tag:blogger.com,1999:blog-7729093380675162051.post6160036402589548739..comments2024-03-28T13:56:47.604+11:00Comments on moyhu: New FEM/LOESS method of integrating temperature anomalies on the globeNick Stokeshttp://www.blogger.com/profile/06377413236983002873noreply@blogger.comBlogger6125tag:blogger.com,1999:blog-7729093380675162051.post-5863727346268252052019-10-17T20:21:07.443+11:002019-10-17T20:21:07.443+11:00Thanks, Clive,
I have just put up a new post which...Thanks, Clive,<br />I have just put up a new post which sets out the maths more consistently. It also does comparisons. My mesh method took about an hour doing all meshing as new; the high resolution methods of FEM/LOESS (eg h4p4) take about five minutes.<br /><br />The dependent variables are the multipliers of each shape function. There are a few underlying FEM tricks here. The basis functions are such that they are 1 on their node, and zero on all the other nodes. So f(z)=ΣaᵢBᵢ(z) is the unique element-wise polynomial which takes values aᵢ at the node locations zᵢ. It is also continuous. In the mesh method, these are the piecewise linear approximants.<br /><br />So it is basically a regression to find which aᵢ bring f(z) to be least squares closest to the datapoints y. Then integrate f(z). <br /><br />There are a whole lot of numbers for doing this which I can post if anyone is interested eg integrals of shape functions.<br /><br />Nick Stokeshttps://www.blogger.com/profile/06377413236983002873noreply@blogger.comtag:blogger.com,1999:blog-7729093380675162051.post-3508429828798292512019-10-17T20:04:18.507+11:002019-10-17T20:04:18.507+11:00I am essentially using your MESH method, but recal...I am essentially using your MESH method, but recalculating the geometry for every monthly time step. It takes about 40mins to process all G4/HadSST3 from scratch which I think is not too bad. I also tried icosahedral grids but then simply averaged all values inside each cell to form an area weighted global average.<br /><br />I am not sure I understand your LOESS method. In order to do a least squares fit you assume a polynomial distribution , but what are the dependent variables? Are they distance from a node point ? Do you only fit measurements inside one cell or what ? <br /><br />Impressive though !Clive Besthttps://www.blogger.com/profile/10486120708699060846noreply@blogger.comtag:blogger.com,1999:blog-7729093380675162051.post-35873247724859034322019-10-17T17:51:51.537+11:002019-10-17T17:51:51.537+11:00Bindidon,
Yes, sorry, those files aren't work...Bindidon, <br />Yes, sorry, those files aren't working at the moment. But the data is in those tables. If you go to the <a href="https://moyhu.blogspot.com/p/latest-ice-and-temperature-data.html#L1" rel="nofollow">head table</a> on the data page, and click the TempLS tag, you'll get 8 different kinds of TempLS output back to start 2014. On the land/sst button, you'll get a further set of TempLS land and ocean. If you go to the active plot <a href="https://moyhu.blogspot.com/p/latest-ice-and-temperature-data.html#Drag" rel="nofollow">here</a>, show TempLSmesh (or sst, land) and click the show data button bottom right, it will bring up a tab with data starting 1900.<br /><br />I have put a file month.csv <a href="https://s3-us-west-1.amazonaws.com/www.moyhu.org/2019/10/month.csv" rel="nofollow">here</a>. It has about 28 datasets back to 1850 (up to date). I'll keep doing that each month.<br />Nick Stokeshttps://www.blogger.com/profile/06377413236983002873noreply@blogger.comtag:blogger.com,1999:blog-7729093380675162051.post-45107694512099298632019-10-17T07:56:41.430+11:002019-10-17T07:56:41.430+11:00Thanks Nick but you misunderstood me I guess.
I o...Thanks Nick but you misunderstood me I guess.<br /><br />I of course did not mean the bunch of major data; that I have all already, including HadISST1, GHCN daily and others.<br /><br />What I mean is a monthly time series<br />year month value<br />of what <b>you</b> produce each month (no: please no R files).<br /><br />When I click on TempLSgrid<br /><br />https://moyhu.blogspot.com/p/templsgrid.txt<br /><br />or on TempLSmesh<br /><br />https://moyhu.blogspot.com/p/templsmesh.txt<br /><br />I obtain<br /><br />"Sorry, the page you were looking for in this blog does not exist."<br /><br />instead of a txt output...<br /><br />Rgds<br />J.-P.<br />Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-7729093380675162051.post-64566481890897864152019-10-16T11:35:07.815+11:002019-10-16T11:35:07.815+11:00Bindidon,
On the regular monthly data table, there...Bindidon,<br />On the <a href="https://moyhu.blogspot.com/p/latest-ice-and-temperature-data.html#L1" rel="nofollow">regular monthly data table</a>, there is a button which switches to land/SST - that is HADSST etc and also .<br /><br />You can also go to the <a href="https://moyhu.blogspot.com/p/latest-ice-and-temperature-data.html#Drag" rel="nofollow">active graph</a> and choose just the datasets you want. Then the Data button will print out the data plotted in ascii.<br /><br />Another option is to download a file<br />https://s3-us-west-1.amazonaws.com/www.moyhu.org/2019/10/month.sav<br />This is an R file with 28 data sets; the basis for the above presentations. Just change the 2019/10 to latest month. I keep copies like that so one can go back and see changes. I'll probably switch to csv format.<br />Nick Stokeshttps://www.blogger.com/profile/06377413236983002873noreply@blogger.comtag:blogger.com,1999:blog-7729093380675162051.post-89229232107246045162019-10-16T09:18:15.672+11:002019-10-16T09:18:15.672+11:00Hello Nick, you have posted inbetween so much info...Hello Nick, you have posted inbetween so much info that it becomes hardly possible to recall where is what.<br /><br />Do you have a corner where you post each month a few monthly time series (land-ocean, land, ocean) ?<br /><br />That would be great help for folks who want to compare what they do with your more professional results!<br /><br />Thx<br />J.-P. alias Bindidon<br />Anonymousnoreply@blogger.com