The generation of 5k land surface forcing dataset in China Xiaogu zheng, Xue Wei
Original data anusplin 5k 3hr data Data flow Data preparation
Original Datasets Five global land surface forcing datasets – Prin( 1d, 3hr, 50yr) – Ncc (1d,6hr, 50yr) – Gswp2 (1d,3hr, 10yr) – Gold ( T62,6hr, 50yr) – NCEP_qian( T62, 3hr, 50yr) 700+ meteorological stations hydrological stations
Variables forcing datasets ( prin, gswp,ncc) – 3hr/6hr T, P,Q,W, PRCP (rate),SW,LW Instantaneous field: T,P,Q,W Average field : PRCP, SW, LW – Different treatment for these two fields when temporal downscaling from 6hr to 3hr for NCC data meteorological stations – Daily values of T,P, RH,PRCP (amount), W hydrological stations – Daily value of PRCP (amount)
1 d mean forcing data Instantaneous fields (t,p,q,w) – If hr=0,6,12,18 1d_mean =(prin + gswp + ncc)/3 – If hr = 3,9,15,21 1d_mean= (prin + gswp)/2 Average fields (sw,lw,prcp) – Downscaling 6hr NCC to 3hr first – 1d_mean = (prin + gswp + ncc)/3
Obs Diurnal cycle Temporal downscaling for daily obs to 3hr – Daily metero Obs (Beijing time 20pm to 20pm) – Forcing data at Greenwich time – Get diurnal range from 1d forcing mean Interpolate forcing to obs location ( no elevation adjustment) Adjusted by obs_daily Previous day 20pm bj Today 20pm gw Previous day 12pm Today 12pm 12219
Splina input format Dimensions, variable, weight – Give same weight 1 to both obs & forcing Can’t calculate predicted error if weight !=1 – Dimension Independent variables (x, y must in km, not degree) Independent covariates varies for each forcing variable, chosen from following pool – x, y, z, t-3 (regression), other relative forcing variables
relations among variables p, t, sw, wind q lw prcp
Downward Short Wave No obs used, only 1d data as splina input sw_new = sw/(s0 *cos(sza)) Set threshold for solar zenith angle (sza) – If cos(sza)< cos(80 degree) cos(sza) = cos(80) f(x,y) -> splina – Test z, negative slope, not add in
Wind Dimensions[ f (x,y,z) + ] -> splina
Specific Humidity (q) Dimensions [ f(x,y) + t + p ] -> splina
Downward Long Wave No obs used, only 1d data as splina input Dimensions [f(x,y) + t + ] -> splina Test q, no obvious contribution
Precipitation Prcp_new = sqrt (prcp) Dimensions [f(x,y,z) + q + ] -> splina Signal/noise = 0.9
Reference Hutchinson M.F., Anusplin version 4.2 User guide Xiaogu zheng and Reid Basher, Thin-Plate Smoothing Spline Modeling of spatial climate data and its application to mapping south pacific rainfalls Reid Basher and Xiaogu zheng, MAPPING RAINFALL FIELDS AND THEIR ENSO VARIATION IN DATA- SPARSE TROPICAL SOUTH-WEST PACIFIC OCEAN REGION
Thanks Thanks to Zuoqi Chen for data plotting