Crop Science Journal of Natural Resources and Life Sciences Education
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via ISI Web of Science (7)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Lizaso, J. I.
Right arrow Articles by Fonseca, A.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Lizaso, J. I.
Right arrow Articles by Fonseca, A.
Agricola
Right arrow Articles by Lizaso, J. I.
Right arrow Articles by Fonseca, A.
Related Collections
Right arrow Crop Growth and Development
Right arrow Crop Models
Right arrow Seed Production
Crop Science 43:892-903 (2003)
© 2003 Crop Science Society of America

CROP PHYSIOLOGY & METABOLISM

Predicting Potential Kernel Set in Maize from Simple Flowering Characteristics

J. I. Lizasoa, M. E. Westgate*,b, W. D. Batchelora and A. Fonsecab

a Dep. of Agricultural and Biosystems Engineering, Iowa State Univ., Ames, IA 50011
b Dep. of Agronomy, Iowa State Univ., Ames, IA 50011

* Corresponding author (westgate{at}iastate.edu)


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Accurate prediction of kernel number per plant is critical for yield simulation in maize (Zea mays L.). Model predictions of kernel number generally are based on empirical relationships between final kernel number and carbohydrate supply or plant growth at silking, without regard to flowering dynamics. The objective of this study was to develop a model for predicting kernel set from seasonal dynamics of pollen shed and silk exsertion. Hybrids were planted in isolated plots at four ratios of male fertile (MF) and male sterile (MS) plants. Dynamics of pollen shed in each plot was described by a Gauss curve and by a population index (Pind) derived from the percentage of plants at beginning, maximum, and ending pollen shed. Percentage of plants at silking followed a sigmoid curve, whereas the number of silks exposed daily per ear followed a monomolecular model. Pollen shed and silk exsertion curves were translated to potential kernel set using a published relationship. Excellent agreement was observed between predicted and measured number of florets per hectare available for pollination. The model overpredicted kernel number per hectare at high pollen shed densities. Predicted kernels per hectare were within 5% of actual kernel numbers at low pollen shed densities when delayed development of subapical ears and asynchronous pollination within ears were considered. These results indicate kernel set in maize can be predicted from simple measures of the flowering process. They confirm that kernel set is limited primarily by factors other than pollen amount when all plants in the population are MF.

Abbreviations: IPAR, intercepted photosynthetically active radiation • MF, male fertile • MS, male sterile • Pind, population index • RMSE, root mean square of error


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
THE DEVELOPMENTAL PROCESSES that determine kernel number per plant in maize can be divided into three consecutive stages for simulation purposes. In the first stage, the reproductive structures are initiated and differentiated into staminate (male) flowers on the apical inflorescence (tassel) and pistillate (female) flowers on one or more lateral inflorescences (ears). In the second stage, the flowers are functionally mature and engage in pollination; male flowers release pollen grains from the anthers and female flowers exsert receptive stigmas (silks) to intercept the airborne pollen. Separate male and female flowers require that the timing of pollen shed coincide closely with silk exsertion and that tassels produce an overabundance of pollen to ensure pollination of exposed silks. In the third stage, pollination is followed by fertilization and kernel formation (Kiesselbach, 1999). Current models for simulating maize yield do not consider the quantitative and dynamic nature of the first two of these flowering stages.

Most efforts to simulate kernel formation in maize have attempted to associate the final kernel number per plant with the current supply of photosynthate or related characteristics such as light interception or plant growth rate around the time of silking (Edmeades and Daynard, 1979; Tollenaar et al., 1992; Andrade et al., 1993; Kiniry and Knievel, 1995; Otegui, 1997; Andrade et al., 1999; Andrade et al., 2000). Edmeades and Daynard (1979) were the first to relate the number of kernels per ear with a calculated rate of photosynthesis per plant at anthesis. They used a rectangular hyperbola to describe this relationship, which was the basis for predicting kernel number per plant in the original version of CERES-Maize (Jones and Kiniry, 1986).

Tollenaar and colleagues (1992) found that kernel number per plant was related to the rate of dry matter accumulation 1 wk before to 3 wk after silking. They observed also that modern hybrids tended to produce kernels on a second (subapical) ear at low plant population densities. Therefore, they proposed a double-hyperbola function to describe the relationship between kernel number per plant and plant growth rate. Otegui (1995), however, reported that among Argentinean hybrids, only prolific types tended to produce kernels on subapical ears at commercial plant densities.

Because canopy photosynthesis is controlled primarily by solar radiation (Christy et al., 1986), a similar relationship between kernel number and light interception would be expected. Andrade et al. (1993) reported that the number of kernels per plant was curvilinearly related to the IPAR per plant during a 31-d period centered on silking. Kiniry and Knievel (1995) used a linear equation to relate kernels per plant with IPAR during 10-d period following silking, as did Otegui (1997), for IPAR during a 30-d period around silking. Andrade et al. (2000) and Lizaso et al. (2001), however, observed that curvilinear functions predicted kernel numbers from IPAR more accurately than did a simple linear function over a broad range of plant growth rates.

There is evidence that kernel number per ear is not simply a function of assimilate supply. Struik et al. (1986) associated the effects of temperature and photoperiod on kernel number with asynchrony between pollen shed and silk emergence. Because these environmental variables are known to alter floral development and function (Herrero and Johnson, 1981; Hall et al., 1982; Struik et al., 1986; Bassetti and Westgate, 1993c), a mechanistic description of the floral biology of maize may be needed to improve forecasting of kernel number per plant across a wide range of environmental conditions. Few studies provide the quantitative information needed to relate floral development and kernel formation.

Bassetti and Westgate (1994) measured kernel number per ear formed at various intensities of pollen shed. In their study, seasonal pollen shed rates followed a normal distribution that peaked at {approx}500 grains cm-2 d-1 two days after anthesis (50% of population shedding pollen). They also determined that daily rates of pollen shed densities at 100 grains cm-2 d-1 or greater reaching the plane of exposed silks were sufficient to achieve maximum kernel number per ear. Struik and Makonnen (1992) reported a similar pattern of pollen shed, but did not relate it to kernel number.

Bassetti and Westgate (1993a)(b) described the dynamics of silk emergence and senescence on individual ears of maize as a progressive process that varies by genotype and environmental conditions. This silk exsertion dynamic for individual plants provides the essential link between the population dynamics of silking (percentage of plants with silks exposed) and the number of silks exposed for pollination on an area basis, which ultimately determines the potential for kernel set.

These few reports on pollen shed and silk emergence indicate the processes of male and female flowering in maize progress in a fairly predictable manner. As such, it should be possible to develop a mechanistic description of kernel set on a field scale based on a quantitative evaluation of synchrony between pollen production and silk exsertion. Sadras et al. (1985a)(b) recognized this possibility in their work, but predicted kernel set only on a relative basis. They also lacked sufficient quantitative information on the flowering and pollination processes to predict kernel number when pollen amount was limiting. The objective of this study, therefore, was to describe mathematically the processes of floral anthesis (i.e., pollen shed and silk emergence) and pollination. In this initial effort, treatments were designed to alter kernel set by limiting pollen amount. The potential impact of environmental limiting conditions on floral synchrony and pollen viability were not considered, but procedures were developed to accommodate the effects of such constraints.


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
 REFERENCES
 
Procedure for Estimating Kernel Number per Plant
The procedure for estimating kernel number per hectare is based on a quantitative description of maize flowering characteristics on a field scale. Mathematical functions were fit to temporal profiles of plant population dynamics for pollen shed and silk exsertion measured in the field. The characteristics described are: (i) amount and temporal distribution of pollen shed, (ii) amount and temporal distribution of exserted silks; (iii) relationship between kernel set and daily pollen shed intensity reported by Bassetti and Westgate (1994). These components were linked mathematically to estimate the total number of receptive silks pollinated each day. This value represents the potential number of kernels that can be formed by the plant population.

Amount and Temporal Distribution of Pollen
We used the field data of Westgate et al. (2003) for pollen shed in different MF treatments to generate a seasonal distribution of pollen production per plant. Daily values for pollen shed (grains cm-2) were normalized to daily pollen shed per fertile plant by correcting for the fraction of MF plants and the population density in each treatment. Gauss curves were fit to these normalized data to generate a genotype-specific pattern of seasonal pollen shed given by:

[1]
where PR is the rate of pollen shed on a fertile plant basis (grains per plant per day), p is the total amount of pollen produced per plant (grains per plant), W is the width of the pollen shed curve measured at half the maximum pollen shed rate (days), and t and tx are the current day and the day of maximum pollen shed. Rates of pollen shed per plant predicted from the Gauss curves were scaled to rates of pollen shed on an area basis (grains cm-2 d-1) using the population densities and fraction of fertile plants in each MF treatment.

The seasonal distribution of pollen shed also was predicted from field observations of male flowering characteristics at the population level. The three stages of tassel development identified in Westgate et al. (2003): beginning of pollen shed (start shed—first anthers visible on main tassel branch), maximum intensity of pollen shed (max shed—all branches of tassel are involved in pollen shedding), and ending of pollen shed (end shed—no new anthers visible) were expressed as a percentage of the plant population on a daily basis. The progress of each stage was described using a sigmoid logistic function:

[2]
where Pop and Popx are the percentage of the plant population at each stage of pollen shed and the corresponding maximum percentage; t and tm are the current day and the day when 50% of the plant population reaches each stage of pollen shed; k is a curve parameter governing the slope of the function.

Using the predicted percentage of plant population at each stage of pollen shed (Pop), a Pind was calculated to simulate the daily rate of pollen produced in the field:

[3]
where Pind is the population index (%), and start shed, max shed, and end shed are the predicted percentages of the plant population that have begun to shed pollen, reached maximum shed intensity, or completed pollen shed. Calculated Pind values were scaled to daily rates of pollen shed on an area basis (grains cm-2 d-1) using the population density and the total pollen production per plant. Details of these procedures and calculations are provided by Westgate et al. (2003).

Amount and Temporal Distribution of Receptive Silks
The distribution of silks exposed for pollination in the field results from developmental processes occurring at the plant level and at the population level. At the plant level, silks emerge from the surrounding ear leaf sheaths (husks) during a period of days following a curvilinear pattern (Bassetti and Westgate, 1993a, 1994). We used a monomolecular model to describe the pattern of silk exsertion on each ear:

[4]
where SN is the cumulative number of exposed silks on one ear, SNx is the potential number of silks per ear equal to the potential number of florets, t0 is the time when the first silk is exserted (day of year), and b is a shape parameter controlling the slope of the curve.

At the population level, the percentage of plants with silks exposed progresses in a sigmoid fashion. We used a sigmoid logistic function, similar to Eq. [2], to characterize this developmental pattern. Daily percentages for each treatment were converted to number of plants silking by multiplying by the plant population density (plants ha-1).

The number of newly exserted silks per unit land area was calculated each day by multiplying the pattern of silk exsertion by individual plants times the pattern of silk emergence for the population. The daily cohort of plants that began to exsert silks was calculated from the population dynamics curves for each treatment. Measurements of silk emergence patterns for the hybrids used in this study indicated silks continued to emerge for up to 9 d on unpollinated ears. Therefore, it was assumed each plant assigned to a daily cohort would exsert silks for the next 9 d (Bassetti and Westgate, 1993a) according to Eq. [4]. The same calculation was made for each subsequent day until 100% of the plants in each treatment were silking. The number of new silks exserted on an area basis was calculated each day as the sum of silks exserted by each of these nine cohorts.

An alternative mathematical approach used to describe the daily emergence of silks was a double Gauss function fit to the seasonal pattern of newly exposed silks in the field:

[5]
where Stot is the daily number of newly exposed silks per land area, P1 and P2 are the area of the Gauss Curves 1 and 2 whose addition represents the total number of exposed silks (or viable flowers) during the season per unit area, W1 and W2 are shape parameters controlling the width of the curves (days), and tx1 and tx2 are the times when the peaks of Curves 1 and 2 are reached (days).

Relationship between Kernel Set and Daily Pollen Shed Intensity
We applied two linear functions in series to the published data of Bassetti and Westgate (1994) to calculate a percentage of kernel formed by flowers with exposed silks from the daily rate of pollen shed. The limits of pollen shed rate for each equation were:

[6]
where ks is the percentage of florets with exposed silks that formed kernels (%), and pr is the daily rate of pollen shed (grains cm-2 d-1). In this formulation, 100 grains cm-2 d-1 is the critical rate of pollen shed below which ks is less than the maximum 96%. Because Bassetti and Westgate (1994) used final kernel set in their analysis, their equations relating pollen shed density to percentage of kernel formed incorporate the natural level of pollen viability inherent in their experiment. Nonetheless, our approach provides the mathematical basis for testing the impact of pollen viability on kernel number per hectare by adjusting the critical rate of pollen shed required for maximum ks.

Daily kernel formation as measured by Bassetti and Westgate (1994) never reached 100% of exposed silks even at the highest rates of pollen shed (Eq. [6]). Kernel abortion was nil in their study since they measured kernel set only at highly receptive floral positions in the middle of the ear. Therefore, we assumed the remaining florets with exposed silks remained receptive. These silks were added to the next day's pool available for pollination. These unpollinated silks remained receptive to pollen for five additional days. Silks that were not pollinated by the sixth day were assumed to senesce and lose receptivity (Bassetti and Westgate, 1993b) and no longer contributed to kernel set.

Evaluation of the Procedure for Estimating Kernel Set
We compared the number of florets per hectare and kernels per hectare calculated by the model with measured floret and kernel numbers in eight isolated field plots that varied widely in pollen shed density. In 1986, MF levels of 100, 50, 20, and 0% were obtained by detasselling a 100% MF hybrid, Pioneer 3978 (hereafter P3978). In 1987, MF levels of 75, 50, 20, and 0% were obtained by mixing MS and MF isolines of Pioneer 3925 (hereafter P3925). P3978 was planted at 7.2 plants m-2, and produced an average of 658 silks per ear. P3925 and the pollen sterile isoline P3925S were planted at 5.7 plants m-2, and produced an average of 621 silks per apical ear. Details of the experimental design and plant culture conditions are provided in the companion paper by Westgate et al. (2003).

In the model, each MF treatment was corrected for plant population density and average number of ears per plant at harvest. Percentage of plants at each stage of pollen shed or reaching silking was recorded every other day by evaluating at least 80 plants in the four center rows of each plot.

Procedures used to document the daily rate of pollen shed and daily progress of silk exsertion are detailed in Bassetti and Westgate (1994) and Westgate et al. (2003). Briefly, passive pollen traps were placed within the canopy at ear level and collected every 24 h at {approx}1700 h for counting. Pollen density on the trap surface was evaluated in the laboratory by image analysis according to Bassetti and Westgate (1994). Every 2 d beginning at first silk appearance, 1 to 2 cm of unpollinated silk tissue was collected from ears of 10 plants. Care was taken to collect all exposed silks and to prevent pollination at all other times by covering the ears with a glassine bag. Silk pieces were fixed in ethyl alcohol-glacial acetic acid-formalin-water (50:5:10:35 by volume) containing 10 g kg-1 analine blue, and subsequently counted by hand.

Effect of Asynchronous Pollination
Within-Ear Synchrony
Several studies indicate that kernel set at flower positions with late emerging silks increases dramatically if all exposed silks on the ear are pollinated synchronously (Freier et al., 1984; Cárcova et al., 2000). Evidently, both the timing of pollination and the number of prior fertilization events affects kernel set by later pollinated flowers. Therefore, the effect of asynchronous pollination of florets on the apical ear was included following the hypothesis that earlier-formed kernels could have a detrimental effect on the success of kernel formation by later-fertilized flowers. The daily calculation of kernel set (Eq. [6]) was modified to account for the cumulative number of kernels set per ear (KS) and asynchronous pollination efficiency (EAP):

[7]
where EAP is a unitless function (0 to 1) that varies with the cumulative number of kernels per ear formed the previous day (KS). Parameter c controls the slope, and KS0 is the intercept for kernel set. If EAP = 1.0, all pollinated flowers will set kernels. If EAP = 0.0, no pollinated flowers will set kernels, regardless of the pollen shed density. Parameters in Eq. [7] (c = 0.013; KS0 = 1.2 x SNx, for apical ear; KS0 = 1.1 x SNx, for subapical ear) were calibrated to data in Cárcova et al. (2000). Asynchrony-adjusted kernels set was calculated daily using Eq. [8]:

[8]
where KS is the adjusted kernel set (kernels ear-1), SN is the cumulative number of silks exposed, ks is the fraction of kernel set estimated with Eq. [6] (%), and EAP is the efficiency factor for kernel set calculated in Eq. [7].

Between-Ear Synchrony
We included the effects of asynchrony between apical and subapical ears of P3925 in terms of delayed silk exsertion and a decreased number of flowers per ear on subapical ears. P3978 was not included in this analysis because it did not produce second ears. Silk exsertion on subapical ears of P3925 was estimated from the pattern of silk exsertion measured on apical ears and the measured relationships between apical and subapical ear development in a similar prolific hybrid (Table 1). The experiment providing data on flowering of subapical ears was planted at the Bruner Research Farm near Ames, IA, in 2000. Silk exsertion was evaluated on the prolific hybrid Asgrow 740 grown at three population densities (1, 8, and 16 plants m-2) supplied with 168 kg N ha-1 and replicated three times in a split-plot design. When 30% of the plants in each plot reached silking, five plants about to exsert silks were selected for sampling. Ears were covered with glassine bags to prevent pollination, and silks were sampled 2, 4, and 8 d after first silks appeared. Two-centimeter segments of silk tissue were cut at husk level, transferred to plastic bags containing 500 g kg-1 ethanol, and stored at 4°C until counted manually.


View this table:
[in this window]
[in a new window]
  		Table 1. Parameters used in Eq. [4] to predict daily silk exsertion for subapical ears of Pioneer 3925 (P3925). Silk exsertion curves were fit to measured field data for the apical ear of P3925 and for apical and subapical ears of Asgrow 740. Parameters for subapical ears of P3925 were calculated relative to the apical ear using the ratio of apical/subapical values for the corresponding Asgrow 740 parameters. Subapical ear parameters for P3925 were adjusted to 5.7 plants m-2 by linear interpolation.

 
Monomolecular functions (Eq. [4]) were fit to measured silk exsertion data for the apical ear of P3925 and for apical and subapical ears of Asgrow 740 (Fig. 1). Parameters SN (maximum silk number), b (shape factor), to (initial day of silk emergence), and duration of silk emergence were generated from each curve. Parameters for subapical ears of P3925 were calculated relative to the apical ear using the ratio of apical/subapical values for the corresponding Asgrow 740 parameters. Subapical ear parameters for P3925 were adjusted to 5.7 plants m-2 by linear interpolation. The delay in initial silk emergence for the subapical ear was calculated as to2 - to1.



View larger version (24K):
[in this window]
[in a new window]
  		Fig. 1. Dynamics of silk exsertion in apical and subapical ears of Asgrow 740 as influenced by plant population density. Symbols are the number of exposed silks per ear and lines are corresponding monomolecular functions fit to the data. Parameters SN, b, to and duration in Eq. [4] to predict silk emergence from subapical ears of P3925 were generated from these curves (see Table 1). Vertical lines are standard errors for five plants. Plants grown at 16 plants m-2 did not produce a second ear.

 
Prediction Evaluation
Deviation of predicted and observed values were compared using the root mean square of error (RMSE):

[9]
where Pi and Oi are predicted and observed values, and n is the number of observations.


   RESULTS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
DISCUSSION
 REFERENCES
 
Dynamics of Pollen Shed
The first step in the procedure to calculate the number of kernels formed on an area basis was to characterize the dynamics of pollen production in the field. The evaluation of daily pollen production showed greater variability of daily rates for P3978 than for P3925 (Fig. 2a and 2b). Also, there was a greater amount of pollen drift into the 0% MF treatment for P3978. These results likely reflected differences in methodology used to reduce the levels of pollen production (detasselling in P3978 vs. MS isoline in P3925). The difference in plant population density (7.2 plants m-2 for P3978 vs. 5.7 plants m-2 for P3925) contributed to the lesser maximum rates of pollen shed for P3925 as well. To compare the response of two hybrids to pollen shed density, pollen shed rates were normalized using plant population density and the fraction of MF plants. Normalized in this way, it was evident that the MF treatments did not affect the pattern or the timing of the pollen shed process (Fig. 2c,d). Pollen production per plant based on the seasonal pollen shed curves fit to the field data were very similar for the two hybrids: 4.2 million grains per plant for P3978 and 4.5 million grains per plant for P3925. The seasonal pattern of pollen shed, however, was 4 to 5 d longer for P3978. The normalized curves for pollen production per plant were scaled to pollen shed rates on an area basis (grains cm-2 d-1) using the population densities and relative seasonal pollen production for each hybrid-treatment combination (Fig. 2a,b). These scaled curves of pollen production accurately described the seasonal pattern of pollen shed for each treatment and were used to predict kernel number per hectare.



View larger version (38K):
[in this window]
[in a new window]
  		Fig. 2. Seasonal dynamics of pollen shed intensity for hybrids Pioneer 3978 (P3978; a,c) and Pioneer 3925 (P3925; b,d). (a,b) Observed (symbols) and predicted (lines) rates of pollen shed at four levels of male fertility (MF). (c,d) Daily pollen shed rates normalized per fertile plant (symbols) and Gauss functions fit to the normalized data (lines). Vertical lines are standard errors for four measurements. Note that a single curve describes the seasonal pattern of pollen shed for each hybrid.

 
The pollen trap method accurately characterizes the dynamics of pollen shed in the field (Bassetti and Westgate, 1994; Fonseca et al., 2002). But, it is a labor and time-intensive process. Therefore, an alternative procedure to describe the seasonal pattern of pollen shed in the field was needed. Westgate et al. (2003) showed that the daily rate of pollen shed could be derived from simple observations of the male flowering process. Using their procedure, we fit sigmoid curves (Eq. [2]) to field observations of percentage of plants beginning pollen shed (anthesis), reaching maximum shed, and ending shed (Fig. 3a,b). Simulated percentages of population at beginning, maximum, and ending pollen shed were used to calculate a Pind of plants shedding pollen (Eq. [3]). To simulate the intensity of pollen shed, the Pind was scaled using the total pollen production per plant (parameter p in Eq. [1]) and the plant population density (Westgate et al., 2003). Figure 3c and 3d show that the rates of pollen shed computed from the Pind were in close agreement with rates measured in the field. Values of RMSE were similar to corresponding values obtained when rates of pollen shed were predicted with the scaled Gauss curves (Fig. 2a,b). These results indicate that the Pind procedure provides a promising alternative to the pollen trap method to characterize the dynamics of pollen shed in the field. If the average pollen production per plant is known, pollen production on an area basis can be distributed realistically in time using simple observations of male flower development.



View larger version (36K):
[in this window]
[in a new window]
  		Fig. 3. (a,b) Normalized percentage of plant population at three stages of pollen shed: beginning shed, maximum shed, and end shed for Pioneer 3978 (P3978; a) and Pioneer 3925 (P3925; b). (c,d) Measured pollen rates (symbols) and predicted pollen rates using the population index (lines) for hybrids P3978 (c) and P3925 (d). The population index was calculated using Eq. [3]. Vertical lines are standard errors of four measurements.

 
Dynamics of Silk Exsertion
The second component required to calculate kernel number per hectare is a measure of the dynamics of silk numbers available and receptive to pollination. This dynamic requires a simultaneous calculation of silks exserted per plant and an estimate of the percentage of plants beginning to exsert silks. We calculated silk numbers per hectare first on a cumulative basis as depicted in Fig. 4a and 4b. The process is described by Eq. [4], which accounts for variations in the maximum number of florets per ear (SNx) and the duration of silk exsertion on an ear (days). The hybrids examined in this study required 8 and 9 d to complete silk exsertion, which is in agreement with Bassetti and Westgate (1993a), who reported differences up to 4 d between hybrids to complete the silking process.



View larger version (27K):
[in this window]
[in a new window]
  		Fig. 4. Seasonal progress of silk emergence for the plant population and individual apical ears. (a,c) Pioneer 3978 (P3978); (b,d) Pioneer 3925 (P3925). (a,b) Cumulative percentage of plant population at silking (solid lines) and cumulative number of silks exserted on the apical ear (broken lines). (c,d) Daily percentage of plant population that has started silking (solid lines) and daily number of silks exserted from the apical ear (broken lines).

 
To calculate the number of newly emerged silks per hectare, we computed the daily components for each plant and for the population (Fig. 4c and 4d). Each day, a new group of plants begins to exsert silk simultaneously with a second group in its second day of silk exsertion, along with those in their third day, and so on. After calculating the number of silks exserted by each group based on their respective day of silking, the values are added to obtain a daily total of newly exposed silks (Fig. 5). A double Gauss model also described the dynamics of silk exsertion for each hybrid. The model explained >99% (r2 = 0.999) of the seasonal variation in silk exsertion per hectare. The maximum rate of silk appearance occurred 2 to 3 d after silking for the population (50% of population with silks exposed), which corresponds closely with the maximum rate of pollen shed for both hybrids. On a population basis, the two hybrids exposed silks during a period of 31 to 34 d (Fig. 5). By the day of silking, the populations had exserted 20 to 25% of their seasonal totals.



View larger version (20K):
[in this window]
[in a new window]
  		Fig. 5. Seasonal dynamics of silk exsertion on an area basis for hybrids Pioneer 3978 (P3978; top) and Pioneer 3925 (P3925; bottom). Calculated values (symbols) were generated from population and silk exsertion curves in Fig. 4a and b. Predicted curves (lines) are a double Gauss function fit to the seasonal pattern of newly exposed silks using Eq. [5].

 
Pollen and Silks Synchrony: Potential Kernel Set
Each exposed silk translated mathematically into a potential kernel, depending on its timely pollination. So the third component in our predictive procedure was to link the dynamics of silk exsertion with the dynamics and intensity of pollen production to generate a cumulative curve of kernel set for each treatment. We relied on two facts about the flowering biology of maize to estimate potential kernel set: (i) kernel set varies with pollen shed density in a manner defined by Bassetti and Westgate (1994); and (ii) each exposed silk remains receptive to pollen for at least 6 d (Bassetti and Westgate, 1993a).

Each day, a new group of silks was exposed for pollination, a certain rate of pollen shed was available to pollinate them, and the daily pollen rate was associated with an expected percentage of fertilized florets (Bassetti and Westgate, 1994). To calculate the number of kernels set each day, we multiplied the number of silks exposed that day times the percentage of florets that should be fertilized given the daily rate of pollen shed (Fig. 6). Figure 7 shows the daily progress of kernel addition for two levels of pollen availability provided to P3978 and P3925. The model predicted that hybrid P3978 would pollinate 96% of the exposed silks when 100% of the plants contributed to pollen shed (Fig. 7a). All exposed silks were successfully pollinated until the daily rate of pollen shed decreased to {approx}100 grains cm-2. Reducing the number of plants shedding pollen to 20% MF decreased the seasonal production of pollen to 30% of the 100% MF rate. Yet, sufficient pollen was available to fertilize {approx}90% of the flowers with exposed silks (Fig. 7c).



View larger version (21K):
[in this window]
[in a new window]
  		Fig. 6. Relationship used to convert daily pollen shed density into potential kernel set adopted from Bassetti and Westgate (1994). Data for pollen shed are from passive pollen traps placed at ear level, and kernel set was measured on ears exposed to pollen for 1 d at the indicated densities. The broken line indicates the predicted kernel set if pollen viability were decreased to 63% of that in the Bassetti and Westgate (1994) study.

 


View larger version (25K):
[in this window]
[in a new window]
  		Fig. 7. Seasonal dynamics of pollen shed and potential kernel set for Pioneer 3978 (P3978; a,c) and Pioneer 3925 (P3925; b, d) at two levels of male fertility (MF). Note that all florets pollinated before maximum pollen shed develop into kernels. At low pollen density, a greater proportion of late emerging silks fail to be pollinated.

 
Hybrid P3925 apparently was not as efficient as P3978 in pollinating receptive silks. With 75% of the plants shedding pollen, only 86% of the florets with exposed silks set a kernel (Fig. 7b). When the fraction of MF plants decreased to 20% MF, seasonal pollen production was reduced to 24% of the maximum, and 71% of flowers with exposed silks produced kernels (Fig. 7d). This supports the notion that pollen production is generally well in excess of pollination requirements for >90% of exposed silks. Maize canopies might fail to pollinate late emerging silks, however, even when all plants are MF.

Calculated values for silking florets per hectare and kernels per hectare were compared with field measured floret and kernel numbers. Our procedure predicted floret numbers with >99.5% accuracy in all treatments, when the specific plant population and average number of ears per plant were considered (Table 2). Predicted kernel numbers were overestimated, however, by up to 15% for P3978 and up to 20% for P3925 even in the 20% MF treatment, which was shown to limit kernel set (Westgate et al., 2003). This consistent overestimation of kernel number suggested that a systematic factor affecting kernel set was not considered in the model.


View this table:
[in this window]
[in a new window]
  		Table 2. Observed and predicted number of florets and kernels per unit land area produced at various levels of male fertility (MF) for Pioneer 3978 and 3925 (P3978 and P3925). Data in parenthesis indicate the floret or kernel value as a percentage of the observed number of florets in each MF treatment. Difference (Diff) = Predicted % - Observed %. Predicted kernel numbers are not corrected for loss due to pollen viability, within-ear pollination asynchrony, or delay in silk exsertion on subapical ears.

 
Pollen Viability
The relationship of Bassetti and Westgate (1994) adopted to predict kernel set (Eq. [6] and Fig. 6) included an unknown level of pollen viability. By using 100 grains cm-2 d-1 as the critical pollen density to assure 96% kernel set in Eq. [6], we tacitly assumed that the level of pollen viability in Bassetti and Westgate (1994) applied to P3798 and P3925. A lower level of pollen viability in our experiment might explain the overprediction of kernel numbers in Table 2. To test this possibility, pollen viability was decreased mathematically by increasing the critical pollen rate in Eq. [6] to values >100 grains cm-2 d-1. For this evaluation, we selected MF treatments in which pollen production was similar for both hybrids and was low enough to limit kernel set: the 20% MF treatment for P3978 and the 50% MF treatment for P3925 (Westgate et al., 2003). Table 3 shows that decreasing pollen viability lowered kernel set in these treatments, but the impact was too small to account for the overprediction observed, even when pollen viability was decreased by 37%, which is typical of high temperature stress (Herrero and Johnson, 1980) or field aging (Luna et al., 2001).


View this table:
[in this window]
[in a new window]
  		Table 3. Predicted kernel number at decreasing levels of pollen viability for the 20% male fertile (MF) treatment for Pioneer 3978 (P3978) and the 50% MF treatment for Pioneer 3925 (P3925). Kernel set was limited by pollen amount in both treatments. The critical pollen rate needed to achieve 96% kernel set in Eq. [6] was increased mathematically to simulate loss of pollen viability below the original level assumed from Bassetti and Westgate (1994). Predicted kernel numbers are presented as a fraction of the observed values. Note that a 37% decrease in relative pollen viability resulted only in a 3 to 5% decrease in kernel set.

 
Pollination Synchrony within Ear and between Ears
The initial calculation of potential kernel set assumed that the opportunity of a receptive silk to be pollinated and develop into a kernel depended only on the rate of pollen shed. It also assumed that silk exsertion on apical and subapical ears were identical in terms of number and timing. Yet Cárcova et al. (2000) showed that naturally pollinated plants often set fewer kernels than did plants whose silks were pollinated synchronous by hand, particularly at low population densities. Also, silk emergence on subapical ears often is delayed relative to apical ears (Jacobs and Pearson, 1991; Otegui and Melon, 1997), with greater delays evident at higher plant population densities (Jacobs and Pearson, 1991; Cárcova et al., 2000).

To take these developmental effects into account, we modified the procedure for calculating kernel number to include the delay in silk exsertion by subapical ears and the effects of asynchronous pollination within and between ears on the same plant (Fig. 1, Table 1). Although these refinements are based on existing evidence in the literature, we did not measure these developmental effects directly in our data set. Therefore, they are considered as hypothetical effects in our analysis that will require experimental confirmation.

Figure 8a shows that our procedure correctly forecast the lower potential of P3925 florets to form kernels (83% maximum set) compared with those of P3978 (65% maximum set) at high pollen shed density. Including the effects of asynchronous pollination and the delayed development of subapical ears decreased kernel set (relative to the number of exposed silks) substantially in both hybrids (Fig. 8b). Predicted kernel set closely matched measured kernel set at all pollen shed densities for P3978, but the procedure only accounted for about half of the overestimate in P3925.



View larger version (21K):
[in this window]
[in a new window]
  		Fig. 8. Kernel set expressed relative to the number of exposed silks at four levels of male fertility (MF) for Pioneer 3978 (P3978) and Pioneer 3925 (P3925). Pollen shed is presented as a fraction of the total seasonal deposition expected in a field of 100% MF plants. Symbols are measured values; lines indicate predicted values. (a) Predicted kernel set is based on silk exsertion and pollen shed dynamics for each MF treatment. (b) Predicted kernel set incorporates potential kernel loss due to asynchronous pollination within ears and asynchronous development between ears.

 

   DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
REFERENCES
 
The purpose of this work was to develop a model to simulate kernel set from the mathematical description of male and female flowering dynamics in a maize field. We attempted to generate general algorithms useful to improve existing models of kernel number (Lizaso et al., 2001). The functions were developed for nonstress conditions, but can be adapted readily to accommodate a range of conditions known to alter flower development and function. Rates of pollen shed density measured at the level of exposed silks were accurately described with Eq. [1] (Fig. 2a,b). Maximum deviation between predicted and measured pollen rates (as expressed by the RMSE) was 92 grains cm-2 d-1 for P3978 at 100% MF. Similar rates of pollen shed were reported earlier (Hall et al., 1982; Struik and Makonnen, 1992; Bassetti and Westgate, 1994). We also calculated daily appearance of silks using Eq. [4] (Fig. 4) following patterns in previous reports (Bassetti and Westgate, 1994; Cárcova et al., 2000). These patterns and timing of pollen shed and silk appearance also can be modified readily to predict kernel set under stress conditions. High temperature from tassel initiation to kernel set, for example, shortens the duration of pollen shedding (Struik et al., 1986). Drought at or before tasseling delays silk emergence (Herrero and Johnson, 1981; NeSmith and Ritchie, 1992). Flooding during the early vegetative growth delays silking more than tasseling (Lizaso and Ritchie, 1997). These environmental responses can be simulated by altering parameters in Eq. [1] and [4].

Our calculations show that the seasonal duration of silk exsertion for the eight populations examined persisted for a period of 31 to 34 d (Fig. 5). The maximum duration of pollen shed was 24 d for P3978 and 17 d for P3925. Rates of pollen shed at or above that required for maximum kernel set (Fig. 6), however, occurred only on eight of these days for P3978 and on five days for P3925. Because silks remain receptive to pollen for 6 d after they first emerge (Bassetti and Westgate, 1994), most of the unpollinated silks exserted early in pollen shedding were pollinated eventually (Fig. 7). The beneficial impact of this prolonged silk receptivity was most evident in the 20% MF treatments in which a large reduction in pollen deposition early in silking resulted in only a 6% (P3978) and 15% (P3925) reduction in kernel set. Silks exserted after the rate of pollen shed was less than required for maximum kernel set constituted the vast majority of unpollinated silks. In this regard, the longer duration of pollen shed for P3978 (Fig. 2c) contributed to its advantage over P3925 in terms of setting a larger proportion of kernels per exposed silk (Fig. 8).

In the treatments that were saturated for seasonal pollen deposition (>=50% MF), 83% and 65% of the pollinated florets developed into kernels for P3978 and P3925, respectively (Table 2). These percentages are in general agreement with reported values (Otegui and Melon, 1997). Yet, our calculations overpredicted the percentage of florets that set kernels both under pollen-abundant and pollen-limited conditions. When pollen is abundant, some degree of overprediction is expected since kernel set is limited by other factors, such as available assimilate supply (Andrade et al., 1999; Andrade et al., 2000; Lizaso et al., 2001). Under pollen-limited conditions ({cong}30% of maximum pollen shed for each hybrid), however, overprediction of kernels per hectare indicates that the model did not consider one or more important factors constraining kernel formation.

We tested the possibility that the level of pollen viability was lower than originally assumed in the Bassetti and Westgate (1994) relationship between pollen shed density and percentage kernel set. Decreasing the efficiency of kernel set at low pollen shed density by 37% (effectively decreasing pollen viability) decreased predicted kernels per hectare by <5% in MF treatments that were already marginal for pollen amount (Table 3). Therefore, it seems unlikely that this factor alone could account for the overestimation of kernel set. This exercise, however, demonstrates the utility of our modeling approach to quantify developmental and environmental effects on kernel set.

A second possibility was that the original calculation of kernel number did not consider the potential for kernel loss due to asynchronous pollination within and between ears. We tested two related components of asynchronous pollination for their impact on kernel set. First was the developmental dominance of basal kernels over apical kernels on the same ear (i.e., within-ear asynchrony). The second was the delayed development of subapical ears relative to apical ears (i.e., between-ear asynchrony).

Cárcova et al. (2000) showed that synchronously hand-pollinated ears set more kernels than did naturally pollinated ears at the same level of assimilate supply. On the basis of their results, we developed an asynchronous pollination efficiency factor (EAP) to account for the dominance of early-formed kernels on predicted kernel number. Originally, the opportunity for a silk to be pollinated and for its floret to develop into a kernel depended only on the daily rate of pollen shed as dictated by Eq. [6]. In the modified procedure, the fraction of kernels set each day decreased as the number of previously set kernels increased (Eq. [7]). This approach is consistent with Freier et al. (1984), who observed that the success of later-formed kernels decreased as the number of prior pollinations on the same ear increased. After taking within-ear synchrony into account, the model predicted kernel numbers per hectare accurately for P3978 at all pollen shed densities (Fig. 8b). The modified procedure, however, still overestimated kernels per hectare for P3925 by {approx}10%.

A number of studies have reported a delay in subapical ear development relative to the apical ear as plant population density increases (e.g., Otegui, 1997). In this study, subapical ears produced kernels on a number of P3925 plants (1.2 ears per plant on average), but none formed on P3978 plants. Because we did not measure silk exsertion on these subapical ears, it was necessary to assume that their silk numbers and the timing of exsertion were identical to that of the apical ears. Measurements of silk exsertion on subapical ears on a similar prolific hybrid (Asgrow 740) grown at several plant densities (Fig. 1) indicated this assumption likely caused us to overestimate the contribution of the subapical ears to kernel set in P3925. When silk exsertion from the subapical ears of P3925 followed the same relative pattern as for Asgrow 740, predicted kernel numbers for P3925 decreased to within 5% of measured values for the three lowest MF treatments (Fig. 8b).

It is important to emphasize the uncertainty associated with incorporating the potential impact of pollination asynchrony and delayed development of subapical ears. These functions likely are affected by population density and by the inherent prolificacy of each hybrid (Otegui, 1997; Cárcova et al., 2000). Yet our results provide evidence that asynchrony between ears and within ears must be taken into account for accurate prediction of kernel numbers. This is particularly important when pollen amount limits kernel number.

Many authors have shown that kernel number per plant depends on IPAR around silking (Andrade et al., 2000; Lizaso et al., 2001). Such results support the view that the supply of carbohydrates around silking imposes the primary limitation on the number of fertilized ovules that ultimately develops into kernels. There are numerous field conditions such as drought (Hall et al., 1982; Edmeades et al., 1993), flooding (Lizaso and Ritchie, 1997), or hybrid seed production in which asynchrony of male and female flowering could impose a limit on pollen availability to fertilize exposed silks (Bassetti and Westgate, 1994). Kernel set under such conditions would be constrained by the number of fertilized ovaries, even though concurrent metabolic responses to stress also contribute to decreased kernel set. The procedure presented herein predicts kernel set based primarily on the synchrony of staminate and pistillate flowering, delayed development of subapical ears, and the detrimental effect of prior fertilization on the success of later-fertilized ovaries. A general procedure to forecast kernel numbers under both assimilate-limited (Lizaso et al., 2001) and pollen-limited conditions is forthcoming.

Received for publication May 15, 2002.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 




This article has been cited by other articles:


Home page
 Crop Sci.Home page
J. I. Lizaso, A. E. Fonseca, and M. E. Westgate
Simulating Source-Limited and Sink-Limited Kernel Set with CERES-Maize
Crop Sci., September 1, 2007; 47(5): 2078 - 2088.
[Abstract] [Full Text] [PDF]

Home page
 Crop Sci.Home page
D. S. Ireland, D. O. Wilson Jr., M. E. Westgate, J. S. Burris, and M. J. Lauer
Managing Reproductive Isolation in Hybrid Seed Corn Production
Crop Sci., May 18, 2006; 46(4): 1445 - 1455.
[Abstract] [Full Text] [PDF]

Home page
 Crop Sci.Home page
A. E. Fonseca, J. I. Lizaso, M. E. Westgate, L. Grass, and D. L. Dornbos Jr.
Simulating Potential Kernel Production in Maize Hybrid Seed Fields
Crop Sci., September 1, 2004; 44(5): 1696 - 1709.
[Abstract] [Full Text] [PDF]

Home page
 Crop Sci.Home page
M. E. Westgate, J. Lizaso, and W. Batchelor
Quantitative Relationships between Pollen Shed Density and Grain Yield in Maize
Crop Sci., May 1, 2003; 43(3): 934 - 942.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via ISI Web of Science (7)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Lizaso, J. I.
Right arrow Articles by Fonseca, A.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Lizaso, J. I.
Right arrow Articles by Fonseca, A.