Stern et al. (2011) explored how new technologies might be harnessed to integrate material from various sources on the web to generate new products. Berners-Lee (2010) nominates linked data as “a great example of (the web's) future promise (and suggests that) today's web is quite effective at helping people publish and discover documents (but that) our computer programs cannot read or manipulate the actual data within those documents". He observes that “as this problem is solved, the web will become much more useful, because data about nearly every aspect of our lives are being created at an astonishing rate (and that) locked within all these data is knowledge about how to cure diseases, foster business value and govern our world more effectively". He notes that “scientists are actually at the forefront of some of the largest efforts to put linked data on the web”.

It is the primary goal of the work presented here to develop within computer programs the aforementioned “capacity to read and manipulate the actual data within web documents ”. By this means, the value of the data is fully realised via the automatic generation of a broad range of more accurate weather and climate forecasts and other products. The specific purpose of the present paper is to provide an update on the work described by Stern et al. (2011), in particular that work dedicated to developing a seamless (across time scales - Day 1-14, monthly, seasonal, decadal) weather and climate forecasting framework.

A "real time" trial of a methodology utilised to generate Day-1 to Day-7 forecasts, by mechanically integrating (that is, combining) judgmental (human) and automated predictions, has been ongoing since 20 August 2005. The verification data demonstrate how the combining process increases the accuracy of the forecasts over that of the official predictions – temperature predictions in 66 of the 70 months; rainfall predictions in 57 of the 70 months, albeit to a lesser extent during the exceptionally wet period between mid-2010 and early-2011. The data also demonstrate that the combining process leads to overall improvement at all lead times between Day 1 and Day-7. The data reflect the greater potential for an increase in accuracy in temperature forecasts during the summer months (when the variability of temperature is larger) than during the winter months. The data also reflect the greater potential for an increase in accuracy in rainfall forecasts during the summer months (when the variability of rainfall is larger) than during the winter months.

Since 20 August 2006, forecasts have also been generated for beyond Day-7 (out to Day-10). Since 18 January 2009, forecasts have also been generated out to Day-14. Regarding the verification data for the Melbourne experimental Day 1-14 forecasts, positive correlation coefficients between forecast and observed minimum and maximum temperature, amount and probability of precipitation, indicate the presence of some worthwhile skill out to Day-10 for all four forecast elements - (but also with some limited skill beyond). Accompanying the Day 1-14 forecasts are both a monthly and a seasonal climate outlook. Preliminary correlation coefficients (forecasts vs observed over the past two years) are +0.23 (rainfall), +0.07 (min temp) and +0.11 (max temp).

Regarding future development, worthy of consideration is a seamless (across time scales) framework whereby predictions are generated for key population centres (e.g. the eight State and Territory capitals, Adelaide, Brisbane, Canberra, Darwin, Hobart, Melbourne, Perth, Sydney, plus Broome - representing northern Western Australia, Alice Springs - representing Central Australia, and Cairns - representing northern Queensland); on a regional forecast district basis; and, on a State by State (& Territory) basis.