A QGIS Plugin to perform Hotspot analysis based on the Python Spatial Analysis Library - PySAL.
NB: QGIS3 version of the plugin is available at: https://github.com/danioxoli/HotSpotAnalysis_Plugin/tree/qgis3
The Hotspot analysis plugin associates the Z-scores and p-values (under Complete Spatial Randomness hypothesis) of the Gi* local statistic (Getis and Ord, 1992; Getis and Ord, 1996), Anselin Local Moran's I (Anselin, 1995) and Local Moran Bivariate (Wartenberg, 1985; Anselin et al., 2002) for each feature of a shapefile, with an assigned projected coordinate system and (at least) an associated numerical attribute. Output layer allows to indentify hotspots (or coldspots) in the input dataset as well as their statistical significance.
For what it concerns Gi* local statistic, positive and statistically significant Z-score indicates intense cluster of high values (hotspot). Negative and statistically significant Z-score indicates intense cluster of low values (coldspot). With respect to the Local Moran's I (and its bivariate counterpart, the Local Moran Bivariate), Z-scores are translated into quadrant values (q) which depict presence of Clusters or Outliers within the dataset. Significance is computed, based on user's choice, against normality assumption or using permutations approach. Please consider the aforementioned litterature references for detailed information.
Spatial relation between point features is modeled using a Fixed Distance Band (expressed with the same unit of measure of the projected coordinate system of the input point shapefile) or optionally using the K-nearest neighboor approach. For polygon shapefile analysis, the spatial relation is modeled using queen's case contiguity matrix. For more information, please refer to: Geospatial Analysis - 5th Edition, 2015 - de Smith, Goodchild, Longley
Dependency Requirements:
PySALNumpyScipy
These libraries are not included in the QGIS core libraries and must be installed prior to the use of the plugin through the OSGeo4W Shell on Windows, or through terminal on Ubuntu and macOS (see the following).
Note:If you are using the OSGeo-Live Virtual Machine, you do not need to install any dependency. You can simply install the plugin from the offcial QGIS Python Plugins Repository. Depending on the pre-installed Pysal version, some of the functionalities might not be available
1) Open OSGeo4W Shell as Administrator and type:
$ curl https://bootstrap.pypa.io/get-pip.py -o get-pip.py
$ python get-pip.pyDownload get-pip.py , to enable PIP functionalities, which is available at this link: https://bootstrap.pypa.io/get-pip.py
Open OSGeo4W Shell as Administrator and change the working directory where the get-pip.py file is stored and type:
$ python get-pip.pyFurther information regarding PIP installation can be found in:
- https://trac.osgeo.org/osgeo4w/wiki/ExternalPythonPackages
- https://packaging.python.org/en/latest/installing/#install-pip-setuptools-and-wheel
2) Download the following packages according to your Python version and your Operating System characteristics:
Numpy : https://pypi.anaconda.org/carlkl/simple/numpy/
Scipy : https://pypi.anaconda.org/carlkl/simple/scipy/
Note: Default Python version for OSGeo4W Shell is 2.7, select the correct packages accordling to your Operating System, for example: the filename "numpy-1.10.0b1-cp27-none-win_amd64.whl" cp27, amd64 depicts python 2.7 and 64 bit Operating System
Change the directory to where the downloaded packages are stored and type the following commands:
$ pip install numpypackagename
$ pip install scipypackagenameNote: extension must be included with the filename
$ pip install pysalIf the pysal installation with pip generates errors you can try with manual installation as follows:
a) Download the compressed pysal package at this link
The QGIS Python packages are store in this floder:
$ C:\PROGRA~1\QGIS2~1.14\apps\Python27\lib\site-packagesNote: ...\QGIS2~1.14... has to be changed according to your QGIS version
b) Remove the existing folder containing pysal. Then, unzip the package you have downloaded; inside the uncompressed folder you will find a sub-folder called pysal. Copy it into the early mentioned path.
3) Open QGIS:
Go to Plugins -> Manage and Install plugins -> Settings -> Show also experimental plugins
In All plugins tab, look for Hotspot Analysis and tick the checkbox.
A new icon for Hotspot Analysis will appear on the QGIS main panel and in the Vector Menu.
4) If you are interested in the latest unreleased version:
Download or clone the GitHub repository given below into QGIS Python Plugins folder:
https://github.com/danioxoli/HotSpotAnalysis_Plugin/archive/master.zip
Note: default Plugins folder is:
$ cd C:\Users\<your_user_name>\.qgis2\python\plugins1) Open a Terminal and type the commands:
$ sudo apt-get install python-numpy
$ sudo apt-get install python-scipy To install PySAL:
$ sudo pip install pysal2) Open QGIS:
Go to Plugins -> Manage and Install plugins -> Settings -> Show also experimental plugins
In All plugins tab, look for Hotspot Analysis and tick the Checkbox.
A new icon for Hotspot Analysis will appear on the QGIS main panel and in the Vector Menu.
3) If you are interested in the latest unreleased version:
Open a Terminal and change directory to QGIS Plugins directory, default is:
$ cd /usr/share/qgis/python/plugins Clone the GitHub repository into the earlier mentioned path:
$ sudo git clone https://github.com/danioxoli/HotSpotAnalysis_Plugin1) Install dependencies:
If you installed QGIS on your machine using this source : http://www.kyngchaos.com/software/qgis , you would find the required dependencies (Numpy - Scipy - Pysal) in the list of the Optional Python Modules. You can install them directly from here: http://www.kyngchaos.com/software/python
Alternatively:
Open a Terminal and type the commands:
$ sudo pip install numpy
$ sudo pip install scipy
$ sudo easy_install pysal2) Open QGIS:
Go to Plugins -> Manage and Install plugins -> Settings -> Show also experimental plugins
In All plugins tab, look for Hotspot Analysis and tick the Checkbox.
A new icon for Hotspot Analysis will appear on the QGIS main panel and in the Vector Menu.
3) If you are interested in the latest unreleased version:
Change directory to QGIS Plugins directory, default is:
$ cd ~/.qgis2/python/pluginsOpen the folder:
$ open . Download and copy here the unzipped repository folder: https://github.com/danioxoli/HotSpotAnalysis_Plugin/archive/master.zip
or Clone the GitHub repository into the earlier mentioned path:
$ sudo git clone https://github.com/danioxoli/HotSpotAnalysis_PluginAn user guide with demo exercises is included here: https://github.com/danioxoli/HotSpotAnalysis_Plugin/blob/master/test_data/Hotspot_Analysis_UserGuide.pdf
Plese cite this as:
Oxoli, D., Prestifilippo, G., Bertocchi, D., Zurbaràn, M. (2017). Enabling spatial autocorrelation mapping in QGIS: The Hotspot Analysis Plugin. GEAM. GEOINGEGNERIA AMBIENTALE E MINERARIA, 151(2), 45-50.
Latest presentation available here: http://www.slideshare.net/danieleoxoli/hotspot-analysis-with-qgis-foss4git-2017
Note: part of this material might be based on the old version of the plugin!
- Enabled the use of k-nearest neighbor spatial weights matrix for point shapefiles
- Minor bugs fixed
- Enabled Anselin Local Moran's I and Bivariate Local Moran computation
- Minor bugs fixed
- The plugin has been officially published on the QGIS plugin repository
- Minor bugs fixed
- Enabled the use of negative numerical attributes Getis and Ord, 1996
- Enabled polygon shapefiles as input using queen's case contiguity spatial weight matrix.
- Enabled the possibility of selecting between normality assumption (default) and standard normal approximation from permutations to compute Gi* Z-scores and associated p-values.
New check botton to eneable the usage of row standardized spatial weights
With this new version, the output layer is displayed with an automatic style which enables hotspot and coldspot visualization. Moreover, a default Fixed Distance Band is dispalyed. This latter represents the minimum distance to ensure at least 1 neighbor to any element of the dataset in order to compute spatial weights for Gi*
The current version does not require Pyshp as well as to specify the feature coordinates as two separate fields in the attribute table of the input shapefile. Only the numerical attribute must be included and selected using the graphical interface on QGIS. Nevertheless, be sure that your input shapefile is projected. The unit of measure in which you express the analysis distance must agree with the one of the projected coordinate system of your input layer.
- Test on the new functionalities addedd
- improve GUI appereance
- Update User guide and documentation for Anselin Local Moran's I and Bivariate Local Moran computation
- Create a new branch of the sotware without external Python dependnecies
Bug tracker and Wiki
The Hotspot Analysis plugin is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation
Copyright © 2017 Daniele Oxoli/ Mayra Zurbaràn/ Gabriele Prestifilippo/ Stanly Shaji - Politecnico Di Milano.
E-mail: daniele.oxoli@polimi.it