topic Re: Weight of evidence in Python Questions

Weight of evidence

paulbonnard — Mon, 16 Jul 2018 12:09:53 GMT

Hello everybody,

I am working on a project where we are using weight of evidence and i have to implement it on python.

I have some question but not about the implementation, maye i am not in the right place, but if someone knows about woe and want to answer to my question it will be very helpful for me.

Thank you !

Re: Weight of evidence

JoeBorgione — Mon, 16 Jul 2018 12:25:25 GMT

I'm not a statistician and just now learned how to spell WOE, but I found this:

woe function | R Documentation

Re: Weight of evidence

paulbonnard — Mon, 16 Jul 2018 13:46:23 GMT

Hi Joe,

To set up quickly my project, i am studying the possibilities of landslide on a given area.

In this area we have several variables like lithology or slope, and in each variable we have several classes, for example for slope we can have in degrees 0-5, 5-15, 15-30 ect wich are 3 different classes of slope

In fact if we read the document you sent me or wikipedia or wathever, if i understood correctly we can have a woe (wich means a posterior probability) only if we have a landslide.

we have

so if (class i inter landslide) = 0 we have post prob = 0 and by definition woe = 0

however :

here there is no landslide but we have a post prob ..

Do you have any idea why ?

Re: Weight of evidence

DanPatterson_Retired — Mon, 16 Jul 2018 13:47:14 GMT

Showing the code and your reference would be useful if you are wanting to check to see whether your implementation is correct and/or optimized.

Re: Weight of evidence

paulbonnard — Mon, 16 Jul 2018 14:33:04 GMT

In fact this is not a problem with the implementation but maybe a problem of comprehension.

I dont understand why the posterior probability isnt equal to zero whereas there is not landslide in this area.

Here is my code, in fact i want to get the same array i sent a post ago

import numpy as np
import rasterio
#import math
import csv
import os
from itertools import product

def raster_to_numpy(fileraster):
   with rasterio.open(fileraster) as src:
      if src.meta['nodata'] == None:
         print("Warnings : Attention, valeur nodata non renseignée dans le fichier entrée")
         array = src.read(1, masked=True)
         nodata = np.array(array.fill_value, dtype=array.dtype)
         array = array.filled()
      else:
         array = src.read(1)
         meta = src.meta
         meta = meta.values()
         nodata = meta[-1]
   return array,nodata
def variable(raster,nodata):
   variable = list(np.unique(raster))
      if (nodata in variable):
         variable = np.delete(variable,variable.index(nodata))
   return variable
FileDossierRaster = r"C:\Users\user\Desktop\Boulot\Dossier charge"
fileglissement = r"C:\Users\user\Desktop\Boulot\landslides_tests\ASCII\scarp_rb.asc"
fileout = r"C:\Users\user\Desktop\Boulot\Fichiers_csv\sdmuc1.csv"
def attributes(FileDossierRaster,fileglissement,fileout):
   liste_raster = []
   liste_nodata = []
   liste_var = []
   for i in os.listdir(FileDossierRaster):
      if not os.path.isdir(i):
         new_path = os.path.join(FileDossierRaster,i)
         raster_var,nodata = raster_to_numpy(new_path)
         liste_raster.append(raster_var)
         liste_nodata.append(nodata)

         v_ras = variable(raster_var,nodata)
         liste_var.append(v_ras)

   raster_gli,nodata_gli = raster_to_numpy(fileglissement)

   taille = [len(i) for i in liste_var]
   n = np.prod(taille)
   valeurs = np.zeros((n,11+len(liste_raster)))
   combi = list(product(*liste_var)) ## ici je fais les combinaisons

   for i in range(n):
      valeurs[i,0] = i+1
      valeurs[i,1:len(combi[0])+1] = combi ## ici je les mets dans mon tableau
      t = 1
      for k in range(len(liste_raster)):
         t *= np.where(liste_raster == valeurs[i,k+1],1,0) ## possibilé erreur au dernier
         c = np.count_nonzero(t == 1)
         g = np.where(raster_gli == 1,1,0)
         s = np.count_nonzero(t*g == 1)
      valeurs[i,len(combi[0])+1] = c
      valeurs[i,len(combi[0])+2] = s
###########writing csv file#################
   entete = ['numero']
   for i in range(len(liste_var)):
      v = 'variable' + str(i)
      entete.append(v)

entete = entete + ['intersection','Cellules_affectees','Proba_posteriori','c','d','e','f','g']
   with open(fileout,'w') as f:
      w = csv.writer(f, delimiter = ';')
      w.writerow(entete)
      for line in          zip(valeurs[:,0],valeurs[:,1],valeurs[:,2],valeurs[:,3],valeurs[:,4],valeurs[:,5],valeurs[:,6],valeurs[:,7]):
      w.writerow(line)
#######################################
attributes(FileDossierRaster,fileglissement,fileout)

Re: Weight of evidence

DanPatterson_Retired — Mon, 16 Jul 2018 15:10:20 GMT

Paul... your column WOE 10**(-15.ish) is about zero. Are you expecting actual 0.0?

Re: Weight of evidence

AbdelhakEl-fengour1 — Sat, 09 Feb 2019 11:58:21 GMT

Hello Paul,

I would appreciate if you could provide me me WofE extension for ArcView, i have been looking for it for a while. my email (elfengourabdo@gmail.com). Thank you in advance.

Re: Weight of evidence

ParthipanS — Mon, 13 Jan 2020 17:19:24 GMT

Hi Everyone,

I'm working on Landslide susceptibility mapping wherein I've planned to use the Weight of evidence. the independent variables that I am using are slope. aspect, LULC, distance from the road, geology, geomorphology, etc I wanted to know what are the tools that I can use for WofE and how to calculate WofE for the above-mentioned variables. kindly, help me to solve this issue. I deeply appreciate your help and consideration

Re: Weight of evidence

ParthipanS — Mon, 13 Jan 2020 17:20:32 GMT

Parthipan S‌

can you help me to solve WofE?

Re: Weight of evidence

paulbonnard — Tue, 14 Jan 2020 08:01:55 GMT

Hello,

If you used to work with python i've made a program that Can calcul wofe or a particular variable (landslide for exemple) with different raster as input.

Pm me and i will send you the code

Re: Weight of evidence

ParthipanS — Tue, 14 Jan 2020 09:58:36 GMT

His Sir,

Thanks for the response. Yeah, I would like to use landslide

areas as dependent variable and slope, aspect, distance from lineaments,

geology and geomorphology as independent variables. can you send me the

python code it will be very helpful ?

Yours Faithfully,

S. Parthipan,

M.Tech, Geoinformatics,

Department of Geography,

University of Madras, Guindy

Chennai-600025.

Ph: 9585445673

Re: Weight of evidence

ParthipanS — Tue, 14 Jan 2020 10:08:00 GMT

Hi,

Thanks for the response. Yeah, can you send me the code? I've considered the Landside points as dependent variable and slope, aspect, distance from the road, geology, geomorphology as the independent variable. kindly help me to solve this issue it will be very helpful

Re: Weight of evidence

paulbonnard — Tue, 14 Jan 2020 12:44:03 GMT

Okay,

Give me your Email and i will send you the code.

Bonnard Paul

Re: Weight of evidence

MichelleMathias — Tue, 14 Jan 2020 13:00:06 GMT

Hello, may I suggest instead of posting your email in this public thread, please consider following/connecting with each other and then you can send a direct message. These docs will help you with those actions:

Connecting to Other Users

https://community.esri.com/docs/DOC-1079-creating-and-sending-messages-to-other-members

Re: Weight of evidence

ParthipanS — Tue, 14 Jan 2020 13:15:25 GMT

Hi Paul Bonnard,

Please connect/follow me so that we can send a private message.

thank & regards

Parthipan.s

Re: Weight of evidence

paulbonnard — Tue, 14 Jan 2020 13:35:04 GMT

Hello, i have already send you a mail, please look at your gmail, and maybe in the spam

Re: Weight of evidence

ArijitArghya — Sat, 13 Nov 2021 21:41:08 GMT

Hello Paul,

I am working on a project for which I need to identify the location of the landslide based on morphological features using the weight of evidence method. Can you please send me the codes for the weight of evidence model which I can apply in ArcGIS pro?

Regards

Arijit

Re: Weight of evidence

NeilFordyce — Sun, 04 Jun 2023 04:23:18 GMT

Sorry it's taken years to reply to this but I am just getting back to WOFE after a break. I was involved in WOFE with ArcVIEW in the mid 1990's and then with ArcMAP in the early 2000's.

To understand WOFE read Graeme Bonham-Carters' book "GIS for Geoscientists". The code examples are very easy to follow. Also Graeme's early papers.

For other easy to understand papers look for papers by Gary L. Raines.

For more recent work than Gary's look for papers by Vesa Nykanen from Geological Survey of Finland (GTK).

Vesa has done a lot of this sort of work and worked with Gary Raines.

Search for Mark Mihalasky and WOFE to find some other papers e.g. WOFE modelling sediment hosted gold deposits. Mark's PhD is a good read for the basics of WOFE.

GTK has released code for WOFE for ArcGIS PRO. See https://github.com/gtkfi/ArcSDM

Older "versions" of the Spatial Data Modeller for ArcGIS 10 can be found at https://www.ige.unicamp.br/sdm/

The logic for WOFE is very robust and it is relatively easy to program. What is difficult is manipulating the and finding appropriate data sets and understanding the contrast curves etc.