Book Image

Hands-On Geospatial Analysis with R and QGIS

By : Shammunul Islam, Brad Hamson
Book Image

Hands-On Geospatial Analysis with R and QGIS

By: Shammunul Islam, Brad Hamson

Overview of this book

Managing spatial data has always been challenging and it's getting more complex as the size of data increases. Spatial data is actually big data and you need different tools and techniques to work your way around to model and create different workflows. R and QGIS have powerful features that can make this job easier. This book is your companion for applying machine learning algorithms on GIS and remote sensing data. You’ll start by gaining an understanding of the nature of spatial data and installing R and QGIS. Then, you’ll learn how to use different R packages to import, export, and visualize data, before doing the same in QGIS. Screenshots are included to ease your understanding. Moving on, you’ll learn about different aspects of managing and analyzing spatial data, before diving into advanced topics. You’ll create powerful data visualizations using ggplot2, ggmap, raster, and other packages of R. You’ll learn how to use QGIS 3.2.2 to visualize and manage (create, edit, and format) spatial data. Different types of spatial analysis are also covered using R. Finally, you’ll work with landslide data from Bangladesh to create a landslide susceptibility map using different machine learning algorithms. By reading this book, you’ll transition from being a beginner to an intermediate user of GIS and remote sensing data in no time.
Table of Contents (12 chapters)
8
GRASS, Graphical Modelers, and Web Mapping

Modeling autocorrelation

We have so far learned how to check for spatial autocorrelation, but we have not yet learned how to incorporate this into our model. We'll now learn how to do this.

Spatial autoregression

The spatial autoregression model considers the dependence of the value upon near regions and integrates that dependence into the data-generation process. We'll be working with a Simultaneous Autoregressive (SAR) model here. One important parameter here is lambda, which indicates the level of spatial dependence, where a positive value indicates positive correlation, a negative value indicates negative correlation, and zero means no spatial dependence.

First, we fit a SAR model without a predictor (that is...