Bias corrction SSEC偏置校正中國石化

1、1satellite winds superobbinghoward bergermary forsythejohn eyresean healyimagecourtesyof uw - cimsshurricane opaloctober 19952outlinebackground/problemsuperob methodologymethodobservation errorresultsconclusions/future workproblem: high - resolution satellite wind data sets showed negative impact (b

2、utterworth and ingleby, 2000) why?suspected that observations errors were spatially correlatedto account for this negative impact, wind data were/are thinned to 2 x 2 x 100 hpa boxes4bormann et al. (2002) compared wind data to co-located radiosondes showing statistically significant spatial error co

3、rrelations up to 800 km. correlationmet-7 w v nh correlationsgraphic frombormann et al.20025question:can we lower the data volume to reduce the effect of correlated error while making some use of the high-resolution data?6proposed solution:average the observation - background (innovations) within a

4、prescribed 3-d box to create a superobservation.7advantages:data volume is reduced to same resolution that resulted from thinning.averaging removes some of the random, uncorrelated error within the data.8superobbingmethod:91) sort observations into 2 x 2 x 100 hpa boxes.28 n16 w26 n18 w102) within e

5、ach box: average u and v component innovations, latitude, longitude and pressures.28 n26 n16 w18 w113) find observation that is closest to average position and add averaged innovation to thebackground value at that observation location.26 n28 n16 w18 w12superob observation error13superobbing removes

6、 some of the random observation error.this new error can be approximated by making a few assumptions about the errors within the background and the observation.superob observation error14superob observation errorassume that within a box:observation and background errors not correlated with each othe

7、r.background errors fully correlated.background errors have the same magnitude.15assumptions (cont):all of the innovations weighted equally.constant observation error correlation.superob observation errortoken evil math slide2()tsewded w2se superob observation errorw vector of weights (1xn)d diagona

8、l matrix of component observation errors (n x n)e observation error correlation matrix (nxn)11aaaeaaa observation correlation matrixa correlation within box. value calculated from correlation function in bormann et al., 20021 11 212 11nnn nccccccc cec cc cijcc correlation of ith observation with jth

9、 observation1800z 10 june, 2003.(20 n - 40 n) (0e 30 e)old observation errorsuperob error19experimental designcontrol run:operational set up plus goes bufr vis/ir/wv windsgoes-9 is still satob formatthinning to 2 x 2 x 100 hpa boxessuperob experimentsame as control run, except winds are superobbed t

10、o 2 x 2 x 100 hpa boxes20trial period: 24 jan -17 feb 20044 analyses and 6-hr forecasts 00z,06z,12z,18z1 analysis and 5-day forecast (12z)experimental design (cont)21token model info slide grid point model (288 e-w x 217 n-s)staggered arakawa c-gridapprox 100 km horizontal resolution (one-half opera

11、tional resolution)38 levels hybrid-eta configuration3d-var data assimilation22results % normalized root mean square (rms) error against control rms differences calculated for:mean sea-level pressure (pmsl) 500 hpa height (h500) 850 hpa wind (w850) 250 hpa wind (w250) in regions:northern hemisphere (

12、nh) tropics (tr) southern hemisphere (sh) for forecast periods of:t+24, t+48, t+72 ,t+96 , t+120 trial statistics24-2-101234pmsl t+24pmsl t+48pmsl t+72pmsl t+96pmsl t+120h500 t+24h500 t+48h500 t+72w250 t+24w850 t+24w850 t+48w850 t+72w250 t+24pmsl t+24pmsl t+48pmsl t+72pmsl t+96pmsl t+120h500 t+24h50

13、0 t+48h500 t+72w250 t+24experiment control rms error (%)tp observationstp analysisnh observationsnh analysissh observationssh analysis25anomaly correlationsvs. forecast rangecompared to analysis500 hpa heightnhshtr26t+24 forecast sonderms vector error 250 hpa windnhtrsh27250 hpa u-componentanalysis

14、increments28250 hpa u-componentanalysis increments29results summarysuperobbing experiment results are small and mixedgenerally more positive in the northern hemisphere than in the southern hemisphere or tropicstime series results are mixed: some forecasts better than control, some worse30implicationsmixed results suggest either:random error not most significant error component of amvssuperobbing set up not ideal to treat random error31future workback to basics approachre-calculate observation errors from innovation statisticsexperiment with