INTRODUCTION
The Amazonian Brazil nut tree (Bertholletia excelsa H.B.K., Lecythidaceae) is gap-dependent (Myers et al. Reference MYERS, NEWTON and MELGAREJO2000, Ortiz Reference ORTIZ, Shanley, Pierce, Laird and Guillen2002), favouring open forest areas for early growth and survival (Cotta et al. Reference COTTA, KAINER, WADT and STAUDHAMMER2008, Kainer et al. Reference KAINER, DURYEA, DE MACEDO and WILLIAMS1998, Mori & Prance Reference MORI and PRANCE1990). Brazil nut's persistence in post-disturbance, successional stages has led some researchers to label it as a long-lived pioneer (Swaine & Hall Reference SWAINE and HALL1987, Zuidema Reference ZUIDEMA2003). Brazil nut populations are often described in terms of groves, due to a distinct clumped distribution pattern at the landscape scale (i.e. aggregates that span over thousands of hectares; Mori & Prance Reference MORI and PRANCE1990, Peres & Baider Reference PERES and BAIDER1997, Peres et al. Reference PERES, BAIDER, ZUIDEMA, WADT, KAINER, GOMES-SILVA, SALOMÃO, SIMÕES, FRANCIOSI, VALVERDE, GRIBEL, SHEPARD, KANASHIRO, COVENTRY, YU, WATKINSON and FRECKLETON2003, Shepard & Ramirez Reference SHEPARD and RAMIREZ2011, Thomas et al. Reference THOMAS, ALCÁZAR CAICEDO, LOO and KINDT2014). Groves can also be defined at the local scale, complicating the determination of niche and dispersal drivers (John et al. Reference JOHN, DALLING, HARMS, YAVITT, STALLARD, MIRABELLO, HUBBELL, VALENCIA, NAVARRETE, VALLEJO and FOSTER2007). The aggregated pattern of Brazil nut groves has been linked to a mix of biophysical and anthropogenic causes (Scoles & Gribel Reference SCOLES and GRIBEL2012, Thomas et al. Reference THOMAS, ALCÁZAR CAICEDO, MCMICHAEL, CORVERA and LOO2015), including (1) short-range seed dispersal by scatterhoarding caviomorph rodents (Haugaasen & Haugaasen Reference HAUGAASEN and HAUGAASEN2010, Haugaasen et al. Reference HAUGAASEN, HAUGAASEN, PERES, GRIBEL and WEGGE2012, Peres & Baider Reference PERES and BAIDER1997); (2) early establishment in natural canopy gaps (Mori & Prance Reference MORI and PRANCE1990); (3) historical abandonment of swidden agricultural fields (Balée Reference BALÉE1989, Scoles & Gribel Reference SCOLES and GRIBEL2011); (4) accidental scattering of seeds along harvest trails (Ribeiro et al. Reference RIBEIRO, JEROZOLIMSKI, DE ROBERT, SALLES, KAYAPÓ, PIMENTEL and MAGNUSSON2014); and (5) formation and maintenance of groves by indigenous groups (Ribeiro et al. Reference RIBEIRO, JEROZOLIMSKI, DE ROBERT, SALLES, KAYAPÓ, PIMENTEL and MAGNUSSON2014).
Anthropogenic disturbances are also thought to be responsible for high densities of juvenile Brazil nut trees in the Eastern Amazon (Scoles & Gribel Reference SCOLES and GRIBEL2012, Reference SCOLES and GRIBEL2015). In Western Amazonia, recent studies have demonstrated enhanced early regeneration of Brazil nut in logging gaps and skid trails (Moll-Rocek et al. Reference MOLL-ROCEK, GILBERT and BROADBENT2014, Soriano et al. Reference SORIANO, KAINER, STAUDHAMMER and SORIANO2012). It is unclear, however, if this positive association with logging gaps continues throughout later life stages. Even though a Brazil nut tree can survive as an intermediate-sized juvenile once the forest canopy closes, development is often hindered due to low light levels (Scoles & Gribel Reference SCOLES and GRIBEL2012). Suppressed Brazil nut trees grow at rates a tenth of those in dominant and co-dominant crown positions (Staudhammer et al. Reference STAUDHAMMER, WADT and KAINER2013), increasing the probability of mortality (Swaine et al. Reference SWAINE, HALL and ALEXANDER1987, Terborgh et al. Reference TERBORGH, FLORES N, MUELLER and DAVENPORT1997). A pronounced gap mosaic, characteristic of selectively logged forests, and where the canopy remains open for several years (Asner et al. Reference ASNER, KELLER and SILVA2004) may allow both increased recruitment and growth rates of Brazil nut trees. For example, growth responses to nearby logging gaps were overwhelmingly positive for several tropical tree species in a selectively logged forest in French Guiana (Herault et al. Reference HERAULT, OAULLET, BLANC, WAGNER and BARALOTO2010a).
Despite the prevalence of Brazil nut-timber management systems (especially in the south-western Amazon region), most scientific work about Brazil nut ecology and population dynamics derives from unlogged forests (Kainer et al. Reference KAINER, WADT and STAUDHAMMER2014, Peres et al. Reference PERES, BAIDER, ZUIDEMA, WADT, KAINER, GOMES-SILVA, SALOMÃO, SIMÕES, FRANCIOSI, VALVERDE, GRIBEL, SHEPARD, KANASHIRO, COVENTRY, YU, WATKINSON and FRECKLETON2003, Staudhammer et al. Reference STAUDHAMMER, WADT and KAINER2013, Wadt et al. Reference WADT, KAINER and GOMES-SILVA2005, Zuidema & Boot Reference ZUIDEMA and BOOT2002). To our knowledge, no published study has yet examined the density and spatial distribution pattern of Brazil nut adult (≥ 40 cm dbh) and juvenile (10 ≤ dbh <40 cm) tree populations in selectively logged forests. In this study we assessed the spatial structure of Brazil nut populations in three selectively logged forests in Peruvian Amazonia. We hypothesized that densities of juvenile Brazil nut stems would be higher than juvenile density in unlogged forests relative to adult stems, due to the long history of timber exploitation along the Interoceánica Sur Highway (IOS), which connects Brazil with Peruvian ports. We also expected to find a positive spatial association between juvenile stems and cut stumps.
METHODS
Study area
Given the potential variability of Brazil nut tree distribution across the landscape, the study was conducted in three Brazil nut concessions (henceforth referred to as Alegria I (290 ha), Alegria II (547 ha), Alegria III (576 ha)) located 5–20 km apart in south-eastern Peru, Tambopata Province, Department of Madre de Dios (11°30′30′′–12°10′0′′S, 69°56′0′′–69°21′0′′W). Surveys were conducted from September 2013 to July 2014 (Figures 1, 2). Sites were selected following a series of meetings (conducted in October 2012) with concessionaire associations in the settlement of Alegría. The selection of the three concessions for inclusion in the study was based primarily on concessionaire interest and site accessibility. The Brazil nut trees located in these concessions belong to a single Brazil nut metapopulation located along the IOS between Puerto Maldonado and Alerta (IIAP-CTAR 2001). In all three concessions, Brazil nut harvesting was taking place on an informal basis prior to the arrival of the current concessionaires (1978, 1979 and 1999, respectively). At present, all three concessionaires claim to harvest up to 90% of the annual Brazil nut annual production, leaving approximately 10% for regeneration and forage for important game taxa such as agoutis (Dasyprocta spp.), the spotted paca (Cuniculus paca) and various primate species.
Location of study sites (Alegria I, Alegria II, Alegria III) in Madre de Dios, Peru. The pale pink shade refers to all Brazil nut concessions not associated with the study. The red shade refers to all participating Brazil nut concessions. The dark line running north-south refers to the Interoceánica Sur Highway.
Maps of the study sites, Alegria I (a, 290.6 ha), Alegria II (b, 547.3 ha), and Alegria III (c, 576.3 ha), in Madre de Dios, Peru. The white stars refer to juvenile Brazil nut trees (10 ≤ dbh < 40 cm). The solid black circles refer to the adult Brazil nut trees (dbh ≥ 40 cm). The black crosses refer to cut stumps (≥ 10 y). The white rectangle refers to an abandoned agricultural field (Alegria II).
Mean annual rainfall in the region ranges from 2500 to 3500 mm, with a pronounced rainy season from December to March (Chávez et al. Reference CHÁVEZ, GUARIGUATA, CRONKLETON, MENTON, CAPELLA, ARAUJO and QUAEDVLIEG2012, Tobler et al. Reference TOBLER, CARRILLO-PERCASTEGUI and POWELL2009). Mean annual temperature is 24°C, but it fluctuates within an extreme range for the neotropics (10–38°C; Tobler et al. Reference TOBLER, CARRILLO-PERCASTEGUI and POWELL2009). The landscape is defined by acidic, well-drained soils of moderate to low fertility and gently undulating to flat topography. Madre de Dios is dominated by a mix of humid tropical forests in both elevated terra firme and low-lying floodplain sites (Asner et al. Reference ASNER, KELLNER, KENNEDY-BOWDOIN, KNAPP, ANDERSON and MARTIN2013). The department claims ~90% forest coverage, which is high in comparison with other regions of the Amazon Basin (Asner et al. Reference ASNER, KNAPP, BALAJII and PÁEZ-ACOSTA2009), but forest cover is expected to decline in the future, following the paving of the IOS, which was completed in 2010 (Perz et al. Reference PERZ, QIU, XIA, SOUTHWORTH, MARSIK, ROCHA, PASSOS, ROJAS, ALARCÓN, BARNES and BARALOTO2013). Prior to the paving, the highway had been in existence for several decades, but was often impassable during the rainy season.
Low-intensity timber extraction was prevalent in Brazil nut-rich forests several decades before the concession system was established and logging was formalized by a government decree in 2004 (Cossío-Solano et al. Reference COSSÍO-SOLANO, GUARIGUATA, MENTON, CAPELLA, RÍOS and PEÑA2011). Over the last decade, officially sanctioned timber extraction intensities have ranged from 1–3 trees ha−1 (~5–10 m3 ha−1; Cossío-Solano et al. Reference COSSÍO-SOLANO, GUARIGUATA, MENTON, CAPELLA, RÍOS and PEÑA2011), but they can exceed this amount, especially when informal practices are the norm (Rockwell et al. Reference ROCKWELL, GUARIGUATA, MENTON, QUISPE, QUAEDVLIEG, WARREN-THOMAS, SILVA, ROJAS, ARRUNATEGUI, VEGA, VERA, HANCCO, TITO, PANDURO and SALAS2015). All three concessionaires associated with our study confirmed the long-term presence of logging in their concessions through first-hand knowledge, the presence of stumps predating their arrival, or communication with previous concessionaires. Logging activity dates in all three cases to at least the late 1960s–early 1970s, when hardwood species such as Swietenia macrophylla and Cedrela odorata (both Meliaceae) were the preferred commercial taxa. Species currently favoured for timber harvesting (given the historic high grading of hardwood species) include Cedrelinga catenaeformis (Fabaceae), Matisia spp. (Malvaceae), Ceiba pentandra (Malvaceae) and Couratari spp. (Lecythidaceae), the latter three all considered softwood species. Although timber removal generally occurs during the dry season, at least two of the concessions in this study were recently subjected to logging at the height of the rainy season. Two of the three concessionaires, however, stated that their logging activities tend to be infrequent.
Field measurements
An exhaustive (100%) tree inventory method was employed in the three concessions from September 2013 to July 2014. Surveys to locate juvenile Brazil nut stems were conducted across linear strips 40 m in width, running north-south within the concession borders. Brazil nut tree juveniles were located and geo-referenced with a Garmin® GPSMap 62 handheld unit (precision 1–10 m). Dbh and physical characteristics such as visible damage (e.g. scraped bark, missing branches) and liana infestation, were recorded for each individual. Concessionaires accompanied field teams to verify locations of planted Brazil nut trees so as not to confuse them with natural regeneration; these were excluded from the analysis. Most planted stems were smaller than our size class of interest (i.e. < 10 cm dbh). Locations of all reproductive Brazil nut trees (≥ 40 cm dbh) in each concession were also geo-referenced with a Garmin® GPSMap 62 handheld unit (Figure 2).
Prior to the selection of the three concessions, we had no information about the intensity nor the spatial distribution of selective logging across these sites. Given the opportunistic nature of logging activities in Madre de Dios, timber has historically been harvested without regard to a regular cutting cycle. Historical cut stumps were inventoried in an exhaustive manner simultaneously with the juvenile Brazil nut tree survey. The ages of the cut stumps were estimated by the field team and confirmed with the concessionaires. To characterize the logging mosaic in each concession and to assess the relationship between cut-stump location on Brazil nut juvenile tree distribution and density, distance to nearest logging gap or cut stump was measured for each juvenile tree. Cut stumps younger than 10 y were excluded from the analysis, as they are unlikely to have had an influence on the regeneration/recruitment of the juvenile stems due to a decrease in passage time (i.e. number of years spent in one size class) as the stems reach a larger size class (Brienen & Zuidema Reference BRIENEN and ZUIDEMA2006). Brazil nut tree diameter growth is highly variable, so estimating the ages of Brazil nut stems on site was untenable. For example, Schöngart et al. (Reference SCHÖNGART, GRIBEL, FERREIRA DA FONSECA-JUNIOR and HAUGAASEN2015) found that Brazil nut trees (10 ≤ dbh < 40 cm) from two sites in Central Amazonia ranged in age from 18–34 y. The oldest cut stumps encountered during the initial surveys in our study were ~30–40 y old. As such, we know that there is a strong likelihood that the older cut stumps relate to logging events contemporaneous to the germination date of some of the juvenile stems included in this study. In light of the variable stem growth rates of Brazil nut, younger cut stumps (10 ≤ and < 30 y) were also included in the analysis. Resulting logging gaps from this period could have feasibly favoured both germination and recruitment rates. All distance measurements were calculated using geo-referenced locations in ArcMap 10.1 (ESRI, Redlands, CA).
Analysis
Ecological point patterns are often characterized by a combination of spatial relationships, thus limiting the use of certain spatial analyses that employ only mean of distances (e.g. nearest-neighbour methods; Haase Reference HAASE1995). In contrast, the spatial point pattern analysis developed by Ripley (Reference RIPLEY1977) uses all point-to-point distances for two-dimensional spatial data, demonstrating how spatial clustering, or dispersion, changes within the area specified. Specifically, Ripley's K function calculates the expected number of neighbours in a circle of a given radius d centred on an arbitrary point in the pattern. The circle uses the given radius as a starting point, eventually encompassing the entire study site (i.e. the whole concession).
In order to describe the spatial patterns of both juvenile and adult tree populations, we used a univariate spatial pattern analysis based on a transformation of the Ripley's K (d) index (a linearized form of Ripley's K function; Ripley Reference RIPLEY1977) using the ads package in R 3.1.1 (http://CRAN.R-project.org/package=ads).
$$\begin{equation*}
L\left( d \right) = \surd \left( {\frac{{K\left( d \right)}}{\pi }} \right) - d
\end{equation*}$$
Under the null hypothesis of complete spatial randomness, L(d) is 0, or a homogeneous Poisson point process (Marcon et al. Reference MARCON, TRAISSAC and LANG2013), against which our observed set of points was tested. We estimated a confidence interval of 99% using the Monte Carlo method. The null hypothesis (i.e. random distribution) was simulated by computing 1000 Poisson distributions (Goreaud & Pélissier Reference GOREAUD and PÉLISSIER1999). For each particular distance (d), the observed L(d) value that differed from the expected L(d) by the greatest amount is the confidence interval. If the observed L(d) is larger than the expected L(d) value at a particular distance, the pattern can be characterized as clustered; if it is smaller than the expected L(d), the pattern is regular. To account for the irregular shapes of one of the concessions (Alegria II), we employed an edge-effect correction that supported the complex triangle assembly associated with our data set (Goreaud & Pélissier Reference GOREAUD and PÉLISSIER1999, Ripley Reference RIPLEY1977, Traissac & Pascal Reference TRAISSAC and PASCAL2014). The use of triangles in the analysis allows the removal of blank zones that would otherwise be used by the Ripley's function to calculate the spatial structure of the species in question.
To understand the spatial relationship between juvenile and adult populations, and between the juveniles and cut stumps, we employed the linearized intertype L12 function (Goreaud & Pélissier Reference GOREAUD and PÉLISSIER2003, Traissac & Pascal Reference TRAISSAC and PASCAL2014):
$$\begin{equation*}
{L_{12}}\left( d \right) = \sqrt[{}]{{\frac{{{K_{12}}\left( d \right)}}{\pi }}} - d
\end{equation*}$$
where L12 (d) is 0 under the null hypothesis of independence between juveniles and adults and juveniles and cut stumps. Under the toroidal shift null model, a positive value of L12 (d) indicates attraction between the two populations and a negative value indicates repulsion at a given distance (d; Traissac & Pascal Reference TRAISSAC and PASCAL2014). In this particular case, the positions of the points of one class remain unchanged while the points of the other class are shifted by the same random vector (Ledo et al. Reference LEDO, CONDÉS and MONTES2011). Accordingly, a confidence interval of 99% was developed from 1000 random toroidal shifts for each analysis (Goreaud & Pélissier Reference GOREAUD and PÉLISSIER2003, Traissac & Pascal Reference TRAISSAC and PASCAL2014). When analysing the spatial patterns between juveniles and reproductive adults, we reduced the sample size of the juveniles to include only those individuals between 10 ≤ dbh < 30 cm in order to avoid the assumption that smaller juveniles are the progeny of larger juvenile stems. Likewise, we limited the analysis of correlation between cut stumps and juvenile trees to Brazil nut stems between 10 ≤ dbh < 30 cm, given the possibility that larger juvenile individuals (30 ≤ dbh < 40 cm) pre-date our oldest cut stumps (~40 y).
RESULTS
Descriptive statistics
Total number of cut stumps ≥10 y old, per concession, were 11 (Alegria I), 42 (Alegria II) and 18 (Alegria III; Figure 2). Two hundred and seventy-six Brazil nut trees (≥ 40 cm dbh) were found in Alegria I, 345 in Alegria II and 336 in Alegria III (Table 1, Figure 2). The location of adult trees in each concession was used to calculate local density (trees ha−1; Table 1). The adult density in Alegria III was the lowest of all three sites (0.58 trees ha−1), while Alegria I contained the highest density (0.95 trees ha−1).
Descriptive statistics of population structure of Bertholletia excelsa in three Brazil nut concessions in Madre de Dios, Peru.
Thirty-six juvenile Brazil nut trees (10 ≤ dbh < 40 cm) were found in Alegria I, 101 in Alegria II and 59 in Alegria III (Table 1, Figure 2). The location of juvenile trees in each concession was used to calculate local density (Table 1). The juvenile density in Alegria III was the lowest of all three sites (0.10 trees ha−1), while Alegria II contained the highest density (0.19 trees ha−1), as well as the highest percentage of juveniles (Table 1).
Spatial patterns
Aggregated distribution patterns were found for the adult tree size class in all three concessions (Figure 3). The strongest peaks of aggregation were found in Alegria I (~700 m) and Alegria II (300 m). At all three sites, spatial aggregation patterns were observed amongst juvenile stems (Figure 4), but with significant peaks of clustering noted solely in Alegria II at 500 m. In Alegria III, juveniles were randomly distributed up to ~600 m, at which point the pattern became spatially aggregated, with one peak of aggregation noted at 800 m. Aggregation patterns peaked at 900 m in Alegria I.
Spatial pattern analyses of adult (dbh ≥ 40 cm) Bertholletia excelsa trees in three sites in Madre de Dios, Peru. The solid black line refers to the L(d) index (observed spatial pattern). The dotted violet line refers to the random (expected) spatial pattern. The dotted green line refers to the confidence interval. The sites include Alegria I (a), with a local density of 0.95 stems ha−1; Alegria II (b), with local density of 0.63 stems ha−1; and Alegria III (c), with local density of 0.58 stems ha−1.
Spatial pattern analyses of juvenile (10 ≤ dbh <40 cm) Bertholletia excelsa trees in three sites in Madre de Dios, Peru. The solid black line refers to the L(d) index (observed spatial pattern). The dotted violet line refers to the random (expected) spatial pattern. The dotted green line refers to the confidence interval. The sites include Alegria I (a), with local density of 0.12 stems ha−1; Alegria II (b), with local density of 0.19 stems ha−1; and Alegria III (c), with local density of 0.10 stems ha−1.
Overall, our results demonstrated a lack of juveniles near mature trees. Average distances between nearest reproductive neighbour and juvenile stems were higher in Alegria II than Alegria I (66.5 m) or Alegria III (68 m), with a median distance of 108.9 m (Table 1). We did not find significant attraction or dispersion between juvenile and adult Brazil nut trees at any given radial distance using the intertype point pattern analysis (Figure 5). No significant attraction or repulsion patterns were discernible between cut stumps ≥ 10 y old and juvenile trees (Figure 6).
Spatial relation between juvenile (10 ≤ dbh ≤ 30 cm) and adult (dbh ≥ 40 cm) Bertholletia excela trees in three sites in Madre de Dios, Peru. The solid black line refers to the L(d) index (observed spatial pattern). The dotted violet line refers to the random (expected) spatial pattern. The dotted green line refers to the confidence interval. The sites include Alegria I (a); Alegria II (b); and Alegria III (c).
Spatial relation between juvenile (10 ≤ dbh ≤ 40 cm) Bertholletia excela trees and location of cut stumps (≥ 10 y) in three sites in Madre de Dios, Peru. The solid black line refers to the L(d) index(observed spatial pattern). The dotted violet line refers to the random (expected) spatial pattern. The dotted green line refers to the confidence interval. The sites include Alegria I (a); Alegria II (b); and Alegria III (c).
DISCUSSION
Much of the literature on Brazil nut tree ecology has confirmed the existence of local and landscape scale groves (Mori & Prance Reference MORI and PRANCE1990, Peres & Baider Reference PERES and BAIDER1997, Salomão Reference SALOMÃO1991). Despite relatively low densities of reproductive adults, results from our sites appear to corroborate those findings – spatial aggregations at the local scale (< 6 km2) were found in all three concessions for this size class (Figure 3). We submit that the scattered gap mosaic conditions of South-Western Amazon forests likely favour grove formation, even in the absence of anthropogenic disturbance. For example, Asner et al. (Reference ASNER, KELLNER, KENNEDY-BOWDOIN, KNAPP, ANDERSON and MARTIN2013) determined that forests in Madre de Dios are subject to large gap-forming processes due to strong winds (or blowdowns (30–2000 ha); Nelson et al. Reference NELSON, KAPOS, ADAMS, OLIVEIRA and BRAUN1994) as well as smaller-scale canopy damage resulting from mass loading of bamboo (Guadua sarcocarpa and G. weberbaueri) culms (Griscom & Ashton Reference GRISCOM and ASHTON2006). Although a direct link has yet to be made between Brazil nut groves and these natural canopy disturbance events, such phenomena would surely introduce elevated light levels to the forest, the most limiting resource for Brazil nut tree growth (Staudhammer et al. Reference STAUDHAMMER, WADT and KAINER2013).
We predicted that juvenile density in these forests would be high, given our knowledge of the positive influence of other anthropogenic disturbances on Brazil nut densities in unlogged forests (Scoles & Gribel Reference SCOLES and GRIBEL2011). Juvenile densities in our three concessions, however, are low when compared with unlogged sites in the Amazon. Indeed, in the cases of Alegria I and Alegria III, juveniles comprise a small proportion of the entire Brazil nut tree population (12.5% and 14.9% respectively) when compared with other Amazonian studies (Scoles & Gribel Reference SCOLES and GRIBEL2011). In spite of enhanced light levels, it is possible that the risk of juvenile mortality is increased in logged sites (Jansen & Zuidema Reference JANSEN, ZUIDEMA, Fimbel, Grajal and Robinson2001); damage to pre-reproductive Brazil nut stems as a result of timber harvesting has been cited as a threat for Brazil nut tree populations (Guariguata et al. Reference GUARIGUATA, LICONA, MOSTACEDO and CRONKLETON2009). It can also be argued that since historical logging activities in Madre de Dios were typically characterized by low logging intensities (and indeed, are still today; Cossío-Solano et al. Reference COSSÍO-SOLANO, GUARIGUATA, MENTON, CAPELLA, RÍOS and PEÑA2011, Rockwell et al. Reference ROCKWELL, GUARIGUATA, MENTON, QUISPE, QUAEDVLIEG, WARREN-THOMAS, SILVA, ROJAS, ARRUNATEGUI, VEGA, VERA, HANCCO, TITO, PANDURO and SALAS2015), the size of the resulting gaps may not have been sufficient in size to enhance Brazil nut regeneration and recruitment over the long term (Fredericksen & Putz Reference FREDERICKSEN and PUTZ2003).
But could low juvenile density also indicate a regeneration failure linked to seed overharvesting and independent of logging activities? Other researchers have noted that long-term NTFP harvests can lead to population decline (Herrero-Jáuregui et al. Reference HERRERO-JÁUREGUI, SIST and CASADO2012, Ticktin Reference TICKTIN2004). In Madre de Dios, the increase in demand for Brazil nut in both the national and international markets and the recent IOS road paving has intensified harvesting pressure on contemporary Brazil nut gatherers. Thus, we suspect that current harvest intensities (~90% of total seed crop, according to concessionaires) are likely higher than in previous decades (Cardó et al. Reference CARDÓ, SEQUEIRA and TORRES PADILLA2002), leading us to conclude that the low juvenile stem density in our sites is not completely atypical and probably not the result of historical harvesting practices. Indeed, the predominance of large, centenarian individuals and scarcity of juveniles are common features of Brazil nut stands (Nepstad et al. Reference NEPSTAD, BROWN, LUZ, ALECHANDRE and VIANA1992, Peres & Baider Reference PERES and BAIDER1997, Scoles & Gribel Reference SCOLES and GRIBEL2012).
Clumping patterns were detected in all three sites for the juvenile size class (Figure 4), but juvenile groves were not related to location of cut stumps. Our knowledge of historical logging is based on location of old cut stumps and interviews with the concessionaires. It is possible that we did not find the locations of all old cut stumps; stump decomposition can take place within a relatively short period of time, especially in sites characterized by a prolonged rainy season (Delaney et al. Reference DELANEY, BROWN, LUGO, TORRES-LEZAMA and BELLO QUINTERO1998). Nonetheless, given historical logging practices, most of the cut stumps found in the concessions are remnants of stems characterized by high wood density and large diameter, characteristics determined by Herault et al. (Reference HERAULT, BEAUCHÊNE, MULLER, WAGNER, BARALOTO, BLANC and MARTIN2010b) to be strong predictors for wood decay. We know from other tropical studies, though, that tree distribution patterns are influenced by many different environmental factors, challenging the process of identifying specific drivers (Traissac & Pascal Reference TRAISSAC and PASCAL2014). For example, our study was not able to account for the role of natural gaps which generally provide favourable conditions for seedling establishment and growth of Brazil nut. If recruitment is enhanced by large-scale canopy openings (e.g. blowdowns), the clustering effect might be seen at larger (and variable) distances, similar to the patterns observed in all three concessions in this study for the juvenile size class.
The clumping patterns detected in the juvenile populations can also be explained by other factors: aggregation patterns of juvenile and adult trees can be influenced by frugivore activity (Bleher & Böhning-Gaese Reference BLEHER and BÖHNING-GAESE2001). In the case of Brazil nut, agoutis normally scatterhoard fruits at limited distances of 5–60 m (Haugassen et al. 2012, Peres & Baider Reference PERES and BAIDER1997). Under such circumstances, one would anticipate that cohorts of half-siblings dispersed by small mammals would be centred at short distances from mother trees (Nathan & Muller-Landau Reference NATHAN and MULLER-LANDAU2000). The lack of a spatial relationship with reproductive stems in this study has several possible explanations, which are not necessarily mutually exclusive: (1) there is a negative relationship between Brazil nut regeneration and proximity to mother trees, a phenomenon first noted by Janzen (Reference JANZEN1970) and Connell (Reference CONNELL, den Boer and Gradwell1971) for other neotropical species, and proposed as one potential explaining mechanism for Brazil nut distribution by Peres & Baider (Reference PERES and BAIDER1997); (2) Post-primary dispersers (e.g. spiny rats (Proechimys spp.)) might camouflage the spatial relationship between a parent tree and its progeny (Jansen et al. Reference JANSEN, HIRSCH, EMSENS, ZAMORA-GUTIERREZ, WIKELSKI and KAYS2012, Peres & Baider Reference PERES and BAIDER1997, Vander Wall et al. Reference VANDER WALL, FORGET, LAMBERT, HULME, Forget, Lambert, Hulme and Vander Wall2005); and (3) the mother trees of the juveniles in question are dead, thus explaining why no pattern between adults and juveniles was found.
Many authors have suggested a strong influence of indigenous activity (both deliberate and accidental) on grove maintenance (Ribeiro et al. Reference RIBEIRO, JEROZOLIMSKI, DE ROBERT, SALLES, KAYAPÓ, PIMENTEL and MAGNUSSON2014, Scoles & Gribel Reference SCOLES and GRIBEL2011, Shepard & Ramirez Reference SHEPARD and RAMIREZ2011, Thomas et al. Reference THOMAS, ALCÁZAR CAICEDO, MCMICHAEL, CORVERA and LOO2015). Even so, the size class of interest (10 ≤ dbh < 40 cm) in this study is likely too young for its clustering pattern to be considered a result of deliberate or accidental Amerindian intervention. Amerindian communities along the IOS began to dissipate around a century ago, coinciding with the arrival of rubber tappers in the late 19th and early 20th centuries (Alexiades & Peluso Reference ALEXIADES, PELUSO and Alexiades2009). And while it is true that the adult populations in our sites are also aggregated, Thomas et al. (Reference THOMAS, ALCÁZAR CAICEDO, MCMICHAEL, CORVERA and LOO2015) have suggested that the regeneration of Brazil nut in Western Amazonia (a region characterized by a sparser pre-Columbian human population than in the Eastern Amazon (McMichael et al. Reference MCMICHAEL, PIPERNO, BUSH, SILMAN, ZIMMERMAN, RACZKA and LOBATO2012)) seems to have been controlled predominantly by natural processes.
It is likely that other anthropogenic (e.g. historical campsites of Brazil nut harvesters, contemporary agricultural fields) and biophysical processes (e.g. blowdowns) have contributed to the local distribution patterns and densities of this size class in our study sites. The strong peaks of aggregation for juveniles and adult Brazil nuts in this study occurred at long distances (300–900 m), pointing to the potential of multiple tree canopy gaps as drivers of spatial patterns, either via natural or anthropogenic sources. In addition to blowdowns, the influence of more contemporary agricultural activities within the concessions could explain the spatial clustering of juveniles, especially in the case of Alegria II (see Figure 2 for depiction of abandoned field in this concession). Other studies have confirmed that planted Brazil nut seedling growth rates are much higher in open areas (e.g. shifting cultivation fields) than in small canopy openings found in undisturbed forest, where light, nutrient and water resources are limited (Cotta et al. Reference COTTA, KAINER, WADT and STAUDHAMMER2008, Kainer et al. 1998). The field in this particular concession is ~30 y old (gradually increasing in size over the last three decades), thus a strong candidate in terms of influencing regeneration and recruitment for the size class in question around the perimeter of the field.
Our data contribute to a more thorough understanding of Brazil nut tree population structure in selectively logged forests in south-western Amazonia, particularly for assessing the status of juvenile trees that represent the future resource base. Even though we found no direct influence of cut stump location on the spatial distribution of juvenile trees, we recognize that the interaction between timber harvests and Brazil nut population dynamics is not static, given the multiple anthropogenic drivers (e.g. logging, Brazil nut harvesting, hunting, small-scale agriculture) occurring in these concessions. Indeed, it is unlikely that these drivers will remain fixed in the near future, either as independent or combined factors (Ticktin et al. Reference TICKTIN, GANESAN, PARAMESHA and SETTY2012).
Considering the relative scarcity of Brazil nut juvenile trees in our sites, we also posit that natural regeneration cohorts would benefit from tending (i.e. clearing of competing species around young trees; D'Oliveira Reference D'OLIVEIRA2000, Peña-Claros et al. Reference PEÑA-CLAROS, FREDERICKSEN, ALARCÓN, BLATE, CHOQUE, LEAÑO, LICONA, MOSTACEDO, PARIONA, VILLEGAS and PUTZ2008) to enhance recruitment into the intermediate size classes. Concessionaires already recognize the importance of enrichment planting of Brazil nut, often taking advantage of logging gaps to plant seedlings provided by local NGOs (pers. obs.). This type of informal integration to maintain local Brazil nut tree populations (motivated by cultural as much as economic considerations) is happening on a frequent basis, providing an opportunity to link best management practices for both forest products. Particularly in the case of smallholder forestry in south-western Amazonia, joint inventories and Brazil nut-specific technical recommendations are still largely absent from timber management plans (Cronkleton et al. Reference CRONKLETON, GUARIGUATA and ALBORNOZ2012, Duchelle et al. Reference DUCHELLE, GUARIGUATA, LESS, ALBORNOZ and CHÁVEZ2012). As such, efforts to manage the Brazil nut resource base in the context of multiple forest use is vital to the development of local conservation and socio-economic goals, and further efforts should be undertaken to focus on the management of non-reproductive trees.
ACKNOWLEDGEMENTS
MRG and CAR conceived the study. MRG and MM designed the initial phase of the work which was later refined by CAR. CAR led the writing in collaboration with MRG. CAR carried out data analysis and data quality control. EWT was responsible for the initial mapping and distance calculations of Brazil nut trees and logging gaps. The rest of the co-authors collected field data. We appreciate the contributions of P.A. Usca, P.L.A. Gallegos, B.A. de Simplicio, G.C.R. Marques, F.R. Surco, the Asociación de Castañeros de Alegría, the Asociación de Productores de Castaña de Alerta, Ing. M.G. Roca, C.N.A. Mamani, M.A. Quispe, D.N.C. Apaza, S.A. Coaquira, L.A. Huaman, L.C. Leon, A.N.C. Rafael, D.D. Salcedo, R.L.M.C. Puma, J.A.G. Juriuchi, N.G. Quispe, F.L.M. Siani, S.M. Ferro, E.S. Ccalluchi, J. Loja, D.A. Huamantica, M. Aleman, R. Isla, M.D. Bernal and C. Murto. We also thank S. Traissac for his valuable insights and development of the R code for the spatial analyses. Our thanks also to C. Baraloto, C. Klimas, S. Traissac and two anonymous reviewers for providing helpful comments on earlier drafts. This work was funded by the United States Agency for International Development and the CGIAR Research Program on Forests, Trees and Agroforestry.
