# GeoPAT2: Entropy calculations for local landscapes

**Rstats on Jakub Nowosad's website**, and kindly contributed to R-bloggers]. (You can report issue about the content on this page here)

Want to share your content on R-bloggers? click here if you have a blog, or here if you don't.

# Introduction

**GeoPAT 2** is an open-source software written in C and dedicated to pattern-based spatial and temporal analysis.
Four main types of analysis available in **GeoPAT 2** are (i) search, (ii) change detection, (iii) segmentation, and (iv) clustering.
However, additional applications are also possible, including extracting information about spatial patterns.

Global landscape diversity (based on Shannon entropy of land cover categories in a 9 by 9 km local window). It shows where only one land cover category dominates, and where there is a large number of lc categories. #sil #gis #landscapeecology pic.twitter.com/E7pVkIwwY7

— Jakub Nowosad (@jakub_nowosad) January 15, 2019

In the above case, I was interested in extracting values of Shannon entropy of land cover categories for local landscapes on a global scale. It had prompted a comment from Steffen Ehrmann:

The LC data I have, but it would be awesome to see the code you use in GeoPAT2, would be a good excuse to learn it!

— Steffen Ehrmann (@DerEhrmann) January 16, 2019

Therefore, this blog post describes how to calculate Shannon entropy of land cover categories using **GeoPAT 2** and R.

# Setup

Installation instructions for **GeoPAT 2** can be found at https://github.com/Nowosad/geopat2#Installation and for R at https://cloud.r-project.org/.
Additionally, a few R packages should be installed.

pkgs = c( "rgeopat2", # helper functions for GeoPAT 2 in R "sf", # vector data classes "raster", # raster data classes "dplyr", # data manipulation "tmap", # spatial visualisation "landscapemetrics" # contains example dataset ) to_install = !pkgs %in% installed.packages() if(any(to_install)) { install.packages(pkgs[to_install]) }

Now you have the necessary packages, the next step is to attach the ones we will use.

library(rgeopat2) library(sf) library(raster) library(dplyr) library(tmap) library(landscapemetrics)

# Input data

We will use the `augusta_nlcd`

dataset from the **landscapemetics** package as an example.
It is a raster representing land cover categories for an area of about 270 km^2^ west from Augusta, Georga.

data("augusta_nlcd") augusta_nlcd = deratify(augusta_nlcd, "NLCD.2011.Land.Cover.Class") dir.create("data") writeRaster(augusta_nlcd, "data/augusta_nlcd.tif", overwrite = TRUE) nlcd_colors = c("#000000", "#00F900", "#476BA0", "#D1DDF9", "#DDC9C9", "#D89382", "#ED0000", "#AA0000", "#B2ADA3", "#68AA63", "#1C6330", "#B5C98E", "#dcca8f", "#fde9aa", "#DBD83C", "#AA7028", "#BAD8EA", "#64b3d5") p1 <- tm_shape(augusta_nlcd) + tm_raster("NLCD.2011.Land.Cover.Class", palette = nlcd_colors) + tm_layout(legend.outside = TRUE) p1

# Calculate Shannon entropy

The `gpat_gridhis`

module of **GeoPAT 2** reads the input raster data from a GeoTIFF file and creates an output based on selected size and signature.
Firstly, we need to choose the size of the local landscapes.
The input data has a resolution of 30 meters, and we decided on local landscapes of 600 by 600 meters (0.36 km^2^) by setting both `size`

(`-z`

) and `shift`

(`-f`

) to 20 (map resolution of 30 * 20 = 600 meters).
Secondly, we need to pick a proper signature - in this case, it is `ent`

(Shannon **ent**ropy).
For this signature, we also must disable any value normalization (`-n 'none'`

).

system("gpat_gridhis -i data/augusta_nlcd.tif -o data/augusta_ent.grd -z 20 -f 20 -s 'ent' -n 'none'")

The output is a binary file and we can easily convert it to text with the `gpat_grid2txt`

module.

system("gpat_grid2txt -i data/augusta_ent.grd -o data/augusta_ent.txt")

# Create a vector grid

The output text file can be next processed in any software.
However, we will use R here, as there is an R package for working with **GeoPAT 2** outputs called **rgeopat2**.
One of its functions, `gpat_create_grid()`

, creates a spatial vector object recreating the grid used in the previous **GeoPAT 2** calculations.

augusta_grid = gpat_create_grid("data/augusta_ent.grd.hdr")

This new object contains a number of regular square polygons (aka local landscapes), each having a size of 600 by 600 meters.

p2 = tm_shape(augusta_nlcd) + tm_raster("NLCD.2011.Land.Cover.Class", palette = nlcd_colors) + tm_shape(augusta_grid) + tm_borders() + tm_layout(legend.outside = TRUE) p2

# Read data to R and connect it to the grid

The **rgeopat2** also has a helper function, `gpat_read_txt()`

, for reading text outputs from **GeoPAT 2**.

augusta_ent = gpat_read_txt("data/augusta_ent.txt", signature = "ent")

The new object `augusta_ent`

is a data frame with five columns describing each local landscape - (i) value of Shannon entropy, (ii) number of land cover categories, (iii) area with values in cells^2^, (iv) column number, (v) row number.

head(augusta_ent) ## Shannon_entropy number_of_categories object_size col row ## 1 1.174216 5 400 1 1 ## 2 2.005147 6 400 2 1 ## 3 1.485236 5 400 3 1 ## 4 2.200218 10 400 4 1 ## 5 2.226918 8 400 5 1 ## 6 1.918720 6 400 6 1

Now, we can combine the spatial object with the data frame.

augusta_grid = bind_cols(augusta_grid, augusta_ent)

The output, `augusta_grid`

is a set of polygons, where each is represented by the value of Shannon entropy.
It allows distinguishing the local landscapes with only one or two land cover categories (low values of Shannon entropy) and the ones with many land cover categories (high values of Shannon entropy).

p3 = tm_shape(augusta_nlcd) + tm_raster(legend.show = FALSE, palette = nlcd_colors) + tm_shape(augusta_grid) + tm_polygons("Shannon_entropy") + tm_layout(legend.outside = TRUE) p3

# Summary

This post has shown how to use **GeoPAT 2** to extract certain metric (Shannon entropy), and how to connect the result with its spatial representation for a relatively small area in Georgia, USA.

tmap_arrange(p1, p2, p3, ncol = 1)

However, **GeoPAT 2** was written to handle large spatial rasters, including ones on continental and global scales.
Therefore, I encourage you to try it on your own study area - regardless of its size.
You can also read more about additional applications of **GeoPAT 2** and see more examples at https://github.com/Nowosad/geopat2.

**leave a comment**for the author, please follow the link and comment on their blog:

**Rstats on Jakub Nowosad's website**.

R-bloggers.com offers

**daily e-mail updates**about R news and tutorials about learning R and many other topics. Click here if you're looking to post or find an R/data-science job.

Want to share your content on R-bloggers? click here if you have a blog, or here if you don't.