LandScriptDeforestMap

Deforestation analysis on classified images.

The following code is aimed to help with deforestation analysis, using classified images - such as from MapBiomas.

Download the package in R/Rstudio

If using windows as an OS (Operating System), it is needed to first download Rtools from the website (https://cran.r-project.org/bin/windows/Rtools/).

Then, to download LandScriptDeforestMap through R, you must download the devtools package beforehand (https://cran.r-project.org/web/packages/devtools/index.html).

To install and open the devtools package using the command line in R, run the following codes:

install.packages("devtools")
library(devtools)

If the package (devtools) ir properly installed and opened, you must then install, and load, the LandScriptDeforestMap package through the github link with the code:

devtools::install_github("Lappicy/LandScriptDeforestMap")
library(LandScriptDeforestMap)

Usage

By defining a vector file (geopackage or shapefile), also called here as a geospatial file, and a folder with the needed images, this package may produce three main results for a proposed mesh (the refinement may be adjusted by an argument on the functions used):

1. A vector (geospatial) file with the frequency for each class inside each grid cell (from the proposed mesh);
2. Compare the deforestation time series with the growth of known classes (agriculture, pasture, urban and mining);
3. A map showing heatmaps for the acumulated, or for a given year, deforestation (or one of the other known classes).

Practical example

As an example, within this github there is data for a study case of the brazilian Protected Area named "Caverna do Maroaga (Presidente Figueiredo)".

Data used

To open the data used here, one must simply write at the console the function "example.files()":

example.files()

The "CavernaMaroaga" is, inside the R environment, a sf object with the boundaries for "Caverda do Maroaga". The "MapBiomas_8_example" is a list, that contains the collection 8.0 from MapBiomas for the years 2013 to 2022 (10 elements on this list). These images were already cut to only have the extent of "Caverda do Maroaga". This was made to make the folder lighter, but using the original images should not affect the results or the processing time by any significant standards.

Other way to acess data

Another way is to manually download everything from this github directory. There is an "Example application" folder, within it you can find a folder entitles "Data" which has 2 subdirectories. One, "GPKG", has a geopackage file with the boundaries of "Caverda do Maroaga", and the folder entitled "Mapbiomas8" has MapBiomas images (collection 8.0) from 2013 to 2022.

The code itself

There are only three main functions that one must run for this practical example. One does the analysis and creates a sf object with all the necessary data for future analysis while the other two are graphical functions - one for a time series graph for deforestation and the possible driving vectors and the other is a map showing the locations of such deforestation or growth of these other vectors.

Analysis with the Growth.Analysis function

To run the analysis (this may take a moment, possibly up to ~10 minutes depending on your computer specifications), one must simply define the geospatial data used (CavernaMaroaga) and the folder on which you downloaded the images or the MapBiomas object created (MapBiomas_8_example). This will output an object with a table like format, named "FinalAnalysis". The first and last two lines of it may be seen after the code.

FinalAnalysis <- Growth.Analysis(geo.file = CavernaMaroaga,
                                 tif.folder = MapBiomas_8_example,
                                 mesh.size = 0.045,
                                 output.folder = "Results/",
                                 output.name = "Analysis_CavernaMaroaga",
                                 MAPBIOMAS = 8)

ID_mesh	Country	Category	Name	Year	Deforestation	Reforestation	Growth_Urban	Growth_Mining	Growth_Pasture	Growth_Agriculture	Forest	NonForest	Water	Others	Urban	Mining	Pasture	Agriculture	0	3	4	6	11	12	15	24	30	33	41
1	Brazil	Environmental Protection Area	Caverna do Maroaga	2013	NA	NA	NA	NA	NA	NA	9.09	0	0	0	0	0	0	0	NA	9.09	NA	NA	NA	NA	NA	NA	NA	NA	NA
1	Brazil	Environmental Protection Area	Caverna do Maroaga	2014	0	0	0	0	0	0	9.09	0	0	0	0	0	0	0	NA	NA	9.09	NA	NA	NA	NA	NA	NA	NA	NA
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
221	Brazil	Environmental Protection Area	Caverna do Maroaga	2021	0	0	0	0	0	0	1.2141	0.0837	0.0252	0	0	0	0	0	NA	1.2141	NA	NA	0.0405	0.0432	NA	NA	NA	0.0252	NA
221	Brazil	Environmental Protection Area	Caverna do Maroaga	2022	0.0144	0	0	0	0	0	1.1997	0.0396	0.0837	0	0	0	0	0	NA	1.1997	NA	NA	NA	0.0396	NA	NA	NA	0.0837	NA

Correlation analysis with the graphical.timeseries() function

Below it is shown the plotted time series of deforestation with the known drivers evaluated (agriculture, mining, urban growth and pasture) for "Caverna do Maroaga". The correlation between the deforestation and these other classes were calculated, with the highest one (Pasture) being shown as a subtitle. To run this one must write:

graphical.timeseries(proxy.table = FinalAnalysis,
                     comparison.names = c("Growth_Agriculture", "Growth_Mining",
                                          "Growth_Pasture", "Growth_Urban"),
                     comparison.color = c("darkorange", "grey50",  "#EA9999", "purple"),
                     save.as = "Results/Deforestation vs Growth.png",
                     title.name = "Analysis for Caverna do Maroaga")

Output map from the map.layout() function

Lastely, we bring the result of the outputed map for the acumulated deforestation in "Caverna do Maroaga". To achieve this, you should run the following code:

map.layout(mesh.data = FinalAnalysis,
           class = "Deforestation",
           year.used = "all",
           col.limits = c(0, 1, 2, 5),
           save.map.as = "Results/Map acumulated deforestation.png")

Other practical examples

The code shown in this github directory has been widely tested throughout the Amazon region, using different spatial resolutions. In the next images we bring some examples for the Guiana Shield Region, which encompasses 6 countries (Brazil, Columbia, French Guiana, Guyana, Suriname and Venezuela). We may see through this map where the deforestation is more agressive, as well as observe the behaviour for other 3 known classes from MapBiomas (agriculture, mining and pasture).

We can take a closer look to the correlations between deforestation and the different classes for the Brazilian part of the Guiana Shield Region with the graphical.timeseries() function. In the subtitle it is already shown which class has the highest correlation with deforestation (+0.70 for Pasture), but this can be visually seen as well. Not only that, we may observe that this relation between deforestation and pasture is very strong in every year except for 1988 and 1989 (which corresponds to the years where the construction of the Balbina hydropower plant took place). This is very important to better understand the deforestation patterns and its evolution throughout the years.