Dear Ahmed: if you have an experimental design, you need at least 3 to 4 replicates, but if you want to do a test such as chi square, you need less number of observations and simple test. When you choose t-test you will need at least 60 observations and up. In this case you will have better conclusion.Cheers.
The best way to determine your sample size is to assess the power of the statistical test that you will be using. The g*power program is designed to answer this question.
In all range of the study, first of all, there should be all samples reflecting the characteristics of the range. If they can be graphical, the sample density should be uniform. Second, we should consider every influencing factor. Third, we should eliminate the samples or factors that do not conform to the law.