261

Given that erosion control structures were built in communal work groups, the rates of 262

ECS construction were only meaningful at the group level. It was not possible to conduct a 263

disaggregated analysis of ECSs/household. Therefore, data were analyzed using a between-group 264

design: a Two-tailed t-Test for Equality of Means; a One-Way ANOVA (assuming equal 265

variance following a Levene's test) or a Brown-Forsythe One Way ANOVA (assuming unequal 266

variance following a Levine's test) with a Games-Howell post-Hoc Multiple Comparison of 267

Means for numerical data; and Person's Chi-Square or Fisher’s Exact Test for categorical data. 268

The scaled-response data in the facilitations and constraints analysis was evaluated for 269

asymptotic significances, using an Independent Samples Mann-Whitney U Test and an 270

Independent Samples Median Test. The Independent Samples Mann-Whitney U Test is used to 271

compare differences between independent groups when the variables are either ordinal or 272

continuous but are not normally distributed. The Independent Samples Mann-Whitney U Test 273

and Independent Samples Median Test are non-parametric tests that can be used to state whether 274

two groups have significant asymptotic differences in their histogram distributions, either 275

through differences in the shape of the distribution or the median. 276

10

4. Results 277

4.1 Participation 278

The HighRMI and LowRMI respondents had similar rates of participation in the 279

watershed rehabilitation and wet meadow restoration program (Table 3). However, on average, 280

HighRMI respondents participated in restoration earlier and built a greater number of erosion 281

control structures (ECSs). This pattern was observed across all types of ECSs, including check 282

dams, terraces, infiltration ditches, slow-forming terraces, diversion ditches, stock ponds, grazing 283

enclosures, micro irrigation, improved pasture plantings (Phalaris sp.), and reforestation. 284

4.2 Motives and perceived benefits 285

The reported motivations for and perceived benefits of land restoration (Table 4) are 286

related to local community livelihoods. The majority of respondents (96.6%) reported one or 287

more benefits from land restoration such as soil and water conservation, increased agropastoral 288

production, community development, and new land restoration knowledge. HighRMI 289

respondents identified a higher number of motivational factors (p=0.062) and perceived benefits 290

(p=0.015) per person compared to LowRMI respondents. Pasture rehabilitation, water "filtration" 291

and increased soil moisture, conflict resolution, and staple food items from the food-for-work 292

program were cited as motivating factors at significantly higher rates by HighRMI respondents. 293

In contrast, trees for housing and firewood was cited as a motivating factor at a significantly 294

higher rate by LowRMI respondents. 295

4.3 Incentives 296

The only incentive to build ECSs in the study area was the food-for-work program. The 297

HighRMI respondents cited the food-for-work program as a motivational factor at a significantly 298

higher rate (Table 4). In the facilitation and constraint analysis (Figure 2), the majority of 299

respondents perceived the food-for-work program to be a moderate facilitating factor for erosion 300

control construction. There are significant asymptotic differences in the histogram distributions, 301

with a higher proportion of LowRMI respondents citing food-for-work as a strong facilitating 302

factor. 303

11

4.4 Conflict and conflict resolution 304

The most common types of conflict reported were disputes over territory and resource 305

use rights, and the tensions that arise from some people in the ayanoka communities residing 306

only part of the year and acting as free riders on the efforts of permanent residents. Conflict 307

resolution was cited as a benefit of land restoration at significantly higher rates by HighRMI 308

respondents. The HighRMI group also responded at a significantly higher rate that there was less 309

conflict now, compared to when the land restoration project began (Table 7). Additionally, a 310

significantly higher proportion of people in the LowRMI group declined to respond or gave an 311

evasive answer (e.g. "yes, the project is very important"). The majority of respondents stated 312

conflict in the community was a significant constraint to building ECSs, whereas conflict 313

resolution was a moderate facilitating factor (Figure 2). There are significant asymptotic 314

differences in terms of the perceived role conflict resolution plays in building ECSs, with a 315

higher proportion of the HighRMI group stating that conflict resolution plays a neutral role in 316

building ECSs. 317

4.5 Land tenancy and other institutional factors 318

The rate of ECS construction is significantly higher in communities with private land 319

tenancy (61.2 ± 9.4) and ayanoka communities (19.3 ± 3.9) (t(186)=4.22, p=0.016). Increased 320

efficiency due to land privatization and resource consolidation was cited as a benefit of land 321

restoration by both HighRMI and LowRMI respondents (Table 4). Land privatization was 322

viewed as a moderate facilitating factor for building ECSs by HighRMI and LowRMI 323

respondents (Figure 2). In contrast, collective land tenancy is the only factor that has a bimodal 324

distribution, i.e. a portion of respondents perceives ayanokas to be a facilitating factor for 325

building ECSs, and a portion perceives ayanokas to be a constraint (Figure 2). There are 326

significant asymptotic differences in the histogram distributions of other institutional factors as 327

well. A higher proportion of LowRMI respondents cited active leaders and community 328

organization as a moderate to strong facilitating factor compared to HighRMI respondents. A 329

similar pattern was observed when the facilitation and constraint data was analyzed by land 330

tenancy groups (ayanoka vs. private): a significantly higher proportion (p<0.05) of ayanoka 331

respondents cited active leadership, community organization, and conflict resolution as being 332

strong facilitating factor, compared to the private land tenancy respondents. 333

12

4.6 Road networks 334

Increased animal, pedestrian, and vehicular access was cited as a benefit of land 335

restoration at similar rates in both groups (Table 4). For example, vehicular access to one 336

community (T'ola Marka) was possible only after gullies were consolidated and residents built a 337

a gabion to create a land bridge across a ravine. The majority of respondents perceived road 338

networks to be a strong facilitating factor for building erosion control structures (Figure 2). This 339

pattern existed regardless of whether a person lived on the main road, a secondary road, or a 340

footpath. However, the relationship between household location and the number of ECSs built 341

reveals a different pattern (Table 3). There is a significantly lower mean total number of ECSs 342

for families living along the Cbba-Oruro highway, compared to families living on secondary 343

gravel roads or on footpaths. The difference in Total ECSs between households located on 344

secondary roads and on footpaths is not significant. 345

4.7 Outmigration and off-farm labor 346

There was a significantly higher percentage of emigrant families in the LowRMI 347

communities compared to the LowRMI communities (Table 3). Moreover, there was no 348

significant difference in outmigration rates between groups in the 2006 census (p=0.28), but 349

there was a significant difference in outmigration rates in the 2012 census (p=0.02). The majority 350

of families (51.9%) engage in some form of off-farm labor. The most common types of off-farm 351

labor in the city are in construction, as a taxi or truck driver, or factory worker. The most 352

common types of off-farm labor in the Ayllu as a merchant, a tailor, weaver or embroiderer, or in 353

trades such as teacher or veterinarian. The LowRMI group worked off-farm an average of 25.7 354

days/yr more than the HighRMI group. However, this difference is not statistically significance 355

due to high variance. Working in the city is perceived to be a strong constraint to building ECSs, 356

whereas working in the local area was considered a moderate constraint by both the HighRMI 357

and LowRMI respondents. 358

5. Discussion 359

5.1 Participation 360

The results regarding participation indicate that although a similar percentage of people 361

invested in land restoration in both groups, the HighRMI group adopted land restoration 362

13

technologies at earlier stages of project development and sustained higher levels of restoration 363

management. This is consistent with the early adopter-laggards theory of diffusion of innovation 364

(Rogers 2003). In the context of this study, the HighRMI group may have had more innovators 365

and early adopters, and the LowRMI group may have had more of the late adopters and laggards. 366

5.2 Motives and perceived benefits 367

The reported motives and perceived benefits indicate that the rural poor participate in 368

land restoration in order to maintain their livelihoods. Motivations for land restoration included 369

soil and water conservation and food and fiber production. Perceived benefits also included 370

broader aspects of community development such as improved mobility and efficiency, 371

increased restoration knowledge, community development, and conflict resolution. This is 372

consistent with other studies that documented the social benefits of community-based restoration 373

projects. For example, communities planted mangroves for storm protection and fish nurseries in 374

south-east Asia (Walters 2000, Badola and Hussain 2005, Walton et al. 2006). Farmers planted 375

trees and managed natural regeneration for timber, non-timber, and agricultural resources in 376

Ghana (Blay et al. 2008), and the increased tree cover led to additional health and psycho-social 377

benefits (Weston et al. 2015). Watershed rehabilitation was seen as a way to secure access to 378

clean drinking water in Bolivia (Schulz and Soliz 2007). Forest and watershed restoration led to 379

job creation and increased organizational capacity in the western United States (Nielson-Pincus 380

et al. 2013). Land restoration can be a tool for securing local livelihoods (Reed et al. 2015) by 381

stabilizing production, decreasing vulnerability to natural disasters, and increasing prospects for 382

economic development. In contrast, ecosystem restoration is regarded as a globally important 383

strategy for biodiversity conservation and carbon sequestration (Millennium Ecosystem 384

Assessment 2005). However, conflicts may arise when managing for both ecosystem services 385

important to rural livelihoods and biodiversity objectives important to the global community 386

(Bullock et al. 2011). These potentially contrasting objectives must be reconciled for land 387

restoration in impoverished rural areas to be successful. 388

5.3 Incentives 389

Credits and incentives can promote land restoration in economically stressed rural 390

populations that may have to overcome significant barriers to long-term investments (Beyene 391

2009, Mekuria et al. 2011, Schiappacasse et al. 2012). The food-for-work program was 392

14

perceived to be a strong motivational factor and a moderate facilitating factor, but was not 393

commonly cited as a perceived benefit in this study. These results suggest that food-for-work 394

helped incentivize terrace construction in the initial stages of project development. However, it 395

may be only marginally effective in ensuring terrace construction and maintenance over the long 396

term. In a related study probing indigenous knowledge and attitudes in the study area, 397

respondents mentioned that "some people engage in land restoration for the wrong reasons, just 398

to get foodstuffs, and then no longer maintain the erosion controls" (Hartman et al. 2016). This is 399

consistent with a food-for-work program for terrace construction in Ethiopia, where communities 400

were observed to construct terraces but not maintain them over time (Amede et al. 2007). Given 401

the high levels of investments required, some form of incentive to stimulate land restoration may 402

be necessary, and payment for ecosystem services (Montagnini and Finney 2011) and carbon 403

credits (Mekuria et al. 2011) are policy alternatives to food-for-work programs. 404

5.4 Conflict and conflict resolution 405

Previous studies have linked civil conflict and land degradation (Thiesen 2008), and 406

conflict resolution and stakeholder coordination have been identified as integral to community-407

based land management (Leach et. al. 1999, Amede et. al. 2007, Beyene 2009). Conflict was 408

perceived to be a strong constraint to building ECSs, while conflict resolution was perceived to 409

be a moderate facilitating factor. There were lower levels of conflict reported in the HighRMI 410

communities, and higher rates of non-responsive respondents in the LowRMI group. This 411

suggests there might be more conflict in the LowRMI group, but that members were reluctant to 412

reveal or discuss conflicts. It may be easier to answer when there is less conflict, but difficult for 413

community members in tightly knit ayllus to answer candidly when there are high levels of 414

conflict. The regional and collective nature of the watershed restoration efforts in this study 415

would suggest conflict resolution and stakeholder coordination would be particularly important 416

(Hodge and McNally 2000, Gebremedhin et. al. 2004). However, while the levels of conflict 417

have influenced land restoration in the study area, it does not appear to be a determining factor. 418

For example, one HighRMI community (Yawri Totora) cited high rates of conflict and they were 419

still successful in building ECSs. There may be a temporal lag between ECS construction and 420

when conflicts arise, communities may be able to compensate for the high levels of conflict, or 421

conflict may stimulate competition between households with private landholdings. Moreover, it 422

15

is possible that the influence of conflict is mediated by other factors, such as land tenancy and 423

community leadership, as described below. 424

5.5 Land tenancy and other institutional factors 425

Previous studies that have identified the strengthening and adaptation of existing 426

organizations, creation of new institutional arrangements, and conflict resolution as prerequisites 427

for effective land management (Agrawal and Gibson 1999, Leach et. al. 1999, Amede et. al. 428

2007). In collectively managed lands strong institutional frameworks are required to achieve land 429

restoration (Hodge and McNally 2000, Gebremedhin et. al. 2004). A transition from communal 430

land tenancy (ayanokas) to private parcels was observed in some study area communities. While 431

the data suggest a positive relationship between privatization and erosion control construction, 432

ayanoka communities can still achieve meaningful levels of land restoration. For example, 433

Qanaw Pallqa has an RMI of 219.6 ECs/km2, and Thaya Laka has an RMI of 123.1 ECSs/km

2. 434

This suggests that the effect of land tenancy may be contingent on other institutional factors. For 435

example, a significantly higher proportion of ayanoka respondents cited active leadership, 436

community organization, and conflict resolution as a strong facilitating factor for land 437

restoration. This can be interpreted by understanding the way in which construction of erosion 438

controls is coordinated in ayanokas compared to private land tenancy communities. In private 439

tenancy communities, people can organize into small work groups to address land degradation 440

and engage in reciprocal work exchanges. In ayanoka communities, coordination of work days 441

and location of erosion control structures requires consensus building within the whole 442

community. This may occur more frequently when there are low levels of conflict, there is a 443

strong and active community leader, and/or there is a high level of community organization. In 444

addition, this study shows that clear resource use rights, roles, and responsibilities appear to play 445

an important role in mediating and overcoming potential constraints from to land restoration on 446

communal lands. 447

5.6 Road networks 448

The economic and community development benefits associated with increased 449

accessibility (Pender 2004) may extend to land restoration as well. According to local reports 450

and CEADB project records, road networks facilitated erosion control construction by increasing 451

vehicular access to carry people in work groups, development personnel, and tools. The first 452

16

check dams were constructed along the main road, starting at the headwaters of each gully and 453

working progressively down slope as gullies were stabilized. The higher rate of ECS 454

construction along the main road is corroborated by a related remote sensing study (Hartman et 455

al. 2016), where the greatest increase in greenness (∆NDVI) occurred in an approximately 1 km 456

buffer zone along the Cbba-Oruro Highway. The benefits of increased accessibility must be 457

balanced with the potential for increased gully erosion as culverts discharge high energy water 458

flows from a single point source (Harden 2001, Jungerius 2002, Nyssen et. al. 2002, Valentin et. 459

al. 2005). Some local people perceived roads to contribute to erosion and increased need for 460

ECSs, with 3% of the HighRMI respondents citing a link between roads and gully erosion. 461

Although accessibility influences where ECSs are built at the community or regional 462

level, a different pattern emerges at the household level. There appears to be a significant 463

disincentive to build ECSs for community members that live along the paved Cbba-Oruro 464

highway. However, the difference in accessibility from living on a secondary road, compared to 465

a footpath, does not influence the rate of ECS construction. The reduced rate of ECS 466

construction for households living along the Cbba-Oruro highway is likely due to the ease with 467

which families can engage in alternative economic activities dependent on a main road (e.g. 468

merchant, taxi driver) and the ease of transportation to cities. A person living along the main 469

road has ready access to buses and can reach the city in 2 - 3 hours. A person living on a 470

secondary road or a footpath has to walk significant distances (up to 18 km) to get to the main 471

road and catch a bus and may have to invest a full day to reach the city. This may influence 472

people's decision-making in terms of the trade-offs between local investment and off-farm labor, 473

as discussed below. 474

5.7 Outmigration and off-farm labor 475

Alternative economic opportunities may distract from local investments in land 476

restoration. This includes outmigration and off-farm labor, which can be responses of rural 477

farmers no longer able to secure livelihood from land degraded by drought, soil erosion, 478

deforestation, or environmental disasters (Gray 2009, Reuveny 2009). Outmigration and off-farm 479

labor are forms of income diversification leading to decreased dependency on agriculture, 480

abandonment of marginal lands, and a decline in labor-intensive cultivation methods. For these 481

reasons a greening trend has been associated with decreased land use pressure in several tropical 482

17

biomes (Grau et. al. 2003, Olsson et. al. 2005, Baptista and Rudel 2006, Hecht and Saatchi 2007, 483

Kull et. al. 2007, Izquierdo et. al. 2008, Calvo-Alvarado 2009). However in other cases, 484

increased outmigration and off-farm labor led to continued net land degradation (Garcia-Barrios 485

2009). 486

People have been moving to urban areas at high rates throughout the Andes (Suarez and 487

Torrelaba 1982; Zimmerer 1993; Gray 2009). Outmigration is a relatively recent phenomenon in 488

the study area, beginning in the late 1990s. There was a decreased rate of outmigration in the 489

HighRMI compared to the LowRMI communities, and this difference only became apparent in 490

the later years of project development between 2006 and 2012. This suggests that although land 491

degradation may have stimulated outmigration, land restoration may have reduced the rate of 492

outmigration. Land restoration may have played a role in families' decision-making process, with 493

a greater proportion of people deciding to stay and invest in local natural resources rather than 494

build a life in the city. The decreased rate of rural-urban migration in HighRMI communities is 495

noteworthy, as the dislocation of indigenous peoples from their territory can lead to a loss of 496

language, culture, and identity (insert cite). If land restoration can help stem the tide of rural-497

urban migration and serve as a vehicle for institution building in indigenous territories, it can 498

contribute to cultural survival. 499

While there is a significant difference in the outmigration rates, the difference in off-farm 500

labor rates is not so clear. Once a family has made the decision to stay in the region, off-farm 501

labor appears to be a tool to supplement economic needs and maintain rural livelihoods. The 502

LowRMI group worked off-farm an average of 25.7 days/yr more than the HighRMI group. 503

Although not statistically significant due to high variance, if the LowRMI group is off-farm for 504

almost a month more than the HighRMI group it could have practical ramifications. There may 505

be a trade-off occurring, as people allocate resources between farm and off-farm activities to 506

meet the families’ economic needs. This trade-off is highly variable among families. While, on 507

average, more ECSs correlates with fewer days spent working off-farm, at least some HighRMI 508

respondents indicated high rates of off-farm labor (300+ days/yr) and still built high numbers of 509

ECSs. This may be due to a temporal lag between erosion control construction and current off-510

farm labor rates, or due to good social networks and participation in strong work groups. The 511

data also suggest local off-farm labor is a more effective tool to manage the trade-offs between 512

off-farm labor and local investments in land restoration. When working locally, a person can 513

18

spend a day laboring, and still attend to the needs of the farm in the mornings and evening, 514

whereas work in the city requires an extended absence. This suggests that policies that promote 515

local economic development may create synergies with land restoration. There was no evidence 516

of the use of remittances to pay laborers to help build erosion controls (Gray 2009). 517

6. Conclusions 518

Results of this study suggest that local and indigenous people invest in land restoration to 519

ensure productivity and economic stability. If local and indigenous people define restoration 520

objectives based on productivity and other ecosystem services, it may contrast with restoration 521

objectives based on biodiversity or carbon sequestration (Bullock et. al. 2011). In a meta-522

analysis of restoration projects designed to increase biodiversity, Rey Benayas (2009) and Barral 523

et al. (2015) found that ecosystem services also increased. However, the effect of restoration 524

projects explicitly designed to restore ecosystem services to enhance rural livelihoods has not 525

been sufficiently evaluated. For example, restoration designed to increase productivity of grazing 526

lands may negatively impact plant diversity. If restoration objectives are defined by outside 527

donor groups, these must be reconciled with local land use priorities for restoration to be 528

successful. 529

This study highlights the interdependencies between social processes and land restoration 530

in impoverished rural environments. This implies that concomitant investments in land 531

restoration and rural development will increase the effectiveness of restoration initiatives. This is 532

analogous to Integrated Conservation and Development Projects, which aim to achieve 533

biodiversity conservation goals through investments in social and economic development (Alpert 534

1995, Hughes and Flint 2001). Further research into these interdependencies may help expand 535

the reach of land restoration into rural development settings. This study suggests that land 536

restoration projects that include rural development components such as credits and incentives, 537

institution building, leadership development, conflict resolution, and increased accessibility will 538

have a higher success rate. Moreover, land restoration can slow rural-urban migration and 539

provide new ways for local people to invest in local resources rather than move to urban areas. 540

Indigenous people – such as the Aymara in this study – have a strong identity with their territory. 541

If land restoration can help slow the loss of people from indigenous territories, it has important 542

implications for the survival, empowerment, and self-determination of indigenous cultures. 543

19

7. References 544

Agrawal, A. and C. Gibson. 1999. Enchantment or disenchantment: the role of community in natural 545 resource conservation. World Development 27(4):629-649. 546

Alpert, P. 1995. Applying ecological research at integrated conservation and development projects. 547 Ecological Applications 5(4): 857-860. 548

Amede, T., H. Kassa, G. Zeleke, A. Shiferaw, S. Kismu, and M. Teshome. 2007. Working with 549 communities and building local institutions for sustainable land management in the Ethiopian 550 Highlands. Mountain Research and Development 27(1):15-19. 551

Anderies, J., M.A. Janssen and E. Ostrom. 2004. A framework to analyze the robustness of social-552 ecological systems from an institutional perspective. Ecology and Society 9(1):18. 553

Aronson, J. J.M. Blignaut, S.J. Milton, D. Le Maitre, K.J. Esler, A. Limouzin, C. Fontaine, M.P. de Wit, 554 W. Mugido, P. Prinsloo, L. van de Elst, amd N. Lederer. 2010. Are socioeconomic benefits of 555 restoration adequately quantified? A meta-analysis of recent papers (2000-2008) in Restoration Ecology 556 and 212 other scientific journals. Restoration Ecology 18(2):143-154. 557

Badola, R. and S.A. Hussain. 2005. Valuing ecosystem functions: an empirical study on the storm 558 protection function of Bhitarkanika mangrove ecosystem, India. Environmental Conservation 32(1):85-559 92. 560

Barral, M.P., J.M. Rey Benayas, P. Meli, and N.O. Meceira. 2015. Quantifying the impacts of ecological 561 restoration on biodiversity and ecosystem services in agroecosystems: a global meta-analysis. 562 Agriculture, Ecosystems and Environment 202:223-231. 563

Baied, C.A. and J.C. Wheeler. 1993. Evolution of High Andean puna ecosystems: environment, climate, 564 and culture change over the last 12,000 years in the Central Andes. Mountain Research and 565 Development 13(2):145-156 566

Baptista, S.R. and T.K. Rudel. 2006. A re-emerging Atlantic forest? Urbanization, industrialization and 567 the forest transition in Santa Catarina, southern Brazil. Environmental Conservation 33(3):195-202. 568

Beyene, F. 2009. Exploring incentives for rangeland enclosures among pastoral and agropastoral 569 household in eastern Ethiopia. 19:494-502. 570

Blangy, S. and H. Mehta. 2006. Ecotourism and ecological restoration. Journal for Nature Conservation 571 14:233-236. 572

Blay, D., M. Appiah, L. Damnayag, F. Dwomoh, O. Luukkanen, and A. Pappinen. 2008. Involving local 573 farmers in rehabilitation of degraded tropical forests: some lessons from Ghana. Environment, 574 Development and Sustainability 10:503-518. 575

Brandt, J.S. and P.A. Townsend. 2006. Land use – land cover conversion, regeneration and degradation in 576 the high elevation Bolivian Andes. Landscape Ecology 21:607-623. 577

Brown, S. and A.E. Lugo. 1994. Rehabilitation of tropical lands: a key to sustaining development. 578 Restoration Ecology 2:97-111. 579

Bullock, J.M., J. Aronson, A.C. Newton, R.F. Pywell, and J.M. Rey-Benayas. 2011. Restoration of 580 ecosystem services and biodiversity: conflicts and opportunities. Trends in Ecology and Evolution 581 26(10):541-549. 582

Burke, S.M. and N. Mitchell. 2007. People as ecological participants in ecological restoration. Restoration 583 Ecology 15(2):348-350. 584

20

Calvo-Alvarado, J, B. McLennan, A. Sanchez-Azofeifa, and T. Garvin. 2009. Deforestation and forest 585 restoration in Guanacaste, Costa Rica: Putting conservation policies in context. Forest Ecology and 586 Management 258(6):931-940. 587

Carter, M., P. Little, T. Mogues, and W. Negatu. 2007. Poverty traps and natural disasters in Ethiopia and 588 Honduras. World Development 35(5):835-856. 589

Chepstow-Lusty, A.J., K.D. Bennet, J. Fjelda, A. Kendall, W. Galliano, and A.T. Herrera. 1998. Tracing 590 4,000 years of environmental history in the Cuzco area, Peru, from the pollen record. Mountain 591 Research and Development 18(2):159-172. 592

Collins, A.E. 2001. Health ecology, land degradation and development. Land Degradation and 593 Development 12:237-250. 594

Delgado Burgoa, J.M.F. 2001. Simbiosis interzonal en las estrategias de autodesarrollo sostenible en 595 ecosistemas de montaña: el caso del Ayllu Mujlli,Departamento de Cochabamba, Bolivia (Doctoral 596 Dissertation). Universidad de Cordoba, España. 597

Ellenberg, H. 1979. Man's influence on tropical mountain ecosystems in South America. Journal of 598 Ecology 67:401-416. 599

Garcia-Barrrios, L., Y.M. Galvan-Miyoshi, I.A. Valdivieso-Perez, O.R. Masera, G. Bocco, and J. 600 Vandermeer. 2009. Neotropical forest conservation, agricultural intensification, and rural out-601 migration: the Mexican experience. BioScience 59(10):863-873. 602

Gebremedhin, B, J. Pender, and G. Tesfay. 2004. Collective action for grazing land management in crop-603 livestock mixed systems in the highlands of northern Ethiopia. Agricultural Systems 82(3):273-290. 604

Grau, H. R., T. M. Aide, J. K. Zimmerman, J. R. Thomlinson, E. Helmer, and X. Zou. 2003. The 605 ecological consequences of socioeconomic and land-use changes in postagricultural Puerto Rico. 606 Bioscience 53:1159–1168. 607

Gisladottir, G. and M. Stocking. 2005. Land degradation control and its global environmental benefits. 608 Land Degradation and Development 16:99-112. 609

Gray, C.L. 2009. Environment, land, and rural out-migration in the southern Ecuadorian Andes. World 610 Development 37(2):457-468. 611

Harden, C.P. 2001. Soil erosion and sustainable mountain development: experiments, observations and 612 recommendations from the Ecuadorian Andes. Mountain Research and Development 21:77-83. 613

Hartman, B.D. 1996. Sociocultural constraints to land management decisions: the case of bofedal 614 restoration in Bolivia. Tropical Resources Institute 15(1):24-28. 615

Hartman, B.D., B. Bookhagen, and O.A. Chadwick. 2016a. The effects of check dams and other erosion 616 control structures on the restoration on Andean bofedal ecosystems. Restoration Ecology 24(5):1-12. 617

Hartman, B.D., D.A. Cleveland, and O.A. Chadwick. 2016b. Linking changes in knowledge and attitudes 618 with successful land restoration in indigenous communities. Restoration Ecology 24(5):1-12. 619

Hecht, S.B, and S.S. Saatchi. 2007. Globalization and Forest Resurgence: Changes in Forest Cover in El 620 Salvador. BioScience 57(8): 663-672. 621

Hughes and Flint (2001) 622

Hobbs, R.J., and D.A. Norton. 1996. Towards a conceptual framework for restoration ecology. 623 Restoration Ecology 4(2):93-110. 624

Hodge, I. and S. McNally. 2000. Wetland restoration, collective action and the role of water management 625 institutions. Ecological Economics 35:107-118. 626

21

Izquierdo, A.E., C.D. DeAngelo, T.M. Aide. 2008. Thirty Years of Human Demography and Land-Use 627 Change in the Atlantic Forest of Misiones, Argentina: an Evaluation of the Forest Transition Model. 628 Ecology and Society 13(2):3. 629

Jungerius, P.D., J. Matundura, and J. Van de Ancker. 2002. Road construction and gully erosion in West 630 Pokot, Kenya. Earth Surface Processes and Landforms 27:1237-1247. 631

Kull, C.A., C.K. Ibrahim and T.C. Meredith. 2007. Tropical forest transitions and globalization: neo-632 liberalism, migration, tourism, and international conservation agendas. Society and Natural Resources 633 20:723-737. 634

La Fuente, M.C.T. 1997. Una propuesta metodologica para la comunicacion para el desarrollo: el uso del 635 Diagnostico Rural Participativo en la investigacion de actitudes. Nur University, Final Thesis for the 636 Licenciatura en Ciencias de Comunicacion Social. Santa Cruz, Bolivia.136 p. 637

Lamb, D., P. Erskine and J. Parrotta. 2005. Restoration of degraded tropical forest landscapes. Science 638 310:1628-1632. 639

Lambin, E., H.J. Geist, and E. Lepers. 2003. Dynamics of Land-use and land-cover change in tropical 640 regions. Annual Review of Environment and Resources 28:205-241. 641

Leach, M., R.Mearns, I. Scoones. 1999. Environmental entitlements: dynamics and institutions in 642 community-based natural resource management. World Development 27(2):225-247. 643

Long, J., A. Tecle, and B. Burnette. 2003. Cultural foundations for ecological restoration on the White 644 Mountain Apache reservation. Conservation Ecology 8(1):4. 645

Mekuria, W., E. Veldkamp, M. Tilahun, and R. Olschewski. 2011. Economic valuation of land 646 restoration: the case of esclosures established on communal grazing lands in Tigray, Ethiopia. Land 647 Degradation & Development 22:334-344. 648

Millennium Ecosystem Assessment. 2005. Ecosystems and human well-being: synthesis. Island Press, 649 Washington, D.C. 650

Montagnini, F. and C. Finney. 2011. Payments for environmental services in Latin America as a tool for 651 restoration and rural development. Ambio 40:285-297. 652

Myers, N. 1997. Environmental refugees. Population and the Environment 19(2):167-182. 653

Nielson-Pincus, M. and C. Moseley. 2013. The economic and employment impacts of forest and 654 watershed restoration. Restoration Ecology 21(2):207-214. 655

Nyssen, J. J. Poessen, J. Moeyersons, E. Luyten, M. Veyret-Picot, J.Deckers, M. Haile, and G. Govers. 656 2002. Impact of road building on gully erosion risk: a case study from the northern Ethiopian highlands. 657 Earth surface processes and landforms 27:1267-1283. 658

Olsson, L., L. Eklundh, J. Ardo. 2005. A recent greening of the Sahel – trends, patterns and potential 659 causes. Journal of Arid Environments 63:556-566. 660

Patz, J.A., P. Daszak, G. Tabor, A. Alonso Aguirre, M. Pearl, J. Epstein, N.D. Wolfe, A.M, Kilpatrick, J. 661 Foufopoulos, D. Molyneux, D. Bradley. 2004. Unhealthy landscapes: policy recommendations on land 662 use change and infectious disease emergence. Environmental Health Perspectives 112:1092-1098. 663

Pender, J. 2004. Development pathways for hillsides and highlands: some lessons from Central America 664 and East Africa. Food Policy 29:339–367. 665

Reed, M.S., L.C. Stringer, A.J. Douglii, J.S. Perkins, J.R. Atlhopheng, K. Mulale. 2015. Reorienting land 666 degradation towards sustainable land management: linking sustainable livelihoods with ecosystem 667 services in rangeland systems. Journal of Environmental Management 151:472-485. 668

22

Reuveny, R., W.H. Moore. 2009. Does environmental degradation influence migration? Emigration to 669 developed countries in the late 1980s and 1990s. Social Science Quarterly 90(3):461-479. 670

Rey Benayas, J.M., A.C. Newton, A. Diaz, and J.M. Bullock. 2009. Enhancement of biodiversity and 671 ecosystem services by ecological restoration: a meta-analysis. Science 325:1121-1124. 672

Rist S., F. Delgado Burgoa, U. Wiesmann. 2003. The role of social learning processes in the emergence 673 and development of Aymara land use systems. Mountain Research and Development 23(3):263–270. 674

Rogers, E.M. 2003. Diffusion of Innovations (5th Edition). New York: Free Press. 576 p. 675

Scherr, S. 2000. A downward spiral? Research evidence on the relationship between poverty and natural 676 resource degradation. Food Policy 25: 479-498. 677

Schiappacasse, I., L. Nahuelhual, F. Vasquez, C. Echeverría. 2012. Assessing the benefits and costs of 678 dryland forest restoration in central Chile. Journal of Environmental Management 97:38-45. 679

Schulz, S. and B. Soliz. 2007. Stakeholder willingness to pay for watershed restoration in rural Bolivia. 680 Journal of the American Water Resources Association 43(4):947-956. 681

Shackelferd, N., R.J. Hobbs, J.M. Burgar, T.D. Erickson, J.B. Fontaine, E. Laliberte, C.E. Ramalho, M.P. 682 Perring, and R.J. Standish. 2013. Primed for Change: Developing Ecological Restoration for the 21st 683 Century. Restoration Ecology 21(3):297-304. 684

Siebert, P. 1983. Human impact on landscape and vegetation in the central high Andes. pp. 261-276 in 685 Holzner, M. J.A. Werger, and I. Ikusima (eds.) Man's impact on vegetation. Dr. W. Junk Publishers, 686 The Netherlands. 687

Squeo, F.A., B.G. Warner, R. Aravena, and D. Espinosa. 2006. Bofedales: high altitude peatlands of the 688 central Andes. Revista Chilena de Historia Natural 79:245-255. 689

Stringer, L.C., C. Twyamn, and D.S.G. Thomas. 2007. Combating land degradation through participatory 690 means: the case of Swaziland. Ambio 36(5):387-393. 691

Suarez, M.M. and R. Torrealba. 1982. Recent trends in human migrations: the case of the Venezuelan 692 Andes. Mountain Research and Development 2:299-306. 693

Temperton, V.M. 2007. The recent double paradigm shift in restoration ecology. Restoration Ecology 694 15(2):97-111. 695

Theisen, O.M. 2008. Blood and soil? Resource scarcity and internal armed conflict revisited. Journal of 696 Peace Research 45:801-818. 697

Valentin, C., J. Poesen, Yong Li. 2005. Gully erosion: Impacts, factors and control. Catena 63:132–153. 698

Walters, B.B. 2000. Local mangrove planting in the Philippines: are fisherfolk and fishpond owners 699 effective restorationsits? Restoration Ecology 8(3):237-246. 700

Walton, M.E.M., G.P.B. Samonte-Tan, J.H. Primavera, G. Edwards-Jones, and L. LeVay. 2006. Are 701 mangroves worth replanting? The direct economic benefits of a community-based reforestation project. 702 Environmental Conservation 33(4):335-343. 703

Washington-Allen, R.A., R.D. Ramsey, N.E. West and B.E. Norton. 2008. Quantification of the 704 ecological resilience of drylands using digital remote sensing. Ecology and Society 13:33 705

Weston, P., R. Hong, C. Kaboré, C.A. Kull. 2015. Farmer-managed natural regeneration enhances rural 706 livelihoods in dryland West Africa. Environmental Management 55:1402-1417. 707

Zimmerer, K.S. 1993. Soil erosion and labor shortages in the Andes with special reference to Bolivia, 708 1953-91: Implications for “conservation-with-development.” World Development 21(10):1659-1675. 709

23

710 Figure 1. Location of the HighRMI, LowRMI, and Non-project control communities, watershed 711

rehabilitation and wet meadow (bofedal) restoration project in the Ayllu Majsaya-Aransaya-712

Urunsaya. The HighRMI, LowRMI and NonProject communities were selected from communities 713

that participated in land restoration after controlling for social and biophysical variables and based 714

on the levels of Restoration Management Intensity, where RMI = No of Erosion Control Structures / 715

km2. Erosion control structures included check dams, terraces, and infiltration ditches in each 716

project community. RMI was not calculated for the NonProject communities. There were two 717

additional communities included in the study (Thaya Laka and Chulpani) that had low to medium 718

levels of RMI. 719

24

HighRMI LowRMI

Food-for-work

p=0.05

Conflict in the

community

RE

SP

ON

SE

FR

EQ

UE

NC

Y

p=0.25

Conflict

resolution

p=0.01

Land

privatization

p=0.44

Collective

lands

p=0.81

Community

organization

p=0.001

Active leaders

p=0.001

Road

networks

p=0.06*

Work in the

city

p=0.42

Work in the

Ayllu

p=0.44

Figure 3. Facilitation and constraints analysis, including perceptions of factors that facilitate or constrain land restoration, as 720

measured by the effect of selected social factors on their ability to construct check dams and other erosion controls. Data from a 721

household survey (n=210), with answers recorded on an ordinal scale from (-2) Strong constraint, (-1) moderate constraint, (0) 722

Neutral, (1) Moderate facilitation, (2) Strong facilitation. Responses are evaluated by an Independent Samples Mann-Whitney U 723

Test and Independent Samples Median Test (non-parametric test used to state whether the two groups have significant 724

asymptotic differences in the histogram distribution). p-values indicate the asymptotic significance at a significance level of 0.05, 725

with (*) denoting conditional significance if p<0.10. 726

0.00 0.00 0.11

0.69

0.18

0.00

0.80

0.00 0.01 0.16

0.52

0.31

0.00

0.80

0.69

0.21 0.10

0.00 0.00

0.00

0.80 0.75

0.16 0.07

0.00 0.00

0.00 0.00

0.27

0.61

0.10

0.00

0.80

0.00 0.00 0.14

0.65

0.19

0.00

0.80

0.00 0.01

0.26

0.46

0.25

0.00

0.80

0.00 0.04

0.30 0.39

0.26

0.00

0.80

0.31 0.14 0.07 0.07

0.31

0.00

0.80

0.23 0.20 0.11

0.19 0.23

0.00

0.80

0.00 0.00 0.16

0.59

0.25

0.00

0.80

0.00 0.00 0.01

0.52 0.47

0.00

0.80

0.00 0.01 0.11

0.49 0.39

0.00

0.80

0.00 0.00 0.03

0.32

0.64

0.00

0.80

0.00 0.03 0.09

0.32

0.57

0.00

0.80

0.00 0.00 0.05

0.25

0.69

0.00

0.80

0.68

0.18 0.14 0.00 0.00

0.00

0.80 0.73

0.13 0.13 0.01 0.00

0.00

0.80

0.33 0.44

0.20 0.03 0.00

0.00

0.80

-2 -1 0 1 2

0.43 0.28 0.24

0.04 0.00

0.00

0.80

-2 -1 0 1 2

Facilitation Constraint Facilitation Constraint

25

Table 1. Characteristics of the study communities, watershed rehabilitation and wet meadow (bofedal) restoration 727

project in the Ayllu Majsaya-Aransaya-Urunsaya. 728

Area Elevation Location

Land

Tenancya

No

Familiesb Erosion Controls (ECs) RMI

c

km2 m a.s.l. Latitude Longitude Check

Dams Terraces Infiltr.

Ditches Total

No ECs

(EC/km

2)

HighRMI Communities

Lakolakoni 32.84 4,126-4,594 17⁰34’46”S 66⁰56’11”W Private 81 3400 2600 1400 7400 225.3

Qañaw Pallqa 17.76 3,846-4,397 17⁰44’31”S 66⁰44’50”W Ayanoka 52 500 1200 2200 3900 219.6

Yawri Totora 15.09 4,019-4,480 17⁰37’55”S 66⁰49’28”W Private 71 1200 1100 900 3200 212.1

Pasto Grande 14.41 4,027-4,480 17⁰39’32”S 66⁰47’40”W Private 38 2000 320 250 2570 178.4

LowRMI Communities

Waylla Tambo 7.21 3,934-4,384 17⁰44’10”S 66⁰47’26”W Ayanoka 38 300 120 110 530 73.5

Japo 28.84 4,074-4,390 17⁰41’45”S 66⁰45’50”W Private 87 1100 420 20 1540 53.4

Tola Marka 14.58 3,986-4,595 17⁰36’22”S 66⁰49’00”W Private 25 110 180 40 330 22.6

Tallija 63.32 3,821-4,430 17⁰40’45”S 66⁰43’49”W Ayanoka 129 580 320 30 930 14.7

Other Project Communities

Thaya Laka 7.47 4,006-4,366 17⁰45’04”S 66⁰47’06”W Ayanoka 24 400 240 280 920 123.1

Chulpani 11.05 3,961-4,375 17⁰47’56”S 66⁰46’23”W Ayanoka 47 420 400 80 900 81.4

aPrivate refers to communities that have allocated titled landholdings (average size=40 Has.) to individual families; Ayanoka refers to communities that have 729 maintained the traditional Aymara communal system of land tenancy 730

bNo of Families resident in each community, from a census conducted within the project area in 2006 to determine outmigration rates 731 cRMI is a measure of the Restoration Management Intensity in each project community. RMI = Total No of Erosion Control Structures (Total ECSs) per Km2, 732

with Total ECSs = sum total of the check dams, terraces, and infiltration ditches in each project community. RMI was not calculated for the NonProject 733 community. 734

26

Table 2. Social variables used in the study. The study variables were selected based on a literature review, prior 735

research conducted in the region, and project records for the watershed rehabilitation and wet meadow (bofedal) 736

restoration project in the Ayllu Majsaya-Aransaya-Urunsaya. Further explanation of the study variables and their 737

measurement is provided in the household survey section. 738

Study variable Measurement

Participation

Percent participation Closed question (Yes/No)

Year begin participation Closed question (Year)

Erosion Control Structures(ECSs)

Check dams (No/family) Closed question (No/family)

Terraces (No /family) Closed question (No/family)

Infiltration ditches (No /family) Closed question (No/family)

Other ß (No /family) Closed question (No/family)

Total ECs (No /family) Closed question (No/family)

Motives and perceived benefits

Motives Open question (No of responses, % response)a

Perceived benefits Open question (No of responses, % response)a

Incentives (food-for-work) Scaled response (facilitation and constraints analysis)b

Institutional factors

Land tenancy Closed question (ayanoka vs. private); scaled response (facilitation and constraints analysis) b

Active leadership Scaled response (facilitation and constraints analysis)b

Community development Scaled response (facilitation and constraints analysis)b

Conflict resolution Open question; scaled response (facilitation and constraints analysis)b

Road networks Closed question (household location); scaled response (facilitation and constraints analysis)b

Outmigration and off-farm labor

Family members living in city Closed question (No/family)

Families emigrated/community Closed question (% families/community)

Working in city Closed question (No days/yr per family); scaled response (facilitation and constraints analysis)b

Work in the ayllu Closed question (No days/yr per family); scaled response (facilitation and constraints analysis)b

Total off-farm labor Closed question (No days/yr per family, aggregated)

a Based on coded responses to open ended questions about the motives and perceived benefits of participating in land restoration, and is expressed as a count 739 (No of responses) and the percent of respondents that cited a given response in the the HighRMI and LowRMI groups. 740

b The Facilitation and constraints analysis measured perceptions of whether selected factors helped, hindered or had a neutral effect on building erosion control 741 structures. For more explanation, see the household survey section below. 742

27

Table 3. Comparison of study variable data in HighRMI and LowRMI communities, watershed rehabilitation and 743

wet meadow (bofedal) restoration project in the Ayllu Majsaya-Aransaya-Urunsaya. 744

HighRMIa (n=114) LowRMIa (n=96) t-Testb,

Chi-squarec

x̄ ± SE x̄ ± SE p-value

Participation

Percent of respondents participating 90.4 83.3 1.21c 0.27

Year respondent began participation 1998.1 ± 0.3 2000.2 ± 0.4 9.32b 0.01

Erosion Control Structures d

Check dams (No/family) 35.0 ± 6.0 7.3 ± 11.3 4.19b <0.001

Terraces (No/family) 23.9 ± 5.2 4.7 ± 1.0 3.35b <0.001

Infiltration ditches (No/family) 15.9 ± 4.4 1.0 ± 0.4 3.12b <0.001

Othere (No/family) 0.4 ± 0.1 0.1 ± 0.03 2.16b 0.001

Total ECSs (No/family) 75.3 ±11.7 13.0 ± 4.1 4.86b <0.001

Conflict and conflict resolutionf

Less conflict (%) 67.5 46.9 9.30c 0.026

Same level of conflict (%) 6.1 10.4

More conflict (%) 6.1 8.3

Did not respond (%) 20.2 34.4

Outmigration

No family members living in city 0.93 ± 0.15 1.08 ± 0.18 0.67b 0.51

Families emigrated/community (%)g 29.1 ± 13.3 70.3 ± 6.4 3.00b 0.02

Off-farm labor /householdh

Work in city (No days/yr) 56.2 ± 10.8 72.4 ± 11.3 1.04b 0.30

Working in the Ayllu (No days/yr) 25.7 ± 6.1 34.6 ± 9.0 0.84b 0.40

Total off-farm labor (No days/yr) 81.7 ± 11.5 107.0 ± 13.0 1.46b 0.15

a The HighRMI and LowRMI communities provided the data frame and were selected after controlling for social and biophysical variables, and based on the 745 levels of Restoration Management Intensity (RMI = No of Erosion Control Structures / Km2). The sampling unit was the household, with a single male or 746 female head of household responding for each household. 747

b Two-tailed t-Test for Equality of Means, following Levene’s Test for Equality of Variances 748 c Pearson’s Chi-Square Test 749 d In Ayanoka communities respondents estimated the No of ECSs in the vicinity of their residence, using an area equivalent to a private parcel (42 ha). 750 e Other erosion control structures included slow forming terraces, diversion ditches, grazing enclosures, improved pasture (Phalaris sp.), and reforestation 751 f Response rates to the question “is there more or less conflict in your community compared to when the land restoration project began” 752 g Families emigrated/community was calculated from the 2006 and 2012 census information. The 2006 census contained numbers of resident families, as 753

well as the number of emigrant families. In 2012, the number of resident families was recorded based on examination of community records and verified 754 by community leaders. Therefore, the No of families emigrated/community (%) = (2006 resident families + 2006 emigrant families) – 2012 resident 755 families)/ (2006 resident families + 2006 emigrant families). Two communities with low- to medium levels of RMI (Thaya Laka and Chullpani) were 756 included in the LowRMI group. 757

h Respondents quantified the No of days the men, women, and youth had worked in the city in the past year. Responses were recorded from 0 – 365 days/yr 758 and summed for the whole family. 759

i Calculated by adding the No of days for off-farm labor in the city and in the Ayllu for each household 760

28

Table 4. Comparison of motives for and perceived benefits of land restoration in HighRMI and LowRMI 761

communities, Ayllu Majsaya-Aransaya-Urunsaya 762

High-RMI (n=103)a Low-RMI (n=81)a

% % p-valueb

Motiveb

To combat soil erosion and the loss of land 83.5 63.0 0.002

Because of the ‘food-for-work’ program 63.1 24.7 0.000

Due to soil desiccation 44.7 64.2 0.008

Due to the lack of water 40.8 39.5 0.86

Due to reduced agricultural production 17.5 23.5 0.32

Due to reduced forage production 10.7 13.6 0.55

Due to the contact with development organizations 3.9 1.2 0.27

Because the leaders made it obligatory 2.9 1.2 0.40

To reduce runoff and sedimentation 2.9 0.0 0.17

To combat river scour and the loss of land 1.9 4.9 0.25

Awareness was raised and the whole community participated 1.0 1.2 0.69

We saw that the project was useful 1.0 2.5 0.41

Due to my own initiative 0.0 1.2 0.44

Perceived benefitb

Pasture rehabilitation 73.7 58.1 0.008

Increased animal mobility and reduced injury 72.8 66.7 0.20

Water ‘filtration’ and increased soil moisture 69.3 34.4 0.000

Increased land restoration knowledge 33.3 39.8 0.43

Increased agricultural production 24.6 31.2 0.36

Bofedal restoration 24.6 18.3 0.23

Resolution of conflicts in community 8.8 2.2 0.04

Trees for housing and firewood 7.9 21.5 0.007

Nutrition from the ‘food-for-work’ program 6.1 9.7 0.38

Soil conservation and erosion control 6.1 8.6 0.54

Increased efficiency (land consolidation) 5.3 3.2 0.45

Reduced runoff and river scour 1.8 6.5 0.27

Increased community organization 3.5 2.2 0.42

Increased vehicle and pedestrian mobility 3.5 0.0 0.09

Did not see a benefit 2.6 4.3 0.41

New crops and new planting areas 1.8 1.1 0.56

The terraces act as windbreaks, frost protection 1.7 0.0 0.54

a The sample size for the motives is lower because the question "why did you participate in the land restoration project" was only asked of respondents 763 that had answered that they had participated in the land restoration project. All respondents were asked if they had seen any benefits from land 764 restoration regardless of whether or not they had participated, so the sample size for perceived benefits is HighRMI (n=114) and LowRMI (n=96) 765

bResponses were checked off from a list of pre-coded motives and perceived benefits that was developed based on semi-structured interviews 766 conducted in 2008 and during the pre-testing of the questionnaire in 2012. If respondents listed more than one motive or benefit, each response 767 was recorded, but if the same motive or benefit was repeated it was only recorded once. ‘Did not respond’ was included as a potential response 768 and ‘Others’ was provided as an additional line to record answers not included in the original list of coded responses. 769

cp-values are reported from a Pearson’s Chi-Square Test. If the number of observations was ≤5, p-values were reported from Fisher’s Exact Test. 770

29

Table 5. Comparison of the Total ECSs/family and household location 771

Accessibility Total ECSs (No / family)

x̄ ± SE df F p-value

Household locationa 200 4.22 0.016

1. Cbba-Oruro Hwyß 7.8 ± 2.3

2. Secondary Road 61.4 ± 10.4

3. Footpaths 51.5 ± 10.4

a The household location was recorded as the name of the rancho and any additional identifying information, including whether 772 the household was located on the paved Cbba-Oruro highway, on a graveled secondary road, or in a remote area that was 773 only accessible by footpaths. 774

ß The mean difference between [1 and 2] and [1 and 3] is significant at the <0.01 level, Brown-Forsythe One Way ANOVA 775 following a Levene's F Test for Equality of Variances, with a Games-Howell post-Hoc multiple comparison of means. The 776 mean difference between 2 and 3 is not significant. 777