| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
a Danish Centre for Forest, Landscape, and Planning, Hørsholm Kongevej 11, DK-2970 Hørsholm, Denmark
b Dep. of Mathematics and Physics, The Royal Veterinary and Agricultural Univ., Thorvaldsensvej 40, DK-1871 Frederiksberg C, Denmark
* Corresponding author (sugl{at}kvl.dk).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Abbreviations: FGP, final germination percentage • MGT, mean germination time • T25-75, time from 25 to 75% germination
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
When comparing germination characteristics of cultivars and seed lots, it is desirable to assess both the FGP and the distribution of the timing of germination. In standard germination tests, germination is usually evaluated by counting the number of normal plants at an interim first count and at a final count at the end of the test (International Seed Testing Association, 1999), but in practical testing of grass seeds, the germination percentage usually differs very little between first and final count (e.g., Nienhuis and Baltjes, 1985). The first count, hence, provides very limited information about differences in the timing and uniformity of germination, and more frequent counting of germination during the germination test is more likely to detect such differences.
Fitted curves can efficiently summarize information from germination time courses, provided they fit the observed data sufficiently closely (Brown and Mayer, 1988b). As an alternative to nonlinear regression, Hunter et al. (1984) analyzed germination frequency using a multinomial distribution, assuming that the seeds germinated independently of each other and that the numbers of seed that germinated in each time interval followed a multinomial distribution. The model, however, required transformation of the time scale before fitting a normal distribution to the germination times, and the applied transformation was of major importance for the fit of the model. Since time to germination is not always normally distributed, often causing positively skewed germination time courses (Nichols and Heydecker, 1968; Campbell and Sorensen, 1979; Cheng and Gordon, 2000), it is relevant to take this into account when analyzing germination data. A generalized hyperbolic distribution is very flexible and has been successfully used to describe asymmetric and heavy-tailed behavior of, for example, turbulence and financial data (Bibby and Sørensen, 2003). A function based on this distribution can also describe the variation in germination times even when the distribution is very skewed, and by fitting this function to germination data, it is possible to summarize germination time courses into characteristics of biological relevance. Such parameters may comprise FGP as an expression of the proportion of seeds with germination capacity, MGT as an inverse expression of overall germination speed or rate for the whole population, and T25-75 as an expression of spread of germination times or germination uniformity (Bewley and Black, 1994).
The aim of this study was (i) to estimate biologically relevant germination characteristics from germination time courses with a curve fitting procedure based on an asymmetric multinomial distribution, and (ii) to study the variation in germination percentage, speed, and uniformity within and among cultivars of red fescue, perennial ryegrass, and Kentucky bluegrass.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
250 g was extracted from which subsamples of 100 seeds were selected for germination.
Germination Tests
Germination tests were performed in December 2000 to January 2001. Each seed lot was germinated according to standard rules for germination testing (International Seed Testing Association, 1999), with four replicates of 100 seeds. Seeds were germinated on top of filter paper (AGF725, Frisenette, Denmark) in small plastic germination boxes with a small water reservoir in each box (Jacobsen apparatus; International Seed Testing Association, 1999). At the beginning of the test, the filter paper was saturated in a 0.2% solution of KNO3, whereas pure water was used in the reservoir (International Seed Testing Association, 1999). The germination boxes were placed in two germination chambers (KBP 6395 LL, Termaks, Norway) with cool white fluorescent light (14000 to 30 000 lux, 400–700 nm) for 8 h per day at 25 ± 1°C and darkness for 16 h per day at 15 ± 1°C. For each species, two replicates were randomized on one shelf and placed in one of the germination chambers, and two replicates were randomized on one shelf in the other chamber. To diminish the potential effect of small temperature gradients within and between the germination chambers, the germination boxes were rearranged daily.
Seeds were considered germinated when the radicle had protruded 2 mm. Germination was recorded once or twice daily for 21, 20, and 23 d for red fescue, perennial ryegrass, and Kentucky bluegrass, respectively. Germination was recorded again at the end of the test at Day 31, 30, and 33, respectively. Germination was recorded for seed lots in the same order as the germination tests were started, and the exact time for inspection of the seeds was noted for each individual seed lot. To ensure that none of the seed lots possessed dormant seeds, seeds that had not germinated at the end of the germination test were exposed to a cold treatment by transferring to darkness and 5°C constantly for 14 d and then retransferring to 15/25°C and darkness/light for 16/8 h d–1 for 7 d. Only five seeds out of the total of 22000 seeds germinated after the cold treatment, indicating that none of the seed lots contained dormant seeds.
Fitting of Germination Time Courses
To derive biologically relevant information from the germination experiment, germination curves were fitted to the germination data. For each replicate from each of the 55 seed lots, an individual germination curve was fitted using an asymmetric multinomial distribution of the number of germinating seeds, giving a total of 220 fitted germination curves. The multinomial probabilities were calculated as the integral across the appropriate time interval of the density of the germination times as described by Hunter et al. (1984). The distribution of germination times within replicates was assumed to be generally hyperbolic (see Bibby and Sørensen, 2003). The fitting of germination curves and the estimation of parameters were accomplished using the software package R (The R Foundation for Statistical Computing, 2003). On the basis of the fitted function for each replicate within each seed lot, FGP, MGT, and T25-75 were estimated. The FGP and MGT were estimated using maximum likelihood with FGP on a logit scale. Estimated standard errors for these estimates were obtained from the observed information matrix. An estimate for the parameter T25-75 was subsequently calculated as
 | [1] |
For illustration of the variation in germination characteristics between cultivars, a mean germination curve was fitted to each cultivar with the average result of all seed lots and replicates within the cultivar. Similarly, the variation between seed lots within a cultivar was illustrated by fitting a germination curve to the average germination result of individual seed lots within a cultivar. This was done for one cultivar within each species, choosing the cultivar with largest variation in MGT among seed lots.
Analysis of Variance
To examine the variation in germination characteristics among seed lots, cultivars, and species, the estimates from the curve-fitting procedure were used in an ANOVA. The estimated values of FGP (percentage), MGT (days), and T25-75 (days), respectively, from each of the fitted curves were analyzed as a nested factorial design with replicates nested within seed lots, seed lots nested within cultivars, and cultivars nested within species. Cultivar differences were analyzed in a model with seed lot variation as a random effect. Species differences were analyzed in a model with cultivar variation as a random effect. The ANOVA was performed with the mixed procedure of the SAS package (SAS Institute, Cary, NC).
To obtain homogeneous variance in the ANOVA, the response variables were transformed and, where possible, the values of the response variables were weighted according to the reliability of the curve fitting. Estimates of FGP were analyzed with logit-transformed values of the FGP, and the estimates were weighted by the reciprocal standard error of the estimate from the curve fitting. Estimates of MGT were analyzed with square root transformed values of mean time to germination, and the estimates were weighted by the reciprocal standard error. Estimates of T25-75 were analyzed with square root transformed values of time from 25 to 75% of final germination, and the estimates were not weighted in the ANOVA.
Pair-wise comparisons of cultivars and species, respectively, were performed by t tests. To diminish the risk of erroneous conclusions in the multiple pair-wise comparisons, comparisons of cultivars were only done within species and not among species. The predicted values and 95% confidence limits were presented on the natural, untransformed scale. To compare the relative contribution of the variance components corresponding to cultivar, seed lot, and replicate for each of the three species, separate analyses of variance were performed for each species. Each variance component is presented on the transformed scale; that is, FGP is presented on a logit scale whereas variances of MGT and T25-75 are presented on a square-root time scale.
In the analysis of MGT, one observation was excluded as an outlier, whereas two observations were excluded as outliers in the analysis of T25-75. During the germination experiment it was noticed that the replicates in question were severely contaminated by fungi, which is likely to have delayed the germination considerably. All other analyses were performed both with and without the contaminated replicates, but the results were unchanged.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
|
There were significant differences in FGP among cultivars within red fescue (p = 0.0271), perennial ryegrass (p = 0.0009), and Kentucky bluegrass (p = 0.0004) when tested against seed lot variation within cultivars, indicating that there were true cultivar differences in FGP (Table 1). In red fescue, ‘Napoli’ had significantly lower FGP than ‘Cinderella’, and in Kentucky bluegrass, ‘Mardona’ had lower FGP than all other cultivars (Table 1). In perennial ryegrass, ‘Taya’ had higher FGP than ‘Merci’ and ‘Allegro’, and ‘Figaro’ also had higher FGP than Merci. The FGP differed least among cultivars of perennial ryegrass (from 93.3 to 97.7%) and most among cultivars of Kentucky bluegrass (from 76.3 to 91.6%).
|
There were significant differences in MGT among cultivars within red fescue (p < 0.0001) and Kentucky bluegrass (p = 0.0089), but not within perennial ryegrass (p = 0.41), indicating that there were true cultivar differences in MGT in two of the species when tested against seed lot variation within cultivars (Table 2). In red fescue, Napoli and ‘Smirna’ germinated more slowly (higher MGT) than Cinderella and ‘Symphony’, and in Kentucky bluegrass, ‘Conni’ germinated more slowly than ‘Broadway’, ‘Andante’, and Mardona (Table 2). The MGT differed most among cultivars of Kentucky bluegrass (from 7.2 to 8.1 d) and least among cultivars of perennial ryegrass (from 3.5 to 3.7 d) and with an intermediate range between red fescue cultivars (from 4.4 to 5.2 d) (Table 2).
|
The T25-75 also varied significantly among cultivars within red fescue (p < 0.0001) and Kentucky bluegrass (p < 0.0001), but not within perennial ryegrass (p = 0.64), indicating that there were true cultivar differences in T25-75 in two of the species when tested against seed lot variation within cultivars (Table 3). The T25-75 differed most among cultivars of Kentucky bluegrass (from 1.5 to 2.3 d), whereas it ranged least among cultivars of perennial ryegrass (from 0.8 to 0.9 d), and with an intermediate range among red fescue cultivars (from 0.8 to 1.4 d) (Table 3).
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Cultivar Differences and Seed Lot Differences
When cultivars are only represented by one seed lot each, variation between cultivars cannot be distinguished from variation between seed lots within cultivars, and if a cultivar happens to be represented by a particularly poor germinating seed lot, false conclusions may be drawn about the general germination characteristics of that cultivar. Gooding et al. (1989) found that the time to 50% germination varied from 6 to 20 d between 30 cultivars of Kentucky bluegrass, and Newell and Bludau (1993) found a corresponding variation from 5 to 14 d between 44 cultivars. These variations are much larger than those found among cultivars in this study (Table 2). In both of the mentioned studies, each cultivar was represented by only one seed lot each, and the slowest germinating cultivars generally had very low FGPs. Since low germination percentage is often accompanied by slow germination (Roberts, 1986; Culleton et al., 1991), the slow germination of certain cultivars is most likely to be partly explained by low seed vigor and quality of the applied seed lot. Considering the MGT ranging from 7.4 to 13.6 d among seed lots of Kentucky bluegrass cv. Conni, at least some of the variation in germination time found by Gooding et al. (1989) and Newell and Bludau (1993) was presumably due to seed lot differences in seed quality which were not representative for the cultivars. Use of more than one seed lot per cultivar and testing cultivar differences against seed lot differences would give more realistic information about true cultivar differences and possibly prevent false conclusions.
Naylor (1982) compared eight perennial ryegrass cultivars, each represented by one to five seed lots, and found significant cultivar differences in MGT but not in FGP. The present study, on the other hand, did not detect cultivar differences in MGT in perennial ryegrass but demonstrated such differences in red fescue and Kentucky bluegrass. The study confirms that true cultivar differences in germination characteristics do occur and can be detected by detailed registration of germination time courses and by summarizing with a flexible function.
The MGT is generally inversely related to FGP (Roberts, 1986) and in this study, the perennial ryegrass and red fescue cultivars with the lowest FGP (Table 1) also had the highest MGT (Table 2). Conversely, in Kentucky bluegrass, Mardona had the lowest FGP and also the lowest MGT whereas Conni had one of the highest FGP values and the highest MGT. Despite the low FGP of Mardona, which would normally indicate low quality, this cultivar germinated relatively fast, supporting the hypothesis that MGT is, in fact, partially under genetic control.
Variance Components and Implications for Experimental Design
Although the relative variation depended on the species and germination characteristic considered, cultivars accounted for 15 to 37% of the total variation, whereas seed lots accounted for 10 to 19% (Table 4). In comparison, Naylor (1981) found that differences between seed lots accounted for 16% of the overall variation in field emergence of Italian ryegrass (Lolium multiflorum Lam.), whereas cultivar differences accounted for 42% of the variation. Together, these estimates suggest that cultivars generally differ more than seed lots in germination characteristics. This conclusion can, to some extent, justify that cultivars are often represented by only one seed lot when searching for cultivar differences (e.g., Ellis et al., 1987); that is, true cultivar differences are likely to be detected because seed lot variation is generally smaller. On the other hand, given the occasionally very large seed lot differences within a cultivar (Fig. 2), there is a risk of failing to detect cultivar differences if only one seed lot is used, especially if this seed lot happens to be of a particularly poor quality. Therefore, studies of cultivar differences should be based on more than one seed lot per cultivar or, alternatively, preliminary experiments of seed lot variation should ensure that a representative seed lot of high quality is chosen from each cultivar.
Considering the variance contribution of the replicate factor, it is striking that this factor accounts for much more of the variation than the seed lot factor and the cultivar factor (Table 4). This may reflect the heterogeneity of seeds within a population, which can affect the variation between replicates. Additionally, the large variation of the replicate factor, especially for MGT, may reflect the difficulties of achieving homogeneous experimental conditions in germination studies. It is difficult to control temperature precisely without small gradients within a germination chamber (Ellis and Roberts, 1980), and even small temperature differences may affect MGT. Regular rearrangement of germination boxes can diminish the effect of temperature gradients, but despite application of four replicates of 100 seeds and daily rearrangement within and between germination chambers, there was still considerable replicate variation. Thus, in addition to considering the number of seed lots per cultivar, the number of replicates per seed lot should also be considered when conducting germination experiments. The choice of experimental design may, however, often depend on the scope of the experiment, and decisions may be aided by previous knowledge about the relative variance contribution of the factors considered.
|
ACKNOWLEDGMENTS |
|---|
Received for publication June 10, 2003.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Related articles in Crop Science:
This article has been cited by other articles:
|
|
 |
|
  A. Gazanchian, N. A. Khosh Kholgh Sima, M. A. Malboobi, and E. Majidi Heravan Relationships between Emergence and Soil Water Content for Perennial Cool-Season Grasses Native to Iran Crop Sci., February 1, 2006; 46(2): 544 - 553. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. U. Larsen and B. M. Bibby Differences in Thermal Time Requirement for Germination of Three Turfgrass Species Crop Sci., August 26, 2005; 45(5): 2030 - 2037. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| The SCI Journals | Agronomy Journal | Vadose Zone Journal | |||
| Journal of Natural Resources and Life Sciences Education |
Soil Science Society of America Journal | ||||
| Journal of Plant Registrations | Journal of Environmental Quality |
The Plant Genome | |||
