Table of Contents
This chapter discuss how to run VISIM from within Matlab.
VISIM is a sequential simulation code based on GSLIB ('Geostatistical Software LIBrary', Stanford Center for Reservoir Forecasting, Stanford University) for sequential Gaussian and direct sequential simulation with histogram reproduction. In addition to classical simulation based on noise-free hard data of point support, VISIM also handles noisy data of mixed support, enabling linear stochastic inversion using the sequential simulation approach.
VISIM can be used to generate samples of the a posteriori distribution of a linear inverse problem.
For detailed information the usage of VISIM visit the VISIM homepage at IMGP. [user manual pdf].
References to the use of VISIM should be made to :
Hansen, T. M., and Mosegaard, K., 2008. VISIM : Sequential simulation for linear inverse problems. Computers andGeosciences, 34(1), pp 53-76. [doi:10.1016/j.cageo.2007.02.003].
The three main files to control VISIM from Matlab are the section called “read_visim” the section called “write_visim” the section called “visim”
When running the section called “visim” an attempt is made to look for a VISIM binary. First the 'bin' sub folder where mGstat is installed is searched, then the system path is searched for 'visim' binary file. By running the section called “visim” with no arguments the path to the use binary (if located) is listed:
visim; Using VISIM binary : ~/bin/visim
If the binary for VISIM is not found you have to edit the 'visim.m' m-file and manually give set the path in the top of the file.
the section called “visim” is used to run VISIM from within Matlab. The input is either a VISIM parameter file (as a string) or a VISIM Matlab structure, as obtained using the section called “read_visim”.
Therefore the following two approached will run VISIM on the
visim_sgsim.par:
V = visim('visim_sgsim.par');
V = visim('visim_sgsim.par');
V = visim(V);
which will result in the following out to the screen :
VISIM Version: 1.000
filename is sgsim_uncond.par
sgsim_uncond.par
Initializing data2vol covar lookup table
Initializing vol2vol covar lookup table
Working on realization number 1
Working on realization number 2
Working on realization number 3
...
Working on realization number 100
V =
parfile: 'sgsim_uncond.par'
cond_sim: 0
fconddata: [1x1 struct]
cols: [6x1 double]
fvolgeom: [1x1 struct]
fvolsum: [1x1 struct]
trimlimits: [2x1 double]
debuglevel: -11
out: [1x1 struct]
nsim: 100
ccdf: 0
refhist: [1x1 struct]
nx: 21
xmn: 0.1250
xsiz: 0.2500
ny: 49
ymn: 0.1250
ysiz: 0.2500
nz: 1
zmn: 0.1250
zsiz: 0.2500
x: [1x21 double]
y: [1x49 double]
z: 0.1250
rseed: 69068
minorig: 0
maxorig: 1
nsimdata: 28
volnh: [1x1 struct]
densitypr: 0
assign_to_nodes: 1
max_data_per_octant: 0
search_radius: [1x1 struct]
search_angle: [1x1 struct]
gmean: 0.1304
gvar: 2.0000e-04
Va: [1x1 struct]
tail: [1x1 struct]
D: [21x49x7 double]
etype: [1x1 struct]
The output of running visim is the Matlab structure listed above,
that contains all parameters of the VISIM used parameter file, as
well as the simulated/estimated data
V.D, and the E-type mean and variance as
V.etype.mean and
V.etype.var
An example of running 9 runs of VISIM while changing the primary anisotropy direction from 0 to 90 is the following
Vorig = read_visim('visim_sgsim.par');
for i=1:4
V{i}=Vorig;
V{i}.Va.ang1=i*10
V{i}=visim(V{i}));
end
Each of the Matlab structures
V{1},V{2},..,V{9}
now contain the simulation result.
To plot a number of generated realizations use the section called “visim”.
The simplest use of the section called “visim” is simple to specify a number of realizations to plot.
V=read_visim('sgsim_cond_1.par');
visim_plot_sim(V,2)
An example of using all the options available for the section called “visim”
nsim=7; % number of realizations to plot cax=[.1 .16]; % scaling for colorbar axis FS = 10; % FontSize nxsub=7; % Number of subplots in the X-direction nzsub=7; % Number of subplots in the z-direction visim_plot_sim(V,nsim,cax,FS,nxsub,nzsub);
In case VISIM is run in simulation mode, visim_plot_etype plots the E-type (the point wise mean and/or variance) of all generated simulations.
cax=[.1 .16]; % scaling for colorbar axis
visim_plot_etype('sgsim_cond_1.par');
The following will plot both the E-type mean and variance.
V=read_visim('sgsim_cond_1.par');
plotAll=1;
visim_plot_etype(V,plotAll);
The following will plot both the E-type mean and variance with the specified colorscales.
V=read_visim('sgsim_cond_1.par');
cax=[.1 .16]; % scaling for colorbar axis E-type Mean
cax_var=[0 1e-4]; % scaling for colorbar axis E-type variance
plotAll=1;
visim_plot_etype(V,plotAll,cax,cax_var);
See more info at the section called “visim_plot_etype”
To plot the histogram for all realizations, compared to the a priori chosen one use
visim_plot_hist :
V=read_visim('sgsim_uncond.par');
visim_plot_hist(V);
and to plot just the histogram of one simulation use
V=read_visim('sgsim_uncond.par');
visim_plot_hist(V,10);
See more info at the section called “visim_plot_hist”
To plot the experimental semivariograms compared to the a priori chosen semivariogram, use
visim_plot_semivar_real :
V=read_visim('sgsim_uncond.par');
visim_plot_semivar_real(V);
See more info at the section called “visim_plot_semivar_real”
To compatre the mean experiemntal computed for a number of realizatoins to the a priori chosen semivariogram, use
V=read_visim('sgsim_uncond.par');
visim_plot_semivar(V,1:V.nsim);
See more info at the section called “visim_plot_semivar”
The sensitivity kernel for all or a specific selection of volume average data can be plotted using :
subplot(1,3,1)
visim_plot_kernel('sgsim_cond_3.par');
subplot(1,3,2)
visim_plot_kernel('sgsim_cond_3.par',11);
subplot(1,3,3)
visim_plot_kernel('sgsim_cond_3.par',11:25);
colormap(1-gray)
This will result in the following :
See more info at the section called “visim_plot_kernel”
See more info at the section called “visim_plot_condtab”
To plot the distribution of the volume average estimates as calculated form the realization and the observed volume average values use:
V=read_visim('sgsim_cond_1.par');
visim_plot_volfit(V);
See more info at the section called “visim_plot_volfit”