Southern Forests: a Journal of Forest Science

ISSN: 2070-2620 (Print) 2070-2639 (Online) Journal homepage: http://www.tandfonline.com/loi/tsfs20

Inferring population trends of angustifolia () using a transition matrix model in an old-growth forest

Giovani F Paludo, Miguel B Lauterjung, Maurício S dos Reis & Adelar Mantovani

To cite this article: Giovani F Paludo, Miguel B Lauterjung, Maurício S dos Reis & Adelar Mantovani (2016): Inferring population trends of Araucaria angustifolia (Araucariaceae) using a transition matrix model in an old-growth forest, Southern Forests: a Journal of Forest Science, DOI: 10.2989/20702620.2015.1136506

To link to this article: http://dx.doi.org/10.2989/20702620.2015.1136506

Published online: 01 Mar 2016.

Submit your article to this journal

Article views: 1

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=tsfs20

Download by: [University of California, San Diego] Date: 04 March 2016, At: 19:15 Southern Forests 2016: 1–7 Copyright © NISC (Pty) Ltd Printed in South Africa — All rights reserved SOUTHERN FORESTS This is the final version of the article that is ISSN 2070-2620 EISSN 2070-2639 published ahead of the print and online issue http://dx.doi.org/10.2989/20702620.2015.1136506

Inferring population trends of Araucaria angustifolia (Araucariaceae) using a transition matrix model in an old-growth forest

Giovani F Paludo1*, Miguel B Lauterjung1, Maurício S dos Reis2 and Adelar Mantovani1

1 Grupo Uso e Conservação dos Recursos Florestais, Universidade do Estado de , Lages, Brazil 2 Núcleo de Pesquisas em Florestas Tropicais, Universidade Federal de Santa Catarina, Florianópolis, Brazil * Corresponding author, email: [email protected]

Matrix population models may generate important information to prevent undesirable outcomes for endangered species. This is the case for Araucaria angustifolia, a Critically Endangered , with little knowledge regarding its life history and trends in development over time. This study sought to investigate life-history trends of an A. angustifolia population in a subtropical forest in Santa Catarina, Brazil. Predictions based on the Lozenge regeneration model were established in order to determine if this model could predict changes in the species’ population dynamics: (1) the established individuals exhibit long persistence and (2) seedling and sapling abundance, as well as population descriptors, should exhibit behaviour that indicates one of the stages prescribed by the model. All A. angustifolia individuals were evaluated within a 5.1 ha plot at the study site over a six-year period. Lefkovitch’s transition model was used and population descriptors were calculated. Both predictions were fulfilled. The population had λ = 0.9977 (0.9864 < λ < 1.0020; CI 95%), indicating a declining stability. The basal area remained stable, whereas tree density tended to decrease, and seedlings and saplings did not promote an increase in λ. These results indicate that the population was in a phase called thinning, defined by the Lozenge model. The results led to three conclusions: recruitment seems insufficient; survival of reproductive individuals is responsible for the longevity of the population; and the predictions did not refute the Lozenge model. According to this model, the population is expected to regenerate in the future. However, the species exhibits declining stability, which aggravates the endangerment situation.

Keywords: Araucaria forest, Brazilian , mixed ombrophilous forest, natural regeneration, population dynamics

Introduction

Population matrix models (PMMs) have been used to saplings of long-lived pioneers are frequently shade- develop conservation strategies (Enneson and Litzguz intolerant and are not able to grow under a dense tree 2008; Portela et al. 2010; López-Mata 2013), sustain- canopy. In some long-lived species and canopy-dominant able exploitation of non-timber forest products (Schmidt et species, the appearance of seedlings and saplings will al. 2011; Venter and Witkowski 2013) and predictions of a depend upon the occurrence of an episodic landscape- population’s future (Crone et al. 2013). PMMs have shown level disturbance, such as the dieback of a large number both successes and failures in predicting tree popula- of mature trees (Boehmer et al. 2013) and landslides or tion behaviour (Bierzychudek 1999; Schodelbauerová wind-throw from storms (Pollmann and Veblen 2004). In Downloaded by [University of California, San Diego] at 19:15 04 March 2016 et al. 2010; Crone et al. 2013). Despite occasional model other long-lived species, the appearance of seedlings and failure, PMMs are still a useful tool in elaborating conser- saplings will depend upon the occurrence of large-scale vation strategies and may be used to reduce chances of disturbances, as was found for Agathis australis in New undesirable outcomes for endangered species (Crone et Zealand, whose behaviour can be described by the stand al. 2013). PMMs are especially important for long-lived replacement model called the Lozenge model (Ogden 1985; species, because population trends are difficult to define for Ogden and Stewart 1995). these species without the use of these models. The Lozenge model states that population regenera- Long-lived species are either climax or pioneer species of tion occurs periodically through a gap-filling event that large size (Turner 2004). In a climax species, seedlings and is structured into cohorts after disturbances (Ogden and saplings are shade-tolerant and grow beneath the canopy Stewart 1995). After a cohort dies, which can occur up to layer, promoting a continuous replacement of older trees 1 000 years after its establishment, it is replaced by a new without severe disturbance regimes (Gutiérrez et al. 2008), generation. This model has three main phases: establish- also referred to as ‘continuous regeneration’ (Ogden and ment, thinning and senescence. The establishment phase is Stewart 1995). However, in long-lived pioneers, it is usual marked by a rise in the population after a major disturbance, to find stands with a high density of mature trees while where the first seedling cohort is recruited aided by pioneer exhibiting complete absence of seedlings and saplings nurse , and the individuals’ density and biomass (Ogden 1985; Enright et al. 1999; Turner 2004), because increases. The second stage is the thinning phase, where

Southern Forests is co-published by NISC (Pty) Ltd and Taylor & Francis 2 Paludo, Lauterjung, dos Reis and Mantovani

the individuals’ density decreases through a self-thinning The study seeks to answer the following questions: What process and biomass remains constant or increases. In this are the population trends in a developed Araucaria forest? phase, Agathis australis had a basal area up to 56 m² ha−1 Will the population remain stable, increase or decrease over (Enright et al. 1999). The forest also exhibits a pattern of time? Are seedling and sapling abundance important for the high-density mature trees and it is often characterised by population’s growth rate? seedling and sapling absence (regeneration gap; Enright et al. 1999). The third stage is the senescence phase, where Material and methods the density and biomass of the first cohort decreases, and now a second seedling cohort begins to replace the Study area and data collection old mature trees through a natural regeneration process. The old growth forest used in this study is located in the There may be small gaps and seedlings present in other Caçador Genetic Forest Reserve (CGFR), in the munici- phases, but no recruitment will occur until the first cohort pality of Caçador, state of Santa Catarina, southern Brazil. enters the senescence phase (Ogden 1985; Ogden The reserve is situated between the coordinates 26°49′ and Stewart 1995). The Lozenge model has also been and 26°53′ S and 50°59′ and 50°53′ W, with an altitude utilised to examine the life history of Araucaria husteinii between 920 m and 1 075 m (Kurasz et al. 2008). It is a in New Guinea and A. laubenfelsi in New Caledonia forest fragment with an area of 1 157.48 ha. The vegeta- (Enright et al. 1999). tion type is Mixed Ombrophilous Forest, also known as The regeneration of A. angustifolia populations, considered Araucaria Forest, and is found in the biome a long-lived pioneer species, can also be described by the hotspot (Myers et al. 2000). The old growth forest in this Lozenge model (Souza 2007; Souza et al. 2008). However, study refers to a forest with numerous large long-standing previously predictions for A. angustifolia using the Lozenge trees. A plot was installed inside the old growth forest in model were tested using stem size distributions (Souza an area with no evidence of selective logging during the 2007), and size distributions are not always the best past 80 years. descriptor for population trends (Feeley et al. 2007; Virillo The target population consisted of all A. angustifolia et al. 2011). Therefore, one might question whether this individuals, from seedlings to adults within a 5.1 ha (300 m model could predict the temporal changes in descriptors of a × 170 m) plot. Each individual was identified, mapped conserved A. angustifolia population. with X and Y coordinates using a topographic survey, and If the Lozenge model is suitable in describing the followed annually from 2007 to 2013 in January, totalling A. angustifolia regeneration strategy, the population seven assessments in six years. dynamics predictions will suggest that a group of established Data were collected on the life status of the individual individuals will exhibit longevity at a site, possibly attaining (alive or dead), diameter at breast height (DBH) for trees hundreds of years until senescence and death (first with height ≥ 1.5 m, and the presence of lateral branches prediction) (Ogden and Stewart 1995). Furthermore, it is along the stem and total height for trees < 1.5 m. During expected that the population’s growth rate, changes in the subsequent years from 2008 to 2013, recruits (new biomass, and sapling abundance follow a pattern indicating individuals and recently germinated individuals) were also its status in one of the proposed model’s three phases mapped and evaluated for the same descriptors as the (second prediction). others. These data were used to determine to which class Moreover, Araucaria angustifolia is one of the most the individual belonged during each year. endangered Araucaria species, as well as one of the most poorly understood regarding stand dynamics (IUCN 2013). Data analysis Its distribution spans the southern part of South America A Lefkovitch matrix model (LMM) (Lefkovitch 1965) was (originally it covered nearly 98.96% of southern Brazil and used to estimate the population’s longevity and growth 1.03% of ; IUCN 2013), under the Araucaria rate, as well as the population trends and the role of

Downloaded by [University of California, San Diego] at 19:15 04 March 2016 Forest domain, which originally spanned an area of seedlings and saplings in population growth. This method 25.4 Mha. Araucaria angustifolia has a great socio-cultural was adapted from age-based matrix models, developed and economic importance throughout its natural distribution mainly for animals (Leslie 1945). The LMM does not because it produces large edible seeds and timber (Reis require knowledge of the individual’s age, but requires and Ladio 2012). Of the original occurrence area of the the population’s division into different stages or classes Araucaria forests in Brazil, only 0.65% are legally protected (Lefkovitch 1965). (Ribeiro et al. 2009). In addition, the seeds are exploited Therefore, individuals were classified into five categories: by local seed collectors with no control throughout its area (1) seedlings (SE) were post-germinated individuals without of distribution, including parks designated for its protection. any lateral ramifications; (2) juvenile I (J1) were individ- Because of the man-made threats to the tree species, uals with height < 1.5 m, possessing lateral ramifications; research on population trends are biologically important. (3) juvenile II (J2) were individuals with height ≥ 1.5 m This study aimed to test the hypothesis of whether and DBH < 10 cm; (4) immature (IM) were individuals Araucaria angustifolia follows the Lozenge regenera- with DBH ≥ 10 cm and < 29.2 cm; and (5) potentially tion model. In order to test this hypothesis, two predictions reproductive (PR) were individuals with a probability were established for the model, which are stated above. >50% of being in the reproductive phase, taking into Furthermore, this study also attempted to analyse popula- consideration DBH ≥ 29.2 cm. This DBH was estimated tion descriptors for an A. angustifolia population over time according to the diametric relationship with the reproduc- through a matrix model. tive phase established in Paludo et al. (unpublished data) Southern Forests 2016: 1–7 3

for the CGFR. The relationship was expressed by the method, each category had its individual frequency fixed, following equation: and the transitions (S, G or D) were randomly selected over 100 000 times with replacement (Maloney 2011). When DBH = (ln((1 − P)−1 − 1) + 7.61299) 0.26070−1 no death appeared in a class within the randomly selected matrices, a mortality rate of 0.001 was assigned. In each where ln is the natural logarithm, P is the probability of an iteration a randomly selected fecundity value was assumed individual being in the reproductive phase, and DBH the from a set containing five field-observed values (n = 5; from diameter at breast height. 2008 to 2013). The elasticity analysis was used to access In old and large trees DBH measurements are susceptible the relative participation of each transition element in λ. to errors such as expansion or contraction of bark thickness Indicators of the species longevity were also calculated induced by atmospheric conditions or simply due to random using each class size divided by the class’s average errors. Initially, in order to avoid such problems, a regression growth rate, suggesting the possible amount of time that an between size and year was calculated, which generated individual tree could remain in the class (Rigg et al. 2010). an equation with the coefficients (b – DBH growth rate) for Basal area was used as an approximation to biomass each individual. Secondly, during each year, the size of each measurement in order to verify increases or decreases in individual was estimated through the calculated regression. biomass for the second prediction. The basal area was This procedure allowed for the estimation of the size of each calculated for 2007 and 2013, and a confidence interval was individual every year, and reduces variations as well as constructed using 100 000 randomly selected iterations for random errors between size measurements. the same amount of individuals in a population corresponding It is important to note that for A. angustifolia, seed to the year in question. According to the asymmetric distribu- quantity was slightly correlated to the size of the reproduc- tion of this descriptor, a non-parametric comparison test was tive tree measured (r = 48% for basal area and r = 58% for used (Zar 2010). A two-sample Wilcoxon test (Zar 2010) was height; Mantovani et al. 2004). used for non-paired samples. Transition probabilities for each class were stasis Analyses were performed using the R software (R Core (remaining within the same class; S), advancing classes (G) Team 2012), in association with the popbio package or death (D). Those probabilities were calculated using the (Stubben and Milligan 2007). observed data and class criteria for each year. Afterwards, five matrices were assembled, one for each annual transition Results between 2008 and 2013. Fecundity was estimated as the number of seedlings found in 5.1 ha, divided by the During the study the population showed an average of 113 number of PR and multiplied by the category’s stasis for the (22 individuals ha−1) SE, 136 (27 individuals ha−1) J1, 107 specified year (Caswell 1989). The transition matrix used (21 individuals ha−1) J2, 99 (19 individuals ha−1) IM and 144 for the analysis was generated with the transition elements (28 individuals ha−1) PR within 5.1 ha, totaling 117 individ- average from the five matrices. The mortality rates (D) for uals ha−1. The generated transition matrix was summarised each class were calculated as follow: in the population’s life cycle (Figure 1). In the study area, only the seeding process produced new individuals. In D = 1 – S − G this unlogged site, no resprouting individuals or vegetative propagation were recorded. Female PR produced seeds The long-term population growth rate was estimated by that generated SE and sometimes J1. A seed could only calculating the population’s finite growth rate (λ), using generate a J1 when the individuals grew fast and grew the dominant eigenvalue of the matrix (Caswell 1989). ‘branches’ prior to the first census after the seeds fell. Confidence intervals with 95% probability were constructed The mortality rates (D) for the categories PL, J1, J2, IM for this value using bootstrap resampling and the percen- and PR were, respectively, 0.795, 0.196, 0.064, 0.022 and

Downloaded by [University of California, San Diego] at 19:15 04 March 2016 tile method (Caswell 1989). To compute the percentile 0.003. The transition between categories followed a similar

0.688 0.071

0.197 0.013 0.009 0.008 SE J1 J2 IM PR

0.008 0.792 0.927 0.970 0.997

Figure 1: Population life cycle of Araucaria angustifolia at the Caçador Genetic Forest Reserve, Santa Catarina State, Brazil. The circles represent the different stages (SE = seedling, J1 = juvenile 1, J2 = juvenile 2, IM = immature, PR = potentially reproductive). The horizontal lines represent stage advancement, the lower lines represent permanence within a class and the upper lines represent fecundity accompanied by their respective probabilities of occurrence between each census, determined by the six annual censuses in a 5.1 ha plot 4 Paludo, Lauterjung, dos Reis and Mantovani

pattern of reduction, where the larger the size the lower the Finite population growth rate (λ) and longevity transition rates (Figure 1). For the A. angustifolia population in Santa Catarina, this

The population’s growth rate (λ) was 0.9977, close study found a λ close to 1 (λ = 0.9977, CI95% = 0.9864– to 1 (95% CI: 0.9864 < λ < 1.002; CI based on 100 000 1.002), but with a large concentration of values <1 bootstrap resampling). The distribution of λ values obtained (Figure 2). In two populations of A. laubenfelsii in New through bootstrap resampling is shown in Figure 2. Caledonia, which is also a long-lived pioneer, a λ consist- Elasticity was mainly concentrated in the stasis of PR ently >1 (from 1.013 to 1.184) was found (Rigg et al. 2010). individuals, and secondly in the stasis of IM individuals, with Furthermore, in New Caledonia in four A. muelleri popula- a low participation from other transitions (Table 1). tions, the average λ was 1.0020, varying from 0.9955 to The growth in height for J1 was 0.0436 m y−1. Whereas 1.0039, with all CI values including the value 1 (Enright et the growth in DBH for J2 was 0.0952 cm y−1, for IM it was al. 2013). For the species A. husteinii and A. cunninghamii 0.110 cm y−1 and for PR it was 0.419 cm y−1. The time in New Guinea, λ values ranged from 1.055 to 1.010, given by the ratio between growth rate and size for each respectively (Enright et al. 1995). The first prediction category was 29.5 years for J1 (considering that individuals (a group of established individuals will exhibit longevity at a of age 1 year have a 0.20 m height), 105.0 years for J2, site, possibly reaching hundreds of years until senescence 174.5 years for IM, and 286.9 years for PR. In total, the and death) is confirmed by the λ found for A. angustifolia, time between germination until the current moment for the plus the time given by the ratio between growth rate and largest individual (149.4 cm) in the plot was estimated to be class size of 595.9 years, where the species exhibits high 595.9 years. longevity. When only considering the demographic rates −1 The population’s basal area in 2007 was 12.1 m² ha (λ = 0.9977, CI95% = 0.9864–1.002), the present popula- (CI 95% using the percentile method: 9.9–14.4 m² ha−1), and tion will persist for a long time at the study site, unless for 2013 it was 12.8 m² ha−1 (CI 95% using the percentile affected by a catastrophic event. As an example of the method: 10.6–15.2 m² ha−1). The difference between the species longevity, a deterministic projection of λ for the years was not significant (W = 52 108; P = 0.67). studied population indicates that 50% (59 individuals ha−1) of individuals (117 individuals ha−1) would remain for Discussion 301 years, and 10% (12 individuals ha−1) for 999 years.

The results presented in this study were constructed based Re-establishment of adults on a six-year period of observation, which is a small interval Araucaria angustifolia at this site exhibited an average when considering the lifespan of a long-lived conifer. Thus, recruitment rate of 0.75 individuals y−1 adult−1. This value caution is required when interpreting the presented data, can be contrasted to A. laubenfelsii whose average rate because it may not be an exact and precise forecast of the was 104 individuals y−1 adult−1 and λ consistently >1 (Rigg studied population’s future. et al. 2010). Araucaria muelleri had an average rate of 0.585 individuals y−1 adult−1, with λ close to 1 (Enright et al. 2013). In contrast, for Carya glabra in the USA, even with (a) (a) No deaths in PR recruitment of 18.5 individuals y−1 adult−1, the projection (b) One death in PR 50000 indicated a population decline (Evans and Keen 2013). (c) Two deaths in PR With the exception of A. muelleri, which shows one of the highest known survival rates (Enright et al. 2013), it can be concluded that the number of recruits for A. angustifolia is low relative to other studies, indicating that these rates 30000 may be insufficient to guarantee the maintenance of adult individuals in the population. Some of the causes for this (b) FREQUENCY Downloaded by [University of California, San Diego] at 19:15 04 March 2016 may be the high demand for A. angustifolia seeds by humans (IUCN 2013) and small rodents (Iob and Vieira 10000

(c) 0 Table 1: Elasticity matrix calculated from an average transition 0.97 0.98 0.99 1.00 1.01 matrix of an A. angustifolia population in a forest area in Santa LAMBDA Catarina State, southern Brazil. The data were obtained from seven annual censuses in a 5.1 ha plot (170 m × 300 m). SE = seedling, Figure 2: Distribution of λ values resampled 100 000 times from J1 = juvenile 1, J2 = juvenile 2, IM = immature, PR = potentially a set containing all observations for t = 1 year, with replacement. reproductive The average abundance between years in each category remained the same (SE = 22, J1 = 113, J2 = 107, IM = 99 and PR = 144) Category at Category at time t and the status of the individual (alive, dead or advanced) was time t + 1 randomly selected for each category separately (SE, J1, J2, IM and SE J1 J2 IM PR PR) from the observed data. Fecundity was randomly selected from SE 0.000003 0 0 0 0.000320 observed values. The intervals constructed based on the percentile J1 0.000320 0.001866 0 0 0.000165 method indicated a λ value close to 1 (λ = 0.9864–1.0020). These J2 0 0.000485 0.006348 0 0 are the possible λ values for a population of A. angustifolia within a IM 0 0 0.000485 0.016860 0 conserved forest and were taken from six annual censuses PR 0 0 0 0.000485 0.972661 Southern Forests 2016: 1–7 5

2008), unsuitable microsites or some other process that Schaaf et al. (2006) found an increase from 11.3 to has not been studied or was not considered, because 16.47 m² ha−1 during a 21-year period for A. angustifolia, in this was not the objective of the present study. It is a forest without clear-cutting, but with previous exploitation possible that the low recruitment is the first ‘bottleneck’ of before the first survey. Considering the predominance of λ A. angustifolia regeneration; this aspect should receive values < 1, the maintenance of basal area overtime, and the more attention and additional research is needed to explain low participation of regeneration in λ, the second prediction the main causes of this pattern. (it is expected that the population’s growth rate, changes in Moreover, in the present study, the recruitment elasticity biomass, and sapling abundance follow a pattern indicating and transition elements of the classes PL, J1 and J2 its status in one of the proposed model’s three phases) is were low. Longevity in the studied population is due to not rejected, because it is possible to identify the population the already initially established trees in the PR category in the thinning phase of the Lozenge model. (Table 1), which possess a low mortality rate and may It is expected that the population’s future behaviour also remain for a long time within the population. The concen- will follow the Lozenge model, because the predictions tration of elasticity in the last class’s stasis is an expected constructed using the Lozenge model (Ogden and Stewart pattern for forest tree species (Watkinson 1997) and 1995) were able to predict the established behaviour. If this for long-lived (e.g. Kwitt et al. 2004; Enright et were to happen in practice, the population would gradually al. 2013), which may indicate a possible characteristic pass into the next phase, which will occur when the current of long-lived species for model matrices. Meanwhile, adult population enters into senescence. Concurrently according to the λ distribution values obtained through with senescence, the appearance of a large number of iterations (Figure 2), λ is basically determined by the seedlings and saplings sufficient to form a new group of annual mortality rate of the PR individuals. A value of λ > 1 reproductive individuals is expected. However, to expect a (for t = 1 year) was observed only in iterations where no new group to re-establish from senescent individuals may PR died (Figure 2a). In iterations where at least one PR seem counter-intuitive. Boehmer et al. (2013) observed a individual died, λ was always less than 1 (Figure 2b and similar situation for Metrosideros polymorpha in Hawaii, c), and the peaks in the simulation distributions coincide where episodic regeneration, denominated ‘waves of with the number of dead PR individuals (Figure 2a–c). regeneration’, were found, even though the adult population This result indicates that the random selection of individual was already in senescence. transitions in all of the other categories did not cause The Lozenge model can be accepted for A. angustifolia, variations in λ greater than those caused by the deaths with hopes that in the future regeneration will appear in of the PR. Araucaria angustifolia seedlings and saplings the population. Nonetheless, the present data suggests were present at the study site (about 70 individuals ha−1), that the current regeneration is insufficient in this study but the results indicate that the seedlings and saplings area in Santa Catarina, and strong pressure exists on this were insufficient to completely replace the deaths of the PR species overall. In the twentieth century, at least 30 Mm3 individuals in the observed conditions. Random selection of A. angustifolia lumber was produced from native forests generated different transition rates for each class and, even (Carvalho 2010), and currently the pressure on the species’ then, this did not promote significant increases in λ. Given future is caused via by excessive seed collection (IUCN that the number of recruits per year was low compared with 2013). Collection occurs across the entire Araucaria other species, such results suggest a regeneration failure forest area, with no locations being free of collection. in the current conditions. According to Turner (2004) for a Moreover, only 0.65% of Araucaria forests in Brazil are long-lived pioneer species, the regeneration categories legally protected (Ribeiro et al. 2009). All these factors might be absent when referring to a mature stand. lead to greater risk of species extinction, especially since The reasons for such a failure were not evaluated in the the species distribution is in a country that has recently present study and are not sufficiently known for A. angusti- approved changes in legislation that increase the reduction

Downloaded by [University of California, San Diego] at 19:15 04 March 2016 folia. Although A. angustifolia seedlings and saplings are of native forest areas (Nazareno et al. 2011). Further able to live under low light availability (Duarte et al. 2002), studies should focus on the causes of the low number of the results of the present study showed negligible transitions individuals in the first developmental classes and the between classes. Several hypotheses can be formulated reasons for a limited entry of individuals. to help explain the regeneration failure, for example (1) the slow growth of individuals with low light availability and Conclusion (2) damage to and death of trees caused by a wide range of herbivores (Crawley 1997). For Carya glabra, herbivory The native population of A. angustifolia studied tends was indicated to be responsible for regeneration failure towards a very slow decline in the number of individuals, (Evans and Keen 2013). Drake et al. (2012) suggest that defined here as declining stability. The decline occurs regeneration failure for A. araucana in Chile, which has because of the low recruitment of new individuals, and characteristics similar to those of A. angustifolia, was seedlings and saplings do not grow to the reproductive stage possibly due to seed predation, low production of reproduc- at rates that are equal or higher than the rate of deaths of tive individuals and/or competition with fast-growing species. the reproductive individuals. Nonetheless, the decline is very slow, and is closer to stability because the individuals in the Lozenge regeneration model reproductive stage exhibit expressive longevity. The basal area for the A. angustifolia population between However, the predictions regarding the Lozenge model 2007 and 2013 did not demonstrate significant differences. were confirmed, and the results did not refute the hypothesis 6 Paludo, Lauterjung, dos Reis and Mantovani

that regeneration of A. angustifolia follows this model. Georgia, USA. Natural Areas Journal 33: 171–176. Thus, the model predicts that regeneration should occur in Feeley KJ, Davies SJ, Noor MNS, Kassim AR, Tan S. 2007. Do the future, when the population enters into the next phase, current stem size distributions predict future population changes? i.e. senescence. An empirical test of intraspecific patterns in tropical trees at two spatial scales. Journal of Tropical Ecology 23: 191–198. Gutiérrez AG, Avarena JC, Carrasco-Farías NV, Christie DA, Acknowledgements — We are thankful to the Fundação de Amparo Fuentes M, Armesto JJ. 2008. Gap-phase dynamics and coexist- a Pesquisa e Inovação do Estado de Santa Catarina (FAPESC) ence of a long-lived pioneer and shade-tolerant tree species for financial support in this study, to the Conselho Nacional de in the canopy of an old-growth coastal temperate rain forest of Desenvolvimento Científico e Tecnológico (CNPq) for providing Chiloe Island, Chile. Journal of Biogeography 35: 1674–1687. a productivity scholarship to MSR, and to the Coordenação de Iob G, Vieira EM. 2008. Seed predation of Araucaria angustifolia Aperfeiçoamento de Pessoal de Nível Superior for providing (Araucariaceae) in the Brazilian Araucaria forest: influence a scholarship to GFP. We also would like to thank the Empresa of deposition site and comparative role of small and ‘large’ Brasileira de Pesquisa Agropecuária for availability of the study mammals. Ecology 198: 185–196. area. Lastly, we would like to thank Alison Paulo Bernardi and IUCN (International Union for the Conservation of Nature and Newton Clóvis Freitas da Costa for helping in the development of Natural Resources). 2013. IUCN Red List of Threatened Species. this research, the Núcleo de Pesquisas em Florestas Tropicais Version 2013.1. Available at http://www.iucnredlist.org [accessed for their assistance, and Anna J Mello for the translating process. 8 July 2013]. The constructive input from two anonymous reviewers is gratefully Kurasz G, Rosot NC, Oliveira YMM, Rosot MAD. 2008. Caracteri- acknowledged. zação do entorno da Reserva Florestal Embrapa/Epagri de Caçador (SC) usando imagem Ikonos. Floresta 38: 641–649. References Kwitt C, Horvitz CC, Platt WJ. 2004. Conserving slow-growing, long-lived tree species: input from the demography of a rare Bierzychudek P. 1999. Looking backwards: assessing the understory conifer, Taxus floridana. Conservation Biology 18: projections of a transition matrix model. Ecological Applications 432–443. 9: 1278–1287. Lefkovitch LP. 1965. The study of population growth in organisms Boehmer HJ, Wagner HH, Jacobi JD, Gerrish GC, Mueller-Dombois grouped by stages. Biometrics 21: 1–18. D. 2013. Rebuilding after collapse: evidence for long-term cohort Leslie PH. 1945. On the use of matrices in certain population dynamics in the native Hawaiian rain forest. Journal of Vegetation mathematics. Biometrika 33: 183–212. Science 24: 639–650. López-Mata L. 2013. The impact of seed extraction on the popula- Carvalho MMX. 2010. Uma grande empresa em meio à floresta: a tion dynamics of Pinus maximartinezii. Acta Oecologia 49: 39–44. história da devastação da floresta com Araucária e a Southern Maloney PE. 2011. Population ecology and demography of an Brazil lumber and colonization (1870-1970). PhD thesis, endemic subalpine conifer (Pinus balfouriana) with a disjunct Universidade Federal de Santa Catarina, Brazil. distribution in California. Madroño 58: 234–248. Caswell H. 1989. Matrix population models: construction, analysis, Mantovani A, Morellato LPC, Reis MS. 2004. Fenologia reprodutiva and interpretation. Sunderland, MA: Sinauer Associates. e produção de sementes em Araucaria angustifolia (Bert.) Crawley MJ. 1997. Plant–herbivore dynamics. In: Crawley MJ O. Ktze. Revista Brasileira de Botânica 27: 787–796. (ed.), Plant ecology (2nd edn). Oxford: Blackwell Publishing. Myers N, Mittermeier RA, Mittermeier CG, Fonseca GAB, Kent J. pp 401–474. 2000. Biodiversity hotspots for conservation priorities. Nature Crone EE, Ellis MM, Morris WF, Stanley A, Bell T, Bierzychudek 403: 853–858. P, Ehrlén J, Kaye TN, Knight TM, Lesica P, Oostermeijer G, Nazareno AG, Feres JM, Carvalho D, Sebbenn AM, Lovejoy TE, Quintania-Ascencio PF, Ticktin T, Valverde T, Williams JL, Doak Laurance WF. 2011. Serious new threat to Brazilian forests. DF, Ganesan R, Mceachern K, Thorpe AS, Menges ES. 2013. Conservation Biology 26: 5–6. Ability of matrix models to explain the past and predict the future Ogden J. 1985. An introduction to plant demography with special of plant populations. Conservation Biology 27: 968–978. reference to New Zealand trees. New Zealand Journal of Botany Drake F, Molina JR, Herrera MA. 2012. An ecophysiographic 23: 751–772. approach for regeneration management. Ogden J, Stewart GH. 1995. Community dynamics of the New Ciencia e Investigación Agraria 39: 159–176. Zealand conifers. In: Enright N, Hill RS (eds), Ecology of the Downloaded by [University of California, San Diego] at 19:15 04 March 2016 Duarte LS, Dillenburg LR, Rosa LMG. 2002. Assessing the role southern conifers. Washington, DC: Smithsonian Institution of light availability in the regeneration of Araucaria angustifolia Press. pp 120–155. (Araucariaceae). Australian Journal of Botany 50: 741–751. Pollmann W, Veblen TT. 2004. Nothofagus regeneration dynamics Enneson JJ, Litzgus JD. 2008. Using long-term data and a stage- in South-Central Chile: a test of a general model. Ecological classified matrix to assess conservation strategies for an Monographs 74: 615–634. endangered turtle (Clemmys guttata). Biological Conservation Portela RCQ, Bruna EM, Santos FAM. 2010. Demography of palm 141: 1560–1568. species in Brazil’s Atlantic forest: a comparison of harvested Enright NJ, Franco M, Silvertown J. 1995. Comparing plant life and unharvested species using matrix models. Biodiversity and histories using elasticity analysis: the importance of life span and Conservation 19: 2389–2403. the number of life-cycle stages. Oecologia 104: 79–84. R Core Team. 2012. R: a language and environment for statistical Enright NJ, Miller BP, Perry GLW, Goldblum D, Jafré T. 2013. computing. Vienna: R Foundation for Statistical Computing. Stress-tolerator leaf traits determine population dynamics in the Reis MS, Ladio A. 2012. Paisajes con en Sudamérica: endangered New Caledonian conifer Araucaria muelleri. Austral construcciones culturales pre-colombinas y del presente para Ecology 39: 60–71. producción de alimento In: Navarro V, Espinosa S (eds), Paisajes Enright NJ, Ogden J, Rigg LS. 1999. Dynamics of forests with Culturales: memorias de las Jornadas de reflexión acerca de los Araucariaceae in the western Pacific. Journal of Vegetation paisajes culturales de Argentina y Chile, en especial los situados Science 10: 793–804. en la región Patagónica. Rio Gallegos: ICOMOS/UNPA/UMAG. Evans JP, Keen EM. 2013. Regeneration failure in a remnant stand pp 224–244. of pignut hickory (Carya glabra) on a protected barrier island in Ribeiro MC, Metzger JP, Martensen AC, Ponzoni FJ, Hirota MM. Southern Forests 2016: 1–7 7

2009. The Brazilian Atlantic Forest: how much is left, and how is logging in southern South America. Acta Oecologica 34: 221–232. the remaining forest distributed? Implications for conservation. Stubben CJ, Milligan BG. 2007. Estimating and analyzing Biological Conservation 142: 1141–1153. demographic models using the popbio package in R. Journal of Rigg LS, Enright NJ, Jaffré T, Perry GLW. 2010. Contrasting popula- Statistical Software 22(11): 1–23. tion dynamics of the endemic new caledonian conifer Araucaria Turner IM. 2004. The ecology of trees in the tropical rain forest. laubenfelsii in maquis and rain forest. Biotropica 42: 479–487. Cambridge: Cambridge University Press. Schaaf LB, Figueiredo-Filho A, Galvão F, Sanquetta CR, Longhi SJ. Venter SM, Witkowski ETF. 2013. Using a deterministic popula- 2006. Modificações florístico-estruturais de um remanescente de tion model to evaluate population stability and the effects of floresta ombrófila mista montana no período entre 1979 e 2000. fruit harvesting and livestock on baobab (Adansonia digitata L.) Ciência Florestal 16: 271–291. populations in five land-use types. Forest Ecology and Manage- Schmidt IB, Mandle L, Ticktin T, Gaoue OG. 2011. What do matrix ment 303: 113–120. population models reveal about the sustainability of non-timber Virillo CB, Martins FR, Tamashiro JY, Santos FAM. 2011. Is size forest product harvest? Journal of Applied Ecology 48: 815–826. structure a good measure of future trends of plant populations? Schodelbauerová I, Tremblay RL, Kindlmann P. 2010. Prediction An empirical approach using five woody species from the vs. reality: can PVA model predict population persistence 13 Cerrado (Brazilian savanna). Acta Botanica Brasilica 25: years later? Biodiversity and Conservation 19: 637–650. 593–600. Souza AF. 2007. Ecological interpretation of multiple population Watkinson AR. 1997. Plant population dynamics. In: Crawley MJ size structure in trees: the case of Araucaria angustifolia in South (ed.), Plant ecology (2nd edn). Oxford: Blackwell Publishing. America. Austral Ecology 32: 524–533. pp 359–400. Souza AF, Forgiarini C, Longhi SJ, Brena DA. 2008. Regeneration Zar JH. 2010. Biostatistical analysis (5th edn). Upper Saddle River: patterns of a long-lived dominant conifer and the effects of Prentice Hall. Downloaded by [University of California, San Diego] at 19:15 04 March 2016

Received 1 February 2015, revised 8 October 2015, accepted 15 November 2015