Applying Non-Parametric Methods on Environmental Data

Scientists of Tallinn University of Technology use nonparametric statistical methods in environmental engineering.

The methods of classical statistics are often less suited for analyses environmental data. The environmental databases need nonparametric statistical methods for analyses. The environmental time series are in general non-stationary. For this case special non-parametric trend test are developed. One of the most known non-parametric test is Mann-Kendall test. The Mann-Kendall test analyses the trends by means of Kendall rank correlation coefficient. This test can also apply on data consisting of blanks.

In recent works, scientists of Tallinn University of Technology used Mann-Kendall test for testing the trends of different loadings of Estonian rivers. The latest results of Estonian rivers loading studies are published in Iital et al., 2010 and 2014. The second problem in environmental data is the unbalance of data. That means, we have to compare the samples which sizes are quite different. For this case the bootstrap method is well suited. Also Wilcoxon signed-rank test and Kruskall-Wallis test are used in the case of unbalanced data.

Contact: Assoc. Prof. Margus Pihlak, Department of Cybernetics, Tallinn University of Technology, E-mail: margus.pihlak (at) ttu.ee

References

Iital, A., Pachel, K., Loigu, E., Pihlak, M., Leisk, Ü. (2010) Recent trends in nutrient concentration in Estonian rivers as a response to large-scale changes in land-use intensity and life-styles. Journal of environmental monitoring, 12(1), pages 178-188.

Iital, A; Klõga, M; Pihlak, M; Pachel, K; Zahharov, A; Loigu, E. (2014). Nitrogen content and trends in agricultural catchments in Estonia. Agriculture, Ecosystems and Environment, 198, pages 44−53