Oise - Non-stationary estimation¶

This tutorial is meatn to prepare the data sets for the estimation of the thickness of sediments along the Oise river. This second phase is used to read the data sets (laready prepared in the first phase) and perfom the non-stationary estimation.

In [1]:
import matplotlib.pyplot as plt
import gstlearn as gl
import gstlearn.plot as gp
import gstlearn.document as gdoc
import numpy as np
import pandas as pd
import scipy as sp
from scipy import interpolate
import os

gdoc.setNoScroll()
In [2]:
#!pip install gstlearn
#!pip uninstall -y gstlearn
#!pip3 install numpy --upgrade

Load Data¶

In [3]:
filename = gdoc.loadData("Alluvial", "Oise_GridVectorFinal.ascii")
grid = gl.DbGrid.createFromNF(filename)
grid
Out[3]:
Data Base Grid Characteristics
==============================

Data Base Summary
-----------------
File is organized as a regular grid
Space dimension              = 2
Number of Columns            = 9
Total number of samples      = 1320000
Number of active samples     = 278129

Grid characteristics:
---------------------
Origin : 630000.0006865000.000
Mesh   :     50.000    50.000
Number :       3300       400
Rotation Angles        =     40.000     0.000
Direct Rotation Matrix
               [,  0]    [,  1]
     [  0,]     0.766    -0.643
     [  1,]     0.643     0.766
Inverse Rotation Matrix
               [,  0]    [,  1]
     [  0,]     0.766     0.643
     [  1,]    -0.643     0.766

Variables
---------
Column = 0 - Name = rank - Locator = NA
Column = 1 - Name = x1 - Locator = x1
Column = 2 - Name = x2 - Locator = x2
Column = 3 - Name = Polygon - Locator = NA
Column = 4 - Name = Migrate.u_interp - Locator = z1
Column = 5 - Name = Migrate.v_interp - Locator = z2
Column = 6 - Name = vec_define - Locator = NA
Column = 7 - Name = angles1 - Locator = nostat1
Column = 8 - Name = res - Locator = sel
In [4]:
#ranges = [500,10000]
ranges = [10000, 500]
oldMethod = False
In [5]:
filename = gdoc.loadData("Alluvial", "Oise_Data.ascii")
data = gl.Db.createFromNF(filename)
data
Out[5]:
Data Base Characteristics
=========================

Data Base Summary
-----------------
File is organized as a set of isolated points
Space dimension              = 2
Number of Columns            = 7
Total number of samples      = 1022
Number of active samples     = 1007

Variables
---------
Column = 0 - Name = rank - Locator = NA
Column = 1 - Name = X - Locator = x1
Column = 2 - Name = Y - Locator = x2
Column = 3 - Name = Epaisseur - Locator = lower1
Column = 4 - Name = Epaisseur_1 - Locator = upper1
Column = 5 - Name = Thickness - Locator = z1
Column = 6 - Name = Duplicate - Locator = sel

Kriging with SPDE¶

In [6]:
myVarioParamBidir = gl.VarioParam()
mydir = gl.DirParam(40,800.,0.5,45.,0,0,np.nan,np.nan,0.,[],[1,1])
myVarioParamBidir.addDir(mydir)
mydir = gl.DirParam(20,400.,0.5,45.,0,0,np.nan,np.nan,0.,[],[-1,1])
myVarioParamBidir.addDir(mydir)
myVarioBidir = gl.Vario(myVarioParamBidir)
err = myVarioBidir.compute(data,gl.ECalcVario.VARIOGRAM)
myVarioBidir.display()
ax = gp.varmod(myVarioBidir,idir=0)
ax.decoration(title="Bi-directional Variogram for thickness(45°)")
ax = gp.varmod(myVarioBidir,idir=1)
ax.decoration(title="Bi-directional Variogram for thickness(45/135°)")

Variogram characteristics
=========================
Number of variable(s)       = 1
Number of direction(s)      = 2
Space dimension             = 2
Variance-Covariance Matrix     8.308

Direction #1
------------
Number of lags              = 40
Direction coefficients      =      1.000     1.000
Direction angles (degrees)  =     45.000     0.000
Tolerance on direction      =     45.000 (degrees)
Calculation lag             =    800.000
Tolerance on distance       =     50.000 (Percent of the lag value)

For variable 1
      Rank    Npairs  Distance     Value
         0  1490.000   223.076     4.288
         1  4135.000   817.513     6.027
         2  4998.000  1597.587     5.765
         3  5136.000  2402.372     7.172
         4  4916.000  3203.179     7.076
         5  5370.000  3996.268     6.865
         6  5948.000  4804.243     6.811
         7  5990.000  5607.329     6.236
         8  5883.000  6392.169     6.132
         9  5766.000  7198.769     7.205
        10  5597.000  8003.510     6.703
        11  4787.000  8802.057     7.668
        12  5403.000  9590.324     6.747
        13  5078.000 10419.574     6.709
        14  5653.000 11189.885     6.865
        15  6134.000 12008.161     6.501
        16  5399.000 12774.078     6.778
        17  4853.000 13605.558     6.502
        18  4516.000 14393.275     6.467
        19  4757.000 15204.141     6.521
        20  4609.000 16000.765     5.898
        21  4440.000 16792.781     6.951
        22  4748.000 17616.270     7.334
        23  6281.000 18422.038     7.215
        24  5730.000 19188.968     7.037
        25  5347.000 19996.444     6.630
        26  5141.000 20791.941     7.823
        27  5458.000 21593.393     7.906
        28  4791.000 22407.613     7.568
        29  5377.000 23190.787     9.347
        30  4959.000 24009.942     7.417
        31  5189.000 24797.152     7.993
        32  5187.000 25608.299     7.774
        33  5398.000 26409.517     8.090
        34  4921.000 27190.774     7.407
        35  4706.000 27991.019     7.817
        36  4497.000 28793.891     7.413
        37  5282.000 29647.849     7.006
        38  5399.000 30398.721     7.503
        39  6070.000 31199.474     7.760

Direction #2
------------
Number of lags              = 20
Direction coefficients      =     -1.000     1.000
Direction angles (degrees)  =    135.000     0.000
Tolerance on direction      =     45.000 (degrees)
Calculation lag             =    400.000
Tolerance on distance       =     50.000 (Percent of the lag value)

For variable 1
      Rank    Npairs  Distance     Value
         0   594.000   119.131     2.777
         1  1399.000   393.836     4.535
         2  1453.000   793.785     6.290
         3  1224.000  1196.400     6.529
         4  1032.000  1596.868     7.385
         5   858.000  1983.161     7.192
         6   697.000  2404.699     7.787
         7   675.000  2794.542     7.751
         8   419.000  3189.263     8.768
         9   354.000  3597.904    11.252
        10   339.000  3996.583     9.995
        11   260.000  4402.766     7.896
        12   232.000  4792.850     7.453
        13   177.000  5178.782     7.800
        14   172.000  5586.833     6.960
        15   100.000  5978.383     9.960
        16    89.000  6394.625     9.218
        17    55.000  6800.319     5.859
        18    30.000  7199.585     9.574
        19    39.000  7641.196     4.828
In [7]:
model = gl.Model.createFromDb(data)


structs = [gl.ECov.NUGGET,gl.ECov.BESSEL_K]

#consNug = gl.ConsItem.define(gl.EConsElem.SILL,0, type = gl.EConsType.UPPER,value = 0.1)

cons1P = gl.ConsItem.define(gl.EConsElem.PARAM,1, type = gl.EConsType.EQUAL,value = 1)
#cons1Rm = gl.ConsItem.define(gl.EConsElem.RANGE,1, type = gl.EConsType.LOWER,value = 100)
#cons1RM = gl.ConsItem.define(gl.EConsElem.RANGE,1, type = gl.EConsType.UPPER,value = 350)

#cons2P = gl.ConsItem.define(gl.EConsElem.PARAM,2, type = gl.EConsType.EQUAL,value = 2)
#cons2Rm = gl.ConsItem.define(gl.EConsElem.RANGE,2, type = gl.EConsType.LOWER,value = 100)
#cons2RM = gl.ConsItem.define(gl.EConsElem.RANGE,2, type = gl.EConsType.UPPER,value = 400)

a = gl.Constraints()
#a.addItem(consNug)
a.addItem(cons1P)
#a.addItem(cons1Rm)
#a.addItem(cons1RM)
#a.addItem(cons2P)
#a.addItem(cons2Rm)
#a.addItem(cons2RM)

err = model.fit(myVarioBidir,structs,constraints=a)


#err = model.fit(myVarioBidir,[gl.ECov.NUGGET,gl.ECov.BESSEL_K,gl.ECov.BESSEL_K], False)

model.display()
ax = gp.varmod(myVarioBidir,model,idir=0)
ax = gp.varmod(myVarioBidir,model,idir=1)

Model characteristics
=====================
Space dimension              = 2
Number of variable(s)        = 1
Number of basic structure(s) = 2
Number of drift function(s)  = 0
Number of drift equation(s)  = 0

Covariance Part
---------------
Nugget Effect
- Sill         =      2.215
K-Bessel (Third Parameter = 1)
- Sill         =      4.700
- Ranges       =   1046.558  1497.411
- Theo. Ranges =    302.115   432.265
- Angles       =     45.000     0.000
- Rotation Matrix
               [,  0]    [,  1]
     [  0,]     0.707    -0.707
     [  1,]     0.707     0.707
Total Sill     =      6.915
In [8]:
if not oldMethod :
    model = gl.Model.createFromDb(data)
    covBessel = gl.CovAniso.createAnisotropic(type = gl.ECov.BESSEL_K,ranges=ranges,sill=4.7,param=1,ctxt=model.getContext())
    covNugget = gl.CovAniso.createIsotropic(type = gl.ECov.NUGGET,sill=2.2,ctxt=model.getContext(),range=1)
    model.addCov(covNugget)
    model.addCov(covBessel)
In [9]:
grid.setLocator("Poly*", gl.ELoc.SEL)
#grid.setLocator("Poly*",gl.ELoc.UNKNOWN)
In [10]:
model.setDriftIRF(order = 0)
model
Out[10]:
Model characteristics
=====================
Space dimension              = 2
Number of variable(s)        = 1
Number of basic structure(s) = 2
Number of drift function(s)  = 1
Number of drift equation(s)  = 1

Covariance Part
---------------
Nugget Effect
- Sill         =      2.200
K-Bessel (Third Parameter = 1)
- Sill         =      4.700
- Ranges       =  10000.000   500.000
- Theo. Ranges =   2886.751   144.338
Total Sill     =      6.900

Drift Part
----------
Universality_Condition
In [11]:
u = grid["Migrate.u_interp"]
v = grid["Migrate.v_interp"]
res = 0. * u
for i in range(u.shape[0]):
    codir = gl.VectorDouble()
    codir.push_back(u[i])
    codir.push_back(v[i])
    res[i] = gl.GH.rotationGetAngles(codir)[0]
In [12]:
grid.deleteColumn("angles*")
if not oldMethod:
    grid["angles1"]= res
    grid.setLocator("angles*",gl.ELoc.NOSTAT)
    nostat = gl.NoStatArray(["M2A"],grid)
    model.addNoStat(nostat)
In [13]:
grid.deleteColumns(["spde*"])
In [14]:
spde = gl.SPDE(model,grid,data,gl.ESPDECalcMode.KRIGING)
uid_result= spde.compute(grid)
grid
Out[14]:
Data Base Grid Characteristics
==============================

Data Base Summary
-----------------
File is organized as a regular grid
Space dimension              = 2
Number of Columns            = 10
Total number of samples      = 1320000
Number of active samples     = 138248

Grid characteristics:
---------------------
Origin : 630000.0006865000.000
Mesh   :     50.000    50.000
Number :       3300       400
Rotation Angles        =     40.000     0.000
Direct Rotation Matrix
               [,  0]    [,  1]
     [  0,]     0.766    -0.643
     [  1,]     0.643     0.766
Inverse Rotation Matrix
               [,  0]    [,  1]
     [  0,]     0.766     0.643
     [  1,]    -0.643     0.766

Variables
---------
Column = 0 - Name = rank - Locator = NA
Column = 1 - Name = x1 - Locator = x1
Column = 2 - Name = x2 - Locator = x2
Column = 3 - Name = Polygon - Locator = sel
Column = 4 - Name = Migrate.u_interp - Locator = NA
Column = 5 - Name = Migrate.v_interp - Locator = NA
Column = 6 - Name = vec_define - Locator = NA
Column = 7 - Name = res - Locator = NA
Column = 8 - Name = angles1 - Locator = nostat1
Column = 9 - Name = spde.Thickness.estim - Locator = z1
In [15]:
#fig,ax = gp.initGeographic()
#ax.raster(grid,"spde*kriging",cmap="gnuplot2", flagLegend = True)
plt.figure(figsize=[20,12])
ax = gp.raster(grid,"spde.Thickness*estim",cmap="gnuplot2", flagLegend = True)

Prepare Export to ArcGIS¶

Kriging data to new grid

In [16]:
raster = gl.DbGrid()
raster.reset([2400,2600],[50,50],[625000,6870000],[0,0])
err = gl.migrate(grid, raster, "spde.Thickness.estim",2,[100,100],True)
In [17]:
# Export Kriging grid to Ascii for Raster format => raster.gridwrite.format("arcgis","filename.txt")
#Kri=raster.getColumn("Migrate.spde.Thickness.kriging")
#len(Kri)
#KriG= np.array(Kri)
#KriG = np.where(KriG > 1000, -999999, KriG)
#KriG= np.where(KriG==0, -999999, KriG) 
#np.savetxt("filename.txt", KriG)