- Email: [email protected]
The role of pre- and post-dispersal seed predation in determining total seed loss
Basic and Applied Ecology 15 (2014) 581–589
ORIGINAL PAPERS
The role of pre- and post-dispersal seed predation in determining total seed loss R.D. van Klinken∗ , A.J. White CSIRO Biosecurity Flagship, EcoSciences Precinct, 41 Boggo Road, Dutton Park, PO Box 2583, Brisbane 4001, QLD, Australia Received 17 December 2013; received in revised form 26 August 2014; accepted 26 August 2014 Available online 4 September 2014
Abstract Most seed predation studies focus on either pre- or post-dispersal predation and may therefore underestimate the role of predation in regulating plant populations. We therefore estimated total seed predation of an invasive tree, mesquite (Leguminoseae: Prosopis spp.), by examining the entire seed pool from tree to seed bank. The spatio-temporal dynamics of total seed predation was examined by sampling across its Australian distribution and through time. The main predator was a host-specialist multivoltine beetle, Algarobius prosopis L. (Bruchidae), previously introduced as a biocontrol agent. Seed predation exceeded 20% in all seed stages (in pods on and off the tree, and seeds within woody endocarps (capsules) and free seeds on and in the ground) but was consistently highest in capsules on the ground (up to 90%). Pre-dispersal predation contributed little. Total seed predation rates were primarily determined by predation rates on the most persistent seed stage, in this case fallen pods if only pods are considered and seeds in capsules for the total seed pool. This pattern was consistent across the surveyed taxa, regions, years and seasonally. Predation rate was relatively unaffected by seed density, potentially because densities were always low (<150 seeds m−2 ). Average total seed predation within a region reached 55%, but we conclude that any population regulation of mesquite by seed predation will principally be through reduced seed bank persistence. Our results highlight the need to consider the entire seed pool, especially the often cryptic and overlooked long-lived stages, when determining seed loss to predation and its likely population consequences.
Zusammenfassung Die meisten Untersuchungen zu Samenprädation konzentrieren sich entweder auf Prädation vor oder auf die Prädation nach der Ausbreitung und unterschätzen deshalb möglicherweise die Bedeutung der Prädation bei der Regulation von Pflanzenpopulationen. Wir bestimmten deshalb die Gesamt-Samenprädation des invasiven Mesquitebaums (Leguminoseae: Prosopis spp.), indem wir den gesamten Samenpool vom Baum bis zur Samenbank erforschten. Die Raum-Zeit-Dynamik der GesamtSamenprädation wurde untersucht, indem wir Proben über sein australisches Verbreitungsgebiet und vier Jahre hinweg nahmen. Der Hauptprädator war ein spezialisierter, plurivoltiner Käfer, Algarobius prosopis L. (Bruchidae), der früher als biologischer Kontrollorganismus eingeführt worden war. Die Samenprädation übertraf 20% in allen Samenstadien (in den Hülsen auf dem Baum und nach dem Herunterfallen, Samen innerhalb des holzigen Endocarps (Kapseln) und freie Samen auf oder im Boden), war aber immer am höchsten in den Kapseln auf dem Boden (bis 90%). Die Prädation vor der Ausbreitung war gering. Die Gesamt-Samenprädationsraten wurden hauptsächlich bestimmt durch die Prädationsraten am langlebigsten Samenstadium, in
∗ Corresponding
author. Tel.: +61 7 38335737; fax: +61 7 3833 5504. E-mail address: [email protected] (R.D. van Klinken).
http://dx.doi.org/10.1016/j.baae.2014.08.012 1439-1791/Crown Copyright © 2014 Published by Elsevier GmbH. All rights reserved.
582
R.D. van Klinken, A.J. White / Basic and Applied Ecology 15 (2014) 581–589
diesem Fall herabgefallene Hülsen, wenn nur Hülsen betrachtet werden, bzw. den Samen in Kapseln für den gesamten Samenpool. Dies Muster war gültig für alle untersuchten Arten, Regionen, Jahre und Jahreszeiten. Die Prädationsrate blieb relativ unbeeinflusst durch die Samendichte, vermutlich weil die Dichten immer gering waren (<150 Samen m−2 ). Die durchschnittliche Samenprädation innerhalb einer Region erreichte 55%, aber wir schließen, dass jedwede Populationsregulation des Mesquitebaumes durch Samenprädation prinzipiell durch reduzierte Langlebigkeit der Samenbank erfolgt. Unsere Ergebnisse belegen die Notwendigkeit, den gesamten Samenpool zu betrachten, insbesondere die oft kryptischen und übersehenen langlebigen Stadien, wenn Samenverluste durch Prädation und ihre Konsequenzen für die Populationsdynamik bestimmt werden sollen. Crown Copyright © 2014 Published by Elsevier GmbH. All rights reserved.
Keywords: Bruchidae; Invasive plants; Plant–insect interactions; Predator escape; Pre-dispersal seed predation; Post-dispersal seed predation; Seed bank
Introduction Seed predation can be an important driver of plant population dynamics, and even community structuring (Crawley 1992; Hoffmann & Moran 1998; Kauffman & Maron 2006; Kolb, Ehrlén, & Eriksson 2007; Lewis & Gripenberg 2008; Ramírez & Traveset 2010). Seed predating insects are particularly diverse and widely studied, but their general importance is still disputed. Most studies of insect seed predators have focussed on seeds that are still on the plant (pre-dispersal predation) (Kolb et al. 2007). Univoltine insects, that pass through a single generation a year, can be well synchronised with typically observed seasonal peaks in seed production (Crawley 1992). In contrast, multivoltine insects, that pass through multiple generations a year, are less able to track sharp seasonal fluctuations in pre-dispersal host resources (Raghu, Wiltshire, & Dhileepan 2005; van Klinken et al. 2008). However, this can potentially be overcome if they continue exploiting more persistent post-dispersal seed stages (Johnson 1981; Lewis & Gripenberg 2008; van Klinken & Flack 2008; van Klinken & White 2011). Post-dispersal seeds are often cryptic and therefore overlooked in seed predation studies. As a consequence the role of multivoltine insects as a seed mortality factor, and ultimately in regulating host populations, may be underappreciated (Briese, 2000; Lewis & Gripenberg 2008). Several closely-related Prosopis species (Leguminosae), together referred to as mesquite, are long-lived shrubs and trees that reproduce entirely by seed and are native to arid and semi-arid Americas. They have also been intentionally introduced around the world where they often become serious weeds (van Klinken, Hoffmann, Zimmermann, & Roberts 2009). Pods usually contain 10–20 seeds (c 0.024 g each, unpublished data) and remain closed at maturity (van Klinken & Campbell 2009). Mesquite seeds offer a particularly dynamic resource for seed predators. The pre-dispersal phase (Fig. 1) is typically brief, with pod maturation being relatively synchronised, and pods falling from the tree soon after maturation (van Klinken & Campbell 2009). Most pods are often consumed, either from the tree or ground, within days of maturation by a wide range of vertebrate herbivores. These herbivores are not seed predators, rather they derive
nutrition from the pods and act as important seed dispersal vectors. Seeds are contained within a thin shell or endocarp (the seed capsule) which can remain intact even after the fleshy mesocarp has broken down during passage through the herbivore or through natural decay (van Klinken & White 2011) (Fig. 1). Mesquite seeds have hard-seeded (physical) dormancy imposed by the hard seed coat, from which most are released within 2–3 years under natural conditions (van Klinken et al. 2009). Most seed predation reported on mesquite in its introduced range is by a multivoltine host-specialist, Algorobius prosopis Le Conte. (Bruchidae), sourced from the Americas in an attempt to regulate mesquite populations (van Klinken et al. 2009). Females lay eggs on mature pods, seed capsules and possibly free seeds, with each larva consuming and killing an individual seed before emerging as an adult (Impson, Moran, Hoffmann, Olckers, & Hill 1999; van Klinken and White 2011). Populations can pass through several generations a year, with development from egg hatch to eclosion taking c 33 days (Hoffmann, Impson, & Moran 1993). High seed predation levels can occur on long-lived seed stages. In South Africa it can kill over 90% of seeds in mature pods within 10–12 months provided vertebrate herbivores are prevented from consuming pods (Impson et al. 1999; van Klinken et al. 2009). In Australia predation rates by A. prosopis can be high
Fig. 1. A schematic of the mesquite seed lifecycle with the typically transient pod stages highlighted. It includes pre-dispersal seeds (tree-pods) and post-dispersal seed stages on and in the ground.
R.D. van Klinken, A.J. White / Basic and Applied Ecology 15 (2014) 581–589
for long-lived seed pools in cattle dung (van Klinken & White 2011). However, no study has systematically considered the role of seed predation across the entire seed pool. Total seed loss to predation will be influenced by both oviposition behaviour and the longevity of each seed stage (Fig. 1; van Klinken, 2005). Oviposition on long-lived stages is therefore expected to have the greatest effect on total predation rates, provided oviposition preferences are similar across stages. Most seed predation studies focus on predation in pods as pods are typically the most conspicuous seed stage for researchers. We compared pre- and post-dispersal seed predation in pods through time, with the expectation that total predation rates in pods will increase through the year provided oviposition continues in fallen pods. We also sampled all seed stages between fruiting seasons (winter), including in the seed bank, to test the relative importance of seed predation in each seed stage in determining total seed loss to predation. Our expectation was that pod will be an ephemeral resource and therefore that pod predation will be limited, and that total seed predation rates in the winter seed pool will depend largely on continued oviposition on cryptic seed stages. Furthermore, we tested the versatility of these predictions through field surveys conducted over several years across the core Australian distribution of mesquite and the best-represented mesquite taxa (van Klinken & Campbell 2009). We also consider the implications of our results for the regulation of mesquite populations, including testing whether seed predation rates were correlated with seed density.
583
Klinken et al. 2009). Sites were in the summer-rainfall zone, with hot summers and cool to warm winters. The Pilbara is the driest and hottest region, and western Queensland (WQ) is slightly drier and hotter than central Queensland (CQ) (Table 1). All properties were used for cattle-grazing. Pigs were common at one site in CQ, and wild emus (Dromaius novaehollandiae (Latham)) and macropods (Macropodidae) were present at most sites, although they were most common in the Pilbara. The sampling design within a region was largely dictated by mesquite distribution (Table 1). In CQ dense Prosopis pallida extends across relatively small pastoral blocks around the town of Hughenden. Sampling was conducted on two properties located 8 km apart. In WQ P. pallida is widely distributed and sampling was conducted on three properties on which it reaches high densities. Sampling of hybrid mesquite in WQ was restricted to a single property as it has been largely managed elsewhere in WQ. In the Pilbara sampling was conducted on the property on which mesquite principally occurs (van Klinken et al. 2009). Where possible multiple sites were sampled on each property (Table 1). In CQ and WQ sampling was conducted in dense mesquite (>90% canopy cover) and under mesquite trees that had non-overlapping canopies. However, results were subsequently combined to obtain sitelevel estimates of seed density and seed predation as seed densities were low and no plant density effects were detected (data not shown).
Seed density and seed predation estimates
Materials and methods Sites and surveys Surveys were conducted in three regions across Australia where mesquite is most widespread and abundant (Table 1). Elsewhere mesquite populations were sparse or under active management during our study period (2005–2008) (van
Surveys of pods were conducted to test the relative importance of pre-dispersal (tree-pods) and post-dispersal (fallen pods) predation for total predation in pods through and between years. This was supplemented by sampling of all seed stages (Fig. 1) in winter to determine total predation across the entire seed pool. Winter is when the seed bank is expected to be most stable, as seed production and germination events are limited then (van Klinken et al. 2009).
Table 1. Details of the surveyed regions and pastoral properties. Region and properties
Location ◦
*
Species
◦
Queensland: Central (CQ) RF: 492.4 mm Tmax : 31.6 C Tmin : 16.6 C Alberts (2 sites; 505 ha) S20◦ 49 E144◦ 11 Lands block (2 sites; 65 ha) S20◦ 49 E144◦ 13
P. pallida P. pallida
Queensland: Western (WQ) RF: 469.2 mm Tmax : 32.4 ◦ C Tmin : 18.4 ◦ C Hampden downs (2 sites; 29,100 ha) S21◦ 28 Beaudesert lane (1 site; 49,200 ha) S21◦ 31 Williams (2 sites; 99,000 ha) S20◦ 36 Colwell (2 sites; 25,500 ha) S21◦ 27
E141◦ 40 E141◦ 05 E141◦ 58 E141◦ 09
P. pallida P. pallida P. pallida Hybrid mesquite
Western Australia: Pilbara Region RF: 270.9 mm Tmax : 33.9 ◦ C Tmin : 18.7 ◦ C Mardie station (5 sites; 220,900 ha) S21◦ 09 E116◦ 04
Hybrid mesquite
* Mesquite populations were either Prosopis pallida (Willd.) Kunth or hybrids (between P. glandulosa Torr., P. velutina Wooton and/or P. pallida) (van Klinken & Campbell 2009).
584
R.D. van Klinken, A.J. White / Basic and Applied Ecology 15 (2014) 581–589
Pods Pod surveys were timed to coincide with when pods were expected to be falling (early summer), to have fallen (mid to late summer) and when cumulative seed predation rates were expected to be highest (in winter). In each survey a maximum of 20 intact pods were collected from on or under individual trees to obtain a pooled sample of at least 100 pods on the tree or ground, respectively, from each site. At the same time, pod–seed density was estimated to calculate total seed predation rates (see below). Pods for bruchid emergence were returned to the laboratory within three days of collection, held in ventilated containers at 25 ◦ C and any emerging insects were removed every 1–2 days to prevent reinfestation and kept for identification. Pods were frozen after all adults had emerged (typically within 40–50 days) to kill any second generation insects. Emergent holes and seeds were subsequently counted to determine total seed predation rates for each sample. Algarobius beetles were identified to species (Kingsolver 1986). Pre-dispersal (tree) pod density was estimated by counting the number of pods on each of 10 permanently labelled large, healthy adult trees at each site and converting counts to density by dividing it by canopy area. Post-dispersal (fallen) pod density on the ground was estimated using 10 (625 cm−2 ) quadrats randomly placed under mesquite canopies at each site. Pods from quadrats were subsequently combined with pooled pod samples for bruchid emergence. Pod counts needed to be converted to seed counts to calculate seed densities and seed predation rates. The number of fully developed seeds per pod was estimated in two ways. For visual pod counts on trees the number of pods was multiplied by the average number of seeds per pod calculated from a random subsample (CQ P. pallida: n = 1631 pods, 15.4 seeds pod−1 ; WQ P. pallida: n = 134, 15.1; WQ hybrid: n = 801, 13.3; Pilbara hybrid: n = 820, 11.4). Seed numbers for collected pods were estimated by counting the external swellings of mature pods and pod fragments, and then adjusting this with the average number of fully developed seeds per swelling determined through dissection of random pods (CQ P. pallida: n = 43 pods; 94.2% of swellings had fullydeveloped seeds; WQ P. pallida; n = 20, 95.7%; WQ hybrid: 30, 89.2%; Pilbara hybrid: 26, 86.2%). Total seed pool Winter sampling of pods (2005–2008) at each site was supplemented with soil coring to estimate seed bank densities and seed predation rates for seeds located both on and in the soil (Fig. 1). At each site a single soil core (625 cm−2 wide × 5 cm deep) was taken from under the canopy of each of 10 isolated large mature adults (midway between trunk and edge of canopy) and 10 soil cores were randomly taken along a 20 to 40 m transect under dense mesquite stands. Surface litter and soil were kept and processed separately for each soil core. All free seeds, seeds in capsules and pods (whole and fragments) were removed from samples through sieving and visual inspection, then frozen for processing.
Seed densities and seed predation rates (the proportion of seeds with adult emergence holes at the time of freezing) were subsequently calculated for each seed stage (Fig. 1) at each site (by averaging soil cores).
Data analysis Seed density and seed predation rates for each site-visit were calculated for each individual seed stage (Fig. 1) and for pooled stages: all pod–seeds, all seeds not in pods (capsules and free seeds), and the total seed pool. Total predation rates for pooled stages were estimated by summing the proportional contribution of seed predation of each seed stage (van Klinken, 2005). Seed density and seed predation rates for each seed stage, and for pooled stages, were analysed following appropriate transformations to satisfy ANOVA conditions. Significant differences were determined on the transformed scale. Analyses included region–species combination (Table 1) as a factor, and year where sufficient data was available (CQ only). Seed predation rates were weighted by the number of seeds collected in the ANOVA. Correlation coefficients were calculated between seed density and seed predation rates (calculated for each site and survey) and tested for significance.
Results Pods Overall predation rates in pods were 9.8% (n = 16,953 seeds) for pre-dispersal seeds (tree-pods) and 18.6% (n = 39,958) for post-dispersal seeds (fallen pods), and was highest in CQ and the Pilbara (Fig. 2A). Most emergents (1053) following field collection were A. prosopis beetles, which were present at all sites. In addition, eleven unidentified bruchid beetles emerged from a single sample, and 10 larval or pupal wasp parasitoids emerged across all samples. Average seed predation was similar on tree and fallen pods in CQ and the Pilbara, but higher in fallen pods in WQ (Fig. 2A). Average pod–seed densities were highest on the ground (Fig. 2B). Together this meant that pre-dispersal predation contributed little to total predation, except in the Pilbara (Fig. 2C) where fallen pod density was high on one occasion and zero on the remaining seven. Seed predation and seed density was not correlated for tree-pods (correlation coefficient = −0.13, P = 0.52, n = 28), but seed density was always low (Fig. 2B). In contrast, predation in fallen pods was positively correlated with seed density on the ground (correlation coefficient = 0.42, P < 0.001, n = 60). CQ was the only region where there were sufficient seeds to examine the effect of season and year on predation rates. Again, predation of pre-dispersal seeds (Fig. 3A) generally contributed little to total predation (Fig. 3C). Pod–seed densities on the tree were consistently low (Fig. 3B), as most pods had matured and fallen by the early-summer surveys
R.D. van Klinken, A.J. White / Basic and Applied Ecology 15 (2014) 581–589
585
Fig. 2. Average (±SE) predation rate (A) and seed density (B) for pre-dispersal (tree) pods and post-dispersal (fallen) pods, and (C) the relative contribution of pre-dispersal predation to total seed predation. Different letters indicate significant difference (LSD test, P < 0.05) within each significant dependent variable.
(unpublished data from litter traps). Total predation rates in pods did vary with time, but there was no evidence that it increased within a season (Fig. 3A). Average predation rates in tree-pods were higher in 2005–2006 (19%) than in 2006–2007 (4%) (P < 0.05), so were inversely proportional to seed densities in the tree (Fig. 3B), but there was no density dependent relationship overall (correlation coefficient = 0.0107, P = 0.38, n = 23). Seed predation levels on fallen pods ranged from 8 to 40% (Fig. 3A), but there was no year or season effect. There was also no correlation between seed densities in fallen pods and seed predation (correlation
coefficient = 0.36, P = 0.16, n = 23), unlike in the regional analysis (see above). Pod densities were generally too low in winter to reliably estimate predation rates (Fig. 3A).
Total seed pool Total predation rate (estimated in winter) was highest in CQ (averaging 54.9%), compared to 20.7 to 25.5% elsewhere (Fig. 4A). Predation in pod–seeds contributed relatively little to total predation rates (Fig. 4C) owing to
586
R.D. van Klinken, A.J. White / Basic and Applied Ecology 15 (2014) 581–589
Fig. 3. Seasonal changes in average (±SE) seed predation rates (A), seed density (B) for pre-dispersal (tree) pods and post-dispersal (fallen) pods in CQ, and (C) the relative contribution of pre-dispersal predation to total seed predation. Asterisk indicates pods were too rare to collect.
lower or similar seed densities (Fig. 4B) and consistently lower predation rates (Fig. 4A). Total seed density was highest in CQ and WQ (hybrid), and was barely detectable using our sampling method at the remaining two locations (Fig. 4B). Total seed predation was not correlated with total seed density (correlation coefficient = 0.14, P = 0.42, n = 36). A between-year comparison of seed predation across all seed stages was only possible for CQ where seed numbers were highest. Total seed predation rates ranged from 36 to 72% (Fig. 5A). Predation of seeds in capsules on the ground contributed most to total seed predation (Fig. 5C), owing to consistently high seed predation rates (Fig. 5A) and its contribution to the total seed pool (Fig. 5B). This contrasted with seeds in pods located on both the tree and ground for which average seed predation rates within a year never exceeded 20% (Fig. 5A), and capsules in the
ground which had high predation rates (Fig. 5A) but relatively low densities (Fig. 5B). The contribution of predation in the pre-dispersal seed fraction on total seed predation was insignificant (Fig. 5C).
Discussion In Australia mesquite presents a challenging host for seed predators as seeds were often at low densities and represented a dynamic resource. Seed predation was almost entirely by the biological control agent A. prosopis, which was present at all sites. This species is a specialist on mesquite species, but a generalist with regards to seed stage (Kingsolver 1986; Impson et al. 1999; van Klinken et al. 2009). Patterns in seed predation rates suggest oviposition occurs on the tree,
R.D. van Klinken, A.J. White / Basic and Applied Ecology 15 (2014) 581–589
587
Fig. 5. Average (mean ± SE) seed predation rates (A) and seed density (B) for all seeds combined and for individual seed stages across years in CQ, and (C) the relative contribution of predation in each seed stage to total seed predation.
Fig. 4. Average (±SE) between-year (winter) predation rate (A) and seed density (B), for seeds in pods, seeds no longer in pods and the total seed pool, and the relative contribution of predation in pod–seeds to total seed predation (C). Different letters indicate significant difference (LSD test, P < 0.05) within each significant dependent variable.
fallen pods, seed capsules and possibly free seeds, and it also oviposits on seeds in dung (van Klinken & White 2011). Such wide oviposition preferences are likely to provide a useful mechanism for maintaining seed predator populations when seed resources vary in abundance and quality through the year (van Klinken & Flack 2008). As predicted for multivoltine insects, total predation rates were primarily determined by predation rates on the most persistent seed stage. For total predation it was seeds on the ground that were no longer in pods, which are rarely considered in seed predation studies.
These patterns held irrespective of season, year, region or mesquite taxon. Total seed predation ranged between 20 and 55%. Predation was relatively high at times for each individual seed stage (Fig. 1), exceeding 40% even on the most transient pre-dispersal stage (tree pods). However, in our study predispersal predation of seeds on the tree had little influence on total seed predation levels, and was a poor predictor of predation levels on the ground, despite being the primary focus of many seed predation studies (Kolb et al. 2007). Rather, in our study it was the long-lived, relatively cryptic, seed stages on which oviposition continued that was principally responsible for total seed predation. Predation was consistently highest for seeds in capsules (averaging 80% in Central Queensland), and seeds in capsules were generally also the most common seed stage between seasons. The role of the pericarp in protecting the seed has been observed in other species (Baskin, Baskin, & Dixon 2006), but its importance for mesquite, and in modifying total seed predation rates more generally, has not previously been recognised. All evidence suggests that high predation rate of seeds in capsules is likely to be the result of the cumulative effect of oviposition rather than oviposition preferences, as similar predation rates have been observed in
588
R.D. van Klinken, A.J. White / Basic and Applied Ecology 15 (2014) 581–589
systems where mesquite pods remain intact for long periods (Impson et al. 1999; Baes, de Viana, & Saravia 2001) and in seeds in dung (van Klinken et al. 2009; van Klinken & White 2011). Seed densities varied with season and region. However, with the exception of fallen pods, we found no strong relationship between seed predation rates and seed density despite expectations that predators would be demographically constrained when seed densities are low (Lewis & Gripenberg 2008). The lack of relationships may in part result from the low seed densities observed in our study (averaging 10 to 99 seeds m−2 ). Our highest recorded levels were comparable to other studies under mesquite canopies in central Queensland (82 seeds m−2 ; B. Lynes unpublished data) and the native range (33 seeds m−2 ; (Brown & Archer 1988)) but considerably lower than recorded in Ethiopia (1932 m−2 ) (Shiferaw, Teketay, Nemomissa, & Assefa 2004). Nonetheless, total seed predation was low (<20%) in the Pilbara Region where seed densities have already been greatly reduced by an introduced leaf-feeding insect (van Klinken et al. 2009), thereby limiting additive or synergistic effects between multiple herbivores that may otherwise have been expected (Hoffmann & Moran 1998; Leimu & Lehtilä 2006). Low seed-densities may suggest that mesquite populations in Australia will be particularly sensitive to seed predation (Crawley 1992), but population-level impacts may nonetheless be modest. Total seed predation rates, even when considering the contribution of predation in all seed stages, was in the commonly-recorded range for seed predators (Kolb et al. 2007; Ramírez & Traveset 2010) and is therefore less than is generally considered necessary to significantly reduce population growth rates (Kolb et al. 2007; van Klinken et al. 2008). Furthermore, most seed predation estimates, including ours, are likely to be maximal estimates as they are generally made at seed sinks, such as under or immediately adjacent to parent trees. Other studies suggest strong reductions in seed predation rates with distance from seed sinks (Fragoso, Silvius, & Correa 2003), providing a mechanism for predator escape by well-dispersed plant species (Lewis & Gripenberg 2008) such as mesquite. However, seed predation may affect population dynamics by reducing seed bank persistence (Kauffman & Maron 2006). In our study it is the persistent seed bank, mainly seeds in capsules, that was most vulnerable to seed predation. Persistent seed banks can buffer plant populations from the effects of pre-dispersal seed predation (Briese 2000; Crawley 1992), especially of longlived species such as mesquite, and where the environment is highly variable between years. Our study demonstrates the need for considering all seed stages when estimating total seed predation, and when considering the population consequences of seed predation. Most studies focus on the readily accessible seed stages (Kolb et al. 2007; Ramírez & Traveset 2010), and therefore overlook potentially important cryptic predation, in our case in seed capsules. Furthermore, we demonstrate the importance of both seed dynamics and oviposition behaviour in determining
total seed predation levels. Different systems will have contrasting seed dynamics (Baes et al. 2001; Raghu et al. 2005) and oviposition preferences (Johnson 1981; van Klinken & Flack 2008). Total seed predation rates will therefore ultimately be determined by the interaction between host species, seed predator and environment (Kauffman & Maron 2006), but can be estimated through time by individually tracking seed density and predation rates for the major seed stages.
Acknowledgements We thank Rob Parr (DAFWA), Linda Anderson (PMMC) and Coby Seaborn (QDEEDI) for field assistance, LWA Defeating the Weed Menace initiative for funds, Saul Cunningham and Raghu Sathyamurthy for comments on draft manuscripts, Areli Mira and Kate Detchon for assistances in processing samples, and Anne Bourne for statistical advice.
References Baes, P. O., de Viana, M., & Saravia, M. a. (2001). The fate of Prosopis ferox seeds from unremoved pods at National Park Los Cardones. Journal of Arid Environments, 48, 185–190. Baskin, J. M., Baskin, C. C., & Dixon, K. W. (2006). Physical dormancy in the endemic Australian genus Stylobasium, a first report for the family Surianaceae (Fabales). Seed Science Research, 16, 229. Briese, D. (2000). Impact of the Onopordum capitulum weevil Larinus latus on seed production by its host-plant. Journal of Applied Ecology, 37, 238–246. Brown, J., & Archer, S. (1988). Woody plant seed dispersal and gap formation in a North American subtropical Savanna woodland: The role of domestic herbivores. Vegetation, 73, 73–80. Crawley, M. J. (1992). Seed predators and plant population dynamics. Seeds: The ecology of regeneration in plant communities. Wallingford: CABI Publishing. Fragoso, J. M., Silvius, K. M., & Correa, J. A. (2003). Long-distance seed dispersal by tapirs increases seed survival and aggregates tropical trees. Ecology, 84, 1998–2006. Hoffmann, J., Impson, F., & Moran, V. C. (1993). Competitive interactions between two bruchid species (Algarobius spp.) introduced in South Africa for biological control of mesquite weeds (Prosopis spp.). Biological Control, 3, 215–220. Hoffmann, J., & Moran, V. (1998). The population dynamics of an introduced tree, Sesbania punicea, in South Africa, in response to long-term damage caused by different combinations of three species of biological control agents. Oecologia, 114, 343–348. Impson, F., Moran, V., Hoffmann, J., Olckers, T., & Hill, M. (1999). A review of the effectiveness of seed-feeding bruchid beetles in the biological control of mesquite, Prosopis species (Fabaceae), in South Africa. In T. Olckers, & M. P. Martin (Eds.), Biological control of weeds in South Africa (1990–1998). African Entomology Memoir No. 1 (pp. 81–88). Hatfield, South Africa: The Entomological Society of Southern Africa. Johnson, C. D. (1981). Interactions between bruchid (Coleoptera) feeding guilds and behavioral patterns of pods of the Leguminosae. Environmental Entomology, 10, 249–253.
R.D. van Klinken, A.J. White / Basic and Applied Ecology 15 (2014) 581–589
Kauffman, M. J., & Maron, J. L. (2006). Consumers limit the abundance and dynamics of a perennial shrub with a seed bank. The American Naturalist, 168, 454–470. Kingsolver, J. (1986). A taxonomic study of the genus Algarobius (Coleoptera: Bruchidae). Entomography, 4, 109–136. Kolb, A., Ehrlén, J., & Eriksson, O. (2007). Ecological and evolutionary consequences of spatial and temporal variation in pre-dispersal seed predation. Perspectives in Plant Ecology, Evolution and Systematics, 9, 79–100. Leimu, R., & Lehtilä, K. (2006). Effects of two types of herbivores on the population dynamics of a perennial herb. Basic and Applied Ecology, 7, 224–235. Lewis, O. T., & Gripenberg, S. (2008). Insect seed predators and environmental change. Journal of Applied Ecology, 45, 1593–1599. Raghu, S., Wiltshire, C., & Dhileepan, K. (2005). Intensity of pre-dispersal seed predation in the invasive legume Leucaena leucocephala is limited by the duration of pod retention. Australasian Ecology, 30, 310–318. Ramírez, N., & Traveset, A. (2010). Predispersal seed-predation by insects in the Venezuelan Central Plain: Overall patterns and traits that influence its biology and taxonomic groups. Perspectives in Plant Ecology, Evolution and Systematics, 12, 193–209.
589
Shiferaw, H., Teketay, D., Nemomissa, S., & Assefa, F. (2004). Some biological characteristics that foster the invasion of (Prosopis juliflora) (Sw.) DC. at Middle Awash Rift Valley area, north-eastern Ethiopia. Journal of Arid Environments, 58, 135–154. van Klinken, R. D. (2005). Total annual seed loss on a perennial legume through predation by insects: The importance of withinseason seed and seed-feeder dynamics. Australasian Ecology, 30, 414–425. van Klinken, R. D., & Flack, L. K. (2008). What limits predation rates by the specialist seed-feeder Penthobruchus germaini on an invasive shrub? Journal of Applied Ecology, 45, 1600–1611. van Klinken, R. D., & Campbell, S. (2009). Prosopis species. Australian weeds series. Melbourne, Richardson: RG & Richardson, FJ. van Klinken, R. D., Hoffmann, J., Zimmermann, H., & Roberts, A. (2009). Prosopis species (Leguminosae). In R. Muniappan, G. Reddy, & A. Raman (Eds.), Biological control of tropical weeds using arthropods (pp. 353–377). Cambridge, UK: Cambridge University Press. van Klinken, R. D., & White, A. (2011). Overcoming seasonally fluctuating resources: Bruchid predation of mesquite (Prosopis) seed in dung. Biological Control, 59, 361–365.
Available online at www.sciencedirect.com
ScienceDirect
ORIGINAL PAPERS
The role of pre- and post-dispersal seed predation in determining total seed loss R.D. van Klinken∗ , A.J. White CSIRO Biosecurity Flagship, EcoSciences Precinct, 41 Boggo Road, Dutton Park, PO Box 2583, Brisbane 4001, QLD, Australia Received 17 December 2013; received in revised form 26 August 2014; accepted 26 August 2014 Available online 4 September 2014
Abstract Most seed predation studies focus on either pre- or post-dispersal predation and may therefore underestimate the role of predation in regulating plant populations. We therefore estimated total seed predation of an invasive tree, mesquite (Leguminoseae: Prosopis spp.), by examining the entire seed pool from tree to seed bank. The spatio-temporal dynamics of total seed predation was examined by sampling across its Australian distribution and through time. The main predator was a host-specialist multivoltine beetle, Algarobius prosopis L. (Bruchidae), previously introduced as a biocontrol agent. Seed predation exceeded 20% in all seed stages (in pods on and off the tree, and seeds within woody endocarps (capsules) and free seeds on and in the ground) but was consistently highest in capsules on the ground (up to 90%). Pre-dispersal predation contributed little. Total seed predation rates were primarily determined by predation rates on the most persistent seed stage, in this case fallen pods if only pods are considered and seeds in capsules for the total seed pool. This pattern was consistent across the surveyed taxa, regions, years and seasonally. Predation rate was relatively unaffected by seed density, potentially because densities were always low (<150 seeds m−2 ). Average total seed predation within a region reached 55%, but we conclude that any population regulation of mesquite by seed predation will principally be through reduced seed bank persistence. Our results highlight the need to consider the entire seed pool, especially the often cryptic and overlooked long-lived stages, when determining seed loss to predation and its likely population consequences.
Zusammenfassung Die meisten Untersuchungen zu Samenprädation konzentrieren sich entweder auf Prädation vor oder auf die Prädation nach der Ausbreitung und unterschätzen deshalb möglicherweise die Bedeutung der Prädation bei der Regulation von Pflanzenpopulationen. Wir bestimmten deshalb die Gesamt-Samenprädation des invasiven Mesquitebaums (Leguminoseae: Prosopis spp.), indem wir den gesamten Samenpool vom Baum bis zur Samenbank erforschten. Die Raum-Zeit-Dynamik der GesamtSamenprädation wurde untersucht, indem wir Proben über sein australisches Verbreitungsgebiet und vier Jahre hinweg nahmen. Der Hauptprädator war ein spezialisierter, plurivoltiner Käfer, Algarobius prosopis L. (Bruchidae), der früher als biologischer Kontrollorganismus eingeführt worden war. Die Samenprädation übertraf 20% in allen Samenstadien (in den Hülsen auf dem Baum und nach dem Herunterfallen, Samen innerhalb des holzigen Endocarps (Kapseln) und freie Samen auf oder im Boden), war aber immer am höchsten in den Kapseln auf dem Boden (bis 90%). Die Prädation vor der Ausbreitung war gering. Die Gesamt-Samenprädationsraten wurden hauptsächlich bestimmt durch die Prädationsraten am langlebigsten Samenstadium, in
∗ Corresponding
author. Tel.: +61 7 38335737; fax: +61 7 3833 5504. E-mail address: [email protected] (R.D. van Klinken).
http://dx.doi.org/10.1016/j.baae.2014.08.012 1439-1791/Crown Copyright © 2014 Published by Elsevier GmbH. All rights reserved.
582
R.D. van Klinken, A.J. White / Basic and Applied Ecology 15 (2014) 581–589
diesem Fall herabgefallene Hülsen, wenn nur Hülsen betrachtet werden, bzw. den Samen in Kapseln für den gesamten Samenpool. Dies Muster war gültig für alle untersuchten Arten, Regionen, Jahre und Jahreszeiten. Die Prädationsrate blieb relativ unbeeinflusst durch die Samendichte, vermutlich weil die Dichten immer gering waren (<150 Samen m−2 ). Die durchschnittliche Samenprädation innerhalb einer Region erreichte 55%, aber wir schließen, dass jedwede Populationsregulation des Mesquitebaumes durch Samenprädation prinzipiell durch reduzierte Langlebigkeit der Samenbank erfolgt. Unsere Ergebnisse belegen die Notwendigkeit, den gesamten Samenpool zu betrachten, insbesondere die oft kryptischen und übersehenen langlebigen Stadien, wenn Samenverluste durch Prädation und ihre Konsequenzen für die Populationsdynamik bestimmt werden sollen. Crown Copyright © 2014 Published by Elsevier GmbH. All rights reserved.
Keywords: Bruchidae; Invasive plants; Plant–insect interactions; Predator escape; Pre-dispersal seed predation; Post-dispersal seed predation; Seed bank
Introduction Seed predation can be an important driver of plant population dynamics, and even community structuring (Crawley 1992; Hoffmann & Moran 1998; Kauffman & Maron 2006; Kolb, Ehrlén, & Eriksson 2007; Lewis & Gripenberg 2008; Ramírez & Traveset 2010). Seed predating insects are particularly diverse and widely studied, but their general importance is still disputed. Most studies of insect seed predators have focussed on seeds that are still on the plant (pre-dispersal predation) (Kolb et al. 2007). Univoltine insects, that pass through a single generation a year, can be well synchronised with typically observed seasonal peaks in seed production (Crawley 1992). In contrast, multivoltine insects, that pass through multiple generations a year, are less able to track sharp seasonal fluctuations in pre-dispersal host resources (Raghu, Wiltshire, & Dhileepan 2005; van Klinken et al. 2008). However, this can potentially be overcome if they continue exploiting more persistent post-dispersal seed stages (Johnson 1981; Lewis & Gripenberg 2008; van Klinken & Flack 2008; van Klinken & White 2011). Post-dispersal seeds are often cryptic and therefore overlooked in seed predation studies. As a consequence the role of multivoltine insects as a seed mortality factor, and ultimately in regulating host populations, may be underappreciated (Briese, 2000; Lewis & Gripenberg 2008). Several closely-related Prosopis species (Leguminosae), together referred to as mesquite, are long-lived shrubs and trees that reproduce entirely by seed and are native to arid and semi-arid Americas. They have also been intentionally introduced around the world where they often become serious weeds (van Klinken, Hoffmann, Zimmermann, & Roberts 2009). Pods usually contain 10–20 seeds (c 0.024 g each, unpublished data) and remain closed at maturity (van Klinken & Campbell 2009). Mesquite seeds offer a particularly dynamic resource for seed predators. The pre-dispersal phase (Fig. 1) is typically brief, with pod maturation being relatively synchronised, and pods falling from the tree soon after maturation (van Klinken & Campbell 2009). Most pods are often consumed, either from the tree or ground, within days of maturation by a wide range of vertebrate herbivores. These herbivores are not seed predators, rather they derive
nutrition from the pods and act as important seed dispersal vectors. Seeds are contained within a thin shell or endocarp (the seed capsule) which can remain intact even after the fleshy mesocarp has broken down during passage through the herbivore or through natural decay (van Klinken & White 2011) (Fig. 1). Mesquite seeds have hard-seeded (physical) dormancy imposed by the hard seed coat, from which most are released within 2–3 years under natural conditions (van Klinken et al. 2009). Most seed predation reported on mesquite in its introduced range is by a multivoltine host-specialist, Algorobius prosopis Le Conte. (Bruchidae), sourced from the Americas in an attempt to regulate mesquite populations (van Klinken et al. 2009). Females lay eggs on mature pods, seed capsules and possibly free seeds, with each larva consuming and killing an individual seed before emerging as an adult (Impson, Moran, Hoffmann, Olckers, & Hill 1999; van Klinken and White 2011). Populations can pass through several generations a year, with development from egg hatch to eclosion taking c 33 days (Hoffmann, Impson, & Moran 1993). High seed predation levels can occur on long-lived seed stages. In South Africa it can kill over 90% of seeds in mature pods within 10–12 months provided vertebrate herbivores are prevented from consuming pods (Impson et al. 1999; van Klinken et al. 2009). In Australia predation rates by A. prosopis can be high
Fig. 1. A schematic of the mesquite seed lifecycle with the typically transient pod stages highlighted. It includes pre-dispersal seeds (tree-pods) and post-dispersal seed stages on and in the ground.
R.D. van Klinken, A.J. White / Basic and Applied Ecology 15 (2014) 581–589
for long-lived seed pools in cattle dung (van Klinken & White 2011). However, no study has systematically considered the role of seed predation across the entire seed pool. Total seed loss to predation will be influenced by both oviposition behaviour and the longevity of each seed stage (Fig. 1; van Klinken, 2005). Oviposition on long-lived stages is therefore expected to have the greatest effect on total predation rates, provided oviposition preferences are similar across stages. Most seed predation studies focus on predation in pods as pods are typically the most conspicuous seed stage for researchers. We compared pre- and post-dispersal seed predation in pods through time, with the expectation that total predation rates in pods will increase through the year provided oviposition continues in fallen pods. We also sampled all seed stages between fruiting seasons (winter), including in the seed bank, to test the relative importance of seed predation in each seed stage in determining total seed loss to predation. Our expectation was that pod will be an ephemeral resource and therefore that pod predation will be limited, and that total seed predation rates in the winter seed pool will depend largely on continued oviposition on cryptic seed stages. Furthermore, we tested the versatility of these predictions through field surveys conducted over several years across the core Australian distribution of mesquite and the best-represented mesquite taxa (van Klinken & Campbell 2009). We also consider the implications of our results for the regulation of mesquite populations, including testing whether seed predation rates were correlated with seed density.
583
Klinken et al. 2009). Sites were in the summer-rainfall zone, with hot summers and cool to warm winters. The Pilbara is the driest and hottest region, and western Queensland (WQ) is slightly drier and hotter than central Queensland (CQ) (Table 1). All properties were used for cattle-grazing. Pigs were common at one site in CQ, and wild emus (Dromaius novaehollandiae (Latham)) and macropods (Macropodidae) were present at most sites, although they were most common in the Pilbara. The sampling design within a region was largely dictated by mesquite distribution (Table 1). In CQ dense Prosopis pallida extends across relatively small pastoral blocks around the town of Hughenden. Sampling was conducted on two properties located 8 km apart. In WQ P. pallida is widely distributed and sampling was conducted on three properties on which it reaches high densities. Sampling of hybrid mesquite in WQ was restricted to a single property as it has been largely managed elsewhere in WQ. In the Pilbara sampling was conducted on the property on which mesquite principally occurs (van Klinken et al. 2009). Where possible multiple sites were sampled on each property (Table 1). In CQ and WQ sampling was conducted in dense mesquite (>90% canopy cover) and under mesquite trees that had non-overlapping canopies. However, results were subsequently combined to obtain sitelevel estimates of seed density and seed predation as seed densities were low and no plant density effects were detected (data not shown).
Seed density and seed predation estimates
Materials and methods Sites and surveys Surveys were conducted in three regions across Australia where mesquite is most widespread and abundant (Table 1). Elsewhere mesquite populations were sparse or under active management during our study period (2005–2008) (van
Surveys of pods were conducted to test the relative importance of pre-dispersal (tree-pods) and post-dispersal (fallen pods) predation for total predation in pods through and between years. This was supplemented by sampling of all seed stages (Fig. 1) in winter to determine total predation across the entire seed pool. Winter is when the seed bank is expected to be most stable, as seed production and germination events are limited then (van Klinken et al. 2009).
Table 1. Details of the surveyed regions and pastoral properties. Region and properties
Location ◦
*
Species
◦
Queensland: Central (CQ) RF: 492.4 mm Tmax : 31.6 C Tmin : 16.6 C Alberts (2 sites; 505 ha) S20◦ 49 E144◦ 11 Lands block (2 sites; 65 ha) S20◦ 49 E144◦ 13
P. pallida P. pallida
Queensland: Western (WQ) RF: 469.2 mm Tmax : 32.4 ◦ C Tmin : 18.4 ◦ C Hampden downs (2 sites; 29,100 ha) S21◦ 28 Beaudesert lane (1 site; 49,200 ha) S21◦ 31 Williams (2 sites; 99,000 ha) S20◦ 36 Colwell (2 sites; 25,500 ha) S21◦ 27
E141◦ 40 E141◦ 05 E141◦ 58 E141◦ 09
P. pallida P. pallida P. pallida Hybrid mesquite
Western Australia: Pilbara Region RF: 270.9 mm Tmax : 33.9 ◦ C Tmin : 18.7 ◦ C Mardie station (5 sites; 220,900 ha) S21◦ 09 E116◦ 04
Hybrid mesquite
* Mesquite populations were either Prosopis pallida (Willd.) Kunth or hybrids (between P. glandulosa Torr., P. velutina Wooton and/or P. pallida) (van Klinken & Campbell 2009).
584
R.D. van Klinken, A.J. White / Basic and Applied Ecology 15 (2014) 581–589
Pods Pod surveys were timed to coincide with when pods were expected to be falling (early summer), to have fallen (mid to late summer) and when cumulative seed predation rates were expected to be highest (in winter). In each survey a maximum of 20 intact pods were collected from on or under individual trees to obtain a pooled sample of at least 100 pods on the tree or ground, respectively, from each site. At the same time, pod–seed density was estimated to calculate total seed predation rates (see below). Pods for bruchid emergence were returned to the laboratory within three days of collection, held in ventilated containers at 25 ◦ C and any emerging insects were removed every 1–2 days to prevent reinfestation and kept for identification. Pods were frozen after all adults had emerged (typically within 40–50 days) to kill any second generation insects. Emergent holes and seeds were subsequently counted to determine total seed predation rates for each sample. Algarobius beetles were identified to species (Kingsolver 1986). Pre-dispersal (tree) pod density was estimated by counting the number of pods on each of 10 permanently labelled large, healthy adult trees at each site and converting counts to density by dividing it by canopy area. Post-dispersal (fallen) pod density on the ground was estimated using 10 (625 cm−2 ) quadrats randomly placed under mesquite canopies at each site. Pods from quadrats were subsequently combined with pooled pod samples for bruchid emergence. Pod counts needed to be converted to seed counts to calculate seed densities and seed predation rates. The number of fully developed seeds per pod was estimated in two ways. For visual pod counts on trees the number of pods was multiplied by the average number of seeds per pod calculated from a random subsample (CQ P. pallida: n = 1631 pods, 15.4 seeds pod−1 ; WQ P. pallida: n = 134, 15.1; WQ hybrid: n = 801, 13.3; Pilbara hybrid: n = 820, 11.4). Seed numbers for collected pods were estimated by counting the external swellings of mature pods and pod fragments, and then adjusting this with the average number of fully developed seeds per swelling determined through dissection of random pods (CQ P. pallida: n = 43 pods; 94.2% of swellings had fullydeveloped seeds; WQ P. pallida; n = 20, 95.7%; WQ hybrid: 30, 89.2%; Pilbara hybrid: 26, 86.2%). Total seed pool Winter sampling of pods (2005–2008) at each site was supplemented with soil coring to estimate seed bank densities and seed predation rates for seeds located both on and in the soil (Fig. 1). At each site a single soil core (625 cm−2 wide × 5 cm deep) was taken from under the canopy of each of 10 isolated large mature adults (midway between trunk and edge of canopy) and 10 soil cores were randomly taken along a 20 to 40 m transect under dense mesquite stands. Surface litter and soil were kept and processed separately for each soil core. All free seeds, seeds in capsules and pods (whole and fragments) were removed from samples through sieving and visual inspection, then frozen for processing.
Seed densities and seed predation rates (the proportion of seeds with adult emergence holes at the time of freezing) were subsequently calculated for each seed stage (Fig. 1) at each site (by averaging soil cores).
Data analysis Seed density and seed predation rates for each site-visit were calculated for each individual seed stage (Fig. 1) and for pooled stages: all pod–seeds, all seeds not in pods (capsules and free seeds), and the total seed pool. Total predation rates for pooled stages were estimated by summing the proportional contribution of seed predation of each seed stage (van Klinken, 2005). Seed density and seed predation rates for each seed stage, and for pooled stages, were analysed following appropriate transformations to satisfy ANOVA conditions. Significant differences were determined on the transformed scale. Analyses included region–species combination (Table 1) as a factor, and year where sufficient data was available (CQ only). Seed predation rates were weighted by the number of seeds collected in the ANOVA. Correlation coefficients were calculated between seed density and seed predation rates (calculated for each site and survey) and tested for significance.
Results Pods Overall predation rates in pods were 9.8% (n = 16,953 seeds) for pre-dispersal seeds (tree-pods) and 18.6% (n = 39,958) for post-dispersal seeds (fallen pods), and was highest in CQ and the Pilbara (Fig. 2A). Most emergents (1053) following field collection were A. prosopis beetles, which were present at all sites. In addition, eleven unidentified bruchid beetles emerged from a single sample, and 10 larval or pupal wasp parasitoids emerged across all samples. Average seed predation was similar on tree and fallen pods in CQ and the Pilbara, but higher in fallen pods in WQ (Fig. 2A). Average pod–seed densities were highest on the ground (Fig. 2B). Together this meant that pre-dispersal predation contributed little to total predation, except in the Pilbara (Fig. 2C) where fallen pod density was high on one occasion and zero on the remaining seven. Seed predation and seed density was not correlated for tree-pods (correlation coefficient = −0.13, P = 0.52, n = 28), but seed density was always low (Fig. 2B). In contrast, predation in fallen pods was positively correlated with seed density on the ground (correlation coefficient = 0.42, P < 0.001, n = 60). CQ was the only region where there were sufficient seeds to examine the effect of season and year on predation rates. Again, predation of pre-dispersal seeds (Fig. 3A) generally contributed little to total predation (Fig. 3C). Pod–seed densities on the tree were consistently low (Fig. 3B), as most pods had matured and fallen by the early-summer surveys
R.D. van Klinken, A.J. White / Basic and Applied Ecology 15 (2014) 581–589
585
Fig. 2. Average (±SE) predation rate (A) and seed density (B) for pre-dispersal (tree) pods and post-dispersal (fallen) pods, and (C) the relative contribution of pre-dispersal predation to total seed predation. Different letters indicate significant difference (LSD test, P < 0.05) within each significant dependent variable.
(unpublished data from litter traps). Total predation rates in pods did vary with time, but there was no evidence that it increased within a season (Fig. 3A). Average predation rates in tree-pods were higher in 2005–2006 (19%) than in 2006–2007 (4%) (P < 0.05), so were inversely proportional to seed densities in the tree (Fig. 3B), but there was no density dependent relationship overall (correlation coefficient = 0.0107, P = 0.38, n = 23). Seed predation levels on fallen pods ranged from 8 to 40% (Fig. 3A), but there was no year or season effect. There was also no correlation between seed densities in fallen pods and seed predation (correlation
coefficient = 0.36, P = 0.16, n = 23), unlike in the regional analysis (see above). Pod densities were generally too low in winter to reliably estimate predation rates (Fig. 3A).
Total seed pool Total predation rate (estimated in winter) was highest in CQ (averaging 54.9%), compared to 20.7 to 25.5% elsewhere (Fig. 4A). Predation in pod–seeds contributed relatively little to total predation rates (Fig. 4C) owing to
586
R.D. van Klinken, A.J. White / Basic and Applied Ecology 15 (2014) 581–589
Fig. 3. Seasonal changes in average (±SE) seed predation rates (A), seed density (B) for pre-dispersal (tree) pods and post-dispersal (fallen) pods in CQ, and (C) the relative contribution of pre-dispersal predation to total seed predation. Asterisk indicates pods were too rare to collect.
lower or similar seed densities (Fig. 4B) and consistently lower predation rates (Fig. 4A). Total seed density was highest in CQ and WQ (hybrid), and was barely detectable using our sampling method at the remaining two locations (Fig. 4B). Total seed predation was not correlated with total seed density (correlation coefficient = 0.14, P = 0.42, n = 36). A between-year comparison of seed predation across all seed stages was only possible for CQ where seed numbers were highest. Total seed predation rates ranged from 36 to 72% (Fig. 5A). Predation of seeds in capsules on the ground contributed most to total seed predation (Fig. 5C), owing to consistently high seed predation rates (Fig. 5A) and its contribution to the total seed pool (Fig. 5B). This contrasted with seeds in pods located on both the tree and ground for which average seed predation rates within a year never exceeded 20% (Fig. 5A), and capsules in the
ground which had high predation rates (Fig. 5A) but relatively low densities (Fig. 5B). The contribution of predation in the pre-dispersal seed fraction on total seed predation was insignificant (Fig. 5C).
Discussion In Australia mesquite presents a challenging host for seed predators as seeds were often at low densities and represented a dynamic resource. Seed predation was almost entirely by the biological control agent A. prosopis, which was present at all sites. This species is a specialist on mesquite species, but a generalist with regards to seed stage (Kingsolver 1986; Impson et al. 1999; van Klinken et al. 2009). Patterns in seed predation rates suggest oviposition occurs on the tree,
R.D. van Klinken, A.J. White / Basic and Applied Ecology 15 (2014) 581–589
587
Fig. 5. Average (mean ± SE) seed predation rates (A) and seed density (B) for all seeds combined and for individual seed stages across years in CQ, and (C) the relative contribution of predation in each seed stage to total seed predation.
Fig. 4. Average (±SE) between-year (winter) predation rate (A) and seed density (B), for seeds in pods, seeds no longer in pods and the total seed pool, and the relative contribution of predation in pod–seeds to total seed predation (C). Different letters indicate significant difference (LSD test, P < 0.05) within each significant dependent variable.
fallen pods, seed capsules and possibly free seeds, and it also oviposits on seeds in dung (van Klinken & White 2011). Such wide oviposition preferences are likely to provide a useful mechanism for maintaining seed predator populations when seed resources vary in abundance and quality through the year (van Klinken & Flack 2008). As predicted for multivoltine insects, total predation rates were primarily determined by predation rates on the most persistent seed stage. For total predation it was seeds on the ground that were no longer in pods, which are rarely considered in seed predation studies.
These patterns held irrespective of season, year, region or mesquite taxon. Total seed predation ranged between 20 and 55%. Predation was relatively high at times for each individual seed stage (Fig. 1), exceeding 40% even on the most transient pre-dispersal stage (tree pods). However, in our study predispersal predation of seeds on the tree had little influence on total seed predation levels, and was a poor predictor of predation levels on the ground, despite being the primary focus of many seed predation studies (Kolb et al. 2007). Rather, in our study it was the long-lived, relatively cryptic, seed stages on which oviposition continued that was principally responsible for total seed predation. Predation was consistently highest for seeds in capsules (averaging 80% in Central Queensland), and seeds in capsules were generally also the most common seed stage between seasons. The role of the pericarp in protecting the seed has been observed in other species (Baskin, Baskin, & Dixon 2006), but its importance for mesquite, and in modifying total seed predation rates more generally, has not previously been recognised. All evidence suggests that high predation rate of seeds in capsules is likely to be the result of the cumulative effect of oviposition rather than oviposition preferences, as similar predation rates have been observed in
588
R.D. van Klinken, A.J. White / Basic and Applied Ecology 15 (2014) 581–589
systems where mesquite pods remain intact for long periods (Impson et al. 1999; Baes, de Viana, & Saravia 2001) and in seeds in dung (van Klinken et al. 2009; van Klinken & White 2011). Seed densities varied with season and region. However, with the exception of fallen pods, we found no strong relationship between seed predation rates and seed density despite expectations that predators would be demographically constrained when seed densities are low (Lewis & Gripenberg 2008). The lack of relationships may in part result from the low seed densities observed in our study (averaging 10 to 99 seeds m−2 ). Our highest recorded levels were comparable to other studies under mesquite canopies in central Queensland (82 seeds m−2 ; B. Lynes unpublished data) and the native range (33 seeds m−2 ; (Brown & Archer 1988)) but considerably lower than recorded in Ethiopia (1932 m−2 ) (Shiferaw, Teketay, Nemomissa, & Assefa 2004). Nonetheless, total seed predation was low (<20%) in the Pilbara Region where seed densities have already been greatly reduced by an introduced leaf-feeding insect (van Klinken et al. 2009), thereby limiting additive or synergistic effects between multiple herbivores that may otherwise have been expected (Hoffmann & Moran 1998; Leimu & Lehtilä 2006). Low seed-densities may suggest that mesquite populations in Australia will be particularly sensitive to seed predation (Crawley 1992), but population-level impacts may nonetheless be modest. Total seed predation rates, even when considering the contribution of predation in all seed stages, was in the commonly-recorded range for seed predators (Kolb et al. 2007; Ramírez & Traveset 2010) and is therefore less than is generally considered necessary to significantly reduce population growth rates (Kolb et al. 2007; van Klinken et al. 2008). Furthermore, most seed predation estimates, including ours, are likely to be maximal estimates as they are generally made at seed sinks, such as under or immediately adjacent to parent trees. Other studies suggest strong reductions in seed predation rates with distance from seed sinks (Fragoso, Silvius, & Correa 2003), providing a mechanism for predator escape by well-dispersed plant species (Lewis & Gripenberg 2008) such as mesquite. However, seed predation may affect population dynamics by reducing seed bank persistence (Kauffman & Maron 2006). In our study it is the persistent seed bank, mainly seeds in capsules, that was most vulnerable to seed predation. Persistent seed banks can buffer plant populations from the effects of pre-dispersal seed predation (Briese 2000; Crawley 1992), especially of longlived species such as mesquite, and where the environment is highly variable between years. Our study demonstrates the need for considering all seed stages when estimating total seed predation, and when considering the population consequences of seed predation. Most studies focus on the readily accessible seed stages (Kolb et al. 2007; Ramírez & Traveset 2010), and therefore overlook potentially important cryptic predation, in our case in seed capsules. Furthermore, we demonstrate the importance of both seed dynamics and oviposition behaviour in determining
total seed predation levels. Different systems will have contrasting seed dynamics (Baes et al. 2001; Raghu et al. 2005) and oviposition preferences (Johnson 1981; van Klinken & Flack 2008). Total seed predation rates will therefore ultimately be determined by the interaction between host species, seed predator and environment (Kauffman & Maron 2006), but can be estimated through time by individually tracking seed density and predation rates for the major seed stages.
Acknowledgements We thank Rob Parr (DAFWA), Linda Anderson (PMMC) and Coby Seaborn (QDEEDI) for field assistance, LWA Defeating the Weed Menace initiative for funds, Saul Cunningham and Raghu Sathyamurthy for comments on draft manuscripts, Areli Mira and Kate Detchon for assistances in processing samples, and Anne Bourne for statistical advice.
References Baes, P. O., de Viana, M., & Saravia, M. a. (2001). The fate of Prosopis ferox seeds from unremoved pods at National Park Los Cardones. Journal of Arid Environments, 48, 185–190. Baskin, J. M., Baskin, C. C., & Dixon, K. W. (2006). Physical dormancy in the endemic Australian genus Stylobasium, a first report for the family Surianaceae (Fabales). Seed Science Research, 16, 229. Briese, D. (2000). Impact of the Onopordum capitulum weevil Larinus latus on seed production by its host-plant. Journal of Applied Ecology, 37, 238–246. Brown, J., & Archer, S. (1988). Woody plant seed dispersal and gap formation in a North American subtropical Savanna woodland: The role of domestic herbivores. Vegetation, 73, 73–80. Crawley, M. J. (1992). Seed predators and plant population dynamics. Seeds: The ecology of regeneration in plant communities. Wallingford: CABI Publishing. Fragoso, J. M., Silvius, K. M., & Correa, J. A. (2003). Long-distance seed dispersal by tapirs increases seed survival and aggregates tropical trees. Ecology, 84, 1998–2006. Hoffmann, J., Impson, F., & Moran, V. C. (1993). Competitive interactions between two bruchid species (Algarobius spp.) introduced in South Africa for biological control of mesquite weeds (Prosopis spp.). Biological Control, 3, 215–220. Hoffmann, J., & Moran, V. (1998). The population dynamics of an introduced tree, Sesbania punicea, in South Africa, in response to long-term damage caused by different combinations of three species of biological control agents. Oecologia, 114, 343–348. Impson, F., Moran, V., Hoffmann, J., Olckers, T., & Hill, M. (1999). A review of the effectiveness of seed-feeding bruchid beetles in the biological control of mesquite, Prosopis species (Fabaceae), in South Africa. In T. Olckers, & M. P. Martin (Eds.), Biological control of weeds in South Africa (1990–1998). African Entomology Memoir No. 1 (pp. 81–88). Hatfield, South Africa: The Entomological Society of Southern Africa. Johnson, C. D. (1981). Interactions between bruchid (Coleoptera) feeding guilds and behavioral patterns of pods of the Leguminosae. Environmental Entomology, 10, 249–253.
R.D. van Klinken, A.J. White / Basic and Applied Ecology 15 (2014) 581–589
Kauffman, M. J., & Maron, J. L. (2006). Consumers limit the abundance and dynamics of a perennial shrub with a seed bank. The American Naturalist, 168, 454–470. Kingsolver, J. (1986). A taxonomic study of the genus Algarobius (Coleoptera: Bruchidae). Entomography, 4, 109–136. Kolb, A., Ehrlén, J., & Eriksson, O. (2007). Ecological and evolutionary consequences of spatial and temporal variation in pre-dispersal seed predation. Perspectives in Plant Ecology, Evolution and Systematics, 9, 79–100. Leimu, R., & Lehtilä, K. (2006). Effects of two types of herbivores on the population dynamics of a perennial herb. Basic and Applied Ecology, 7, 224–235. Lewis, O. T., & Gripenberg, S. (2008). Insect seed predators and environmental change. Journal of Applied Ecology, 45, 1593–1599. Raghu, S., Wiltshire, C., & Dhileepan, K. (2005). Intensity of pre-dispersal seed predation in the invasive legume Leucaena leucocephala is limited by the duration of pod retention. Australasian Ecology, 30, 310–318. Ramírez, N., & Traveset, A. (2010). Predispersal seed-predation by insects in the Venezuelan Central Plain: Overall patterns and traits that influence its biology and taxonomic groups. Perspectives in Plant Ecology, Evolution and Systematics, 12, 193–209.
589
Shiferaw, H., Teketay, D., Nemomissa, S., & Assefa, F. (2004). Some biological characteristics that foster the invasion of (Prosopis juliflora) (Sw.) DC. at Middle Awash Rift Valley area, north-eastern Ethiopia. Journal of Arid Environments, 58, 135–154. van Klinken, R. D. (2005). Total annual seed loss on a perennial legume through predation by insects: The importance of withinseason seed and seed-feeder dynamics. Australasian Ecology, 30, 414–425. van Klinken, R. D., & Flack, L. K. (2008). What limits predation rates by the specialist seed-feeder Penthobruchus germaini on an invasive shrub? Journal of Applied Ecology, 45, 1600–1611. van Klinken, R. D., & Campbell, S. (2009). Prosopis species. Australian weeds series. Melbourne, Richardson: RG & Richardson, FJ. van Klinken, R. D., Hoffmann, J., Zimmermann, H., & Roberts, A. (2009). Prosopis species (Leguminosae). In R. Muniappan, G. Reddy, & A. Raman (Eds.), Biological control of tropical weeds using arthropods (pp. 353–377). Cambridge, UK: Cambridge University Press. van Klinken, R. D., & White, A. (2011). Overcoming seasonally fluctuating resources: Bruchid predation of mesquite (Prosopis) seed in dung. Biological Control, 59, 361–365.
Available online at www.sciencedirect.com
ScienceDirect