.svd File

RFM File Formats: SVD-compressed LUT


Input(s), associated with SVD Flag.

SVD (Singular Vector Decomposition)-compressed Look Up Tables of absorption coefficient [m2/mole] for particular spectral range and absorber, generated for the MIPAS/Envisat instrument. (For other purposes, the general .tab look-up tables are recommended).

Specified in *SVD section of the Driver Table

(also described in ESA document PO-TN-OXF-GS-0011.txt)
File Structure
File Header
DATE TIME Single record
#CMNT Single record
MWCODE ID[.ISO] TAB Single record
NL NV V1 NP P1 DP NT T1 DT Single record
U matrix
For IV = 1, NV Loop over wavenumber
    U(IV,1) U(IV,2) ... NL values for each wavenumber
    ... U(IV,NL)
K matrix
For IX = 1, NP×NT Loop over p,T grid points
    K(1,IX) K(2,IX) ... NL values for each p,T grid point
    ... K(NL,IX)

FieldTypeDescription Units Range
DATE C11 File creation Date (DD-MMM-YYYY) [1] (ignored)
TIME C15 File creation Time (HH:MM:SS.MMMMMM) [1] (ignored)
#CMNT C80 Comment record (starts with '#') [1]
MWCODE C8 Microwindow Label Matches LABSPC
IDX I HITRAN Index of absorber ∈ {List of Absorbers}
.ISO I (optional) '.' then HITRAN Local Isotope#[2] 1:12
TAB C3 Tabulation Function ∈{'LIN', '4RT','LOG'}
NL I No. of Singular Values >0
NV I No. of Wavenumber Points >0
V1 D Lowest Wavenumber cm-1 >0
DV D Wavenumber Increment cm-1 >0
NP I No. of −ln(p) Points >0
P1 R Lowest −ln(p) p in hPa
DP R −ln(p) increment p in hPa >0
NT I No. of Temperature Points >0
T1 R Lowest Temperature K ≥0
DT R Temperature increment K >0
U(:,:) R (NV,NL) U Matrix[3]
K(:,:) R (NL,NX) K Matrix (NX=NP×NT) [4]
Type: I=Integer; R=Real; D=Double Precision; Cn=character string, length n.

  1. The File Header (ESA rather than RFM format) consists of 2 records, the first of which is the 'time stamp' and the second is equivalent to an RFM comment record. For this reason, the first two records of any .svd files are written to the rfm.log file rather than just the usual first (comment) record.

  2. An isotopic SVD is identified by a '.' character in position 12. The record is then either read as
    • (A8,X,I2,X,A3) (not isotopic)
    • (A8,X,I2,X,I1,X,A3) (isotopic)

  3. This is effectively UT (with Fortran indexing) — this keeps the records the same length (NL numbers) as for the K matrix.

  4. The absorption coefficient is intrinsically a 3 dimensional table with wavenumber, pressure and temperature axes. However for the purposes of Singular Value Decomposition the p,T indices are combined into a single X index: (P1,T1), (P1+DP,T1), ... ( P1+(NP-1)*DP, T1+(NT-1)*DT )

  5. To reconstruct the absorption coefficient first the matrix F is reconstructed as
    Fν,x = Uνl Kl,x
    The relationship between F(ν,p,T) and the absorption coefficient k(ν,p,T) is determined by
    TAB. For MIPAS, TAB is LOG, so Abs.Coeff (in m2/mole) k = exp(F)

CS_PT__0511_01.DAT pT Microwindow (MWCODE=PT__0511) absorption coefficient for H2O (IDX=01).