Functional groups of woody species in riparian forest in the Brazilian semi-arid region: phenological and morphophysiological evidence


 Relationships between morphophysiological and phenological traits of functional groups of tree species from seasonally dry tropical environments are reasonably known. However, such information is scarce for riparian forests in semi-arid environments, especially regarding the representativeness of these groups in the community. This study aimed to evaluate the relationship of the morphofunctional characteristics of leaves and stem in the identification of functional groups and how these traits vary in these groups in a riparian forest in the Brazilian semi-arid region. Traits of leaf phenology (specific leaf area, leaf area, leaf thickness, and leaf dry matter content), wood density, and amount of saturated water in the stem and their association with soil water availability, relative air humidity, precipitation, temperature, photoperiod, and vapor pressure deficit were evaluated in 23 species. Cluster analysis showed three functional groups: deciduous low wood density species (9%), deciduous high wood density species (48%), and non-deciduous species, which included semideciduous (17%) and evergreen species (26%). Deciduousness was positively related to specific leaf area and negatively related to leaf dry matter and leaf area. Deciduous and semideciduous species (55.1% and 7.5% of the individuals in the community, respectively) showed a strong dependence on abiotic factors linked to precipitation and temperature, while the evergreen species showed a strong dependence on the vapor pressure deficit to trigger their phenophases. It was evident that the studied functional traits influenced the leaf habit, which is essential to determine the abundance of functional groups of woody species from riparian forests in the semi-arid region.


Introduction
The phenology of seasonally dry environments is determined by the duration and intensity of seasonal drought (Kushwaha et al. 2011), being related to the environmental factors and morphophysiological traits of plants found in these regions, especially those related to water control, such as stem and leaf traits (Fu et al. 2012). Phenological patterns are strongly in uenced by water use and conservation strategies, which in turn are controlled by the combination of the physiological and architectural traits of the plant (Butz et al. 2017), together favoring the formation of functional types ). However, the complex interaction between phenology and morphofunctional traits is still little explored, especially in riverbank ecosystems (riparian forests) located in semi-arid regions, which are targets of strong anthropogenic pressure (Hughes and Rood 2003). In addition, the representativeness of functional groups in these forests is unknown.
In semi-arid regions, phenological patterns are highly in uenced by precipitation (Amorim et al. 2009; Leite  However, studies have shown that besides being correlated with abiotic factors, the functional and morphological characteristics of plants are also correlated with their phenology, providing mechanisms that allow them to deal with strong seasonality (Reich and Borchert 1984;Borchert 1994; Borchert and Rivera 2001;Westoby et al. 2002;Poorter and Markesteijn 2008). These mechanisms of resistance to drought re ect on the use and control of water by plants, for example, regarding their ability to store or access soil water (Singhand Kushwaha 2016).
The interaction between these complex mechanisms allowed the emergence of different drought resistance strategies, which can be evaluated from functional traits. Deciduousness, for example, which allows the plant to avoid drought (Borchert 1994;Chaturvedi et al. 2013), is negatively related to leaf dry matter content and leaf area and positively related to speci c leaf area (Poorter and Markesteijn 2008). Other features, such as deep roots, greater stomatal control, smaller leaf area (Slot and Poorter 2007) and speci c leaf area, and greater leaf thickness also reduce water loss from plants (Chaturvedi et al. 2013). In addition to these characteristics, a low wood density allows the plant to store large amounts of water in its stem and thus avoid large variations in water potential in regions with low water availability (Lima et al. 2012). Therefore, the interaction between different plant water-use strategies, such as between stem (Chave et al. 2009) and leaf (Wright et al. 2004) functional traits, are essential to evaluate the phenological behavior of the species and, consequently, of the functional groups (Borchert 1994;Lima et al. 2012;Oliveira et al. 2015).
Functional groups, in turn, help to understand the functioning of ecosystems by simplifying their oristic complexity (Prentice et al. 1992) and to monitor the effects of global changes on community structure and ecosystem processes (Lavorel et al. 2007). In addition, since the seasonal variation in water availability in the plant is one of the determining factors affecting phenology (Borchert 1994), the use of functional groups allows testing how functionally equivalent species respond to water availability and also allows comparison with other similar forests (Lavorel et al. 2007). However, although some studies have identi ed general patterns of plant functional types, the available data represent only a part of the wide range of ecosystems found in nature (Castellanos-Castro and Newton 2015).
In seasonally dry tropical environments, for example, functional groups related to plant phenology and morphofunctional These studies evidenced the predominance of deciduous species with high wood density certainly due to the better hydraulic resistance provided by higher wood density, while evergreen and deciduous species with low wood density were less abundant. In this case, it is possible that the evergreens are more limited to microsites where they can develop the root system and access underground water (Sobrado and Cuenca 1979;Wright et al. 2021). However, there is little information on the in uence of morphofunctional traits on the con guration of functional groups in riparian forests in seasonally dry environments due to deforestation and the almost complete suppression of these forests (Nilsson and  This study aimed to evaluate the relationship of the morphofunctional characteristics of leaves and stem in the determination of the functional groups under conditions of a riparian environment, seeking to understand the possible mechanisms behind the patterns found. For this, we analyzed how abiotic factors in uence the morphofunctional traits and the phenology of these groups. Due to the greater water availability in riparian forests even when located in the semiarid, we sought to nd functional traits that may allow the maximization of the use of this resource. Thus, we hypothesized the predominance of species with high wood density, which can support higher levels of water stress, as well as with leaves presenting characteristics associated with greater longevity, such as smaller speci c leaf area and higher biomass content. Therefore, we expected greater abundance of evergreen species. It is possible that deciduous species are also well represented in the community, although retaining their leaves for a longer period than in drier environments and presenting high values of speci c leaf area and low values of leaf area, leaf thickness, and leaf dry matter content.

Study area
The study was carried out in a riparian forest located in the banks of the Boi Morto stream, Mata da Pimenteira State Park (07°53'21" and 07°57'36"S; 38°18'42" and 38°17'7"W), municipality of Serra Talhada, Pernambuco State, Brazil (Silva and Almeida 2013). The vegetation in the park is mostly deciduous, thorny, and arboreal-shrubby, which is typical of the caatinga (Melo et al. 2013). The soils found in the park are cambisols associated with red-yellow podzolics, planosols, and solonetz, and the area is situated in an altitude ranging from 500 to 700 meters (Silva and Almeida 2013). In the riparian forest area, the vegetation has a larger size, greater diversity (A.P.A. Silva, unpublished data 2018), and longer leaf longevity. The climate of the region according to the Köppen system is classi ed as BSwh' (hot semi-arid), with average annual rainfall of 653.2 mm and average temperature of 26°C (Silva and Almeida 2013). The rainy season starts in December and extends until May, concentrating 69.5% of the annual precipitation (Silva and Almeida 2013), although presenting great interannual variability (Sampaio 1995). Data of precipitation, relative air humidity, and average temperature were obtained from a meteorological station located 11km away from the park through the Pernambuco Water and Climate Agency (www.apac.pe.gov.br/). The photoperiod was obtained from Lammi (2015). During the rst year of the study, the rainfall was 611.8 mm, thus being slightly below the historical average, and there was no rainfall during the last six months of the study period. In addition, during the rst year of study the mean temperature was 25.9 ºC, with a minimum of 21.6°C and a maximum of 29.1°C; the average relative humidity was 61.95%, with a minimum of 44% and a maximum of 78.5%; the gravimetric soil water content ranged from 1.04-6.81%; and the annual variation of the photoperiod was 55 minutes (Fig. 1). In this period, the vapor pressure de cit (VPD) ranged from 0.63 Kpa to 2.24 Kpa, with an average of 1.33 Kpa. During the last semester of the study, the average temperature was 26.7°C, with a minimum of 24.5°C and a maximum of 29.5°C; the average relative humidity was 50.4%, with a minimum of 42% and maximum of 59.6%; the gravimetric soil water content ranged from 0.7-1.6%; and the average VPD was 1.8 Kpa, with a minimum of 1.3 Kpa and a maximum of 2.4 Kpa. The annual variation of the photoperiod was 50 minutes (Fig. 1).

Data collection and analysis
Phenology For phenological monitoring, a stretch of one kilometer of riparian forest on each side of the bank, with each stretch presenting a width of ve meters (A.P.A. Silva, unpublished data 2018) were established as the experimental plots. In these areas, 23 species were selected, representing 64% of the total number of species species and corresponding to 89% of the total number of individuals in the community (APA Silva, unpublished data 2018). Out of these, nine adult individuals of each species with stem diameter at ground level > 3cm were marked , considering taller individuals with apparent canopies for better visualization. Phenological observations were carried out monthly for 18 months, between June 2017 and November 2018. Budding phenophases, leaf shedding (by observing yellow or brown leaves and when leaves were partially or totally lost by the individual) and proportion of leaves in the canopy (all leaves in the canopy) were observed. The intensity of phenological events was estimated according to Fournier (1974), with the intensity of the phenophases being quanti ed from zero to four for each individual (Fournier intensity index), where zero represents the absence of phenophase, one represents a phenophase between 1-25%, two represents a phenophase between 26-50%, three represents a phenophase between 51-75%, and four represents a phenophase between 76-100%.
The phenophase intensity for each species is given through the formula: (∑ Fournier / 4N) * 100, where ∑ Fournier is the value of the Fournier index for each individual and N is the number of individuals of the species. The deciduousness of species was based on Williams et al. (1997), considering as evergreens the especies that remained with more than 50% of leaves in their canopy during the 12 months, as semi-deciduous the species that lose more than 50% of leaves in their canopy in the dry season, although not completely losing their leaves, and as deciduous the especies that lose their leaves during the dry season and remain lea ess for at least a month.

Wood density
Wood density was determined from ve individuals of each species through samples of the branch with at least three cm in diameter. The samples had their barks removed and then were immersed in water for ve days for maximum saturation (Chave et al. 2009). After this period, the samples were removed and remained out of the water for 10 minutes to remove excess water. Subsequently, for measuring the displacement of liquid, each sample was immersed in another container with distilled water to determine their volume. Then, to measure dry matter, the samples were dried in an oven at a temperature of 103 ºC for 72 hours and weighed (Truguilho et al. 1990). Basic wood density was determined through the relationship between the dry matter of the sample and its saturated volume (g/cm 3 ) (Pérez-Harguindeguy et al. 2013). The amount of saturated water in the stem (Qwsat. %) was obtained by the formula: Qwsat = 100 (V-DM)/DM; where V is the displaced volume and DM the dry matter (Borchert 1994). Species were considered as having low wood density when presenting a density lower than 0.5 g/cm 3 and as having high wood density when presenting a density higher or equal to 0.5 g/cm 3 (Borchert 1994).

Leaf traits
The functional traits of leaf thickness, leaf area, speci c leaf area, and leaf dry matter content were determined according to the protocol by Pérez-Harguindehuy et al. (2013), who recommended collecting ten leaves per individual in ve individuals of each species. Leaf thickness was obtained using a stainless hardened digital caliper (MTX-316119). To obtain the leaf area (mm 2 ), each leaf was digitized using a atbed scanner and their dimensions (not including the petiole) were determined using the software Image J ®. The speci c leaf area was calculated using the ratio between the leaf area (mm 2 ) and the dry weight (mg) obtained in an oven at 70 ºC for 72 hours. Leaf dry matter content was obtained by dividing the dry weight of each leaf by its wet weight (mg.g − 1 ).

Soil water content
Soil samples were collected monthly on the same collection dates used for phenological data. Collections were carried out at a depth of 20 cm using an auger at ten random points ( ve on each side of the banks). To avoid moisture loss, the samples were stored in closed plastic bags for subsequent weighing in the laboratory to obtain the wet weight and dry weight after drying in an oven at 105 º C for 24 hours. The gravimetric soil water content was obtained through the equation: % (G): (wet weight -dry weight) * 100/ dry weight (Embrapa 1997).

Statistical analysis
For multivariate analysis, a data matrix was elaborated containing the species as descriptors and the phenological data and functional traits as objects. Box-Cox transformation was used with λ = 0.5 (Box and Cox 1964). Then, a paired group with a Euclidean similarity measure was used as the method for linking the objects and the descriptors in a cluster. Then, the average value of the group was used, and non-metric multidimensional scaling (MDS) analysis was performed from the triangular matrix in order to identify the main groups to which the descriptors are ordered (Mardia et al. 1980).
After identifying the functional groups, principal component analysis (PCA) was performed including the functional traits, whose groups were compared using the ANOSIM test with 9999 permutations to verify which morphofunctional traits had the greatest in uence on the separation of the functional groups. Subsequently, canonical correlation analysis (CCA) was performed with the phenological data of each functional and abiotic group for each vegetative phenophase (Clarke 1993), in order to identify which abiotic variables had the greatest in uence on the triggering of phenophases.

Page 6/20
Out of the studied species, 57% were deciduous, 26% evergreens, and 17% semideciduous (Table 1). Thus, 43% of the species remained with their leaves throughout the year, although with different leaf fall intensities. Basic wood density (WD g/cm³) and water storage capacity in the stem (Qwsat%) varied between species and were inversely proportional (r²= 0.775) ( Table 1). WD ranged from 0.28 g/cm³, as recorded for Jatropha mollissima, to 0.81 g/cm³, as observed for Libidibia ferrea, while the Qwsat ranged from 20.18-222.6% of the dry weight (Table 1). Cluster analysis considering leaf functional traits, vegetative phenology, basic wood density, and water storage capacity in the stem showed the following three functional groups: deciduous species with low wood density (DLWD), deciduous species with high wood density (DHWD), and non-deciduous species. The latter group comprised semideciduous (SD) and evergreen (EG) species (Fig. 2). The formation of these groups was con rmed by the multidimensional scaling analysis (MDS) (Fig. 2).
The DLWD group included 9% of the studied species. Regarding the phenological behavior, the DLWD group sprouted before the rst rains and started to shed leaves at the end of the rainy season, with the species remaining with less than 15% of their leaves for ve months in the dry season. (Fig. 3). The DHWD group started budding with the rst rains and leaf fall increased early in the dry season. However, this group remained with less than 15% of their leaves only during the last two months of this season (Fig. 3). In the third group, the semideciduous species also started budding with the rst rains, although the peak of leaf fall occurred at the height of the dry season, just before the wet season. (Fig. 3). On the other hand, evergreen species exchanged leaves constantly, with budding and leaf fall occurring simultaneously throughout the year (Fig. 3).
According to the result obtained by PCA (Fig. 4), it can be observed that the variables that most in uenced the species were LA, Qwsat, WD (main component 1 -PC1), and SLA (main component 2 -PC2), respectively. That is, the leaf and wood functional traits had a greater in uence on the distribution of species in the PCA than phenological traits. In addition, it can be stated that species that were close in distribution were similarly in uenced by such variables. PC1 explained 57.7% of the group variation and PC2 explained 28.8%.
The canonical correspondence analysis (CCA) (Fig. 5) demonstrated with signi cative difference (p<0.001) for all analysis that budding in the DLWD group was mainly correlated to increased photoperiod, temperature, and vapor pressure de cit (VPD). In the DHWD and SD groups, the increase in relative air humidity, soil moisture, and precipitation served as a trigger for this phenophase. Budding in EG was shown to be inversely related to VPD.
The increase in photoperiod, temperature, and VPD was positively related to leaf fall for the DLWD and SD groups. The conservation of the canopy of evergreen species was also related to those variables (Fig. 5). In DHWD species, the maintenance of leaves in the canopy was mainly in uenced by the increase in relative air humidity, soil moisture, and precipitation, while leaf fall was mainly triggered by the decrease in these variables (Fig. 5).

Discussion
As expected, the analyzed functional characteristics were good predictors for the determination of functional groups in the riparian forest studied, demonstrating that the traits of leaf area, wood density, amount of water in the stem, and ). This behavior can be considered a strategy to anticipate the growth period, maximizing the production of photosynthates (Rivera et al. 2002). The literature reports that this phenological pattern is caused by the large amount of water that these species store in their stems (Borchert 1994;Borchert and Rivera 2001;Lima and Rodal 2010;Lima et al. 2012), which is used for the subsequent production of owers and fruits (Rivera et al. 2002). In addition, the high leaf area index of these species allows a greater amount of light energy, maximizing assimilation in the short period when the species remains with leaves (Rivera et al. 2002). Furthermore, since DLWD present a lower investment in wall material for the cell wall structure, their hydraulic system becomes more vulnerable to collapse when exposed to stressful conditions Deciduous species with high wood density, which corresponded to 48% of the total, were highly dependant on precipitation, soil water availability, and relative air humidity to trigger their phenophases for presenting low water storage capacity in their stems, which is similar to what is reported in the literature (Borchert 1994 The high percentage of non-deciduous species (43%), including semideciduous and evergreen species, and the fact that they remained with leaves in both the dry and rainy seasons, may be related to greater water availability in the riparian environment, which creates a milder microclimate (Azevedo et al. 2014), together with the high water holding capacity of the soil and the greater resistance of the tissues of the stems and leaves. These species presented higher values of wood density, and it is known that this trait favors a hydraulic structure that is more resistant to large variations in water potential during the dry season, allowing higher tensions in the xylem water column and minimizing possible damage to the hydraulic system, such as embolism ( Thus, it can be concluded that the functional groups determined in this study are similar to those found in other areas of seasonally dry environments, although the abundance of species that make up each group varies according to the water availability of the location. As an example, the in uence of VPD on evergreen species is highlighted, while precipitation and soil moisture act on deciduous and semideciduous species. The set of wood and leaf morphofunctional traits were essential to explain the phenological behavior of each functional group, as well as the speci city of each group regarding environmental variables. As predicted, there was a predominance of species with high wood density, which is an important attribute for a more e cient exploration of soil water, minimizing damage to the hydraulic system (Hacke et al. 2001; Wright et al. 2021). The prevalence of this functional trait seems to optimize the presence of these species in the riparian environment in the semi-arid to the detriment of species in the low wood density group, which was represented by only two species, a much lower value than that observed in other seasonally dry environments (Lima et al. 2012;Oliveira et al. 2015). This is corroborated by the greater abundance of evergreen and semideciduous species, which presented higher values of wood density and structurally more resistant leaves. It can also be stated that although the studied forest is located in a seasonally dry environment, almost half of the species had their phenology in uenced by environmental factors inherent to the riparian environment. Thus, it can be concluded that the morphofunctional and phenological characteristics of woody species from riparian forests are strong indicators of functional groups and should be considered in forest restoration plans. Figure 1 Monthly abiotic data of Mata da Pimenteira State Park, Serra Talhada, PE, Brazil. VPD: Vapor pressure de cit.