The extent, frequency and ecological functions of food wasting by parrots

Anecdotic citations of food wasting have been described for parrots, but we lack a comprehensive knowledge about the extent of this behaviour, and its ecological and evolutionary implications. Here, we combine experimental and observational approaches to evaluate the spatial, temporal, typological and taxonomic extent of food wasting by parrots, to identify the ecological and evolutionary factors driving food wasting, and to assess the incidence of two ecological functions derived from food wasting, such as food facilitation to other animal species and secondary seed dispersal. We found that food wasting is a widespread behaviour found in all the studied parrot species. However, the proportion of food wasted differed among species and throughout the year. Parrots wasted more food during the non-breeding season, when they relied on exotic plants and on unripe fruits or seeds. We also recorded 86 animal species feeding on the food wasted by parrots, 27 of which potentially acted as secondary seed dispersers. Overall, our study emphasizes the universality of food wasting among parrots, and the important implications that this behaviour may have for the species involved (i.e., the parrot, the plant, the other species feeding on wasted food), and for the functioning of the whole ecosystem.

because they have had shorter evolutionary times together to get used to them. However, frugivores may also deliberately drop low-quality fruits such as those parasitized 11 , unripe, or those with low energetic and nutritional content 19 . In these cases, if waste is not accidental, individuals may be able to adjust this behaviour according to food availability, as dropping food when it is abundant does not have a large energetic cost.
Despite the uncertainty about why parrots waste food, several studies have suggested that this behaviour may benefit other species 9,10,13 . For example, dropped fruits or seeds become available to ground dwelling animals that otherwise cannot benefit from them 9,10,13 . Besides eating them, these species can also act as secondary dispersers [9][10][11][12][13] , increasing the number of different dispersal modes for the plant involved and, thus, the chances of effective dispersal. However, we do not know to what extent wasted food is consumed or secondarily dispersed by other species 20 .
Parrots (Psittaciformes order) are an evolutionarily old and diversified animal group 21 where waste has been anecdotally observed, but never quantified. Therefore, we combined experimental approaches with fieldwork conducted in five continents to understand the importance of this behaviour across parrot species, its drivers and the consequences for other species. Table 1 summarizes the hypothesis tested about waste by parrots, as well as the associated predictions. Briefly, we evaluate the spatial, temporal, typological and taxonomic extent of waste by parrots. If waste is an anecdotal behaviour, we expect to find it only in few parrot species. Otherwise, if waste is a widespread behaviour, we would find it in most parrot species, independently of their evolutionary history and their region of origin.
Then, we assessed if waste is a random process, independent of parrot life traits. To so do, we looked for differences in waste occurrence and amount of food wasted among species to identify the ecological factors driving waste. We hypothesized that this behaviour can be either accidental (i.e. plant parts fall because of difficulties during handling) or deliberate (i.e. parrots decide to waste food). Accidental waste may happen with a higher frequency when individuals forage in large groups, because of conspecifics disturbance, or with exotic plants, because parrots are less used to handle them. Contrarily, if waste is deliberate, this behaviour should happen more frequently with low-quality fruits and seeds (e.g. parasitized and unripe), and when energetic requirements are softer (e.g. outside the reproductive season, or when food availability is high). We also predict that smaller species will waste less food because they have a faster metabolism than large species (e.g. 22 ) and need to optimize food intake. Lastly, we hypothesized that wasted food can benefit other species, so we assessed the extent and incidence of two possible ecological functions derived from waste: food facilitation to other animal species and seed dispersal. If wasted food is beneficial for the plant and for other species, we expect that the number of seeds underneath the trees where parrots have wasted food will be large and will attract many species. We also expect that waste will benefit a large number of different species with different potential dispersal modes. Finally, we discuss some possible ecological functions of waste and their ecological and evolutionary consequences.

Results
food wasting extent and randomness. Waste by parrots is a widespread behaviour in many aspects (P1, Fig. 1). It was found in all the study sites visited (35 biomes, 17 countries and 5 continents), in native and introduced parrot ranges. Waste was observed throughout the year, in both the breeding and non-breeding season. Parrots wasted mainly fruits and seeds, but also flowers, leaves, twigs, stems, sprouts, parasites and bark. We observed 103 parrot species (40 species in the experimental approach, 75 in the field and 12 in both the experiment and the field) from 38 genera wasting food from 336 plant species belonging to 80 families (see Supplementary Tables S1 and S2 for complete lists of parrots wasting food and plants where waste occurred).
Following our expectations (P2), waste was independent of the evolutionary history of the species, as it did not show a significant phylogenetic signal using any of the 100 phylogenies used (mean K: 0.104, range: 0.007-0.165, all P > 0.409). Besides, we found that the percentage of waste by parrots was large (P3), as it averaged 21.2% (SD = 13.0) of the total food provided to individuals during an in-captivity experiment (range: 0.4-72.7%). In the wild, we found that parrots wasted 11.8% (SD = 25.1) of the fruits and 14.6% (SD = 20.3) of the seeds that they handled, ranging from lack of waste to a waste of all the fruits or up to 80% of the seeds.
Moreover, we found strong among-species differences in waste occurrence in the wild (δAIC with null model 2882) and in the proportion of food wasted during the experiment (δAIC with null model 12, Fig. 2), suggesting that food wasting is not a random process (P4). factors driving food wasting. Using a large database of field observations about waste occurrence we identified some factors affecting the occurrence of this behaviour. Fruit and seed wasting were more frequent during the non-breeding season (P10), in exotic plant species (P6) and on unripe fruits or seeds (P9) for all the species. We also repeated the analyses considering only data from P. krameri, which is the parrot species with more detailed information, and we observed the same pattern, both including all observations collected across the world and only those observed in Seville, the population with more information ( Table 2, Fig. 3). This last analysis corroborates results obtained using the whole data. However, the analyses for all species but excluding P. krameri indicate that the season was the only factor driving waste occurrence.
Results obtained using experimental data showed that species size (i.e. average body mass, P7), food availability (i.e. fasting, P8) and conspecifics disturbance (i.e. number of individuals in the cage, P5) did not affect the proportion of food wasted (all p-values > 0.14, Supplementary Table S3). Finally, we also found that the total proportion of wasted fruits that were parasitized was very low (<4%, n = 176, P11). ecological functions of food wasting. By checking the amount of food wasted under the trees where parrots had been wasting food, we found that this was large in many cases, as expected (P12). We found an average (±SD) of 53.4 ± 52.2 wasted fruits and 42.0 ± 54.9 wasted seeds under the trees (Supplementary Table S4). The maximum number of seeds found under a single tree (239) was larger than the number of fruits (164).
The number of species that benefited from the food wasted by parrots was also large and widely distributed, supporting P13, including 86 species of birds, mammals, reptiles, fishes, and ants ( Table 3) that widely ranged in body size (0.002-750 kg). Also, we detected 28 different species from 26 genera potentially acting as secondary dispersers of the fruits and seeds wasted, including ants, birds, mammals and reptiles. As predicted (P14), these species ranged from very small to very large body sizes, and their average dispersal distances were mainly large (>30 m), but also very large (>100 m) for two species.

Discussion
In this paper, we show that waste is not an anecdotic behaviour among parrots, but a widespread habit present in at least the 103 studied species. Waste is mainly focused on fruits and seeds, although it also involves other plant parts, and occurs throughout the year and through all the regions where parrots have native or introduced populations. We also show that waste has implications for other species that can forage or secondarily disperse wasted seeds and fruits. Thus, we can consider that parrots are interacting with plants along mutualism-antagonism continuums, as they can both prey upon several plant parts but also act as seed-dispersers, pollinators or plant healers 12,23 .
Besides waste being a widespread behaviour in parrots, we found that the proportion of food wasted changed between species, suggesting that some ecological or evolutionary factor may be modulating this behaviour. However, our tests to disentangle the accidental or deliberate nature of waste were not conclusive. We found that parrots waste more food from non-native plant species than from native ones, maybe because they had a shorter co-evolutionary history to adjust their handling techniques to the fruits and seeds of these novel species. Also, exotic plant species are often located closer to human infrastructures and thus, waste may be a consequence of human disturbance (i.e. parrots suddenly flying away and dropping food when humans get close to them). We also found that waste is more frequent during the non-breeding than during the breeding season, which is probably related to the higher nutritional requirements that individuals have while raising chicks 24 . If individuals need to gather more food for themselves and their chicks during the same time, they may choose to reduce food waste to increase foraging efficiency, suggesting that waste may also be deliberate. Also, parrots seem to waste with a higher frequency low quality (i.e. unripe) fruits, showing some type of selectivity. However, we did not find differences in food waste in relation to species body size or food availability (i.e. fasting), suggesting that metabolic differences among species and food availability are not driving this behaviour. Finally, some studies had suggested that parrots might preferentially waste or consume parasitized fruits, modulating the effect of parasites on plant fitness 11,25 . However, our observations did not show any preferential consumption of parasitized fruits. www.nature.com/scientificreports www.nature.com/scientificreports/ Food wasted by parrots benefits many animal species, as seeds and fruits become more readily available and during a longer temporal window after being thrown from the tree by parrots. Also, wasted food is less often rotten or dry, as happens when food naturally falls from the trees. Parrots waste plenty of half-eaten fruits, reducing the original size of the fruit and breaking hard fruit parts, thus allowing their consumption by smaller species. For example, Douglas et al. 26 found that the quantity of parrot frugivory increased habitat quality for Bananaquits (Coereba flaveola), because of parrot ability to open hard fruits and leave them partially consumed and thus www.nature.com/scientificreports www.nature.com/scientificreports/ available for other species. Besides, fruits become available for ground-dwelling species that otherwise cannot access the resource (as already suggested in 9,10,13 ).
Fruits and seeds wasted by parrots can also be subject to secondary dispersal by many animals. This may be vital for some plant species, as many plants rely on secondary dispersal in their reproductive cycle 27,28 . It is important to underline that even if many of the wasted fruits and seeds are unripe and partially consumed, a large proportion of them can ripen and germinate after dispersal 20,29,30 . Also, having seed dispersers with different ecological characteristics (e.g. home ranges, places to defecate, dispersing methods) may increase the chances of effective seed dispersal 28 . The list of species either consuming food discarded by parrots or acting as secondary dispersers is very large, and includes many species from different taxonomic groups and with very different body sizes and dispersal distances. It is worth mentioning that our list is very conservative, as many of the species that we observed consuming the wasted fruits and seeds can also potentially disperse them. Thus, the impact of this wasteful behaviour on a large list of animal and plant species may be large given its universality.
Besides food facilitation and secondary seed dispersal, waste may have other ecological functions. If this behaviour has been maintained over the evolutionarily history of this old and diversified animal group 21 , it may be because it has some beneficial consequences for parrots 12 . One suggestion is that waste may benefit parrots by increasing the availability of high quality (e.g. large-sized, sugar-rich and nutrient-rich) fruits. In horticulture, it is widely known that such large high-quality fruits can be obtained through fruit and flower pruning 31,32 . Interestingly, pruning needs to be directed to unripe fruits to affect the quality of the non-wasted fruits [33][34][35] , coinciding with our result that waste occurs with a higher frequency in unripe fruits. Moreover, fruit and flower pruning are also known to reduce gaps between fructifications 36 or reduce biennial bearing 32,37,38 . Thus, parrots may also be extending the fruiting period of the trees and increasing their predictability (i.e. shorter gaps between fructifications and transition from biennial to annual bearing). For any of these hypotheses to be possible, parrots www.nature.com/scientificreports www.nature.com/scientificreports/ need to be able to make intertemporal choices (i.e. sacrifice short-term satisfaction to obtain a higher reward in the future), which have already been detected for this bird group 39 . Overall, waste may have longer-time effects on wasted plants than expected, but further studies are needed to validate this hypothesis.
Besides the apparent negative effect that fruit and flower wasting may have for the plant, it is widely known that many plant species naturally produce more flowers and fruits than they can set 36 . The overproduction of juvenile fruits may be evolutionarily adaptive as these can satiate pre-dispersal fruit and seed predators 40,41 and plants may be able to produce larger yields than average in occasional years of plentiful resources 42,43 . Plants may also compensate this overproduction by aborting flowers or fruits during the growth process 36 . As wasted fruits are preferentially unripe, their loss can be partially compensated by reducing fruit or flower abortion, reducing the negative effect of fruit thinning for the plant.
Waste may also have important consequences for the soil, as large amounts of organic matter accumulate over short time periods under a single tree. This behaviour may also expedite nutrient cycling, as wasted plant parts may enter the decomposition phase faster than if they were to stay in the tree for longer time periods 12 . Overall, our study wants to emphasize the universality of waste in parrots and the important implications this behaviour may have for the species involved (i.e. plants and parrots), but also show how other species benefit from the wasted food and for the functioning of the ecosystem.
Two specifications about our study need to be done before concluding. First, it is important to notice that the experimental data cannot be directly compared with what happens in the natural environment as the food provided and the way parrots obtain is markedly different. However, the waste experiment was designed to control for several factors that would be very challenging to account for in the wild. For example, it is very difficult to test among-species differences on the wild because the species are never found under the same circumstances (e.g., environmental conditions, food species). Another important factor that cannot be tested in the field but was easily addressed in the experiment was food availability. Therefore, the experiment offers complementary information www.nature.com/scientificreports www.nature.com/scientificreports/ to that provided by the data gathered in the field. Also, in our predictions 7-11, we assume that an individual wasting less food has a higher foraging efficiency than an individual wasting more food. However, it may be more efficient to feed by partially eating and wasting various fruits than to handle the same fruit for a long time, which will become smaller and more difficult to eat. This does not invalidate our predictions, as about half of the fruits and seeds wasted are intact (see Supplementary Table S3).
Despite the several aspects that deserve more research, this study adds to the growing evidence that parrots have a much more important role for the conservation of the ecosystems than previously thought 23,44 . Some of the functions attributed to parrots, such as primary and secondary dispersal [ 12,45 , this study], may have a very important role in the current scenario of global change where tropical forests are being fragmented at high rates and forest recovery may depend on the efficiency of seed dispersal. However, many parrot species are highly endangered, with several species functionally extinct and many of the remaining species under strong human pressure (e.g., 28% of extant species are classified as threatened under IUCN 46 ). Thus, management efforts to conserve parrot species should considerate the ecological interactions of food waste in the conservation planning strategies of habitat protection and population recovery (e.g. selection of priority areas for conservation and selection of release sites).

Datasets used.
To test our hypotheses about waste, we combined information gathered using different approaches and unified in five different datasets (All of them available as Supplementary Material). Dataset 1. Experiment in captivity. We first quantified waste experimentally using captive individuals. We performed 362 experimental trials using 130 individuals from 40 parrot species belonging to 24 genera (body weight range: 33-550 g) to measure the proportion of food wasted by the species. Individuals were kept in groups of 1-4 individuals, simulating their typical aggregated foraging behaviour. As cages used for the experiment always have the same size, experiments done including a larger number of individuals simulate situations of higher conspecific disturbance. Parrots were acclimated to the cages and diet for 10 days prior to the experiment. In a third of the experiments, individuals were fasted 24 hours before the start of the experiment, to simulate low food availability. Each species was fed with the typical food used in captivity (e.g. mixtures of millet, birdseed, sunflowers, and peanuts). Food was weighted (±0.1 g) and provided ad libitum to the parrots. Cages were set so that wasted food could not be eaten by individuals after falling to the cage floor. Each group of birds was tested approximately five times, and each trial lasted 24 hours. Wasted food was weighted, including only those seeds that were intact (i.e. half consumed seeds were not included). Finally, we calculated the proportion of food wasted by each group of parrots in each trial. All experimental protocols are in accordance with the relevant guidelines and regulations. Birds were kept in captivity under permit SGYB/FOA/AFR from the Consejería de Medio Ambiente, Junta de Andalucía, in the authorized centre for experimental avian research SE/16/U (REGA ES410910008016).
Dataset 2. Field transects. We collected observations of waste in different fieldwork campaigns performed in 17 countries and 5 continents (Fig. 1a). We looked for groups of foraging parrots in pre-defined roadside and walking transects (see details about transects on 47 ). Every time we detected a foraging group, we observed focal parrot groups or individuals for 5-10 minutes and annotated the occurrence of food wasting (1/0), the parrot species, the plant species consumed, the part of the plant that was wasted (flowers, fruits, seeds, bark, leaves, twigs, sprouts, Large (>1 kg) 16 4 Medium (>0.1 kg) 29 12 Small (>0.01 kg) 23 6 Very small (<= 0.01 kg) 5 3

Dispersal distance
Very large (>100 m) -2 Large (>30 m) -11 Medium (>10 m) -9 Small (<= 10 m) -5 Table 3. Number of species detected consuming or as secondary dispersers of the fruits wasted by parrots by taxonomy, functional group, body size and dispersal distance. *This is a conservative number of species due to the hard species identification. www.nature.com/scientificreports www.nature.com/scientificreports/ stems, resin or invertebrates), flock size, the season when the observation was taken (breeding/non-breeding), the ripening stage of the fruits/seeds (unripe/ripe), date, site and the origin of the plant where the observation was taken (native/exotic). This dataset includes a total of 6253 observations of foraging parrots observed between 2011 and 2019 in 37,612 km of transects.
Dataset 3. Waste quantification. We quantified the proportion of food wasted in the field by observing the foraging behaviour of individuals detected handling fruits or seeds during the field transects. For this and the following datasets, and for the statistical analyses in this study, we focused on fruits or seeds and excluded other wasted plant parts because fruits and seeds are the main food types wasted by parrots (see Fig. 1). Foraging individuals were observed from a distance with binoculars or telescopes. We identified the bird and plant species, and we counted the flock size and the number of fruits/seeds each individual ate or wasted. We then calculated the proportion of food wasted as the number of wasted fruits/seeds divided by the total number of fruits/seeds handled. We compiled 412 observations of individual birds from 20 species in Bolivia, Costa Rica, Namibia, Brazil, Peru, Argentina and Spain, between 2014 and 2019. Data was collected during six different months in both the breeding and non-breeding season, and birds handled 1841 fruits and 934 seeds. As some studies suggest that parrots may be wasting parasitized fruits or seeds (e.g. 11 ), we also counted the total number of wasted fruits with worms for 176 fruits under 7 different tree species during fieldwork in the Brazilian cerrado in 2017.

Dataset 4.
Waste under tree. We estimated the number of fruits/seeds a group of parrots could waste per individual tree. To do so, we counted the total number of intact and wholly or partially eaten fruits/seeds under a tree after a group of parrots foraged on it. We also identified the plant species. When the number of fruits was very large or the area was hard to screen because of the dense vegetation, we counted half of the area under the tree and then doubled the number of fruits/seeds. We compiled information on 98 trees from 29 species in Australia, Peru, Ecuador, Bolivia and Brazil between 2013 and 2017.
Dataset 5. Camera traps and direct observations. We used 96 camera traps to monitor the animal species using fruits and seeds wasted by parrots. Cameras were located under the plant, in front of a bunch of fruits/seeds. They stayed activated 5-7 days during 24 hours. Data was gathered in Brazil and Bolivia, under four different plant species where waste had been observed: Attalea totai, A. barreirensis, A. speciosa and Mauritia flexuosa. From the pictures, we separated species that consumed the fruits/seeds and those that took entire fruits/seeds out of the camera, thus being possible secondary seed dispersions. We combined this information with 293 direct observations of food facilitation and secondary dispersal taken in Australia, Spain, Puerto Rico, South Africa, Argentina and Sri Lanka between 2012 and 2019. food wasting extent. To test our first prediction (P1) and describe the spatial extent of waste, we identified all the areas around the world where we had observed waste in the wild. The temporal extent of waste was described as the total number of waste events recorded per month. We also calculated the total number of waste events for ring-necked parakeets Psittacula krameri, the species with the largest number of waste events detected. For the temporal extent we used the Field transects dataset, as this is the largest compilation of waste events taken using a standardized method. This same dataset was used to identify the typological extent of waste by counting the total number of waste events found for each plant type (flower, fruit, seed, bark, leaves, twigs, sprouts, stems or invertebrates). Finally, the taxonomic extent of waste by parrots was quantified by identifying the total number of parrot species that was found wasting food and the total number of plant species that were subject to waste by parrots in any of our datasets, and in non-systematic observations performed during fieldwork.
Our second prediction (P2) that waste is independent of the evolutionary history of the species was explored using data from the Experiment. We assessed if there was a phylogenetic signal in the proportion of food experimentally wasted by the different species using the descriptive statistics K 48 . When K < 1, the relatives resemble each other less than expected under the Brownian motion evolution, while when K > 1 close relatives are more similar than expected under the Brownian evolution. We evaluated the statistical significance of the phylogenetic signal by comparing the observed variance of independent contrasts of the proportion of food experimentally wasted by the different parrot species to a null model of shuffling taxa labels across the tips of the phylogeny. We calculated K and the statistical significance of the phylogenetic signal for 100 bird phylogenies from Jetz et al. 49 using the picante package 50 in R version 3.5.3 51 .
Finally, we evaluated the proportion of food wasted (P3) using two datasets: the Experiment and the Waste quantification. We used the average (±SD) proportion of food wasted by each group of parrots in each trial for the Experiment data and the average (±SD) number of wasted fruits/seeds by each individual parrot for the Waste quantification data.
Waste differences among species. If food wasting is a not a random process (P4) waste occurrence and quantity should differ among species. We used the Field transects data to compare waste occurrence (1/0) among species and the Experiment dataset to compare the proportion of food wasted among species. To test if waste occurrence differed between species, we fitted a Generalized Linear Model (GLM) in R with species as a predictor variable and occurrence as a response variable, using a binomial distribution. We then compared the proportion of food wasted (response variable) among species (predictor variable) in the experiment by means of Generalized Linear Mixed Models (GLMM) using a beta distribution with the glmmADMB library 52 . Because the same individuals were used for different trials, we included individual (or group of individuals) as a random factor in all the models. In both cases, we compared the model including species as predictor variable with a null one. Models with a difference in AIC smaller than 10 were considered equally supported. www.nature.com/scientificreports www.nature.com/scientificreports/ factors driving food wasting. Waste may be driven by different factors, depending on its accidental or deliberate behaviour (predictions 5-11, Table 1). Because the different factors affecting waste may be related, we fitted multivariate models including several predictor variables at the same time. We ran one model for waste occurrence (1/0) (Field transects dataset) and one for the proportion of wasted food (Experiment dataset). Experiment data were used to relate the number of individuals in the cage during the trial (P5), parrot body size (mean weight in g, P7) and the reduction in food availability (simulated by a fasting period, P8) (predictor variables) with the proportion of food wasted (dependent variable) by means of GLMMs using a beta distribution. We included individual (or group of individuals) as a random factor in all the models, nested within species. The weight of the birds was standardized before modelling (i.e. transformed to have a mean of 0 and standard deviation of 1).
Then, we evaluated factors affecting waste occurrence (1/0) in relation with the season when the observation was taken (breeding/non-breeding, P10), the ripening stage of the fruit/seed (unripe/ripe, P9) and the origin of the plant where the observation was taken (native/exotic, P6). We also included the number of individuals in the flock as a covariate in the models to control for the potential effect of larger flocks having a higher chance of showing waste. To test the consistency of the results, we performed the analyses for all the species, but also for the species with the largest number of observations (P. krameri); additionally, only for P. krameri in the study site with a larger number of observations (Seville, Spain, where the species is introduced) and finally, for all species excluding P. krameri. We fitted Generalized Linear Models (GLM) in R using a binomial error distribution for the models for P. krameri and GLMMs with species as a random term for the model with all the species and for the model with all the species, but excluding P. krameri.
As some studies suggest that parrots may be wasting parasitized fruits or seeds (P11), we calculated the proportion of wasted fruits or seeds that were parasitized for 176 fruits found underneath seven different tree species. ecological functions of food wasting. To test if waste is large under the tree (P12), we used the Waste under tree dataset. We calculated the mean (±SD) number of intact and partially eaten fruits/seeds under a tree. We also identified the maximum number of intact and partially eaten fruits and seeds found.
We finally evaluated two possible ecological functions of waste by parrots, facilitation of food to other species and secondary seed dispersal (P13 & P14), using the Camera traps and direct observations dataset. For each species benefiting from wasted food, we identified its taxonomic group (i.e. ant, bird, mammal, reptile or fish) and estimated its body size (i.

Data availability
All data generated or analysed during this study are included in this published article (and its Supplementary  Information Files).