A comparison of systematic versus stratified-random sampling design for gradient analyses: a case study in subalpine Himalaya, Nepal

Depending upon the scale and purpose of the study, different sampling designs vary in their advantages and disadvantages. In contrast to the thorough debates on the numerical methods used in gradient analyses, sampling design has not been discussed and evaluated to the same degree. We assess the difference between results obtained by systematic and stratified-random sampling designs regarding (i) heterogeneity of species assemblages and (ii) relationships between vegetation, environment, and geographical space. We compared the statistical attributes of two data sets of the same size obtained using a systematic (S) and a stratified-random (SR) sampling design, along the same physiognomically defined spatial gradient from grassland to old forest, via a thicket of shrub and pioneer forest in subalpine zone of Nepal Himalaya. There is not much pronounced difference between the two sampling designs regarding species heterogeneity; however, systematic sampling captures slightly greater heterogeneity of species assemblages, whereas greater redundancy is revealed by SR sampling design. The species-environment correlation is significantly higher for data collected using the S sampling design than that for the SR data set, whereas the variance explained by environmental variables and space together is higher for the SR dataset. The two sampling designs yield data that are not very different with respect to common multivariate statistics, such as eigenvalue and gradient length. However, systematic sampling is found to be more efficient than stratified-random sampling not only in terms of effort and time but better results especially regarding species environment correlations are also obtained by this technique during ordination and gradient studies.