‹header›
‹date/time›
Click to edit Master text styles
Second level
Third level
Fourth level
Fifth level
‹footer›
‹#›
Seasonal-to-interannual prediction was one of our main interests ie. initial conditions for NSIPP
NOAA’s GDAS (Global Data Assimilation System) forcing used for both land surface models. Plot is for 31 March 2001; output from the Global Land Data Assimilation System project
Areas needing soil moisture are areas with the greatest difference in LSM predictions in slide 5 This slide is for Northern Hemisphere summer! Australia shows greater need during Southern Hemisphere Summer.
Passive Microwave -> all weather capability, less affected by roughness, topography and vegetation
Brightness Temp – soil moisture and temperature
SMMR -> AMSR (6.6GHz) and TRMM (10.7GHz)
AMSR launched in March 2002!
L-band ideal
We use the extended Kalman filter; an application of Bayes Theorem
Requires covariance forecasts
Prohibitive for large systems without simplifications
Requires tangent linear model for traditional form
May be unstable for highly non-linear problems
Basis for most sequential methods ie OI, nudging, PSAS etc
We will concentrate on the Kalman filter in this presentation
We use a 1D Kalman filter – computational efficiency, likely little spatial correlation (through topography, soil properties and atm forcing), 1D calculations
Key innovation is the shape of the land surface element; catchment
3 moisture prognostic variables
-> stressed, unstressed and saturated areas
-> 2cm, 1m, 1 to 3.5m
Parameters – soil, topographic and vegetation
Topographic: mean, standard deviation and skewness of CTI; Hydro1K scaled after Wolock and McCabe to 100m
Soil: porosity, wilting point, Ks, Clapp and Hornberger b, saturated matric potential, total soil depth; dominant soil texture from 5’x5’ FAO digital soil map of the world and parameters from Cosby et al 1984 -> partitioning and timescale parameters
Vegetation: vegetation type (ISLSCP1 using SiB classification), greenness fraction & LAI (climatologies from AVHRR NDVI at 1degx1deg from 1982 to 90)
Assumed:
Constant values for single scattering and roughness
Equal horizontal and vertical optical depth
Wang and Schmugge dielectric model
Repeat coverage at mid-latitudes every 3-4 days
Pixels affected by water bodies, temp less than 1degC and large optical depth excluded
Error estimates from error propogation theory
No calibration; compared well with 10cm measurements in Illinoois
Simulations are for 1979; other years soon forthcoming
 From mid March there are notable differences in central North America; possibly due to unaccounted irrigation
Illinois – 81 to 96
Iowa 72 to 94
New Mexico 82 to present
Vast improvement in May/June; no improvement earlier due to snow
Converged back to openloop after June; due to land surface model and obs
near-surface obs drier than field; smmr underestimated during summer? (disparity of data depths and averaging)
due to high veg biomass of corn site?
increased error not adequately represented in the obs covariance data
Still a net improvement though
Highlights the need for unbiased data and adequate characterisation of the error
White circle SMMR
Red circle NMM data
Red dotted line swe
Yellow solid line assimilation
Green dashed line open loop