Introduction
It is widely acknowledged that in fruit crops such as Coffea arabica L., active vegetative and reproductive growth occur simultaneously, resulting in the coexistence of multiple strong sinks within the same plant. In mature coffee trees, photoassimilate allocation is regulated by competing demands from developing fruits, ongoing vegetative growth (new leaves and branches), and perennial storage organs such as roots and older woody tissues (Bote and Vos, Reference Bote and Vos2016). Among these sinks, fruits generally act as the dominant physiological sink, exerting strong control over carbohydrate partitioning to other organs. Consequently, biomass allocation between vegetative and reproductive growth in coffee is commonly characterised by a negative relationship, as documented across multiple studies (Amaral et al., Reference Amaral, Da Matta and Rena2001; Bote and Vos, Reference Bote and Vos2016; Cannell, Reference Cannell, Clifford and Wilson1985; Franck et al., Reference Franck, Vaast, Génard and Dauzat2006; Vaast et al., Reference Vaast, Angrand, Franck, Dauzat and Génard2005, Reference Vaast, Bertrand, Perriot, Guyot and Génard2006) and similarly reported for other perennial fruit crops such as olive (Fernández et al., Reference Fernández, Ladux and Searles2015), apple (Meland, Reference Meland2009), cherry (Cittadini et al., Reference Cittadini, de Ridder, Peri and van Keulen2008), and citrus (Martínez-Fuentes et al., Reference Martínez-Fuentes, Mesejo, Reig and Agustí2010; Martínez-Alcántara et al., Reference Martínez-Alcántara, Iglesias, Reig, Mesejo, Agustí and Primo-Millo2015). Nevertheless, vegetative growth also provides essential support for reproductive performance: greater leaf assimilation capacity and improved water relations have been shown to sustain flowering intensity and fruit set in coffee, particularly under stress conditions (Rakocevic et al., Reference Rakocevic, Malagodi Braga, Batista, Nunes Maia, Santos Scholz and Filizola2020).
From a production perspective, high fruit load is a desirable trait in the current harvest. However, excessive allocation of resources to reproductive growth decreases vegetative growth, with negative consequences for yield in subsequent seasons. Reduced vegetative biomass, especially leaves, is associated with limited source capacity and lower bud development, both of which are critical for sustaining future yields in coffee (Cannell, Reference Cannell, Clifford and Wilson1985; DaMatta et al., Reference DaMatta, Ronchi, Maestri and Barros2010; Vaast et al., Reference Vaast, Angrand, Franck, Dauzat and Génard2005), a crop characterised by strong interannual feedback between vegetative growth and fruit load (Moscardini et al., Reference Moscardini, Souza, Mazzafera and Favarin2020; Vaast et al., Reference Vaast, Angrand, Franck, Dauzat and Génard2005). Although the influence of fruit load on biomass partitioning in coffee has been qualitatively described, a quantitative framework that relates active vegetative growth to fruit load across the growth cycle remains lacking.
The classical harvest index (HI) was originally proposed by Beaven (Reference Beaven1920), and it represents the proportion of total biomass allocated to the harvested product at maturity. In annual seed crops, this concept is closely associated with the so-called ‘migration coefficient’, defined as the ratio of grain yield to the total biomass of the entire aboveground plant, which has been successfully implemented in crop simulation models (Holzworth et al., Reference Holzworth, Huth, deVoil, Zurcher, Herrmann, McLean, Chenu, van Oosterom, Snow, Murphy, Moore, Brown, Whish, Verrall, Fainges, Bell, Peake, Poulton, Hochman, Thorburn, Gaydon, Dalgliesh, Rodriguez, Cox, Chapman, Doherty, Teixeira, Sharp, Cichota, Vogeler, Li, Wang, Hammer, Robertson, Dimes, Whitbread, Hunt, van Rees, McClelland, Carberry, Hargreaves, MacLeod, McDonald, Harsdorf, Wedgwood and Keating2014; Jones et al., Reference Jones, Hoogenboom, Porter, Boote, Batchelor, Hunt, Wilkens, Singh, Gijsman and Ritchie2003; Raes et al., Reference Raes, Steduto, Hsiao and Fereres2011; Stöckle et al., Reference Stöckle, Donatelli and Nelson2003; Williams et al., Reference Williams, Jones, Kiniry and Spanel1989). However, applying this concept to coffee presents important challenges. Unlike annual crops with determinate growth habit, which exhibit a clear transition from vegetative to reproductive development, coffee is a perennial species in which vegetative growth, fruit development, and storage processes occur concurrently rather than sequentially (Camargo and Camargo, Reference de Camargo and de Camargo2001; DaMatta et al., Reference DaMatta, Ronchi, Maestri and Barros2010; Millard and Grelet, Reference Millard and Grelet2010). Consequently, biomass accumulation cannot be readily partitioned into distinct vegetative and reproductive phases, and destructive measurement of total plant biomass at harvest is impractical. These characteristics limit the direct application of the classical HI concept to coffee.
The application of HI to perennial crops such as coffee has therefore received limited attention, and to our knowledge, a coffee-specific HI has not been previously described. Nevertheless, developing a practical method to characterise active vegetative growth and biomass partitioning as a function of fruit load would represent a significant advance for coffee research. Here, we propose a methodological framework to quantify active vegetative growth in relation to fruit load throughout the growth cycle, allowing characterisation of dry matter partitioning in coffee trees and providing a basis for estimating HI in this perennial system. We hypothesised that increasing fruit load would progressively reduce vegetative biomass accumulation and elevate HI up to a physiological ceiling, and that these relationships could be quantified using regression models across contrasting site-year conditions.
Material and methods
Plant material and growth conditions
Three field experiments were conducted in the Cerrado-Savanna transition zone of Minas Gerais state, southeastern Brazil, during two growing seasons – 2015/2016 and 2016/2017 (Supplementary Material Fig. S1). Two experiments were established at Jacuí-MG (21°6′32.32″S, 46°41′11.79″W, 1098 m a.s.l.) using Coffea arabica L. cv. Catuaí Amarelo in the third production cycle, with tree spacing of 3.5 × 0.7 m (4082 trees ha−1). One experiment was established at Ibiá-MG (19°33′00.16″S, 47°22′32.8″W, 895 m a.s.l.) using Coffea arabica L. cv. Topázio in the fifth production cycle, with tree spacing of 3.5 x 0.8 m (3571 trees ha−1). Both cultivars rank among the most productive genotypes cultivated in Brazil and are adapted to a wide range of environmental conditions. The Köppen climate classification of both locations is Cwb, humid subtropical with dry winters and temperate summers (Alvares et al., Reference Alvares, Stape, Sentelhas, Gonçalves and Sparovek2013). Annual rainfall was 1812 mm (Jacuí-MG, 2015/2016), 1234 mm (Jacuí-MG, 2016/2017), and 1350 mm (Ibiá-MG, 2016/2017), with mean temperatures of 21.4°C, 21.2°C, and 21.7°C, respectively (Supplementary Material Fig. S3). Both experiments were conducted under rainfed conditions. Soils at both sites were classified as Oxisol (Soil Survey Staff, 1999); chemical characterisation was performed prior to experiment installation (Supplementary Material Table S1), and fertilisation was applied according to Minas Gerais state recommendations (Guimarães et al., Reference Guimarães, Garcia, Alvarez, Prezotti, Viana, Miguel, Malavolta, Corrêa, Lopes, Nogueira and Monteiro1999).
Experimental design
The experimental design was completely randomised with controlled fruit load in six levels (100%, 80%, 60%, 40%, 20%, and 0%) and five replicates in Jacuí-MG and eight replicates in Ibiá-MG. Each tree consisted of an independent experimental unit. The trees were selected for the fruit load manipulation based on uniformity and vigour. After flowering in early November of each season, when the fruits were in the pinhead development stage (Camargo and Camargo, Reference de Camargo and de Camargo2001), six fruit loads were manually imposed on all reproductive branches of the respective tree. Fruit load was adjusted by counting the number of fruits per node and selectively removing excess fruits to achieve the target fruit-load percentages. The 100% treatment corresponded to the maximum natural fruit set per node observed at the pinhead stage, and trees were verified to carry comparable initial fruit numbers prior to manipulation to ensure experimental uniformity. The insertion point of the last fully expanded leaf pair was labelled on all plagiotropic and orthotropic branches of each tree for further evaluation of vegetative development. At harvest, all newly produced tissues above the labelling point – including stems, leaves, and fruits on both plagiotropic and orthotropic branches – were collected and separated into their respective components. This approach captures the active annual growth of each tree while excluding pre-existing woody tissues and roots.
Crop yield and measurements
In early June of each season, the fruits were in harvest maturity (Supplementary Material Fig. S2A). Then, branches were cut from the label and characterised. The number of leaves was counted, and leaf area was measured using a LI-3100 area metre (Li-Cor Biosciences, Lincoln, NE, USA) for the Jacuí-MG experiment 2015/2016. Total leaf area per tree was obtained by summing the area of all newly produced leaves above the labelling point; mean individual leaf area was subsequently calculated by dividing total leaf area by the number of leaves. Stem, branches, and leaves, which compose the vegetative biomass; and fruit samples of all experiments were dried (Fig. S2B) until constant mass at 60°C. After drying, fruits were de-husked (Fig. S2C) and mass corrected to 12% water concentration (commercialised bean moisture) to obtain ready-to-be-roast coffee beans, hereafter called green beans.
The leaf-to-fruit ratio (cm2 g−1) was calculated as the ratio of the leaf area (cm2) per plant of active vegetative growth and the fruit dry yield (g) (Equation (1)).
$${\rm{Leaf}} - {\rm{to}} - {\rm{fruit}}\;{\rm{ratio}} = {{{\rm{Leaf}}\;{\rm{area}}\;({\rm{c}}{{\rm{m}}^2}\ {\mkern 1mu} {\rm{per}}\;{\rm{plant}})} \over {{\rm{Fruit}}\;{\rm{dry}}\;{\rm{yield}}\;({\rm{g}}\ {\mkern 1mu} {\rm{per}}\;{\rm{plant}})}}$$
HI (dimensionless) was calculated as the ratio of the fruit dry yield (g) to the total aboveground plant biomass (g) produced over the growth cycle (Equation (2)).
$${\rm{HI}} = {{{\rm{Fruit}}\;{\rm{dry}}\;{\rm{yield}}\;({\rm{g}}{\mkern 1mu} \ {\rm{per}}\;{\rm{plant}})} \over {{\rm{Total}}\;{\rm{plant}}\;{\rm{biomass}}\;({\rm{g}}{\mkern 1mu} \ {\rm{per}}\;{\rm{plant}})}}$$
It is important to note that HI as defined here is based exclusively on the aboveground biomass of newly produced tissues during the evaluation period. Perennial woody tissues, roots, and branches below the marking point were not included; therefore, this index reflects the partitioning of active annual growth only.
Data analysis
All statistical analyses were performed in the R environment (R Core Team, 2022). Initially, residuals from each site-year dataset were tested for normality (Shapiro–Wilk test) and variance homogeneity (Levene’s test). All assumptions were met, and no data transformation was necessary. For the variables leaf area, number of leaves of active vegetative growth, and leaf-to-fruit ratio, treatment means were fitted to simple linear regression models. For vegetative dry matter, fruit dry yield, green beans biomass, and harvest index, linear mixed-effects models were fitted using the ‘lme4’ package (Bates et al., Reference Bates, Maechler, Bolker and Walker2015), with fruit load as a fixed categorical factor and site-year as a random effect. Two model structures were compared: one treating site-year as a fixed effect and one treating it as random. Both were significant for all traits; the random-effects model was selected based on the lower AIC criterion. Regression analyses were subsequently used to characterise the quantitative relationships between variables as shown in Figures 1 and 3. The linear-plateau model for harvest index was fitted using the ‘nlme’ package (Pinheiro et al., Reference Pinheiro, Bates, DebRoy and Sarkar2021).
Vegetation biomass (a) and green beans (b) per tree regarding fruit dry yield. Data are all observations of each coffee plant under coffee-growing conditions (Jacuí-MG 2015/2016, 2016/2017 and Ibiá-MG 2016/2017) and manipulation of the fruit load (100%, 80%, 60%, 40%, 20%, and 0%). The linear models are vegetation biomass: f(y) = −0.392x − 1383.83 (R 2 = 0.40, RMSE = 322.11) and green beans: f(y) = 0.487x + 17.69 (R 2 = 0.93, RMSE = 88.37).
Harvest index regarding fruit dry yield (a) and green-bean biomass (b) of coffee. Data are all observations of each coffee plant under coffee-growing conditions (Jacuí-MG 2015/2016, 2016/2017 and Ibiá-MG 2016/2017) and manipulation of the fruit load (100%, 80%, 60%, 40%, 20%, and 0%). The models are HI f(y) = 0.00054x for x ≤ 1402 and f(y) = 0.75 for x > 1402 (a), and HI f(y) = 0.00104x for x ≤ 724 and f(y) = 0.75 for x > 724 (b). Dashed vertical lines indicate the breakpoint (TFY = 1402 g in a; 724 g in b).
Effect of fruit dry yield and vegetative dry matter on new leaf area (LA), number of leaves in new branches, and leaf-to-fruit ratio per tree of field-grown coffee plants. f(y) = −0.0023x − 8.08 with R 2 = 0.26, RMSE = 1.68 (a); f(y) = −0.608x − 2292.00 with R 2 = 0.14, RMSE = 640.85 (ns, p = 0.062) (b); f(y) = 52.2 × x ^ −1.4147 with R 2 = 0.89, RMSE = 26.13 (c); f(y) = 0.0063x + 1.67 with R 2 = 0.82, RMSE = 0.82 (d); f(y) = 2.2725x + 197.61 with R 2 = 0.84, RMSE = 274.52 (e) and f(y) = 0.2177x + 55.94 with R 2 = 0.61, RMSE = 48.47 (f).
Results
Fruit load effects on vegetation and coffee production
Vegetative dry mass of newly produced tissues above the labelling point ranged, on average, from 1371 g with 0% fruit load to 291 g per tree with full fruit load (Figure 1). With 100% fruit load treatment, fruit dry mass was about ten times the matter accumulated by vegetation over the annual production cycle for both seasons and sites (Figure 1a). Under the same field-growing conditions, green bean production ranged, on average, from no yield with 0% fruit load to 1570 g per tree with full fruit load (Figure 1b).
Coffee harvest index according to fruit load
Harvest index was significantly affected by fruit load (Figure 2). Among the loads, the maximum HI was approximately 0.75, while the minimum HI was 0.1. In general, the HI plateau was reached at 1402 g of fruit dry yield (Figure 2a) or 724 g of green beans per tree (Figure 2b) and remained stable until the full fruit load of 3000 g of fruit dry yield (Figure 2a) or 1570 g of green beans per tree (Figure 2b) over the annual production cycle for both seasons and sites. In contrast, the lowest fruit load resulted in the lowest HI with 97 g of fruit dry yield (Figure 2a) or 52 g of green beans per tree (Figure 2b) under the same field-growing conditions.
Fruit load effects on leaf area, number of leaves, and leaf-to-fruit ratio
Fruit load resulted in strong alterations in the leaf area, number of leaves of active vegetative growth, and leaf-to-fruit ratio (Figure 3). These variables were measured exclusively on newly produced tissues above the labelling point, consistent with the active annual growth framework described in Section Experimental design, and only for the Jacuí-MG 2015/2016 experiment. Increasing fruit load progressively reduced leaf area, number of leaves, and leaf-to-fruit ratio (Figure 3a–c). Total leaf area per tree declined from 10 m2 with no fruit load to 3 m2 with full fruit load (Figure 3a), while the number of leaves declined from 2800 to 870 per tree (Figure 3b). The mean area of individual leaves was 36 cm2 on trees with no fruit load and 34 cm2 on trees with full fruit load, indicating that fruit load had little effect on individual leaf size despite the large reduction in total leaf area per tree. Leaf area, number of leaves, and leaf-to-fruit ratio were all positively and linearly related to vegetative dry matter (Figure 3d–f), with increments of approximately 0.8 m2 of leaf area and 236 leaves per tree for every 100 g of vegetative dry matter produced (Figure 3d and e, respectively). The leaf-to-fruit ratio declined sharply with increasing fruit load following a power function (Figure 3c) and increased positively with vegetative biomass (Figure 3f), ranging from approximately 2 cm2 of leaf area per gram of fruit under high fruit load to values exceeding 100 cm2 per gram under the lowest fruit loads (Figure 3c).
Discussion
Fruit load drives dry matter partitioning in coffee
This study demonstrated that fruit load is a primary regulator of dry matter partitioning in coffee, strongly constraining vegetative growth, canopy development, and HI. Fruit thinning altered reproductive demand early in development and consistently stimulated branch elongation and leaf production, whereas increasing fruit load progressively reduced vegetative biomass accumulation (Figure 1). These responses confirmed that reproductive and vegetative growth acted as competing sinks within the same growth cycle, with fruits exerting dominant control over assimilate allocation, consistent with previous findings for coffee and other perennial fruit crops (Amaral et al., Reference Amaral, Da Matta and Rena2001; Bote and Vos, Reference Bote and Vos2016; Cannell, Reference Cannell1971; Chaves et al., Reference Chaves, Martins, Batista, Celin and DaMatta2012; DaMatta et al., Reference DaMatta, Cunha, Antunes, Martins, Araujo, Fernie and Moraes2008; Franck et al., Reference Franck, Vaast, Génard and Dauzat2006; Kumar and Tieszen, Reference Kumar and Tieszen1976; Vaast et al., Reference Vaast, Angrand, Franck, Dauzat and Génard2005).
Physiological limits to harvest index in coffee
A key finding of this study was the stabilisation of HI at approximately 0.75 under high fruit loads (>1402 g tree−1), which indicated the existence of a biological ceiling to reproductive allocation in coffee (Figure 2). This value reflected the partitioning of active annual aboveground growth and should not be interpreted as a whole-plant harvest index; inclusion of roots and perennial woody tissues in the denominator would necessarily yield lower values. Although fruits receive priority in carbohydrate partitioning (Bote and Vos, Reference Bote and Vos2016; Moscardini et al., Reference Moscardini, Souza, Mazzafera and Favarin2020), they remain dependent on the maintenance of a minimal vegetative structure that sustains photosynthesis, hydraulic continuity, and assimilate transport (Basile et al., Reference Basile, Bryla, Salsman, Marsal, Cirillo, Johnson and Dejong2007; Millard and Grelet, Reference Millard and Grelet2010). The HI plateau suggests that once vegetative biomass approached a critical lower threshold, further diversion of assimilates to fruits became physiologically constrained, such that additional fruit load no longer resulted in proportional gains in reproductive biomass. This response reflects a structural constraint that even under high sink demand, coffee trees must maintain a minimum vegetative ‘baseline’ to support fruit filling (Rakocevic et al., Reference Rakocevic, dos Santos Scholz, Pazianotto, Matsunaga and Ramalho2023). Thus, while fruits retained allocation priority, this hierarchy operated only within physiological limits imposed by the source tissues.
Trade-offs between yield, vegetative growth, and bienniality
While high fruit load and elevated HI maximise current season yield, they carry substantial costs for future productivity. High fruit loads reduced active vegetative biomass by up to 79% compared with the no-fruit treatment (from a mean of 1371 g to 291 g per tree) (Figure 1a), sharply decreasing leaf area and branch development (Figure 3). Quantitatively, the production of each kilogram of fruit dry mass was associated with a reduction of approximately 392 g of vegetative dry mass (Figure 1a). Because vegetative organs provide both the photosynthetic area and the formation of new productive nodes required for future fruiting, such reductions are likely to impair carbohydrate reserves and yield formation in subsequent seasons. In this context, HI values approaching 0.75 represent a high-risk condition associated with increased susceptibility to alternate bearing and long-term tree degeneration, a pattern widely reported in coffee (Cannell, Reference Cannell, Clifford and Wilson1985; DaMatta et al., Reference DaMatta, Ronchi, Maestri and Barros2010). In contrast, growth profiles characterised by moderate fruit load and sustained vegetative development may better support long-term productivity and system stability. These findings suggest that maximising HI in coffee should not be viewed as an agronomic objective per se, but rather as a trade-off between short-term yield and long-term system resilience.
Leaf area reduction overrides photosynthetic compensation
Vegetative traits, particularly leaf area and the leaf-to-fruit ratio, were strongly constrained by increasing fruit load. The active annual leaf area declined sharply from approximately 10 m2 per tree under low fruit load to less than 3 m2 per tree under high fruit load, indicating a pronounced reduction in photosynthetic surface (Figure 3). This structural limitation was further reflected in the marked decrease in the leaf-to-fruit ratio, which declined sharply following a power function from values exceeding 100 cm2 g−1 under the lowest fruit loads to approximately 2 cm2 g−1 under the highest fruit loads (Figure 3c). Assuming two fruits per gram of dry yield (Magalhães et al., Reference Magalhães, Couto, de Queiroz and de Andrade2000), this corresponds to more than 50 cm2 of new leaf area per fruit under low fruit load and only 1 cm2 per fruit under high fruit load, values well below the 15–20 cm2 estimated to support fruit growth without constraining vegetative development in coffee (Cannell, Reference Cannell, Clifford and Wilson1985; Vaast et al., Reference Vaast, Angrand, Franck, Dauzat and Génard2005). Consistently, Rakocevic et al. (Reference Rakocevic, dos Santos Scholz, Pazianotto, Matsunaga and Ramalho2023) reported leaf-to-fruit ratios as low as 17 cm2 berry−1 in heavily bearing coffee trees across contrasting genotypes and water regimes, confirming that trees operating near or below this threshold face significant source limitation regardless of growing conditions. These low ratios indicated that trees carrying high fruit loads operate under severe source limitation. It should be noted that these ratios were calculated from active annual leaf area only; inclusion of pre-existing leaves on older branches would increase the absolute values, though the relative differences across fruit load treatments would remain. Although increased sink demand can induce compensatory increases in leaf-level photosynthetic rates, such responses have primarily been reported under experimental source-sink manipulations (e.g., defoliation or defruiting; DaMatta et al., Reference DaMatta, Cunha, Antunes, Martins, Araujo, Fernie and Moraes2008; Franck et al., Reference Franck, Vaast, Génard and Dauzat2006; Vaast et al., Reference Vaast, Angrand, Franck, Dauzat and Génard2005) or under combined fruiting and irrigation treatments (Almeida et al., Reference Almeida, Ávila, Pérez-Molina, Barbosa, Marçal, de Souza, Martino, Cardoso, Martins and DaMatta2021). In contrast, when canopy reduction arises naturally from high reproductive demand, as observed here, photosynthetic compensation appears limited or absent (Chaves et al., Reference Chaves, Martins, Batista, Celin and DaMatta2012). Consistent with this, Almeida et al. (Reference Almeida, Ávila, Pérez-Molina, Barbosa, Marçal, de Souza, Martino, Cardoso, Martins and DaMatta2021) showed that in field-grown coffee, fruiting reduced individual leaf area by approximately 30% regardless of irrigation, and that branch mortality was substantially higher in fruiting trees; effects attributable to the structural and hydraulic costs of high reproductive demand. Collectively, our findings and the convergent evidence from the literature demonstrated that under high fruit load, dry matter accumulation in coffee is governed primarily by structural source limitation, which is driven by the reduction in active photosynthetic surface area, rather than by impaired biochemical photosynthetic capacity.
Harvest index as a dynamic trait in a perennial crop
Harvest index varied widely (0.1–0.75) across fruit load levels, reflecting the fundamental variability of assimilate partitioning in perennial systems. Unlike annual seed crops, in which vegetative growth largely precedes grain filling (Smith et al., Reference Smith, Rao and Merchant2018), coffee simultaneously supports fruit development, active vegetative growth, and reserve accumulation (Camargo and Camargo, Reference de Camargo and de Camargo2001; Millard and Grelet, Reference Millard and Grelet2010). The observed HI range suggested that coffee trees modulated assimilate partitioning across years: in low-production seasons, plants preferentially invested in vegetative growth (annual branches and leaves) and storage organs (roots and perennial branches), while in high-production seasons, reproductive allocation dominates at the expense of vegetative growth (Chaves Filho and Oliveira, Reference Chaves Filho and de Oliveira2008; Lima Filho and Malavolta, Reference Lima Filho and Malavolta2003). Consequently, HI in coffee is inherently dynamic and context-dependent, responding to the relative intensity of competing sinks rather than converging toward a fixed value. When interpreted in relation to fruit load, vegetative status, and production phase, HI provides meaningful insight into source-sink regulation but should not be interpreted as a static efficiency parameter.
Limitations and future perspectives
The HI reported here was deliberately defined as the ratio of fruit dry yield to the biomass of newly produced aboveground tissues during the annual growth cycle, rather than to total plant biomass. This approach captures the partitioning of active annual growth – the fraction of biomass most directly relevant to seasonal yield formation and management decisions – and represents the first quantitative characterisation of a dynamic HI for coffee. It is important to acknowledge, however, that roots and perennial woody tissues constitute major reservoirs of carbon and nitrogen in coffee trees that buffer seasonal imbalances between assimilation and demand (Millard and Grelet, Reference Millard and Grelet2010), and their dynamics, along with belowground biomass and reserve mobilisation, remain important avenues for future investigation. Additionally, branch die-back and leaf abscission were not systematically quantified; although field observations suggested these were minor, they could marginally affect biomass allocation estimates under high fruit load conditions. Nevertheless, even when calculated from active annual leaf area only, the leaf-to-fruit ratios reported here fell within the range reported for heavily bearing coffee trees in independent studies using whole-canopy measurements (Rakocevic et al., Reference Rakocevic, dos Santos Scholz, Pazianotto, Matsunaga and Ramalho2023), suggesting that the relative differences across fruit load treatments are biologically meaningful despite the methodological scope. Future studies could further extend this framework to include bean quality parameters such as bean size and density, which are known to respond to fruit load and would provide additional practical guidance for crop management. Overall, these results showed that coffee operates under a constrained source-sink system in which reproductive dominance is substantial but physiologically bounded, highlighting the trade-off between maximising current yield and maintaining sufficient vegetative structure to support both present and future productivity in perennial cropping systems.
Conclusion
This study demonstrated that fruit load is a key regulator of active vegetative growth and harvest index in Coffea arabica grown in the Cerrado-Savanna transition zone of southeastern Brazil. Increasing fruit load strongly suppressed vegetative biomass accumulation by up to 79% relative to the no-fruit treatment. The coffee HI, which has not been previously quantified, varied widely from 0.1 to 0.75 as a function of fruit load. Importantly, fruit loads exceeding 1402 g of fruit dry yield did not result in further increases in HI, which stabilised at approximately 0.75. This plateau indicated a physiological upper limit to reproductive allocation, reflecting the need to maintain a minimum vegetative structure to sustain fruit development. Overall, these findings advanced the understanding of source-sink regulation in coffee by quantitatively linking fruit load, vegetative growth, and HI within a perennial framework. By providing a practical means to characterise biomass partitioning as a function of reproductive demand, this study offered valuable insights for managing fruit load to balance short-term yield and long-term vegetative sustainability in coffee production systems.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0014479726100398.
Data availability statement
The data that support the findings of this study are available from the corresponding author upon reasonable request.
Author contributions
L.T.S.: conceptualisation, methodology, formal analysis, investigation, data curation, writing – original draft. J.L.C.B.: data curation, writing – review & editing. M.V.O.N.: formal analysis, writing – review & editing. A.F.B.R.: methodology, formal analysis, writing – review & editing. D.D.N.: supervision, funding acquisition, writing – review & editing. J.L.F.: supervision, conceptualisation, project administration, writing – review & editing.
Funding statement
This work was supported by the Brazilian Federal Agency for Support and Evaluation of Graduate Education (CAPES) (grant number 33002037003P0) and the State of Sao Paulo Research Foundation (FAPESP) (grant number 2017/00315-2). CAPES and FAPESP had no role in the design of the study, the analysis or interpretation of data, or the writing of this manuscript.
Competing interests
The authors have no relevant financial or non-financial interests to disclose.
Open access
