Geostatistical regional trend detection in river flow data

Many studies have identified global warming and climate change as some of the biggest challenges facing Canada. In this paper, the regional temporal trend in river flows is investigated using a space-time model. Though the primary focus is the time component, the spatial relationship among monitoring stations in a region is used to develop a space-time model that is composed of a random time trend as a function of space, and a random error term as a function of both time and space. The estimate of regional time trend is a linear combination of the differenced observations that minimizes the variance of estimated errors. Data from 248 river stations in the Reference Hydrometric Basin Network (RHBN) established by Environment Canada is analysed. These hydrological monitoring stations are grouped into ten non-overlapping homogeneous regions covering all of Canada. An estimate of trend, along with its variance, is calculated for each region. Some significant trends are found for the annual mean, maximum and minimum flows, as well as for the mean monthly flows for July and December, and are consistent with those detected in other Canadian studies. Copyright (C) 2001 John Wiley Sons, Ltd.