Hello. This is my first correspondence on this forum. Forgive me as I haven't used WIKI for quite a few years and need to re-educate myself.
I dabbled in some reanalysis data a few years ago and had little to do with it regularly until a few months ago. I am very impressed with the progress made, not that I knew much about it all beforehand, and warmly compliment all involved. The free access to this wonderful resource coupled with the opportunities freely available packages such as 'R' allow to one to 'play' with the data, make it a fun opportunity for enquiring minds to understand and educate others in this fascinating subject.
The main reason I drifted back to this resource was because of the paper 'Evidence linking Artic amplification to extreme weather in mid-latitudes' by Jennifer Francis and Stephen Vavrus. They used the reanalysis data in their paper.
In locating this URL I stumbled across this piece that appeared on the Internet just today, so the research is understandably worthy of attention given what is going on with the world's weather.
However I would strongly urge anybody interested in this important issue to visit the page:
You can read the narrative, watch the episode and even download a copy. It's a good summary of what issues are importantly at play with regards to climate change and Francis's research is considered at the end of it.
The reason for my correspondence here deals with the fact that I downloaded some reanalysis data to have a look at the issue myself. Unfortunately I stumbled across some data I was uncomfortable with. For various reasons I'm not the most motivated person when it comes to writing up my calculations and my short private correspondence on the matter died. I then found this website, so I thought I might try this avenue if you don't mind. I'll try and keep it to a minimum and make the data and 'R' code I have used available to others to re-produce, correct and add to, if there is an interest. If done correctly and in the right spirit it might be a good opportunity to help educate some folks who are starting out on a journey to explore the endless opportunities that exit with this resource. Especially as the data set deals with the world and we can all learn from others experiences with the circumstances elsewhere.
As a starter, I downloaded annual daily temperatures for the world [144 x 73 grid points] from the US reanalysis I data set for the years 1980-2012. From this, I used 'R' to compute the annual average values for each grid point and I have saved into a netCDF3 file which I will gladly make a link to, if anybody wants a copy to look into this data themselves and doesn't want to re-do what has already been done. The first thing I was interested in was to derive a simple linear regression between the annual temperatures and the year for all grid points, to gauge what % of grid points are rising and falling. I decided to plot the regression slope against the correlation so that I could get an idea of the size of the trend and the strength of the trend (as measured by the correlation). The plot I obtained was:
I was rather taken by the plot. The majority of the points lay in the top right of the plot signifying a rising trend in the values, which is what I expected, but the size of some of these and particularly a few of the falls, are surely too large. The annual slopes range from -0.2852 to 0.2089. When multiplied by 32 to account for the 32 year period of the data, this comes to -9.1 to +6.7 deg changes over just 1/3 of a century. Further exploration is in order.
I won't say anything more for the moment but have obtained EC and US2 data sets for the area where the temperature falls are notably contrary to the rest of the data set. These apply to the mountainous region of South America. The most significant decline here occurs at the grid point 70W, 30S and the figure below shows how it arose.
In concluding this contribution, I've listed the values below for anybody who wishes to have a 'play'. If you average the values from 1980-1998, it means 9.92 with a standard deviation of just 0.47; its very consistent. For the period 1999-2012 the average drops to 3.96 with a standard deviation of 1.43. A t-test on these 2 series, assuming an unequal variance, gives a t-value of 15 and a 95% CI of the difference as [5.11,6.81]. Clearly there is something at play here and it isn't climate related. I will discuss this further in subsequent correspondence if it is OK to do so.
I would also like to point out that I'm also dabbling in some 'perfect prog' calculations on rainfall probabilities in Australia using the reanalysis data set. The results very nicely connect the indicence of rain [we don't get snow here much...] with basic atmospheric conditions and the data analysis would be of some interest to those folk who have ventured to this wonderful resource and would like to get a feel for what absorbing things they could do with it.