Spatial and temporal patterns in the frequency of events exceeding three times the median flow (FRE3) across New Zealand

• Published in: Journal of hydrology, New Zealand New Zealand, 2013-01, Vol.52 (1), p.15-39
• Main Author: Booker, Doug J.
• Format: Article
• Language: English
• Subjects: Autocorrelation; Climatic changes; Ecohydrology; Ecology; Environmental aspects; Floods; Forest hydrology; Freshwater ecology; Hydrogeology; Hydrology; Islands; Mathematical models; Regression; Statistical median; Statistical variance; Stream measurements; Time series; Trends; Watersheds; Windows (intervals)
• ISSN: 0022-1708; 2463-3933

FRE3 has been described as the average number of high flow events (freshes or floods) per year that exceed three times the median flow. It has previously been used as an index of flow-driven disturbance in ecological studies of in-stream biota such as periphyton, macrophytes, invertebrates and fish. Although the above description is expedient, clear instructions describing calculations are required for methods to be fully replicable. Instructions are provided for a replicable method for calculating FRE3 that deals with gaps in the record and optionally not counting events that occur within a specified time window of previous events. Annual series of FRE3 were calculated for many sites throughout New Zealand to assess the interannual variability and stationarity of FRE3. Shapiro-Wilk tests of normality showed that although the majority of inter-annual distributions were normally distributed, some could not be considered to be normal. Significant partial autocorrelations were found in several annual FRE3 series, suggesting the presence of repeating patterns. Trends were identified within annual FRE3 time-series covering an arbitrary 30-year period (1980-2009) using Mann-Kendall tests for trend in non-parametric data. During this period, flow time-series from the South Island of New Zealand exhibited downward trends in FRE3, whereas time-series from the North Island exhibited both upward and downward trends. This indicates that FRE3 cannot be considered to be stationary, implying that calculations of FRE3 should be applied over as long a time-series as possible when aiming to characterise FRE3 at a site. A random forest regression model was fitted to predict FRE3 across New Zealand from available catchment characteristics.