tag:blogger.com,1999:blog-9093436161326155359.post6364276326543636280..comments2024-03-28T06:43:02.954+00:00Comments on Variable Variability: A real paper on the variability of the climateVictor Venemahttp://www.blogger.com/profile/02842816166712285801noreply@blogger.comBlogger5125tag:blogger.com,1999:blog-9093436161326155359.post-60517041216932931972013-01-23T22:38:42.939+00:002013-01-23T22:38:42.939+00:00Dear Ari Jokimäki. Thank you very much! Those are ...Dear Ari Jokimäki. Thank you very much! Those are promising titles. <br /><br />Knowing your blog, I should have mailed you before, :-) as expert for the scientific literature.<br /><br />Old literature is great, then you can see who cited them in the Web of Science.Victor Venemahttps://www.blogger.com/profile/02842816166712285801noreply@blogger.comtag:blogger.com,1999:blog-9093436161326155359.post-31784079040961024962013-01-23T22:35:04.261+00:002013-01-23T22:35:04.261+00:00@Zeke, thank you for the link, that fits perfectly...@Zeke, thank you for the link, that fits perfectly to this blog and it is a very good post <a href="http://rankexploits.com/musings/2013/a-defense-of-the-ncdc-and-of-basic-civility/" rel="nofollow">on the quality of the NOAA homogenisation algorithm</a>. (Unfortunately, Blogger does not automatically detect links.)<br /><br />It is hard to understand how people can defend the misbehaviour by Anthony Watts in the comments to your post. The public discourse in the US is completely out of whack. :-(<br /><br /><br />It would be interesting to see the difference between the homogenized and the inhomogeneous data. Unfortunately, the influence of inhomogeneities on the grid box averages will be largest where there are not many stations, but this is also where relative homogenization will fail to find many inhomogeneities. Thus you will probably not see the full effect.<br /><br />That little bit of pinching for some decades is also something I would love to understand. Maybe the variability of the grid box averages of the last decade is larger because the "recent" data of many stations is not yet in the global database? This leads to more uncertainty in the mean and possibly to inhomogeneities for the gridded fields for the last decade.<br /><br />I would personally prefer to perform such studies on weather variability first on the station data itself, before trying the more difficult case of gridded data.Victor Venemahttps://www.blogger.com/profile/02842816166712285801noreply@blogger.comtag:blogger.com,1999:blog-9093436161326155359.post-26263103277459613842013-01-23T21:28:53.433+00:002013-01-23T21:28:53.433+00:00Figure 4 in my memo does show some pinching with t...Figure 4 in my memo does show some pinching with the use of a very recent baseline. I'll have to look into the homogenization issue; it should be relatively easy to recreate the graphs in my paper using the homogenized GHCN data rather than the unhomogenized data, though I suspect that the results won't change that much.<br /><br />On a somewhat unrelated note, you might find this interesting: http://rankexploits.com/musings/2013/a-defense-of-the-ncdc-and-of-basic-civility/Zekehttps://www.blogger.com/profile/09757819498566612533noreply@blogger.comtag:blogger.com,1999:blog-9093436161326155359.post-79861727245227948972013-01-23T20:04:58.810+00:002013-01-23T20:04:58.810+00:00I mean Figure 9 of Hansen et al. (2012).
The mean...I mean Figure 9 of Hansen et al. (2012).<br /><br />The mean and the variance used to compute the anomalies are here derived from two climate normal periods (1951-1980 (left panel) and 1981-2010 (right)) and a 60-year period (1951-2010 (middle)). Not for for every decade separately.<br /><br />Maybe I am overreaction, but what I do not understand is why all widths of the distributions are the same if you normalise by the 1981-2010 period (right panel). <br /><br />If the change in variability was "real", a change in the local variability of the weather, the period used to compute the anomalies should not matter and the left panel should look like the right panel and the last period should still have the largest variability.<br /><br />If the increase in variability is due to computational problems suggested by you, Tamino and Wickenburg, the right panel should look like a mirrored version of the left panel and the last period should have the smallest variability. Wickenburgs pinch effect should work both ways.<br /><br />It is possible that these two phenomena exactly cancel each other out and that consequently almost all decades have about the same variability in the right panel, but that would be quite coincidental. <br /><br />Unfortunately Wickenburg did not write in his memo how large the changes were that he assumed to reproduce the left panel. That would be needed to see whether the assumption is reasonable. <br /><br />Especially as in praxis the changes would probably have to be even stronger: Wickenburg computes the influence of having various trend slopes and of have various variabilities separately. However, both the slopes and the variability are largest in high-latitudes. Thus seen in terms of Hansens anomaly time series, the difference in the trends will be less large and the distortion of the distribution thus smaller.<br /><br />Problems not discussed in the reviews I have read are that GISS data is not homogenized. This is especially important for the variability. And even if the station data was homogenized, the gridded product would still be inhomogeneous as stations move in and out of the dataset. <br /><br />For the global average temperature, it may not increase the error too much not to homogenize the data (and only to correct for urbanization), but for this study the data was analysed per grid box. On a grid box scale (250x250km), inhomogeneities will more often go in the same direction for a certain period and lead to biases in the trends and will add variability. I would be curious if this would change the result. If one were interested in variability, also the uncertainty in the grid average values should be subtracted.Victor Venemahttps://www.blogger.com/profile/02842816166712285801noreply@blogger.comtag:blogger.com,1999:blog-9093436161326155359.post-22609487998963614982013-01-23T19:30:54.757+00:002013-01-23T19:30:54.757+00:00Was the final figure really that confusing? It was...Was the final figure really that confusing? It was just intended to show the results where each decade shown has an anomaly calculation of the decade in question.Zekehttps://www.blogger.com/profile/09757819498566612533noreply@blogger.com