List of abbreviations
BA branch area;
BE leaf and berry expansion phenophase;
BH berry maturation and harvest phenophase;
ILA individual leaf area;
IR irrigation;
LA leaf area;
LAI leaf area index;
NI rainfed conditions;
PPFD photosynthetic photon flux density.
Introduction
Among about 130 species of genus Coffea (Davis and Rakotonasolo, Reference Davis and Rakotonasolo2021), only two dominate the entire coffee trade, Coffea arabica – Arabica coffee and C. canephora – Robusta coffee (Cassamo et al., Reference Cassamo, Mangueze, Leitão, Pais, Moreira, Campa, Chiulele, Reis, Marques, Scotti-Campos, Lidon, Partelli, Ribeiro-Barros and Ramalho2022), both originated from African continent (Davis et al., Reference Davis, Govaerts, Bridson and Stoffelen2006). In its natural evolution, C. arabica became allotetraploid as a result of a natural cross of two parental diploid species: C. canephora and C. eugenioides (Bardil et al., Reference Bardil, Almeida, Combes, Lashermes and Bertrand2011). Today coffee is produced in tropical and subtropical areas in about 80 countries over the world, with Brazil being the greatest producer of C. arabica (CONAB, 2023). Wild arabica coffee has been harvested for millennia and used as a food, while cultivated only for several hundred years for being used as a beverage (Davis et al., Reference Davis, Challa, Williams, Baena, Gole and Moat2018). Arabica and Robusta coffees are unusual among major crop plants, analyzing the short period over which domestication occurred, and the level of their domestication, i.e., genetic distance from wild to cultivated types is short, except in the cases of interspecies hybrids (Leroy et al., Reference Leroy, Ribeyre, Bertrand, Charmetant, Dufour, Montagnon, Marraccini and Pot2006). In Brazil, the plant breeding of C. arabica toward high-productive genotypes started in the 1930s (Guerreiro-Filho et al., Reference Guerreiro-Filho, Ramalho and Andrade2018). However, berries of wild genotypes of both coffee species can be harvested and processed to produce coffee with high chemical (Rakocevic et al., Reference Rakocevic, dos Santos Scholz, Pazianotto, Matsunaga and Ramalho2023) and organoleptic (Geeraert et al., Reference Geeraert, Berecha, Honnay and Aerts2019) qualities, which are similar or even indistinguishable from cultivated and domesticated types (Davis et al., Reference Davis, Chadburn, Moat, O’Sullivan, Hargreaves and Lughadha2019).
C. arabica is a relatively small evergreen tree (up to 5–9 m tall) that originated from highlands of Afromontane forests of Southwest Ethiopia and the Boma Plateau of Sudan in Africa (Anthony et al., Reference Anthony, Bertrand, Etienne, Lashermes and Kole2011). The flowering and fruiting phases occur in a two-year period, biennial cycle (Majerowicz and Sondahl, Reference Majerowicz and Sondahl2005). Arabica coffee phenology is complex, englobing six vegetative and reproductive phenophases, some of them occurring simultaneously: (1) branch and flower bud induction and formation; (2) maturation of flower buds; (3) anthesis; (4) berry and leaf expansion; (5) berry maturation; and (6) winter rest (Camargo and Camargo, Reference Camargo and Camargo2001). Coffea spp. tree architecture can be described by the Roux’s model, characterized by continuous growth of all axes and dimorphism of axes, being orthotropic in 1st order, and plagiotropic at higher orders – 2nd to 5th branching order (Hallé et al., Reference Hallé, Oldeman and Tomlinson1978). The 4th and 5th branching orders appear three to four years after planting or after severe training (Rakocevic and Androcioli-Filho, Reference Rakocevic and Androcioli-Filho2010). At each node of orthotropic axis, two opposite plagiotropic axes of 2nd order are usually emitted, even though sometimes no branch, or just one, develops (Matsunaga et al., Reference Matsunaga, Tosti, Androcioli-Filho, Brancher, Costes and Rakocevic2016).
In a scope of ongoing climate change, synchronous climate hazards pose an increasing challenge to global coffee production due to more frequent warm and dry events related to El Niño-like sea-surface temperatures in the Pacific Ocean, or the frequency and severity of frosts in some continental and maritime subtropical regions related to La Niña-like signatures (Richardson et al., Reference Richardson, Kath, Byrareddy, Monselesan, Risbey, Squire and Tozer2023). Drought is considered one of the major environmental factors limiting coffee plant growth and yield in most coffee-growing areas (DaMatta and Ramalho, Reference DaMatta and Ramalho2006). Under water deficit, reduction in growth has been highlighted in C. arabica, C. canephora, and C. liberica, but the highest reductions for tree height, leaf length, and leaf width, and the most significant physiological modifications were observed in C. arabica (Vu et al., Reference Vu, Park, Tran, Bui, Vu, Jang and Kim2018). Therefore, irrigation is highly recommended for great part of coffee areas in Brazil, as rainfed conditions negatively impact plant growth traits, such are relative growth rate or leaf area (Dias et al., Reference Dias, Araujo, Moraes, Barros and DaMatta2007; Souza et al., Reference Souza, Guimarães, Colombo, Sant’Ana and Castanheira2016), but no differences are observed in the number of leaves or number of 2nd order branches due to irrigation (Chemura, Reference Chemura2014). C. arabica genotypes show reduction in physiological and productive traits when cultivated under rainfed compared to irrigated conditions in Brazilian Cerrado, a very dry environment, with Iapar 59 cultivar highlighted as the most productive and a wild Ethiopia accession ‘E237’ as the less productive one (Silva et al., Reference Silva, Junior, Ramos, Rocha, Veiga, Silva, Brasileiro, Santana, Soares, Malaquias and Vinson2022). As a perennial crop, coffee plants exposed to multiple drought cycles can express differential acclimation capabilities, associated with their defense mechanisms, allowing them to be maintained in a ‘state of alert’ to successfully deal with new drought events (Menezes-Silva et al., Reference Menezes-Silva, Sanglard, Ávila, Morais, Martins, Nobres, Patreze, Ferreira, Araújo, Fernie and DaMatta2017).
As C. arabica is one species that originated from a deep forest shade, some studies have shown that only coffee beans produced from natural forests of Ethiopia are qualified as specialty coffee following the Specialty Coffee Association of America’s standards, suggesting that the most important drivers of deteriorating coffee quality include light level and biotic interactions (Geeraert et al., Reference Geeraert, Berecha, Honnay and Aerts2019). In shaded Arabica coffee cultivated in agroforestry system, the use of shallower soil water prevailed during wet season in comparison to drier periods (Muñoz-Villers et al., Reference Muñoz-Villers, Geris, Alvarado-Barrientos, Holwerda and Dawson2020). Under natural shading (∼50% light reduction), Arabica coffee plants show enhanced photosynthesis and increased leaf area index (LAI), resulting in a better performance than under full sun light (Bote and Struik, Reference Bote and Struik2011). Also, coffee plants have greater area of individual mature leaves under shaded conditions than under full sun light (Pompelli et al., Reference Pompelli, Pompelli, Cabrini, Alves and Ventrella2012). On the other hand, even in dense coffee monoculture plantations (up to 10,000 plants ha−1), self-shading is present, inducing a light environment for the bottom canopy stratum like one found in deep forest understorey (Rakocevic et al., Reference Rakocevic, Batista, Pazianotto, Scholz, Souza, Campostrini and Ramalho2021a, Reference Rakocevic, dos Santos Scholz, Pazianotto, Matsunaga and Ramalho2023). At the initial stages of development, young coffee leaves are not equipped to withstand rapid light switches from low to high levels (Campa et al., Reference Campa, Urban, Mondolot, Fabre, Roques, Lizzi, Aarrouf, Doulbeau, Breitler, Letrez, Toniutti, Bertrand, La Fisca, Bidel and Etienne2017). From a morphological perspective, leaf elevation angles vary significantly over the vertical profile of Arabica coffee canopy (Unigarro-Muñoz et al., Reference Unigarro-Muñoz, Jaramillo and Flórez2017), which likely changes light incidence and availability inside canopy. Low light conditions modify some leaf gas exchange responses in the bottom stratum (Rakocevic et al., Reference Rakocevic, Batista, Pazianotto, Scholz, Souza, Campostrini and Ramalho2021a) and morphological characteristics, such as increased metamer length (Rakocevic et al., Reference Rakocevic, Matsunaga, Baroni, Campostrini and Costes2021b), green bean’s chemical characteristics (Rakocevic et al., Reference Rakocevic, Scholz and Kitzberger2018a, Reference Rakocevic, dos Santos Scholz, Pazianotto, Matsunaga and Ramalho2023), or beverage sensorial quality (Scholz et al., Reference Scholz, Kitzberger, Durand and Rakocevic2018) compared to the upper canopy strata, where light is not limiting.
The functional structural plant models aim at modeling the establishment of the plant structure and the functioning of the organs, considering properties and functions of each organ and the distribution of biomass produced by organs in the whole-plant scale (Sievänen et al., Reference Sievänen, Godin, DeJong and Nikinmaa2014). Such models generate an explicit 3D geometric plant representation (virtual plants), obtained for various species after parametrization, such as one used in GreenLab modeling formalism for coupled interactions with diseases in coffee (Triki et al., Reference Triki, Ribeyre, Pinard and Jaeger2023). Virtual coffee plants are built to assist in plant architecture analysis and quantification of various morphological traits (Rakocevic et al., Reference Rakocevic, Ribeiro, Marchiori, Filizola and Batista2018b, Reference Rakocevic, Matsunaga, Baroni, Campostrini and Costes2021b), to couple coffee structure to its functioning (Rakocevic et al., Reference Rakocevic, Ribeiro, Marchiori, Filizola and Batista2018b), or analyze the berry distribution over the plant strata (Rakocevic et al., Reference Rakocevic, dos Santos Scholz, Pazianotto, Matsunaga and Ramalho2023) using VPlants modeling platform (Pradal et al., Reference Pradal, Boudon, Nouguier, Chopard and Godin2009).
To our knowledge, the impact of the dry period on coffee leaf architectural traits over the vertical profiles of Arabica coffee cultivars and wild genotypes had never been analyzed over the phenological phases, nor the general tree form modification, based on strata that are building coffee plants. Herein, we hypothesized that coffee cultivars have a canopy structure less sensitive to drought than wild genotypes due to breeding process of the former ones. To address this prediction, plant and leaf architecture were analyzed over the Arabica coffee strata of two cultivars (Iapar 59 and Catuaí 99) and two wild Ethiopia accessions (‘E083’ and ‘E027’) grown under rainfed conditions and irrigation, during leaf and berry expansion and harvest phenophases of two consecutive productive years.
Material and Methods
Plant material and experimental conditions
From a total of about hundred Ethiopian accessions, two accessions named ‘E083’ and ‘E027’ were chosen for our experiment, because of their outstanding architectural characters noticed previously. The accession ‘E083’ had elongated leaves, a lot of flowering, and quick vegetative space occupation by 3rd-order branching, while the ‘E027’ visually presented big and large leaf blades, branched structure, and very few flowers. Also, two largely used coffee cultivars were planted, Iapar 59 and Catuaí 99. Iapar 59 is originated from the cross between the cultivar Villa Sarchi CIFC 971/10 and hybrid of Timor CIFC 832/2, representing C. canephora introgression by spontaneous specific cross with C. arabica (Anthony et al., Reference Anthony, Quiros, Topart, Bertrand and Lashermes2002), while Catuaí 99 is a high-productive C. arabica cultivar with high cup quality (Pérez-Molina et al., Reference Pérez-Molina, Picoli, Oliveira, Silva, Souza, Rufino, Pereira, Ribeiro, Malvicini, Turello, D́Alessandro, Sakiyama and Ferreira2021). The first ‘Catuaí’ was selected in 1949, with the hardiness, vegetative vigor, and yield potential of ‘Mundo Novo’ and small plant size and compact architecture of ‘Caturra’ (Guerreiro-Filho et al., Reference Guerreiro-Filho, Ramalho and Andrade2018).
Coffee seedlings were planted in field conditions (23°18’37”S, 51°09'46''W, 620 m a.s.l.), where coffee rows were oriented East–West, in 2.5 m distance among them and 0.5 m between plants in the row (planting density of 8,000 plants ha−1). The soil was dusky-red dystrophic latosol, with 790 g clay, 160 g silt, and 50 g sand per kg of soil in 0 to 0.20 m depth layer (Nunes et al., Reference Nunes, Cortez, Zaro, Zorzenoni, Melo, Figueiredo, Aquino, Medina, Ralisch, Caramori and Guimarães2021). Climate of the region is subtropical, Köppen–Geiger climate type Cfa, with average annual precipitation of about 1,585 mm, ranging from 55 mm in the driest month (August) to 245 mm in wettest one (January). The limiting factors for C. arabica survival in Cfa climate are defined by low minimum winter temperatures causing strong frost in some years (Meireles et al., Reference Meireles, Camargo, Pezzopane, Thomaziello, Fahl, Bardin, Santos, Japiassú, Garcia, Miguel and Ferreira2009), and dry periods. Coffee plants were cultivated under rainfed (not irrigated – NI) and with additional irrigation (irrigated – IR). For this latter treatment, drip irrigation was implemented, being triggered based on soil water balance method and aiming to restore the soil water field capacity (Cesanelli and Guarracino, Reference Cesanelli and Guarracino2011). The irrigation flow rate of each dripper was 3.5 L h−1. The irrigation was applied in the morning and satisfied a maximum evapotranspiration of 2.7 to 4.7 mm day−1 (winter and summer seasons, respectively) calculated for drip irrigated coffee culture. Drippers were distributed at 0.5 m of linear distance and hoses were near to coffee trunks. NPK (20:5:15) fertilization was added at 1,000 kg ha−1 year−1, split into four equal applications at flowering, two at berry and leaf expansion and one at the beginning of berry maturation.
The period of experiment comprised the two first productive years, starting from the January of the 2nd year after planting (experimental ‘year 1’) to July of the 3rd year in the field (experimental ‘year 2’) – (Supplementary Material Fig. S1). Based on the climatic water balance (Rolim et al., Reference Rolim, Sentelhas and Barbieri1998), there were four dry periods during the experiment, three in autumn/winter seasons as expected, and an unusual period in the hot summer, during February-March of the ‘year 1‘ (Fig. S1). Four plants of each accession and cultivar were grown under each water regime. Among ‘E083’ plants under rainfed conditions, one died after the dry winter of ‘year 1‘, while all other plants including two cultivars and ‘E027’ were killed by the frost of ‘year 2’ (Fig. S1), with only four ‘E083’ plants surviving under irrigated conditions (Rakocevic and Matsunaga, Reference Rakocevic and Matsunaga2018).
Light irradiance, measured as photosynthetic photon flux density (PPFD, μmol photons m−2 s−1), was obtained from stational photodiode sensors Hamamatsu G1118 (Hamamatsu, Japan) distributed along the plant canopy strata, starting from 20 cm from the soil level (average height where the 2nd-order axes appeared), and positioned at every subsequent 40 cm of the plant height (at 20, 60, 100, and 140 cm), and about 10 cm from the orthotropic trunk at South side. Those sensors were installed in one representative plant of each genotype and each water regime. Hamamatsu sensors were calibrated with LI–190R light sensors (LICOR, Lincoln NE, USA). One LI–190R light sensor was installed in the middle of the field, as a reference, at 2 m height (above the plant canopy). PPFD was measured in two-month periods (January–February and June–July) for both years of experiment, representing the observed phenophases described below. Initially, only two Hamamatsu sensors were positioned in the plant canopy, because of low plant height, and their number increased over the formation of strata. For data acquisition and storage, dataloggers CR21X and multiplexeres AM416 (Campbell Scientific, Logan UT, USA) were used. Data were collected every 60 s and integrated as mean values for each 15 min interval. The mean values were obtained from sensors installed in four plants of each water regime, irrespective of genotype. The accumulated intercepted light from the bottom to the top of the canopy was calculated for each observed stratum.
Plant coding and computational processing
Plant coding was performed in four coffee phenophases (Camargo and Camargo, Reference Camargo and Camargo2001), taking into account the leaf area expansion and berry production: 1) expansion of leaf area and berries during long days, in January–February of ‘year 1’ (BE1); 2) berry harvesting of the first production, in June–July of experimental ‘year 1’ (BH1), when there is reduced growth and emission of small leaves of short lifespan during short days (Rakocevic and Matsunaga, Reference Rakocevic and Matsunaga2018); 3) expansion of leaf area and berries of the second production during long days, January–February of the experimental ‘year 2’ (BE2); and 4) end of the second production season, June–July of the ‘year 2’ (BH2).
For architectural reconstructions and variable extractions, the plant canopy was divided into up to four strata (S) of 40 cm of thickness, starting from the soil level: S1:20–60 cm or <60 cm; S2:61–100 cm; S3:101–140 cm; and S4: >140 cm. We assumed an average height of 20 cm for the coffee trunk without plagiotropic branches. Topological and geometric coding of coffee trees in such actions was performed in three botanical scales – metamers, branches, and plants (Rakocevic and Androcioli-Filho, Reference Rakocevic and Androcioli-Filho2010), in multiscale tree graphs, MTGs (Godin and Caraglio, Reference Godin and Caraglio1998) following the VPlants modeling platform (Pradal et al., Reference Pradal, Boudon, Nouguier, Chopard and Godin2009).
The orthotropic axes were always described at metamer scale, collecting a maximum number of variables, as length of each metamer, leaf cardinal orientation, length, width (cm) and elevation (λ, angle in relation to horizontal plane, ranging from 0° to 90°), orientation, and total length of all plagiotropic branches inserted in the orthotropic axis, mortality/vivacity of 2nd-order plagiotropic axes terminal apex, and total number of berries. Four representative 2nd-order plagiotropic axes were sampled (one for each cardinal point) representing each 40 cm-thick stratum of the vertical plant profile (Matsunaga et al., Reference Matsunaga, Tosti, Androcioli-Filho, Brancher, Costes and Rakocevic2016). The sampled 2nd-order plagiotropic axes were described in detail following the same logic as orthotropic axes. In addition, the 3rd to 4th-order plagiotropic lateral branches belonging to the sampled and decomposed 2nd-order branches were also described in detail. Coding respective to each phenophase took about one month and four people to be completed.
The 3D reconstructions and visualizations followed the strategy to reconstruct Arabica coffee plant structure proposed by Matsunaga et al. (Reference Matsunaga, Tosti, Androcioli-Filho, Brancher, Costes and Rakocevic2016) using CoffeePlant3D software, which integrates the VPlants platform and modules: (1) AmostraCafe3D for branches and leaves interpolations using probabilistic and statistical models; (2) VirtualCafe3D for adjusting the MTG files with corrected branches angulation and phyllotaxy; and (3) Cafe3D for reconstructing the MTG coffee plants in 3D and attributing different colors for each leaf blade according to the axes orders. Each MTG element was rendered according to the geometry of defined objects, in which each coffee leaf blade was constructed from 16 triangles, while the metamers/branches from cylinders were formed by five rectangles (Rakocevic and Androcioli-Filho, Reference Rakocevic and Androcioli-Filho2010). These reconstructed plants were visualized in 3D in PlantGLViewer (Pradal et al., Reference Pradal, Boudon, Nouguier, Chopard and Godin2009).
Extraction of leaf and branching traits
The Red Green Blue options were attributed to leaves of four axes orders (black: 1st order; dark blue: 2nd order; pink: 3rd order; and light blue: 4th order) to differentiate leaves of several branching orders. The 3D reconstructions were converted in ‘vgx’ (Pradal et al., Reference Pradal, Boudon, Nouguier, Chopard and Godin2009), a tab-separated text file containing all (x, y, z) vertices coordinates and faces. The extraction of information was performed using R (R.4.2.1 software, R Development Core Team, 2022), considering 16 lines containing information (triangles) about each leaf blade object. It is worthy mentioning that we analyzed 176,399 leaves, each described in 16 lines. The z–values of Cartesian coordinates in ‘vgx’ files helped to separate leaf distributions over the 40 cm-thick strata during data extraction. The leaf ‘point 1’ (x1, y1, z1) was considered as leaf height. For leaf elevations, the main nerve vector for each leaf blade was formed by two points, initial ‘point 1’ (x1, y1, z1) and its tip localized at ‘point 9’ (x9, y9, z9). The λ of each leaf was calculated using the equation: arcsin(z9–z1)/Norm(p9–p1) in R software (R Core Team, 2022). When leaf elevation values were negative in MTG for 3D reconstructions, they were adjusted to positive signs for adequate calculations of elevation averages, following Ma et al. (Reference Ma, Zheng, Eitel, Magney and Moskal2017). For each leaf, the individual leaf area (cm2) was also extracted summing the surface of triangles from 16 corresponding subsequent lines. During coding and to verify the liveliness and mortality of branches, the terminal buds were coded, using virtual values of length and width of 0.1 cm and λ of 0°. In elevation and leaf area calculations, those virtual buds were excluded, as their surface was infinitely small, and their presence could distort the reconstructed average λ and individual leaf size. The total leaf area of each branching order was presented as relative participation in total plant leaf area. The branch area was obtained from diameter and average internode length of each branch (cm2). The orthotropic axis length, number of orthotropic metamers, 2nd-order axes number, and their elevation were extracted from the original MTGs of coding in the field. Based on the number of 2nd-order branches and 1st-order metamers, we calculated the probability of branching. In order to schematize branching orders and leaf elevations, free software GLOUPS (https://greenlab.cirad.fr/GLUVED/html/P3_Tools/Tool_simul_003.html) was used.
Validation of architecture modeling
Firstly, the validation was made at leaf scale by comparing the reconstructed leaves with 16 triangles with leaf measurements. The measurements of individual leaf area (ILA) were made using a leaf area (LA) meter LI–3100 (LICOR, Lincoln NE, USA), and parallelly measuring the leaf length and width using a ruler. In each of four observed phenophases, ten leaves of each genotype (‘E083’ and ‘E027’, Iapar 59, and Catuaí 99) and of each water regime (IR and NI) were measured. Based on leaf and width, the individual leaf area was reconstructed under VPlants. Secondly, the validation was made at plant scale by comparing measured and reconstructed plant leaf area. The measurements based on whole leaf area of each plant of each water regime and during each of four phenophases were performed using a LAI–2000 plant canopy analyzer (LICOR, Lincoln NE, USA). The procedure for coffee plant reconstruction was executed exactly as established by Matsunaga et al. (Reference Matsunaga, Tosti, Androcioli-Filho, Brancher, Costes and Rakocevic2016). The reconstructed plant leaf area obtained from VPlants was multiplied by 0.8, to obtain the reconstructed LAI.
Statistical analyses
Architecture analyses were performed at leaf, plant, stratum, and branching order scales, always having plant as an experimental unit. The experimental design was completely randomized, and the number of replications was four (only in the last two observed phenophases, ‘E083’ was represented by three plants in rainfed conditions). The R software (R Core Team, 2022) libraries and functions were used for statistical analyses.
The accuracy of modeling was estimated by linear regressions, at leaf and plant scales. The comparison of measured and reconstructed ILA was performed by linear regressions for each genotype in each water regime, while the comparison of measured and reconstructed LAI was performed by linear regression for all genotypes, water regimes, and phenophases. The RMSE (package ‘qpcR‘), equations, trend lines, R2, bias, and 1:1 line are shown in charts.
To analyze the radial and vertical distribution of berries in coffee trees, the resulting ‘vgx’ files of 3D reconstructions were explored, using leaf positions in the Cartesian coordinate system (x, y, z). The cumulative empirical distributions weighted by the inverse of the total number of leaves per plant were estimated by ‘Ecdf’ function from the ‘Hmisc’ package (Harrel, 2023). The ‘ks_test’ function from the package ‘Ecume’ (Bezieux, Reference Bezieux2021) that estimates the weighted Kolmogorov–Smirnov two-sample test (Monahan, Reference Monahan2011) was used to compare two water regimes for each genotype, separately for each of four observed phenophases.
ANOVA was applied to analyze factor effects on plant scale variables (leaf area, branch area, LAI, orthotropic trunk height, number of orthotropic metamers, number of 2nd order plagiotropic branches, and probability of branching), stratum scale variables (LA of stratum, ILA per stratum, leaf elevation per stratum, and 2nd-order elevation per stratum), or branching order variables (percent of LA per each branching order and leaf elevation per branching order). Two-way ANOVA considered a mixed linear model ('nlme’ package) and maximum likelihood of differences concerning the variable values depending on the levels of the following factors: (a) two water regimes (IR and NI), and four genotypes (‘E083, ‘E027’, Iapar 59, and Catuaí 99) at plant scale; (b) two water regimes and strata (2–4, number dependent on phenophase and plant) over the vertical plant profile for each of four genotypes; (c) two water regimes and leaf orders (2–4). If no significant interaction between factors was found, the reduced model was applied and fitted again. One-way ANOVA was additionally applied to analyze the responses (i) over four phenophases of each genotype, and (ii) compare genotypes at scales of plant, stratum, or branching order over all phenophases. For comparing average values estimated by the ANOVA models, the Tukey HSD test (p = 0.05) was used, supported by ‘lsmeans’, and ‘multcompView’ packages. The estimated means and standard errors are shown in charts.
Results
Modeling and validation at leaf and plant scales
A large fraction of collected coffee leaves ranged from 20 to 60 cm2, with some outliers that reached about 105 cm2 (Fig. 1a–d). The reconstructed individual leaf surface of coffee genotypes was slightly inferior when compared to measured ones, dimensioning this difference in a range of 0.9 to 3.4% (based on linear regression equation), with a bias ranging from 0.452 to −2.320 cm2, and RMSE from 0.332 to 2.811, while R2 was very near to 1, about 0.99. The leaves of NI plants were reconstructed with slightly lower bias and RMSE than of IR ones in all four genotypes. Well-adjusted area of individual leaves was the first validation step of 3D reconstructions. The adjustments of plant leaf area over soil area (leaf area index, LAI) measured and reconstructed were the second step in coffee plant reconstruction validations (Fig. 1e). The LAI estimated from 128 plant mock-ups was slightly underestimated compared to the measured, on average for 1.8% (based on linear regression equation), or for −0.122 m2 based on bias. This slight underestimation of total LAI was in accordance with individual leaf area reconstruction (Fig. 1a–d).
Measured and reconstructed leaf area of four Coffea arabica genotypes. Individual leaf area (cm2) of ‘E083’(a), ‘E027’(b), Iapar 59 (c), Catuaí 99 (d), and measured and reconstructed leaf area index (LAI, m2 m−2) of four genotypes (e). Leaves collected from irrigated (IR) and rainfed (NI) plants. Data from four phenophases were pooled. Linear regression equations, R2, RMSE, bias (n = 39–40 for individual leaf area, and 126 for LAI), and 1:1 line are shown.
Architectural traits at plant scale
Leaf (LA) and branch (BA) area per plant increased with tree structure development, being always significantly higher at a given date/phase than in the previous measurement or phenophase (Fig. 2a). The positive impact of irrigation on LA and BA was a general response of all genotypes and throughout all phenophases (BE1 to BH2). Under both water regimes, Iapar 59 showed the highest initial LA at BE1 (Fig. 2a). BA did not differ among genotypes under rainfed conditions, but ‘E083’ invested more in branching structure than ‘E027’ at BE1. At the time of first harvest (BH1), ‘E083’ had lower LA than the other three genotypes, irrespective of water regime, whereas no difference was found among ‘E027’, Catuaí 99, and Iapar 59. On the other hand, BA was significantly greater in Catuaí 99 than in Iapar 59 at BH1 and BE2, while the two wild accessions showed intermediate values in both water regimes. In the second year (BE2 and BH2), the average LAs of all four genotypes were similar, when considering each water regime separately. BA was also similar among genotypes at BH2 in each water regime.
Plant architectural traits of four Coffea arabica genotypes grown under irrigation (IR) and rainfed (NI) conditions and measured during leaf/berry expansion of (BE1 and BE2) and harvest (BH1 and BH2) during two consecutive years. Estimated mean ± SE and P-values (n = 3–4) for leaf and branch areas reconstructed from mock-ups (a), measured orthotropic trunk height (b), and counted number of emitted plagiotropic branches of the 2nd order (c). For each graph, different lowercase letters indicate significant differences among four genotypes in a given water regime and phenophase, while different uppercase letters indicate significant differences between the water regimes for a given genotype and phenophase. The uppercase green letters near the phenophase abbreviations indicate differences among the phenophases.
The height of the orthotropic trunk increased progressively from BE1 to BH2 (Fig. 2b). Under irrigated conditions, ‘E083’ orthotropic axes were the tallest among the four genotypes. In fact, the three genotypes other than ‘E083’ had similar orthotropic trunk height at BE1. Irrigated Iapar 59 had the lowest orthotropic trunk at harvest (BH1 and BH2), while both cultivars had shorter orthotropic trunks than the wild accessions at BE2 under irrigated conditions. The orthotropic axis height differed among genotypes under IR more than under NI conditions, and only one situation of significant variation was found at BH1 when rainfed Iapar 59 had shorter orthotropic trunk than other genotypes under the same water regime. Water availability affected the orthotropic trunk height only in ‘E083’, while the other three genotypes had similar trunk heights under both water regimes. It is worthy highlighting that the trunk height up to the first plagiotropic branching varied from 18 to 40 cm (average 29 cm), but nonsignificant differences were found among genotypes (p = 0.7325) or water regimes (p = 0.6676), and this trait had no significant impact on total orthotropic height.
The number of emitted 2nd-order plagiotropic axes gradually increased over the time (Fig. 2c), with effect of water regimes being significant only at the end of experimental period. At BH2, rainfed plants presented lower 2nd-order branch number than irrigated ones. The highest 2nd-order branch number was emitted by ‘E083’ and Catuaí 99, while the lowest by ‘E027’ during BE1 and BH2. Iapar 59 had similar 2nd-order branching as ‘E027’ during BH1 and BE2.
The lowest number of 1st-order metamers (Supplementary Material Fig. S2A) and the lowest number of emitted 2nd-order plagiotropic axes were observed in ‘E027’ at BE1, which lead to the lowest probability of the 1st-order axis branching (Fig. S2B). These two traits were the lowest in ‘E027’ during the following observed phenophases, but with similar values to those of Iapar 59 for the 1st order metamer number (Fig. S2A). The highest number of metamers on orthotropic trunk was observed in ‘E083’, which also had the tallest trunk (Fig. 2b). The irrigated plants had higher number of 1st-order metamers than NI ones, starting from BH1 to BH2 (Fig. S2A). The trunk branching was initiated lately in newly emitted trunk metamers of IR plants, considering that the probability of 1st-order branching was higher in NI than in IR plants from BE1 to BE2, and insensitive to water regime at BH2 (Fig. S2B).
Leaf distribution over vertical and horizontal plant profiles
In all phenophases, the leaf distribution along the canopy profile differed between the two water regimes (P < 0.05), in all four genotypes (Supplementary Material Fig. S3A). Rainfed plants accumulated more leaves in lower canopy strata when compared to irrigated ones, mainly of ‘E083’. While water availability changed the leaf distribution between the bottom and upper canopy strata of ‘E027’, it affected the height reached by the upper stratum in Iapar 59. Under irrigated conditions, ‘E083’ and Catuaí 99 showed the highest leaf positions over the vertical canopy profile at BH2, about 200 and 180 cm, respectively. In the radial distribution (horizontal distribution over x-axis and y-axis), leaves attained up to 80 cm distance from the trunk at BE1, and up to 100 cm in the three following phenophases (Fig. S3B). The leaf distribution over the plagiotropic branches of ‘E083’ was slightly delayed in rainfed as compared to irrigated conditions. Significant change in radial leaf distribution of ‘E027’ due to water regimes was noticed at harvest (BH1 and BH2). The radial leaf distributions of both coffee cultivars were not impacted by water regimes, except for Catuaí 99 at BH1 (leaves of rainfed plants attained longer distance from the trunk than ones of irrigated plants) and Iapar 59 at BE2 (leaves of irrigated plants attained longer distance from the trunk than ones of rainfed plants).
Architectural traits over plant strata
The 3D reconstruction snapshots of one irrigated plant with strata indications were presented for BH1 and BH2, together with branching order visualization (Fig. 3). The two coffee cultivars formed generally higher LA per stratum than the wild accession ‘E083’, while ‘E027’ presented intermediate LA values per stratum (P < 0.001). The mean LA per stratum increased from BE1 to BH2, not differing significantly over the two last phenophases for Catuaí 99 and ‘E027’. At BH1, the 2nd stratum (61–100 cm) had the most abundant leaf area in all genotypes under irrigation (Figs. 4 and 5), with such pattern occurring in Catuaí 99 up to BH2 (Fig. 5b). In wild accessions (Fig. 4) and rainfed Iapar 59 (Fig. 5a), the 2nd and the 3rd strata had similar LA, starting from BE2. The 4th stratum (>140 cm) appeared at BH1 or BE2 and was the less abundant in LA among the strata, sometimes being similar to the 1st stratum. Those LA distributions indicated the general conical or near-spherical forms of irrigated coffee crowns and were illustrated by snapshoots following the LA distribution over the vertical plant profile. The rainfed plants had lower LA of near all strata over the vertical profile compared to irrigated plants at some phenophases, as ‘E083’ at BH1, ‘E027’ at BH2 (Fig. 4a), or Catuaí 99 from BE1 to BH2 (Fig. 5b). In Iapar 59, the 2nd (at BE2) and 3rd (at BH2) stratum of plants under rainfed conditions had reduced LA compared to irrigated plants (Fig. 5a). Interestingly, no differences in plant stratum size between water regimes were observed in ‘E027’ from BE1 to BE2, which presented a near ellipsoidal tree crown (Fig. 4b). The bottom plant stratum had been better preserved in rainfed than in irrigated plants in some phenophases, as in Iapar 59 at BE2 and BH2 (Fig. 5a). The largest bottom stratum caused a near conical tree form (snapshoots).
The snapshoots of 3D reconstructions: scheme of branching orders (1st to 4th) and leaf angles (a), and strata distribution at BH1 (three strata) and BH2 (four strata) with black leaves corresponding to 1st, blue to 2nd, pink to 3rd, and light blue to 4th order axes, while berries are yellow (b).
Leaf area (LA, cm2) per strata of 40 cm (S1: <60 cm., S2:61–100 cm; S3:101–140 cm; S4 >140 cm) of the two wild Coffea arabica genotypes: ‘E083’(a), and ‘E027’(b) grown under irrigation (IR) and rainfed (NI) conditions and measured during leaf/berry expansion (BE1 and BE2) and berry harvest (BH1 and BH2) during two consecutive years. Estimated mean ± SE and P-values (n = 3–4) are shown. For each graph, different lowercase letters indicate significant differences among the strata in a given water regime and phenophase, while uppercase letters indicate significant differences between the water regimes for a given stratum and phenophase. The uppercase green letters near the phenophase abbreviations indicate differences among the phenophases. The snapshoots of 3D reconstructions of representative plants are shown at right of average LA per stratum with black leaves corresponding to 1st, blue to 2nd, pink to 3rd, and light blue to 4th order axes.
Leaf area (LA, cm2) per strata of 40 cm (S1: <60 cm., S2:61–100 cm; S3:101–140 cm; S4 >140 cm) of the two test cultivars of Coffea arabica: Iapar 59 (a), and Catuaí 99 (b) grown under irrigation (IR) and rainfed (NI) conditions and measured during leaf/berry expansion (BE1 and BE2) and berry harvest (BH1 and BH2) during two consecutive years. Estimated mean ± SE and P-values (n = 3–4) are shown. For each graph, different lowercase letters indicate significant differences among thestrata in a given water regime and phenophase, while uppercase letters indicate significant differences between the water regimes for a given stratum and phenophase. The uppercase green letters near the phenophase abbreviations indicate differences among the phenophases. The snapshoots of 3D reconstructions of representative plants are shown at right of average LA per stratum with black leaves corresponding to 1st, blue to 2nd, pink to 3rd- and light blue to 4th order axes.
The average individual leaf area per stratum was from 12.2 to 35.4 cm2 (Supplementaty Material Fig. S4). The smallest ILA in all genotypes was found at BE1 (plants with still well-lighted plant strata), compared to the next three phenophases. Among genotypes, the smallest ILA over the vertical plant profiles was found in ‘E083’ (Fig. S4A). Under rainfed conditions, reduced ILA was observed over nearly all phenophases and genotypes (Fig. S4A–C), with exception of Catuaí 99 (Fig. S4D). The ILA per stratum did not change due to water availability in Catuaí 99, even with the total leaf area per plant and general LA per stratum being reduced under rainfed conditions (Figs. 2a and 5b). These findings and also the insensitivity of ILA to water regime in Catuaí 99 (Fig. S4D) indicated decreases in leaf number per stratum under rainfed conditions (data not shown).
The average elevation values of 2nd-order branches per stratum varied from 16.8 to 78.0° (Fig. 6a–d). The highest average 2nd-order branch elevations were observed in decreased genotype order: Iapar 59, ‘E027’, ‘E083’, and Catuaí 99. Changes in the 2nd-order branch elevation per stratum along phenophases increased over time, being the highest at BH2 and the lowest at BE1/BH1, in all genotypes. The irrigation reduced the branch elevations only in ‘E083’ compared to more erected branches of rainfed plants at BE2 and BH2 (Fig. 6a). The tree structure was being more complex when the 4th stratum appeared, starting from BH1 or BE2 (Figs. 3–5). As one general response of coffee plants irrespective of phenophases or water availability, the lowest 2nd-order branch elevation angles were noticed in the lowest 1st stratum, while the highest branch elevations were attributed to the upper canopy strata (3rd and 4th), positioning the bear leaves higher than the orthotropic trunk (Fig. 6a–d). From BH1, even the 2nd stratum (61–100 cm) had the branches of significantly lower elevation angle than the upper two strata (3rd and 4th) in all genotypes.
2nd order axes elevation (°) per strata of 40 cm (S1: <60 cm., S2:61–100 cm; S3:101–140 cm; S4 >140 cm) analyzed for four Coffea arabica: ‘E083’ (a), ‘E027’ (b), Iapar 59 (c), and Catuaí 99 (d) grown under irrigation (IR) and rainfed (NI) conditions and measured during leaf/berry expansion (BE1 and BE2) and berry harvest (BH1 and BH2) during two consecutive years. Estimated mean ± SE and P-values (n = 3–4) are shown. For each graph, different lowercase letters indicate significant differences among the strata in a given water regime and phenophase, while uppercase letters indicate significant differences between the water regimes for a given stratum and phenophase. The uppercase green letters near the phenophase abbreviations indicate differences among the phenophases, while the red ones on right indicate general differences among the genotypes (P Gen = 0.0001).
The average leaf elevation per plant stratum was from 14 to 49° (Supplementary Material Fig. S5A–D). Generally, the lowest leaf elevation angles (tending to planophile) were observed in ‘E083’ compared to other genotypes. The highest leaf angle elevation per stratum (49°) was observed at BE1 and the lowest one was at BH1 irrespective of genotype. In ‘E083’, the lowest elevation values of the 1st stratum were observed at BH2 of rainfed plants and of 4th stratum at BH1 irrespective of water regime. The 4th stratum of ‘E083’ rainfed plants had a very high leaf angle, about 45° at BE2 and BH2 (Fig. S5A). In ‘E027’, the lowest leaf elevation of the 1st stratum among the strata happened only in well-developed tree structure at BH2, while a leaf elevation of the 4th stratum was very low only when this stratum appeared, at BH1 (Fig. S5B). In the two bred cultivars under irrigation, leaf elevation of the 1st stratum was lower than upper 3rd stratum at BE1 or the lowest among the strata at BE2 (Fig. S5C,D). Low leaf angles of new appeared 4th stratum were only observed in NI plants of Iapar 59 at BE2 (Fig. S5C).
Light intercepted over vertical plant profile
The superior strata intercepted the great part of incoming light energy, and irrigated plants intercepted more light than rainfed plants (Supplementary Material Fig. S6). In terms of diurnal average, the 2nd and 3rd strata localized above the sensor positioned at 60 cm, intercepted 70–77% of solar energy in irrigated and about 52–55% in rainfed plants at BE1 and BH1 (Fig. S6A,B). One smaller portion of incoming light was intercepted by the 1st stratum (20–60 cm) at those two phenophases, about 13–24% in irrigated and 22–33% in rainfed plants. More complex light distribution along canopy strata occurred in the second experimental year (Fig. S6C, D), due to an additional 4th leaf stratum as compared with the first year. Also considering the diurnal average at BE2, the increased LA provoked the interception of about 96% and 93% of incoming light in irrigated and rainfed plants, respectively (Fig. S6C). More than half of PPFD was intercepted by the 3rd and 4th strata (above the sensor at 100 cm) in IR plants, and even more in NI plants, depending on day time. The diurnal curve was slightly asymmetric because of the sensor position in South. At BH2, the 3rd and 4th strata had lower leaf area to intercept the incident light in NI than in IR plants, as revealed by PPFD incoming to the sensor at 100 cm (Fig. S6D), in line with the 3D reconstructions and LA of strata (Figs. 4 and 5).
Leaf distribution and elevation over branching orders
The 1st order leaves (emitted and maintained on orthotropic trunk) represented a very small fraction of the total plant LA (Fig. 7a–d). At BE1, all genotypes built the tree architecture by 1st, 2nd, and 3rd order leaves, irrespective of water regime. ‘E027’ and Iapar 59 did not modify their pattern of leaf order participation over the time or due to water availability, having the most LA built by 3rd order followed by 2nd > 4th > 1st order leaves (Fig. 7C). At BE1, the 3rd order LA was significantly higher than the 2nd order (∼35–40% of 2nd order plus 60–65% of 3rd order LA, Fig. 7b–d) in all genotypes but the rainfed Catuaí 99 (Fig. 7d) and ‘E083’ irrespective to water regime (Fig. 7a) where the opposite LA distribution over branching orders (∼75% of 2nd order and ∼25% of 3rd order LA) was observed. Irrigated ‘E083’ preserved this LA distribution at BH1, meaning the preservation of investments in radial leaf space occupation by existing and newly emitted 2nd-order leaves. Under rainfed conditions, ‘E083’ showed higher participation of the 3rd-order leaves than the 2nd-order ones at BH1, as similarly found in all other genotypes (Fig. 7b–d). At BH1, the ‘symphony’ of synchronized events started to be complex including very slight 4th order axes participation in ‘E027’, Iapar 59, and Catuaí 99. At BE2, the leaf order participation in total LA was 3rd > 2nd > 4th > 1st stratum irrespective of genotype or water regime. This was preserved at BH2 in all genotypes but not in rainfed ‘E083’, which formed the tree crown equally by 2nd- and 3rd-order leaves.
Percent of plant leaf area formed by each branching order per strata of 40 cm (S1: <60 cm., S2:61–100 cm; S3:101–140 cm; S4 >140 cm) analyzed for four Coffea arabica genotypes: ‘E083’ (a), ‘E027’ (b), Iapar 59 (c), and Catuaí 99 (d) grown under irrigated (IR) and rainfed (NI) conditions and measured during leaf/berry expansion (BE1 and BE2) and berry harvest (BH1 and BH2) during two consecutive years. Estimated mean ± SE and P-values (n = 3–4) are shown. For each graph, different lowercase letters indicate significant differences among branching orders in a given water regime and phenophase, while different uppercase letters indicate significant differences between the water regimes for a given branching order and phenophase.
The lowest leaf elevation by branching order was observed in Catuaí 99 and the highest in ‘E027’ and Iapar 59 (Supplementary Material Fig. S7A–D). It increased from BE1 to BH2 phenophases, primarily due to increased 1st-order leaf elevation values. The 2nd- to 4th-order leaf elevations were on average 40–45°, differing from leaves of the 1st-order axis that were positioned from ∼0° at BE1 and near 90° at BE2 and BH2.
Discussion
In this study, we revealed the effects of water deficit and ageing on canopy architecture of Arabica coffee plants. Then, acclimation to repeated episodes of drought depends not only on an orchestrated reprograming of plant metabolism (Menezes-Silva et al., Reference Menezes-Silva, Sanglard, Ávila, Morais, Martins, Nobres, Patreze, Ferreira, Araújo, Fernie and DaMatta2017) but also on leaf responses over the vertical plant profile, which were less sensitive to water availability in bred cultivars than in wild accessions.
Leaf area responses to drought at plant and stratum scale
From an ecological point of view, leaf investment strategies are closely associated with growth, survival, and light requirement of the species, and are linked to variations in whole-plant growth and survival (Poorter and Bongers, Reference Poorter and Bongers2006). As an adaptative strategy to drought, leaf area is generally reduced, parallelly increasing leaf thickness and tissue density (Werner et al., Reference Werner, Correia and Beyschlag1999), indicating morphological and anatomical plant responses to increase water use efficiency in dry environments (Yang et al., Reference Yang, Lu, Wang, Wang, Liu and Chen2021). Under drought, the expansion time and lifespan of coffee leaves are decreased, with higher leaf senescence, mainly in winter season (Rakocevic and Matsunaga, Reference Rakocevic and Matsunaga2018). In our experiment, the individual leaf area of wild ‘E027’ accession grown in rainfed conditions was reduced in all strata at all phenophases, with exception of BE1 (Fig. S4B), while similar ILA was observed in wild ‘E083’ between the two water regimes only in rainy summer of BE2 (Fig. S4A). On the opposite, the drought did not reduce ILA of Catuaí 99 in any phenophase (Fig. S4D) and this finding could clearly confirm our initial hypothesis of lower drought impacts on cultivars than wild accessions. However, Iapar 59 showed similar ILA sensitivity to drought (Fig. S4C) as the wild accessions. Regardless of genotype and water regime, ILA increased gradually over the plant structure development (Fig. S4).
At plant scale, coffee LA is diminished even under moderate water deficit (DaMatta and Ramalho, Reference DaMatta and Ramalho2006), which is due to reductions in leaf number and ILA (Souza et al., Reference Souza, Guimarães, Colombo, Sant’Ana and Castanheira2016), similar to our findings (Fig. 2a). Coffee plant LAI shows strong coupling to atmospheric conditions and high sensitivity to large vapor pressure deficits and air/leaf temperatures (Marin et al., Reference Marin, Angelocci, Righi and Sentelhas2005). The quantification of leaf trait responses at scale of one species and/or genotype helps to understand their capacity and adaptability to environmental challenges (Oktavia and Jin, Reference Oktavia and Jin2020; Duan et al., Reference Duan, Jia, Li and Wu2022). At stratum scale, LA of the 2nd stratum was not reduced under rainfed conditions, with exception of ‘E083’ at BE1, ‘E027’ at BH2 (Fig. 4), Iapar 59 at BE2, and Catuaí 99 at BE2 and BH2 (Fig. 5). This means that coffee trees reduced LA of the oldest and/or the newest canopy strata due to increased leaf senescence, reduced new leaf emissions, or reduced ILA, investing more carbon in the most developed stratum (by new metamer emission or/and branching) and changing tree crown form from near-spherical or conical (in irrigated plants) to near ellipsoidal with high vertical diameter (rainfed plants). Interestingly, the stratum size in wild ‘E027’ did not change due to water regimes up to BH2, with plants showing a near ellipsoidal crown irrespective of water regime (Fig. 4b). At this point, the leaf traits showed lower sensitivity to drought in one wild accession compared to cultivars, but in meantime, total plant LA always differed between the two water regimes (Fig. 2a), suggesting that architectural responses to episodic drought at the leaf, stratum, and plant scales changed in terms of magnitude and over phenology.
Trees growing under low light or low water availability tend to exhibit leaf traits with slow returns on resource investment (such as low specific leaf area, low leaf nitrogen and phosphorus concentrations, and high leaf dry matter content) and have conservative ecological strategies (Vogel et al., Reference Vogel, Manning, Cadotte, Cowles, Isbell, Jousset, Kimmel, Meyer, Reich, Roscher, Scherer-Lorenzen, Tilman, Weigelt, Wright, Eisenhauer and Wagg2019). Leaf traits vary on a continuum trading off from short-term carbon gain against long-term leaf persistence, as shown in Arabica coffee (Rakocevic and Matsunaga, Reference Rakocevic and Matsunaga2018). The continuum in leaf traits is linked to variation in whole-plant growth and survival (Poorter and Bongers, Reference Poorter and Bongers2006), which could be translated into varying light distribution within plant canopy. High growth in highlighted exposed strata is promoted by short-lived and physiologically active leaves of 2nd and 3rd branching orders of upper strata in C. arabica, while shaded leaves of long lifespans are found in 3rd and 4th order branching orders of bottom canopy stratum (Rakocevic and Matsunaga, Reference Rakocevic and Matsunaga2018), as reported here (Fig. 7a). Considering that branch growth is inversely related to fruit load in coffee trees (Bote and Vos, Reference Bote and Vos2016), the vegetative growth must be evaluated taking into account the reproductive growth, with consequent carbon investments in berry production (Rakocevic et al., Reference Rakocevic, Braga, Batista, Maia, Scholz and Filizola2020). Among the four genotypes, ‘E027’ has the lowest berry production irrespective of water regime, while ‘E083’ is well productive as Iapar 59 under irrigation (Rakocevic et al., Reference Rakocevic, dos Santos Scholz, Pazianotto, Matsunaga and Ramalho2023). Under rainfed conditions, two coffee-bred cultivars are more productive than the wild accessions. Here, we must point out that coffee breeding is primarily directed at increasing berry production (Guerreiro-Filho et al., Reference Guerreiro-Filho, Ramalho and Andrade2018). During berry elongation and maturation (phenophase studied herein), shoot elongation and branch biomass are generally reduced due to the high carbon demand for filling coffee berries (Vaast et al., Reference Vaast, Dauzat and Génard2002). So, carbon allocation to berries or leaves varies over phenology and is modified by drought.
At plant scale: orthotropic axis and branching orders
Decreased plant height is a common response to drought in various C. arabica experiments (Avila et al., 2020; Vu et al., Reference Avila, Almeida, Costa, Machado, Barbosa, Souza, Martino, Juárez, Marçal, Martins, Ramalho and DaMatta2018), and here the orthotropic axis had been reduced only in wild ‘E083’ under rainfed conditions (Fig. 2b). Historically, the breeding process was guided by searching low height, structurally compact, and productive cultivars (Guerreiro-Filho et al., Reference Guerreiro-Filho, Ramalho and Andrade2018). So, insensitivity of orthotropic height in Iapar 59, Catuaí 99, and wild ‘E027’ indicated one morphological characteristic that is stable in cultivars and not decisive to coffee breeding (Silva et al., Reference Silva, Junior, Ramos, Rocha, Veiga, Silva, Brasileiro, Santana, Soares, Malaquias and Vinson2022). On the other hand, even mild drought (Fig. S1) reduced the number of metamers at 1st order axis and in number of 2nd-order branches of all genotypes at BH2 (Figs. 2c and S3A). The probability of branching was not impacted by drought, being near 1 in cultivars and showing lower values (the lowest about 0.83) in wild genotypes (Fig. S3B). Under mild-water deficit, irrigation did not impact branch growth, but markedly reduced branch mortality in fruiting trees of Catimor, an Arabica coffee hybrid (Almeida et al., Reference Almeida, Ávila, Pérez-Molina, Barbosa, Marçal, de Souza, Martino, Cardoso, Martins and DaMatta2021). The number of 2nd-order plagiotropic branches is supposed to be an important trait used in coffee breeding (Silva et al., Reference Silva, Junior, Ramos, Rocha, Veiga, Silva, Brasileiro, Santana, Soares, Malaquias and Vinson2022), and our findings about the 2nd-order branching dynamics support this assumption. About the partitioning of LA among branching orders, the prevalent 3rd-order leaf space occupation in both coffee cultivars and in wild ‘E027’ began from the first phenophase BE1 and in ‘E083’ from BH2 under rainfed conditions (Fig. 7). Again, ‘E083’ differed from other genotypes in terms of space occupation, vertical growth, and investments into new 2nd order axes in the upper canopy strata, but parallelly invested more carbon into berry production than ‘E027’, similarly to two bred coffee cultivars (Rakocevic et al., Reference Rakocevic, dos Santos Scholz, Pazianotto, Matsunaga and Ramalho2023).
Leaf and branch elevations over plant strata
Plants have developed several strategies to reduce the interception of excessive radiation by foliage and then avoid photoinhibitory damage while maintaining positive carbon balance (Werner et al., Reference Werner, Correia and Beyschlag1999). In general, the geometric trends of foliage distribution describe leaf arrangement distribution: planophile (the highest frequency at 0–30°); plagiophile (the highest frequency at 31–60°); erectophile (the highest frequency at 61–90°); and extremophile (two peaks of leaf elevation angles that are found at 0° and 90°), as proposed by Goudriaan (Reference Goudriaan1988). In our experiment, leaves of 2nd to 4th-order branching were roughly plagiophile (Figs. 3a and S5A–D), similar to progeny CX2391 with horizontal leaf elevations (Unigarro-Muñoz et al., Reference Unigarro-Muñoz, Trejos Pinzón and Acuña-Zornosa2021), while the 1st order leaves were classified as extremophiles (Fig. S5A–D). The angles of the 2nd-order branch elevation define the ‘Caturra’ coffee cultivar as planophile (Castillo et al., Reference Castillo, Arcila, Jaramillo and Sanabria1996), with a predominance of elevations between 0 and 30°. In our experiment, the planophile branch elevation was noticed at bottom canopy stratum in all genotypes, plagiophile at the 2nd stratum due to plagiotropism and gravity (weight of leaves and berries), and erectophile at the upper 3rd and 4th strata, (Fig. 6a–d). Regarding the angle of branch elevations, the planophile distribution is predominant in both large- and small-size genotypes, with the only exception being the large-size Ethiopian accession (‘E070’), which presents an erectophile distribution (Unigarro-Muñoz et al., Reference Unigarro-Muñoz, Jaramillo, Ibarra and Flórez2016). Branch elevations of two wild Ethiopian accessions did not differ from the two cultivars in our experiment, even with ‘E083’ being classified as large-size one.
Leaf angle adjustment can be a strategic response to changing light, heat, or water conditions (Yang et al., Reference Yang, Li, Jablonski, Stovall, Kim, Yi, Ma, Beverly, Phillips, Novick, Xu and Lerdau2023). A ten-degree change in the leaf angle from 50 to 60° can result in an instantaneous 22% difference in intercepted direct radiation, which can, in turn, have impacts on photosynthesis, leaf temperature, transpiration, and energy balance (Falster and Westoby, Reference Falster and Westoby2003). In our experiment, the average leaf elevation was plagiophile. However, plant age modifies leaf and branch structures (Niinemets, Reference Niinemets2010) and, in this sense, the quantitative geometric description of the canopy becomes more complex with ageing (Campbell, Reference Campbell1990). Among various Mediterranean species, Cistus spp. structurally regulates light interception through leaf elevation, from a more horizontal elevation in spring (<35°) to a more vertical elevation in summer (>70°) (Werner et al., Reference Werner, Correia and Beyschlag1999). As a general trend, the Arabica coffee leaf elevations were the highest in very early berry and leaf expansion phenophase during the summer (BE1), and the lowest at harvest (BH2) (Fig. S5). Most large-size genotypes cultivated in Colombia are defined as plagiophile, while the small-sized genotypes have a planophile leaf distribution (Unigarro-Muñoz et al., Reference Unigarro-Muñoz, Jaramillo, Ibarra and Flórez2016). In general, leaves are more vertical in dry, hot, and light-exposed conditions, while more horizontal in mesic and light-limited conditions (Yang et al., Reference Yang, Li, Jablonski, Stovall, Kim, Yi, Ma, Beverly, Phillips, Novick, Xu and Lerdau2023). Regardless the genotype, coffee leaves were planophile at the bottom canopy stratum in our study (Fig. S5). Near planophile leaves were also found in some of the newest 4th stratum, irrespective to water regime. This revealed that leaf elevation in Arabica coffee was more influenced by light than water availability.
Conclusions
Modeling and structural analyses of Arabica coffee plants revealed changes in some architectural traits due to ageing and water availability, such as variations in leaf area, individual leaf area, and leaf elevation per canopy stratum. The orchestrated reprograming of leaf responses in the vertical plant profile was less sensitive to water availability in bred cultivars than in wild accessions. The individual leaf area was less sensitive to water deficit in Catuaí 99, while the 2nd order axis elevation was less sensitive to drought in both bred cultivars than in ‘E083’, partially proving our hypothesis that leaf/branch responses over the vertical plant profile were less sensitive to water availability in cultivars than in wild accessions, suggesting the effect of selection.
Plants of all genotypes under the two water regimes showed a general tendency from planophile branch elevation at the bottom canopy stratum, to plagiophile at the 2nd stratum, and finally to erectophile at the upper 3rd and 4th strata. Leaves of 2nd- to 4th-order branching were roughly plagiophile, while the 1st-order leaves were classified as extremophiles. When the coffee leaves were planophile, irrespective of genotype, this distribution was noticed at the lowest plant stratum and the newest low-developed 4th stratum. This pattern was not obligatorily related to water regime, suggesting that the leaf elevation in Arabica coffee was more influenced by light distribution within canopy than water availability.
As expected, rainfed trees had their crowns with much lower leaf area than irrigated trees, with leaf area showing an increasing trend over the four phenophases studied. The increase in individual leaf area over phenophases was due to the increase of tree structure and increased shading over the vertical plant profile in coffee cultivars, while such increased individual leaf area in wild accessions was reserved for the bottom strata. The dynamics of architectural changes, coupled with functioning of the meristems and/or functional responses as carbon assimilation, would give us a more complete image of how wild and bred genotypes respond to changes in water availability.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/S0014479724000036
Data availability statement
The authors can provide the experimental data for all interested researchers.
Author contributions
Conceptualization, M.R.; methodology, M.R. and E.C.; software, F.T.M.; validation, M.R., E.C., R.V.R., and J.C.R.; formal analysis, R.A.A.P. and M.R.; investigation, M.R. and F.T.M; resources, M.R.; data curation, R.A.A.P. and M.R.; writing – original draft preparation, M.R.; writing – review and editing, J.C.R., R.V.R, and E.C.; visualization, M.R., E.C., R.V.R., and J.C.R.; supervision, M.R.; project administration, M.R.; funding acquisition, M.R. All authors have read and agreed to the published version of the manuscript.
Funding statement
The research work was carried out with the support of Consórcio Pesquisa Café (Grants (02.09.20.008.00.00 and 02.13.02.042.00.00). Authors acknowledge the FAPES for awarded fellowship for M.R. (2022-M465D). R.V.R. is a fellow of the National Council for Scientific and Technological Development (CNPq, Brazil, Grant 304295/2022-1). Funding support from Fundação para a Ciência e a Tecnologia I.P., Portugal, to J.C.R. through the units CEF (UID/04129/2020), GeoBioTec (UIDP/04035/2020), and associated laboratory (LA/P/0092/2020) is also greatly acknowledged.
Competing interests
The authors declare no conflict of interest.
