- Home Antje Moffat http://www.bgc-jena.mpg.de/~antje.moffat/ Gap Filling Comparison - Home - More Details (password protected) - Forum (password protected) More Details - Gap Filling Comparison Discussion Forum http://www.bgc-jena.mpg.de/bgc-systems/pmwiki/pmwiki.php/GapFilling/GapFillingHome The seven golden sites These forest sites and years have been chosen for their very good density of real data and describe a variety of species and climates: Prefix Sitename Country Species Type Climate Years BE1 Vielsalm Belgium Fagus sylv./ Pseudotsuga menz. dbf/enf temperate/continental 2000/2001 DE3 Hainich Germany Fagus sylvatica dbf temperate/continental 2000/2001 FI1 Hyytiala Finland Pinus sylvestris enf temperate/continental 2001/2002 FR1 Hesse France Pinus sylvestris enf temperate/suboceanic 2001/2002 FR4 Puechabon France Quercus ilex ebf mediterranean 2002 IL1 Yatir Forest Israel Pinus halepensis enf semi arid 2002 IT3 Roccarespampani Italy Quercus cerris dbf mediterranean 2001/2002 The provided NEE (= flux plus storage of CO2) is ustar filtered and quality checked for low variability and spikes. Furthermore the following meteorological variables are provided: Name Description Measurement Unit NEE FC+SC carbon dioxide umol m -2 s-1 Time time Julian Day ustar friction velocity m s-1 Rg global radiation W m-2 PPFD photosynthetic photon flux density umol m-2 s-1 Ta air temperature °C Ts soil temperature medium °C SWC soil water content medium %vol Rh relative humidity % precip precipitation mm WS Wind horizontal speed m sec-1 WD Wind direction degrees -9999 is used as a gap filler Filename NEE ustar Rg PPDF Ta Ts SWC Rh precip WS WD be1_2000checked.txt 72 97 94 95 95 95 95 95 88 97 97 be1_2001checked.txt 71 97 96 96 96 96 96 96 96 97 97 de3_2000checked.txt 66 86 100 100 100 99 100 99 100 98 98 de3_2001checked.txt 67 92 100 100 100 100 100 98 100 97 97 fi1_2001checked.txt 65 86 87 88 100 99 22 98 89 100 100 fi1_2002checked.txt 66 88 96 96 100 98 19 96 97 99 100 fr1_2001checked.txt 79 94 100 100 100 100 0 0 100 100 97 fr1_2002checked.txt 78 96 100 100 100 100 0 0 100 100 97 fr4_2002checked.txt 64 92 99 100 100 0 0 100 100 100 100 il1_2002checked.txt 53 88 92 92 87 88 0 70 19 92 92 it3_2001checked.txt 55 97 73 73 99 100 91 91 29 98 98 it3_2002checked.txt 69 95 98 98 95 98 96 51 100 95 95 Please upload the files here: gap_filling_comp1.zip. For more information on the sites, please check http://gaia.agraria.unitus.it/database/. Key File Description Superimposition scheme Real data: flag_real=0 for real data and flag_real=1 for real gaps. Artifical key: flag_artificial=0 for the data to use and flag_artificial=1 for artificial gaps. REAL ARTIFICIAL 0 0 real data -> use for gap filling procedure 0 1 artificial gap -> fill 1 0 real gap -> ignore 1 1 real gap (and artificial gap) -> ignore Types and number of gaps The key file will have different keys for 4 different TYPES of gaps plus 1 mixed version. The count of gaps will equal about 10% of the half-hourly values (hh) TYPE LENGTH COUNT AMOUNT OF HH v: very short term = 1 hh = 1752 gaps = 1752 s: short term = 8 hh (4 hours) = 219 gaps = 1752 m: medium term = 64 hh (~1.5 days) = 27 gaps = 1728 l: long term = 576 hh (12 full days) = 3 gaps = 1728 mixed: mix of the above: 400 vs, 50 s, 6 m, 1 long = ~1760 Ten different samples of random scenarios are taken per gap type. Randomness The starting point of each gap is chosen randomly out of the whole set of half-hourly values. Then all the half-hours of the gap are marked as a gap. Furthermore the same amount of half-hours are blocked off after and twice the amount of half-hours are blocked off before to avoid overlapping. For the long term gap of 12 days the starting day is chosen and four times this period is blocked off to distribute the long gaps over the whole year. For the mixed version, all starting points are chosen randomly and no blocking before and after. The gaps could potentially overlap and the total amount of half-hourly values of gaps might here be a bit less. In the mixed version the long term gaps among the different scenarios are not independent, again to make sure that the gaps are distributed evenly over the year. Here the period is blocked off only once. Key file text The produced key file key_file.txt (space delimitered) looks like this (v0 - very short term scenario 0, ...) hh v0 v1 v2... s0 s1 ... 1 0 0 0 0 0 2 0 0 1 0 0 3 1 0 0 0 1 4 0 0 0 0 1 5 0 0 0 0 1 6 0 0 0 0 1 The total number of different sets is 50. This means there will be 50 times 10 golden files => 500 files to fill. Upload Upload the key file here: keyfile.txt and keyfile.xls. Timeline First submission of gap filled results Deadline 31st of July 2005!!!! Comparison of results Performance: short, medium, long, mixed gaps Statistical Analysis: RMSE, R-squared, ratio of variance, bias Sums: Daily and yearly sum with uncertainties 1st Draft of the paper 15th of September --> 2nd Draft of the paper --> Submission Webmaster: Silvana Schott updates by: Antje Moffat last update: 30 May 2005