User Tools

Site Tools


start:hype_file_reference:assiminfo.txt

AssimInfo.txt

General

The AssimInfo.txt file contains additional model settings to info.txt and is therefore located in the same folder as info.txt. The AssimInfo-file is used to define what kind of data assimilation to be done if assimilation is turned on in info.txt. The file information is divided into four groups:

  1. general settings,
  2. setting of control variables,
  3. observation settings, and
  4. meteorological forcing data settings.

The settings are usually kept in their four groups, but that is not necessary. The general settings all start with the letters G_, while the control variable settings start with A_, observations settings with O_ and forcing settings with F_. The control variable settings determine which group of HYPE states the filtering will be applied to.

File format

The basic format in the info file is simply a row-wise code-argument(s) combination:

!! <comment>
<code 1.1> [<code 1.2>] <argument 1> [<argument 2>] ... [<argument n>] 
<code 2.1> [<code 2.2>] <argument 1> [<argument 2>] ... [<argument m>] 
...

Comment rows can be added anywhere and are marked with double exclamation marks, i.e. !!, or '!!' followed by a space. For other rows, the first code string decides what information is to be read. The code can be written within or without apostrophes ('…'). Codes are not case sensitive. Maximum 18000 characters can be read on a single line.

General settings

Code Argument Description
G_NEIntegerEnsemble size, number of ensemble members (default=100)
G_MVRealMissing value for the assimilation routine (default=-9999) (not useful for HYPE, must be -9999?)
G_MEANOUT0/1mean(1) or median(0) value printed in ordinary output files (default=mean)
G_STATOUT0/1/2/3/4/5/..Extra output files for statistics (0-5) and ensemble members (6-). 1 give minimum as _002, 2 give min and max (as _003), 6 and up give ensemble members 1 and up to maximum 5+NE (as _007 and up). Note: 3-5 is not implemented. They should in the future give: 3 give previous and 0.025-perc (as _004), 4 give previous and median (as _005), 5 give previous and 0.975-perc (as _006) (default=0)
G_XYLOCRealHorizontal length scale [m] for covariance localization (distance with ~90% covariance reduction) (default=1000000)
G_ZLOCRealVertical length scale [m] for covariance localization (default=100000)
G_USEBINX0/1/2Use bin-files to hold state ensembles (0=no, 1=one bin-file, 2=several bin-files) (default is no)
G_USEBINFA0/1/2Use bin-files to hold forcing and auxiliary ensembles (0=no, 1=one bin-file, 2=several bin-files) (default is no)
G_INIBIN0/1Initialise assimilation run with previously saved bin files for state ensemble (0=no (default),1=yes)
G_STOP0/1Stop simulation when cholesky factorisation fails (0/1) (default=0=not stop)
G_CNC0/1Collapse non-controlled states to ensemble mean(or median) (0==no, 1=yes) (default=0)
G_TRANSTAT0/1Transform state variables (and some outvar) before the EnKF analysis (0=no, 1=yes) (default=0). If yes, then statevariables with physical range [0,+inf] will be log-transformed, and variables with range [0,1] will be logit transformed. Once implemented, Yeo-Johnson transform will be used on variables with unbounded physical limits (such as temperatres)
G_TRANEPSRealepsilon = minimum value used for log and logit transforms (used for state variables with physical range [0,+inf] and [0,1], respectively) (default=0.000001)

Control variable settings

These are the variables controlled by assimilation. They are set as a group by category or separate by name. The categories and names are specific to the HYPE model. If they are turned off, the analysis is NOT applied to variables in this category. Instead, they are re-initialized to the ensemble mean (or median depending on G_MEANOUT) after each time step. If they are turned on, the analysis IS APPLIED whenever there are observations available. No re-initialization.

Format of control variable lines: They start with A_ followed by include_ or exclude_ followed by bycategory or byname followed by [category] or [name]. To identify the variable by name, the category of the variable need to be set on a line directly before the variable. Example:

A_INCLUDE_BYCATEGORY SNOW 
A_INCLUDE_BYCATEGORY SOIL 
A_EXCLUDE_BYCATEGORY GLACIER 
Code Argument Description
A_INCLUDE_BYCATEGORYCategoryCategory is defined in the HYPE code (see table below)
A_EXCLUDE_BYCATEGORYCategoryCategory as above
A_INCLUDE_BYNAMENameName is state variable name in HYPE code (see table below).
A_EXCLUDE_BYNAMENameName as above
Category Names
SNOWsnow csnow snowage snowdepth snowcov snowmax snowheat snowliq
GLACIERglacvol
LAKEICElakesnow lakesnowage lakesnowdepth lakeice lakebice lakeicecov lakeicepor
RIVERICEriversnow riversnowage riversnowdepth riverice riverbice rivericecov rivericepor
SOILwater temp deeptemp conc humusN fastN partP fastP humusP fastC humusC PPrelpool Srelpool oldgrw partT1 surface icelens
AQUIFERwater conc lastrecharge clastrecharge nextoutflow cnextoutflow
RIVERWTwater temp conc TPmean temp10 temp20 Psed qqueue cqueue cwetland Qdayacc Q365 Qmean T1sed Ssed
LAKEWTwater temp conc TPmean temp10 temp20 uppertemp lowertemp volfrac
MISCtemp5 temp30 temp10 temp20 gdd gsbegin nextirrigation cnextirrigation updatestationsarcorr floodwater cfloodwater partT1sf nexttransfer cnexttransfer

Observation settings

The observation settings determine which observations should be assimilated. The observations settings are given as a table with one observation variable per line. The settings include which HYPE outvar variables to compare, the ensemble generation model, minimum and maximum values allowed, standard deviation parameters, and parameters for generation of spatially correlated perturbations. The columns are in the order given in the table below.

Column number Column name Type Value range Description
1 ObservationCharacterO_nnn Beginning with the code for observation setting (“O_”), the following characters (nnn) are a description for the user
2 IDobs4 charactersHYPE variable ID The 4 letter code for the observation as used by HYPE
3 IDmod4 charactersHYPE variable ID The 4 letter code for the corresponding simulated variable as used by HYPE for output
4 EnsTypeInteger0-4Ensemble generation model (following Turner et al). EnsType definition: 0=not used, 1=unrestricted, 2=semi-restricted(minimum), 3=semi-restricted(maximum), 4=constrained (max and min)
5 MinReal-Minimum value allowed (EnsType 2,3,4). Perturbations outside this range will be truncated to the min value.
6 MaxReal-Maximum value allowed (EnsType 2,3,4). Perturbations outside this range will be truncated to the max value.
7 MinsigmaReal-Standard deviation parameter. Minsigma is minimum allowed standard deviation
8 SigmaReal-Standard deviation parameter. Sigma is constant standard deviation used for EnsType = 1, also used as minimum allowed standard deviation for EnsType = 2-4
9 SemiMetaReal-Standard deviation parameter. SemiMeta is relative standard deviation used for EnsType = 2 & 3
10 RestMetaReal-Standard deviation parameter. RestMeta is relative standard deviation for EnsType = 4
11 LscaleReal-correlation length (horizontal)
12 GridSizeReal-cellsize (x and y dir) in the 2D grid used for the 2D spatially correlated random fields (interpolated to the model coordinates)
13 CorrTypeInteger1-3correlation function: 0 none 1 Gaussian, 2 Compact 5th degree polynomial, 3 Power law
14 CoordidInteger1-4spatial domain of observation (1=subbasin, 2=upstream area (ie. COUT), 3=aquifer, 4=outregions)
15 TransformInteger0-3kind of transformation to be applied to the variable before filtering (0=none, 1=log, 2=Yeo-Johnson (not implemented yet), 3=logit
16 epsilonReal-minimum value used to avoid 0 in log or logit transform
17 ClassGroupCharacter-Optional. If a class group variable is used, the class group name (as defined in info.txt) is given.

Forcing data settings

The meterological forcing data settings determine which forcing data should be perturbed and included in assimilation. The settings are given as a table with one forcing variable per line. The settings include the ensemble generation model, minimum and maximum values allowed, standard deviation parameters, and parameters for generation of spatially correlated perturbations. The columns are in the order given in the table below.

Column number Column name Type Value range Description
1 ForcingCharacterF_nnn Beginning with a code for forcing data setting (F_), the following characters (nnn) are a description for the user
2 IDobs4 characters - A letter code for the forcing as used by HYPE. It is the filename without the file ending, e.g. Pobs
3 EnsTypeInteger0-4Ensemble generation model (following Turner et al). EnsType definition: 0=not used, 1=unrestricted, 2=semi-restricted(minimum), 3=semi-restricted(maximum), 4=constrained (max and min)
4 MinReal-Minimum value allowed (EnsType 2,3,4). Perturbations outside this range will be truncated to the min value. Note: TMIN and TMAX is handled as deviations from Tobs in the DA, thus their range is the range of the deviation (negative for TMIN, positive for TMAX).
5 MaxReal-Maximum value allowed (EnsType 2,3,4). Perturbations outside this range will be truncated to the max value. Note: TMIN and TMAX is handled as deviations from Tobs in the DA, thus their range is the range of the deviation (negative for TMIN, positive for TMAX).
6 MinsigmaReal-Standard deviation parameter. Minsigma is minimum allowed standard deviation
7 SigmaReal-Standard deviation parameter. Sigma is constant standard deviation used for EnsType = 1, also used as minimum allowed standard deviation for EnsType = 2-4
8 SemiMetaReal-Standard deviation parameter. SemiMeta is relative standard deviation used for EnsType = 2 & 3
9 RestMetaReal-Standard deviation parameter. RestMeta is relative standard deviation for EnsType = 4
10 LscaleReal-correlation length (horizontal)
11 GridSizeReal-cellsize (x and y dir) in the 2D grid used for the 2D spatially correlated random fields (interpolated to the model coordinates)
12 CorrTypeInteger1-3correlation function: 0 none 1 Gaussian, 2 Compact 5th degree polynomial, 3 Power law
13 TauReal-perturbation memory coefficient (fraction of perturbation propagated from previous timestep)
start/hype_file_reference/assiminfo.txt.txt · Last modified: 2022/09/27 14:18 by cpers