Figure F horizontal color slice and contour slice of the pressure together with the map of
the HIRLAM domain
Figure F iso surface of the wind at knots together with the orography of the HIRLAM domain
APPENDIX F EXAMPLE IMAGES
Figure F contour surface of a D eld without handle and a D eld with handle
Figure F legenda of the temp erature
App endix F
Example Images
Figure F vertical color surface of the temp erature together with the orography of the
HIRLAM domain
APPENDIX E VOGL IMPLEMENTATION DETAILS
Note This routine do esn t yet work in the VIS D distribution of VOGL It is prob
ably obsolete as the line style is not device dep endent any more The linestyle
must b e used when scan converting lines
For full details the reader is referred to the X driver The le X c contains the complete
sp eci cation of a VOGL device driver
E VOGL DEVICE DRIVER STRUCTURE
Note This routine is not available in the original VOGL distribution
It is the resp onsibility of the device to approximate bit color when a frame
bu er of only bits is available
DEV mapcolor int LUTindex int r int g int b
Changes a colormap index to a given RGB value
DEV font char fontfile
Sets up a hardware font This should also set vdevice hwidth and vdevice hheight
which are the width and height of the current hardware font in pixels VOGL assumes
that hardware text is of a xed width
DEV char char c
Prints a character of hardware text This routine must do any neccessary moving to
make sure the current drawing p osition on the device is correct and it must leave the
device in graphics mo de
string char s DEV
Prints a string of hardware text This routine must do any neccessary moving to make
sure the current drawing p osition on the device is correct and it must leave the device
in graphics mo de
DEV fill int n int x int y
Draws lled p olygons devices which do not supp ort this should just do an outline
Note This routine is obsolete in the VIS D distribution of VOGL It is never called
but included to retain backward compatibility
backbuf DEV
Initializes double bu ering by selecting the back drawing bu er and p erforming any
other initializations If no double bu ering is supp orted then should b e returned
Note The double bu ering used is in fact pseudo double bu ering Drawing is done
is a backbu er which is copied into the front bu er when requested
DEV swapbuf
Swaps the front and back bu er What it really do es in most device drivers is copy the
back bu er into the front bu er
frontbuf DEV
Switches drawing into the front drawing bu er
DEV setlw int w
Sets the line width
Note This routine do esn t yet work in the VIS D distribution of VOGL It is prob
ably obsolete as the line width is not device dep endent any more The linewidth
must b e used when scan converting lines
DEV setls int lss
Sets the line style of line drawings
APPENDIX E VOGL IMPLEMENTATION DETAILS
There must also b e a function in the device driver to copy the device driver structure in
the global used device entry This function must b e called conditionally from drivers c
as should the device name print statement the device are printed if VOGL is used but no
device is set
The nineteen functions required for a complete device driver are
DEV init
Device initialization routine A routine which enables graphics on the device sets the
default color map and sets vdevice maxS xy and vdevice minS xy to the window
size in pixels
Note In the VIS D distribution of VOGL this routine also initializes the palette to
colors distributed in the RGB space when the frame bu er is only bits
exit DEV
Do es the necessary cleaning up to allow VOGL to exit leaving the device in a usable
form
DEV draw int x int y
Draws a line from the current device p osition to a p oint x y in vogle device co ords
note these assume that is the b ottom left hand corner This routine must up date
the current device p osition
Note This routine is obsolete in the VIS D distribution of VOGL It is never called
but included to retain backward compatability
DEV point int x int y
Draws a single pixel in the current device color
Note This function is not available in the original VOGL distribution
DEV getkey
Gets a single character of input from a device capable of providing it
DEV locator int wx int wy
Returns the mouse p osition for the device in VOGL device co ordinates returned in the
arguments and returns a bit pattern giving which buttons were down at the time of
the call
DEV clear
Clears the current viewp ort to the current background color
DEV color int LUTindex
Changes the current device color to the sp eci ed LUT index
DEV RGBcolor int r int g int b
Changes the current device color to the sp eci ed RGB value
E VOGL DEVICE DRIVER STRUCTURE
void lmbind short target short index
lmbind binds one of the eleven light controlling resource of the GL to a resource de ned
by lmdef target can b e one of MATERIAL BACKMATERIAL LIGHT or LMODEL
void shademodel long model
Sets the shading mo del Two shading mo dels are supp orted FLAT and GOURAUD
long getsm void
Returns the current shading mo del The value returned can b e either FLAT or GOURAUD
E Depthcueing functions
Private functions
void calc depthcue Vertex v
Calculates the color of the given vertex v using the Screen Z co ordinate of v and the
current RGB and Z range
Public functions
void depthcue int onoff
Turns depth cueing on or o If onoff is zero depthcueing is turned o Otherwise it
is turned on
void lRGBrange short rmin gmin bmin rmax gmax bmax long znear zfar
Sp eci ed the RGB color and Z range to b e used for depth cueing The values rmax
gmax and bmax are mapp ed to znear
E VOGL device driver structure
VOGL device drivers contain the following information
Three character p ointers giving
The name of the device
The name for the small hardware font
The name for the large hardware font
function p ointers giving the functionality of the device If a device is not capable
of some function eg color changing a no op function should b e provided which has a
return value of
APPENDIX E VOGL IMPLEMENTATION DETAILS
COLOR RGB color commands will set the current color If a color is the last LMC
thing sent b efore a vertex the vertex is colored If a normal is the last thing sent
b efore a vertex the vertex will b e lighted LMC COLOR is the default mo de
LMC EMISSION RGB color commands will set the EMISSION color prop erty of
the current material
AMBIENT RGB color commands will set the AMBIENT color prop erty of LMC
the current material
LMC DIFFUSE RGB color commands will set the DIFFUSE color prop erty of
the current material Alpha the fourth color comp onent sp eci ed by RGB color
commands will set the ALPHA prop erty of the current material
LMC SPECULAR RGB color commands will set the SPECULAR color prop erty
of the current material
LMC AD RGB color commands will set the DIFFUSE and AMBIENT color prop
erty of the current material Alpha the fourth color comp onent sp eci ed by RGB
color commands will set the ALPHA prop erty of the current material
NULL RGB color commands will b e ignored LMC
IMPORTANT lmcolor do es NOT change the data structures holding the material
de nitions but is provided to supp ort fast changes to the current material without
changing the de nition of the currently b ound material Thus lmcolor changes are lost
whenever a new material is b ound only the MATERIAL prop erties can b e changed not
those of the BACKMATERIAL
void n len type type normal
These functions are named identical as the color functions pre xed with an n instead
of an c and are used to sp ecify vertex normals len can currently only b e type can b e
s short i long d double or f oat Vertex normals must b e normalized in order
for the shading calculation to b e meaningful b ecause the dot pro duct of the vectors is
used as a measure of the angle b etween the vectors See also the function nmode
void nmode long mode
Tell the GL when normal vectors must b e normalized mode can b e
NAUTO normalize vectors only when the current Mo delView matrix is not orthonor
mal
NNORMALIZE always normalize vectors disregarding the current Mo delView ma
trix
Even in hardware re normalization can have an adverse e ect on the p erformance
void lmdef short deftype short index short np float props
De nes or mo di es a material light source or lighting mo del deftype can b e one of
DEFMATERIAL DEFLIGHT and DEFLMODEL index is the index of the resource to de ne
or mo dify each resource typ e has an unique array of de nitions np is the numb er
of entries in the props array props contains the various attributes appropriate to the
typ e of resource The kinds of attributes available for each resource typ e were listed in section
E VOGL ENHANCEMENTS
void ccall short r short g short b short a
Generic color pro cessing functions All public RGB color functions call this function
The arguments are clipp ed to a limit of Then dep ending on the color mode see
section the current color is set or a sp eci c material characteristic is set
Public functions
void cpack unsigned long argb
Sp ecify a true red green blue color using the packed ARGB format the color com
p onents are packed into one unsigned long as xAARRGGBB The A comp onent is
the alpha transparency comp onent and is currently not used in VOGL
void RGBcolor short r short g short b
Equal to cpack only sp eci es no alpha comp onent and red green and blue separately
void c len type type colors
These are various color sp eci cation functions which all take a vector an array as input
len determines the length of the vector exp ected and can b e array contains red
green and blue values or alpha comp onent also included type determines the
typ e of the array and can b e s short i long or f oat Colors are clipp ed to a
maximum of If the oat variant of the function is used values are exp ected to b e
in the range For the other functions the range is
E Lighting functions
The following functions are implemented for GL lighting supp ort
Private functions
void calc shade Vertex v
Calculates color of vertex v The supplied Vertex struct should contain the trans
formed vertex p osition in Eye co ordinates When lighting is on the color is calculated
using the vertex normal and lighting mo del Otherwise the vertex is colored using the
current color
void init lightmodule void
Initializes the light mo dule by binding all eleven lighting controlling resources to zero
The shading mo del lighting and normalization mo des are also initialize d
Public functions
void lmcolor long mode
Sp eci es the color mo de The color mo de determines the e ect of color commands like
cpack and the c calls The following color mo des are p ossible
APPENDIX E VOGL IMPLEMENTATION DETAILS
Private functions
zbuftype WtoSz Vector v
Returns the Screen Z co ordinate of the Vector v which must b e in Clip co ordinates
int zbuffer condition Scoord x Scoord y zbuftype z
Eight zbu er test functions which all return TRUE and up date the Z bu er if condition
is TRUE which can b e never less equal lequal greater notequal gequal and always
Selection of the functions is done by zfunction which sets a function p ointer to
one of the eight functions a case statement is probably to o slow b ecause the selected
function must b e called for every pixel in every geometry
swapbuffers void void zbuffer
Swaps the Z bu ers only when double bu ering is supp orted and activated
Public functions
void zbuffer Boolean onoff
Turns Z bu ering on or o
void zfunction int type
Sp eci es one of eight Z bu er comparison functions that determine if a pixel must b e up
dated or not Possible values for type are ZF NEVER ZF LESS ZF EQUAL ZF LEQUAL
GREATER ZF NOTEQUAL ZF GEQUAL and ZF ALWAYS For instance if ZF LESS is cho ZF
sen a pixel is up dated if the Z value of the incoming pixel is less than the existing Z value
in the Z bu er
void lsetdepth unsigned long near unsigned long far
Sets the range of Z bu er values that values in clip co ordinates are mapp ed to
void zclear unsigned long zfar
Clears the Z bu er to the sp eci ed value
void czclear unsigned long color unsigned long zfar
Clears b oth the Z bu er and the color frame bu er This function has b een provided by
the GL b ecause clearing b oth bu ers at once can often b e done in hardare faster than
clearing the bu ers separately using the functions clear and zclear This is not
the case in VOGL
E RGB color functions
Since VOGL only supp orted colormap mo de the following functions had to b e implemented
in order to enable the sp eci cation of RGB colors
Private functions
App endix E
VOGL implementation details
This app endix describ es the enhancements made to VOGL and the structure of VOGL device
drivers Of the private functions only the most imp ortant are discussed The reader is referred
to the source co de for further detail The public functions are all discussed
E VOGL enhancements
The list of typedefs of VOGL have b een extended with the following new typedefs
zbuftype This is the typ e used for each entry in the Z bu er It can b e changed
but then also the define for MAXZ must b e changed MAXZ represents the maximum
numb er that ts in the typ e zbuftype
Vertex This is a structure which holds information ab out vertices It contains the
following memb ers
v vertex p osition in Ob ject co ordinates as sp eci ed by the user
eye vertex p osition in Eye co ordinates
clip vertex p osition in Slip co ordinates
sx vertex x p osition in Screen co ordinates
sy vertex y p osition in Screen co ordinates
sz vertex z p osition in Screen co ordinates
r g b a vertex color and transparency when in RGB mo de
color vertex color when in colormap mo de currently not used
E Z bu er functions
The following functions have b een implemented to supp ort Z bu ering
APPENDIX D VIS D MODIFICATIONS
D VOGL RELATED MODIFICATIONS
vslice is altered it is of course necessary to change When the interface of extract
all the function calls as well The changed vertical slice routine is not enough however
to visualize wind slices correctly The wind vectors are scaled with a constant factor
in the function calc hwind and calc vwind The scaling must b e changed when
pressure levels are used
hwind and calc vwind calc
Change the scale factor for the W comp onent of the wind
D VOGL related mo di cations
In order to use VOGL with vis d the following mo di cations must b e made
x h
In x h the VOGL include le vogl h must b e included
graphics c
The graphics mo dule contains all D graphics library sp eci c co de Each function in the
graphics mo dule contains system sp eci c parts for the various graphics libraries used by VIS
D In order to use VIS D with VOGL some functions must b e extended with a co de section
sp eci c to VOGL while other functions can use the same co de as the SGI and IBM systems
do es The last co de sections are enclosed by sgi or ibm or VOGL prepro cessor statements
and can b e easily identi ed Function which are di erent for VOGL with resp ect to the SGI
are
init graphics Initialization is split in a system sp eci c part and a VOGL sp eci c
part The rst part initialize s the variables ScreenWidth and ScreenHeight to the size
of the screen and the second part sets variables which indicate the functionality that
is available in VOGL such as High Quality rendering and screendo or transparency
make window This function creates the D window and initializes the graphics mo dule
VOGL must b e instructed to use the window created in this function This is done by
calling the X driver sp eci c function vo xt window Furthermore VOGL do esn t
supp ort screendo or transparency which do esn t have to b e initialized
color Sets the current color and transparency The latter is not set for VOGL set
polytrinorm Due to strangeness in the IBM hardware it is necessary to break
p olytriangle strips into individual triangles and test to make sure the normal vectors
are correct Also triangles with zero area must b e culled out Apparently VOGL
su ers from the same b ehaviour b ecause the same precautions are necessary when
using VOGL
Other changes are necessary to disable calls to transparency functions
APPENDIX D VIS D MODIFICATIONS
grid c
old grid read
LEVELS Set the LevelType variable to HEIGHT
read new grid
LEVELS Set the LevelType variable to HEIGHT
Note that in the function read varNL grid the LevelTyp e is read from the le This
is not conditional the leveltyp e is always available in this typ e les ID x
load dataset grid
Dep ending on the leveltyp e read from the le or set to the default value ini
tialize the variable pairs BottomCoordinate TopCoordinate and BottomBound
TopBound to the appropriate values For height levels the co ordinate pairs are
equal i e BottomCoordinate BottomBound equals the height of level and
TopCoordinate TopBound equals the height of level MaxNL For pressure
levels BottomBound must b e set to instead and TopBound to MaxNL A warning is
LEVELS is used instead printed if an unknown leveltyp e is encountered and HEIGHT
get grid value
Implementation of the same typ e interp olation as describ ed in section
render c
draw box
When real domain b ounds must b e printed near the b ox corners use the variables
BottomCoordinate and TopCoordinate instead of the variables BottomBound and
TopBound used previously
slice tick draw
Use BottomCoordinate and TopCoordinate to calculate the level values to place
at slice ticks
work c
extract hslice
Use linear interp olation for HEIGHT LEVELS and other levels default When
LEVELS however interp olation must b e p erformed using LevelType is PRESSURE
the log of the level pressures The pressure of the level to extract is calculated using
linear interp olation b etween the two adjoining levels and the resulting pressure is
used to calculate new interp olation fractions as describ ed in section
extract vslice
Logarithmic interp olation of vertical slices is currently not yet implemented but
the places are marked where mo di cations must b e made A metho d that can b e
used is describ ed in section
D LOGARITHMIC INTERPOLATION MODIFICATIONS
D Logarithmic interp olation mo di cations
The C prepro cessor statements have the following form
ifdef LEVELTYPES
newcode
else
oldcode
endif
A new global variable int LevelType is declared to contain the leveltyp e of the data Cur
rently LevelType can have the following values
HEIGHT LEVELS Metres are used as unit of the vertical co ordinate This is the default
leveltyp e and used for old COMP les VIS D should b ehave as b efore for data de ned
at height levels
PRESSURE LEVELS Pascal is used as unit of the vertical co ordinate If data with
this leveltyp e is visualized vertical interp olation is p erformed using the logarithm of
the vertical pressure co ordinate
Furthermore b ecause the vertical co ordinate stored in the Height array do esn t have to
b e the height of the levels in metres anymore these values can t b e used for scaling of the
top ography Two new variables are used BottomCoordinate and TopCoordinate represent
the b ottom and top of the domain in units of the vertical co ordinate The already existing
variables BottomBound and TopBound represent the top and b ottom of the domain in kilome
tres The rst pair of variables is used to calculate values to place at slice ticks i e the real
vertical co ordinates The second pair of variables is used to scale the height of the top ography
etc to the correct height
The following changes have b een made to vis d
globals h
Declare variables
extern int LevelType
extern float BottomCoordinate
TopCoordinate
and prepro cessor constants
LEVELS HEIGHT
PRESSURE LEVELS
globals c
Idem as globals h but only the variables and without the extern class
APPENDIX D VIS D MODIFICATIONS
static int compress iso verts vx vy vz
cvx cvy cvz verts numlev
int numlev
This function is obsolete and only changed to preserve consistency
hslice extract
New interface for extract hslice
static float extract hslice grid level colmajor numlev
int numlev
where numlev is the numb er of levels of the parameter for which to extract a slice
It is used as an upp er b ound for the level to extract
hslice calc
Mo di ed function call to extract hslice passing the extra parameter NL parm
vslice calc
Set the numb er of rows in the vertical slice to the numb er of levels of the parameter
calc chslice
hslice passing the extra parameter NL parm Mo di ed function call to extract
cvslice calc
Set the numb er of rows in the vertical slice to the numb er of levels of the parameter
calc hwindslice
hslice passing the extra parameter WindNL Mo di ed function call to extract
for U V and W
Calculate the Z co ordinate of the horizontal wind slice using WindNL
calc vwindslice
Set the numb er of rows in the vertical slice to the numb er of windlevels WindNL
calc traj
Convert from grid to graphics co ordinates using WindNL and check if WindNL is zero
D D FIELDS MODIFICATIONS
The new slice p osition is scaled using MaxNL but clipp ed to the extra parameter
maxlevel
move vslice
vslice New interface for move
static int move vslice numlevels
curx cury r c r c corner
int numlevels
where numlevels is a new input parameter and holds the numb er of levels of the
parameter for which a vertical slice is moved
Set the numb er of levels of the vertical slice to move to the supplied value numlevels
move slice
hslice and move vslice with the new supplied parameter times Call move
tra j c
trace
Set ql the vertical b ox dimensions to the numb er of levels of the wind parameters
WindNL
work c
calc surface
march with the correct numb er of levels NL parm Call main
Pass an extra parameter to compress iso verts The numb er of levels must b e
used in that function to scale the Z co ordinates of the calculated iso surface
iso verts compress
New interface for compress iso verts
static int compress iso verts vx vy vz vpts
cvx cvy cvz verts numlev
int numlev
where numlev holds the numb er of levels of the parameter for which the iso surface
generated must b e compressed If numlev is one a warning is given This should
never o ccur b ecause iso surfaces cannot b e calculated for single level elds
compress iso verts
New interface for compress iso verts
APPENDIX D VIS D MODIFICATIONS
chslices draw
If a horizontal color slice is drawn for a single level eld initialize the Z co ordinate
of the slice to Zmax instead of doing the division which will b e a division by zero
Disable the drawing of slice ticks for single level elds color slices
draw hwind
If a horizontal wind slice is drawn for a single level eld initialize the Z co ordinate
of the slice to Zmax instead of doing the division which will b e a division by zero
Disable the drawing of slice ticks for single level elds wind slices
draw slice tick
Calculate the slice tick Z p osition from MaxNL and make sure that MaxNL is greater
than MaxNL could b e if only single level elds are in the data set
Idem as ab ove now to calculate the slice tick level value
save c
save
Calculate the correct numb er of bytes to save using NL ip
Write the correct numb er of ga gb and grid values using NL ip
restore
Read the correct numb er of ga gb and grid values using NL ip
slice c
to hslice distance
Calculate the normalized slice level using MaxNL and check if MaxNL is greater than
one
find nearest slice
Skip single level eld slices when searching for the slice nearest to a given cursor
p osition These slices have no slice tick so they can t b e moved
Do the same for color slices
Do the same for wind slices
move hslice
A new interface
static int move hslice curx cury level corner maxlevel
int maxlevel
with a new variable which holds the upp er b ound on the slice movement
D D FIELDS MODIFICATIONS
Use NL ip instead of Nl as upp er b ound for the for lo ops
decompress
The mo di ed decompress interface
static int decompress ip data compdata ga gb
int ip
with an added ip parameter to use as index into the NL array
Use NL ip as upp er b ound for the for lo op
gui c
create widgets
Create no buttons for single level elds for iso surfaces and vertical slices
button row add
When a new button row is created for clone and analysis variables disable the
iso surface and vertical slice buttons if the variable has only one level
init new var
Initialize the horizontal slice level to NL newvar This will b e zero for
single level elds which is correct
Do the same for horizontal color slices
main c
initialize
Initialize the horizontal contour and color slice levels of all physical parameters to
levels as in gui c
Initialize the horizontal wind slice levels to WindNL
Initialize the D cursor p osition to the middle of the b ox MaxNL
render c
init box
Initialize the scale of the b ox using MaxNL instead of Nl
box draw
When plotting grid dimensions at the corners of the b ox MaxNL should b e used
draw hslices
If a horizontal contour slice is drawn for a single level eld initialize the Z
co ordinate of the slice to Zmax instead of doing the division which will b e a division
by zero
Disable the drawing of slice ticks for single level elds contour slices
APPENDIX D VIS D MODIFICATIONS
Use a new more di cult calculation to nd the le address in a McIDAS GRID
le This is more di cult b ecause the grids can vary in size First the total size
of one time step is calculated Then the o set in the current time step is found
and is added to timestepnumber timestepsize Returned by the function is the
sum of the value calculated and the size of the header which is calculated in
read variableNL grid
from disk read
Add the new le format When a variableNL typ e COMP le is used it is p ossible
to read the original data from the McIDAS GRID le on disc
Use the numb er of levels of the current parameter to read the ga gb values and
to allo cate the correct grid size
Read the correct numb er of bytes from the le the o set is calculated by
grid address
grid get
Allo cate a bu er of the right size using NL ip ip was already available as a
parameter thankfully
Get the address of the ga gb values for the grid to compress using the new macros
Decompress the data using a mo di ed version of the function decompress The
parameter numb er ip is passed as an extra parameter
release grid
Deallo cate is called with the mo di ed grid size as an extra parameter
grid value get
If the lev value to nd a value at exceeds the total numb er of levels available for
the requested parameter return a missing value
Mo dify the upp er b ound check of k the level grid index of the p oint to obtain
and if the parameter has only one level available set b oth level grid indices to zero
Interp olation in a cub e is not p ossible in this case
allocate clone variable
Initialize the NL array entry for the new clone variable to the value of the original
variable the one b eing cloned
Use the new numb er of levels as upp er b ound in the for lo op
analysis variable allocate
Set the NL array entry for the new variable to b ecause no levels are yet
calculated by the external analysis function
new grid install
Use the macro COMPSIZE to calculate the grid size
D D FIELDS MODIFICATIONS
old grid read
When the program is compiled with the prepro cessor symb ol VARIABLE NL de ned
it must still b e p ossible to read the old COMP formats see app endix A for a
description of the various le formats VIS D can handle Therefore the old
functions must b e patched in the following way
Read the single value tempNL from the le and copy it into every used entry
of NL MAXPARMS Also initialize MaxNL to tempNL Initialization of WindNL is
delayed until the parameters are scanned for existance of the U V and W
parameters later in grid c
Allo cate the bu ers for ga and gb changed in size use MaxNL instead of Nl
When reading the ga and gb values from the le use NL ip as upp er b ound
of the for lo op instead of Nl b ecause Nl no longer exists
Allo cate a bu er for the grid data size has changed now uses the COMPSIZE
macro
Read the grid from the le size has changed now uses the COMPSIZE macro
read new grid
Read the single value tempNL from the le and copy it into every used entry of
NL MAXPARMS Also initialize MaxNL to tempNL Initialization of WindNL is delayed
until the parameters are scanned for existance of the U V and W parameters later
in grid c
The new grid format contains a Height array of values given the height values of
the grid levels The size of this array is MaxNL
new grid are analog to the changes to The rest of the changes to the function read
old grid from the ga gb bu er allo cation onwards read
read variableNL grid
A new function is written to read the new format COMP les describ ed in ap
p endix A This function is just a straightforward implementation that reads the
new le format The functions read old new variableNL grid are much alike
data set load
This is the main function that calls one of the functions read old grid
new grid or read variableNL grid Prior to loading the data set the read
whole NL MAXPARMS array is set to This can make p ossible bugs go real wrong
and is intended to sp ot p ossible bugs more easy
variableNL grid func Another if id is inserted to call the new read
tion The FileFormat variable is given the value the rst unused value
After reading the data set the variable WindNL is set to the minimum of the numb er
of levels of the parameters U V and W if they are available If none of them is
available WindNL will have the value again to make bugs more apparent
Set the TopBound of the domain using MaxNL instead of Nl
grid address
APPENDIX D VIS D MODIFICATIONS
where numlevels is the new parameter and is used seven times in an upp er b ound
on the numb er of iterations of for lo ops
get mcgrid
mcgrid interface The changed get
get mcgrid file grid data numlevels
int numlevels
where numlevels is the numb er of levels to read from the McIDAS GRID le This
value is used to verify the read op eration to check if the correct numb er of levels
is read
globals c globals h
Declaration of the following global variables
int NL MAXPARMS MaxNL WindNL
These are describ ed earlier in this section
grid c The following COMP or grid formats are mentioned
old grid format equal numb er of levels for each grid grid by grid compression
grid format equal numb er of levels for each grid layer level by layer compres new
sion
VariableNL grid format variable numb er of levels for each grid layer level by layer
compression
In the grid mo dule bu ers are created to store the ga and gb values of each level
for all parameters and all time steps Entries from this bu er are set and obtained
by two macros which are mo di ed to the following
define GA IT IP LEV Ga IT MAXPARMS IP MaxNL LEV
define GB IT IP LEV Gb IT MAXPARMS IP MaxNL LEV
where IT is the time step IP the parameter and LEV the level to obtain or set a ga or
gb value for The array contains gaps b ecause for every parameter MaxNL values are
stored to simplify the indexing Because every grid has it s own size it is imp ossible
to use the variable CompSize which was initialized to Nr Nc Nl the
numb er of oats in a grid Therefore a new macro is de ned
define COMPSIZE ip Nr Nc NL ip
This macro must b e used instead of the old variable CompSize If the parameter
numb er ip is not available it must b e provided by mo difying function interfaces
D D FIELDS MODIFICATIONS
INTEGER FUNCTION VARNLUSERFUNC OUTGRID MAXNL OUTNL
INGRID SUMNL
NR NC NL NLIDX NP
NAMES DATE TIME
SOUTH WEST BOTTOM
DELTALAT DELTALON DELTAHGT
C ARGUMENTS
INTEGER MAXNL OUTNL SUMNL NR NC NP
INTEGER NL NP NLIDX NP
REAL OUTGRID NR NC MAXNL
REAL INGRID NR NC SUMNL
CHARACTER NAMES NP
INTEGER DATE TIME
REAL SOUTH WEST BOTTOM
REAL DELTALAT DELTALON DELTAHGT
References to the output grid are not changed outgrid ir ic il references
grid lo cation row ir column ic level il as normal References to the input grid
are changed in the following way The old intuitive manner of referencing ingrid
to read grid lo cation ir ic il of parameter ip
ingrid ir ic il ip
now b ecomes
ingrid ir ic NLIDX ip il
which is only slightly less intuitive
The compress routine compress grid must know the numb er of levels to com
press so this value is passed as an extra integer parameter The new interface is
describ ed b elow in this mo dule
The proto col of the output grid sent back to vis d has changed First the numb er
of levels in the output grid is sent then the grid data The numb er of ga and gb
values sent back also changed b ecause there is one such value for each level in the
output grid
compress grid
The changed compress grid interface
compress grid data comp min max levga levgb numlevels
int numlevels
APPENDIX D VIS D MODIFICATIONS
The receive proto col is changed The external functions sends the numb er of levels
in the grid that it has computed Dynamic allo cation of two bu ers one for the
uncompressed grid received the other for a compressed grid calculated is done
using this value After receiving the grid the compression values ga and gb are
received one of each for every level in the output grid
cursor c
cursor event
Scale the cursor Z co ordinate using MaxNL the size of the b ox expressed in numb er
of levels
extmain c This mo dule is separated from vis d in that it is not linked with vis d but is
linked instead with every external analysis functions The external analysis function is
called from extmain c
Declare the following new global variables
int NL MAXPARMS Number of levels for each parameters
MaxNL Largest number of levels in NL
SumNL Sum of all levels in NL
NLindex MAXPARMS Cumulative sum of levels in NL
call user function
Declare a new integer variable to pass to the external analysis function After
execution of the external analysis function it contains the numb er of levels in the
output grid
Receive an array of Nl s one for each parameter instead of one These values are
stored in NL MAXPARMS and MaxNL SumNL and NLindex are calculated during the
reception of the NL array
Allo cate storage for all the input grids of one time step and the output grid A total
time step has Nr Nc SumNL oats The output grid must b e allo cated worst case
ie the largest numb er of levels MaxNL b ecause it is not known in advance how
many levels the external analysis function will compute Also worst case bu ers
are allo cated for the ga and gb values and the compressed grid
Storing of the input grids received from vis d is done using NLindex If the
input bu er of size Nr Nc SumNL is seen as an array of levels then NLindex gives
the start index for each parameter in this array
mcgrid which now Read a grid from a McIDAS GRID le using the function get
takes the numb er of levels to read as an extra parameter change describ ed b elow
in this mo dule
The interface of the external analysis functions has b een changed It now lo oks like
D D FIELDS MODIFICATIONS
D D elds mo di cations
The C prepro cessor statements have the following form
ifdef VARIABLE NL
newcode
else
oldcode
endif
The following extra or new global variables are declared
int NL MAXPARMS to hold the numb er of levels for each parameter Note that the low
ercase l in the old single integer Nl has b een changed delib erately to make the compiler
generate a lot of errors which has help ed in nding the places where mo di cations had
to b e made
int MaxNL contains the maximum of all the numb ers of levels of all physical variables
MaxNL represents the size of the viewb ox in levels and must b e used in every op eration
that needs the height of the b ox
int WindNL contains the minimum numb er of levels of the three parameters U V and
W which are used for wind tra jectories and wind slices This variable is intro duced for
safety only In practice the user could present VIS D with a le containing di erent
sized U V and W grids Using the WindNL variable VIS D will only use the levels
available for all three parameters
The following list of mo di cation descriptions is alphab etically ordered by mo dule and the
mo di cations made to each mo dule are listed from top to b ottom Line numb ers have b een
omitted b ecause these are sub ject to change Instead mo di ed functions are listed b etween
parentheses
analysis c
compute analysis variable
Send a Nl value numb er of levels for each parameter instead of just one at the
b eginning to the external analysis functions The co de that handles reception of
the data can b e found in the mo dule extmain c
Send a variable sized grid to the external analysis function The size in oats is
given by the formula Nr Nc NL ip where Nr is the numb er of rows of the grid Nc
the numb er of columns and NL ip the numb er of levels of the current parameter
ip
Declare a new integer variable to receive an extra integer from the external function
the numb er of levels in the grid it will send thereafter
APPENDIX D VIS D MODIFICATIONS
The appropriate MAKECOLOR macro must b e de ned For HP systems this the same
macro as for SGI and IBM systems
extmain c
On HP systems the signed keyword is invalid in the typ edef for int
globals h
TASK De ne the symb ol SINGLE
De ne a new typ e int as char
graphics c
graphics c contains all D graphics library related co de Most changes are related to
VOGL see app endix D Currently the following parts are HP sp eci c
Initialization of integers holding the current screen size
Setting of le format ags VIS D can save X Windows dump for HP
main c
main
CORE is not de ned for HP Therefore the core dump disabling The symb ol RLIMIT
has b een turned o for HP
The font command line option is available on HP so notify the user when the
command line options are printed
Scan the command line arguments for the font option
Call the main pro cessing function main loop for HP b ecause HP is a single
threaded system
matrix h
Include x h to de ne the Matrix typ e
x h
Include the graphics library and X Windows header les
App endix D
VIS D mo di cations
This app endix contains high level descriptions of the mo di cations made to VIS D The
accent lies on the reasons for the mo di cations For details ab out the implementation of the
changes the reader is referred to the source co de of vis d All changes have b een marked by
prepro cessor statements for easy nding of the changes and backward compatibility an old
version of VIS D without this mo di cations can easily b e restored this way should problems
arise due to the mo di cations The changes are ordered alphab etically by mo dule name
The name of the mo di ed function is listed b etween parentheses
D HP p ort mo di cations
D LUI changes
lui c
MoveResizeWindow LUI
Set XSizeHints of the window by calling XSetNormalHints and
XSetStandardProperties The p osition and size must b e set but also the max
imum size of the window Otherwise all VIS D windows can b e resized to full
screen size
LUI ResizeWindow
Same as the change for LUI MoveResizeWindow but only the size and maximum
size have to b e set not the p osition
D vis d changes
cw c
APPENDIX C LIBRARIES
C RUL LIBRARIES
C grib vis
cmdline arguments Parse the command line arguments and store the results in pro cess
global variables If a variable is not set a default value will b e taken if appropriate
otherwise an error is rep orted
grid header Allo cation and initialization where p ossible of the GRID le header init
write grid header Write the GRID le header to a le
grid Write a grid to the current output le write
print grid info Print selected contents of a sp eci ed grid information structure
timesteps Determine the numb er of time steps in a sp eci ed time range count
APPENDIX C LIBRARIES
C cutil
The headers of the cutil functions can b e found in the header le cutil h
yymmdd centuryday Convert the date in yymmdd format to century day
centuryday yymmdd Convert the century day to the date in yymmdd format
hhmm minutes Convert time in hhmm format to minutes since midnight
minutes hhmm Convert the minutes since midnight to the time in hhmm format
iallo c Allo cate oat bu er and terminate if error o ccured
ifree Free oat bu er and warn if error o ccured
icheck Check if p ointer equals NULL and warn when this is the case
reorder matrix Reorder the input matrix from row ma jor to column ma jor
mirror matrix Mirror the matrix around the horizontal axis
sort Sort an array of integers in ascending or descending order
rep ortminmax Calculate the minimum and maximum of data and rep ort them
getstring Read a string enclosed in double quotes from bu er
C futil
These are functions that are not in the KNMI libraries or are a small section of functions
from the KNMI libraries
calcq Calculate the relative humidity
iyyddd Conversion from century day days since januari to the date in yyddd
format
setfortrancommon Set the variables to indicate the presence or absence of physical pa
rameters This function should rst b e called b efore using the interp olation routine
intpre Physical variables not present will b e calculated by intpre if p ossible
conphys De ne the physical constants used by the interp olation routines
ab ort A routine to simulate the ab ort function which is not known to the HP FORTRAN
compiler
getarg A routine to simulate the command line argument pro cessing function getarg not
known to the HP FORTRAN compiler
C RUL LIBRARIES
C grib
The headers of the grib functions can b e found in the header le grib h
grib parnumvalid Private test whether a passed physical variable is within the range of
the current parameter table size Returns TRUE when the parameter numb er is valid
FALSE otherwise
shortname Return p ointer to the short parameter name if the parameter numb er is grib
valid Otherwise a p ointer to N A Not Available is returned
unit Return p ointer to the unit of the passed parameter if the parameter is within grib
range Otherwise a p ointer to N A is returned
grib parname Return p ointer to the full parameter name if the parameter numb er is valid
Otherwise a p ointer to N A is returned
listparameters Print a list of physical variables as de ned in section of the grib grib
de nition
grib parnumfromname The parameter name that is passed invoking the function is re
turned as the parameter numb er if it exists
C interp ol
The headers of the interp ol functions can b e found in the header le interpol h For more
information ab out co ordinate systems see section C
calcmo delparameters Calculate the half and full mo del level parameters and interp olate
the pressure half and full mo del level parameters
interp olate calcindexarray Calculate index eld that maps output gridp oints to input
gridp oints with the help of the KNMI routine defind
init Initialize all global variables used for the interp olation The information interp olate
ab out the gridsize is passed in the Asimof Information structure The parameters ini
tialized are the data grids of the surface pressure the surface temp erature and the full
level parameters temp erature U V and the sp eci c humidity Also the mo del level
parameters are calculated These parameters are used by the interp olation
interp olate destroy Free all allo cated blo cks and unset all variables set by the routine
init interp olate
interp olate pressure Interp olate a D data grid to pressure levels according to the passed
Asimof Variable structure input grid and output grid with the help of the KNMI routine
intpre
APPENDIX C LIBRARIES
C asimof
The information needed for calling the functions loadfd and getfd is stored in the Asimof
Variable AsimofVars structure The grid information from an ASIMOF le is stored in the
Asimof Information AsimofInfo structure The exact headers of the asimof routines in this
section can b e found in the le asimof h The following routines are implemented in the
mo dule asimof c
allo cvars Allo cate and initialize an Asimof Variable structure with the parameter asimof
values If an input parameters is the default value from asimof h is used The
output is a p ointer to an allo cated and lled AsimofVars structure
asimof freevars Free all blo cks used by the Asimof Variable structure and memory used
by the structure itself
asimof op en Op en the ASIMOF grib database with the help of the function loadfd from
the KNMI p ort library and ll the Asimof Variable structure with all relevant data from
the ASIMOF le Returns the errorco de as returned by loadfd
asimof close Close the ASIMOF grib database connected to the device describ ed by the
Asimof Variable structure with the help of the function asimhc Returns the errorco de
as returned by asimhc
get eld Get a eld from the ASIMOF database le with the help of the function asimof
getfd Returns the errorco de as returned by getfd
asimof imdi Returns the key entry not used indicator
asimof setkey Fill the given key with imdi key entry not used values and insert values
given as routine parameters The function has a variable numb er of arguments
destagger Destagger data to a grid de ned by the Asimof Variable structure Destaggering
of the data is necesssary when the grid co ordinates are not the same as the grid data
co ordinates used by vis d
asimof getdata Extract data elds from the ASIMOF le The actual reading of the
getfield The parameter and level typ e of data in done with the function asimof
the physical variable to extract can b e sp eci ed with the function parameters When
no level numb er is sp eci ed this function extracts all data matching the parameter
and level typ e When more than one level numb er is found the levels are sorted from
b ottom level to top level Also the vis d grid information structure is lled with values
known at that p oint When necessary the data array is destaggered with the function
destagger see section C Optionally the data array is reordered to row ma jor and
is mirrored across the horizontal axis see section C
asimof getco ordinates Calculate the GRIB co ordinates of an input eld using the integer
section with the help of the function gbg fc Returns the numb er of grid p oints found
C RUL LIBRARIES
pint Vertical interp olation of geop otential from a hybrid mo del level to pressure levels
intpre Organization of the interp olation to pressure levels during p ost pro cessing
omint Vertical interp olation of OMEGA vertical velo city
p ostpp Main routine for p ost pro cessing to pressure levels and to mo del levels This routine
also handles the extraction of surface parameters It calls at ab out all vertical inter
p olation routines describ ed in this library Because the many parameters involved in
calling this function we prefered to call the separate interp olation functions
tpint Vertical interp olation of the temp erature to pressure levels
vineta Vertical interp olation from one level into another level
xpint Vertical interp olation from mo del levels to a pressure level
xtreta Organization of the extraction of mo del level data during p ost pro cessing
C grw
as ddr Op en an ASIMOF le and contruct the DDR from it
grclos Interface for the function asimhw
gread Read blo ck and blo ck from an ASIMOF le and print the DDR
grop en Interface to the routine as ddr and ASIMOF
gwclos Interface for the routine asimhc
gwop en Load blo ck and blo ck from an ASIMOF le and contruct the DDR
gwrite Prints the DDR for a GRIB co ded HIRLAM eld le and writes the blo cks to an
ASIMOF le
priddr Prints the DDR for a GRIB co ded HIRLAM eld le
C RUL libraries
Information and de nition of the GRID le header and the D grid information blo ck can
b e found in the header le grid d h
APPENDIX C LIBRARIES
grib ex Co ding and deco ding of GRIB le format messages
minmax Calculate minimum and maximum values from an array of oating p oint numb ers
C util
cd dat Converts century day numb er of the day in this century to day month year format
ddmmyyyy
convdt Converts day month year format ddmmyyyy to century day dddd
diasim Diagnose an ASIMOF le diasim can b e run with two options The option h will
print an extensive header and s will print the statistics of the eld data No option
will just print the GRIB section values of all variables
gregor Converts Julian day numb er into Gregorian normal date year month day
ic ymd Converts century day to day month and year
idat c Converts a given day month and year into century day
C vari
gb gcc Transform co ordinates describ ed in input section GRIB format to co ordinate
system describ ed in output section
gb llc Transform from one latitude longitude grid to another This function is mainly used
to transform from or to the geographical system
gb lp c Transform from geographical grid to p olar stereographical grid
gb plc Transform from p olar stereographical grid to latitude longitude grid
gb slc Shift p ole of regular latitude longitude grid
gb tlc Stretch regular latitude longitude co ordinates toward the north p ole
C prp o
de nd De nition of a p ointer eld to b e used for the extraction of grid p oint values in a
sub area from a grid p oint eld de ned in a larger area
destag Destaggering of wind comp onents to t the analysis program conventions This
program is used when the grid p oints of the wind comp onents are not the same as the
grid p oints of the system
etaeta Vertical interp olation from one level into another level
App endix C
Libraries
In this app endix the most imp ortant routines from the KNMI and RUL libraries are listed
For more detail the reader is referred to the inline do cumentation of the libraries
C KNMI libraries
C p ort
asimhc Close an ASIMOF le
asimhr Read an array from the ASIMOF le If the key passed along do es not exists in the
database an error is returned
asimhw Write an array into the ASIMOF le If the key passed along is not unique an error
is returned
getfd Get a co ded eld from the ASIMOF database and deco de a eld from GRIB This
function calls the functions asimhr and degrib see app endix C
loadfd Op en a direct access le as GRIB database If the le exists the contents will b e
extracted else the le will b e op ened with an empty table of contents
putfd Convert a eld into GRIB format and place it in an ASIMOF database This function
calls the functions asimhw and engrib see app endix C
C gco d
degrib Deco ding of data or identi cation sections section and from GRIB format
enco de Enco ding of data in GRIB format
APPENDIX B VIS D SOURCE CODE STRUCTURE
B LITTLE USER INTERFACE LUI LIBRARY
get d Read a GRID le to obtain the grid info structures
iggt d Read a data grid from a GRID le returning the numb er of data p oints read This
function is used from within vis d
For a description of the VIS D utilities see section
B Little User Interface LUI library
The LUI library contains the following mo dules
browser File browser mo dule
button Provides functions to create and maintain all sorts of buttons like toggle buttons
one shot buttons etc Buttons can b e group ed in button pads
dial Provides functions to create and maintain dials
dialog Provides functions to create dialog b oxes which can contain a message and OK and
Cancel buttons
event Event dispatcher Each window created by the LUI library can sp ecify events and a
call back function This call back function is called only for the sp eci ed events
lab el Provides text lab els and typ e ins
lui Main LUI mo dule that contains the LUI initialization and most functions dealing with
windows It also contains various color transformation functions
pixmaps Contains pixmap de nitions for radio widgets
p opup Handles p op up menus
slider Slider supp ort The user can use sliders to adjust a value used by the program
APPENDIX B VIS D SOURCE CODE STRUCTURE
INTEGER DATE Datestamp of variables in HHMMSS format
i e hours minutes seconds
INTEGER TIME Timestamp of variables in YYDDD format
i e year century day
REAL SOUTH WEST BOTTOM Geographic location of grids SOUTH
latitude in degrees WEST longitude
in degrees BOTTOM coordinate in
kilometres
REAL DELTALAT DELTALON
DELTAHGT Grid increments DELTALAT and DELTALON
in degrees DELTAHGT in kilometres
A user function should return if completed successfully and a non zero value if an error
o ccured
The extmain mo dule contains co de to receive the input data from vis d through a so cket
and send the output grid back to vis d The user functions is called from extmain
B Utilities
The most imp ortant utility mo dules include the following
comp d Conversion of McIDAS GRID les to COMP les the compressed vis d le for
mat There are several comp d programs one for each typ e of COMP le see ap
p endix A The structure of the comp d programs is as follows
Read the rst time step grid info structures checking that every physical variable
has the same numb er of levels etc
Read the following time steps grid info structures checking if the same physical
variables are encountered as in the rst time step
Allo cate memory to hold input grids
Read input grids from input GRID le s
Compress grid data
Write grids to output COMP le
Write header of output COMP le
Note that the time steps must b e ordered chronologically and the order of the physical
variables must b e the same for each time step
compinfo Read and display contents of COMP les
The rest of the programs can only handle the McIDAS GRID le format The most imp ortant
subroutines are
B EXTERNAL USER FUNCTIONS
slice Slice handling functions This mo dule pro cesses all events that o ccur when vis d is
in slice mo de The most imp ortant function is to nd the correct slice and moving it
when the user drags from a sp eci c p osition in the D window
so cketio Provides functions to send and receive data through a so cket These functions are
used to send to and receive from the external analysis functions
sync Provides semaphore functions for systems that allow multi threaded execution
top o Functions for reading and drawing the top ography
tra j Tra jectory mo dule Tra jectories are traced using this mo dule
vtmcp Mo dule dedicated solely to the calculation of iso surfaces
work This mo dule contains the actual routines to extract contour color and wind slices an
interface to the iso surface and tra jectory mo dule
zcont Contouring mo dule Given an input slice of data and a contour interval it calculates
the p osition of contours on the slice and outputs vector arrays containing the contour
p ositions
The mo dules gui c and cw c contain all the user interface functions making use of the LUI
library For the direct reading of McIDAS GRID les vis d uses the McIDAS GRID le
library
B External User functions
The Fortran mo dules must all contain a function called USERFUNC with the following in
terface
INTEGER FUNCTION USERFUNC OUTGRID INGRID NR NC NL NP
NAMES DATE TIME
SOUTH WEST BOTTOM
DELTALAT DELTALON DELTAHGT
REAL OUTGRID NR NC NL Output grid calculated by the function
REAL INGRID NR NC NL NP Input grids all physical variables
one time step
INTEGER NR Number of latitude points rows in
input grids
INTEGER NC Number of longitude points columns in
input grids
INTEGER NL Number of levels in input grids
INTEGER NP Number of physical variables
CHARACTER NAMES NP Names of variables Padded with spaces
APPENDIX B VIS D SOURCE CODE STRUCTURE
function of vis d needs a data grid it must b e requested from the grid mo dule Grids
are stored in a compressed form in memory and are decompressed when requested The
decompression routine uses two values which are also stored in a compressed grid le
ga and gb These values are used to scale and map oating p oint values from to bit
integers This conversion cannot b e done without losing precision vis d is able to read
the original GRID les with the oats as well see also app endix A
gui The GUI Graphical User Interface mo dule takes care of the creation and maintenance
of all windows except for the D window which is created in the graphics mo dule
GUI pro cesses all user input from the mouse and the keyb oard and up dates appropriate
global variables contained within the globals mo dule which determine how the data set
is displayed GUI makes extensive use if the User Interface Library LUI discussed in
section B
main Entryp oint of vis d After initialization pro cessing command line parameter reading
of the data set etc an eternal lo op is executed pro cessing user input
map Contains functions to read a map le draw the map in the D window and up dat
ing the display if necessary A map le usually contains a numb er of line segments
de ning land sea b orders and country outlines The elevation is calculated from the
top ography
matrix vis d uses its own library routines of matrix op erations to create and mo dify the
transformation and pro jection matrices used to map ob ject co ordinates into the D
window see section Matrix multiplication concatenation copying inversion
vector normalization and vector matrix multiplication is provided
memory Memory management mo dule vis d functions needing large amounts of memory
should call one of the functions from this mo dule vis d tries to retain all requested
graphics in memory When allo cation of a requested blo ck fails however the memory
management mo dule will try to create enough free memory by deallo cating the least
recently viewed graphics A numb er of functions is provided Memory allo cated using
pallo cate p ermanent allo cate will never b e deallo cated contrary to memory allo cated
by allo cate
queue Implementation of a task queue When only one pro cessor is available the requested
graphic will b e calculated immediately On systems with multiple pro cessors multi
threaded execution is p ossible Each requested graphic is put into the task queue
Pro cessors will calculate the graphics in the queue multi threaded Access to graphics
is made exclusive through the use of semaphores
render The render mo dule contains the drawing functions which are all based on the graph
ics mo dule The b ox slices tra jectories slice ticks clo ck etc are drawn from within
the render mo dule
save VIS D allows the user to save the current state of the D window to a le Data
saved includes top ography colors lab els and graphics with their colors and p ositions
Such a le can later b e restored
App endix B
VIS D source co de structure
B VIS D main source co de
vis d version is split into several mo dules The most imp ortant mo dules include in
alphab etical order
analysis Activation of activate external user analysis functions which can calculate new
grids from the available data Analysis programs must reside in a sp ecial directory
which is scanned by vis d When an analysis function is activated all available data
one time step at the time is sent through the so cket to the analysis program The
analysis program can calculate a new grid of data derived from the existing data This
new grid is then sent back through the so cket to the analysis mo dule of vis d When
successful the new grid is installed and can b e visualized as usual
extmain Wrapp er for vis d external analysis co de It is linked separately with every ex
ternal analysis function It receives the grid description and data through the so cket
connection with vis d and calls the user analysis function The grid description and
data are passed as parameters to the analysis function The interface is describ ed in
section B
globals Contain declarations of globally accessible variables and constants The most im
p ortant variables include the grid dimensions and the grid data p ointers
graphics All D graphics library related co de is contained in the graphics mo dule It
contains functions to make the D window initialize the graphics mo dule library draw
p olylines text triangle meshes quadrilateral meshes etc All functions in the graphics
mo dule are based up on functions from the D graphics library XFDI for the Stellar
GL for the SGI and IBM Functions from other mo dules call functions from the graphics
mo dule if they want something drawn in the D window
grid All grid management functions are contained in the grid mo dule grid will determine
the le format of an input le and read the data sets contained within the le If a
APPENDIX A F ILE FORMATS
int offset Offset in byte ints from start of file
int length Length of segment data in byte ints
Co ordinate information for each segment The p osition and length of the data is ob
tained from the directory structure The data consists of an array of l eng th l at l on
5
pairs in degrees
A VIS D top ography le format
A VIS D top ography le contains the surface elevation and land sea mask of a sp eci c part
of the earth surface Each le starts with the following header
char id ID string TOPO
int westlon West longitude in s of degrees
int eastlon East longitude in s of degrees
int northlat North latitude in s of degrees
int southlat South latitude in s of degrees
int rows nlat Number of rows latitude
int columns nlon Number of columns longitude
Positive latitude degrees are north p ositive longitude degrees are west
After the header the top ography data follows
short data rows columns
Row corresp onds to the north edge of the top ography Row ROWS corresp onds to the
south edge Column corresp onds to the west edge and Column COLUMNS corresp onds
to the east edge The shorts of the top ography data consists of the surface elevation in
metres multiplied by and with added if the lo cation is water
A VIS D MAP FILE FORMAT
int GRIDTIMES Maximum numb er of time steps
int NumTimes Numb er of time steps
int NumParms Numb er of parameters
int Nr Nc Horizontal grid dimensions global i e
for all grids
int NL NumParms A numb er of levels for every parameter in the le
oat LatInc LonInc Horizontal grid increments in latitude and
longitude direction
int LevelTyp e Typ e of vertical co ordinate
oat Height Nl Levelvalues unit dep ends on leveltyp e
char ParmName GRIDPARMS Names of parameters
oat MinVal NumParms Minimum values of parameters
oat MaxVal NumParms Maximum values of parameters
int TimeStamp GRIDTIMES Timestamp of time steps
int DayStamp GRIDTIMES Daystamp of time steps
oat NorthLat GRIDTIMES North latitude of domain
oat WestLon GRIDTIMES West longitude of domain
The data grids vary in size The size can b e calculated using the NL values from the array
and are again rounded up to a multiple of four
for each time step
for each parameter
int McFile McIDAS GRID file number with original data
int McGrid Grid number in GRID file
float ga NL ip
float gb NL ip Decompression values for each level ip is
the parameter index
float griddata Data for this grid consisting of Nr Nc NL ip
floats rounded up the nearest multiple of
A VIS D map le format
A VIS D map le holds information of land sea b orders etc as a numb er of continuous
p olylines map segments A map le has the following structure
Numb er of map line segments
int nsegs
Directory structure containing for every segment
int d d d d Not used
APPENDIX A F ILE FORMATS
int GRIDTIMES Maximum numb er of time steps
int NumTimes Numb er of time steps
int NumParms Numb er of physical variables
int Nr Numb er of rows latitude p oints
Nc Numb er of columns longitude p oints
Nl Numb er of levels
oat LatInc LonInc Horizontal grid increments in latitude and
longitude direction in degrees
oat Height Nl Heights of levels in km
char ParmName GRIDPARMS Names of physical variables
oat MinVal NumParms Minimum value of physical variables
oat MaxVal NumParms Maximum value of physical variables
int TimeStamp GRIDTIMES Timestamp of time steps HHMMSS format
int DayStamp GRIDTIMES Daystamp of time steps YYDDD format
oat NorthLat GRIDTIMES North latitude of domain in degrees
oat WestLon GRIDTIMES West longitude of domain in degrees
There is now slightly more data for each grid than in the case of the old style format due to
the extra decompression values
for each time step
for each parameter
if ID x
int McFile McIDAS GRID file number with original data
int McGrid Grid number in GRID file
fi
float ga Nl gb Nl Decompression values for each level
float griddata Data for this grid consisting of Nr Nc Nl
rounded up to the nearest multiple of floats
A Variable NL format
This is the new style format made to enable the visualization of single level elds The header
lo oks like
A COMP FILE FORMAT
Variable NL format new style format with layer by layer compression GRID le read
and a variable numb er of levels for each parameter
A Old style format
Grid compression is done grid by grid entire grid at the time The maximum numb er of time
steps GRIDTIMES is for ID x and for ID x The maximum
numb er of parameters GRIDPARMS is for ID x and for ID x
The following typ e lengths are used char is byte int is bytes float is bytes
The header lo oks like
int NumTimes Numb er of time steps
int NumParms Numb er of physical variables
int Nr Numb er of rows latitude p oints
Nc Numb er of columns longitude p oints
Nl Numb er of levels
oat LatN LonW Geographic lo cation in degrees
oat HeightTop Height of highest grid level in km
oat LatInc LonInc Horizontal grid increments in latitude and
longitude direction in degrees
oat HeightInc Vertical grid increment in km
int DayStamp GRIDTIMES Daystamp of time steps in YYDDD format
Note that the le always contains GRIDTIMES
DayStamps despite the numb er of time steps
int TimeStamp GRIDTIMES Timestamp of time steps in HHMMSS format
char ParmName GRIDPARMS Names of parameters
Then the data follows in the same order as the time stamps and variable names
for each time step
for each parameter
float ga gb Grid decompression values
float griddata Data for this grid consisting of Nr Nc Nl
rounded up to the nearest multiple of floats
A New style format
Compression of the grids is done layer by layer and extra information is contained within the
le to retrace the origin of the grids the McIDAS GRID le and grid numb er in the le to
enable vis d to read the original uncompressed grid data only for ID x The
header lo oks like
APPENDIX A F ILE FORMATS
co ordinate unit to vis d Therefore the eld pad is replaced by
int leveltyp e Indicator of unit of vertical co ordinate system on
which grid is de ned
int pad Reserved now one integer less
The following values for the leveltyp e are currently valid
LEVELS This is the default leveltyp e and set for old HEIGHT
style COMP les The unit of the vertical
co ordinate is km
LEVELS Indicates equidistant pressure levels Vertical PRESSURE
interp olation b etween levels will b e done using the
logarithm of the pressure The unit of the vertical
co ordinate is Pa
The contents of the top and topinc entries must b e changed These will contain the pressure
level of the highest level and the pressure level increment b oth multiplied by if the
leveltyp e is PRESSURE LEVELS Otherwise the contents will b e the same as b efore height
level and height increment
The data grids are stored as byte oating p oints numb ers A data value greater than e
indicates missing data usually the value e is used
A COMP le format
The COMP compressed le format is the only format that can b e read directly into vis d
VIS D is supplied with skeleton example programs for converting data into COMP le format
and utilities for converting the GRID le format see app endix A into COMP les During
this conversion all data p oints in the grids are compressed from byte oating p oints values
to bit integers hence the name of the format Because this compression results in loss of
accuracy it is p ossible to read grids from the uncompressed GRID les when they are needed
for calculations by vis d The compression uses two oating p oint values g and g These
a b
(g +g )
max 250
min
the range and g g the o set values are equal to g
b a a
(g g ) 2
max
min
of the input grid data New compressed data values are then calculated using the formula
g g g g
new a old b
There are several COMP le formats which are identi ed by an ID eld in the rst bytes
in the le The following ID elds are valid
Old style format with grid by grid compression ID s x and x
New style format with layer by layer compression and GRID le read and the new format
A GRID FILE FORMAT
char ident File description identi c ation
int npro j Pro ject numb er
int creation date Date created in YYDDD format
int maxsize Numb er of data p oints in largest D grid
int numgrid Numb er of D grids in this le
int rstgrid Lo cation of rst grid s data as o set
in byte words from start of le
int pad Pad to bytes
The gridinfo structures contain the following elds each
int size Numb er of data p oints words of data
nr nc nl
int nr Numb er of rows latitude p oints
int nc Numb er of columns longitude p oints
int nl Numb er of levels
int iword Lo cation of data as o set in byte
words from start of le
int date Grid date stamp in YYDDD format
int time Grid time stamp in HHMMSS format
int pad Reserved
char name character name of physical variable
char units character unit description
int pad Reserved
int ityp e Always
int latn North latitude
int lonw West longitude
int latinc Latitude increment
int loninc Longitude increment
int pad Reserved
int ihtyp e Always
int top Top height
int topinc Height increment
Because the grid info structure contains the lo cation of the grid data in the le it is p ossible
to create les with a gap b etween the grid info structures and the data This way space can
b e reserved for future headers
For an original McIDAS GRID le the gridinfo structures are padded to bytes in length
by adding
int pad Reserved
In order to make vis d p erform the correct typ e of vertical interp olation dep ending on the
co ordinate system automatically it must know the leveltype i e the unit of the vertical
co ordinate on which the grids are de ned Because the conversion from GRIB to COMP
takes the indirect route via GRID to retain the functionality of the utilities supplied with
VIS D it is necessary to mo dify the GRID format to pass information ab out the vertical
APPENDIX A F ILE FORMATS
predetermined bit map provided by the centre that made the GRIB message The rest of the
section contains the bit map This section is optional
Section The binary data section starts with three o ctets describing the length of the
section The next o ctet contains a ag setting the representation of the binary data Octet
contain the scale factor for the binary data Octet are set to the minimum of the
binary data The following o ctet has the numb er of bits containing each packed value Then
from o ctet on the actual data is describ ed dep ending on the ag values in o ctet
Section The end section ending the GRIB message is always o ctets long and contains
the character co ded message
A ASIMOF le format
The ASIMOF le format was designed as a database le system at KNMI as a standard
interface b etween their numerical mo dels and presentation systems It was adopted as the
standard for the reference HIRLAM system It is used at KNMI many other HIRLAM
institutes use it or are considering its use For more detailed information ab out the format
see the inline do cumentation of the ASIMOF routines of the HIRLAM system The ASIMOF
le format is computer dep endent
A GRID le format
The GRID le format is used by the McIDAS system VIS D is a stand alone subsystem
of McIDAS see Smit GRID les cannot b e read directly by vis d but they can b e
converted to the COMP le format app endix A which is currently the only format vis d
can read There exists a compile time option Read On Demand that enables vis d to read
GRID les but even then the header of the COMP le that is made from the GRID le s
must b e available
The global structure of a GRID le is
header size is bytes
numtimes numparms gridinfo headers size of each is bytes numtimes is the numb er
of time steps numparms is the numb er of physical variables
space for future headers
numtimes numparms data grids each data p oint is a byte oat
The following typ e lengths are used char is byte int is bytes float is bytes
The McIDAS GRID le header consists of the following entries
A GRIB FILE FORMAT
Octet No Contents Description
Numb er of o ctets in section
No vertical parameters
No list present
Rotated latitude longitude grid
Points on parallel
Points on meridian
Latitude rst grid p oint
Longitude rst grid p oint
Direction increments given
Earth assumed spherical
Positive U and V in easterly and northerly directions
Latitude last grid p oint
Longitude last grid p oint
Latitude direction increment
Longitude direction increment
First grid p oint left down b ottom
Latitude southern p ole
Longitude southern p ole
Angle of rotation
Table A GRIB format Section example values for HIRLAM
APPENDIX A F ILE FORMATS
Octet No Contents
Length of section
NV Numb er of vertical co ordinate parameters
PV Lo cation of list of vertical co ordinate
parameters if present or
PL Lo cation of list of numb ers of p oints in each
row if o ctet if present or
if neither are present
Data representation typ e grid typ e
Numb er of p oints along a parallel
Numb er of p oints along a meridian
Latitude of rst grid p oint in millidegrees
Longitude of rst grid p oint in millidegrees
Resolution and comp onent ags
Latitude of last grid p oint in millidegrees
Longitude of last grid p oint in millidegrees
Latitude direction increment in millidegrees
Longitude direction increment in millidegrees
Scanning mo de direction of increments
Reserved set to zero
Latitude of southern p ole in millidegrees
Longitude of southern p ole in millidegrees
Angle of rotation
Reserved for stretched latitude longitude grid
PV List of vertical co ordinate parameters
PL List of numb ers of p oints in each row
Table A GRIB format Section grid description section for rotated latitude longitude
grid typ e as used by HIRLAM
A GRIB FILE FORMAT
Octet No Contents
Length of section
Version numb er currently
Identi cation of centre
Generating pro ces identi cation numb er
Grid de nition numb er according to
Volume B of publication WMO No
Flag signaling if section is included bit
and if section is included bit
Indicator of parameter
Indicator of leveltyp e
Height pressure etc of levels
Year of century
Month
Day
Hour
Minute
Indicator of unit of time range
Numb er of time units
Same as o ctet or time interval undergoing
averaging or accumulation
Time range indicator
Numb er included in average if o ctet or
otherwise set to zero
Numb er missing from averaging or accumulation
Century of reference time of data
Reserved set to zero
Units decimal scale factors
Reserved for future use may not b e present
nn Reserved for originating centre use
Table A GRIB format Section pro duct de nition section
APPENDIX A F ILE FORMATS
Section The indicator section contains o ctets where the rst o ctets are character
co ded as GRIB the next o ctets contain the length of the entire GRIB message and the last
o ctet represents the GRIB edition numb er
Section The pro duct de nition section starts with three o ctets describing the length of
this section The rest of this section contains the information of when where by whom etc
the GRIB message has b een made More sp eci c information ab out this section can b e found
in table A
While working with HIRLAM data we encountered the following level typ es o ctet
Isobaric surface Grid values are given at a xed pressure level Octet
contain the pressure in hPa
Mean sea level Grid values are given at mean sea level Octet are set to
zero
Sp eci ed altitude Grid values are given at a sp eci c altitude Octet contain
the altitude in metres
Sp eci ed height level ab ove ground Grid values are given at on a sp eci c height
ab ove the ground Octet contain the height in metres
Hybrid level Grid values are given at mo del levels see section Octet
contain the level numb er
Section The grid description section starts with three o ctets indicating the length of this
section The rest of this section describ es the grid as set in o ctet numb er of section
Because HIRLAM uses a rotated latitude longitude grid o ctet o ctet contain
sp eci c information for that typ e of grid More information ab out section can b e found in
table A For HIRLAM an example section is given in table A
The rotated latitude longitude co ordinate system that is used by HIRLAM formed by a
general rotation of the earth can b e describ ed by three parameters
The geographic latitude in degrees of the southern p ole
p
The geographic longitude in degrees of the southern p ole
p
The rotation of the new p olar axis p ositive in easterly direction This new p olar axis
can b e obtained by rst rotating the sphere through degrees ab out the geographical
p
p olar axis and then rotating through degrees so that the southern p ole moved
p
along the previously rotated Greenwich meridian
Section The bit map section starts with three o ctets used for the length of this section
Octet is used for giving the numb er of unused bits at the end of section The next
two o ctets are the table reference if they are zero a bit map follows else it refers to a
App endix A
File formats
In this app endix all relevant le formats are explained HIRLAM pro duces les in the GRIB
le format which is the world standard format for storage of meteorological data The
ASIMOF le format is used at KNMI to store several GRIB elds in one le The remaining
le formats are related to VIS D GRID COMP map and top ography le formats
A GRIB le format
GRIB gridded binary as describ ed in FM IX Ext GRIB WMO is a le format for
pro cessed data in the form of grid p oint values expressed in binary form In this paragraph
the GRIB edition format will b e explained
In GRIB the data are represented in a sequence of an even numb er of o ctets o ctet bits
This representation is indep endent of any particular machine representation The o ctets of a
GRIB message are group ed in sections
Section Indicator section
Section Pro duct de nition section
Section Grid description section
Section Bit map section
Section Binary data section
Section End section
Every section contains an even numb er of o ctets If not the section is app ended with bits set
to zero
BIBLIOGRAPHY
Bibli og raphy
Baas Baas S A Meteorological Visualization using apE a to olkit for visualization
Internal rep ort Leiden University
GLman Fisher S M Heinrich Graphics Library programming guide do cument ver
sion Silicon Graphics Incorp orated
Gour Gouraud H Continuous shading of curved surfaces IEEE Computer Graphics
and Applications C pp
Hibb Hibbard W L R Kraus J T Young D Weather Displays Using McIDAS
Preprints ICI IPSMOH Los Angeles American Meteorological So ciety
Hibb Hibbard W D Santek Visualizing large data sets in the earth sciences IEEE
Computer Graphics and Applications pp
Hill Hill Jr F S Computer Graphics Maxwell MacMillan International Editions
LoCl Lorensen W E H E Cline Marching Cub es a high resolution D surface
construction algorithm Pro ceedings Siggraph Computer Graphics
pp
Smit Smith E A The McIDAS System IEEE Transactions on Geoscience Electron
ics GE
WMO Secretariat of the World Meteorlogical Organization Manual on Co des Volume
I Annex II to WMO Technical Regulations International Co des Part B
Binary Co des edition Supplement No Geneva Switzerland
CHAPTER FUTURE WORK
It would b e more realistic to insert missing values b elow the orography instead of
extrap olating the data values This is not appropriate for all variables In general all
full mo del level elds should b e substituted with missing values b elow the orography
but not the pressure at at mean sea level
The vertical co ordinate of the top ography should b e converted from metres to Pascal
Because the pressure is not constant at the earth surface this implies a change to vis d
to calculate a di erent top ography every time step
Logarithmic interp olation of vertical slices is not yet implemented
Enhance VIS D in such a way that the user can determine the place of each colormap
legends in the same way as the lab els can b e p ositioned
For HP hardware platforms at Leiden University the Starbase Graphics Library can b e
tested for use with VIS D Starbase is a HP sp eci c library and is probably faster than
VOGL but not p ortable
Chapter
Future work
The recommendations listed in this chapter can b e used as a guide to future work They
include required features see section for a list that are not yet implemented
Extend the D p oint prob e in VIS D to a D curve prob e A D curve prob e should
enable the user to manipulate a spline curve in the domain b eing visualized The
value of the physical variables are then plotted as a function of the displacement along
the curve
In meteorology wind vectors are often visualized using shafts and various attachments
The shafts line segment are all equal in length and p oint in the opp osite direction of
the wind vector The sp eed of the wind is visualized by attaching several line segment
and triangles to the shaft degrees rotated This can easily b e added to VIS D
When multiple vertical co ordinate units are supp orted it is recommended to print the
unit in the D window near the height axis to avoid misinterpretation of the data A
unit for every physical variable would b e useful to o
A visualization technique not currently supp orted by VIS D is volume rendering Fu
ture versions of VIS D will include this technique but will use transparency blending
features from the graphics library These are currently not supp orted by VOGL and
should b e implemented if the sp eed of VOGL allows it
The visualization of D elds is currently restricted in the way that every D slice is
placed in the b ottom of the b ox Slices can b e visualized for arbitrary height it is the
placement that is always at the b ottom The D elds of the KNMI not p ositioned at
the earth surface include the metre temp erature and metre wind vectors With
the current vertical resolution of the vertical slices it is not useful to mo dify VIS D to
draw the elds at the correct place moreso b ecause only one equidistant set of vertical
co ordinates can b e used all physical variables must b e de ned on the same grid p oints
The lighting mo del implemented in VOGL must b e enhanced For instance sp otlights
are not supp orted at the moment
CHAPTER CONCLUSIONS
on any hardware platform as long as a C compiler is available Note however that in
order to use VOGL with VIS D other system requirements see section must also b e satis ed
Chapter
Conclusions
In order to visualize the output of a numerical weather forecasting mo del HIRLAM the
visualization package VIS D was chosen VIS D is suitable although a numb er of mo di
cations were necessary Based on our exp eriences with VIS D the following conclusions can
b e drawn
Advantages of VIS D are
VIS D has a very intuitive user interface and enables the user to combine all supp orted
visualization techniques and all available physical variables easily into one image
VIS D is free under the GNU licence
the source co de is supplied with VIS D
1
go o d supp ort For the growing list of users a mailing list end e mail to Bill Hib
bard whibbard vms macc wisc edu or Brian Paul bpaul vms macc wisc edu if you
would like to b e added to the VIS D mailing list is available and the authors are very
resp onsive when presented with questions and suggestions
VIS D is still in development which means that future versions will have more func
tionality and keep up to date with resp ect to new visualization techniques
A disadvantage of VIS D is
VIS D relies heavily on the existence of a D graphics library If such a library is not
available VIS D cannot b e used This problem has b een solved for HP Series by
extending the library VOGL software implementation of the SGI GL libraries with
the functionality needed by VIS D However a software implementation of a hardware
library can never match its sp eed so VIS D is a bit slow the reader is referred to
chapter for concrete timing values when using VOGL VOGL by itself can b e used
1
S
CHAPTER RESULTS
lines of co de and ca lines were added
Images showing some of the capabilities of VOGL are included in app endix F
Because VOGL is quite slow a numb er of sp eed tests were executed on a HP RX The
drawing of color slices is tested b ecause these take the longest time to render The following
times resulted
Rendered graphics Time seconds
Empty b ox ca
One horizontal color slice ca
Two horizontal color slices ca
To determine how much time each op eration p erformed by VOGL costs the op erations are
split into the following categories
drawing the drawing of pixels on the screen and setting of colors
scan conversion calculation of which pixels to draw
p olygon clipping
miscellaneous the remaining op erations
The same sp eed tests were run without execution of one of the listed categories which resulted
in the following
VOGL categories Time relative to normal
Normal op eration
No drawing
No scan conversion no drawing
No clipping
No clipping No scan conversion
From this results the conclusion can b e drawn that drawing and clipping are the categories
that should b e considered rst to sp eed up VOGL
Colormap legends are implemented The only restriction is currently the space of the D
window Dep ending on the width of the values printed next to the legends up to ca
legends can b e displayed An image displaying a colormap legend is given in app endix F
Chapter
Results
This chapter presents the results obtained by using VIS D for the visualization of the
HIRLAM output
The required visualization techniques see section are all p ossible with VIS D To obtain
a discretized color mapping the user must form the colored bands by changing the color map
in the RGB widget window of VIS D Wind vector slices can only b e visualized if the U and V
wind comp onents are present if the vertical wind comp onent W is present it is used otherwise
it is substituted with zeros To visualize wind tra jectories all three wind comp onents U V
and W must b e present and more than one time step must b e available Currently HIRLAM
do es not calculate the vertical wind comp onent W although the equations to do so are
present which means that wind tra jectories cannot b e used for the HIRLAM data
Of the additional required features see section only the D data prob e is not yet
available but will b e included in future versions of VIS D
The HIRLAM data can b e visualized using pressure for the vertical co ordinate This is
realized by interp olating the HIRLAM data from HIRLAM mo del levels to pressure levels
and writing the resulting data to a le that can b e further pro cessed by VIS D Due to
the exp onential decrease of pressure with height logarithmic interp olation must b e used For
large pressure values Pa the linear and logarithmic fractions are nearly equal
however
The VIS D mo di cations describ ed in chapter are all implemented succesfully except for
the logarithmic interp olation of vertical slices The source co de of vis d consists of ca
lines of co de The mo di cations add ca lines of co de all mo di cations including the
graphics mo dule mo di cations This numb er of lines is exaggerated b ecause the old co de is
retained and every mo di cation is emb edded within C prepro cessor statements for backward
compatability
VIS D is p orted to the HP hardware platform for use at Leiden University VOGL a software
implementation of the SGI Graphics Library is used for all graphics related functions VOGL
is enhanced with functionality necessary for VIS D The original VOGL consisted of ca
CHAPTER VIS D MODIFICATIONS
the gravitational acceleration at the earth surface The resulting geop otential height cor
resp onds with the geometrical height height ab ove mean sea level The land sea mask
variable leveltyp e height can b e extracted directly from the HIRLAM climate le
The values must b e rounded to the nearest integer is land is sea
VIS D is adapted to display the b ox height co ordinates in Pascal but the vertical co ordinate
of the top ography is still in metres If the vertical co ordinate should b e changed to pressure
co ordinates vis d should also b e enhanced to calculate a new top ography every time step
b ecause it dep ends on the non constant surface pressure
Map conversion
vis d has a built in function to display the map of the world This map consists only of coast
lines not the b orders b etween countries The data are extracted from the default map le
OUTLSUPW which is supplied with VIS D
This map le has the same grid problems as the VIS D top ography le as describ ed in
section the data grid p oints are in a latitude longitude format while the HIRLAM grid
p oints are in a rotated latitude longitude format Because the climate le do es not contain
a map of the HIRLAM area the default map le had to b e converted to the correct grid
In the standard VIS D package the map mo dule handles reading and drawing of the map
This mo dule is the starting p oint to construct a function that makes a map of the HIRLAM
area in the correct grid format
The conversion of grids is p ossible with the ASIMOF library routine
gb gcc pxold pyold pxnew pynew k kb i pb i kb o pb o
where pxold and pyold denote the grid values in the longitude and latitude direction resp ec
tively pxnew and pynew the converted longitude and latitude grid values k is the numb er of
longitude and latitude values kb i and pb i are the values of the integer and real blo ck of
the GRIB section of the input grid and kb o and pb i are the section integer and real
blo cks of the output grid
The HIRLAM map can b e made by typing at the commandline
hirlammap
or
hirlammap input map file output map file
Running hirlammap without commandline arguments assumes OUTLSUPW as the input map
le and HIRLAM MAP as the output map le
SCALING VERTICAL WIND COMPONENT
is visualized on pressure levels logarithmic interp olation must b e used instead The change
necessary to vis d uses the same logarithmic fraction as derived for the horizontal slice
extraction see section
Scaling vertical wind comp onent
When a vertical co ordinate system di erent from metres is used the scaling of the vertical
wind comp onent W must b e adjusted Currently VIS D scales W from the unit m s to
gridb oxes s The unit of the vertical wind comp onent if calculated by HIRLAM is Pa s
instead of m s which VIS D assumes The required scaling can easily b e p erformed by
dividing the values of the vertical wind comp onent by the height of a grid b ox in Pascal
These mo di cations must b e p erformed in the trajectory and work mo dules for the tra jec
tory calculation and horizontal and vertical wind slice calculation resp ectively
Top ography conversion
VIS D has a built in function to display the top ography of the world The VIS D top ogra
phy consists of two items for every grid p oint
Surface elevation height ab ove sealevel
Land sea mask
The word top ography assumes that there is more information than just the height and the
land sea mask available for instance the ow of rivers roads and the use of land The correct
word should b e orography but since VIS D refers to top ography when meaning the orography
this will b e used in the remainder of this section
The top ography le format is describ ed in app endix A The grid of the default top ography
le EARTH TOPO is in a standard latitude longitude format whereas the HIRLAM grid is
a rotated latitude longitude grid With VIS D a sample le maketopo c is supplied that
enables the user to make his own top ography data le This sample le has b een adjusted
to make a correct HIRLAM top ography The top ography le can b e made with the program
hirlamtopo by simply typing
hirlamtopo input climatefile outputfile
The information necessary to create the top ography in the correct grid format can b e ex
tracted from the HIRLAM climate le The height ab ove the ground is calculated by extract
2
ing the geop otential variable leveltyp e height and dividing by G m s 0
CHAPTER VIS D MODIFICATIONS
To intro duce logarithmic interp olation into the iso surface calculation the following equation
must b e mo di ed
isov al ue nodev al
1
f
nodev al nodev al
2 1
where isov al ue is the value for which the iso surface is calculated nodev al and nodev al are
1 2
the data values at the edge no des b etween which is linearly interp olated and f is the fraction
i e the p osition of isov al ue within the interval nodev al nodev al
1 2
In order to use logarithmic interp olation a three step calculation must b e p erformed First
the linear fraction f is calculated using formula Then the pressure value corresp onding
with isov al ue in the given interval can b e found by using the following formula for f
l n P l n P
1 iso
f
l n P l n P
1 2
where P and P are the pressure levels for the edge no des P must b e calculated b ecause
1 2 iso
this value determines the lo cation of isov al ue in the interval nodev al nodev al when inter
1 2
p olating using the logarithm of the pressure levels Solving equation using equation
yields
(ln(P ) f (ln(P ) ln(P )))
1 1 2
P e
iso
Third and last the values of P P and P are used to calculate the following fraction
iso 1 2
P P
1 iso
f
iso
P P
1 2
Fraction f corresp onds to the logarithmic intersection p oint b etween the two no des
iso
The mo di cations necessary to vis d use the ab ove equations in a new macro LOGcalcNode
This macro is used only to calculate intersections of vertical cub e edges
Wind tra jectories
To calculate a smo oth wind tra jectory VIS D linearly interp olated in a D b ox three spatial
and one time dimension The interp olation in the vertical co ordinate should b e p erformed
logarithmically if pressure levels are used The change necessary to vis d uses the same
logarithmic fraction as used for the horizontal slice extraction see section
A second mo di cation is necessary to the tra jectory to scale the vertical wind comp onent
correctly see section
Data prob e
The data prob e is used to obtain single values from the data grids If the data prob e is
p ositioned b etween two levels vis d linearly interp olated b etween the levels When data
LOGARITHMIC INTERPOLATION
where
P the pressure of the upp er level of the adjoining pair
abov e
P the pressure of the lower level of the adjoining pair
below
P the pressure of the level that must b e computed
w anted
The new data values are now calculated in the normal way using the logarithmic fraction
The pressure value P is calculated from the slice p osition and P and P by linear
w anted abov e below
interp olation
Vertical slices
A vertical slice is calculated by horizontal interp olation of the data values nearest to the slice
for every level of data available No interp olation b etween the levels is done so a vertical
slice has the same numb er of data p oints in the vertical direction as there are levels There
are several ways to intro duce the logarithmic interp olation b etween the levels
Extract the slice as normal For a contour slice the contouring algorithm must b e mo d
i ed so that intersections of edges spanning di erent levels are calculated by logarithmic
interp olation For a color slice the graphics library must interp olate the colors b etween
the vertices logarithmically For the wind vector slices no change is necessary b ecause
only the data values extracted from the grid p oints are used
Extract extra levels in b etween the levels Calculate these extra data values in b etween
by the same typ e interp olation as describ ed in section This way the contouring
algorithm do esn t have to b e mo di ed provided the vertical resolution is high enough
The color slices can b e rendered as normal using linear interp olation b etween the ver
tices
The rst metho d cannot b e used b ecause the GL library do es not supp ort logarithmic color in
terp olation the GL man pages refer to p ossible future implementation of this feature though
see GLman Only the second option can b e used but has the disadvantage that more data
p oints are extracted which increases the time needed to calculate and render the slices
Implementation of this metho d is not yet nished
Iso surfaces
The iso surfaces are calculated using the Marching Cub es algorithm LoCl The data grid
is divided into cub es with eight data values at the corners of each cub e For every edge of a
cub e iso surface intersections are determined The iso surface intersects an edge of the cub e
if the values at the cub e no des lie on opp osite sides of the iso value the data value for which
the iso surface is calculated The intersections of all twelve edges together determine the
part of the iso surface that lies within the cub e A D p olygon is generated
The intersections of vertical cub e edges must b e interp olated logarithmically whilst the in
tersections of horizontal cub e edges must b e interp olated linearly as normal
CHAPTER VIS D MODIFICATIONS
Floating levels D elds at a certain altitude are not allowed All single level elds
are assumed to b e surface elds
The levels must still b e equidistant
Each physical variable must have the same numb er of levels in each time step
For a detailed description of the changes see app endix D
Logarithmic interp olation
The co ordinate system chosen for visualization uses the pressure as the vertical co ordinate
unit see section for a description of p ossible co ordinate systems Because in nature
the pressure decreases exp onentially as a function of the altitude vertical interp olation of
the physical variables b etween the data levels must b e p erformed using the logarithm of
the pressure level values Currently vis d can only interp olate linearly Because the correct
interp olation metho d to use dep ends on the co ordinate system an extra entry in the vis d
COMP and GRID le format is necessary an integer leveltype The GRID and COMP
le formats and the mo di cations are describ ed in app endix A The leveltype variable
indicates what kind of vertical co ordinate system is used Currently these can b e height
and pressure values In the following sections the changes to vis d are describ ed Only the
calculation of graphics and the data prob e i e the extraction of data from the grids has to
b e mo di ed In the following sections the mo di cations are explained
Horizontal slices
When the user selects a horizontal slice for a level that is not in the data grid but b etween
existing levels vis d calculates the values on the requested slice by interp olation of the data
values of the two adjoining levels The current interp olation function of vis d calculates
linear fractions f and f which are used to calculate the data values on the requested
slice
data f data f data
v alue below abov e
where
data data value on lower adjacent slice
below
data data value on upp er adjacent slice
abov e
This formula is evaluated for every grid p oint on the horizontal slice
The most easy way to p erform the logarithmic interp olation is to add an alternative way to
compute the fractions The logarithmic interp olation can b e expressed using the following
formula for the logarithmic fraction
l n P l n P
below w anted
f
log
l n P l n P
below abov e
VISUALIZATION OF D FIELDS
No redundant data
Disadvantages
Only a color mapping can b e visualized no contour or wind vector slices
Mo di cation of VIS D is necessary
The D elds must b e stored in separate les
Recoloring the top ography with a physical variable can b e confusing to the user
who loses his p erception of the actual height of the top ography
Mo dify VIS D to eliminate the restriction of the trivial height dimension
Advantages
No redundant data
All visualization techniques as far as useful are available for each physical vari
able Visualization techniques that are not applicable to D elds can b e turned
o in the control window
User friendly Everything concerning grid dimensions can b e hidden from the user
but the user can never forget the numb er of levels due to the restricted slice
movement etc vis d can simply prohibit op erations that are illegal for D
elds
All combinations of physical variables can b e visualized simultaneously
Each eld has an unique button row in the control window and can thus b e easily
identi ed
Disadvantages
Requires the most work of all available solutions
The complexity of VIS D increases
After examination of the source co de of VIS D it was decided to cho ose solution b ecause
It is the most elegant and user friendly metho d
No redundant data needs to b e stored This is quite imp ortant b ecause vis d already
needs a lot of memory it is wise to use it sparingly
Various changes to VIS D were already necessary for other reasons see section
The amount of work was limited in the following ways
Only the trivial dimension height the numb er of levels is eliminated In other words
all data grids must still have the same numb er of rows and columns The numb er of
rows and columns must each b e at least
CHAPTER VIS D MODIFICATIONS
Mo di cation of VIS D is not necessary
Disadvantages
The user has to keep track of where the D elds are p ositioned in the D grid
A horizontal slice must b e p ositioned exactly on the real D eld otherwise the
D eld is interp olated with a neighb ouring D eld which is incorrect This
cannot b e solved by inserting dummy levels with missing data indicators b etween
each pair of levels b ecause the slice interp olation will then return missing data
values for a slice adjacent to the missing value slice
Vertical slices and iso surfaces are incorrect but can still b e visualized
For each time step dummy values must b e used to ll the D grid in which the D
elds are stacked if there are not enough D elds to ll the D eld completely
A separate physical variable is used for every D eld and all levels in a D grid are
set to the same values as in the D eld
Advantages
All physical variables can b e viewed simultaneously
Mo di cation of VIS D is not necessary
The slice p osition is not relevant all p ositions will result in the correct values b eing
obtained
Disadvantages
A lot of redundant data
Vertical slices and iso surfaces are still meaningless and incorrect
Split the physical variables to visualize into two separate COMP les one with the D
elds and another one with the D elds The D elds are duplicated so that they
are two levels high eliminating the trivial height dimension of one Visualizing of the
D elds constructed this way can b e done as normal
Advantages
Not many redundant data all D elds are only duplicated
Vertical slices and iso surfaces are incorrect but less confusing b ecause of the
reduced volume of data
Mo di cation of VIS D is not necessary
Disadvantages
Combinations of D and D elds cannot b e visualized
Color the top ography using one of the required D elds The D elds are stored in
a separate le
Advantages
Chapter
VIS D mo di cations
In this chapter the mo di cations to VIS D are discussed which were necessary to make
VIS D able to visualize the HIRLAM output The following mo di cations are discussed
D eld visualizations VIS D could only cop e with D elds
logarithmic interp olation of the vertical co ordinate when pressure levels are used
creation of top ography and map les of the current domain of HIRLAM
creation of colormap legends
The D prob e mo di cations have not yet b een studied
Visualization of D elds
In order for VIS D to b e useful for visualizing the HIRLAM data a solution must b e found
to visualize D elds VIS D has b een written sp eci cally for the visualization of D data
elds Fields with only one row column or level cannot b e visualized Another restriction is
that every physical variable must have the same numb er of p oints in all four dimensions
Possible solutions to visualize D elds with VIS D are
Stack all D elds which must b e visualized in D data elds and ll the remaining
levels with arbitrary dummy values
Advantages
All physical variables can b e viewed simultaneously rememb er that multiple vi
sualizations of one physical variable are p ossible by cloning the variable using the
NEWVAR option
CHAPTER INTERFACING HIRLAM WITH VIS D
GRID TO COMP CONVERSION
GRID to COMP conversion
In order to visualize the GRID les created by xg v they must b e converted to COMP les
which can b e read by vis d This can b e done with the utility comp d which is supplied with
VIS D Due to the changes to VIS D which will b e describ ed in chapter the standard
comp d program can not b e used anymore b ecause it terminates with an error if the data
grids in the input GRID le s do not have equal dimensions To imp ort the new GRID les
a new program varNLcomp d has b een derived from the existing comp d program with the
following mo di cations
When reading the rst time step the numb er of levels and height of each grid in the
time step is stored in an array and kept for later use
When reading the remaining time steps the numb er of levels and height of each grid is
compared with the appropriate value from the arrays
The compression is mo di ed to handle variable sized grids
CHAPTER INTERFACING HIRLAM WITH VIS D
Figure window with list of data elds
CONVERSION FROM GRIB TO GRID
Starting xg v will op en the main data le window see gure In the top of this window
the following buttons are available
Options Op ens or closes the options window see gure In this window the default
conversion options can b e mo di ed
Create Create the output le as selected by the user
Exit Exit xg v
Under this row of buttons the window displays a list of les found in the sp eci ed search
directories and matching the sp eci ed lename pattern
The lenames are sorted in chronological order which results in the time steps b eing written
to the output le in the correct order Each data le has a separate button which are all
turned o by default The left mouse button can b e used to select and deselect data les A
highlighted button will mean that the corresp onding data le is included in the output le
The middle mouse button is used to op en close a new window displaying the contents of the
corresp onding data le as a list of data elds see gure The names of the data elds
consist of the following information
GRIB variable numb er according to WMO sp eci cations
A short name of the variable
Leveltyp e numb er according to WMO sp eci cations
Levelrange
Fields of equal variable numb er and leveltyp e are group ed under one button Using the left
mouse button data elds can b e selected and deselected The data elds are read from the
le the rst time the user clicks on a data le button in the main window
xg v will make sure that the selection of les and elds remains consistent by
Prohibiting the selection of new data les that do not contain all the selected data elds
Prohibiting the selection of new data elds that are not present in every selected data
le
If such a selection is attempted by the user xg v will rep ort this in the standard output
window
CHAPTER INTERFACING HIRLAM WITH VIS D
Figure main window of xg v
Figure options window of xg v
CONVERSION FROM GRIB TO GRID
More information ab out the RUL libraries can b e found in app endix C
The RUL library functions were implemented in C except for the functions in futil f for
the following reasons
C provides dynamic memory allo cation
C provides b etter string handling than FORTRAN
The routines in futil f are provided b ecause a FORTRAN common block cannot b e refer
enced from C
In order to call FORTRAN functions from a C program the following should b e noted
FORTRAN passes the address of every function parameter Therefore when calling a
FORTRAN function from C of every parameter that is not a p ointer the address must
b e passed
FORTRAN has no string termination character When a character string is passed to
a FORTRAN function exp ecting a CHARACTER the string length must b e passed as
an extra parameter after the list of remaining formal parameters
Arrays in C are row ma jor opp osed to arrays in FORTRAN b eing column ma jor
The index of the rst element in a C array is in FORTRAN this is
Conversion program
The program xg v is resp onsible for the conversion of GRIB elds stored in ASIMOF les to
McIDAS GRID les xg v must b e started with the command
xg v options
The following options are available
Option Description Default
le Sp ecify wild carded input lename pattern fc
path Sp ecify comma separated list of directories
out Override the default output lename GR D
climate le Override the default climate le cl
pro jectname Override the default pro ject name grib vis test le
pro jectnumb er Override the default pro ject numb er
outputleveltyp e Override the default output level typ e
outputlevelrange Override the default output level start and
end separated by a minus sign
numoutputlevels Override the default numb er of output levels
v Increases the amount of information printed
to the screen during conversion
CHAPTER INTERFACING HIRLAM WITH VIS D
util Provides routines for date conversion and ASIMOF information programs
vari Provides conversions from one co ordinate system into another as well as routines to
convert U and V to several co ordinate systems
prp o Provides routines used for p ostpro cessing vertical interp olation and destaggering of
the wind comp onents of the data
grw Provides functions that are used as an interface to ASIMOF les The DDR data
description record is the same as the keys of a GRIB message blo ck and blo ck The
parameters are kept in a common FORTRAN blo ck
More information ab out the KNMI libraries can b e found in app endix C
Using the KNMI libraries not only reduces the work necessary but also ensures future com
patibility if the KNMI libraries are changed or extended
RUL libraries
The RUL libraries were created to provide an interface b etween the KNMI libraries and
the main GRIB to VIS D conversion program xg v X Windows GRIB to VIS D for the
following reasons
The KNMI libraries use very cryptic function and variable names
The function interfaces of the KNMI libraries are sometimes very complex function
parameters is no exception The RUL libraries hide unnecessary detail but still retain
enough exibility to provide the same functionality as the KNMI libraries The param
eters that need not b e altered are placed in a C struct which is passed to the functions
in the RUL libraries
The following RUL mo dules are implemented
asimof Handles the op ening and closing of ASIMOF les and the extraction of data from
these les
grib Provides functions to convert b etween GRIB physical variable names and numb ers
Also a unit and short name is de ned for every physical parameter b ecause the GRID
le format reserves only characters for the names of the physical variables
interp ol Provides functions to interp olate GRIB data from levels to pressure levels
cutil Provides general purp ose utility C functions
futil Provides general purp ose utility FORTRAN functions
xg v Main program mo dule providing a X Windows based user interface for the conversion
CONVERSION FROM GRIB TO GRID
level upwards This means that when a physical variable is extracted with more than one
level from an ASIMOF le the levels must b e sorted b efore writing them to an output GRID
le Because each GRIB message contains one level is it most easy to extract the levels
in the correct order This has the additional advantage that it is easy to extract elds in
reverse order which is necessary for vertical co ordinates such as pressure where the p ositive
co ordinate direction is downwards The HIRLAM mo del levels are also de ned downwards
level is the top level
For each eld read from an ASIMOF le two additional reordering op erations must take place
First the array has to b e converted from column ma jor to row ma jor b ecause the eld is read
by a FORTRAN function and is further pro cessed by a C function The second op eration is a
mirror op eration around the horizontal axis b ecause the origin of the HIRLAM domain is the
south west b ottom grid p oint while VIS D exp ects the origin to b e the north west b ottom
grid p oint
Not all physical variables within HIRLAM are calculated on the same grid p oints Some
variables are translated a half grid p oint distance with resp ect to the grid for which the
surface geop otential is de ned this grid is used as a reference grid for the remaining elds
b ecause the top o is derived from it and used as reference by the user Because VIS D can
only cop e with variables de ned on equal grid p oints the physical variables will b e lineair
interp olated to the reference gridp oints read from the surface geop otential eld This
pro cess is called destaggering
After the data elds are extracted and converted they are written to a GRID data le The
grid info structure which contents is completely known after the conversions is written also
Since single level elds cannot b e vertically interp olated they will b e written to the GRID
le without b eing interp olated but reordering mirroring and destaggering are p erformed on
al l elds
Variables that are already interp olated to pressure levels will also b e directly written to the
GRIB le
Conversion from GRIB to GRID
KNMI libraries
The KNMI has several libraries available that provide functions to manipulate the ASIMOF
les to p erform various co ordinate transformations to transform time and date formats etc
The following libraries are available
p ort Provides routines used for the ASIMOF le handling
gco d Provides GRIB related routines such as enco ding and deco ding retrieving sections
from GRIB messages printing these sections and checking the parameters of the sections etc
CHAPTER INTERFACING HIRLAM WITH VIS D
The pressure decreases exp onentially with the height and therefore cannot b e in
terp olated linearly in height
The vertical velo city must b e converted from Pa s to m s in order to keep the
tra jectory mo dule functioning correctly
Extrap olating variables b elow the orography takes sp ecial care
Considering the advantages and disadvantages of these three co ordinate systems the chosen
vertical co ordinate representation in VIS D is the pressure level co ordinate system b ecause
it is the representation that corresp onds b est with the meteorological practice As a result
this means that leveltyp es other than pressure levels will have to b e interp olated to pressure
levels
ASIMOF to GRID conversion sp eci cation
In order to b e able to interp olate variables to pressure levels the following physical variables
must b e present in every ASIMOF le
Pressure at ground level
Temp erature at ground level
Temp erature at mo del full levels
U comp onent of the wind at mo del full levels
V comp onent of the wind at mo del full levels
Sp eci c humidity at mo del full levels
In addition the geop otential at ground level variable numb er level typ e height
must b e read from a climate le which is an ASIMOF le and contains the average of several
variables over a long p erio d usually a year of time
If one of these variables is not available it is imp ossible to interp olate to pressure levels and
the conversion program will stop with an error message
Furthermore the full mo del vertical level parameters A and B are needed to interp olate cor
rectly A and B represent the contribution of the orography and pressure to the co ordinate
resp ectively Between the full mo del levels of HIRLAM half mo del levels are de ned There
is one half level more than there are full levels The lowest and highest half levels corresp ond
with the b ottom and top of the domain of HIRLAM resp ectively The half mo del level pa
rameters can b e calculated from the full mo del level parameters From the mo del half and
full level parameters the pressure at half and full mo del levels can b e derived
The ASIMOF le format imp oses no ordering on the GRIB elds contained within the le
However VIS D exp ects the grids in the GRID and COMP les to b e given from the b ottom
VERTICAL COORDINATE
The largest change in pressure p er height are in the lowest levels of the mo del
This results in the highest resolution in the area of most interest
Pressure levels corresp ond b est with meteorological practice
Disadvantages
The orography i e the surface pressure is not constant in time
The vertical interp olation of variables must b e done in the logarithm of the pressure
b ecause of the exp onential decrease of the pressure with height
The p ositive co ordinate direction is downwards
Sp ecial care must b e taken to extrap olate variables b elow the orography
Model level numbers The HIRLAM mo del levels are used directly as an equidistant
co ordinate system
Advantages
The most simple of all co ordinate systems
The levels are equidistant so no initial interp olation is needed
No orography is needed at orography b ecause the b ottom mo del level follows
this orography
Best representation for a mo del b ehaviour study
Disadvantages
The vertical velo city is de ned as the change in pressure Pa s This has to b e
transformed to a velo city expressed in mo del levels p er second to b e able to use
the tra jectory mo dule
The p ositive co ordinate direction is downwards
Vertical interp olation of the physical variables cannot b e done linear but must b e
done in the logarithm of the pressure
The e ects of the orography are b eing masked b ecause the used orography is at
Geopotential height levels with the same geop otential i e the same p otential energy of
an air particle
Advantages
Easy to calculate using the HIRLAM p ost pro cessing package postpp
The orography is constant and de ned as the geop otential height of the surface
All variables except the pressure can b e interp olated in height
The p ositive co ordinate direction is upwards
Disadvantages
The levels are not equidistant The highest resolution is given and needed at
the surface of the earth Interp olation to equidistant levels will result in loss of
essential information in the most interesting part of the atmosphere
CHAPTER INTERFACING HIRLAM WITH VIS D
Only a single conversion step resulting in quicker conversion and less disc space
usage
Two step conversion First convert GRIB data elds to a GRID le then convert the
GRID le to a COMP le with the already existing VIS D utilities This has the
following advantages
The VIS D package includes several utilities that allows the user to p erform var
ious op erations on GRID les such as interp olation b etween di erent time steps
resampling of the grid data etc see section for an overview of available
utilities
The vis d program includes prepro cessor statements RD ON DEMAND which en
ables the read on demand le access feature vis d will read the COMP le header
information as usual but when a data grid is requested the data is read from the
GRID le with the original uncompressed data if p ossible The advantage of this
feature is that the data can b e visualized without the loss of accuracy that hap
p ens when converting GRID to COMP data values are compressed from float to
one byte integer see also app endix A
Considering the two p ossibiliti es the choice was made to convert the HIRLAM output to the
GRID le format b ecause in that way the functionality of the VIS D utilities and direct
GRID le access remains available
Vertical co ordinate
The vertical co ordinate used in the HIRLAM system is the so called coordinate The
co ordinate system also referred to as hybrid co ordinate system combines the orography
surface elevation with pressure levels Close to the ground the levels follow the orography
Going upward the relative contribution of the orography decreases while the relative contri
bution of the pressure increases The relative amounts of each contribution are given for each
level by the HIRLAM mo del vertical parameters Furthermore the levels of the co ordinate
systems are not equidistant The largest resolution is close the ground which is the part
of the atmosphere of most interest In order to visualize the HIRLAM data with VIS D a
co ordinate system must b e used with equidistant levels b ecause VIS D can not cop e with
non equidistant levels
In general p ossible vertical co ordinate systems for visualization are
Pressure levels levels given at sp eci c pressure values
Advantages
Pressure levels can easily b e calculated from the surface pressure and the vertical
mo del parameters which are b oth directly available from the GRIB messages
The vertical velo city is correctly de ned in Pa s
Chapter
Interfacing HIRLAM with VIS D
HIRLAM output
The HIRLAM output data les are available in the ASIMOF le format see app endix A
ASIMOF les consist of one or more GRIB messages see app endix A with a directory
structure giving information ab out the lo cation and characteristics of the GRIB messages
The GRIB messages in the le contain the data Each GRIB message can contain information
and data of one level eld of a physical variable Information such as time and date stamp of
the eld level typ e and value are included
The ASIMOF data or history les have a name like fc pp where
fc le typ e fc forecast an analysis in initialised analysis
start of forecast year month day hour
length of forecast
pp status identi er pp p ostpro cessed not p ostpro cessed
In the following sections the conversion from the ASIMOF le format to the COMP le format
will b e discussed
Conversion to GRID versus COMP
To view the HIRLAM data using VIS D a conversion from the ASIMOF le format the
GRIB database le format to the VIS D COMP le format is necessary Two choices can
b e made for converting GRIB data to the COMP le format
One step conversion Convert GRIB data immediately to the COMP le format This
has the following advantages
CHAPTER PORTING VIS D
MISCELLANEOUS PROBLEMS
Fortran function call a version with and one without an underscore attached to the
function name is available
Due to di erences in the X versions on the SGI IBM Stellar and the HP it is neces
sary to set the new p osition and size of a window when it is op ened moved or resized
using X Windows SizeHints
The rest of the p ort is just a matter of following the instructions in the p orting guide supplied
with VIS D See app endix D for a description of HP p ort mo di cations
CHAPTER PORTING VIS D
Scan conversion
Because every draw op eration must b e p erformed on a p er pixel basis for correct functioning
of the Z bu er lines and p olygons must b e scan converted by VOGL and for every pixel the
Z bu er must b e checked Two typ es of scan conversion are necessary Line scan conversion
and p olygon scan conversion Line scan conversion is implemented using the Bresenham line
drawing algorithm see Hill Two versions are supplied The rst is just a pure line
drawing algorithm The second takes extra color parameters for the start and endp oint of
the line The line color is interp olated linearly b etween these values Polygon scan conversion
is more di cult The algorithm used is a variant of the p olygon scan conversion algorithm
as describ ed in Hill For each scanline the intersections with the p olygon edges are found
and sorted in increasing x values Then for each pair of sorted x values a run of pixels is
lled The color of a run is determined by the color of the start and endpixels of the run
Colors along the run are linearly interp olated b etween these values Start and endcolors of
each run are found by linearly interp olated the vertex colors over the p olygon edges
Due to the manner of geometry sp eci cation of the GL p oints and colors are sp eci ed by a
sequence of function calls a change to VOGL is necessary so that vertices are shaded as so on
as they are sp eci ed b ecause the current color and normal used for shading are changed by
the next color or normal sp ecifying function call
VOGL device drivers
The device dep endent parts of VOGL are separated from the rest of the co de into device
drivers A device driver gives VOGL access to certain device sp eci c features such as draw
ing keyb oard and mouse read op erations device initialization and clean up The use of
device drivers makes VOGL very p ortable b ecause only the device driver needs to b e altered
when p orting VOGL to a new system The structure of a VOGL device driver is given in
app endix E
Miscellaneous problems
The following miscellaneous problems were solved when p orting VIS D to HP
The C prepro cessor used at Leiden University release do es not understand elif
which has b een changed to else if statements
The HP FORTRAN compiler do es not understand the typ e length combination
INTEGER the typ e BYTE must b e used instead
For compilers on some systems it is necessary to attach an underscore to function
names of Fortran functions that are called from C This is not the case for HP
Series systems A new prepro cessor symb ol has b een used underscored and of each
D GRAPHICS LIBRARY
Sp ecular
f
I I r u u
sp s s r v
where I intensity of sp ecular re ected light
sp
I intensity of light source
s
r material sp ecular re ectance co e cient
d
u normal vector p ointing in the direction of
r
re ected light R
u normal vector p ointing in the direction of the
v
vertex to viewp oint vector
f shininess co e cient
Because the cosine of the angle b etween two vectors is prop ortional to the dot pro duct b e
tween their normalized versions the latter is used to avoid cosine calculations This has the
consequence that all vectors used in the calculations must b e normalized Normalization of
normal vectors is necessary if the supplied normal vectors do not have unit length or if the
current Mo delView matrix which is used to transform the normal vectors to eye co ordinates
b efore doing lighting calculations is not orthonormal The GL can b e noti ed by the user
when to normalize normal vectors
Two shading mo dels have b een added to VOGL FLAT and GOURAUD shading Gour
When FLAT shading is used the shade of the last vertex of a geometry determines the shade
of the whole geometry When GOURAUD shading is used each vertex of a geometry is
shaded and colors are interp olated linearly b etween the vertices
All lighting functions have b een implemented in a new mo dule called light The functions of
the light mo dule are listed in app endix E
Depth cueing
Depth cueing has b een added to VOGL b ecause while working with vis d without depth
cued lines in combination with orthogonal pro jections it b ecame apparent that it is very
hard to determine the orientation of the b ox After implementation of the Z bu er the only
mo di cations necessary were to implement functions to set the depthcueing color and Z range
and to turn depthcueing on or o These are describ ed in app endix E
To simplify the implementation of the depthcueing there are currently two restrictions to
the depthcueing First only lines can b e depth cued Second lighting and depthcueing are
mutually exclusive that is when depth cueing is on lighting commands are ignored In fact
they are not ignored but the colors calculated by the lightmo del are overwritten by the
depthcueing color
CHAPTER PORTING VIS D
eye
n n light source v light source θ s r s
diffuse reflection specular reflection (highlights)
Figure Vectors in the lighting calculations
Light source color scaled by the material di use re ection and a function of the angle
b etween the vertex surface normal and the vertex to light source vector
Light source color scaled by material sp ecular re ectance and a function of the angle
b etween the vertex to viewp oint vector and the re ected light vector
For each light source two typ es of re ection due to the light source incident light are usually
computed
Di use re ection scattering This o ccurs when some of the incident light slightly
p enetrated the surface and is reemitted uniformly in all directions
Sp ecular re ection These kind of re ections are more mirrorlike and highly directional
These are the re ections that make the surface lo ok shiny
The various vectors used for lighting calculations are depicted in gure n is the surface
normal vector s is the vertex to light source vector v is the vertex to viewp oint vector and
r is the direction of re ected sp ecular light
Di use and sp ecular re ection can b e expressed using the following formulas
Di use
I I r u u
d s d s n
where I intensity of di use re ected light
d
I intensity of light source
s
r material di use re ectance co e cient
d
u normal vector p ointing in the direction of s
s
u normal vector p ointing in the direction of n n
D GRAPHICS LIBRARY
A typical GL lighting session go es as follows
De ne prop erties for each material light and lighting mo del wanted
Activate by binding to one of the eleven resources the de nitions
Draw the scene providing the surface normals and vertices for all primitives to b e
lighted
The reader is referred to GLman for more details In the following sections the enhancements
made to VOGL are discussed
Z bu ering
The use of a Z bu er has a great impact on the current structure of VOGL The calculation
has to b e done on a p er pixel basis b ecause every pixel of a geometry can b e b ehind or
b efore the existing geometry VOGL uses high level drawing primitives to draw lines and
lled p olygons which are implemented in the device drivers Drawing of these primitives
must now b e done on a p er pixel basis so the device driver routines cannot b e used anymore
Instead a new device driver routine must b e added to draw a single pixel this routine did
not exist a single pixel was drawn by drawing a line with equal start and endp oints
The Z co ordinate used for Z bu ering is the screen Z co ordinate of the pixel This value is
obtained by a new function which maps a vertex in clip co ordinates to screen Z co ordinates
using the current Z bu er depth range A new typ e typedef zbuftype is added to VOGL
This typ e is used for each entry in the Z bu er Currently an bit Z bu er is used but
this can b e changed easily by changing the zbuftype typedef and the de ne for MAXZ the
maximum p ossible value to store in the Z bu er
The reader is referred to app endix E for a description of implemented functions
Lighting
GL lighting calculates the color at each lighted vertex by summing the total ambient light
scaled by the material ambient re ectance the material emitted light and the contributions
of each light source
The total ambient light is the sum of the ambient light asso ciated with each light source and
the ambient light asso ciated with the scene as given by the lighting mo del
The contribution of each light source is the sum of
Light source ambient color scaled by the material ambient re ectance
CHAPTER PORTING VIS D
GL lighting p erforms the shading calculation on vertices in eye co ordinates This has the
consequence that GL lighting calculations are only p ossible in multi matrix mode and not in
single matrix mode
GL lighting is controlled by de ning various instances of three typ es of resources
materials surface prop erties
light sources
lightmo del environment prop erties
These de nitions can then b e b ound to one of eleven light control ling resources a material
a backmaterial up to eight light sources and a lighting mo del GL lighting is inactive if no
material or no lighting mo del has b een b ound
A material de nition consists of the following prop erties
transparency
ambient re ectance
di use re ectance
material emitted light
sp ecular re ectance
sp ecular exp onent shininess
A light source de nition consists of the following prop erties
ambient light asso ciated with the light source
light source color
light source p osition
sp otlight direction if the light source is a sp otlight
exp onent and spread of sp otlight cone
A lighting mo del de nition consists of the following prop erties
ambient light asso ciated with the entire scene
various attenuation factors to reduce direct light on ob jects and make the light intensity
vary prop ortional to the ob ject light source distance
D GRAPHICS LIBRARY
Z bu ering
A Z bu er is a set of integers asso ciated with pixels in the framebu er The integers represent
the distance of the pixel to the viewp oint When Z bu ering is activated for each pixel x y
to draw the Z bu er is checked to see if the new pixel is closer to the eye than the pixel already
present at p osition x y If the new pixel is closer the pixel is drawn and the Z bu er is
up dated VOGL do es not supp ort Z bu ering rendering
Depthcueing
When lo oking at ob jects in real life the eye s ability to p erceive depth gives awareness of
the D nature of ob jects and judgement of the relative distance of ob jects The illusion of
depth p erception can b e created on D screens by depth cued images Depth cueing mo di es
an ob ject s color based on the distance to the viewer The screen Z co ordinate is used to
determine the depth of an ob ject This value is translated to a corresp onding Z bu er value
By sp ecifying a color range and depth range the user can determine what color the Z values
maps to For a prop er depthcueing e ect ob jects farther from the viewp oint should b e colored
darker Depthcueing is not supp orted by VOGL
RGB and colormap mo de
The GL supp orts two colormo des RGB and colormap mo de In colormap mo de a Color
Lo okUp Table LUT is used Colors are sp eci ed by entries in the LUT single integer
values The LUT contains the red green and blue color values In RGB mo de true color red
green and blue values are sp eci ed directly Not all op erations of the library are p ossible in
colormap mo de the most p owerful of the two is RGB mo de Other op erations need sp ecial
actions For example when drawing depthcued lines in colormap mo de the LUT must rst
b e loaded with a color ramp b efore the line can b e displayed correctly VOGL only supp orts
colormap mo de but VIS D needs RGB mo de b ecause all colors in VIS D are represented
as bit color
GL lighting
In GL lighting as in the real world the characteristics of the light source determine the color
and intensity of incident light on a surface The characteristics of the ob ject geometry and
its surface material determine the direction intensity and color of re ected light How light
is re ected dep ends on the interaction b etween the incident light and the surface material
The interaction b etween light and matter is far more complicated than can b e simulated in
real time Lighting in the GL achieves a balance b etween realistic app earance and real time
drawing sp eed The GL achieves this balance by p erforming lighting calculations only at
geometry vertices rather than calculating lighting for each pixel rendered
CHAPTER PORTING VIS D
The eye coordinate system is the result of transforming through matrix multiplication the
geometry co ordinates by the contents of the mo deling and viewing ModelView matrix The
eye co ordinate system is the system in which lighting calculations are p erformed internally
Points in eye co ordinates are transformed to clip coordinates by matrix multiplication with
the Projection matrix The next system is the Normalized coordinate system which is the
result of limiting x y and z to the range w x y z w clipping and dividing x y and
z by w The result are normalized co ordinates in the range x y z The space
of normalized co ordinates is called the D unit cub e The x and y co ordinates of this D
unit cub e are scaled directly into the Window coordinate system The pixel at the lower left
corner has window co ordinates
Window co ordinates mo di ed by a window o set that represents the window s lo cation on
the screen represents the Screen coordinate system which corresp onds to pixel values Screen
co ordinates are typically thought of as D but in fact all three dimensions of the normalized
co ordinates are scaled The screen Z co ordinate can b e used for hidden surface removal and
depth cueing
The co ordinate systems are all present in VOGL except for the eye co ordinate system this
is skipp ed b ecause the Mo delView and Pro jection matrices are concatenated The eye co
ordinate system is necessary for lighting calculations Furthermore the incorrect term World
coordinates is used instead of Object coordinates In fact if the Mo delView matrix would
b e separated in a Mo del matrix and a View matrix the co ordinate system b etween these
matrices would b e correctly referred to as the World coordinate system
Transformation matrices
The graphics system of the GL maintains three transformation matrices the Mo delView Pro
jection and Texture matrices The Mo delview and Pro jection matrices are already discussed
The Texture matrix transforms texture co ordinates directly from ob ject to clip co ordinates
Its transformation is unrelated to that sp eci ed by the Mo delView and Pro jection matrices
Op erations on the matrices are p erformed dep ending on the current matrix mode Four matrix
mo des are available In single matrix mode only one transformation matrix is maintained
which transforms p oints directly from ob ject to clip co ordinates This is the only matrix
mo de available in VOGL but cannot b e used when lighting calculations are necessary b ecause
these must b e done in eye co ordinates The other three matrix mo des are viewing projection
and texture matrix mo de which sp ecify the matrix that is mo di ed by op erations such as
loadmatrix Other functions will always a ect the Pro jection matrix disregarding the
current matrix mo de Examples are perspective and ortho which are used to de ne the
view volume that is mapp ed to the D unit cub e Multi matrix mo de not available in VOGL
is necessary to obtain ob jects in eye co ordinates which are used for lighting calculations A
more detailed description is b eyond the scop e of this rep ort the reader is referred to GLman for details
D GRAPHICS LIBRARY
Object coordinates
ModelView matrix
Eye coordinates
Projection matrix
Clip coordinates
Divide by w
Normalized coordinates
Viewport transform
Window coordinates
Window offset
Screen coordinates
Pixel values
Figure Co ordinate systems used by the SGI GL
An intro duction to the SGI GL
In this section parts of the existing SGI Graphics Library GL are discussed that are imp or
tant in relation to the mo di cations and enhancements that are necessary to VOGL
Geometry sp eci cation
Sp eci cation of the geometry to draw is done by issueing function calls that each send a
vertex p osition vertex normal or vertex color to the GL For instance when a line is drawn
the GL must rst b e set to line drawing mo de After that separate function calls are issued
to set the color supply normals and to supply vertices which are used to draw and color or
shade the line After the line is completely sp eci ed the line drawing mo de must b e unset
Geometries drawn this way include lines triangle meshes and quadrilateral meshes
Co ordinate systems
The co ordinate systems used by the GL are imp ortant b ecause each function acting up on
geometry do es so in one of the available co ordinate systems The co ordinate systems used
by the GL are displayed in gure It starts with a D system de ned in right handed
cartesian oating p oint co ordinates called the object coordinate system Geometry vertices
sp eci ed by the user as x y z triplets are in this co ordinate system There are no limits to
the size of co ordinates in this system other than the largest legal oating p oint value
CHAPTER PORTING VIS D
b e written providing functions to draw lines text set color etc For a description of a
VOGL device driver the reader is referred to app endix E
VOGL resembles the SGI GL the subset implemented so mo di cations to VIS D are
hardly necessary
VOGL February has enough functionality to give an adequate display
of contour slices and wind tra jectories
VOGL is free and can b e mo di ed and used for every purp ose as long as the original
COPYRIGHT notice remains intact
Disadvantages of VOGL are
Slow compared with the hardware SGI GL The sp eed of the hardware GL can of
course never b e matched by VOGL This is made worse by the trade o b etween p orta
bility and sp eed Every geometry is drawn by device driver calls To keep VOGL
p ortable device dep endent optimalizations should only b e used in the device drivers
and not in the main co de section
Not all GL functionality used by VIS D are provided by VOGL GL functionality used
by VIS D but not available in VOGL include
Z bu ering
shading
depthcueing
transparency
textures
Z bu ering and shading are vital The others are not but can enhance the rendered
images and increase the ease with which the images are interpreted by the viewer
It was decided to use VOGL for p orting VIS D for the following reasons
Portability which is a great advantage b ecause VIS D must b e p orted to HP and DEC
Using VOGL will probably also reduce the work necessary for future p orts of VIS D
The resemblance to the SGI GL makes changes to VIS D hardly necessary which
reduces complexity of VIS D Also in co op eration with the KNMI we had access to
the manuals and do cumentation of the SGI GL
The subset of the GL already implemented reduces the amount of work to do
The mo di cations necessary to vis d for using VOGL are describ ed in app endix D In the
following sections the enhancements made to VOGL are discussed after a short intro duc
tion to the SGI GL which is necessary to understand why the mo di cations to VOGL were
necessary
Chapter
Porting VIS D
VIS D do es not work on HP Series and DEC hardware platforms see section for
a list of system requirements VIS D had to b e p orted to HP systems to b e able to use
VIS D at Leiden University The KNMI is interested in a p ort of VIS D to a DEC Alpha
station The D graphics library requirement is the most serious requirement for a p ort
to HP and DEC VIS D uses the XFDI library for Stellar and the SGI Graphics Library
GL for SGI and IBM hardware platforms see GLman The XFDI library uses very
high level drawing primitives like XFDIVnormPolyTri which draws a p oly triangle strip
a p olygonal mesh built of triangles using normals for lighting calculations The GL uses
low level primitives and draws a p oly triangle strip by separately sending the normal vectors
and vertex p ositions to the GL The function calls that supply the normals and vertices must
b e enclosed by function calls that notify the GL that a p oly triangle strip must b e drawn
These two examples are fairly representative for the remaining part of the libraries The D
library is further discussed in the next section
D graphics library
In order to p ort VIS D to HP and DEC systems a D graphics library is necessary for b oth
systems An ideal solution would b e a portable graphics library which can b e used for b oth
1
target systems A low cost solution is available in the form of VOGL which stands for Very
Ordinary GL like Graphics Library VOGL is a software implementation of a subset of the
hardware SGI Graphics Library GL
Advantages of VOGL are
VOGL is p ortable The device dep endent source co de is separated from the main source
co de into device drivers To use VOGL with a new device only a device driver has to
1
VOGL was written by Eric H Echidna at the University of Melb ourne echidna munnari OZ AU It may
b e freely used and mo di ed as long as the original COPYRIGHT notice remains intact
CHAPTER OVERVIEW OF VIS D
Figure Control window
VIS D SOURCE CODE ORGANIZATION
Figure Example contents of the VIS D D window
CHAPTER OVERVIEW OF VIS D
External user functions directory userfuncs One way of deriving new physical vari
ables from existing ones is to have an external user function calculate the new variable
In the userfuncs directory the user can install Fortran mo dules which are all compiled
and linked into separate programs vis d scans the userfuncs directory and the user
can execute them from within vis d More information ab out external user functions
is provided in app endix B
Utilities for managing and analyzing ve dimensional data grids and converting data
into VIS D le format directory util In the VIS D package several utilities see
section are included for various op erations on GRID and COMP les see ap
p endix A and A Skeleton example programs reside also in the utilities directory
These programs show how to imp ort data into vis d The util directory also contains
a library of utility functions McIDAS library for reading McIDAS GRID les The
main co de of vis d makes use of the McIDAS library for the direct reading of GRID
les
LUI Little User Interface library directory lui The LUI library is based up on X
Windows and provides convenient functions for all sorts of manipulation and interaction
needed in a window environment VIS D uses this library for all graphical user interface
functions
The main program vis d relies on the LUI library for all X Windows based functions and
the McIDAS library functions for reading GRID les For a more detailed overview of the
available source co de the reader is referred to app endix B
VIS D SOURCE CODE ORGANIZATION
isosurfaces horizontal contour slices vertical contour slices horizontal colored slices and
vertical colored slices of the data The buttons are arranged in a matrix Each row corresp onds
to a physical variable in the data set Each column corresp onds to one typ e of graphic listed
b elow
To display the graphic click the left mouse button on the widget button A small control
window will app ear that is di erent for each typ e of graphic The following graphic typ es are
supp orted
D contour surface isosurface The isosurface control window that app ears is used to
set the iso level value for the contour surface The desired value can b e selected by
dragging the slider to the correct place When the correct value is selected click on the
NEW SURF button The right mouse button can b esides to change the colors of the
surface b e used to change the transparency The transparency value can slide b etween
invisible and opaque
Contour slice After selecting a contour slice with the left mouse button a contour slice
control window containing a typ e in widget will app ear This typ e in is used to sp ecify
the interval b etween contour lines These intervals can b e typ ed in while p ointing the
mouse to the widget A RETURN at the end will change the values When a negative
value is entered all contour lines with a negative value will b e dashed and all p ositive
values will b e drawn in solid lines Optionally after the interval value a range of values
a b can b e sp eci ed which will cause only contour lines b etween a and b to b e
drawn
Color slice The color slice control window that turns up contains a function graph for
red green and blue that maps data values to colors To change the red green and
blue function press the left middle and right mouse button resp ectively and drag the
mouse to draw a new function Also the keyb oard cursor arrow keys can b e used
to change the function curves Press right left to move the function right left Press
up down to change the shap e of the curve
VIS D source co de organization
In order to succesfully make mo di cations to VIS D a clear understanding of the source
co de is necessary The source co de which is written in b oth C and FORTRAN total ca
lines can b e split into four parts
vis d main source co de directory src The source co de of vis d is written in a mo du
lar fashion The rst systems vis d was programmed for included the Silicon Graphics
and Stellar Stardent using multi threaded execution The parts of the program that
mo dify exclusive resources are enclosed within semaphore statements Single threaded
execution is also p ossible Due to the sp eed of these systems the source co de is easy
to understand b ecause it is programmed in a very straightforward way This makes
mo di cations and enhancements to vis d more easy
CHAPTER OVERVIEW OF VIS D
bigger or smaller resp ectively When pressed and moved left or right the clipping plane
i e plane of invisibili ty moves towards or away from the user resp ectively When the
right mouse button is pressed the b ox in the window can b e translated
Tra jectory This mo de is used for creating and displaying wind tra jectories This is
only p ossible if the data set contains U V and W wind comp onents and more than one
time step First a p osition for the cursor must b e selected This can b e done with the
right mouse button Then a time step must b e selected with the STEP button The
tra jectory will b e traced forward from the chosen time step to the last time step and
backward through time to the rst time step Finally the middle mouse button must
b e pressed to activate the tra jectory Wind tra jectories can b e displayed in two ways
as line segments D and as ribb ons D The tra jectory window also contains two
buttons lab eled STEP and LENGTH The STEP button is used to control the step size
used in the tra jectory tracing algorithm The LENGTH value is used to control the
length of the tra jectories
Slice This mo de is used to rep osition horizontal and vertical slices contour slices
colored slices and wind vector slices To move a slice p oint the mouse at any slice
corner and drag the slice to its desired p osition while holding the right mouse button
Horizontal slices can b e p ositioned in any way as long as they are parallel to the b ottom
of the b ounding b ox Vertical slices can have any orientation as long as they remain
vertical They can b e moved parallel by p ointing to the middle of the top or b ottom
edge and then dragging to the new p osition while holding the right mouse button
Lab el This mo de is used to create and edit text lab els in the D window First
p osition the mouse p ointer in the D window and press the left mouse button A
vertical bar cursor will app ear where the text can b e typ ed in The middle mouse
button can b e used to move the text to another p osition To delete a text p oint the
cursor at the text and press the right mouse button
Prob e Used to insp ect individual grid values by moving a D cursor through the
D grid For each physical variable the value is printed along the left edge of the D
window
Wind slice buttons
If the data set contains U and V wind comp onent variables there will b e a row of four wind
slice buttons if the W wind comp onent is present it is used otherwise the W comp onent is
assumed to b e zero for all grid p oints
The lo cation of the slice plane can b e changed with the mouse while in Slice mo de The
color of the slice can b e changed by selecting the widget button with the right mouse button
The wind slice control window contains two typ e in widgets to control the scale and density
of the wind vectors The scale parameter is used to multiply the length of the vectors drawn
The density parameter controls how many wind vectors are b eing displayed
Display buttons
The last set of widget buttons in the control window is used to control the display of D
USER INTERFACE
ANIMATE Toggle animation on or o Left and middle mouse button
will result in forward animation right mouse button in backward
animation
STEP Step ahead one time step left mouse button go to rst time
step middle mouse button or step backward one step
right mouse button
NEW VAR Duplicate physical variables or invoke external analysis functions
Two kinds of new variables can b e added Clone variables copies
of existing variables and external function variables calculated
as a function of the existing variables see also
app endix B
EXIT Exit the program The decision must b e veri ed
NEW SURF Compute a new D contour surface
TOP Set D window to a top view left and middle mouse button or
to a b ottom view right mouse button
SOUTH Set D window to a south view left and middle mouse button
or to a north view right mouse button
WEST Set D window to a west view left and middle mouse button or
to a east view right mouse button
TOPO Toggle display of top ography Right mouse button will edit the
top ography color The commandline option topo topographyfile
can b e used to load a top ography le Default le is EARTH TOPO
MAP Toggle display of map The commandline option map mapfile
can b e used to load a map le The default map is OUTLSUPW
BOX Toggle the display of the D b ox
CLOCK Toggle clo ck on or o
SAVE Save current graphics and colors left mouse button or only
save colors right mouse button
RESTORE Restore the data set last saved with the SAVE button
GRID s Toggle b ounds along the edges of the b ox in latitude longitude
and height co ordinates or grid co ordinates
CONT s Toggle contour lab eling on or o
PRETTY Toggle the pretty rendering option on or o
REVERSE Toggle white b ox on black background to black b ox on white
background
SAVE PIC Save picture in the D window to a le
PERSPEC Toggle b etween p ersp ective and orthogonal viewing pro jections
Mouse b ehaviour buttons
This group of radio buttons only one may b e selected at the time is used to determine the
e ect of pressing a mouse button while the mouse p ointer is p ositioned in the D window
To the immediate right of these buttons the mouse button legend is displayed The following
mouse b ehaviour buttons are available
Normal This mo de is used to view the graphics in the D window By pressing the
left mouse button and dragging the mouse the D image can b e rotated The center
mouse button has two functions When pressed and moved up or down the b ox gets
CHAPTER OVERVIEW OF VIS D
Graphical user interface
When starting vis d it op ens two windows a D window on the right of the screen and a
control panel on the left
The D window is used to view and interact with the data In its upp er left corner a
analog digital clo ck is displayed which indicates the current time step In the center of the
window a b ox is drawn when selected by the user which depicts the domain of the D data
All graphics will b e displayed in this b ox To the left side on the b ottom the values of all
physical parameters are displayed when the Prob e button is selected In the right b ottom
corner the name of the program is shown An example is shown in gure
The control window contains several groups of buttons that are used to control various func
tions The graphical user interface is an X Windows based interface that is device indep en
dent Another p ossibility is to op en the D window with maximum size full screen and to
op en the control window and all other windows on a second terminal
The control window see gure is used to control any graphic by clicking on its widget
button with the left mouse button A small control window will app ear To bring up the
graphics control window without toggling the display use the middle mouse button To
change the color of the graphic click with the right mouse button on the widget button
Some functions may not b e present on the control window This can happ en b ecause the
function is not available on the hardware platform the button do es not apply to the data set
or the button con icts with a command line option that is set
The buttons of the control window can b e divided into four groups see gure
control buttons
mouse b ehaviour buttons
wind slice buttons
display buttons
Control buttons
These buttons are used to toggle activate or control various functions
USER INTERFACE
vis d command line options
After pro ducing a compressed le the data can b e visualized with the command
vis d compfile options
The following command line options are available
map mapfile Use a map le other than the default map le
le OUTLSUPW This default le is a world map in
a normal latitude longitude grid The reader is
referred to app endix A
for a description of the map le format
topo topographyfile Use a top ography le other than the default
top ography le EARTH TOPO This default le
is a world top ography in a normal
latitude longitude grid with heights and a
land water mask The reader is referred to
app endix A for a description
of the top ography le format
box x y z Sp ecify the asp ect ratio of the D b ox in the
D window The default ratios of this b ox are
font fontname height Set the font used in the D window On the
SGI system the height of the font can also b e
sp eci ed
mbs n Override the assumed system memory size of Mb
rate ms Change the default animation rate to ms
microseconds b etween frames The default
animation rate is ms
full Op en the D window as a b orderless
full screen sized window
wdpy xdisplay Make the control window app ear on a di erent
X display
alpha Use alpha blending instead of screen do or
transparency This feature is only available on
SGI systems
wide w Set width of line segments in pixels The
default width is pixels
An overview of the command line options is given on the screen by running vis d without
arguments
CHAPTER OVERVIEW OF VIS D
lat XLATS XLATN
which sets the latitude range b ounds south north and
lon XLONE XLONW
which sets the longitude range b ounds east west and
hgt XHGTB XHGTT
which set the height range b ounds b ottom top for the destination grid
Before viewing the data set the GRID les must b e converted to a compressed le structure
COMP le This can b e done using the command
comp d N M F
where N is the rst D grid le numb er M is the numb er of grid les in the data set and F is
the name of the compressed grid le
If the data set contains wind vector comp onents U V and W the horizontal wind SPD can
b e included in the COMP le using the command
comp d N M F wind SPD U
This will result in including SPD and U in the COMP le not V and W
To create D grid le numb er N which allows D grids of up to M p oints each use the
command
igu d make N M
To get information ab out the COMP le made use the command
compinfo COMP file
that prints the grid dimensions numb er of time steps numb er of variables etc of the COMP
le
The command
help NAME
can b e used to get a quick overview of the parameter format of the VIS D utilities
USER INTERFACE
where grid numb er K is pro duced by interp olating b etween grid numb ers I and J all in D
grid le numb er N Grid numb er K will b e assigned day D in YYDDD format and time T in
HHMMSS format The relative weighting of grids I and J is calculated from this time and
date assuming linear time interp olation If grid K is not b etween grids I and J in date and
time igg d prints an error message
The second way to interp olate in time is to use the command
igg d int I D T setdel S M lag U V gr df N
This will put a grid in the next empty slot of D grid le numb er N assigned to day D in
YYDDD format and time T in HHMMSS format This grid will b e interp olated from a
sequence of grids all in le numb er N at grid numb ers I I S I S I M S This
sequence of grids should b e ascending in date and time igg d will search the sequence and
linearly interp olate b etween the two consectutive grids from the sequence which bracket day
D and time T Furthermore the interp olation will b e done in a co ordinate system moving at
constant velo city U V where U and V are in m s with V p ositive for motion from south
to north and U p ositive for motion from west to east The two bracketing grids from the
sequence will b e shifted in latitude and longitude to their p ositions at day D and time T and
the result interp olated b etween these two spatially shifted grids Furthermore if the grids in
the sequence are identi ed in their directory entries with variable name U or V then the
corresp onding comp onent of the velo city U V will b e subtracted from the grid values
There is a built in p ossibility to resample the GRID le This is done with the commands
gg d samp N M I J
and
gg d ave N M I J
where a N M grid is resampled to a I J grid The samp option will linear interp olate b etween
source grid p oints and is appropriate for increasing resolution The ave option averages
multiple source grid p oints and is used to decrease the resolution of a grid
Starting gg d without keywords will do a straight copy op eration gg d can b e invoked with
the keyword
size NLATS NLONS NHGTS
which will cause the destination grid dimensions to di er from the source grid dimensions
NLATS NLONS NHGTS are the destination grid latitude longitude and height dimension
resp ectively Other keywords that can b e used are
CHAPTER OVERVIEW OF VIS D
le format see app endix A D GRID le names consists of GR Dxxxx where xxxx is the
D GRID le number the numb er is assumed to b e digits A D GRID le consists of
one or more D data grids which are numb ered in the order in which they app ear in the
GRID le starting with Grids in a particular GRID le are referenced by the utilities
using this grid number
VIS D utility commands
The directory information of the grids in the GRID les can b e viewed with the command
igg d list I J gr df N
where N is the D grid le numb er and I and J give the range of grid numb ers to list
A short overview of the data in the GRID les such as the minimum and maximum values
the mean and the standard deviation and the numb er of grid p oints with missing data can
b e viewed with the command
igg d info I J gr df N
with I J and N the same as ab ove
The igg d command also provides options for copying and deleting D grids and interp olat
ing D grids in time Sequences of D grids are copied using the command
igg d get N I J M K
where N is the source D grid le numb er I and J are the range of source grid numb ers M
is the destination grid le numb er and K is the starting grid le numb er
A single grid can b e copied using the command
igg d copy I J gr df N
where N is the D grid le numb er I is the numb er of the source grid and J is the numb er
of the destination grid The same way a numb er of grids can b e deleted with the command
igg d del I J gr df N
There are two ways to interp olate in time The rst way is use the command
igg d ave K I J D T gr df N
CAPABILITIES
The reader is referred to chapter for information ab out p orting VIS D to other hardware
platforms
Capabilities
vis d can handle physical variables in single time steps and multiple time steps It is also
p ossible to pro duce a new variable as a function of the existing variables All these data can
b e easily visualized manipulated and lo oked at from every p ossible viewp oint
vis d is capable of visualizing data in the following ways
D contour surface isosurface a surface that shows the D volume b ounded by a
particular isovalue This is the D equivalent of a D contour line
contour slice a slice with lines on which values in the data set are equal
colored slice a slice with data values mapp ed to colors
wind vector slice a slice that depicts the motions of the wind by drawing arrows
which p oint in the direction of the wind The length of each line segment indicates its
magnitude The tail of the line segments are all anchored within a horizontal or vertical
plane through the D b ox
wind tra jectories tracing of the motion of air through the D volume much like the
trailing of line smoke in a wind tunnel
prob e insp ection of individual grid values
All these visualization techniques can b e shown simultaneously during one or more time steps
All graphics can b e manipulated interactively even during animation
With equal ease text lab els can b e put in the D window and a window dump can b e made
from this window and saved to a le or printed
User interface
In VIS D there are two ways of controlling the data The rst way of doing this is using
command line options when starting vis d or making and analyzing the data with the utility
commands The second way is with the use of the graphical user interface within the program
Command line interface
With vis d additional programs are supplied that enable the user to make and analyse the
data The grids are assumed to b e in D GRID le format see app endix A or in COMP
CHAPTER OVERVIEW OF VIS D
VIS D is o ered under the terms of the GNU General Public License There is no warranty
for the VIS D package
VIS D is b eing distributed as a stand alone system free of charge However it is also a
subsystem of a much larger ca two million lines of FORTRAN co de system named McIDAS
Man computer Interactive Data Access System see Smit and Hibb
System requirements
VIS D version works on the following hardware platforms
Stardent GS or GS with
true color display i e planes
MB RAM
optional SpaceBall is supp orted
Stellix version or later is suggested
Silicon Graphics with
bit color and Z bu er or bit Indigo
MB RAM
IRIX version or higher
multiple pro cessors are used when present
IBM RS
mo del H or higher
requires bit color and Z bu er GTO desirable
MB RAM
AIX version or later is suggested
To b e able to use VIS D on a sp eci c target machine the following resources should b e
available
C Compiler
FORTRAN Compiler
X Windows
D graphics library with Z bu er
And preferable
MB of memory
bit color display with high p erformance D graphics hardware
Chapter
Overview of VIS D
1
This chapter gives an overview of VIS D as originally distributed by ftp The latest avail
able version of VIS D August is used as the basis of this chapter The mo di ca
tions made in order to visualize the HIRLAM data with VIS D are describ ed in chapter
Background
VIS D is a software system for visualizing data made by numerical weather mo dels and
similar sources VIS D works on data in the form of a ve dimensional rectangle That
is the data are real numb ers at each p oint of a grid or lattice which spans three space
dimensions one time dimension and a dimension for enumerating multiple physical variables
VIS D works p erfectly well on data sets with only one variable or one time step i e no time
dynamics However the data should have some depth in all three spatial dimensions This
makes visualization of D elds imp ossible
The VIS D package includes the vis d visualization program several programs for manag
ing and analyzing ve dimensional data grids and instructions and sample source co de for
converting your data into its le format The VIS D source co de is also included so it can
b e mo di ed or extended There are also sample data sets from the LAMPS Limited Area
Meso Scale Prediction System mo del and from Rob ert Schlesinger s numerical thunderstorm
mo del available
VIS D version was written by Bill Hibbard and Dave Santek of the University of Wis
consin Space Science and Engineering Center supp orted by the NASA Marshall Space Flight
Center and by Marie Francoise Voidrot Martinez of the French Meteorology O ce Later
version enhancements were written by Bill Hibbard Brian Paul and Andre Battaiola Dave
Kamins and Je Vro om of Stellar Computer Inc provided substantial help and advice in
using the Stellar software libraries
1
VIS D can b e obtained via anonymous ftp from vis d ssec wisc edu in the directory pub vis d
CHAPTER INTRODUCTION
VIS D is restricted to height as the vertical co ordinate
VIS D do esn t supp ort the creation of color map legends
VIS D supp orts only a D p oint prob e instead of a D curve prob e
Solving these problems requires a relatively small e ort compared to the e ort required to
implement a complete visualization package
VISUALIZATION TOOLS
Furthermore the following features should b e supp orted
supp ort for various vertical co ordinate systems
colormap legends for color mappings a legend should b e optionally displayed giving
information ab out the corresp ondence b etween colors and data values
data prob e to insp ect the value of one or more physical variables along a path through
the data domain
Visualization to ols
To visualize the output of HIRLAM a visualization to ol is needed Visualization programs
can b e divided into two categories First there are programs written with a sp eci c purp ose
in mind These kind of programs contain sp eci c routines to calculate and display an often
restricted range of images Secondly there are visualization to olkits These consist of a
collection of mo dules all containing only a small part of the total visualization pro cess By
combining mo dules the user can construct larger applications suited to his visualization needs
Advantages of a to olkit are exibili ty and the ease with which the to olkit can b e enhanced
with new mo dules For this reasons the required visualizations were rst attempted using
apE I I I a to olkit for visualization Due to the large numb er of shortcomings and bugs
see Baas apE was abandoned in favour of another package called VIS D Chapter
will give an overview of VIS D
VIS D was chosen for the following reasons
VIS D is written sp eci cally for the visualization of meteorological data
free under the GNU licence
the source co de is available
supp ort an active VIS D mailing list is available
VIS D is written with p ortability in mind
the KNMI evaluated VIS D and are very enthousiastic ab out it
To make VIS D able to visualize the HIRLAM output as describ ed in section the fol
lowing problems had to b e solved
VIS D only runs on SGI Stellar Stardent and IBM RS systems
VIS D can t read the HIRLAM output data format
VIS D can t visualize D data sets
CHAPTER INTRODUCTION
Pro ject description
Currently the physical variables pro duced by HIRLAM include
wind direction and sp eed
temp erature
surface pressure
air density
humidity
For the purp ose of visualization the output variables can b e divided into three categories
D scalar elds surface pressure
D scalar elds temp erature air density humidity
D vector elds wind
The wind is actually a D eld of D vectors b ecause the vertical comp onent of the wind
vectors physical variable W is currently not present it is not calculated by HIRLAM but
this can easily b e changed
The following visualization techniques stationary or as time lo ops are required
For D scalar variables and D slices taken from D scalar variables
contour images lines connecting the gridp oints where the variable has the same
value the contours lab eled with corresp onding values
color mappings data values are mapp ed to colors according to a sp eci ed col
ormap
discretized color mappings the data is rst mapp ed into intervals using a step
function so that colored bands are formed
For D scalar variables
iso surfaces D equivalent of a contour line
For D vector variables
D slices with arrows representing wind vectors
wind tra jectories streaklines depicting the path of particles through a stationair
p oint moved by the wind
Chapter
Intro duction
Pro ject background
In collab oration with the KNMI the Royal Dutch Meteorological Institute the High Per
formance Computing Division of the Department of Computer Science Leiden University
RUL researches e cient implementations of a numerical weather forecasting mo del on var
ious parallel computer systems This mo del called HIRLAM HIgh Resolution Limited Area
Mo del is the result of a co op erative pro ject b etween Denmark Finland Iceland Ireland
The Netherlands Norway and Sweden The purp ose of the HIRLAM pro ject is to design a
state of the art weather forecasting system HIRLAM is already used by some of the partici
pating countries to make predictions of the weather The horizontal domain of HIRLAM as
used at KNMI is displayed in gure
Figure Horizontal domain of HIRLAM
The output of the atmospheric circulation mo del of HIRLAM consists of large volumes of
four dimensional data currently several time steps each consisting of ca by by
calculated data p oints representing several physical variables Visualization is a natural way
to obtain insight in these data
CONTENTS v
C futil
C grib vis
D VIS D mo di cations
D HP p ort mo di cations
D LUI changes
D vis d changes
D D elds mo di cations
D Logarithmic interp olation mo di cations
D VOGL related mo di cations
E VOGL implementation details
E VOGL enhancements
E Z bu er functions
E RGB color functions
E Lighting functions
E Depthcueing functions
E VOGL device driver structure
F Example Images
iv CONTENTS
VIS D mo di cations
Visualization of D elds
Logarithmic interp olation
Horizontal slices
Vertical slices
Iso surfaces
Wind tra jectories
Data prob e
Scaling vertical wind comp onent
Top ography conversion
Map conversion
Results
Conclusions
Future work
A File formats
A GRIB le format
A ASIMOF le format
A GRID le format
A COMP le format
A Old style format
A New style format
A Variable NL format
A VIS D map le format
A VIS D top ography le format
B VIS D source co de structure
B VIS D main source co de
B External User functions
B Utilities
B Little User Interface LUI library
C Libraries
C KNMI libraries
C p ort
C gco d
C util
C vari
C prp o
C grw
C RUL libraries
C asimof
C grib
C interp ol
C cutil
Contents
Preface i
Intro duction
Pro ject background
Pro ject description
Visualization to ols
Overview of VIS D
Background
System requirements
Capabilities
User interface
Command line interface
Graphical user interface
VIS D source co de organization
Porting VIS D
D graphics library
An intro duction to the SGI GL
Z bu ering
Lighting
Depth cueing
Scan conversion
VOGL device drivers
Miscellaneous problems
Interfacing HIRLAM with VIS D
HIRLAM output
Conversion to GRID versus COMP
Vertical co ordinate
ASIMOF to GRID conversion sp eci cation
Conversion from GRIB to GRID
KNMI libraries
RUL libraries
Conversion program
GRID to COMP conversion iii ii
Preface
This is a master thesis ab out the visualization of meteorological data on b ehalf
of Leiden University and the KNMI Royal Dutch Meteorological Institute We
would like to thank a numb er of p eople who help ed us with the many problems
that arose during the research
In the rst place we would like to thank our sup ervisor Lex Wolters Leiden
University for his supp ort during the research and the time he has sp ent in
correcting this thesis
Sp ecial thanks go es to Gerard Cats KNMI To on Mo ene KNMI and Ben Wich
ers Schreur KNMI for their help in explaining the sometimes complex working
of meteorology Furthermore we would like to thank Nies Huijsmans Leiden Uni
versity for his advice on the sub ject of computer graphics i
Meteorolog i cal
Visualization
using VIS D
Simon Baas Hans de Jong
Leiden University August