Complementarity between herbaceous and woody plants in providing resources for bees 1 in a semi-arid tropical climate : insights for conservation 2 3


 The distribution of floral resources in time and space varies according to vegetation strata and to the influence of precipitation and photoperiod on flowering patterns. There are scarce studies jointly investigating the spatio-temporal distribution of plant resources for bees and at the community level, which are critical for planning strategies for their conservation. Here we investigated how herbaceous and woody strata integrate to provide resources (e.g. nectar, pollen, oil) for bees inhabiting fragments of Caatinga dryforest in the northeastern Brazil. We assessed herbaceous and woody strata species composition, abundance and their floral phenodynamics for 18 months in three 0.5-ha plots. We registered bee’s plant visits and collected data on abiotic factors to examine how they correlate to flowering patterns. Bees visited approximately 80% of the plant community, composed by a variety of growth (e.g. trees, shrubs, lianas) and life (e.g. phanerophytes, chamaephytes, hemicryptophytes, and therophytes) forms. Water availability was the main variable correlated to arboreal and herbaceous flowering, but photoperiod also played a role in annual and interannual flowering patterns. Both woody (52%) and herbaceous components (48%) similarly contributed to nectar and pollen supplies, while trees were also a source of resin. Herbaceous life forms flowered sequentially during the rainy season, keeping continuous supply until the beginning of the dry season. While growth forms in the woody stratum provided resources relatively constantly throughout the year. The phenological complementarity between different life and growth forms should be considered in bee conservation strategies, aiming to optimize resource availability throughout the year.

The study area is located in the northeastern Brazil semi-arid region (Fig. S1A), known as the   (Fig. S2A). During the rest of the year, there is water deficiency in the area that can last from 7 to 11 months. The 175 temperature is high, with an annual average of 28.3 °C (Fig. S2A), which results in high 176 potential evapotranspiration (on average, 1,500 mm per year).

178
The Caatinga vegetation shows a woody stratum mostly composed of deciduous tree species, 179 many of them bearing thorns. The treetops do not form a real canopy, whose average opening 180 is 18% in the rainy season and 89% in the dry season when the trees lose almost all their 181 leaves (Fig. S3). This canopy opening facilitates the growth of an herbaceous layer composed 182 mainly of annual plants (therophytes), whose richness can be two to three times higher than 183 that of the woody component (see Costa et al. 2007). Succulent plants, especially Cactaceae, 184 are common in both herbaceous and woody strata. As the study area is located near the Curu 185 River, it also has floristic elements of riparian vegetation (see Moro et al. 2015). We selected three Caatinga fragments with more than 30 years of natural secondary 190 regeneration and installed a 0.5 ha plot (50x100 m) in each, considering 2-km equidistance 191 between them (Fig. S1B). This distance between plots was required to ensure independence 192 when sampling bee data, as 1-2 km radius is the foraging area for most bee species (Greenleaf 193 et al. 2007). We collected data on floral phenodynamics for 18 months (August 2017-194 February 2019) in these three established plots.      We collected data on rainfall (mm/month), photoperiod (h/day), and water availability in the and drying on floral phenodymamics and bees' resource supply. Rainfall data were obtained 249 from the meteorological station installed on the farm (Fig. 1A, Fig. S1B). The rainfall 250 distribution pattern from September 2017 to February 2019 followed the historical record for 251 the past 50 years (Fig. 1A, Fig. S2A). Although the annual rainfall for 2018 and 2019 felled      We then built correlograms (cross-correlation graphs) to visualize pairwise cross-correlations. 319 In the graph, the negative lag indicates that the phenological event occurred after the change   (Table S1).  The morphospecies with the highest cover percentage were Ruellia paniculata (15%),  (Table S1).

350
The total flora analyzed in the three plots included 82 species (44 herbaceous and 38 woody).

351
Of these, 69 were visited by bees: 40 (90%) herbaceous morphospecies and 29 (76%) woody 352 species, representing 84% of the total species found in the community. However, in the 353 individual analysis by plot, the woody and herbaceous strata showed similar relative 354 importance in the proportion of species/morphospecies visited by bees (Fig. 2, Fig. S4).

367
Nectar and pollen availability were analyzed separately per growth and life forms by stratum. 368 We found similar proportions of species in the two strata (51% woody and 49% herbaceous) 369 providing pollen and a relatively higher proportion of woody species providing nectar (59% 370 woody and 41% herbaceous) for bees (Fig. 2). This difference reflects a great proportion of 371 tree and liana species providing nectar in the woody stratum. We recorded bees visiting lianas only for nectar (Fig. 2). Among herbaceous plants, the proportion of chamaephytes and 373 therophytes species providing nectar was similar (~ 20%), but the proportion of therophytes 374 providing pollen was higher (29%) than chamaephytes (12%). Hemicryptophytes only 375 provided pollen (Fig. 2). Only the woody component (trees) provided resins, and a therophyte 376 species provided oils (Table S1). to the dry season in 2018 (Fig. 3, Fig. 4). The woody stratum showed activity in 15 out of the 384 18 months and the herbaceous stratum in 10 out of the 18 months analyzed. Simultaneous 385 activity of both two strata occurred in periods of four to six months, which defined the peak of 386 floral resources (Fig. 3, Fig. 4). The period of the highest resource availability (more species 387 active and more intense flowering) was shorter for nectar (April to August 2018, Fig. 3, Fig.   388 4) than the period of pollen availability, which lasted from January to August 2018 (Fig. 3, 389 Fig. 4). Oil was available only in June 2018, with a cover percentage below 1% of the plot 390 (Table S1).  (Fig. 3). Trees were the main resource supply in the 399 rainy season and in the beginning of the dry season, when we recorded higher species 400 flowering activity (Fig. 4). While shrubs and lianas stood out during the last months of the dry 401 season (October-December) and in the transition between the dry and rainy seasons 402 (December-January) (Fig.4). Together, these different growth forms kept flowering supply for (55% of species) (Fig. 3, Fig. 4). The analysis by life forms revealed that the first species 409 bearing flowers were hemicryptophytes (maximum activity in February), followed by 410 terophytes (maximum activity in May), and chamaephytes (maximum activity in July), whose 411 flowering extended until September, with R. paniculata being the only chamaephyte species 412 active in that month (Fig. 3, Table S1). lag), indicating a rapid response to the first rains, i.e., those that started before the peak of 428 water availability (maximum rainfall and soil moisture) of the rainy season (Fig. 5). Their 429 correlation with photoperiod was positive but not significant. water availability but not in relation to the decrease in photoperiod (Fig. 5).

437
Hemicryptophytes did not show a time lag in relation to the peak of water availability but 438 showed a time advance (three months) in relation to the minimum photoperiod (Fig. 5). 439 Finally, therophytes showed an intermediate pattern between cryptophytes and 440 hemicryptophytes, with two months lag in relation to the peak of water availability and one 441 month in advance before the minimum photoperiod (Fig. 5).        Figure S4 shows the individual data per plot.

Assemblage
Life forms Growth forms