Water Journal : Water Journal April 2012
catchment management technical features 120 APRIL 2012 water flows and commencement of water transfers can be seen in the flow exceedance curves (see Figure 3). Generalised Linear (GLM) and Generalised Additive (GAM) models for trend Since it was clear that variation in flow affected water quality in the Hawkesbury- Nepean River, a series of base statistical models were developed that sought to relate variation in water quality parameters at various sites over time, after allowing for variation in flow. These base models also included terms that attempted to allow for seasonality in the response variable. The GLM used for trend assessment had the general mathematical form: Y = µ+ α*logfow + β*logfow_lag1 + δ*loginfow + γ*time + η*cos_time + φ*sin_time + ε Where logflow was the log10-transformed flow at Penrith Weir on the day of sampling; logflow_lag1 was the log10-transformed flow at Penrith Weir on the previous day; loginflow was the log10-transformed flow at the closest tributary on the day of sampling; cos_time and sin_time were seasonal components; and time was a linear data series (in years) from the start of the data record. The error structure (ε) was assumed to be normally distributed with 0 mean. Models were fitted using the GLM procedure in the SAS statistical software (Enterprise Guide V4.1; SAS 2006). The GAM used for trend assessment had the general mathematical form: Y = μ+ α*s(logfow) + β*s(logfow_lag1) + δ*s(loginfow) + γ*s(time) + η*cos_time + φ*sin_time + ε Where s(logflow) was a non-linear smoothed term for log10-transformed flow at Penrith Weir on the day of sampling; s(logflow_lag1) was a non-linear smoothed term for log10-transformed flow at Penrith Weir on the previous day; s(loginflow) was a non-linear smoothed term for log10- transformed flow on the day of sampling at the closest tributary; cos_time and sin_time were seasonal components; and s(time) was a non-linear smoothed term for time (in days) from the start of the data record. The error structure (ε) was assumed to be normally distributed with 0 mean. Models were fitted using the mgcv package (Wood, 2006) in R Version-2.5.1 (The R Foundation for Statistical Computing 2007). Trends in water quality were identified at a variety of sites using these statistical models (see DECC, 2009 for a full description). In general, the results of the GLM and GAM provided similar conclusions, although the GAM was preferred where trends through time were clearly non-linear. A comparison of predictions from the GLM and GAM models for chlorophyll-a levels at North Richmond is illustrated in Figure 4. Discussion Trends in flow Analysis of the Hawkesbury-Nepean flow data indicated that there has been a significant reduction in flow over Penrith Weir (and at many other gauging stations) in recent times, with current river flows being much less than the long-term average. This is a result of both climate variability (lower rainfall in recent time periods) and river regulation. The construction of dams in the Hawkesbury- Nepean catchment has had a major effect on river flows (Sammut and Erskine, 1995; DECC, 2009). At Penrith Weir this is particularly evident for medium flows and the smoothed trend line for flow at Penrith Weir now consistently falls below that of the unregulated Colo River for the first time since records began (DECC, 2009). It is also relevant to note that in recent years extremely low flows at Penrith Weir have also been eliminated, due to a requirement to ensure that a minimum flow of 50ML/day is maintained over Penrith Weir (NSW Government, 2006). The exact environmental flow rules that will apply to Warragamba Dam in the future are still under consideration. Upstream of the dams, the 2010 Audit generally identified a return to periods of higher rainfall than had been experienced in previous audit periods where drought conditions had prevailed over much of the last decade. This return to wetter conditions was reflected in higher stream flows in some, but not all, areas. Continued declines in flow compared to longer-term statistics were noticeable in some areas (e.g. Werriberri Creek) but there was insufficient time during the 2010 audit to relate observed changes and/or trends in flow back to varying rainfall patterns across the catchment. Some areas in the Hawkesbury-Nepean catchment are experiencing continued rainfall deficits when compared to the long-term rainfall records (e.g. Russell et al., 2010). Upstream of the dams, the influence of water transfers on flow statistics was particularly important, with sites in the Wingecarribee and Upper Nepean Rivers experiencing the greatest effects. The greatest increase in flows occurred in the Wingecarribee River at Sheepwash Bridge (downstream of Wingecarribee Dam), where the median flow during the 2010 audit period (July 2007--June 2010) was almost seven times its historic long-term median. Median flows had been even higher in the preceding three years (2004--2007). Large increases in flow due to water transfers have the potential to exert geomorphic stress on both the Wingecarribee and Upper Nepean River systems. Flows can also be affected by increased water extractions and these effects were also considered in the 2010 Audit (DECCW, 2010). The NSW Office of Water (NOW) has recently exhibited their Draft Water Sharing Plan for the Greater Metropolitan Region, which covers the Hawkesbury-Nepean River system (NOW, 2010). This Plan identified a significant existing allocation of water in the Hawkesbury-Nepean River catchment. It is notable that NOW has assessed most valleys to currently be at or close to their limit of sustainable water extraction (NOW, 2010). Water Quality Trends Since many water quality variables are significantly affected by flow, assessments of changes and/or trends in water quality necessarily need to consider variation in flow. This was achieved for sites downstream of the dam using generalised linear and generalised additive models (DECC, 2009). Upstream of the dams trend assessments were Figure 3. Flow exceedance curves at Penrith Weir (top) and Wingecarribee River at Sheepwash Bridge (bottom).
Water Journal May 2012
Water Journal December 2011