In the last decades, many studies have tried to determine how different floral traits (e.g., color, shape, odor) or pollinator syndromes reflect plant adaptations to particular pollinator species (review ^{Rosas-Guerrero et al. 2014}, ^{Murúa & Espíndola 2015}). Although pollination specialization exists, most of the flowering plant species are visited by a diverse number of pollinator species, which has shown that generalist pollination is a predominant plant-animal interaction in nature (^{Waser et al. 1996}, ^{Herrera 2005}, ^{Ollerton et al. 2006}). In those cases, the relative importance of pollinators as selective agents is not necessarily obvious. Since, the action of selection can vary depending on the diversity (species number), abundance (visitation frequency), and effectiveness (pollen carry over by visit) of each pollinator (^{Aigner 2001}, ^{Ne’eman et al. 2010}, ^{Sahli & Conner 2011}). In this respect, ^{Stebbins (1970}) proposed “The Most Effective Pollinator Principle”, which raises that flower phenotype would be molded by the most frequent and effective pollinator species. So, even if a plant species presents a diverse pollinator assemblage, only some of them would have a greater impact on the reproductive success of the plant (^{Kay & Sargent 2009}).

Alstroemeria ligtu L. var. simsii (Spreng.) Ehr. Bayer (Alstroemeriaceae) is a perennial and hermaphroditic self-incompatible herb, which is completely dependent on pollinators for reproduction (^{Arroyo & Uslar 1993}). Along its range of distribution, their populations show a high spatial variation in their pollination assemblages, being visited by approximately 24 different species of insects belonging to the Diptera, Hymenoptera and Lepidoptera orders (^{González et al. 2014}). Despite this, recent investigations have shown that most of the A. ligtu populations are mainly visited by the same pollinator species (e.g., Lasia corvina Erichson (1840) and Centris nigerrima Spinola (1851)), which are consistent among years (^{González et al. 2014}, ²⁰¹⁵), suggesting that they may engage in a more specialist, rather than generalist, plant-pollinator interaction. However, there is still unknown how effective are these species in transferring pollen to A. ligtu. According to this, it is possible to ask whether the same floral visitor could be equally effective in the pollination service provided to these populations. In order to answer this question, here we characterize the pollinator assemblages and determine the pollination effectiveness of the main pollinator species in three A. ligtu populations.

The study was conducted during the flowering season of 2014 (November - December) in three populations of central Chile Cuesta Zapata, Río Clarillo and Termas del Flaco, where A. ligtu flowers dominated the area (Fig. 1A-B). All populations are characterized by a Mediterranean-type climate with strong rains during wintertime (^{Di Castri & Hajek 1976}). The population Cuesta Zapata (ZA; 33º39’19” S; 70º19’23” W) occurs in Coastal hills at 529 masl, where the site is dominated by Quillaja saponaria Molina and different herbs species such as Alstroemeria pulchra Sims, A. angustifolia Herb., Carduus pycnocephalus L., Sisymbrium officinale (L.) Scop., Loasa tricolor Ker Gawl., Clarkia tenella (Cav.) F.H. Lewis & M.E. Lewis, Oxalis rosea Jacq., Eschscholzia californica Cham., Calceolaria sp., Schizanthus tricolor Grau & Gronbach, and Papaver sp. (pers. obs.). Río Clarillo population (RC; 33º46’80’’ S; 71º49’72’’ W) is located in the pre-Andean mountains at 1153 masl, where individuals of A. ligtu are inhabiting with a diverse assemblages of trees, scrubs and herbs species like Cryptocarya alba (Molina) Looser, Quillaja saponaria, Escallonia pulverulenta (Ruiz & Pav.) Pers., A. angustifolia, Carduus pycnocephalus, Conium maculatum L., Leucheria sp., Madia chilensis (Nutt.) Reiche, Stellaria chilensis Pedersen, Geranium berteroanum Colla, Sisyrinchium sp., Stachys grandidentata Lindl., Leucocoryne ixioides (Hook.) Lindl., Pasithea caerulea (Ruiz & Pav.) D. Don, Loasa tricolor, Clarkia tenella, and Schizanthus tricolor (pers. obs.). Whereas the third population occurs on road to Termas del Flaco (TF; 34º60’87’’ S; 71º49’72’’ W) at 585 masl, where plants are mainly surrounding by herbs such as Hypericum perforatum L., Sisyrinchium sp., Anagallis arvensis L., Rubus ulmifolius Schott and Calceolaria sp. (pers. obs.).

In each population, we selected three random patches of plants separated by 3 meters. A total of 482 plants were studied (ZA=200, RC= 200, TF=82). All individuals were tagged and monitored during 2 to 4 days per months. In order to characterize flower size one flower per plant were cut and photographed from a frontal view. Corolla area (mm²) were measured from digitalized photos and analyzed using Image J Laucher program version 1.45 (available online: https://imagej.nih.gov/ij/). Pollinator visitation regimens were quantified through focal observations during 15 min per plant, in sunny days between 10:00 and 15:00 hours. Visitation rate was calculated as the number of visits per flower per hour (visits*flower^-1*hour^-1). In addition, the handling time (i.e., the time speeded in a flower during a single visit) of each pollinator species was recorded. Only the most frequent pollinators were captured and stored in Eppendorf tubes with 1 ml of ethanol. The insects were moved to laboratory for pollen load counting. All captured pollinators were photographed, and their body lengths were measured in ImageJ Laucher program. Pollen was removed from pollinator´s body by shaking tubes in a vortex for 1 minute. From each tube, we took one aliquot of 0.3 ml and we put in a Neubauer chamber for quantification. In this way, under microscope we identified and quantified the pollen grains of A. ligtu and the other plant species based on the description made by ^{Arredondo-Núñez (2010)}. This procedure was carried out three times and pollinator pollen load (PL) was estimated as the average number of pollen grains transported. Since the number of pollen grains carried on the body has been related to the pollinator body length (^{Griffin et al. 2009}) we standardized our data by dividing the pollen grains by pollinator’s total body length (mm).

To identify the main pollinator species of each population the five species with the highest visitation frequency were selected using the results of an ANOVA test. Once the main pollinator per population was identified, their pollinator effectiveness was estimated as VR*PL (^{Ne’eman et al. 2010}). Finally, to determine statistical differences in visitation rate, pollinator effectiveness and handing time of main pollinator species a GLM with Gaussian distribution was performed. All variables were transformed to log (x+1) previously to the analysis in R package (Core Team 2018).

Corolla size measurements for all A. ligtu populations are presented in Table 1. The largest corolla areas were observed in plants inhabiting RC, followed by ZA, and the smallest ones were quantified in TF. Respecting to pollinator visitation regimens, a total of 798 visits by 22 pollinator species were registered in approximately 60 h of observation per population (RC: 63.5 h, TF: 69.5 h, ZA: 66.8 h; Appendix 1). Hymenoptera was the order with the highest number of species (12 species), followed by Diptera (8 species) and Lepidoptera (2 species). When only the most frequent pollinators were considered per population, the ANOVA test showed significant differences in the visitation rates of the five species with the highest visitation frequency in each A. ligtu populations (Table 2). The species of the dipteran genus Lasia were the most frequent pollinators in the three populations. Lasia corvina was the most frequent pollinator species in RC and TF, with the 62% and 85% of the total visits, respectively (Fig. 1C). While, Lasia aenea Philippi (1865) was the responsible of 50% of the visits in ZA population (Fig. 1D).

L. corvina showed a higher visiting rate in RC than in TF, and L. aenea in ZA (Fig. 1E). L. corniva showed thirteen times higher effectiveness in RC than in TF, while that L. aenea in ZA was also higher than L. corvina in TF (Fig. 1F). While, pollinator handling time species was mostly the same in each A. ligtu populations (Fig. 1G). Finally, GLM analysis showed significant differences in the visitation rate and pollinator effectiveness, but not in the handling time of main pollinator species among sites (Table 3).

Results of this study reveal that not only the visitation rate, but also the pollinator effectiveness can vary among A. ligtu populations, even if the same pollinator genus or species are the responsible of most of the visits in a population. Here, we found that Lasia corvina was the main pollinator in two of the three study populations (RC and TF) and L. aenea was in the third one (ZA). Where, they showed different visitation frequency and pollinator effectiveness, which seems not to be related to pollinator handling time. Pollinator choices are based on different floral cues which together favored their performance and gain in the foraging process (^{Waser 1983}). In RC population, L. corvina visited more frequently and effectively the flowers of A. ligtu than those in TF and ZA, but in the later L. aenea showed higher pollination effectiveness than L. corvina in TF. Pollination foraging behavior and consequently the success of the

pollination process depend on several non-exclusive factors both at the individual and community scale. At the individual level, flower size is an important visual signal in pollinator attraction, where it is known that plants with larger corolla sizes could have a higher chance to be visited than smaller ones (^{Conner & Rush 1996}). In RC, flowers have larger corolla areas than those in TF and ZA (Table 1), which might increase the probability of pollen recollection. In the same way, plants in ZA showed a higher corolla area than plants in TF, which could explain the higher mean visitation rate of L. aenea. At the community scale, an important factor that can also compromise pollination effectiveness is the composition of co-flowering species within a plant community (^{Hegland & Boeke 2006}). Pollination sharing can lead to pollen competition by visitation loss or interspecific pollen transfer, where plant species can steal pollinators reducing visitation frequency and/or they could perform mixed visits increasing pollen exchange between

plants in the community (^{Morales & Traveset 2008}). Here, we observed that L. corvina carried on average a larger amount of A. ligtu pollen and less pollen grains of other plant species inhabiting in RC than in TF, and ZA (Table 1), which could explain in part the highest pollination effectiveness estimation. Contrary, L. aenea in ZA carried more pollen of others plant species than L. corvina in TF, but this pollinator still showed a higher effectiveness (Table 3). This difference could be explaining by individual preferences and the different modes that pollinator species have to interact with the flower. Different species could be explaining the differences in the pollinator effectiveness, however, there is another important factor, and that is how the pollinator interact with the flowers (i.e, pollinator behavior). L. corvina and L. aenea showed almost the same average handling time (RC= 3.4 s/visit, ZA=3.5 s/visit) and they have higher pollinator effectiveness respect to the main pollinator of TF, which also showed the lowest HT (TF=2.8 s/visit). The time that pollinators spend in a flower can be tightly related to the amount of pollen grains that they are able to extract from the flower (^{Ohashi 2002}). Therefore, it is expected that the more time the insect spend on the flower; the more pollen it may extract. Nevertheless, this must be explored more carefully in order to determine if this effectively has the any impact on pollinator effectiveness.

FIGURE 1 Geographical distribution of studied populations, Alstroemeria ligtu flower detail, pollinator species and differences in the visitation rate, pollination effectiveness and handling time of the main pollinator species of each A. ligtu study populations. A) Populations locations, B) A. ligtu flower, C) Lasia corvina, D) L. aenea, E) Visitation rate, F) Pollination effectiveness, and G) Handling time. / Distribución geográfica de las poblaciones estudiadas, detalle de la flor de Alstroemeria ligtu, especies de polinizadores y diferencias en las tasas de visita, efectividad de polinización y tiempo de manipulación del principal polinizador de cada población en estudio de A. ligtu. A) Ubicación de las poblaciones, B) Flor de A. ligtu, C) Lasia corvina, D) L. aenea, E) Tasa de visita, F) Efectividad de polinización, y G) Tiempo de manipulación. 

In summary, this work has revealed that the same genus and/or pollinator species can have different effectiveness across A. ligtu populations. In the light of this evidence, future investigations must focus in determining the pollinator efficiency (^{Schupp et al. 2017}), that is to say, how the quality of pollination performance by a species (e.g., visitation and pollen extraction) could affect the quantity of pollination (i.e., pollen deposition and seed production), and consequently plant population fitness. Under a scenario of spatial variation of pollinator assemblage and effectiveness in A. ligtu, it is possible that local conditions can drive local plant adaptation, which could stimulate the generation of a mosaic of phenotypic plant selection across their range of distribution.

TABLE 1 Corolla size of Alstroemeria ligtu, Principal pollinator species, Pollinator’s proboscis length, Pollen grains of A. ligtu and from other plant species carried by the principal pollinator species in each study population. / Tamaño de la corolla de Alstroemeria ligtu, Polinizador principal, Largo de la proboscis del polinizador, Granos de polen de A. ligtu y de otras especies de plantas acarreados por el polinizador principal en cada población en estudio. 

TABLE 2 Visitation rate differences of the five most frequent pollinator species in each Alstroemeria ligtu populations after ANOVA analysis. Populations: Río Clarillo (RC), Termas del Flaco (TF) and Cuesta Zapata (ZA). / Diferencias en la tasa de visita de los cinco polinizadores más frecuentes de cada población de Alstroemeria ligtu despúes de un analisis ANDEVA. Poblaciones: Río Clarillo (RC), Termas del Flaco (TF) y Cuesta Zapata (ZA). 

TABLE 3 Pollination effectiveness (PE), Visitation rate (VR), and Handling time (HT) difference of the most frequent pollinator species of each Alstroemeria ligtu populations after a GLM model with Gaussian distribution. Populations: RC= Río Clarillo, TF= Termas del Flaco and ZA= Cuesta Zapata. / Diferencias en efectividad de polinización (PE), Tasa de visita (VR), y Tiempo de manipulación (HT) de la especie de polinizador más frecuente de cada población de Alstroemeria ligtu después de un modelo GLM con distribución gausiana. Poblaciones: RC= Río Clarillo, TF= Termas del Flaco y ZA= Cuesta Zapata.