HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Figures Only Full Text (PDF) Free Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in Web of Science Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Web of Science (4) Citing Articles via Google Scholar Google Scholar Articles by Tang, S. Articles by Rucker, M. Search for Related Content PubMed Articles by Tang, S. Articles by Rucker, M. Agricola Articles by Tang, S. Articles by Rucker, M.

SEED PHYSIOLOGY, PRODUCTION & TECHNOLOGY

Survival Characteristics of Corn Seed during Storage

I. Normal Distribution of Seed Survival

^a Dep. of Agronomy, University of Kentucky, Lexington, KY 40546-0091 USA

dtekrony{at}ca.uky.edu

	ABSTRACT

TOP NOTES ABSTRACT INTRODUCTION Materials and methods Results Discussion REFERENCES

 
Understanding the survival characteristics of hybrid corn (Zea mays L.) seed during storage is necessary to predict seed deterioration. This investigation tested one key assumption of the Ellis and Roberts viability equation, namely, that seed survival is normally distributed. Eleven corn seed lots (six hybrids) with little mechanical injury and a wide range in initial vigor were stored in various combinations of constant temperatures (20, 30, 40, and 50°C) and seed moisture contents (100, 120, 140, and 160 g kg^-1, fresh weight basis). Seed-survival curves were constructed by conducting successive germination tests at frequent intervals during storage. The ² goodness-of-fit test was used to evaluate the normality of survival curves constructed from either full or truncated (germination between 95 and 5%) data sets. When the data were truncated, the majority (79%) of the 187 survival curves analyzed were classified as normal (P > 0.05) or near-normal [P < 0.05 but relatively small ², heterogeneity factor (H = ²/df) < 10]. Only 57% of the curves from the full data set followed a normal or near-normal distribution. Seed moisture and storage temperature had no consistent effect on the shape of the survival curves. Survival of low-vigor seed lots was more likely to be normally or near-normally distributed than was survival of high vigor seed lots. The assumption that seed survival is normally distributed was generally valid for truncated data sets of hybrid corn seed in constant storage environments.

Abbreviations: ISTA, International Seed Testing Association

	INTRODUCTION

TOP NOTES ABSTRACT INTRODUCTION Materials and methods Results Discussion REFERENCES

 
MATHEMATICAL APPROACHES to evaluating the rate of seed deterioration have been extensively investigated (Roberts, 1972, 1973; Ellis and Roberts, 1980a, 1980b, 1981; Moore and Roos, 1982; Moore and Jolliffe, 1987). The goal was to develop a means of predicting seed germination in germplasm repositories or during commercial seed storage. From a quantitative viewpoint, seed deterioration can best be defined as an increased probability of death of an individual seed as deterioration proceeds. Seed death is indicated by the failure to germinate and seed longevity is the period until seed death occurs.

Seed-survival curves (germination percentage plotted against time) have reverse sigmoid shapes resembling a negative cumulative normal distribution (Roberts, 1972, 1973). Such a shape is expected if (i) under any given set of storage conditions, a seed lot has a particular mean viability period and a random distribution of the viability period of the individual seeds around this mean value and (ii) the distribution of the viability period of individual seeds in a population can be described by a normal distribution

(1)

where y is the relative frequency of deaths occurring at time p, is the mean viability period, and is the standard deviation of the distribution of deaths in time.

Assuming that seed survival follows a normal distribution, Ellis and Roberts (1980a) developed a viability model (Eq. [2]) to predict declines in seed germination during storage as a function of storage temperature, seed moisture and initial seed quality.

(2)

In this equation, v is the probit or normal equivalent deviate of germination (%), K_i is an initial seed quality constant (on the probit scale), p is the storage period (days), m is seed moisture content (%, fresh weight basis), t is temperature (°C), and K_E, C_W, C_H, and C_Q are constants whose values are assumed to be the same for all seed lots of a species.

A reliable application of the viability equation is dependent on several assumptions. One key assumption is a normal distribution of seed deaths. Visual examination of seed-survival curves in many grain crop species, such as barley (Hordeum vulgare L.), wheat (Triticum aestivum L.), and rice (Oryza sativa L.), suggests that they follow a normal distribution (Roberts, 1972, 1973; Ellis and Roberts, 1980b, 1981; Ellis et al., 1982, 1990). However, there are several reports in which ² tests for normality indicated that seed-survival curves deviated from the normal distribution (Moore and Roos, 1982; Moore and Jolliffe, 1987; Wilson et al., 1989; Fabrizius et al., 1999). Wilson et al. (1989) reported that analysis of survival curves of field bean (Phaseolus vulgaris L.) seed lots with low initial germination always resulted in significant ². They suggested that adjustment for the initial proportion of nongerminable seeds would result in seed-survival curves that were more likely to follow a normal distribution (Wilson et al., 1989), but Ellis et al. (1990) refuted the need for correction, stating that correction was unnecessary because of the arbitrary start of seed storage, and can be misleading by distorting the estimate of K_i.

Two factors may be responsible for a significantly large ² (Finney, 1971). First, individual seeds in the seed lot may not deteriorate wholly independently of one another, resulting in random heterogeneity. Secondly, a significant ² may occur because the underlying mathematical model is incorrect, which tends to produce a systematic deviation of observed from expected (Finney, 1971). In the former case, the ² is inflated and the heterogeneity factor (H = ²/degrees of freedom) must be calculated and used to adjust the variance. Unfortunately, it is difficult to distinguish clearly which form of heterogeneity, random or systematic, is causing a large ². Finney (1971) stated that a large ² should be treated with extreme caution and a combination of ² and careful inspection of a graph of the data is essential.

Little information is available about the distribution followed by corn seed survival during storage. Consequently, the objectives of this investigation were (i) to determine whether the seed-survival curves of corn follow a normal distribution in various storage conditions of constant temperature and seed moisture content and (ii) to determine the effects of initial seed vigor and the storage environment (seed moisture and temperature) on the shape of corn seed-survival curves.

	Materials and methods

TOP NOTES ABSTRACT INTRODUCTION Materials and methods Results Discussion REFERENCES

 
Three storage experiments were conducted in constant environmental conditions (temperature range 20–50°C; moisture contents from 100–160 g kg^-1, fresh weight basis) from 1995 to 1997 using several hybrid corn seed lots. Seed germination was measured at frequent intervals to provide seed survival data for subsequent analysis.

Seed lots from six hybrids that had low levels of mechanical injury and disease infection, but high germination (89%) and a range in seed vigor were selected (Table 1) . All seed lots were treated with Captan 400 [N-(trichloromethythio)-cyclohex-4-ene-1,2-dicarboximide]. The initial seed moisture content ranged from 110 to 130 g kg^-1 (fresh weight basis) as determined by drying ground seeds at 130°C for 4 h (International Seed Testing Association [ISTA], 1993). The average weight per seed was determined by weighing at least four replications of 100 seeds. Initial seed quality of all seed lots was determined by standard germination, accelerated aging (45°C and 72 h), and cold tests as described by ISTA (1993, 1995). Normal and abnormal seedlings were evaluated according to the criterion described by ISTA (1993) and Association of Official Seed Analysts (1992). Initial seed vigor was determined by averaging accelerated aging and cold test germination.

Incubators were used for long-term storage at 20 and 30°C (± 1.0°C), while a water bath was used for shorter storage periods at 40 and 50°C (± 0.5°C ). During storage, foil packets were removed at regular intervals from hours to weeks and seed moisture content (20–25 seeds or 7–10 g) and standard germination (ISTA, 1993) determined immediately. Sampling continued until seed germination declined to <5% or all seed packets were exhausted. Data from successive germination tests were used to construct seed-survival curves.

Experiment 1
Four seed lots (Lots 1–4) from four hybrids (Table 1) were stored in 13 of the 16 treatments in a four by four factorial arrangement of four temperatures (20, 30, 40, and 50°C) and four seed moisture levels (100, 120, 140, and 160 g kg^-1). The combinations of 20°C and 100 g kg^-1, 20°C and 120 g kg^-1, and 30°C amd 100 g kg^-1 were omitted due to the slow seed deterioration at these low temperatures and low moisture levels. The seeds were first divided into two replications and conditioned to the desired moisture levels separately.

Experiment 2
Five seed lots (Lots 1–5, Table 1) were evaluated using a rapid aging test at 40°C and 160 g kg^-1 seed moisture content. The seeds were conditioned to 160 g kg^-1 before they were divided into two replications. Samples were collected at 1- to 3-d intervals for periods of 18 to 25 d. In addition, samples of Lot 5 were collected at 8- and 24-h intervals to investigate the effect of a shorter sampling interval and sample size. The germination tests were conducted with three replications of three sample sizes: 100, 200, and 400 seeds.

Experiment 3
Seeds from six seed lots (Lots 6–11) of two hybrids with a range in seed vigor and flat and round seed with similar seed size (Table 1) were stored in all combinations of a three by three factorial arrangement of temperature (30, 40, and 50°C) and seed moisture content (120, 140, and 160 g kg^-1). Seeds were conditioned to three moisture levels and sealed in foil packets before the foil packets were divided into two replications.

Data Analysis
After plotting germination percentage from each replication against storage time, 187 complete survival curves (germination declined to 50%) were available for analysis. These curves were examined by visual inspection and by goodness-of-fit tests in conjunction with probit analysis to determine if they followed a normal distribution. The PROBIT procedure (SAS Institute, 1988) transforms germination percentages into the probit scale, which linearizes seed-survival curves if survival is normally distributed, and calculates ², which was used to test for normality. The analysis was performed on all of the germination data (full data set) or after dropping observations where the germination percentage was >95 and <5% (truncated data set). The ² was first used to classify the curves as essentially normal (P > 0.05) or non-normal (P < 0.05). The heterogeneity factor (H = ²/degree of freedom; Finney, 1971), together with visual examination of the fit between the survival curves from the probit analysis and the observations, was then used to further classify non-normal curves. Curves with large ² values (P < 0.05), but relatively small heterogeneity factors (H < 10) were recategorized as near-normally distributed. The remaining curves (both P < 0.05 and H > 10) were described as non-normal distributions.

Contingency tables were constructed to test the relationship between distribution of seed survival and seed moisture content, storage temperature and initial seed vigor. The normal and near-normal survival curves were combined as a group in comparison to non-normal curves.

A separate analysis, involving Abbott's formula (Finney, 1971), was employed for seed lots with lower initial germination (90–91%, Lots 6, 7, and 11; Table 1). The survival data derived from these seed lots were subjected to a correction rate of 10% before the probit analysis to determine if the correction affects the fit of the curves to normal distributions. The correction was calculated by , where g^*_i and g_i are observed and corrected germination percentage, and c is correction rate. For example, an observed 90% germination was corrected to 91, 40 to 45.5%, etc., and analyzed as before.

	Results

TOP NOTES ABSTRACT INTRODUCTION Materials and methods Results Discussion REFERENCES

 
Survival Curves
In accordance with the assumption underlying the viability equation (Eq. [2]), negative cumulative normal distributions were fit by probit analysis to the 187 seed-survival curves. Most of the survival curves exhibited a symmetrical sigmoid shape as illustrated for normal, near-normal, and non-normal curves in Fig. 1 .

View larger version (18K): [in this window] [in a new window]   Fig. 1 Seed-survival curves of corn seed lots stored in constant environments. The solid line represents the full data set and the dashed line the truncated data set (only germination percentages <95 and >5%). (A) Typical survival curves following a normal distribution (storage environment 40°C, 160 g kg-1 seed moisture content). (B) Typical survival curves following a near-normal distribution (storage environment 40°C, 140 g kg-1 seed moisture content). (C) Typical survival curves following a non-normal distribution (storage environment 40°C, 160 g kg-1 seed moisture content)

Probit analysis of the survival curves in Fig. 1 resulted in large differences in ² and H (Table 2) . Seed lots with high germination percentages tended to have large ² when the full data set was evaluated (e.g., Lots 8 and 10 in Fig. 1B or Lots 1 and 4 in Fig. 1C), confirming that high germination percentages may disturb the ² goodness-of-fit test.

Sample Size and Interval
Experiments were conducted using one corn seed lot (Lot 5) to determine if the shape of the seed-survival curve was affected by sampling interval or the number of seeds tested for germination. The shape and the fit of the survival curves to a normal distribution were not changed by sampling at 8-h compared with 24-h intervals during 7 d of storage at 40°C and 160 g kg^-1 seed moisture content (Fig. 2A) . Likewise, there was little effect of using 100, 200, or 400 seeds for germination tests on the shape of survival curves (Fig. 2B). The survival curve derived from a germination test of 200 seeds was indistinguishable from that of 400 seeds, suggesting that 200 seeds were adequate when constructing seed-survival curves and testing the fit of the curves to a normal distribution.

View larger version (16K): [in this window] [in a new window]   Fig. 2 Corn seed-survival curves generated with (A) different sampling intervals and (B) various sample sizes for the germination test. Lines are the fitted normal distribution. The curve for 400 seeds was identical to the curve for 200 seeds

	Discussion

TOP NOTES ABSTRACT INTRODUCTION Materials and methods Results Discussion REFERENCES

 
The Ellis and Roberts viability equation (Eq. [2]) is based on the assumption that seed-survival curves can be accurately described by a negative cumulative normal distribution (Ellis and Roberts, 1980a,b, 1981; Ellis et al., 1982). The results of this study with hybrid corn, using the ² goodness-of-fit test, generally support this assumption in a wide range of constant storage conditions. We evaluated 187 survival curves and a majority of the curves (79%) were found to be normally (P > 0.05) or near-normally (P < 0.05 but relatively small ², H < 10) distributed, although only 15% of the survival curves followed a normal distribution, based solely on the non-significance of ². Our results are in good agreement with the conclusions of Roberts (1972), Ellis and Roberts (1980a, 1980b, 1981), Ellis (1988), and Ellis et al. (1982, 1990) for numerous crop species, including barley, wheat, rice, and corn, although their conclusions were based primarily on visual examination of survival graphs.

Deviation from a normal distribution and a significant ² may be attributed to three sources (Ellis et al., 1990): (i) errors in determining and maintaining constant storage environments (in particular, seed moisture); (ii) errors in sampling and testing seeds for germination; and (iii) errors in defining the survival distribution, i.e., assuming normality when, in fact, the distribution is non-normal. If the mathematical model (normal distribution) is not appropriate, the observed germination percentages will lie on a curve (systematic deviation) while the other errors result in large random scatter about the fitted curve. Thus, a significant ² does not necessarily mean that the distribution is not normal. Although a large ² often occurred in probit analysis of full and truncated data sets (Tables 2, 3, and 4), the majority of the survival curves did not show systematic deviation of observations from the normal curve as illustrated in Fig. 1A and 1B, suggesting that random errors were the primary sources of large ²s.

When a large ² resulted from random errors, H was calculated. An H of 10 was as an effective criterion to classify curves with large ² (P < 0.05) into two groups, i.e., near-normal and non-normal distributions. An H of 10 was identified by inspecting the fit between the normal curve and the actual data (Fig. 1B and 1C; Table 2). There seemed to be little difference between the fitted curves and the observations for H < 5 and for 5 < H < 10 (Fig. 1B and Table 2). In comparison, the large deviation of the data points from curves with H > 10 (Fig. 1C and Table 2) was easily observable. Despite a significant ² at relatively small P values, all curves with relatively small H values fitted the normal curve very realistically, demonstrating the capability of combining probabilities and H in ² goodness-of-fit tests for validating the assumption of normality.

Insufficient or unbalanced sampling (data sparsity) and variability of germination tests contribute to non-normality (Ellis et al., 1990; Fabrizius et al., 1999), because they increase the deviation of observations from the expectation. If only one or two data points were available in the most useful portion of the survival curve, i.e., germination percentages between 5 and 95%, probit analysis would be less certain (Wilson et al., 1989). When only a few data points occurred within the critical range, there were large differences in the fit of normal distributions between full and truncated data sets (Tang, 1998); however, shortening the sampling interval to increase the data points between 5 and 95% germination to more than five did not increase the number of survival curves testing normal (Fig. 2A). Thus, at least five or more germination data points should occur in the critical region of the survival curve, i.e., between 95 and 5% germination, to provide a curve that accurately reflects seed deterioration. It appears that no further improvement can be expected from more frequent sampling.

Variation in germination test results can contribute to a nonsignificant ², so an increase in the number of seed tested may reduce the variation. However, increasing sample size with one corn seed lot in this study did not affect the distribution of the survival curve (Fig. 2B). The curve from germination tests with 400 seeds was almost the same as that with 200 seeds. Thus, there appears to be little reason to increase seed number beyond 200 for the germination test when constructing corn seed-survival curves.

Although most corn seed-survival curves were normally or near-normally distributed, approximately 20% of the curves did not follow normal or near-normal distributions. Roberts (1972) suggested that non-normal distributions may occur in extreme conditions when all seeds are dead after a few days of storage. Moore and Roos (1982) and Moore and Jolliffe (1987) reported non-normal distributions in high temperature and relative humidity storage conditions confirming Roberts' (1972) suggestion. Non-normal distributions occurred in our experiments in all storage conditions with neither moisture nor temperature showing a consistent effect on the distribution of the survival curve (Table 4). Thus, the rapid aging test, which employs high temperature and high seed moisture content as storage conditions, does not necessarily lead to a non-normal distribution of seed survival. On the other hand, initial seed vigor seemed to significantly influence the distribution of the curves as measured by ² tests (Table 4). The curves for low-vigor seed lots followed normal or near-normal distributions more frequently than those of medium- and high-vigor seed lots. High-vigor seed lots may exhibit the two phase deterioration curve described by Bernal-Lugo and Leopold (1998) where a period with no seed deterioration is followed be rapid deterioration that follows a normal distribution.

Our research does not support the suggestion of Wilson et al. (1989) and Wilson (1995) that correction for low initial germination improves the fit of seed-survival curves to the normal distribution. The normal distribution is symmetric, but an asymmetric incomplete survival curve can also follow a normal distribution. In one seed lot of foxtail millet (Setaria italica L.) with initial germination only marginally >50%, the survival curve was precisely described by an incomplete negative cumulative normal distribution (Ellis et al., 1990). Similar results were found here in that those seed lots with low initial germination followed a normal or near-normal distribution. The seeds which died before storage was initiated were actually part of the population (seed lot) under study because the start (zero time) of the storage experiment is somewhat arbitrary, the storage experiment is only a continuation of progressive deterioration beginning at physiological maturity. The results of this study, together with evidence from studies of Roberts (1972), Ellis and Roberts (1980a, 1980b), and Ellis et al. (1990) support the conclusion that correction for low initial germination in the probit analysis is not helpful.

The Ellis and Roberts viability model (Eq. [2]) is based on the assumption that seed-survival curves can be described by negative cumulative normal distributions. Evaluation of this assumption with 187 corn seed-survival curves suggests that the assumption is generally valid (79% of the curves were normal or near-normal) when only germination percentages between 95 and 5% were included in the analysis. The distribution of the survival curves was not affected by storage conditions, but there was a tendency for high quality seed lots to have more non-normal distributions. Overall our results suggest that the distribution of seed-survival curves does not limit the use of the Ellis-Roberts viability equation (Ellis and Roberts, 1980a) to predict changes in corn seed germination during storage or to estimate the initial quality (K_i) in a rapid aging test, provided care is taken in the construction of the curve.Association of Official Seed Analysts 1992; International Seed Testing Association 1995

	ACKNOWLEDGMENTS

 
This research was supported in part by a grant from the American Seed Research Foundation. We want to thank Adler's Seeds, Sharpsville, IN; Beck's Superior Hybrids, Atlanta, IN; Caverndale Farms, Danville, KY; Huber Seeds, West Lebanon, IN; and Pioneer Hi-Bred International Inc., Johnston, IA for supplying the seed lots used in these experiments.

	NOTES

TOP NOTES ABSTRACT INTRODUCTION Materials and methods Results Discussion REFERENCES

 
Contribution from Kentucky Agric. Exp. Stn. No. 98-06-119.

Received for publication August 8, 1998.

	REFERENCES

TOP NOTES ABSTRACT INTRODUCTION Materials and methods Results Discussion REFERENCES

 

• Association of Official Seed Analysts. Seedling evaluation handbook. NE: Lincoln, 1992 No. 35. AOSA..
• Bernal-Lugo I., Leopold A.C. The dynamics of seed mortality. J. Exp. Bot. 1998;49:1455-1461.[Abstract/Free Full Text]
• Ellis R.H. The viability equation, seed viability nomographs, and practical advice on seed storage. Seed Sci. Technol. 1988;16:29-50.
• Ellis R.H., Hong T.D., Roberts E.H. Moisture content and the longevity of seeds of Phasolus vulgaris. Ann. Bot. 1990;66:341-348.[Abstract/Free Full Text]
• Ellis R.H., Osi-Bonsu K., Roberts E.H. The influence of genotype, temperature and moisture on seed longevity in chickpea, cowpea and soya bean. Ann. Bot. 1982;50:69-82.[Abstract/Free Full Text]
• Ellis R.H., Roberts E.H. Improved equations for the prediction of seed longevity. Ann. Bot. 1980;45:13-30 a.[Abstract/Free Full Text]
• Ellis R.H., Roberts E.H. The influence of temperature and moisture on seed viability period in barley (Hordeum distichon L.). Ann. Bot. 1980;45:31-37 b.[Abstract/Free Full Text]
• Ellis R.H., Roberts E.H. The quantification of aging and survival in orthodox seeds. Seed Sci. Technol. 1981;9:373-409.
• Fabrizius E., TeKrony D., Egli D.B., Rucker M. Evaluation of a viability model for predicting soybean seed germination during warehouse storage. Crop Sci. 1999;39:194-201.[Abstract/Free Full Text]
• Finney D.J. Probit analysis, 3rd ed Cambridge, UK: Cambridge University Press, 1971.
• International Seed Testing Association. International rules for seed testing. Rules 1993. Seed Sci. Technol. 1993;21:1-287 (suppl.)..
• International Seed Testing Association. Handbook of vigour test methods, 3rd ed Zurich, Switzerland: ISTA, 1995.
• Moore F.D., Jolliffe P.A. Mathematical characterization of seed mortality in storage. J. Am. Soc. Hortic. Sci. 1987;112:681-686.
• Moore F.D., Roos E.E. Determining differences in viability loss rates during seed storage. Seed Sci. Technol. 1982;10:283-300.
• Roberts E.H. Storage environment and control of viability. In: Roberts E.H., ed. Viability of seed. Syracuse, NY: Syracuse Univ. Press, 1972:14-58.
• Roberts E.H. Predicting the storage life of seeds. Seed Sci. Technol. 1973;1:499-514.
• SAS Institute. 1988. SAS/STAT user's guide. 6.03 ed.SAS Inst., Cary, NC.
• Tang S. Survival characteristics of hybrid corn seed during constant storage. University of Kentucky, Lexington: Ph.D. diss. (Diss. Abst. 99-07726), 1998.
• Wilson D.O., Jr. Storage of orthodox seeds. In: Basra A.S., ed. Seed quality: Basic mechanisms and agricultural implications. New York: Food Products Press, 1995:173-208.
• Wilson D.O., Jr., McDonald M.B., Jr., St. Martin S.K. A probit planes method for analyzing seed deterioration data. Crop Sci. 1989;29:471-476.[Abstract/Free Full Text]

This article has been cited by other articles:

				S. Tang, D. M. TeKrony, D. B. Egli, and P. L. Cornelius An Alternative Model to Predict Corn Seed Deterioration during Storage Crop Sci., March 1, 2000; 40(2): 463 - 470. [Abstract] [Full Text]

				S. Tang, D. M. TeKrony, D. B. Egli, and P. L. Cornelius Survival Characteristics of Corn Seed during Storage: II. Rate of Seed Deterioration Crop Sci., September 1, 1999; 39(5): 1400 - 1406. [Abstract] [Full Text]

This Article Abstract Figures Only Full Text (PDF) Free Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in Web of Science Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Web of Science (4) Citing Articles via Google Scholar Google Scholar Articles by Tang, S. Articles by Rucker, M. Search for Related Content PubMed Articles by Tang, S. Articles by Rucker, M. Agricola Articles by Tang, S. Articles by Rucker, M.