The coming zero snow depth season #4

One of a series; also see editions: #1, #2, #3.

To finish this series, it remains to translate the method we’ve developed across to Australia’s best-known snow depth record at Spencers Creek near Charlotte Pass, New South Wales (midway between Perisher Valley and Thredbo; data from Snowy Hydro Limited). At about 1830 m elevation, Spencers Creek is Australia’s highest snow depth record, so zero depth there may well mean no natural snow cover anywhere in Australia.

Unfortunately the Bhend et al¹ (2012) report we’ve been using makes no predictions for NSW (it’s a Victorian study), and the older Hennessy report² (2003) is rather dated and doesn’t provide the necessary detail. Instead we need to resort to correlation, which fortunately is quite strong between Spencers Creek and Bhend et al’s prediction site at Rocky Valley Dam near Falls Creek, Victoria:



The correlation provides an approximate means of translating the Bhend et al predictions between the two sites.  Of course, inherent in that is the assumption that the past correlation continues to apply to future declining depths. I cannot say whether that is valid or not, but at least the Bhend et al prediction bands fall mostly within the range of historical variability (so little extrapolation is required):



It’s then simple to compute a Spencers Creek peak depth moving average³ as before and fit a second order trend⁴ to that and the midpoints of the translated Bhend et al prediction bands for A1B and A1FI⁵:



Once more, it’s striking how good the fit is, both to the moving average trace and the translated Bhend et al prediction bands. The trend hits zero in about 2100. That is for predictions using pretty much “business as usual” IPCC greenhouse gas emission scenarios (A1B and A1FI), consistent with the observed warming trend⁵. If there were to be substantial emissions reduction over the ensuing 86 years, the snowpack reduction would not be as severe.

As before, we can estimate when to expect the first-ever zero snow depth season (perhaps across the whole of Australia), by proceeding to compute the residuals from the trend (differences between raw data and trend), fit a distribution⁶ and extrapolate the resulting band of variability assuming that the residuals are 50% proportional to the trend⁷:



And then integrate (add up) the intersection of that variability band with zero:



That’s about 50/50 by 2070 (56 years from now) and very likely by 2090 (note 8).




1. Climate change impacts on snow in Victoria. J Bhend, J Bathols and K Hennessy, Centre for Australian Weather and Climate Research (a CSIRO / Bureau of Meteorology partnership), 2012.

2. The impact of climate change on snow conditions in mainland Australia. K Hennessy, P Whetton, I Smith, J Bathols, M Hutchinson and J Sharples, CSIRO Atmospheric Research, August 2003.

3. The 20-year moving average, extended a little by progressively reducing the averaging interval to 10 years at the leading end (the dashed extension). The moving averages are plotted centred.

4. A second order polynomial, also called a quadratic or parabola, which is a parsimonious choice for a dataset with known acceleration. As before, the fit is to the extended moving average trace plus the mid points of both the A1B and the A1FI prediction bands at 2020 and also at 2050 (both are included to increase their weighting in the regression).

5. The IPCC SRES emission scenarios A1B (medium) and A1FI (high). These SRES scenarios were superseded for the most recent IPCC Fifth Assessment Report (AR5 WG1, 2013) by “representative concentration pathways”, with RCP6.0 (medium) and RCP8.5 (high) generally corresponding in intent.

6. A Pearson type III distribution, to allow for the slight positive skew in the residuals.

7. To allow for some heteroscedasticity, as previously argued should apply to this record.

8. As before, that is a “best estimate”; it doesn’t attempt to incorporate estimation uncertainty as opposed to natural variability.  The Bhend et al estimation bar lengths are clearly meant to imply uncertainty in their report, but they seem to derive directly from the spread of the GCM outputs used to drive their model. It’s questionable whether that fairly represents the overall uncertainty in their analysis. (It might even overstate it, based on the extrapolation of the moving average trend to 2020.)

3 comments to The coming zero snow depth season #4