Trend Analysis of Hydrometeorological Time Series under the Scaling Hypothesis

Abstract

Mann-Kendall (MK) trend test is one of the commonly used statistical tests for detecting changes
in hydrometeorological time series. The test is derived as a function of the ranks of observations
making it distribution free as well as less sensitive to outliers and non-homogeneous time series.
However, climatic variability which is present in such series in the form of autocorrelation violates
its independent observations assumption and may lead to erroneous conclusions. The scaling
hypothesis has been proposed for modeling such variability in natural time series. This study
analyzed 50 years (1971-2020) observed data of rainfall, minimum temperature, maximum
temperature and wind speed (obtained from NiMet Kano airport station) for statistically significant
trends at monthly, seasonal and annual timescales. The data was analyzed using MK trend test and
its modified versions for the effect of autocorrelation and scaling/anti-scaling behavior on the trend
tests. The numbers of significant trends were found to reduce from 9, 7, 8 and 2 (when Mk trend
test was used) to 9, 7, 6, and 0 (when both autocorrelation and scaling/anti-scaling was considered)
for minimum temperature, maximum temperature, total rainfall and average wind speed
respectively. Thus some significant trends under independent observations assumption turn out to be spurious trends when the effect of dependence was considered. Moreover all significant trends
were found to be positive for rainfall and temperature series except for August that shows a
negative trend in maximum temperature. On the other hand, no change in wind speed was observed
at all timescales. The results thus indicated a warming and wetter climate for Kano, which will
most probably influence other variables including evapotranspiration and stream flow. The results
also showed the importance of considering the effect of climatic variability in climate change
studies.