The use of Monte Carlo simulations in the detection of nonstationarity in regional peak annual streamflow time series of British Columbia

• Main Author: Stonewall, Adam
• Language: English
• Published: 2009
• Online Access: http://hdl.handle.net/2429/10972

The stationarity of historical hydrological time series is a fundamental assumption of water resources management. When this assumption proves valid, a time series is regarded as a stochastic process with constant statistical parameters. Anthropogenic activities may invalidate this assumption. Land use changes within a watershed may alter hydrologic characteristics. On a more regional scale, increased levels of atmospheric CO2 could potentially result in perturbations of climatic patterns. In addition, evidence of cyclical patterns exists in historical climate records. Provided these changes are not offsetting, they may manifest themselves as nonstationarities in the statistical parameters of hydrologic time series. The objective of this study was to investigate the assumption of stationarity in long-term peak annual streamflow records across British Columbia. Twenty-five peak annual streamflow records were grouped into five delineated regions. Nonparametric statistical tests were applied to each streamflow series within the region. Each region was then 'collapsed' into a single figure representing the test parameter for that region. These parameters were tested for trends and jumps in location, dispersion and overall distribution characteristics of each region. Finally, the regional results were contrasted against various Monte Carlo simulations to evaluate the probability that the magnitude of nonstationarity signified by each regional parameter could have resulted from random climatic variation. Evidence of drastic temporal decreases in mean peak annual floods suggested that stationarity is not a valid assumption for gauging stations located in the designated "East" region of British Columbia. A similar conclusion can be inferred for the region designated as "Mid North," where all flood frequency distribution parameters exhibit notable magnitudes of nonstationarity; Other regions showed inadequate evidence or do not contain enough information to dismiss the assumption of stationarity.