Four treatment sites included coffee farms shaded by either Senna siamea (Lam.). Irwin and Barneby (Senna), Leucaena leucocephala (Lam.) de Wit (Leucaena), Durio zibethinus L. (Durian), and one unshaded farm (Open), were used to determine the influence of different shade tree types on light intensity, flowering, fruit set, and fruit fall in Robusta coffee. Measurements were taken during December 2014 (the harvesting period) to the end of May 2015 (the pinhead stage of fruit). These three shade tree species are evergreen and long-lived with an open canopy and good drought tolerance [30] . The slope of the land at all sites was less than 3%; the altitude is ranging between 634 - 700 m a.s.l. Total shade cove per ha was 34% for Leucaena site (60 trees ha⁻¹), 17% for Senna (35 trees ha⁻¹) and 14% for Durian (46 trees ha⁻¹).

Shade tree layout in regard to coffee bushes and the location of light measurements are provided in Figure 1. Each treatment (shade tree type) was designed with five plots/blocks, each one containing one shade tree (see Figure 1 for an example plot). Three coffee bushes north and south of the stem of the shade tree were assessed in each plot, to give six coffee bushes allocated evenly in two directions of the shade tree. Only the two coffee bushes, in either direction, closest to the shade tree were shaded; the others was unshaded.

The position at the stem of the shade tree was marked as 0 m and the distance from this to the nearest coffee bush, in both directions, was 1.5 m. The distance between coffee bushes was 3 m and hence the distance from the stem of the shade tree to the farthest coffee bush was 7.5 m. The positions of the three coffee bushes, ranging from the nearest to farthest from the shade tree, were therefore labeled 1.5, 4.5 and 7.5 m respectively. At the Open site, five plots with six coffee bushes per plot were randomly allocated. For each coffee bush, measurements were also taken in the vertical plane on both the north and south side of each plant. These were taken from three levels, branches on the top, middle and bottom (Figure 1).

An Electric Calibration Line Quantum² sensor (model MQ-306 with 6 Sensors, Apogee Instruments) was used to measure light intensity (μmol∙m⁻²∙s⁻¹). Readings were taken at between 11:00 am and 1:30 pm on clear, sunny days once the sun was at a right angle (90˚) to the ground [31] at three times: 2 February, 24 March and 14 April 2015. Light measurements were taken north and south of each shade tree at the three shade tree sites and Open site. Measuring light intensity under shaded trees, each direction, light was measured at 0.0 m (shade tree stem), 1.5 m, 3.0 m, 4.5 m, 6.0 m and 7.5 m from the shade tree to outside of the shade canopy (as illustrated in Figure 1).

Five coffee bushes were selected from under the canopy of the shade trees at Leucaena and five bushes at Open to determine light variation caused by coffee self-shading and shade trees. At Leucaena, coffee bushes located at

4.5 m on the north side of the shade tree were used. Because the height of trees was consistent at 180 cm, each coffee bush was measured at six locations every 30 cm starting at and above 30 cm above the ground (Figure 1). Readings were taken at three times: 4 February, 25 March and 14 April.

Counts were conducted 3 - 5 days after watering. The total number of flowers per branch was calculated. Measurements were taken from coffee bushes chosen to measure light intensity (three bushes each on the north and south side of the shade tree) (Figure 1). Three randomly chosen branches; one each at the top (around 170 cm above ground), middle (90 cm above ground) and bottom (45cm above ground) of the coffee on both the north and south side of each coffee bush were used to give a total of six branches per bush. Total flower buds per branch were counted separately (Figure 1).

The percentage fruit set and fruit fall was calculated from branches also used for measuring number of flowers per branch. Fruit set was measured in March when flowers had finished the pollinating process; approximately one month after blossoming. First fruit drop was measured during May at the “pinhead” stage; 3 - 4 months after the first stage of fruiting [32] [33] . The fruit set proportion equals total fruit set/total flowers and fruit fall proportion equals total fruit/total fruit set.

Figure 4 shows that there was no significant difference in the number of flowers branch⁻¹ bush⁻¹ between

shaded and unshaded sites. Coffee shaded by Durian which had 278 flowers branch⁻¹ bush⁻¹, tended to have a little higher than others sites.

In terms of vertical position, shaded and unshaded sites both produced the highest number of flowers at the top of coffee bush (336 and 327 flowers branch⁻¹), followed by the middle (251 and 252 flowers branch⁻¹) and the bottom (175 and 183 flowers branch⁻¹). Furthermore, data indicated that there was a significant difference in the amount of flowers branch⁻¹ bush⁻¹ on the vertical positions (P < 2e⁻¹⁶).

The average proportion of fruit set branch⁻¹ bush⁻¹ for all treatments was above 82% (84.1%, 83.3%, 82.9% and 82.6% for Leucaena, Senna, Open and Durian, respectively) (Figure 5), and there were no significant differences between any of the treatments (P = 0.4). There was also no effect of direction (P = 0.13) (N or S) nor position (P = 0.55) from the shade tree on the proportion of fruit set, however, vertical position was critical (P = 4e⁻¹¹) with the top and middle of coffee bushes recording more fruit set than the bottom.

The effect of shade tree type on the first fruit drop for the four sites was highly variable, ranging between 2.4% to 4.7% (Figure 6). Average fruit drop was significantly higher in full sun (4.7%) compared to shaded sites (3.1%) with P = 2.43e⁻⁸, however, there was similar fruit fall rate in the three shade tree types; Leucaena (2.4%), Senna (3.6%) and Durian (3.4%). Vertical position did have a significant effect (P = 4.9e⁻¹¹), fruit drop showed upward trends following increasing the vertical height. While fruit fall was about 2.3% at the bottom, it increased to 3.9 and 4.4% at the middle and the top, respectively.

In terms of flowering, there were no significant differences in the average number of flowers per branch per bush between shaded and unshaded coffee. This finding contrasts with previous work that found that shaded coffee produced fewer flowers as a result of fewer flower buds forming per branch, which related the reduction in light intensity [12] - [16] . Our finding also contrasts with, but in the opposite way, the work of Lin [4] , who found that coffee grown under shade had more flowers per bush and flowers per node than coffee grown in full sun.

A possible reason for the difference in our findings with the work of others, is that our study was conducted in Vietnam where light intensity is relatively high [30] . Thus under shade trees, the light intensity at the top of coffee bushes (approximately 1000 - 1200 μmol∙m⁻²∙s⁻¹) can still be higher than the optimum suggested for coffee growth (500 - 600 μmol∙m⁻²∙s⁻¹) [15] [32] [33] . Additionally, because the coffee bushes were grown at dimensions of 3 m × 3 m, shading by adjacent coffee bushes is unlikely to occur, as can happen in bushes with close spacing (3 m × 2.5 m) [32] [44] . Other reasons for the lack of difference between flowering at the open and shaded sites, is that upright growth was prevented due to height control through pruning [32] [45] and lateral branches or secondary growth were regularly pruned. This meant that all bushes were evenly spaced and sufficiently far apart to allow light and air to penetrate the foliage [46] . Finally, intensive coffee cultivation in the sites can also could contribute to successful flowering, resulting in insignificant different in flowers [32] [44] . However, in our case, the number of flower per node per branch and amount of node per branch are factors building yield was not investigated.

With respect to vertical position in the coffee canopy, both shaded and unshaded coffee followed the same trend, whereby the number of flowers per branch per bush was similar for the top and middle of the canopy, but these were significantly higher than for the bottom. This result is similar to Lin [4] , who reported that self-shad- ing caused significant light limitations and was highly likely to reduce flower bud initiation; the lowest branches and innermost branch nodes produced significantly fewer flowers than the outermost nodes and high and medium height branches [47] .

A decrease in light intensity was also shown to inhibited flower bud initiation by Can nell [14] . This could be reasonable in our study, because the light intensity at 30 - 90 cm aboveground bushes in our study was about 400 μmm⁻²∙s⁻¹, which is slightly lower than the saturation light intensity for coffee production. Because the lower branch position (inside canopy and close to ground) grows slowly and poorly as a result of light and nutrient competition [33] [48] , it forms fewer fruit nodes and young branches, which results in lower yield compared to positions above [21] . This further affects yield in this position because flowering is initiated at only one time on the young and strong branch nodes (growth of which is limited in the bottom position) formed the previous year [15] [32] [33] [44] [48] .

Fruit set is an important process in achieving good coffee yield. Studies recognised that coffee, in general, has a low proportion of fruit set, for example, in Brazil, fruit set of Arabica coffee has been reported at around 29% [45] to 57% [49] . Robusta, on the other hand, has an even lower percentage of fruit set than Arabica. Some authors report Robusta fruit set as less than 30% [45] , although others have reported that it can be up to, but still less than, 50% [50] . In contrast, in Vietnam, fruit set has been reported to be 90% - 95% in Arabica and 60% - 79% in Robusta [32] . The proportion of fruit set in Robusta in this study was consistent for both the open and shaded sites and averaged 82%, which is slightly higher than that reported by Tiem [32] and much higher than that reported for other countries [45] [50] .

Whereas Arabica coffee is self-fertilising, Robusta requires cross-fertilisation, therefore, compared to Arabica coffee, the pollination process in Robusta coffee can be affected by many additional factors. The two main factors that strongly influence pollination are genetic and external conditions [30] . External factors affecting pollination include cold or hot temperatures, wind speeds, heavy rains, insect species [51] - [53] , the number of flower openings [32] ; the number of leaves and flowers on branch, and flower atrophy [15] [32] [33] [45] . Flower numbers and flower opening time could explain the high fruit set proportion observed in our study, as demonstrated in other studies that found that a large number of flowers and > 90% flower bud opening was positively correlated with fruit set [27] [29] .

The different fruit set proportions observed for the different vertical branch positions (top, middle and bottom) was also found by Reis and Arruda cited by [45] , whereby the lower the branch position has less the proportion of fruit set. This may be a result of restrictions associated with pollen transport due to fewer flowers [32] , pollinators [51] [53] and branch and leaf cover [15] [32] .

After pollination, fruit experiences many growth periods before finally forming a bean at the ripening stage. Depending on coffee cultivar and region, in Vietnam, fruit drop for Robusta is defined by three periods. The first period occurs 2 - 4 months after flowering and is commonly known as the pinhead drop. The second stage is endosperm filling at 3 - 5 months, and the last is fruit growth after 6 months [32] [33] . Fertilisation failure, and hence first fruit fall, can be caused by climate conditions, such as high or low light intensity [15] , high temperature [4] and humidity [15] [27] [32] , and also the number of pollinators [51] . In our study, significant difference in fruit fall occurred between the top, middle and bottom of the bush, however, this was most likely related to over-production of flower buds [15] rather than other condition as these would have largely consistent across the sites.

Average fruit fall at the four sites in our study ranged between 2.4% and 4.7%, with less fruit fall in the shaded sites compared to the Open site. This contradicts the study of Campanha, Santos [21] , who found that Arabica coffee grown under shade in an agroforestry system, hada higher proportion of fruit fall per branch (48%) compared to Arabica grown without shade (11%). However, the amount of fruit set and defoliation in coffee may be unrelated to pinhead drop because during this stage fruit growth is negligible as fruits are dormant [15] [32] .

Because fruit fall is expected to be at its lowest in the first fruit fall period (pinhead drop), resulting from pollination failure [43] . The majority of fruit fall actually occurs during the second period as the endosperm fill between June and July [32] ; and because fruit set may be unrelated to pinhead drop [15] [32] , our measurement offruit fall at this stage is a limitation of our research. Therefore, it would be of benefit to include other stages in further research, especially for purposes of relating fruit fall with fruit set and yield.