**3. Results**

### *3.1. Forest to Pasture Conversion*

Log distributions of sedimentary charcoal accumulation (CHAR) revealed that fire was uncommon until the early Neolithic, and results from colluvial records that began between ca. 18,000 BP (Mulhedoy) and ca. 14,000 (Ihitsaga) indicated that CHAR values were not detectable (<0.0001 mm<sup>2</sup> cm<sup>−</sup><sup>2</sup> yr<sup>−</sup>1) at two locations (Ihitsaga, Mulhedoy) until the Middle Holocene (between 6 and 7000 BP). Sporadic fires were evident prior to 9500 BP at a third location (Ibarandoua), ye<sup>t</sup> appeared insignificant until ca. 8000 BP. The evidence suggests that low-severity burning activity progressively increased over several millennia and a human grazing-fire regime was not established in all catchments until about 6000 BP. The site of Mulhedoy registered the earliest CHAR peak ca. 3850 cal BP (middle Bronze Age), followed by the site of Ibarrandoua some 1550 years later at ca. 2280 cal BP (late Iron Age), while the site of Ihitsaga ended with a subtle peak ca. 1900 cal BP (Antiquity). We sugges<sup>t</sup> that the cycling between pronounced peaks and declines in CHAR levels indicate the transitional tipping points between forest and grassland states at a landscape scale (Figure 4).

Archaeological survey located over 100 prehistoric/early historic sites and nearly 200 features (i.e., hunting blinds, cabins, corrals, tumulus). There are only six archaeological radiocarbon dates earlier than 3000 BP from sites discovered in Larrau previous to our own research. CHAR peaks at the sites of Ibarandoua and Mulhedoy fell within the confidence intervals for two early composite archaeological radiocarbon distributions (Figure 4). While suggestive of an association between burning activities and human occupation of upland areas, the highest density of archaeological radiocarbon dates occurred in the last millennia following a 1000-year decline in CHAR values. Thus, while burning and human presence are circumstantially related, the evidence is not su fficient to confirm or deny associated land use, much less social and economic organization.

**Figure 4.** Charcoal accumulation (Log scale) for the Holocene period with the number of radiocarbon dates from archaeological contexts by probability distribution [83].

### *3.2. Land Use Change*

Event analysis of parcel- and etxe-level land use change between 1830 and 1950 showed a pattern of increasing farm size, decreasing land use intensity, in tandem with household abandonment. About 50 percent of parcels that started in etxe that were abandoned ended up transferring to etxe that increased their landholdings. Yet, etxe that expanded their farms did not display preferences for particular parcel qualities, rather, they opportunistically absorbed (or invested in) adjacent parcels following the abandonment of neighboring farms, maintaining their own previous capacities for crop production (Figure 5). However, parcel quality (based on its 1830 land use tax value) did statistically buffer against etxe expansion. In other words, some etxe with a relatively small amount of high-quality, arable land remained tied to more traditional subsistence strategies and did not seize opportunities to expand or specialize in market-oriented production. Etxe with lower quality farmland either invested and expanded their estate, or abandoned farming altogether.

**Figure 5.** Etxe household estate expansion 1830–1958. Example of investor household, blue landholdings 1830, by 1933 expanding to adjacent yellow landholdings.

Results also indicate that the specific nature and location of land use change was directed by etxe-specific demographic cycles and was contingent on the pre-1830 composition of the etxe estates and the estates they subsumed. Considering that plowing, planting, and harvesting crops is the most labor-intensive component of the farming system, the areal proportion of an etxe's crop fields to hay meadows, pastures, and woodlots was constrained by its labor capacities. These, in turn, were determined by household demographic cycles in the ratio of consumers (young children, elderly, and infirm) to producers (workers). In a system where etxe farm sizes were fine-tuned to demographic cycles from at least the 15th century, the absorption of a neighbor's parcels entailed the transition of parcels to less labor-intensive uses, i.e., crop fields to hay meadows, hay meadows to pastures, and pastures to woodlots.

This transition of etxe land use was not as significant for the communal high-elevation pastures. Many of the currently extant pastoral syndicates were indicated as active cayolar on the 1830 cadaster. Others were abandoned and consolidated or re-located as roads were constructed during the midto late-20th century. Post 1830 construction efforts (mostly carried out after the Second World War), elaborated on existing infrastructure and provided cheese makers with modern sanitation equipment. Our cadastral data for 1830 show a total of 329 cayolar shares (where 1 share is equal to between 45 and 60 ewes) were held by etxe households for communal pastures in Larrau. Thus, we estimate that in 1830, the minimum stocking rate in the communal pastures of Larrau ranged between 14,805 and 19,740, depending on share/ewe equivalency. Agricultural subsidy records for the commune of Larrau show that the average etxe milch ewe herd size increased from 49 in 1975 to 160 in 2010. Agricultural census data from 1984, 1993, 2000, and 2008 for Larrau's pastures show an annual average of 22,422 sheep (including ewes, rams, and lambs). Thus, in spite of increasing etxe herd size, we sugges<sup>t</sup> the total number of sheep has stayed relatively stable over the last 200 years.

Our examination of ecological components of the system also lends support to the hypothesis that grazing pressure has remained stable over the past few hundred years. Paired forest-pasture soil pits revealed that pasture A horizons exhibit three times the thickness in comparison to forested soils. They have higher concentrations of organic matter and significantly lower bulk densities than forests soils. Indeed, when compared with the soils of forested slopes of similar degree and aspect, soils in pastures appear more resilient to ecological disturbances. Our studies on pastoral fire use also revealed a pattern of persistence without negative impacts to forests. For example, evidence of fire scarring in tree trunks at treeline was relatively scant. Dendrochronological dating of trees cored along two pasture edges confirmed that those woodland-pasture ecotones have been relatively stable over at least the last ca. 200 years, in spite (or because) of the regular use of pastoral fire (Figure 6). Lastly, Bayesian modeling of the interactions of fire use suitability and land use change from 1830 to 2003 suggested that although areas that are the least suitable for fire managemen<sup>t</sup> experienced the highest afforestation on privately held lands, pasture commons used and managed by cayolars appear to have buffered against the afforestation.

**Figure 6.** Map and photos of pasture treeline study sites showing dendrochonological sample trees and pair forest-pasture soil pits [60]). (**A**) Bizkarze, a south-facing slope. Photo of pastoral burns in April, 2011, near location of sampled treeline. (**B**) Oronitz, a north-facing slope. Photo of recently burned heath, May, 2012, along sampled treeline.

### **4. Discussion and Conclusions**

### *4.1. Cause and E*ff*ect in Forest-Pasture Conversion*

Archaeological results for the Ibarrandoua colluvial site provide a good illustration of the di fficulty of empirically establishing cause-e ffect generalizations, even from the place-based pairing of archaeological and paleoenvironmental archives. Ibarrandoua is close to four major prehistoric mortuary sites, of which two (Amelstoy and Milgate) are ca. 750 m away. Amelstoy is a mortuary cave on a steep slope to the south of the colluvial catchment [96] that yielded human bone dated to ca. 3632-3514 cal BP. CHAR for this period registers around 0.014 mm2cm−2yr−1—well below the Holocene average, ye<sup>t</sup> suggestive of some anthropogenic burning. Milgate is a large, multi-component cromlech-tumulus site located on an exposed east-west ridge overlooking the Ibarrandoua colluvial site from which two features (Milgate 4 and 5) were excavated and dated by a previous investigator [97,98]. Milgate 5 (ca. 2872-2764 cal BP, late Bronze or early Iron Age) coincides with an uptick in CHAR ca. 2877 cal BP, while Milgate 4 (ca. 2292-2001 cal BP, late Iron Age) overlaps with the steep CHAR peak ca. 2283 cal BP. Another site, Behastoy, is on the flanks of Pic d'Orhy opposite the Ibarrandoua colluvial site distant ca. 4.6 km, and it overlaps with the peak in CHAR as well. We interpret this peak as marking the forest-pasture transition point at Ibarrandoua that was followed by a greater than 1000-year decline in burning. CHAR values are sustained from 950 BP forward to the present and coincide with the establishment and use of three dated archaeological features (most likely representing seasonal pastoral shelters) directly down slope from the colluvial sample site (99.7% CI range: 905 to 0 BP [post 1950 common era]).

These results speak to the onset of anthropogenic burning without directly addressing the origin of the so-called 'Neolithic package' giving rise to the pastoral economy in the mountains. From the point of view of archaeological evidence, the most intensive and clear use of the area follows well after forest-pasture conversion of the catchment. Nevertheless, the evidence for land clearance and erosion processes in Soule during the Long Anthropocene indicate heterogeneous, non-synchronous outcomes at the landscape level [70]. We are beginning to resolve some aspects of the source area for the herders and livestock using the upland pastures for periods prior to 1000 CE from our examination of land use change in Larrau and the Soule Valley.

### *4.2. Land Use Change*

Our event analysis results are consistent with the assertion that land use change between 1830 and 1950 in the Soule Valley was driven by the opening of dairy markets through improved transport and industrial creameries [55], as well as by outmigration that constrained locally available labor. For example, in 1902, Roquefort established a cheese facility in Tardets, the market town at the center of the Soule Valley, thus creating an annual market for ewe's milk and coinciding with the demise of long-distance transhumance beyond the confines of the Soule Valley, as historically described [17]. Many etxe began specializing in the production of ewe's milk, while their strategic response entailed abandoning traditional mixed-intensity agropastoralism, increasing the size of sheep herds, expanding hay meadows (thus, augmenting the capacity to stall-feed ewes over the winter), and collectively improving infrastructure surrounding seasonal transhumance between etxe and communal mountain pastures [55].

Our analysis revealed that the importance of the hay meadow contribution to etxe landholdings increased relative to the contribution from crop fields, but this shift in land use was not accomplished through a simple replacement of an etxe's crop fields with hay meadows. Instead, event analysis revealed that market-oriented etxe increased their hay production by increasing in size (land area) as other etxe were abandoned [26]. Thus, at the local level, the pace and character of land use change was significantly constrained by the social and spatial relationships between etxe [26]. In concert with the results of our soil and dendrochronological investigations, we sugges<sup>t</sup> that institutional persistence in Larrau has bu ffered against degradation and constrained land use at a sustainable level.

Adaptation to a novel socioeconomic opportunity did not result in abandonment and collapse of the system, but a shift from a mixed-intensity agropastoralism to a more specialized pastoral land use. Such flexibility may be inherent to the system of agropastoral households in the Pyrenees, enabling the system to persist through centuries of shifting socioeconomic and environmental change. While not all etxe were equally able or willing to meet the diverse opportunities they encountered, the response diversity of their decisions facilitated the persistence of the remaining etxe and the institution itself. In Larrau, etxe that increased landholdings during the 19th century and maintained high fertility into the 20th century were more likely to persist beyond 1958 than those etxe that experienced an earlier transition to lower fertility and did not reorient production toward emerging markets. The local historical and spatial contingencies in Larrau mediated the influence of exogenous forces and guided the direction of change taken by individual etxe.

### *4.3. Boundary, Scale, and Flow in Agropastoral Response Diversity*

Agropastoralism has deep roots in prehistory and much of the focus to date in the western Pyrenees has been on the regional onset of the Neolithic rather than the socioecological process of anthropogenic landscape transformation. Transitions are not enforced; they vary in rate and character through time and across space, because the cycle of households and the rhythms of the landscape are dynamically linked. A comprehensive consideration of the Long Anthropocene must contemplate how livestock managemen<sup>t</sup> strategies and pastoral response diversity help to explain the place-specific trajectories of landscape transformation. Mobility and exchange are critical to pastoralism, ye<sup>t</sup> also do not typically leave behind material signatures, since they are highly variable over time and across space and occur between rather than at places. A pastoral landscape includes various types of sites and features, e.g., cabins, pathways, and corrals, along with natural features such as caves, springs, and overhangs. It is these places and the relationships between them that structure pastoralism and serve as the arena in which repeated circulation and activities produce meaningful material patterns [99,100]. The multi-sited nature of pastoralism thus defines the spatiotemporal distribution of movement and settlement at varying social scales [101].

Pastoralists and agriculturalists, rather than forming divergent groups, may constitute sub-communities within the same identity group in the Soule Valley. Members of each group effectively retain the flexibility to shift between productive sectors over time, a trait that appears to be associated with pastoralism generally [102,103]. This suggests that assumptions about the pastoral lifestyle must be tempered by understanding response diversity among pastoralists who sometimes behave like agriculturalists. There is clear evidence for coexistence of divergent groups in Soule by AD 750, as Basque members of the valley republic coexisted with religious communities from Leyre and Sauvelade, and noble houses associated with various princes and monarchs. Pastoralism may not be, as often portrayed, an adaptation to a marginal environment, but a flexible adaptation to a shifting political-economic landscape resulting from the rise and fall of states and empires [104].

The second half of the 20th century witnessed a rapid disintegration of smallholder farming systems across European mountain landscapes [21] closely associated with rural population decline, agricultural industrialization, participation in non-local labor markets, and reforestation of abandoned lands [105–107]. In France it is referred to as the post-World War II rural crisis [108,109], which implies a change in lifestyle with implications for the future. Reforestation, for example, encroaches on continuing agropastoral land uses [110,111], reduces biodiversity, and leads to other conditions that threaten the future availability of ecosystem services [112,113]. There are serious e fforts in Europe to preserve the pastoral lifestyle [114], even though there is still only a rudimentary understanding of the interplay between households as the fundamental unit of production and the forces responsible for disintegration of smallholder systems.

The boundaries, scale, and flow of the response diversity of human agents in pastoralism relate to di fferences in herd composition and labor availability. While the livestock portfolio of an etxe may change in response to environmental stochasticity, they also express preferences derived from experience, knowledge, and contingencies. As a consequence, individual herders representing the interests of their etxe household differentially evaluate the risks and opportunities they confront. Some etxe may follow an aggressive herding strategy, while others may follow a more cautious strategy, minimizing their exposure to risk. Response diversity is thus a multi-level undertaking within and among individuals, households, and villages [79], and the effects of response diversity at one level might act synergistically with or counter those at another level. Places and the relationships between them define the spatiotemporal context of pastoralism, while response diversity translates into the manifest consequences for a landscape [115,116].

### *4.4. Finding Sustainability in the Long Anthropocene*

It has been suggested, with some finality, that the expansion of mountain grasslands and the creation of new upland pastures occurred through intentional landscape conversion and degradation (e.g., slash-and-burn practices [9,117]). While a plausible explanation, this rests on the search for a "golden spike" that could mark an abrupt transition in land use that aligns with conventional archaeological periodization schemes [118]. It assumes, a priori, that land use transitions represent stepwise intensification of human penetration and conversion of pristine landscapes that, in combination, represent an unsustainable trajectory of degradation. The contemporary conversion of tropical forests into degraded rangelands is a clear example of unsustainable anthropogenic landscapes that can influence how the past is interpreted from an Anthropocene position that rests on total human impacts on the whole earth system, while eschewing social, temporal, and spatial aspects of human-environmental interaction.

While one could argue that specialization and intensification of pastoral land use explains the motives for intentional forest to pasture conversion in the Pyrenees, our radiocarbon dating of charcoal from seasonal livestock cabins in Larrau suggests that intensive agropastoral land use followed only after forest to pasture conversion rather than vice versa [65]. Florescu et al. [119] identified a similar pattern in the Carpathians. In their study, charcoal and pollen archives derived from lake sediment showed an increase in subalpine and treeline fire activity between 8000 and 5000 BP that declined with the increase in the archaeological evidence for settlements. Indeed, as settlement and land use intensified after 2000 BP in lower elevation coniferous forests, fire activity remained low in the subalpine because forests were already transitioned to grasslands.

Given our analyses of the interplay between demography and agropastoral institutions in the Pyrenees [26,66], it seems difficult to imagine that mixed agropastoralists would intentionally invest the time and effort to convert forests to pasture in areas that are difficult to defend and ultimately served only as one component of agropastoral livelihood activities. In any case, agropastoral societies known to have inhabited the western Pyrenees could not have mounted the surplus labor required to intentionally convert forests to pastures in the rugged terrain of the region. Even under market pressures and technological enhancements of 20th century specialization and intensification in the use of mountain pastures, household-level labor and scheduling constraints continued to depend on collective efforts to support transhumant grazing. Furthermore, aside from the pooled labor efforts of the cayolar (and, post-1950s, the modern grazing syndicate), labor-saving managemen<sup>t</sup> activities such as the application of pastoral fires are necessary just to retain existing pastures even in the face of continued grazing pressures [65].

As discrete events, pastoral fires are antithetical to converting forests to pastures. They are low-severity fires spatially confined by topography and previous burning activity that are set in winter or spring when soil moisture is high [65]. However, when set repeatedly, but with varying frequency, over decades and centuries, they may, unintentionally, tip the balance from relatively closed canopy forests with grass-dominated gaps to open canopy forests with grass understory, and finally, to open grasslands [83]. Over centuries, such processes could have transformed an entire landscape, perhaps without significant degradation of ecosystem services. In fact, the persistence of agropastoral practices over the long-term would sugges<sup>t</sup> a sustainable co-evolution of land use and landscape.

The Pyrenees show little evidence of significant degradation and are better described as productive and predictable rangelands, and the hypothesized socioecological arrangements responsible for the conversion of upland landscapes are still under debate [120]. Our multi-proxy evidence does not empirically support the notion of agropastoral "impacts" and degradation. In the French Alps, Doyen et al. [121] explained the intermittent "intensification" of tree clearing from about 6000 BP onward as a function of creating and then abandoning arable and pastoral landscapes, perhaps as a way to avoid degradation. This and other evidence sugges<sup>t</sup> that the construction of the agropastoral niche across Europe was a long-term, non-linear process of slow, cumulative change, a persistent ecological press, largely devoid of ultimate human intentionality [122,123]. Elsewhere in Europe, the temporal resolution of burning, soil loss, forest clearing, and agropastoral infrastructure (e.g., [124–126]) have made it di fficult to distinguish between cause and e ffect. Yet, these and other studies do show that the evolution of agropastoral landscapes were regionally and locally asynchronous, and while ultimately giving rise to the Anthropocene, the phase shift cannot be independently explained or understood outside the details of the Long Anthropocene.

Deterministic narratives about human landscape transformation can be satisfyingly simple, while failing to explain the process itself. Human-environmental interactions and landscape history are not merely a function of population size, reducible to the insights a fforded by a single archive or the opinion of a single agent. By examining the boundaries, scale, and flow of the response diversity of human agents to the contingencies they confront, it becomes possible to supersede the limitations of quasi-immobile history, as well as answer questions about desirable future end-states about rural lifestyles [127–129]. For places like the western Pyrenees, research that focuses on the how and why of the complex co-evolution of anthropogenic landscapes could be the key to understanding the nature of sustainability in the Long Anthropocene.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2071-1050/12/9/3882/s1, Supplementary Materials S1: Geographic and temporal sketch of the Pyrenees. Figure S1.1: Relief, average temperature and average precipitation across the Pyrenees Mountains and surrounding area. Figure S1.2: Distribution of key paleoarchives for the western Pyrenees-Cantabrian mountain belt and dated archaeological sites by major periods from the LGM through the Middle Holocene.

**Author Contributions:** This was a collaborative e ffort, we contributed equally in conceptualizing the research, seeking funding to support the research, and directing di fferent aspects of the research. We most often worked together in the field and assisted each other with lab work, while simultaneously discussing our findings and the direction in which we wanted to take the research. T.L.G. carried out the original field assessment of the Soule Valley, and conceptualized the initial research project. M.R.C. carried out extended on-site research on fire and settlement, and in conjunction with T.L.G. carried out the pedestrian archaeological survey of the commune of Larrau. D.S.L. identified colluvial and paired forest-pasture locales, and collected all sediment cores, then analyzed or supervised others, including T.L.G. and M.R.C., in analyzing the sediment samples. T.L.G. and M.R.C. collected documentary records from various local and regional repositories across France and Spain. M.R.C. performed the fire probability analysis, while T.L.G. performed the historical analysis of Basque land use; maps and figures were done by T.L.G. and M.R.C., T.L.G. drafted the initial text with substantial revisions by M.R.C. and D.S.L. All authors contributed equally to prepare the final draft. All authors have read and agreed to the published version of the manuscript.

**Funding:** Partial support for the research has been provided by: the National Geographic Society (9573-14); a STAR Fellowship Assistance Agreement (FP917243); the Coweeta Long Term Ecological Research program funded by the National Science Foundation (DEB-0823293); a Partner University Fund award to the University of Georgia and the Université de Pau; the Chaire d'Attractivité IdEx Program at the Université de Toulouse (UMR 5608); and, a 2017 UGA Center for Integrative Conservation Research (CICR) Faculty Research Grant.

**Acknowledgments:** We gratefully acknowledge o fficials and residents of the commune of Larrau, France. Logistical support was provided by Pascal Palu and members of the Laboratoire ITEM, Université de Pau et du Pays de l'Adour.

**Conflicts of Interest:** The authors declare no conflict of interest.
