SEEDEATERS AND SEEDS AT A TECOMA SAVANNA IN THE SOUTHERN PANTANAL, BRAZIL

· Seedeaters ( Sporophila spp.) comprise a rich Neotropical bird group of seed consumers common in open habitats. In this study, we documented the feeding habits of seedeaters and seed produc?on at a Tecoma savanna (dominated by Tabebuia aurea ) in southern Pantanal, Brazil. We also analyzed the rela?onship between seed oﬀer (abundance, richness, and diversity) and the number of seedeaters foraging across seasons. Six species ( Sporophila angolensis , S. caerulescens, S. collaris S. hypoxantha, S.leucoptera, and S. lineola ) occurred in the savanna, mainly in the height of the wet season, when seed produc?on increased abruptly, a)rac?ng seedeaters. Seedeaters used 14 of 16 grass species that produced seeds. Indeed, the number of foraging seedeaters paralleled the abundance

adapted to consume alterna?ve resources during seed decline (Schluter 1982a(Schluter , 1982b)).Vertebrates such as granivorous birds owen search for seeds at diverse spa?al scales, responding to local produc?onpa)erns (Schluter 1982a, Jahn et al. 2020).Indeed, the local abundance of granivorous birds owen mirrors seed produc?on in a given area (Marone 1992).Although granivorous birds have strong beaks and gizzards that help to process the seeds (Grant & Grant 2006), they owen forage on flowers, fruits, and arthropods to supplement their diet, mainly during breeding periods or when seeds decline (Price 1987, Poulin et al. 1994).
Seedeaters of the genus Sporophila comprise a prominent Neotropical bird group common in open grassy areas; they are rela?velysmall bodied (ranging from 6 to 16 g) with stubby, conical bills adapted for feeding on seeds (Ridgely & Tudor 1994, Sick 1997).Seedeaters are owen temporally abundant in a given area in response to seed produc?on (Areta et al. 2013, Rosoni et al. 2019).However, few studies document seedeaters moving locally or regionally tracking seeds (Areta 2012, Areta et al. 2013).For example, some seedeater species breed in southern Brazil during summer while spending the winter in central Brazil, where and when the food supply is adequate (Cintra & Yamashita 1990, Da Silva 1999).
Seedeaters' mobility and diet plas?city may influence how they respond to changes in seed produc?on under anthropogenic pressures (Machado & Silveira 2010).However, the accelerated habitat loss and degrada?on of Brazilian open areas have suppressed grass species, which seeds are key resources for seedeaters (Da Silva 1999, Pozer & Nogueira 2004).
In Brazil, seedeaters are common in grassy areas of several biomes, including the Pantanal (Ridgely & Tudor 1994, Sick 1997).The Pantanal is a large seasonal floodplain comprised of a vegeta?onmosaic of dense gallery forests, patches of dry forests, savannas, and open fields (Silva et al. 1998).In the southern Pantanal, savanna-like habitats are prominent, and one such habitat is known as the Tecoma savanna (dominated by Tabebuia aurea; Silva et al. 1998).Locally, in the spring and summer, up to six seedeaters species may co-occur in this habitat type (Sporophila angolensis, S. caerulescens, S. collaris, S. hypoxantha, S. leucoptera, and S. lineola; J. Ragusa-Ne)o pers.observ.).The Tecoma savanna is seasonally flooded and includes a diversity of grass species (Soares & Oliveira 2009) that seedeaters may exploit according to their availability, as expected for granivorous birds with reduced or no dependence on a par?cular food source (Walker 2007).
We studied the seed produc?on and the feeding habits of six species of Sporophila seedeaters at a Tecoma savanna in the Pantanal.We also analyzed the rela?onship between seed offer (seed abundance and diversity, and the number of grass species seeding) and the number of seedeaters foraging across seasons.2001).We took as a proxy of seed produc?on the total number of seed spikes belonging to all individuals of each species within a plot.Every month, one of us (Ieda M. N. Ilha, hereawer IMNI) counted all seed spikes present at each plot and grouped them per species to es?mate the contribu?on of every species to seed produc?on.Then, our monthly index of seed produc?on resulted from the sum of all seed spikes present in the 60 plots sampled.

METHODS
Seedeaters food resource use.We appraised seedeaters' feeding habits through direct observa?ons of them foraging.
Monthly, IMNI employed 20 h walking the trails described to sample seed produc?on.We randomized the ini?a?on point of the sampling walks and the direc?on to be followed.IMNI carried out these observa?ons in the morning at ?mes when seedeaters were owen feeding, from sunrise to 10:00 h and in the awernoon from 15:00 h to sundown (GTM -04:00).Whenever at least one feeding seedeater was spo)ed, IMNI recorded: a) plant species (whether na?ve or exo?c), b) food resource (flower or seeds), and c) species and the number of seedeaters feeding.When a seedeater captured a small animal (taking the bill length as reference, owen ≤ 1cm), IMNI recorded the taxon as arthropod.Assuming that seedeaters were equally likely to be seen feeding on any food source, IMNI avoided resampling feeding birds during an observa?on period by walking trails in only one direc?on.Seedeaters were not individually marked, therefore, to avoid pseudoreplica?on, IMNI used only the ini?al observa?on of a feeding bird, recording the first inges?on of a specific food item eaten by each seedeater (i.e., sequen?al observa?ons were discarded; Hejl et al. 1990).To improve assessments of the extent of food source use, we provide the percentage of individuals of each Sporophila species feeding on each food item from the total of feeding records of the species (Table 1).

Analyses.
To analyze the rela?onship between the monthly total number of seedeaters feeding and a) seed abundance, b) the number of species producing seeds, and c) seed diversity, we used the Pearson correla?on in which data were logtransformed to improve linearity.Therefore, we took the monthly indexes of seed produc?on, the number of grass species seeding (only those eaten by seedeaters), and the number of seedeaters observed feeding on seeds every month as explanatory variables.We adopted that approach because our interest was to show how seedeaters used the Tecoma savanna as a feeding area instead of showing species-specific rela?onships.Moreover, we did that because no seedeater species intensely used seeds from any grass species (see results below).The three explanatory variables were collinear, so we took the residuals of each regression to assess the presence of confounding effects on the correla?ons with the number of seedeaters.Then, we correlated each factor with residuals of the number of seedeaters.In addi?on, we performed a Power test on the correla?on analyses to assess the poten?al of Type I error, and we assumed that a Power greater than 0.8 (α = 0.05) conformed to correla?on robustness.We correlated the accumulated monthly rainfall with the monthly index of seed offer, as rainfall pulses may posi?vely affect the magnitude of seed produc?on (Pol et al. 2010).We used the Simpson index (D), the reciprocal of Simpson's original formula (Simpson 1949), to describe the food diversity offered to seedeaters.The Simpson index (and its deriva?ves) is sensi?ve to changes in the common species, whereas the more widely used Shannon index is more sensi?ve to changes in rare ones (Peet 1974).We used the Simpson index to minimize the influence of rarely produced foods and stress-out changes in the common ones since granivorous birds owen use abundant food resources (Schluter 1982a(Schluter , 1982b)).We conducted a correspondence analysis (CA) to explore the varia?on in the use of items (arthropods, flowers, and seeds of grass species) by every seedeater species (Table 1).Like other ordina?onmethods, CA a)empts to place similar samples in similar posi?ons in the ordina?onplot.The measure of distance between sam-ples is propor?onal to the Chi-squared sta?s?c.Samples considered for this ordina?onprocedure consisted of the number of individuals of each seedeater species observed feeding on every item/species (Table 1).Here, a species of seedeater took a posi?on in the graph due to its diet dominance (feeding item/species) rela?ve to the diet of other species.

RESULTS
Seed producaon.A total of 16 grass species were recorded seeding in the phenology plots (Figure 1).Seed produc?on started in October, the transi?on from the dry to the wet season, peaking in January-February.From March to May, seed produc?on began to decrease leading to the absence of seeds from June to September (Figure 1).Rainfall totaled 1015.9 mm during the study year, with ca.90% occurring from October to March.Seed produc?on paralleled rainfall along the year (r = 0.88, P = 0.0001, Figure 2A).Four species were important for seed produc?on, making up almost 80% of spike abundance (Setaria geniculata, 37.1%; Andropogon hypogynus, 22%; Panicum laxum, 10.5%; and Paspalum hidrophylum, 9.4%; N = 12076 seed spikes).Much of the monthly seed offer resulted from the produc?on of these four species.On the other hand, species such as Pennisetum nervosum contributed slightly to seed totals (0.5%; Figure 1).In December, S. geniculata spikes dominated the phenology sample comprising 58% of the total produced (N = 2977).Setaria geniculata and A. hypogynus, when grouped, dominated seed produc?on in January (67%, N = 4196 spikes) and February (55%, N = 3242; Figure 1).In March, when seed abundance declined, S. geniculata and P. laxum comprised 61% of seed produc?on (N = 645).Finally, Cynodon dactylon and Eleusine indica, unused by seedeaters, produced minor propor?ons of spikes (2.2% and 0.6%, respec-?vely), being.
Seedeaters consumed seeds from 14 grass species but foraged more owen on five (A.hypogynus, S. geniculata, P. laxum, Axonopus paraguayensis, and P. nervosum, Table 1).Both S. collaris and S. leucoptera were extreme in consuming 13 species, whereas S. angolensis fed on only four, among which A. hypogynus was the most important food item (Table 1).The other three seedeaters' species consumed five to seven species, of which Hymenachne amplexicaulis and Digitaria cuyabanensis were taken in higher propor?ons(Table 1).
Seedeaters foraged on flowers of only six herbaceous species in the transi?on from the dry to the wet season (October and November).Sporophila caerulescens and S. lineola used only one species, while S. angolensis and S. hypoxantha consumed no flowers.On the other hand, S. collaris consumed flowers of three species, whereas S. leucoptera foraged on  five, among which HypEs lappacea was an important food item (Table 1).Seedeaters mostly foraged on arthropods during the dry season.In the wet season, when seeds were plen?ful, they composed 37.5% of the seedeaters' feeding items (N = 30 birds ea?ng).Although all seedeater species ate arthro-pods, S. leucoptera fed on this item more owen (47.0% of arthropods eaten by all seedeaters).This species mostly foraged on arthropods during the dry season, as only 7.1% of their consump?on(N = 14) occurred in the wet season (Table 1).Seedeater species exhibited different associa?ons with species/items eaten.Axis 1 of correspondence analysis (CA 1 = 37.8% of the variance associated with bird species and respec-?ve food items/species) described the progressive decrease of seeds consump?on from S. caerulescens to species that foraged on seeds, arthropods, and flowers (mainly S. leucoptera, and S. collaris; Figure 3, Table 1).On the other hand, axis 2 (CA 2 = 34.1%)separated the peculiar feeding habits of S. hypoxantha from that of S. angolensis.In between these extremes were species, including arthropods, flowers, and seeds (Figure 3, Table 1).
Seedeater occurrence in the Tecoma savanna varied pronouncedly throughout the year.They were abundant during the rainy months, mainly in January-February, and scarce across the dry season.Indeed, the number of seedeaters feeding on seeds paralleled seed produc?on (r = 0.97, P < 0.0001; Power = 0.999; Figure 2A, D).Besides the abundance, the number of grass species bearing seeds varied, peaking in January (12 species) and dropping to zero from June to September.The number of seedeaters feeding on seeds paralleled the varia?ons in the number of grass species exploited by seedeaters (r = 0.96, P < 0.0001; Power = 0.997; Figure 2 B,   D).Finally, seed diversity fluctua?ons were abrupt and correlated to the number of seedeaters feeding in the Tecoma savanna (r = 0.86, P < 0.0004; Power = 0.986; Figure 2 C, D).To reinforce the significance of correla?ons, the residuals of the monthly number of seedeaters feeding had no rela?onwith seed abundance (r = 0.0008, P = 1), the number of frui?ng grass species (r = -0.009,P = 1), and seed diversity (r = 0.0003, P = 1).

DISCUSSION
Seed producaon.In the southern Pantanal, most rainfall occurs from January to March, triggering intense frui?ng in herbaceous species (Tannus et al. 2006).However, in this area, the predominance of sandy soils augments dry condi-?ons,restraining the length of the frui?ng period (Pozer & Nogueira 2004).Indeed, grasses from highly seasonal areas need significant more water input to trigger seed produc?on depending on intense rainfall periods (Schwinning & Sala 2004).This was the case for the grasses, which produced seeds massively only during the we)est period of the year, as documented elsewhere (Pol et al. 2010).Indeed, in the southern Pantanal, the most eaten species are widespread across a gradient of soil humidity (Pozer & Nogueira 2004).Not surprisingly, such species (A. hypogynus, S. geniculata, P. laxum, A. paraguayensis, and P. nervosum) comprised much of seed produc?on, making the bulk of seedeaters' foods.
Seedeaters' feeding habits.In the Tecoma savanna, seedeaters consumed various items, owen comprised less than 10% of any species' diet.Each species consumed seeds of Figure 3. Associa?on between each seedeater species (black circle) and food item eaten on CA (Correspondence Analysis; axis 1 = 37.8% of the variance, and axis 2 = 34.1%).The first three le)ers represent generic and species names of grass species with seeds (see Table 1).
at least three species besides flowers and/or arthropods.They owen overlapped in the intake of seeds from several grass species, for example, in a moderate use of the abundant S. geniculata and A. hypogynus seeds.The most common foragers, S. collaris and S. leucoptera, ate seeds of 14 species in reduced propor?ons,far from specialized seed use (Areta & Cockle 2012).Moreover, A. hypogynus seeds were the main food item for both S. collaris and S. leucoptera, although forming less than 17% of their exploited items.Then, in principle, the Tecoma savanna emerges as a feeding area where seedeaters foraged on a diverse seeds.The use of many seed species by seedeaters may be a strategy to increase nutri?onal reward because varied seeds may fulfill their requirements in terms of minerals and essen?al amino acids (Díaz 1996, Cueto et al. 2006).Further studies on the content of consumed seeds should clarify their contribu?onto the nutri?onal spectrum of seedeaters in this peculiar area in southern Pantanal.
Seedeaters were abundant in the we)est months when seed supply was abundant, and these birds foraged intensely on seeds of several grass species.Conversely, they were rare in the driest months when seed supply was reduced, and they owen fed on flowers and arthropods.This varia?on in abundance resulted from the high mobility of Sporophila seedeaters tracking seasonal seed patches (Da Silva 1999, Areta et al. 2013).Therefore, the parallelism between seed offer and the number of seedeaters feeding suggests these birds moved to the Tecoma savanna due to the abundant seed supply instead of tracking par?cular seeds (Areta et al. 2013).Besides feeding, seedeaters might also be breeding in the Tecoma savanna.Indeed, seedeaters' movements toward feeding areas during spring and summer coincide with their breeding period, mainly for seedeaters species with the southern breeding distribu?on(Franz & Fontana 2013, Repenning & Fontana 2016, Rosoni et al. 2019).
In seasonal cerrado areas, blossoms are common during the dry and transi?on to the wet season (Tannus et al. 2006).
The herbaceous plants' flowering during the dry season emerged as an alterna?ve resource for seedeaters.During the dry season, seeds became increasingly scarce, and the few persistent seedeaters fed on alterna?ve items to overcome this period (Table 1).In this respect, studies have documented the importance of flowers for granivorous birds during the dry season (Smith et al. 1978, Tebbich et al. 2004).Despite their low nutri?onalcontent, consump?on of flowers might be advantageous due to their abundance (Symes et al. 2008), when seedeaters more owen ate arthropods.Thus, lacking seeds, the remaining seedeaters, at least partly, switched their feeding habits un?l seeds became plen?ful in the Tecoma savanna, similar to that documented in the Atlan?c Forest (Areta et al. 2013).
To conclude, besides seed abundance and diversity, the number of seed species consumed influenced the seedeaters' use of the Tecoma savanna.This highlights the importance of a rich collec?on of sca)ered food patches for the abundance of seedeaters.Thus, the conserva?on of this bird group in the Pantanal depends on preserving key habitats having food resources, at least as documented here.The subs?tu?on of the na?ve for exo?c pastures (Seidl et al. 2001) impoverishes feeding areas that provide the adequate food supply seedeaters require (Da Silva 1999, Machado & Silveira 2010).
In areas where na?ve grasses were extensively subs?tuted for exo?c ones, seed diversity decreased, and seedeaters declined (Da Silva 1999, Filloy & Bellocq 2006).Another serious concern emerges from the recent occurrence of more frequent and extensive fires.Indeed, the intense degrada?on of feeding areas, among other reasons, may cause a severe decline in seedeater popula?ons in the Neotropics (Filloy & Bellocq 2006).Management plans for seedeaters should aim at preserving wide areas, including abundant and diverse seed patches in sites where seedeaters also breed during the wet season (Areta et al. 2013, Franz & Fontana 2013, Repenning & Fontana 2016, Rosoni et al. 2019).Further studies may document the use of the Tecoma savanna as a stopover for the species studied here, which also breed to the south of our study area (Franz & Fontana 2013).
Study site.This study was carried out in the South Pantanal flood plain in a Tecoma savanna cut by the Miranda River (19°35'S, 57°2'W, al?tude ± 100 m, Municipality of Corumbá, State of Mato Grosso do Sul).The vegeta?on in the area is a mosaic conformed by dense gallery forest in the Miranda River, patches of deciduous forest interspersed with open grassy areas, and palm savannas (dominated by Copernicia alba) and Tecoma savannas dominated by T. aurea; Soares & Oliveira 2009).The largest area of the Tecoma savanna in the South Pantanal covers 63,779 ha (Silva et al. 1998).Tabebuia aurea trees are 4-8 m height, although some individuals may reach 12 m.They occur on small mounds (± 0.5-1.0m) interspersed with open grassy areas.Individuals of Byrsonima orbignyana are owen intermixed with T. aurea, although in smaller numbers (Soares & Oliveira 2009).Annual rainfall is around 1000 mm, mostly from November to March (70-80%, wet season [source: meteorological service of Corumbá Interna?onal Airport, Brazil]).In this area of Pantanal, flood pulses typically occur from January to March.During floods, standing water in the Tecoma savanna is up to 1.0 m.Seed producaon.To assess seed offer in the Tecoma savanna, we established three 2 km long trails, 1 km apart.Every trail had 20 plots (2 x 2 m, 100 m apart) designed to sample monthly seed produc?on (from December 2000 to November

Figure 1 .
Figure 1.Total monthly (Dec 2000 -Nov 2001) number of seed spikes produced by 16 grass species in 60 plots of the Tecoma savanna (southern Pantanal, State of Mato Grosso do Sul, Brazil; 2000 and 2001).

Figure 2 .
Figure 2. From top to bo)om: A) Total seed produc?on (seed spikes monthly recorded [see methods], and monthly rainfall (mm; line)), B) Monthly number of species with seeds, C) Monthly seed diversity, and D) Monthly number of Sporophila seedeaters feeding on seeds (N = 174 [the number of individuals feeding on arthropods and flower are not represented]) in the Tecoma savanna (southern Pantanal, State of Mato Grosso do Sul, Brazil; Dec 2000 -Nov 2001).