0. Introduction
The seven group files are relatively compact alternatives to the full Monthly Summary Trimmed and Untrimmed (MST or MSU) formats, intended for studies using only a few variables or statistics. Eight important statistics for each of four related variables are grouped together in each file using a packed binary format. Thus five files are needed to represent all 19 MST variables, and two files are needed to represent all eight MSU variables. The statistics were chosen to bring together information that can be used to analyze the variability of the data and inhomogeneities of their distribution in time and space.
Cross reference is made to supp. A for standardized unpacking information, and the same notation for variables and statistics is followed or extended, also using the same type of two-dimensional table presentation.
1. Monthly Summary Trimmed Groups (MSTG)
The five trimmed groups were derived from MST (described in
supp. A). Each MST was split into five MSTG records;
these were written out onto the five separate group files even if every
αβ (the value of the statistic α for the
variable β) was missing. Thus the record structure is
identical for all the groups and their parent MST file. The five
trimmed groups are numbered 3-7 to distinguish them from the untrimmed
groups (numbered 1-2). Groups 3-7 contain four variables each: 3 = (S,
A, Q, R), 4 = (W, U, V, P),
5 = (C, R, X, Y), 6 = (D, E,
F, G), and 7 = (I, J, K, L).
Table B1-1a shows the bit layout in common to any
MSTG, and Tables B1-1b through B1-1f show the bit layout of each of
the 64-bit or 16-bit sections of groups 3 through 7, respectively, in
sequential bit-order reading from top to bottom. An example showing
the bit-order is given following Table B1-1f.
# α Statistic Bits ----------------------------------------------------- rptin 16 year 8 month 4 2° box 14 10° box 10 identification checksum 12 ----------------------------------------------------- 11 3 3/6 sextile (the median) 64 6 m mean 64 5 n number of observations 64 15 e standard deviation estimate 64 1 d mean day-of-month of observations 16 2 ht fraction of observations in daylight 16 3 x mean longitude of observations 16 4 y mean latitude of observations 16 ----------------------------------------------------- total 384
# β Variable Bits Bits ---------------------------------------------- 1 S sea surface temperature 16 4 2 A air temperature 16 4 8 Q specific humidity 16 4 9 R relative humidity 16 4 ---------------------------------------------- total 64 16
# β Variable Bits Bits ---------------------------------------------------------- 3 W scalar wind 16 4 4 U vector wind eastward component 16 4 5 V vector wind northward component 16 4 6 P sea level pressure 16 4 ---------------------------------------------------------- total 64 16
# β Variable Bits Bits ---------------------------------------------------------- 7 C total cloudiness 16 4 9 R relative humidity 16 4 14 X WU 16 4 15 Y WV (14-15 are wind stress parameters) 16 4 ---------------------------------------------------------- total 64 16
# β Variable Bits Bits -------------------------------------------------------------------------------- 10 D S - A = sea-air temperature difference 16 4 11 E (S - A) W = sea-air temperature difference * wind magnitude 16 4 12 F Qs - Q = (saturation Q at S) - Q 16 4 13 G FW = (Qs - Q)W (evaporation parameter) 16 4 -------------------------------------------------------------------------------- total 64 16
# β Variable Bits Bits ---------------------------------------------------------- 16 I UA 16 4 17 J VA 16 4 18 K UQ 16 4 19 L VQ (16-19 are sensible and latent heat transport parameters) 16 4 ---------------------------------------------------------- total 64 16
((αβ, β = S,A,Q,R), α = 3,m,n,e,d,h,x,y)
which defines the following matrix, with 4 rows and 8 columns:
| α | 3 m n e d h x y ---|---|-------------------------------------- β | # | 11 6 5 15 1 2 3 4 ---|---|-------------------------------------- S | 1 | 3S mS nS eS dS hS xS yS A | 2 | 3A mA nA eA dA hA xA yA Q | 8 | 3Q mQ nQ eQ dQ hQ xQ yQ R | 9 | 3R mR nR eR dR hR xR yR
stored in the order:
column 1, row 1,..., row 4; column 2, row 1,..., row 4; ...; column 8, row 1, ..., row 4.
2. Monthly Summary Untrimmed Groups (MSUG)
The two untrimmed groups were derived from MSU (described in
supp. A). Each MSU was split into two MSUG records;
these were written out onto the two separate group files even if every
αβ (the value of the statistic α for the
variable β) was missing. Thus the record
structure is identical for all the groups and their parent MSU file.
Groups 1-2 contain four variables each: 1 = (S, A, P,
Q), 2 = (W, U, V, C).
Table B2-1a shows the bit layout in common to any MSUG, and
Tables B2-1b and B2-1c show the bit layout of each of the 64-bit or
16-bit sections of group 1 and group 2, respectively, in sequential
bit-order reading from top to bottom.
# α Statistic Bits ----------------------------------------------------- rptin 16 year 8 month 4 2° box 14 10° box 10 identification checksum 12 ----------------------------------------------------- 11 3 3/6 sextile (the median) 64 6 m mean 64 5 n number of observations 64 15 e standard deviation estimate 64 1 d mean day-of-month of observations 16 2 hu mean hour of observations 16 3 x mean longitude of observations 16 4 y mean latitude of observations 16 ----------------------------------------------------- total 384
# β Variable Bits Bits ---------------------------------------------------------- 1 S sea surface temperature 16 4 2 A air temperature 16 4 6 P sea level pressure 16 4 8 Q specific humidity 16 4 ---------------------------------------------------------- total 64 16
# β Variable Bits Bits ---------------------------------------------------------- 3 W scalar wind 16 4 4 U vector wind eastward component 16 4 5 V vector wind northward component 16 4 7 C total cloudiness 16 4 ---------------------------------------------------------- total 64 16
The coded and true value ranges, the units, and the base for the
fields that are unique in representation to the groups are given in
Table B3-1.
# α Statistic True value Unit s Base Coded ------------------------------------------------------------------------------- 1 d mean day-of-month of observations 2≤30 2 days 0.0 1≤15 2 ht fraction of observations in daylight 0.0≤1.0 0.1 -1 1≤11 2 hu mean hour of observations 1≤23 2 hours -0.5 1≤12 3 x mean longitude of observations 0.1≤1.9 0.2° -0.5 1≤10 4 y mean latitude of observations 0.1≤1.9 0.2° -0.5 1≤10 -------------------------------------------------------------------------------
The group number, 1 for (S, A, P, Q), 2 for (W, U, V, C), 3 for (S, A, Q, R), 4 for (W, U, V, P), 5 for (C, R, X, Y), 6 for (D, E, F, G), or 7 for (I, J, K, L), must be added into the usual checksum prior to the modulo for proper identification. For example, supposing that the coded group 3 data matrix is available in an array MSTG, the checksum CK is computed and verified against the stored checksum CKS in FORTRAN as follows:
INTEGER CK,J,I,MSTG(4,8),YEAR,MONTH,BOX2,BOX10,CKS CK = 0 DO 500 J = 1,8 DO 400 I = 1,4 CK = CK + MSTG(I,J) 400 CONTINUE 500 CONTINUE CK = CK + YEAR + MONTH + BOX2 + BOX10 + 3 CK = MOD(CK,4095) IF(CK .NE. CKS) THEN PRINT *,'ERROR. CK = ',CK,' .NE. CKS = ',CKS STOP ENDIF
Instead of the standard deviation about the mean (statistic 7, s), this robust estimate is provided from the fifth and first sextiles: e = (s5 - s1)/2. (This was computed using integer truncating division on the coded quantity s5 - s1,i.e., rounding down.) For unpacking purposes, e is treated exactly like the corresponding standard deviation of each respective variable.
The centroids of observational location in time and space are shortened in length and precision from their representation in MST and MSU. "Nice" true values are reported for d, h, x, and y using the aforementioned units and base. Because of successive rounding steps, these values actually represent intervals whose trimmed or untrimmed behavior is shown in Tables B3-2a or B3-2b. The lowest and highest intervals are always exceptions to the behavior of the central intervals, so these extreme intervals are shown explicitly with the deviation (δ) of the actual midpoint of this interval from the reported true value. Coded values greater than one and less than the maximum coded value correspond to central intervals that can be obtained by subtracting the minus value and adding the plus value to the reported true value, yielding an inclusive lower and exclusive upper bound. The actual midpoint of each central interval can be obtained in any of three ways: by taking the mean of the upper bound and the lower bound, by plugging the basex shown into the usual formula in place of base, or by adding the midpoint deviation (δ) to the reported true value. For example, the intervals and actual midpoints corresponding to true hu values 1, 3, 5,..., 23 are [0,2.05), [2.05,4.05), [4.05,6.05),..., [22.05,23] and 1.025, 3.05, 5.05,..., 22.525.
-------|-------------------|-------------------------------|-------------------- | Lowest | Central | Highest # α | Interval δ | ± reported Basex δ | Interval δ -------|-------------------|-------------------------------|-------------------- 1 d | [1,3.1) +0.05 | [-0.9,+1.1) 0.05 +0.1 | [29.1,31] +0.05 2 ht | [0,0.055) +0.0275 | [-0.045,+0.055) -0.95 +0.005 | [0.955,1] -0.0225 3 x | [0,0.205) +0.0025 | [-0.095,+0.105) -0.475 +0.005 | [1.805,2] +0.0025 4 y | [0,0.205) +0.0025 | [-0.095,+0.105) -0.475 +0.005 | [1.805,2] +0.0025 -------|-------------------|-------------------------------|--------------------
-------|-------------------|-------------------------------|-------------------- | Lowest | Central | Highest # α | Interval δ | ± reported Basex δ | Interval δ -------|-------------------|-------------------------------|-------------------- 1 d | [1,3.05) +0.025 | [-0.95,+1.05) 0.025 +0.05 | [29.05,31] +0.025 2 hu | [0,2.05) +0.025 | [-0.95,+1.05) -0.475 +0.05 | [22.05,23] -0.475 3 x | [0,0.205) +0.0025 | [-0.095,+0.105) -0.475 +0.005 | [1.805,2] +0.0025 4 y | [0,0.205) +0.0025 | [-0.095,+0.105) -0.475 +0.005 | [1.805,2] +0.0025 -------|-------------------|-------------------------------|--------------------