Pollen harvest by Apis mellifera L . ( Hymenoptera : Apidae ) in the Dourados region , Mato Grosso do Sul state ( Brazil )

ABSTRACT – (Pollen harvest by Apis mellifera L. (Hymenoptera: Apidae) in the Dourados region, Mato Grosso do Sul state (Brazil)). We present data on the pollen harvest by Apis mellifera L. at a central-western Brazil site not yet studied. Corbiculae pollen loads were collected during one year in front of the hive, acetolysed and slides mounted for optical microscopy. Identiﬁ cation followed comparisons with a local pollen collection. Forty-two pollen types were utilized by the bee; the most important families were Myrtaceae, Asteraceae, Euphorbiaceae, Brassicaceae and Poaceae. The genus/species level ranking of relative importance showed Eucalyptus (19%), Raphanus raphanistrum (13%), Poaceae type 2 (7.5%), Jatropha cf. gossypiifolia (7%) and Sapium glandulatum (6.5%). Overall, the majority of pollen types (80%) were not abundantly harvested, and a minority (4%) reached values of 10%. Given that the study area has cultivated plants and not natural vegetation, we indicate most of the identiﬁ ed species as manageable and point out that weedy vegetation can be an important food source for bees since it frequently occurred in the pollen spectra even within a totally cultivated area.


Introduction
Pollen is an essential food element for bees as it is their main protein source (Michener 2000).Flower use by visitors may include the collection of both pollen and nectar, which are some of the ways plants have to guarantee reproduction (Barth 1991;Proctor et al. 1996).This insectplant relationship may go beyond what can be expected and show many details, especially when trophic niches are studied.Every bee has its own behavior.Apis mellifera L. is a generalist collector that effectively seeks out fl owering plants, instead of using a few sources (Roubik 1989;Marques-Souza et al. 1993;Carvalho et al. 1999).
The honey bee (Apis mellifera) was introduced in Brazil in 1839 by the priest Antônio Carneiro (Camargo 1972) and since then has been potentially competitive with native bees of Neotropical habitats (Ramalho et al. 1990), mainly after hybridization with introduced African species.Abiotic features can infl uence the collection and data on the autoecology of bees since collecting behavior, fl ight activity, distance from the source, abundance and fl oral preference are important to understand its trophic niche (see Roubik 1989;Wilms et al. 1996;Wilms & Wiechers, 1997).
Conservation of pollinators has largely been discussed in the Brazilian Pollinators Initiative due mainly to the great concern on how important pollinators are for food production worldwide (Imperatriz-Fonseca et al. 2004).Despite this, studies on fl ower sources utilized by visitors are scanty in the region where this work was carried out: in a review of melissopalinology surveys made in Brazil, Barth (2004) listed no data for the Central-West region.Actually Pott & Pott (1986) have studied the bee fl ora of Pantanal and Manente-Balestieri (2001) the pollen and nectar collected by bees, including Apis mellifera, in Corumbá, both in Mato Grosso do Sul state.These results were not included in Barth's (2004) review because they were not published in scientifi c periodicals.More recently, some new results have been published by Soares et al. (2007) for Mato Grosso state and Polatto & Alves (2008) and Vieira et al. (2008) for Mato Grosso do Sul state.Given the lack of data, it is more than evident the need for developing such studies in the region, not only for conservation purposes but also to make scientifi c knowledge available for Apiculture and Agriculture.The aim of this survey was to identify the bee fl ora utilized by Apis mellifera for the collection of pollen.

Material and methods
The study was carried out in the state of Mato Grosso do Sul, region of Dourados municipality (22º 14' S and 54º 49' W, 452 m altitude), on the campus of the Universidade Federal da Grande Dourados (UFGD).Regional climate is altitudinal tropical with two seasons, rainy from October to June (mean annual rainfall from 1500 to 1700mm) and a sharp dry season from July to September.The average temperature ranges from 19 to 24 ºC (Mato Grosso do Sul 1990).
Despite the fact that the local vegetation was originally semi-deciduous seasonal forest there is no native forested area on the campus but a few individual trees remain.Rather, there are plenty of monoculture plantations, for instance soya, corn, sunfl ower and others, an orchard, a medicinal plants garden, horticultural fi elds and fl ower gardens.
From August 2005 to July 2006 corbicular pollen loads were sampled from worker bees that were returning from the fi eld into the hive.This was done daily between 10:00 am and 12:00 pm and then samples were subjected to the method of acetolysis (Erdtman 1960), which makes grains hollow and thus more appropriate to study detail.Every single day of collection is a subsample within a month, which made counting grains in the slides unnecessary, as is usually done in surveys where pollen is sampled in pollen storage pots.Measurements of pollen grains followed Erdtman (1952) so as any correlation between grain size and frequency of the collected species could be tested.Plant species were identifi ed according to comparisons of bee pollen slides to the pollen database of the Laboratório de Abelhas Nativas (LAN) of UFGD.
Meteorological data were obtained from the university's meteorological station on the campus and distance to some pollen sources utilized by bees from the hive could easily be calculated as the survey area is fl oristically homogeneous due to cultivated plants, which made possible and quite accurate the mapping of some plant individuals, except for non-cultivated species, like weeds, and unidentifi ed pollen types.The diversity index used was Shannon (H') and a clustering analyses of euclidean distance using Ward's method was performed using the software PAST version 1.89 (Hammer et al. 2001).

Results
The pollen types collected by Apis mellifera are found in Table 1, they comprise 42 plants belonging to 17 families and fi ve unidentifi ed pollen types.The Myrtaceae family was the most important with only one type (Eucalyptus) while the Asteraceae, Euphorbiaceae, Brassicaceae and Poaceae appear nearly identical in harvest percentage, with eight, four, two and three species respectively.Among these, the most frequent pollen types were Eucalyptus (19%), Raphanus raphanistrum (13 %), Poaceae type 2 (7,5%), Jatropha cf.gossypiifolia (7%) and Sapium glandulatum (6,5%).The Poaceae family was considerably representative, with three different species, one of which was utilized by the bees during most of the year (eight months).Our data set reveals 80% of the species being collected between 0.5% and 2 %, an amount of 16% being collected between 2.5% and 7.5%, and only 4% more than 10%.
A seasonal approach reveals discrepant resource diversities: spring (25 pollen types), summer (18 pollen types), autumn (10 pollen types) and winter (12 pollen types).In the spring we point out four predominant pollen types, Emilia sonchifolia, Jatropha cf.gossypiifolia, Sapium glandulatum and Eucalyptus.The summer is represented mainly by Eucalyptus, Jatropha cf.gossypiifolia and Poaceae pollen types.During the winter Eucalyptus and Raphanus raphanistrum are important species.A Poaceae pollen type is the most important in the autumn but no Poaceae type is present in the winter.
Meteorological data and diversity indexes are plotted in Fig. 1.Trends are observed with temperature, precipitation and diversity increasing from August to October, when highest diversity is reached.A contrary trend is found in April and May when lowest diversities are seen.From May to July the three variables behave in the same manner.
Clustering (Fig. 2) divided the entire study period mainly in two major groups.The fi rst is characterized by wintery elements, namely Tropaeolum majus, Raphanus sativus, Raphanus raphanistrum, Vicia sativa and also Helianthus annuus and Eucalyptus.The second clusters March, December and January in one subset with Jatropha cf.gossipiifolia, Ocimun basilicum, Eucalyptus and Poaceae pollen types 1 and 2; the other subset (October, April, February and May) is simply represented by higher percentages of Eucalyptus and Poaceae pollen type 2.
With respect to grain size, the large class obtained 19%, very large 7% and small and very small only 2.3% each.The medium class, which comprises axes between 25 and 50 μm, was the most frequent with 29 pollen types (69%).The correlation "pollen diameter X frequency" was not signifi cant (r²=0.086).The species collected by Apis mellifera in this survey were from a few meters to about 450 meters distant from the hive.The colony hive was placed beside several Eucalyptus trees (i.e.~0 m distant), the distance from Raphanus raphanistrum and Raphanus sativus was 275 meters and most of the plants, including for instance the Lamiaceae, Aloysia triphylla, Tropaeolum majus, Citrus, Artocarpus heterophyllus, Jatropha cf.gossypiifolia were about 100 meters away.For Vicia sativa and Helianthus annuus the distance was about 450 meters.We assume weeds were not too far away because they are naturally expected to be more widespread.Indeed, we saw bees on small groupings of Emilia sonchifolia about 100 meters from the hive but we cannot safely infer that location was the source of this pollen since we assumed the collected weeds to be widely distributed over the campus.Among these weeds we can identify the Asteraceae species (except Helianthus annuus and Mikania cordifolia), Mimosa, Sida, Poaceae, Talinum and Richardia brasiliensis.If we consider the distance of such weeds and Eucalyptus is from a few to no more than 50 meters, we turn to have 43.5% of the total pollen collected at this distance range; 17% up to 100 meters, 14.4% up to 275 meters and 6.5% up to 450 meters.The remaining 18.5% are unknown.

Discussion
It is agreed that the spectrum of pollen collection by Apis mellifera tends to present dominant and accessory pollen types.This is seen as we have few pollen types collected more than 10% and most of them less than that, such results are the same as those reached by Cortopassi-Laurino & Ramalho (1988) and Vanderhuck (1995), which may be seen as temporary specialization on a fl oral source (Villanueva-G. & Roubik 2004).The major dominant types here are Eucalyptus and Raphanus raphanistrum.Many different species of Eucalyptus are cultivated, including hybrids, and their pollen morphology does not guarantee reliable identifi cation, thus it is very likely that the bees harvested more than one species.In addition, it is foreseeable that the Eucalyptus population within the study site being composed of different species and hybrids allows for a prolonged fl owering period.What is more, the proximity of the trees to the hive greatly infl uences the choice of this pollen source and the fl oral morphology displays no diffi culty for collection.Eucalyptus is also a nectar source for the Africanized honey bee (Santos 1961;Vanderhuck 1995;Bastos et al. 2003).
In contrast to Eucalyptus being collected most of the year (9 months), Raphanus raphanistrum was less constant (5 months) but abundant when used.This is clear as this species is mainly cultivated during the winter in-between plantations.Indeed, some Cruciferae species (Brassicaceae) are well known as good alternatives for green manuring like Raphanus raphanistrum and Raphanus sativus (Lima et al. 2007) as well as vetch (Vicia sativa).In this period, when food resource availability may decline due to the weather, such species were important options for the honey bee; this is a common practice among beekeepers, who move their hives to places where green manuring is used or where Raphanus sativus is cultivated, despite knowing that such practice may lead to nasturtium-fl avored honey, not so appreciated, but which guarantees the hive's survival during low foodresource availability (P.G.Rubert personal communication).
The Asteraceae was the fi rst family in number of species, with eight species found in ten months.Helianthus annuus occurred in November, December, June and July, when it was cultivated on the campus.Emilia sonchifolia was also an important source, mainly in November.The Poaceae types were used during nine months (13%) and were only absent in the winter.Other weed species were less frequent, namely Mimosa, Richardia brasiliensis, Talinum, Sida and Cecropia.We infer from our data that weeds are important pollen sources for A. mellifera and suggest a hierarchic trend for pollen exploitation of weeds by the honey bee on the campus; (1) Poaceae pollen production is massive and when there are favorable conditions like in the warm, wet summer it is common to see grass aggregates (e.g.colonião) bursting around, which is why no grass pollen was found in the winter.
(2) Asteraceae species also spread fast over disturbed areas and similarly to the grasses form large populations, and despite their pollen production not being as massive as that of the Poaceae, they provide bees with good quantities of pollen.(3) A third group is formed by less copious pollen producers and is occasionally harvested.Second-growth forest did not exist on the UFGD campus during the survey period due to regular weeding, thus the intervals from one  weeding to another only allowed pioneer herbs to succeed and reach fl owering.
The pollen types identifi ed here are most commonly found in bee fl ora inventories, so a brief comparison of the species and genera co-occurring in such surveys is possible.For instance, Machado & Carvalho (2006) concluded that 60% of the bee fauna visiting sunfl ower capitula are Apis mellifera and that it is an effective pollinator of this plant.Marques-Souza et al. (1993) found a great quantity of Cecropia pollen in loads of the honeybee in the Amazon, and argued that the harvest intensity on this plant may be strongly correlated with the presence of disturbed areas.Carvalho et al. (1999) evidenced the Africanized honeybee harvesting Eucalyptus, Sida, some Poaceae, Asteraceae and Mimosoideae.Manente-Balestieri (2001) argued the facility that bees have to harvest Mimosa species, being effective pollinators, what was indeed seen during fi eld observations in this study when Apis mellifera harvested Leucena leucocephala (Lam.) de Wit., though not found in the pollen spectrum.Furthermore, beekeepers already know that honey bees forage on "angicos" (Anadenanthera, Piptadenia, Parapiptadenia) (P.G.Rubert personal communication), once they eventually observe them in the fi eld.Santos et al. (2005) point out that the pollination of species of the Jatropha genus (Jatropha mutabilis and Jatropha molissima) can be done effectively and occasionally by Apis mellifera and that not only pollen but also nectar is collected.These species like many Euphorbiaceae necessarily need biotic vectors for pollination since the fl owers are unisexual, and thus, regarding the importance of Jatropha species as oil producers, it is interesting to learn which species are the most effective pollinators so that management can be better exploited.
According to results of Marques-Souza et al. (1993) and Manente-Balestieri (2001) Arecaceae is an important food resource for Apis mellifera as the dense infl orescences provide pollen in an accessible way and that Curculiondae beetles which visit and effectively pollinate do not show agonistic behavior against bees.Nevertheless, our data set contains no Arecaceae.This may be explained by two main points: in fi eld observations we testifi ed Trigona spinipes bees behaving agonistically against other Apidae on the fl owers of a palm species.In addition to this, there were few palm trees fl owering during the study period on the campus, and in this case the distance from the hive was approximately 700 meters, which seems not to have been the commonest distance for external activities of A. mellifera according to other values found.
In a Cerrado area (Cassilândia municipality-MS), Vieira et al. 2008 performed fi eld observations in order to list the plants visited by bees and out of 38 plants visited by the honey bee only Vernonia is a similar result to the data presented in this study.This is interesting because considering the mosaic-like vegetation of the region we clearly see our data are far from a complete range of possible exploited fl oral sources, regionally speaking.Soares et al, (2007) found Mimosa, Cecropia, Poaceae, Astronium and Vernonia present in honey samples of the Africanized bee in Mato Grosso state.
In a synthesis of melissopalynological studies in Brazil, Barth (2004) commonly mentions the use of Eucalyptus by the honey bees as both pollen and nectar source.Other cited taxa also present in our dataset are Astronium, Vernonia, Mikania, Cecropia, Croton, Mimosa, Artocarpus, Citrus and Poaceae.Pollen grains between 25 and 50 μm were preferably collected, which is in agreement with results of Pernal & Currie (2002), who saw a higher collection on particles below 48 μm.On the other hand, Silva & Absy (2000) did not fi nd positive correlation (pollen diameter X frequency) when analyzing pollen from honey samples of Apis mellifera.It seems that a positive correlation is more likely to be found when one analyzes the collection of pollen, not nectar, due to characteristics inherent in pollen handling (Pernal & Currie 2002), however this was not seen here.In relation to distance from the source plants Paranhos et al. (1997) and Pacheco et al. (1985) saw that the Africanized honey bees prefer to harvest over distances between 50 and 100 meters but can go further.Data presented here are in agreement with this fi nding.
As shown in Fig. 1, the relationship of visited plant numbers and rainfall ranges in direct proportion from August to October, which is the highest rainfall period and in May there is a sharp decrease in all variables.In this month the collection had the lowest richness, possibly due to low temperatures recorded (mean of 17.5 °C) and short food source.Thus, the bees based their collection on Eucalyptus and Poaceae, exactly the fi rst and third respectively most collected plants in total.When capturing bees for pollen sampling we noticed there was no activity on very cold days.Contrarily, on warm days after constant rain the bees were extremely active, but on such occasions we waited over an hour and saw no worker bee carrying pollen loads.
Finally, to get to a bigger picture of the collection dynamics ruling Apis mellifera activities we claim a set of arguments is needed to explain the bee's autoecology.Resource availability, increase in collection effi ciency and recruitment, temporary specialization and behavioral features are undoubtedly major traits governing pollen choice (Pernal & Currie 2002;Villanueva-G. & Roubik 2004;Ramalho et al. 2007).Hence, it is clear that Apis mellifera behaves in such a way as to satisfy colony needs, with worker bees harvesting most of the time near the hive but extending this harvest distance range when the food source is low or if there is any appreciated plant farther away.As a model of generalist harvesting behavior for bees (Roubik 1989) our results are unsurprisingly in agreement with theory.Recommendations for regional apiculture may stem from all data that have been accumulating in literature plus the present ones.It is well known that the region is an extremely fragmented landscape of plantations and forest fragments which include semi-deciduous and cerrado vegetation, occasionally called Mata de Dourados.Nevertheless, our dataset better expresses the matrix areas rather than natural vegetation cover.A plant list may not be enough since the area where a bee colony is placed is usually chosen according to plant composition, not the contrary; hence it seems indicatively workable that hives are placed near orchards (e.g.Citrus and Artocarpus) or mixed gardens (e.g.Mikania, R. sativus, Jatropha, Ocimum, Marsypinathes chamaedrys, Talinum, Talinum majus and Aloysia triphylla), plantations (Eucalyptus and Helianthus annuus but also Jatropha, Raphanus sativus and cultivated grasses) and green manuring (Raphanus and Vicia).Obviously we do not deny the importance of natural vegetation and weeds, and indeed recommend land users to avoid weeding when possible, this is indicated by pollen sources that are widespread as pioneer herbs (Asteraceae, Poaceae, Sida, Mimosa, Lamiaceae, Rubiaceae and others) and trees (Cecropia and Sapium).Ruderal vegetation, for instance, is usually found bordering plantations and may thus be additional or accessory pollen sources for honey bees, not to mention nesting sites for native bees.The knowledge of nectar sources will increase the ability for good management, and combining bee products with pollination service will progressively enhance more ecologically fair practices.

Figure 1 .
Figure 1.Diversity of pollen collection plotted together with precipitation and temperature data for the study period.

Figure 2 .
Figure 2. Similarity dendrogram of euclidean distances for the study period.

Table 1 .
Pollen types collected by Apis mellifera L. in Dourados (MS), Brazil, and their monthly and annual percentages.