2.7 Characters for selection

2.7.1 Objective

Remember that the aim is to select parent palms for the production of tenera offspring with high yield of oil and kernels per hectare. Basically, this means developing tenera which optimize the transfer of available resources of the physiological environment (solar radiation, temperature, humidity, soil water and nutrients, etc.) into oil and kernels. Among these resources, interest in oil palm breeding focuses mainly on optimizing the supply of assimilate, produced in the process of photosynthesis, and its transfer into economic products. An understanding of the carbon budget, as reported by Breure (1987) for the oil palm, is therefore crucial to develop the desired tenera ideotype, i.e. a biological model which is expected to perform in a predictable manner (Donald, 1968).

2.7.2 Partition of carbohydrates

Briefly, carbohydrates produced in the process of photosynthesis are, as a first priority, used for the maintenance of existing biomass (maintenance respiration). Priority is then given to the production of vegetative dry matter. Once requirements for maintenance respiration and vegetative growth are satisfied, carbohydrates are allocated to bunch production. Oil-and-kernel production thus strongly benefits from increasing photosynthetic production above a certain threshold level and reducing carbohydrate requirements for maintenance and vegetative growth. Maintenance requirements can be calculated from the protein and mineral content of plant tissue and their metabolic activity (van Kraalingen et al., 1989). But this method is not suitable to quantify, and thus to select for, maintenance respiration in breeding work. Therefore, only selection for reduced vegetative requirements is considered in this present report.

2.7.3 Harvest Index

The greatest scope for increasing yield appears to be by selecting for high Harvest Index (HI), i.e. the proportion of dry matter used for the production of oil-and-kernels (Hardon et al, 1972; Breure & Corley, 1983). Little progress can be achieved by direct selection for HI because of its low heritability (Breure & Corley, 1983; Breure & Bos, 1992).

HI can be increased, according to the carbon budget, in two ways: (i) by reducing carbohydrate requirements for Vegetative Dry Matter production (VDM) and (ii) by increasing photosynthetic production above a certain threshold level. Both options are achieved if the reduction of VDM is restricted as much as possible to non-photosynthetic tissue, that is, by increasing Leaf Area Ratio (LAR), defined as the ratio of the new total leaf area produced to new VDM.

Breeding for speed of photosynthesis without an accompanying increase in VDM can also be achieved through increasing magnesium content in the leaves. It is, however, not obvious that magnesium per se increases photosynthetic production.

Magnesium content in oil palm leaves could be accompanied with other decisive photosynthetic components, such as those factors involved during the biosynthetic process of photosynthesis. This mechanism comprises different biochemical reactions during photosynthesis, which are regulated by enzymes during CO2 fixation. Shibles (1993) found in annual crops that the photochemical process in C3 plants, like oil palm, is not the bottleneck in terms of photosynthtic efficiency. It is therefore possible that besides rising magnesium content, a parallel rise in relevant enzymes also occur into the mesophylic cells; it may be actually this change which is measured as leaf-Mg level. Nevertheless, Peaslee & Moss (1966) demonstrated that magnesium concentration in the leaves is closely related to photosynthetic production, and Breure (1986) showed that parent selection for magnesium content in the leaves indeed increases HI.

More recently Breure & Bos (1992) confirmed in a multiple regression analysis that HI benefits from selection for high GCA values of LAR and leaf-Mg. They also found that reducing Leaf Production (LPR), the main component of VDM, positively affects HI. This is not surprising as reducing LPR will not affect light interception (and thus photosynthetic production), because a rather constant number of crown leaves is maintained through regular leaf pruning for harvesting the bunches. It does, however, diminish the proportion of carbohydrates allocated to vegetative growth to the benefit of bunch yield. In the same vein, reducing vertical trunk growth is expected to increase HI as the trunk is not contributing to photosynthesis; but this could not be substantiated in the study of Breure & Bos (1992). They argued that the gain in diminishing carbohydrate requirements is probably outweighed by the associated reduction in competitive ability for light and thus photosynthetic production when progenies are planted in a mixture. Thus in a more uniform population, the advantage of slow height increment would probably be greater. In progeny tests the benefit of slow height increment may better show up when progenies are grouped in the field according to the vigour established in the nursery, as is explained in section 2.10.

HI can thus be increased by selection for low GCA values for VDM, with LPR as the main component, low values of height increment, and high values of LAR and Leaf-Mg.

2.7.4 Crown expansion

The canopy may take up to 6 years to close (Squire & Corley, 1987). At that stage photosynthetic production per hectare, and thus bunch yield, has reached its maximum, and Breure (1985) showed that selection for quick canopy closure indeed benefits early yield.

The effect of crown expansion on the production and partitioning of carbohydrates is a crucial aspect of oil palm breeding.

Before canopy closure the rate of increase in the area of individual leaves, or when taking into account the number of green leaves in the crown, the Leaf Area Index (LAI), i.e. the total leaf area per unit ground area, is directly related to photosynthetic production per hectare, and thus carbohydrates allocated to bunch yield. Crown expansion is influenced by husbandry practices and, as shown by Breure (1985), also by genotype.

Once the canopy is practically closed, photosynthetic production per unit area has reached its maximum, and both bunch production and vegetative growth by and large stabilise (Breure, 1988). The yield pattern in the period following canopy closure seems to depend, among other things, on the rate at which the crown continues to expand (Breure, 1988). Prolonged expansion does not further contribute to photosynthetic production per unit area, as the canopy is closed already, but does increase carbohydrate requirements for vegetative growth and maintenance respiration at the expense of those allocated to bunch production. The result is a decrease in HI with age.

The search is thus for ideotypes which stabilize at LAI-value which maximises yield of oil and kernels per hectare. In terms of the carbon budget this means when the gain in light interception does not become outweighed by accompanying losses in carbohydrates for vegetative requirements and maintenance respiration. This optimal LAI should be reached as soon as possible after field planting in order to maximize the proportion of incident solar radiation intercepted by the oil palm canopy.

This, albeit over-simplified model, implies that the trend in crown expansion affects yield during the entire economic life of a planting.

Expansion of mean leaf area per palm with time fits a logistic growth curve (Breure, 1985). Basically a logistic growth function is of the following form:

f(t) = A/(1 + B*e^Ct ), where A, B and C are positive constants and t is the time of growth. C describes the rate of growth and A is the asymptotic maximum of f(t) which is approached when t runs to infinity. At the start (t=0), f(t) = A/(1 + B). The inflexion point of the logistic growth function is at t = -(1/C)*ln(1/B) = (1/C)*lnB when the function has reached a value of A/2.

The parameters of the logistic growth function can be estimated by the Least Squares Method. However this procedure is now much more difficult than in the case of a linear model, because we have here a non-linear regression problem. The Normal Equations can now only be solved iteratively. Statistical computer packages such as SAS, SPSS, SYSTAT, BMDP and GENSTAT have a module for the Least Squares Method for non-linear regression problems. Also the computer package CADEMO (Computer Aided Design of Experiments and Modeling) has a module "Growth Curves" where the parameters can be estimated; furthermore in this module for a new experiment an optimal design is advised to estimate the logistic growth function (or other growth functions).

Breure (1985) describes the mean leaf area (in m²) as a function of the time (in months) after field planting as a logistic growth function but he used in his publication the following function:

f(t)= L_m /[(1 + (L_m - L_i)/L_i)) e^-kt], with

A= L_m, B= (L_m - L_i)/ Li and C= k or

L_m= A, Li= A/(1+B) and k= C,

where k = the relative rate of growth of the mean leaf area,

L_m = asymptotic maximum leaf area,

L_i = leaf area at field planting.

For selection purposes this is conveniently expressed as the time to reach 95% of the maximum leaf area (t_0.95 ), hence f(t_0.95)= 0.95*L_m, as follows:

For an example of a calculation see section 4.5.6.

Thus, in order to minimize the period of sub-optimal LAI, selection should aim at low t_0.95 and thus high k-values. Final leaf area (L_m) is one of the components of LAI at maturity and is thus an important characteristic in breeding for optimal LAI.

Practically, leaf area can first be measured when new leaves have emerged, about 6 months after field planting.

Yield recording stops after 90 months in the field, so this is the final age at which measurements can be made.

The logistic growth function is determined by three parameters (A, B and C); hence, three measurements on different leaves is the absolute minimum to determine the logistic growth function.

More measurements to estimate the parameters of the growth function will clearly improve the precision, but resources are limited. For practical reasons we propose to measure four times for step 2 of parent palm selection which is meant for a first screening. For step 3 of parent palm selection five measurements is recommended.

From the results of 52 progenies planted at Dami Oil Palm Research Station, New Britain, Papua New Guinea, 11 annual leaf area measurements in m² averaged over 6 replications of 4 palm plots (24 palms), were done at month 12, 24, 36, 48, 60, 72, 84, 96, 108, 120 and 132 months from planting (See Table 2 ).

Using the Least Squares Method the parameters of the fitted logistic growth curve per progeny are given in Table 3. The average of these parameters, together with the standard error of the average and the 95%-confidence interval for the parameter, is given in Table 4.

Using the logistic growth curve with these average parameters we can derive a design for the optimal time of measurements from 6 to maximal 90 months after planting for n = 4 (step 2 of parent palm selection) and for n = 5 (step 3 of parent palm selection or for a sample of palms in step 2 to compare groups of families derived from one source of parents).

The design for optimal time of measurements is such that the determinant of the asymptotic covariance matrix of the parameter estimators is a minimum (D-optimality). Unfortunately this optimum design depends on the parameters of the logistic growth function and it gives therefore only a local optimum for the given parameters (not an optimum for all possible parameters). The theory of the D-optimality design for a logistic growth function is described by Rasch (1992) and the Growth Curve analysis module of the computer package CADEMO (Computer Aided Design of Experiments and Modeling) has been used to find the determinant of the asymptotic covariance matrix of the parameter estimators. For the calculation of the determinant of the asymptotic covariance matrix we used the average values of the parameters A, B and C as given in Table 4. The results are presented in Table 5. For the proposed time measurements we will choose the design which is feasible, in terms of the restricted period of recording, and has a determinant equal to or near the least determinant of the asymptotic covariance matrix for the specified number of time measurements.

Results show that it is crucial to include one measurement before 18 months in the field.

For n=4, the set of measurements at 6, 12, 54, and 90 months after planting gives the best design. The design is improved if a second measurement is included in the first 18 months after planting; we found the best design then at 6, 12, 42, 66 and 90 months after planting.

Note, that for n=4 the design was only slightly less precise when measurements were done at 6, 42, 66 and 90 months, that is, those which are all included in the best design for n=5 measurements.

Also that, the last three times of measurements correspond approximately to the end of the 1st, 3rd and 5th year of production which is a convenient timing for calculating growth parameters (see section 4.3).

This set of four measurements (6, 42, 66, and 90 months after planting) is therefore adopted for a step 2 progeny test; an additional measurement at 12 months is recommended for all palms of a step 3 test or for a sample of palms of a step 2 test when the sole objective is to evaluate leaf expansion of different sources of planting material.

2.7.5 Incidence of crown disease

Another way to increase yield through improving crown expansion is to screen genotypes for crown disease, a disorder which appears as bending of the opening spear leaves during the early years after field planting (Breure & Soebagjo, 1991). Crown disease is the most frequently occuring disorder in oil palm. Breure & Soebagjo (1991) showed that losses of oil yield of susceptible material can amount to 4.5% during the first six years of production.

Breure & Soebagjo (1991) showed that the screening of progenies for crown disease is the most effective control; parental GCA values for the incidence of crown disease is therefore an important aspect in selection. They recorded crown disease both by scoring the severety on newly emerged leaves and the percentage of affected palms. The former is quite labour intensive, and for parent selection, recording can be restricted to the percentage of affected palms (rate of incidence).

2.7.6 Tolerance of light competition

Yield recording is usually restricted to the first five years of production; harvesting starts, in favourable environments, about 30 months after field planting. Yield components are fixed as early as two years before harvest (Breure & Menendez, 1990; Breure & Corley, 1992), so that yield for virtually the entire recording period is determined under conditions of low competition for light. The ranking for yield of the progenies may change, however, when components are determined under a closed canopy, that is, when inter palm competition for light has reached its maximum. This drawback in selection efficiency can be partly circumvented by selection for characters associated with tolerance to light competition.

In this way it is more likely that high-yielding progenies, identified during the conventional period of yield recording (low light competition), continue to perform well at mature light competition. Remember that, according to the carbon budget, bunch production is more sensitive to light competition than vegetative growth. There is, however, evidence of genotypic differences in adopting vegetative requirements to available light (Corley & Donough, 1992).

These genotypes are characterized by a high plasticity of vegetative growth, and are thus expected to maintain a high HI under a mature canopy.

One option is direct selection for trends in HI, but this is not practical. One of the components of HI is bunch yield which tends to show pronounced annual fluctuations. Meaningful HI values require therefore two or three year yield records which, in the limited period of recording, does not yield a sufficient number of values to estimate trends with age.

Vegetative dry matter production is the other characteristic determining HI; this component, on the other hand, is less prone to annual fluctuations than bunch yield.

Among the components of vegetative growth, leaf production is particularly sensitive to light competition (Breure, 1982; Corley & Donough, 1992). This component also responds quickly to changes in the amount of light, as shown by Breure (1994). Genotypes which maintain a high HI at maturity are therefore expected to show a strong decrease in leaf production with the increase of LAI with age. Corley (personal communication) indeed found highly significant differences between clones in the slope of the regression of leaf production on LAI, indicating that selection for trend of rate of leaf production with age may be feasible.

2.7.7 Height

Height records are relevant in the selection procedure for two reasons.

Firstly, vertical stem growth is one of the parameters to estimate dry matter incorporated in the trunk which is one of the components of VDM. Selection for slow height increment therefore positively affects HI.

Secondly, as fruit bunches must be cut for harvesting, height at the level of the bunches affects the cost of harvesting. At a certain height, usually about 12 m above ground level, harvesting is not economically feasible anymore, and replanting is necessary.

The stem first forms a wide base without internodal elongation; vertical stem increment thereafter increase, until it reaches a by and large constant value. Conventionally, trunk height is measured from ground level to a reference point on a standard leaf in the crown, usually leaf 25 or leaf 41, counted from the youngest fully opened leaf (leaf 1); height increment is then calculated as the difference between sequential measurements.

This method for determining height increment is not preferred. The area around the palm trunk is usually not flat (holes left from removal of the tree crop, slopes etc.), palms may be leaning, and it also proves difficult to define the reference point on the petiole when the lower leaves are still attached to the stem. Moreover, rate of leaf production differs between palms and also declines with age after reaching a maximum in the second year after planting (Breure, 1987). Vertical increment for a certain interval, determined from measurements between a standard leaf in the crown, does therefore not correspond to the actual increment at the level of the growing point.

Reliability can be improved by measuring height to the insertion of the leaf base of known opening date; this corresponds to the level of the growing point. A further improvement can be made by measuring from a reference point on the stem instead of from ground level.

Note, however, that the bases of leaves produced during about the first 18 months after field planting cannot be used as a reference because these are concealed under the expanding trunk base (Breure & Powell, 1987). Only leaves which open about 30 months after planting remain clearly visible on the stem.

Stem increment to estimate trunk dry matter production are thus best measured on the stem.

Regarding the height at which fruit bunches can still be conveniently harvested, however, it is clearly not feasible to actually measure the height of palms at the time of reaching the critical height for replanting.

So the search is for height measurements during the early life of the palms which closely corresponds to mature height.

From 65 palms representing different progenies from three distinct seed sources of planting material, planted in North Sumatra (Indonesia), the latest fully opened leaf was marked from the start of bunch production, six months later, and then at four successive annual periods. Table 6  presents the height measured from the insertion of the first marked leaf base, that is the leaf which opened at the stage of harvesting (see Fig. 12 for the method of recording). These records per palm fit a logistic growth curve, f(t) = A/(1 + B * e ^Ct), as previously described for leaf area (see section 2.7.4). The characteristic parameters A, B and C, and also the time to reach the inflexion point of the growth function, t_0.50 = (1/C)lnB, vary considerably across palms (see Table 7); in other words, there are quick and slow starters. This indicates that increment values obtained between sequential measurements during a restricted period cannot be used to reliably compare palms for actual height at a certain age.

Indeed, correlations between height increments presented in Table 7 see for the periods 6 to 30 months after the start of harvesting and those of 30 to 54 months was very low and none were significant (correlation coefficients of 0.23 (11 d.f., P-value 0.453), 0.28 (16 d.f., P-value 0.258) and 0.33 (14 d.f., P-value 0.207) respectively for the three palm sources).

The correlations are higher and significant for height increment during 6 to 30 months and actual height at 54 months (correlation coefficients of 0.72 (11 d.f., P-value 0.005), 0.60 (16 d.f., P-value 0.009) and 0.46 (14 d.f., P-value 0.071) respectively for the three palm sources).

Mean height values to the base of leaf 25 (in cm) of 24 palms per plot of three different seed sources from the same experiment (see Table 8 ) show a similar relationship. The correlation coefficient of height increment, measured to the base of leaf 25, from year 1 to 5 after the start of harvesting and the actual height at 11 years was 0.58 (43 d.f., P-value 0.0001).

It is therefore concluded that the actual height to a reference point in the crown at the end of the recording period should be measured to compare progenies for height at maturity.