2020/03/18 Replace NCEP GTS with ICOADS R3.0.2 after (including) March 2020 Change NetCDF3 into NetCDF4 (2008-present) Implemented in operational production starting on April 3, 2020. 2020/03/13 Revise gtsqc.f by adding if(kk.ne.1.and.kk.ne.2) go to 800 to skip cman data starting from 2020/01 2019/06/06: hfsst.f in code_hist and code_hist.1985 line 193: open(27,status=st,form=fm, file='../data/gts2d/argo.avg2.mon.'//cyr//'.dat') needs to be revised in future update, which uses buoy rather than argo obs numbers. The impact to final SSTA is small. 2018/10/05: nmo=130 is corrected to nmo=140 in subroutine intts in hfsst.f and sst2d.f 2018/03/15: oiv2.ano2.1982.2011 was generated and was used in ersstv5 historic run but oiv2.ano2.1982.2002 is used (should not be the case) in operational run this should be updated whenever possible. 2018/02/09: Add iod calculation 2018/02/04: Bug fixing in argoqc.f by defining mon=imonth-1, when imonth=1, mon=12 2017/04/17: Switch lf95 compiler to gfortran compiler: No visibile difference but random noise at order of 1.e-8 2017/02/24: hfsst.f: e1(3) revised as e1(3)**2, although tests show little difference in 0.001C in RMSD 2017/02/23: QC and data ingest start from the current year in gtsqc.f and argoqc.f therefore the data in previous years will not be changed due to recurring FG. NC output remains starting from 2001. 2016/11/14: Remove Fortran NetCDF writing (netcdf.f) and reading (in regional_average.f) add GrADS NetCDF writing. We have migrated ERSST production to a new computer system starting from February 3, 2017. The migration is fine but the NetCDF writing became a problem. Previously, FORTRAN code is used and integer format data are written in NC files. But this does not work or I have to spend long time to fix it. In stead, I am using GrADS and ncks to write the NC files and real format data are written in NC files. Therefore the data format has been changed for all NC files from 1854, otherwise people will not be able to read the files from 1854 to present. The data themselves should not be changed beyond rounding off between real and integer format. Let me know if this is not the case. The data may be changed in the last 3 months because of a 3-month filer applied in ERSST. For example, for a specific location in Nov, Dec, and Jan, observed data are 10, xx (missing), and 12 degrees. In December calendar month update, xx is filled with 10 because the January data has not yet available, while xx is filled with 11 [(10+12)/2] in January update. ERSST is run from 1854 to 2014, rerun every month from January 2001 to the present month (January 2017), and data after 2008 are updated in ftp site. Therefore data may have some changes since the first guess of the previous run is used. But the difference should not be large 2016/09/20: Testing ERSSTv5 setting e1(3) and wt1(3) need to be implemented in HF-SST 2016/09/01: Historic (1854-2015) ERSST reconstruction has been done. ==================================================================================== v4 ==================================================================================== ==================================================================================== 2015/11/17: add unadj.f to calculated unadjusted SST for uncertainty estimation replace nys=2012 by nys=2015 in gtsqc.f 2015/11/17: add ersst.cp.sh at the end of ersst.v4.sh. This is to copy data for the uncertainty estimation in ersst-prod. 2015/10/20: change ftp to sftp using specified credential: sftp -o IdentityFile=/home/boyin.huang/.ssh/id_ersst ersstsftp@ftp0 add chmod 644 in ersst.ftp.sh 2015/09/24: change user name to boyin.huang 2015/06/30: subroutine intrp1, goto 200 is revised into goto 10 2015/05/26: data are set to be official v4 release. 2015/04/02: a bug fix in nino_sst.f using SST rather than SSTA 2014/12/29: a bug in ersst_netcdf.f fixed: iday_num = iw3jdn(iy,mon,ida)-IW3JDN(IYRref,IMONref,IDAYref) is correct iday_num = iw3jdn(iydo,mondo,ida)-IW3JDN(IYRref,IMONref,IDAYref) was wrong 2014/10/18: NetCDF global attributions revised. 2014/10/03: ERSST.v4 Beta Version data release to public ftp site 2014/10/01: codes are revised based on ERSST.v4 Code Review 2014/07/30: put_cdf_ersst.f has been revised so that the standardized error variance is removed in the CDF file. The metadata convention has been upgraded from CF-1.0 to CF-1.6 2013/07/25: add parameter selections in ersst.v4.sh add ersst.v4.cron.sh, delete oiv2mon1t2.f 2013/03/04: remove oisst procedure, the data can be referred to /home/bhuang/gpfs/oisst 2013/02/07: replace lfsst.f by lfsst.space.f, lfsst.fill.f, lfsst.time.f zero-filling is replaced by near-by filling in both operational and historic runs rerun in 02/25/2013 to confirm data/warmcore.hadice result 2013/01/22: Codes are updated to /home/bhuang/ersst.test/b.ann.eot.lag1.std.damp0 exception: 2a.lfsst.space.f, 2b.lfsst.fill.f, 2c.lfsst.time.f minor corrections: nino_anm.f and nino_sst.f: add cj weighting dati2.upd.f: add filter to missing value in sqrt(e2) 2012/10/26: SST STD revision: stdev1d.oisst.fill.1982-2011 from oisst TO replace stdev1d-coads3-fill that may be calculated from patch coads data: the difference is very large 2012/10/25: lag-1 autocorrelation updated using weighting coefficients of 1854-2011 the correlation is higher than old one, but similar if 1971-2011 coefficients are used 2012/08/30: code clean up with some standard 2012/08/29: change clim.rcntr.1971.2000.dat into clim.even.1971.2000.dat replace clim.71.00.gdat-fill.dat by clim.even.1971.2000.fill.dat constructed from clim.even.1971.2000.dat since clim.rcntr.1971.2000.dat is no longer the latest. the pdo change is about -0.175 to -0.095 smaller in comparison with pdo magnitude of -3.5 to 3.5. 2012/08/28: Bug found in ssta.merg.f in calculate merged observation number c cs(i,j)=cs(i,j)+cb(i,j)*wt1(2) !byh: bug has includded additional 6.8 x buoy observation number. impacts: buoy ssta weighting may have been exaggerated. 2012/08/27: change SSTA calculation: original gtsqc.f: calculate SST, original ssta.merg.f: calculate and merge SSTA new gtsqc.f: calculate SSTA, new ssta.merg.f: merge SSTA The difference: SSTA is more accurate 2012/08/01: Bias correction file is replaced by bias.annual.fit.2x2.dat using new ICOADS R2.5 SST and MOHMAT NMAT. 2012/03/01: Ingest SST from ICOADS R2.5 2012/09/13: clean and standardize the codes in code_hist/ and code/ replace bias correction of v3b (1854-1941) by new bias correction (1855-2006) in code/ and code_hist/ 2012/08/27: generate ship.anm2.mon.yyyy.dat in data/gts2d generate buoy.anm2.mon.yyyy.dat in data/gts2d 0.gtsqc.f revise 1.ssta.merg.f to merge SSTA 2011/09/14: Rerun sea ice correction using seasonal (Jan-Dec) and coverage (0.0-1.0 per 0.1 increase) parameters using code_hist/ice/adj.ncep_ice2.mon.cyc.f and adj.gsfc_ice2.mon.cyc.f (whose data are preempted by HadIce). The total ice coverage change is visible between Jan-Aug 2011 due to changes in ice correction. Overall, NCEP ice is reduced due to correction. The correction largely happens in the ice margin areas. The SST increases 0.8C south of Greenland due to change of ice correction. 2011/09/12: manually run ersst4 from Jan-Jul of 2011 to control "date" paramter reactivate ftp nq files for operational run set cron job for ersst4 compare result of 2001-2010 between historic run and operational run test ice correction for NCEP ice 2011/08/30: prepare for operational run for ERSST4 nq files should not be ftp-ed from NCEP to avoid over-written NCEP ice need to be corrected historic run from 1854-2010 operationa run is set from 2001-present, need more than 7 yr for LF component year parameters need to changed: 1978 => 1994, LF starting point for LF 1985 => 2001, starting operational 1969 => 1985, for initializing LF 1978-1984 => 1994-2000 in HF 1970-1984 => 1986-2000 in txt note need to change file names of output. need to change names and contents of their ctl files The last 10-yr SST data from historic run are read and re-written as sst2d.2001.last.dat, which is used for operational run starting from 2001. /home/bhuang/ersst_v4/code_hist/first.guess.f 2011/08/23: Run sampling error estimate Run standard output routines for NetCDF, ASCII, PDO, NINO 2011/07/19: Comparison between SSTs using Had ice and Had(GSFC+NCEP) ice Comparison between ersst4 and ersst v3b 2011/06/27: HadISST downloaded, ice (1870-2010) used for ERSST4, and SST used for test global integrated icea area s fine run historic ERSST4 from 1854-2010 at /home/bhuang/ersst4/code_hist with step 0-4b, save it at warmcore.hadice/ 2011/06/14: Had+GFSC+NCEP ice copied from ERSST v3b operational notice a jump in global integrated icea area between GFSC (NCEP) and Had (1980) run historic ERSST4 from 1854-2010 at /home/bhuang/ersst4/code_hist with step 0-4a, save it at warmcore.3ice/ to simplify and speed up the historic run, ERSST v3b SST data are copied and used as an initial first guess for GTS QC 2011/06/09: ICOADS GTS data translated into NCEP abbreviated format (not exactly same) comparison in observation number