A STATISTICAL APPROACH TO DETERMINE SAMPLE SIZE FOR MOISTURE CONTENT DETERMINATION IN RECALCITRANT FOREST TREE SEEDS

(A case study using Shorea leprosula and Shorea parvifolia)

by

B. Krishnapillay & M. Marzalina
Forest Research Institute Malaysia
Kepong
52109 Kuala Lumpur
Malaysia

INTRODUCTION

For the seeds of most agricultural and horticultural plants and certain tree species, sampling procedures including sample size for moisture content determination have been prescribed in the International Rules for Seed Testing (ISTA, 1976; ISTA, 1985) and the recent publication - A Handbook on Tree & Shrub Seed Testing (1991). However, neither the ISTA rules nor the Handbook discuss the procedure for the testing of tropical recalcitrant seeds.

Most recalcitrant forest seeds are large and the numbers available for testing are limited during most periods of collection. For these reasons it is not possible to follow the sampling rules recommended by ISTA for small sized orthodox seeds. On the other hand, it is also important that the sample size used for any test be statistically significant while being economical at the same time.

Various sample sizes have been used for moisture content determination of large seeds by different researchers (Rees, 1963; Yap, 1986; Tompsett, 1987). Chin (1988) and Berjak (1989) have suggested the use of a minimum of 20 seeds, while Bonner (1991) has suggested enough seed pieces to equal the weight of 5 seeds. These suggestions however are largely arbitrary and related to the experience of the researchers concerned. What is really required is the dertermination of a scientifically defined and tested minimum sample size.

The work reported in this paper attempts to use a statistical approach to obtain precise information on required sample size for moisture content determination using variation in moisture content between individual seeds as a basis.

APPROACH

During 1990, there was mast fruiting in the forests of Peninsular Malaysia. Seeds of ten species of trees belonging to the family Dipterocarpaceae were collected for this study. To illustrate the methodology used for moisture content determination, work carried out on two species namely Shorea leprosula and Shorea parvifolia is discussed below.

For each of the species, three well separated ( >80 km apart) population of fruiting trees were selected. From each population 1 500 fully mature seeds were collected. The seeds were packed in polythene bags and transferred to the laboratory on the day of collection. Four replicates of 25 seeds each were taken randomly from each population and dewinged. The moisture content of individual seeds were determined for each replicate. Each seed was marked, weighed, sliced into two halves and oven dried at 103°C for 20 hours (Krishnapillay et al., 1991). The oven dried seeds were cooled in a desiccator over silica gel and weighed. The moisture content was expressed as a percentage of the fresh weight. The variability of moisture content expressed as standard deviation for the three populations was used to determine the minimum sample size.

The sample size for moisture content determination was calculated using the formula developed by Cochran (1953) which is as follows:

where n = number of seeds to be sampled; t = the value of the normal deviate corresponding to the desired confidence level; s = the standard deviation of the factor to be tested; d = the level of precision or the margin of error.

The size of sample required at different confidence probabilities and the relavant levels of precision was calculated using this formula for varying standard deviations. Where the sampling fraction of the finite seedlot used in this study was more than 5% of the population (in our case 6.67%), a correction to the sample size was applied. This was done using the correction formula below (after Cochran, 1953):

where n' = corrected sample size; n = uncorrected sample size; N = population of seed lot.

Using the two formulae, a table showing the sample sizes necessary for moisture content determination at desired confidence probability and precision was constructed. Once the actual standard deviation of the variability of moisture content among the three populations of the species was known, the sample size could be determined.

In this study also, to determine variability in seed size and weight within the three populations, two replicates of 25 seeds each from the three populations were taken at random. For each seed, the length and width of the seed, wings and their respective weight were also tabulated.

RESULTS

Results based on measuring individual seeds from the three population to evaluate the variability in size and weight of seeds for Shorea leprosula and S. parvifolia, are shown in Tables 1 & 2 respectively.

Intact seed weight for S. leprosula from the three populations varied from 0.42 to 1.10 g with a mean of 0.71 ± 0.12 g, while in S. parvifolia it ranged from 0.28 to 0.52 g with a mean of 0.40 ± 0.06 g. The seeds that were dewinged weighed between 0.38 – 1.00 g and 0.24 – 0.46 g for the two species respectively. The average seed size for S. leprosula was found to be 0.92±0.13 cm in length and with a width of 0.34±0.04 cm at its widest circumference, while for S. parvifolia length was 0.97±0.09 cm and width 0.47±0.04 cm.

Table 1. Average weight and size of individual seeds of Shorea leprosula taken from three different populations

Table 2. Average weight and size of individual seeds of Shorea parvifolia taken from three different populations

Table 3 gives the calculated size of the sample required for testing the moisture content of seeds at various probability and levels of precision. This table could be further extended using Cochran's formulae described earlier if a larger standard deviation of seed moisture content were to be encountered.

The variability of moisture content expressed as standard deviation for the three populations over four replicates is presented in Table 4 for S. leprosula and in Table 5 for S. parvifolia.

Table 4 related to S. leprosula shows that the variability in moisture content expressed as standard deviation ranged between 2.34 and 2.52. Referring to Table 3, for this amount of variation, at a precision level of 0.5% and at confidence probability of 0.95, between 94 – 112 seeds would be required for moisture test; while at the 1.0% precision level and at the same confidence probability, only 25–30 seeds would be required. Similarly, Table 5 for S. parvifolia shows that the deviation in moisture content varied between 2.38 and 2.57 among the three populations studied. In this case, too, at 0.5% precision level, 94–112 seeds are required while at 1.0% precision level, a sample of about 25–30 seeds is sufficient. (Table 3).

Table 3. A table for determining the size of sample required for testing the moisture content of Shorea leprosula and S. parvifolia seeds at various levels of confidence and precision (based on a seed lot of 1 500 seeds)

Table 4. The variation of moisture content in relation to the three different population of Shorea leprosula

Note: a) R1–R4 = replicates; b) moisture content is expressed as a percentage on wet weight basis; c) values in parenthesis are the variability in moisture content expressed as standard deviation.

Table 5. The variation of moisture content in relation to the three different populations of S. parvifolia

Note: a) R1–R4 = replicates; b) moisture content is expressed as a percentage on wet weight basis; c) values in parenthesis are the variability in moisture content expressed as standard deviation.

DISCUSSION

Seed moisture is one of the most important factors affecting viability and storability of seeds. It is essential therefore that proper and sufficiently precise methods be devised for seed testing. To date, very little work has been done for establishing standards for large seeds of tropical species. The available information and guidelines used are generally those that have been developed for orthodox seeds (ISTA, 1976; 1985; Gordon et al., 1991)

The average seed size and weight of recalcitrant seeds usually far exceeds those or orthodox seeds (Roberts et al., 1984). Generally the 1 000 seed weight of recalcitrant seeds exceeds 500 g (Chin et al., 1984). Such large seed weights occur due to both the high moisture content of the seed, which ranges from 20–70% (on a fresh weight basis), and also due to their large sizes. In this study the mean 1 000 seed weight of intact seeds of S. leprosula was approximately 710 g, while that of S. parvifolia was 400 g. It was generally observed that there was a wide variation in size and weight both within and between populations. The varying standard deviation showed that for these two species there is no uniformity of size or weight at one location or between locations (Tables 1 and 2). On the other hand, while varying between populations, moisture content showed fairly uniform standard deviations within a given species (Tables 4 and 5). This uniformity in standard deviation is most important because it enables a definite sample size to be adequately defined for moisture content testing for the species.

The study on minimum sample size required for moisture content determination showed that the standard deviation over the three populations for S. leprosula ranged between 2.34 and 2.52, while for S. parvifolia it ranged between 2.38 and 2.57. Over this range, for a probability of 0.95 and at a precision level of 1.0%, a sample size of 25–30 seeds was found to be sufficient to accurately determine the moisture content in these two species. Taking the upper limit of deviation for both species (i.e. 2.75, see Table 3), a sample size of 30 seeds for each of these two species is recommended. A 1.0% level of precision is selected for the moisture test is a destructive process and only 30 seeds will be used at this level. If a 0.5% precision level is used, this would destroy 112 seeds, which may be undesirable in tropical species where scarcity of seeds and the small quantity available are major constraints. Bonner (1982), in working with recalcitrant seeds of temperate species, suggested that a precision to tolerance of variability of up to 2.5% is acceptable for large seeds with moisture content of more than 12%. Hence, in the present study, a 1.0% level of precision was considered sufficiently accurate for the sample size of S. leprosula and S. parvifolia.

Mok (1972), working with oil palm seeds, Chin & Lassim (1987), working with recalcitrant fruit seeds and Krishnapillay et al (1991) working with forest species, were able to establish the minimum sample size for their respective species using a similar statistical approach attempted in this study.

CONCLUSION

This study indicates that 30 dewinged seeds taken in three samples of ten or six replicates of five seeds each, after cutting each seed in two prior to drying, can be used for accurate moisture content determination for seeds of Shorea leprosula and S. parvifolia.

ACKNOWLEDGEMENTS

The authors wish to thank the Director General of the Forest Research Institute Malaysia for permission to publish this paper and Mr Louis Retnam for critically reviewing the manuscript.

Author's Note

The senior author welcomes valuable suggestions and criticisms to further improve on the approach attempted in this paper.

References

Berjak, P. (1989). International Seed Testing Association. Working Group on Recalcitrant Seeds of The Seed Storage and Moisture Committee. A Communicae, 8 p.

Bonner, F.T. (1982). Measurement and Management of Tree Seed Moisture. Danida Forest Seed Centre, Technical Note.1, 28pp.

Bonner, F.T. (1991). Measurement of Moisture Content. In: Tree Shrub Seed Handbook. A.G. Gordon, P. Gosling and B.S.P. Wang eds. Chapter 12. International Seed Testing Association, Zürich, Switzerland.

Chin, H.F. (1988). Recalcitrant Seeds - A Status Report. International Board for Plant Genetic Resources (IBPGR), Rome, 28 pp.

Chin H.F., Y.L. Hor and M.B. Mohd Lassim (1984). Identification od Recalcitrant Seeds. Seed Sci. & Technol., 12: 429–436

Chin, H.F. and M.B. Mohd Lassim (1987). Determination of Moisture Content of Recalcitrant Seeds by Microwave Technique. Acta Hort. 215: 159–164

Cochran, W.G. (1953). Sampling Techniques. pp. 50–56. John Wiley and Sons, New York.

Gordon, A.G., P. Gosling and B.S.P. Wang eds. (1991). Tree Shrub Seed Handbook. International Seed Testing Association, Zürich, Switzerland.

International Seed Testing Association (1976). International Rules for Seed Testing. Seed Sc. & Technol., 4: 51–177

ISTA (1985). International Rules for Seed Testing. Seed Sci. & Technol., 13: 299–355

Krishnapillay, B. Marzalina Mansor, Yap, S.K., Chua, L.S.L., Ang, K.C., Siti Asha Abu Bakar and Zaiton Salleh (1991). Determination of Seed Testing Standards-Moisture Content of Hopea odorata seeds. J.Trop For Sci 4(2): 170–178

Mok, C.K. (1972). Sample Size for Moisture and Viability Testing of Oil Palm (Elaeis guineensis JACQ) Seeds. Proc. Int. Seed Test. Ass. 37(3): 751–761

Rees, A.R. (1963). A Large Scale Test of Storage Methods for Oil Palm Seeds. J. W. Afr. Oil Palm Res. 4:46–51

Roberts, E.H., M.W. King and R.H. Ellis (1984). Recalcitrant Seeds: Their Recognition and Storage. In: Crop Genetic Resources: Conservation and Evaluation. Allen and Unwin, London pp 38–52.

Tompsett, P.B. (1987). Desiccation and Storage Studies on Dipterocarp Seeds. Annals of Applied Biology. 110: 371–379

Yap S.K. (1986). Effect of Dehydration on Germination of Dipterocarp Fruits. Proceedings International Union of Forestry Research Organisation (IUFRO) Symposium. Report No. 12.F, pp 168–181

Forest Genetic Resources Information no. 21. FAO, Rome (1993)
Manuscript received June 1993