Artificial-intelligence-based detection tests for the identification of shifts and trends in Canadian hydrometric data

Abstract

Two artificial-intelligence-based detection tests for the identification of shifts and trends in data sequences are described and applied to Canadian hydrometric data. These tests are based on the Kohonen neural network and fuzzy c-means approach. They are applied for the detection of shifts and trends in annual mean and daily maximum streamflow data from 43 Canadian hydrometric stations. The results of the tests are compared with those from conventional detection tests, such as the Mann–Whitney test for shifts and the Mann–Kendall test for trends. These results support conclusions from previous studies about the presence of trends in Canadian hydrometric data. As a whole, the artificial-intelligence-based and conventional tests may be used to confirm one another. In some cases, given their respective strengths and weaknesses, the tests may complement one another. One should therefore consider the use of more than one detection test for determining the presence or absence of anomalies in data sequences.Key words: hydrometric data, detection tests, shifts, trends, artificial intelligence.