2017 - Mid Year Forcasts

As we now have temperatures for the first half of 2017 for all the data sets, I thought I might again try my hand at some simplistic statistical analysis - trying to forecast the final 2017 temperature. Last year I was trying to estimate the probability of a new record being set. This doesn't make much sense this year as there was never going to be a realistic chance of that this year. Therefore this year I'll be putting more emphasis on what the final total will mean for the ongoing warming.

This year, I'm using a slightly different, hopefully improved, model which takes into account the underlying trend as well as the temperatures for the year so far.. Using the same model as last year would be forecasting the satellite data as slightly cooler, by around 0.05 °C, and surface data as almost unchanged.

Will 2017 be a record warm year?

Short answer, no. Or at least I'd be very surprised if it is - and pretty worried. For all the satellite data it is all but certain 2017 won't be warmer than 2016. My model puts the chance as zero to two decimal places.

For surface temperature data the situation is a little more uncertain, as statistically I still give all sets a small chance, but I doubt even that small chance is realistic. GISSTemp is least likely to set a record, with only around a 2.5%, around 1 in 40. Other data sets have slightly more chance, but apart from HadCRUT all are under 7.5%. HadCRUT is the odd one out with a surprisingly high ~12.5%.

The usual caveats about these not being based on any climate models, and could be very different if a new La Niña, El Niño, or nuclear war develops.

The Graphs

Here are some of the forecasts in context. Each graph shows the annual temperature from 1979 to 2016 in blue, with the forecast for 2017 in red. The red circle shows the expected anomaly, whilst the red line shows the 95% prediction interval. That is there should be only a 1 in 20 chance that the actual value will be outside the red line.

Note the anomalies shown are all using the respective base period for each set, and the scales of each graph may not be identical.

UAH 6

I'll start with the UAH 6 satellite data. This is the data that will be used by anyone wanting to believe in a pause as it shows the least warming.

It is most likely UAH 6 will finish 4th warmest, but could be 5th or 3rd. It definitely won't be warmer than either of the strong El Niño years and will more likely than not be cooler than 2010. However, it's still looking like 2017 will be a warm year, and there seems little chance we will see a return to Monckton's Great Pause this year.

RSS 4

RSS has now updates their Lower Troposphere data to the new 4.0 version. This is also likely to show 2017 as 5th - 3rd warmest, but with more likelihood of being at the top of that range.

What both satellite sets suggest is that 1998 will have been far more of an outlier than 2016.

GISSTemp

Moving onto surface data, here's GISSTemp. In this set, 2017 is quite likely to be the second warmest, and almost certain to be at least the 3rd warmest.

HadCRUT

I'll finish with HadCRUT as it shows the most likelihood of setting a new record. There's a curious difference between the last two years here and in the GISS data, and this means that whilst HadCRUT has more chance of finishing as the warmest, it also has more chance of finishing 3rd. It's very unlikely to finish 2nd because the last two years were so close together.

For context here's HadCRUT since 1850>

Conclusion

Regardless of the ranking, it is clear that 2017 is on track, in all data sets, to be another very warm year. Those arguing the pause is real and only temporarily disappeared because of the 2016 El Niño may have to change their minds - though more likely they will just start claiming a pause started just before 2016.

Appendix

Here are the figures for all the predictions. This shows the forecast anomaly for each set in °C along with the upper and lower bounds for the 95% probability. As with the graphs, all anomalies are given using the respective data set's own baseline.

Data Set	Forecast	Lower	Upper
UAH 5.6	0.39	0.30	0.49
UAH 6.0	0.30	0.21	0.39
RSS 3.3	0.40	0.30	0.49
RSS 4.0	0.57	0.48	0.67
NOAA	0.88	0.79	0.96
GISSTemp	0.91	0.82	0.99
HadCRUT	0.72	0.63	0.81
BEST	0.90	0.81	0.99

And here's the data as an anomaly of the 1981 - 2010 base period.

Data Set	Forecast	Lower	Upper
UAH 5.6	0.39	0.30	0.49
UAH 6.0	0.30	0.21	0.39
RSS 3.3	0.30	0.21	0.40
RSS 4.0	0.44	0.35	0.54
NOAA	0.44	0.35	0.53
GISSTemp	0.48	0.39	0.56
HadCRUT	0.43	0.33	0.52
BEST	0.52	0.42	0.61