GSI/WRF Regional Data Assimilation System and Its Application in theWeather Forecasts over Southwest Asia - Data Assimilation for Atmospheric, Oceanic and Hydrologic Applications - page 604

Table 25.1 The averaged height of topography (Hgt: meter), vegetation type (Veg) and soil type
(Soil) in the nine sub-regions (defined as Fig. 25.1 ) over SWA
ABC

D E F GHI

Hgt 328 2,557 737 3,833 4,839 958 67 0 75

Veg Barren Grass Barren Shrub

land

Wooded
tundra

Barren Barren Water Mixed

Dryland/
Cropland

Soil Loam Loam Clay loam Loam Loam Sandy

loam

Loam Water Loam

The nine sub-regions effectively represent the heterogeneity of complex terrain
in SWA region. Table 25.1 displays the average elevation of the topography (Hgt),
vegetation type (Veg) and soil type (Soil) over these nine regions. Except for the
water type in the Arabian Sea (marked H), the soil types in all other eight regions
are loam; and the vegetation types include barren, grass, shrub land, wooded land,
mixed dry/crop land and water. Three regions (B, D, E) with terrain above 2,500 m
are covered by short plants with grass (B), shrubland (D) and wooded tundra (E).
Three regions with terrain under 1,000 m (A, C and F) and the two plains regions
(G and I) are practically free of any plants.
This evaluation is designed to present the model errors of surface temperatures,
precipitation, wind speeds and upper atmospheric variables for both 24-h (hour)
and 48-h (hour) (e.g. the second 24-h) forecasts. The statistical measures used
to quantify model forecast performance are bias (forecast - observation), mean-
square error (MSE), and standard deviation (SD) error. The MSE represents
the total model forecast error including contributions from both systematic and
nonsystematic/random errors. Systematic error may be caused by a consistent
misrepresentation of physical parameters such as radiation or model convection.
Nonsystematic errors are caused by uncertainties in the model initial conditions or
unresolvable differences in scales between the forecasts and observations ( Nutter
and Manobianco 1999 ).
If X represents any of the parameters under consideration for a given time and
vertical level, then the forecast error is defined as X 0 D X f X a ,wherethe

subscripts denote forecast and analyzed/observed quantities, forecasts and

analyses, the bias is computed as

f and a

N X

Bias D X 0 D N X i (25.1)

i D 1

the mean-square error is computed as

N X

MSE D 1
N

X i 2

(25.2)

i D 1