Appearances can be deceiving. A site that currently hosts high biodiversity may not provide sufficient conditions for long-term persistence of species. Although habitat loss and land-use changes are the main causes of local extinctions in the tropics, species extinction may not be immediate (Hanski, 2013; Hylander and Ehrlén, 2013). When habitat loss is sudden, there may be a time-lag before extinction and colonization dynamics reach a new equilibrium with current habitat distribution – non-equilibrium dynamics (Krauss et al., 2010; Ovaskainen and Hanski, 2002; Tilman et al., 1994). As consequence, sometime after disturbance some species may be found in fragments where persistence is unlikely, even with no further land use change (the so-called “extinction debt”) (Hanski, 2000; Tilman et al., 1994).

A snapshot in time may therefore be a misleading representation of the capacity of a site to sustain biodiversity in long term (Krauss et al., 2010; Metzger et al., 2009; Vellend et al., 2013). Monitoring over time is therefore required for assessing effects of land use changes on local dynamics (Hanski, 2013). One way to monitor species persistence in fragmented landscapes is modeling extinction and colonization rates of habitat patches, using metapopulation theory. Metapopulation models link the proportion of occupied habitat patches to local extinction and colonization rates (Hanski, 1998, 2000). In a monitoring context, the rate of change in proportion of occupied patches is an important indicator of species viability (Ovaskainen and Hanski, 2002). Furthermore, assessing occupancy trends over time also enables focusing on mechanisms underlying site occupancy dynamics (Hylander and Ehrlén, 2013; Kuussaari et al., 2009; MacKenzie et al., 2003).

In this paper, we assessed the main drivers of changes in occupancy of two primate species, the black penciled marmoset Callithrix penicilatta and the black-fronted titi monkey Callicebus nigrifrons. New World marmosets of genus Callithrix are a small-bodied (<500g) group of monkeys that play a unique role in forests of Central and South America, because they are specialized on tree gouging and feeding on gum exudates (Rylands et al., 2009). Marmosets also have high reproductive rates (usually twins, twice a year), early sexual maturity (around one year old), and are relatively short-lived (up to 10 years in wildlife) (Mittermeier et al., 2013; Mustoe et al., 2012).

Marmoset small body size, rapid reproductive rates and systematic exploitation of plant exudates allow them to occupy several habitat types, even relatively inhospitable environments (Rylands et al., 2009). Indeed, black-penciled marmoset is not only able to tolerate habitat fragmentation, but is usually found in disturbed and secondary forests and also edge habitats (Kinzey, 1997; Vilela and Del-claro, 2011), being listed as “Least Concern (LC)” by IUCN Red List (www.iucn.org).

Titi monkeys of genus Callicebus have larger body size (1–2kg) (Roosmalen et al., 2002), slower reproduction rates and longer generation times, compared to marmosets (Kinzey, 1997). Female first birth happens at around 3.7±1.3 years, breeding females give birth to a single offspring annually (Valeggia et al., 1999), and they live up to 25 years (Mittermeier et al., 2013).

Black-fronted titi monkey (C. nigrifrons) originally had a wide distribution in Brazilian Atlantic forest and Cerrado. However, fragmentation and urbanization of forests within its range caused remaining populations to be isolated and generally small (Roosmalen et al., 2002). Due to 20–25% population declines in past 25 years, black-fronted titi monkey is listed as “Near Threatened (NT)” by IUCN Red List.

We created a sampling scheme that allowed us to infer the effect of a massive land use change in southeastern Brazil – the implementation of a hydroelectric reservoir (Fig. S1) – on primates’ occupancy, extinction and colonization probabilities. The effect of forest patch size and connectivity on extinction and colonization dynamics was also assessed. We tested the following hypothesis: (1st) time-lags can be found if observed occupancy is different from equilibrium expectations; (2nd) the faster a species responds to environmental disturbances, the less we expect it to exhibit time-lags; (3rd) occupancy is determined by patch size and/or connectivity in a positive relation; (4th) reductions in size and/or connectivity increase extinction and reduce colonization rates.

We created a sampling scheme that allowed us to infer the effect of a massive land use change in southeastern Brazil – the implementation of a hydroelectric reservoir (Fig. S1) – on primates’ occupancy, extinction and colonization probabilities. The effect of forest patch size and connectivity on extinction and colonization dynamics was also assessed. We tested the following hypothesis: (1st) time-lags can be found if observed occupancy is different from equilibrium expectations; (2nd) the faster a species responds to environmental disturbances, the less we expect it to exhibit time-lags; (3rd) occupancy is determined by patch size and/or connectivity in a positive relation; (4th) reductions in size and/or connectivity increase extinction and reduce colonization rates.

There was a significant reduction in forest fragment size after deforestation of the area destined to Funil hydroelectric dam (t-value=2.47; p<0.05). However, there was an overall increase in connectivity for both titi monkey (t-value=3.84; p<0.05) and marmoset (t-value=4.23; p<0.05).

For marmoset, single-season models with current landscape covariates better explained nowadays occupancy than past landscape models (Table 1). Most supported dynamic model indicated by ΔQAICc suggested that extinction events were mainly driven by changes in site connectivity (Table 2). However, the constant model was similarly supported (ΔQAICc<2). Also, the high Akaike's weight of constant model (QAICc Wt=0.24) provides little support for site-specific variation in parameters. Therefore, local occupancy, extinction, and colonization processes of marmoset were not strongly related to any covariates. Further, no covariates had effect on marmoset initial occupancy, extinction probabilities or colonization probabilities (Table 3).

Based on model averaged estimates of extinction (¿=0.24±0.18) and colonization probability (γ=0.62±0.41) for marmoset, occupancy expected in equilibrium is 0.72. That value is within those obtained for marmoset observed occupancy at present-day (Ψobs=0.77±0.18). Marmoset occupancy estimates also did not change through time (Ψt=0.81±0.11; Ψt+1=0.83±0.20). Further, Titi monkey current occupancy is still affected by initial forest patch size, as suggested by top-ranking model (Table 1). Most supported model indicated that initial occupancy was strongly affected by initial connectivity (Table 2). Connectivity and size before flooding of reservoir were positively related to initial occupancy, but no other covariate exhibited a strong relationship with extinction probability (Table 3). Extinction probability was low, but differed from zero (¿=0.08±0.07).

Colonization probability for titi monkey could not be modeled because we observed no colonization event. As calculation of metapopulation occupancy expected in equilibrium requires estimation of both extinction and colonization rates, we also could not estimate expected equilibrium values for titi monkey. Titi monkey occupancy estimates slightly decreased within primary sampling periods (Ψt=0.62±0.15; Ψt+1=0.57±0.14).

Species distribution in a mosaic of forest patches is the ultimate result of extinction and colonization dynamics (Ovaskainen and Hanski, 2002). Those processes are affected by habitat patch size and connectivity and can be time-delayed from habitat degradation (Hanski, 2013; Tilman et al., 1994). We aimed at disentangling the effects of forest patch size and connectivity on occupancy and extinction and colonization probabilities of two primate species, subject to a major habitat loss ten years ago.

Marmoset occupancy seems to be in equilibrium with present-day landscape, because current occupancy is equal to that expected in metapopulation equilibrium. Also, the better fit of marmoset occupancy to present-day landscape illustrates absence of time-lag effects. This corroborates our predictions that time-lag effects should only be observed if occupancy was not in equilibrium with current landscape – 1st hypothesis. Marmoset rapid reproductive rates arise through early sexual maturity and ovulation immediate to postpartum (Mustoe et al., 2012). Also, evolution of multiple litters (twins, twice a year) increases reproductive capacity (Mustoe et al., 2012). Such increased reproductive rate in marmosets, coupled with efficient use of unlimited gum resource, can be related to a faster response to environmental changes. Therefore, observed equilibrium with current landscape and absence of time-lag effects in marmoset occupancy may be related to a fast response to land use changes, as a result of rapid reproductive rates, which supports our 2nd hypothesis.

We found an indication of time-lags for titi monkey, because current occupancy is related to a landscape attribute (forest patch size) dated before a major habitat loss. Moreover, we did not observe any colonization event of previously unoccupied sites. If extinction rate is greater than zero and colonization is zero, the population is expected to decline until entire depletion. That is an overly pessimistic scenario due to our small sample size. However, that may be suggestive of long-term decline in titi monkey local population. We hypothesize that titi monkeys did not colonize any site, despite the increase in landscape connectivity, due to their dependence on canopy structures (Lawler et al., 2006; Norconk, 2007). That canopy dependence may reduce dispersal ability in non-forest matrix and disturbed patches, thus inhibiting colonization events.

The strong effect of initial connectivity on titi monkey occupancy – 3rd hypothesis – may also be a result of a delayed response. Species with long generation times, relatively stable populations and exhibiting apparent low extinction risk, such as titi monkey, are predicted to face highest probability of delayed extinctions/colonizations (Kuussaari et al., 2009). Those transient dynamics may overestimate the conservation status of populations remaining in habitat patches (Kuussaari et al., 2009; Ovaskainen and Hanski, 2002). It may be therefore a negative bias to assume that all remaining species in recently fragmented landscapes have viable populations (Hanski, 2013; Metzger et al., 2009), such as in titi monkey case.

Contrary to our expectations, variation in size and connectivity were not good predictors of extinction and colonization probabilities – 4th hypothesis. Absence of an effect of size and isolation on extinction and colonization parameters is suggestive that drivers of occupancy dynamics for marmoset and titi monkey may be other than landscape-scale processes, as also suggested by other studies (Sales et al., 2015). Extinction and colonization events can be driven by many factors apart from size and connectivity. Local factors, such as species invasions, can impact communities as much as land-use change and more than habitat loss/fragmentation (Simberloff, 2007).

Investigating the influence of local factors such as species interactions, resource distribution and habitat quality on extinction and colonization dynamics is an intriguing extension of our analysis. Use of dynamic occupancy models to assess drivers of extinction and colonization rates, and also to calculate deviance from equilibrium expectation, can also be incorporated in study of other time-lag effects.

The authors declare no conflicts of interest.