Materials and methods

Our data included annual yields from grower-run trials using a completely randomized design conducted by the Citrus Research and Development Foundation (CRDF), Lake Alfred, FL. There were three trials in different locations within Florida; two sites were in central Florida (Lake Placid and Babson Park) and one site was in southwest Florida (Felda). Each site was managed by a different grower. At each site, the following rootstocks were evaluated: UFR-2 (a complex allotetraploid from somatic hybrids {[Citrus clementina × (Citrus paradisi × Citrus reticulata)] + Citrus grandis × (C. reticulata + Poncirus trifoliata)}; UFR-3 and UFR-4 (same pedigree as UFR-2); UFR-16 (Citrus maxima × Citrus depressa); US-812 (C. reticulata × P. trifoliata); and US-942 (C. reticulata × P. trifoliata). Rootstock US-897 (C. reticulata × P. trifoliata) was also evaluated at the Babson Park site. The acronyms UFR and US denote rootstocks developed by the University of Florida and the USDA, respectively. Table 1 shows the list of rootstocks and their corresponding release date and number of propagations during 2019–20 statewide (Florida Department of Agricultural and Consumer Services, 2020), which together account for ≈53% of the total citrus propagations in the state.

Table 1.

List of citrus rootstocks with the corresponding release date and number of propagations during 2019–20 that were evaluated using annual yields from grower-run trials. The acronyms UFR and US denote rootstocks developed by University of Florida and USDA, respectively.

View Table

Most trees at each location were planted in 2015. However, trees on rootstocks UFR-3 and UFR-16 in Lake Placid and Babson Park were planted between 2 and 12 months later than those on the other rootstocks. Because of such differences in planting time, we did not analyze such rootstocks (in Lake Placid and Babson Park) because of concerns regarding the variability induced by the different growing conditions (i.e., weather and growing time) those trees were subjected to relative to all other trees.

The trials also included trees on traditional commercial rootstocks used in each region, which were used as a control. Therefore, in Lake Placid, the rootstock used as control was sour orange (Citrus aurantium); in Babson Park, it was Carrizo citrange (C. sinensis × P. trifoliata); and in Felda, it was Swingle citrumelo (C. paradisi × P. trifoliata). Each rootstock was replicated five times; therefore, there were 25 plots in Lake Placid, 30 plots in Babson Park, and 35 plots in Felda. In Lake Placid and Babson Park, yield data were collected during three seasons, 2017–18, 2018–19, and 2019–20; however, in Felda, data were also collected in 2016–17.

The trials in Lake Placid and Babson Park included 144 trees per plot (12 rows wide and 12 trees in each row). However, the trial evaluation focused on a subplot within the 144-tree main plot. The subplots at those two sites consisted of 64 trees (eight rows wide and eight trees in each row). The site in Felda was bedded and the main plots consisted of 126 trees (seven rows wide and 18 trees in each row). The data were also collected from a subplot, which, in this case, consisted of 70 trees (five rows wide and 14 trees in each row). The tree planting density for each site was as follows: Lake Placid, 240 trees per acre (22 × 8.25 ft [between rows and in rows, respectively]); Babson Park, 303 trees per acre (18 × 8 ft); and Felda, 202 trees per acre (21.5 × 10 ft). Table 2 shows a summary of the site descriptions. All trees were budded with ‘Valencia’ sweet orange (clone 1-14-19). ‘Valencia’ is the predominant late orange cultivar produced in Florida and has accounted for ≈55% of the bearing area of oranges grown in the state during the past few years.

Table 2.

Description of the trial sites within Florida, evaluated citrus rootstocks, planting date of trees, tree spacing, tree density, and soil series. The acronyms UFR and US denote rootstocks developed by University of Florida and USDA, respectively.

View Table

By having the same scion grafted on different rootstocks, we can make inferences regarding the tolerance of the rootstocks to HLB based on the differential yield and revenue they provide. However, the statistical analyses and comparisons we conducted were performed based on each individual site because groves at each site were managed by different growers who used different inputs and input levels (i.e., fertilizer, foliar nutritionals, insecticide sprays, etc.). Additionally, other sources of variability across sites included soil type, irrigation practices, tree density, weather, and level of asian citrus psyllid population, resulting in different HLB infection rates. By conducting our analyses by site, the only differing treatment across plots within a site was the rootstock because all other major sources of variability were kept constant. By performing our inferences in this way, we avoided the shortcomings that have been noted by Bowman et al. (2016), who pointed out that because the performance of citrus rootstocks is affected by many different factors, the relative rootstock field performance will often vary considerably between trials in different locations.

Throughout the duration of the trial, annual data regarding the CLas infection was also collected by conducting polymerase chain reaction (PCR) analyses of leaf samples from sentinel trees in each plot. There were eight sentinel trees per plot. The total number of plots was that for the full trial, which included the rootstocks we did not include in the statistical and economic analyses because of delayed planting. Therefore, there were 35 plots in Lake Placid and Felda and 40 plots in Babson Park. The cut-off cycle threshold (Ct) values used to determine whether a tree was CLas-negative was ≥36.

To analyze the yield data, we used regression analysis. In particular, we used ordinary least squares (OLS) with clustered ses (by rootstock and plot) because, if there is a positive correlation of the errors within a cluster, then an additional observation in the cluster does not provide a completely independent piece of new information. Therefore, using default ses can lead to using ses that are too small, resulting in incorrect inferences by over-rejection of the true null hypotheses (Cameron and Miller, 2015). Instead, by using clustered ses, we obtained robust estimates of correlations among errors of the same rootstock/plot and heteroscedasticity over time.

Results

Table 3 shows the progression of CLas infection by site and sampling date. In the first sampling, the percentage of infected trees was quite heterogeneous among the three sites. In Babson Park and Felda, the sampling was conducted ≈9 and 10 months after planting, and the levels of infection were 22% and 14%, respectively. In contrast, even though the sampling was performed 6 months after planting in Lake Placid, the percentage of infected trees was 51%. Such heterogeneity in infection rates across sites could be attributed to the differences in asian citrus psyllid spraying programs among the collaborating growers (as well as that of their neighbors). Importantly, however, in the fourth sampling, which occurred 43, 46, and 47 months after planting, the level of infection was more than 90% at all three sites. To the best of our knowledge, those field estimates are the first to show how quickly the disease progresses within a new grove in a commercial setting and provide evidence of how important it is for growers to find ways to optimize grove practices and management to manage the disease.

Table 3.

Candidatus Liberibacter asiaticus (CLas) date of sampling, number of sentinel ‘Valencia’ sweet orange trees sampled, and infection percentage of sentinel trees by site within Florida.

View Table

Table 4 shows the yield data by site and season. Table 5 shows the regression results by site. Importantly, the high R² (of at least 87%) in all three regressions indicate that most of the variations in yield are accounted for by the covariates included in the model. The regression coefficients of each of the intercepts denote the average yield during the first season across all rootstocks, and the coefficients for each season dummy represent the difference in yield of the control rootstock through time relative to that average. The coefficients of each of the interaction dummy variables represent the average differential yield of rootstock i with respect to the control in season t. For example, in Lake Placid, the coefficient for rootstock UFR-2 season 2018–19 denotes that such rootstock attained, on average across all replications, 229 lb/acre solids less relative to sour orange that season. All coefficients in the regression for Lake Placid were statistically significant either at the level of 1% or 5% level, except that for rootstock US-942 season 2019–20. In Babson Park, the only statistically significant coefficients were those for rootstock UFR-2 during both seasons and for rootstock UFR-4 season 2018–19 (all of which were negative). In Felda, the coefficients that were statistically significant were those for rootstock UFR-3 season 2019–20 (which was negative) and US-812 and US-942 during season 2018–19 (which were positive).

Table 4.

Annual ‘Valencia’ sweet orange yield data by site within Florida and harvest season. The treatments were the different citrus rootstocks at each site. The control citrus rootstocks in Lake Placid, Babson Park, and Felda were sour orange, Carrizo citrange, and Swingle citrumelo, respectively. The acronyms UFR and US denote rootstocks developed by University of Florida and USDA, respectively.

View Table

Table 5.

Regression results of the differential ‘Valencia’ sweet orange yields in pounds of solids per acre (1 lb/acre = 1.1209 kg⋅ha⁻¹) relative to the control by citrus rootstock, season, and site within Florida. The treatments were the different citrus rootstocks within each site. The control citrus rootstocks in Lake Placid, Babson Park, and Felda were sour orange, Carrizo citrange, Swingle citrumelo, respectively. The acronyms UFR and US denote rootstocks developed by University of Florida and USDA, respectively.

View Table

View Table

Figure 1 illustrates the results of the regressions for each site through time, which is useful to visualize the differential yields of the evaluated rootstocks through the different seasons. Figure 1 panels B and C evidence that the positive yield differentials of the U.S. rootstocks relative to the control observed during season 2018–19 became statistically nonsignificant in 2019–20. Such results could be attributable to decreased performance of the rootstocks or decreased input use by growers as a response to lower fruit prices (Singerman, 2020), or a combination of both. We were unable to determine the underlying cause of those yield changes with the data we had available. However, Table 4 shows that in Felda, the yield of US-812 and US-942 decreased in 2019–20 relative to that of the previous season, whereas that of Swingle citrumelo increased slightly. This would imply that if the cause was a decrease in input alone, then US-812 and US-942 were more sensitive to such change compared with Swingle citrumelo. Table 4 also shows that in Babson Park, the yield of all rootstocks increased relative to the previous season. Even though the yield of US-812 was still the highest among all rootstocks in that site in 2019–20, its increase in 2019–20 relative to the previous season was not as high as that of Carrizo citrange, thereby making the differential yield nonstatistically significant. This would suggest that US-812 is a rootstock that attains higher yields (and revenue) sooner but that Carrizo citrange catches up after a few seasons. Although these are inferences, they could be valuable considerations for growers to use to decide which rootstock to use. To be able to effectively establish the driver of the changes in yield during the last season, it is important to continue data collection for these (and other) trials.

View Full Size

View Full Size

Illustration of regression results of the differential ‘Valencia’ sweet orange yield in pounds of solids per acre relative to the control by citrus rootstock, season, and site within Florida: (A) Lake Placid, (B) Babson Park, and (C) Felda. The treatments were the different citrus rootstocks within each site. The control rootstocks at each site were sour orange, Carrizo citrange, and Swingle citrumelo, respectively. The acronyms UFR and US denote rootstocks developed by the University of Florida and USDA, respectively. Statistical significance: *** P < 0.01 and ** P < 0.05. 1 lb/acre = 1.1209 kg⋅ha⁻¹.

Citation: HortTechnology hortte 31, 4; 10.21273/HORTTECH04784-20

• Download Figure
• Download Figure
• Download figure as PowerPoint slide

Revenue differentials: Statistical significance and economic importance

Based on the yield data, we also computed the differential revenue by season of each of the rootstocks relative to the control at each site. Table 6 shows the differential revenue of each rootstock with respect to the control. The prices we used to calculate revenue were the actual average prices for ‘Valencia’ sweet oranges during each season as reported by the Florida Department of Citrus (Florida Department of Citrus, 2017, 2018, 2019, 2020). Therefore, the prices for 2016–17, 2017–18, 2018–19, and 2019–20 were $2.85, $3.09, $2.42, and $1.25 per pound of solids, respectively. By using the market prices for the seasons for which we had data, we were able to provide estimates of the actual revenue resulting from the yield attained by each rootstock during these trials. Using a 5-year average of prices instead, for example, would not change the ranking of the rootstock based on revenue; it would just provide a different revenue amount.

Table 6.

Differential ‘Valencia’ sweet orange revenue by citrus rootstock, season, and site within Florida relative to the control. The treatments were the different citrus rootstocks within each site. The control rootstocks in Lake Placid, Babson Park, and Felda were sour orange, Carrizo citrange, Swingle citrumelo, respectively. The acronyms UFR and US denote rootstocks developed by University of Florida and USDA, respectively. Estimates of revenue were obtained using actual average prices for ‘Valencia’ sweet oranges during each season as reported by the Florida Department of Citrus (2017, 2018, 2019, 2020). The prices for 2016–17, 2017–18, 2018–19, and 2019–20 were $2.85, $3.09, $2.42, and $1.25/lb solids, respectively ($1/lb = $2.2046/kg).

View Table

Table 6 shows the differential revenue for all rootstocks, not just those that we found to have statistically significant differences in yield relative to the control. The reason for doing so is because even though the differences in yield of some of the rootstocks were not statistically significant, the differences in revenue they attained relative to the control are large enough to have practical importance for growers. As argued by McCloskey (2005), statistical significance should not be used as an all-purpose way of judging whether a number is large enough to matter. As pointed out by Kirk (1996), “while statistical significance is concerned with whether a result is due to chance or sampling variability; practical significance is concerned with whether the result is useful in the real world” (p. 746).

Table 6 illustrates the aforementioned point for UFR-16 in Felda. None of its coefficients attained a statistically significant difference in yield relative to the control (Table 5). However, as shown in Table 6, the cumulative revenue UFR-16 attained over the course of three seasons was −$761/acre compared with that of Swingle citrumelo. In contrast, rootstock US-812 in Felda attained a statistically significant difference in yield relative to the control only in 2018–19, but its cumulative revenue over the course of three seasons was $759/acre. Although none of the coefficients for UFR-16 was found to be statistically significant during all four seasons, and although only one coefficient was found to be statistically significant for US-812, there is still a meaningful economic difference between the two rootstocks that should not be ignored; a grower would not be indifferent between obtaining $760/acre more than the control as opposed to $760/acre less than it. Although the data available are only those of the first few seasons, early differences in yield and revenue are particularly important to growers’ profitability and sustainability at a time when margins are becoming particularly narrow, if positive at all.

Table 6 shows that in Lake Placid, both UFR-2 and UFR-4 attained a lower level of revenue compared with sour orange during each of the seasons for which we had data; therefore, there was negative cumulative revenue over the course of three seasons of $866/acre and $715/acre. In contrast, US-812 and US-942 attained cumulative revenues of $506/acre and $468/acre more than the control, respectively. Despite the similar cumulative revenue, US-812 seems to be a better choice because, even though in 2019–20 both rootstocks attained a lower positive difference with respect to the control, that of US-812 was larger. It is also noteworthy that negative and positive revenue differences for all evaluated rootstocks decreased in 2019–20 compared with 2018–19 because of lower yield or lower fruit prices.

In Babson Park, rootstocks UFR-2 and UFR-4 attained negative revenues during all three seasons compared with Carrizo citrange, with cumulative revenues of −$1793/acre and −$974/acre, respectively. Therefore, it would be more beneficial if growers used Carrizo citrange to graft ‘Valencia’ rather than UFR-2 or UFR-4 when using such a soil profile and management. Although US-897 performed somewhat better compared with the UFR rootstocks, it also attained negative cumulative revenue. Both US-812 and US-942 attained positive revenue differences compared with Carrizo citrange during all three seasons. However, at $1012/acre, US-812 attained a cumulative revenue that doubled that of US-942.

In Felda, trees on rootstock UFR-2 showed performance similar to that of Swingle citrumelo, but with an intrinsic higher risk because its performance through time is still unknown. The cumulative revenue over the course of four seasons of trees on rootstocks UFR-3 and UFR-16 were similar; they obtained revenues of −$776/acre and −$761/acre relative to Swingle citrumelo, respectively. Therefore, it would benefit growers to use Swingle citrumelo to graft ‘Valencia’ when using such a soil profile and management. US-812 and US-942 resulted in the largest differences in yield and revenue compared with the control, but the differences decreased considerably in 2019–20 because of the combined decrease in yield differentials and prices. Although UFR-4 did not attain large differences, increases over the course of seasons would indicate that it is worthwhile to continue data collection to evaluate its performance over longer time periods.

Discussion

We compared ‘Valencia’ sweet orange yield data from side-by-side grower-run trials performed at three locations in Florida. The treatments consisted of scions grafted onto different rootstocks recently developed by the University of Florida and the USDA, as well as traditional rootstocks that were used as controls. Although the data were from trees during their first bearing years, the results of our analysis indicate that rootstocks can be a significant factor affecting yield and economic return. Moreover, finding rootstocks that yield returns (and revenue) sooner can be particularly valuable for growers to achieve breakeven (and profits) more quickly. Therefore, our estimates should prove useful to citrus growers during their decision-making process when determining which rootstock to use (and which not to use) in the current HLB-endemic environment in Florida.

The rootstocks that attained statistically positive differences in yield relative to the corresponding control included US-812 and US-942 in Lake Placid during two seasons and one season, respectively; rootstock US-812 in Babson Park during one season; and rootstocks US-812 and US-942 in Felda during one season. The rootstocks that attained statistically negative differences with respect to the corresponding control included rootstocks UFR-2 and UFR-4 in Lake Placid during two seasons; rootstocks UFR-2 and UFR-4 in Babson Park during two seasons and one season, respectively; and rootstock UFR-3 in Felda during one season. Importantly, some of the rootstocks attained meaningful positive and negative differences in revenue compared with traditional rootstocks even when none, or just one, of their coefficients attained a statistically significant difference in yield relative to the control. Those cases are illustrated by the positive (negative) differences in revenue attained by rootstock US-812 (UFR 4) in Babson Park and rootstocks US-812 and US-942 (UFR-3 and UFR-16) in Felda.

Our analysis had some limitations. The positive differences in performance attained by some of the evaluated rootstocks decreased during the last season for which we had data. Therefore, it remains to be seen whether the observed trend in performance will continue beyond the first few seasons. Follow-up studies that include more seasons are needed to address such an issue. It should also be clear that our findings have no implications for how well (or not) the rootstock evaluated would perform with other scions (e.g., early season and midseason) or in other locations. Even for the rootstocks/scion combinations evaluated here, our findings are the result of specific growing conditions (e.g., grower practices and weather) and should not be viewed as a guarantee of performance.

Units

View Table