120 likes | 273 Views
Assessment of apparent non-stationarity in hydroclimatic series: A case study from Western Australia. Climate Adaptation. Bryson Bates (CSIRO, Australia) Richard Chandler (UCL, UK) Steve Charles & Eddy Campbell (CSIRO). Background. Study region: southwest corner of Australian mainland
E N D
Assessment of apparent non-stationarity in hydroclimatic series: A case study from Western Australia Climate Adaptation Bryson Bates (CSIRO, Australia) Richard Chandler (UCL, UK) Steve Charles & Eddy Campbell (CSIRO)
Background • Study region: • southwest corner of Australian mainland • Mediterranean-like climate – hot dry summers; cool wet winters • ~ 80% rainfall occurs in winter half-year (May-October) • low orography (max height 582 m above mean sea level) • Integrated Water Supply System supplies water for 1.6 million in Perth & surrounding areas • Previous research findings (IOCI): • declines in number of wet days & extreme amounts; confined mainly to May-July • increase in MSLP in winter & decline in atmospheric moisture in winter & spring important in explaining declines • reduction in intensity of cyclogenesis across southern Australia (esp. SWA) • Infrastructure investment: $A921 million invested in source development 1996-2006; another $A1 billion after 2006 • Our approach: application of a suite of modern, model-based statistical approaches to build explicit representation of changes in surface water availability and associated climatic drivers 11th IMSC Edinburgh 12/7/2010
Data Flow Diagram Trends & Explanatory Covariates (SLR & GLM) At-Site Daily Rainfall Annual Inflow Atmospheric Predictors Weather State Probability Series (NHMM) Trends or Change Points (LLR) Key Atmospheric Predictors (ANOVA) Period of interest: 1958 to 2007 11th IMSC Edinburgh 12/7/2010
Transformed Annual Inflow Series 11th IMSC Edinburgh 12/7/2010
Significance Traces 11th IMSC Edinburgh 12/7/2010
Sub-region used for Rainfall Analysis Dots indicate main IWSS dams 11th IMSC Edinburgh 12/7/2010
Results from Generalised Linear Modelling • Trend in log mean wet-day rainfall amounts per decade: • –0.0289 – (0.0222 x Lat1) – (0.0498 x Long1) • where Lat1 and Long1 first-order Legendre polynomial • transformations of latitude & longitude • Interpretation: rainfall decline strongest in the north and east • Trend in daily rainfall occurrence: • Interpretation: decline strongest in the west at low altitudes 11th IMSC Edinburgh 12/7/2010
Changes in Atmospheric Circulation 11th IMSC Edinburgh 12/7/2010
Rainfall Occurrence & Synoptic Patterns (NHMM) 11th IMSC Edinburgh 12/7/2010
Weather State Probability Series May-June-July 'Dry' 'Mixed' 'Wet' Little evidence of changes in persistence 11th IMSC Edinburgh 12/7/2010
|b | Sensitivity of Weather State Probability Trends • Factorial experiment • fwd and bwd reordering of atmospheric predictor series for NHMM (intraseasonal structure preserved) • reversal of sequence for dominant factor, reversal in sign of trend • 24 factor combinations • 3 NHMM simulations per combination • compare differences in frequencies for 1st & last 10 years • Linear model coefficients represent effects of factors individually & in combination Influential predictors 11th IMSC Edinburgh 12/7/2010
Conclusions • Little evidence of discontinuity in the mean of annual dam inflow series (smooth, nonlinear declining trend) • Substantial regionally-varying declines in daily rainfall occurrence & wet-day amounts in vicinity of dam catchments • Increasing/decreasing trends in probabilities of 'dry'/'wet' synoptic types for MJJ (bleak short-term prognosis) – little evidence of changes in persistence • Increase in MSLP & DTD in MJJA; decrease in PG in MJJ – all important & all favour drier conditions • Temporal orderings of MSLP, PG and DTD have demonstrable impact on trends in the weather states probability series – trends driven by individual factors & not their interactions • Tangible benefit in using a multifaceted approach to the study of the nature and drivers of non-stationarity in hydrologic series 11th IMSC Edinburgh 12/7/2010