August 2008 Update on Global Temperature – NCDC
Posted by The Diatribe Guy on August 28, 2008
I haven’t posted in a little while, as other things in life have swamped my time (both for good and bad – got some R&R in there, as well). The baby’s doing well, we’ve bought a new camper, we’re trying to get in some summer activities before the fall, my fantasy football drafts are coming up… it just goes on and on.
But, I’ve slowly but surely tackled the NCDC spreadsheets. The reason it took me so long to get comments up is because I have once again refined the methodology to be even more robust, I’ve done additional parameter testing (which will be an ongoing issue for a while) and I really like the way it’s shaping up. I’ll give a little update here, but more will come on that in the future.
I just haven’t had the time to update all the charts yet, but let me throw out my general observations:
The July anomaly for NCDC, found here was 49.19. This actually bumped up the 12-month average anomaly a bit to 45.8. The anomlay was an oulier compared to the other data sets in that it was the only data set to show both an increase over the previous month (47.13) and over previous year July (46.54). While neither value is remarkably higher, the value – at least temporarily – dampened the measures that indicate cooling patterns. The cooling period still only extends back to January 2001, though one data point was added to the end with the new month. Also, the year-over-year run of cooler months was stopped.
Overall trend, January 1880 – current: Slope = 0.04267 (in terms of 0.01 degree Celsius), which corresponds to overall warming of +0.512 degrees per century.
Last 30 year trend: 0.137521, corresponding to 1.650 degrees per Century. This 30-year slope value has trended down contuously since the period ending December 2003, where it was 0.153552 (1.84 degrees per century). The current slope is the lowest since the period ending May 2002. Predicted anomaly based on the 30-year trend model for next month is 61.84, which would lower the slope to 0.13712. Based on my own observation, I doubt we’ll get there, so I anticipate a more significant drop in slope than that after next month’s results. Predicted slope a year from now looks to be 0.131750. The 30-year slopes move slowly, since it is a long trend line that is shifting.
Last 20 year trend: 0.156328, corresponding to 1.876 degrees per Century. This 20-year slope value has trended down with some intermittent peaks and valleys since the period ending March 2004, where it was 0.183568 (2.203 degrees per century). The current slope is the lowest since the period ending December 2006, so there hasn’t been a lot of movement in the last couple years either way. Predicted anomaly based on the 20-year trend model for next month is 57.22, which would lower the slope to 0.156147. Anything below that value will drop the slope and it looks like that is entirely possible. Predicted slope a year from now looks to be 0.156170, only slightly lower. This model is predicting fairly steady anomalies at this point, without much movement in the trend line. I haven’t vigorously parameter-tested any of the models above the 5-year model. That will be coming in the future.
Last 10 year trend: 0.118418, corresponding to 1.421 degrees per Century. This 10-year slope value has been fairly steady over the last couple years, but has increased from a trough in the period ending January 2008. The current slope is the highest since the period ending October 2006. Predicted anomaly based on the 20-year trend model for next month is 37.56, which would increase the slope to 0.118852. Predicted slope a year from now looks to be 0.097838, with slopes staying around current levels through May 2009, and then trending down after that.
As of now, the 60-month trends have been parameter tested most vigorously. The current 60-month slope is -0.161608, which corresponds to a cooling rate of 1.939 degrees Celsius per century. Nobody should extrapolate a five-year trend that long, of course, but it places the value in perspective relative to the warming trends indicated by the longer trend periods. Clearly, there is a short-term cooling trend evident if it is likewise argued that there are longer-term warming trends. The 5-year trend line has been heading in this direction since the period ending February 2004, where the peak positive value was 0.456552. Of special note is the last time the negative slope on the five-year trend has reached a value this low: January 1995. The 5-year model predicts an August anomaly of 55.97 and a slope after next month of -0.159326. The model actually predicts that the slope will increase to almost -0.10 over the next few months, peaking in March, followed by a fairly sharp downturn such that the slope a year from now will be -0.191542.
I’d have to get into model detail to explain the parameter testing I’ve done, but I’ll at least note that I’ve tested different ways of assessing changing slopes of fixed time lengths (at some point, I will vary these more as well). To date, the minimum least squares testing value corresponds with taking a look at how the 60-month tend line shifts, along with how the 14-month trend of 60-month slope values shift. From here, I have derived formulas to more easily populate and calculate and test projected anomalies, so that I can forecast them well into future periods.
This has no bias whatever introduced depending on personal belief about warming or cooling. It is entirely data-driven, and selections are made based only on which parameters generate a minimum least squares.
Of the 6 models using different base trend periods and assessing changes in trend, five of them forecast a long-term cooling of global temperatures, while one shows slight warming. The only model where I have significantly tested various lengths of times of slope trends is the 60-month model. The minimum least squares value is produced using a slope trend length of 14 months, and this model actually produces the most significant cooling scenario.
More testing needs to be done, however, and then I will want some stats freaks out there to peer review it to make sure I’m not doing something stupid.