Skip to content

Rwanda’s Nyungwe Forest

Nyungwe Forest land use changes

The land use changes in Rwanda’s Nyungwe Forest have been minimal since 2017 according to Esri’s Sentinel-2 land use series of processed Landsat satellite images.

Conservation Efforts

In 1933 the first conservation efforts began for Rwanda’s Nyungwe Forest area. This national treasure has continued to be preserved via presidential decrees (the Presidential Decree of 26th /04/1974) and the involvement of the Rwanda Development Board (RDB) in 2008 via the Organic Law N° 53/2008 of 02/9/2008. But, legal dictates are one thing, forest preservation must be evident in actual conservation of the woodland area.

The monitoring of these conservation efforts has recently been greatly simplified with the release of the Sentinel-2 10m land cover time series. This was produced by Impact Observatory, Microsoft, and Esri (see the ArcGIS REST Services Directory here). This time series shows current land cover / land use as determined by satellite imagery and an artificial intelligence (AI) categorization algorithm in 10m pixels.

In this exciting new dataset, each 10m area of the world is identified as either: 1) Water, 2) Trees, 3) Flooded vegetation, 4) Crops, 5) Built Area, 6) Bare ground, 7) Snow/Ice or 8) Rangeland. If no valid observations have been made due to persistent cloud cover, the area is categorized as 9) Clouds. These 9 area types are color coded for each 10m pixel of the GeoTIFF files that cover the earth’s land surface using Esri’s AI algorithm. Esri offers these tiles in annual series at their Sentinel-2 site – and these tiles are available for each year since 2017.

Has the forest really been preserved?

For the question we ask in this post – Has the forest really been preserved? – we can use this Sentinel-2 series to answer our question.

First, we stipulate that for our analysis, we will use Esri’s categorizations of land use as indicated above. If land use hasn’t changed since 2017, then we would expect Esri’s Sentinel-2 tiles to have the same color-coded pixels in the Nyungwe Forest in the 2022 series as it had in the 2017 series.

To test this hypothesis, we use a very simple fact about color-coded images. Each colored pixel in an image is really represented as a number (actually, usually as set of three numbers, one for it’s red-value, one for it’s green-value and one for it’s blue-value). We can just call this set of three values the pixel’s numerical value. And, we can determine if the numerical value of the pixels is the same in subsequent images of the same area. We just subtract one image from another.

Finding the forest pixels

In order to select the right pixels in our image for our subtraction – we need to create a geographic shape that has very specific latitude and longitude coordinates. In this case, I rummaged around the Internet and found images of the Rwandan portion of the Nyungwe Forest. I extracted the outline of this area as a geographic shape file (available here) and used this shape to identify those pixels in the Sentinel-2 data series that match the Nyungwe Forest area.

Nyungwe Forest land use changes


Changes in Rwanda’s Nyungwe Forest

The image above (and at the top of this article) is the result of these subtracted images (the image from 2017 subtracted from the image from 2022, both presented below). The color black in this image represents the value 0.0 – exactly what one would expect if the numerical values of the pixels in the two images are exactly the same. Any colored pixels represent a difference in the two images.

What is obvious from the image above is that most of the pixels are black. Only 1.57% of the pixels in the forest area have any color in this image. This means that there was no change in the Esri categorization land use for 98.4% of the Nyungwe Forest area between 2017 and 2022.

It is pretty clear that land use change is very slow in Nyungwe Forest.

How much is still forest?

Even though the change in Nyungwe Forest is slow – we still need to determine how much of Nyungwe Forest is still forest. Did it change from it’s pristine state?

To answer this question, we accept the Esri definition of forest. Any 10m pixel that is categorized as Trees we will call forest. That is, “Any significant clustering of tall (~15 feet or higher) dense vegetation, typically with a closed or dense canopy; examples: wooded vegetation,  clusters of dense tall vegetation within savannas, plantations, swamp or mangroves (dense/tall vegetation with ephemeral water or canopy too thick to detect water underneath).” – according to Esri’s definition.


Sentinel-2 data for Nyungwe area 2017


Sentinel-2 data for Nyungwe area 2022

The two images above illustrate the actual Esri categorization of the Nyungwe Forest – on the left, from 2017 and on the right from 2022. Although there are slight differences between the two images (captured in the image at the top of this article), how much is still forest?

To answer this question, we just have to count the pixels in the Nyungwe area that are the color green (the code for Trees – or what we call forest).

Using the outline of Nyungwe Forest in Rwanda, I was able to count the pixels that are not categorized as “Trees” by Esri. In 2017 a total of 2.96% of the pixels were not categorized as Trees. In 2022, a total of 3.33% of the Nyungwe Forest pixels were not categorized as Trees.

How much variation occurs year to year?

We can examine each of the Esri tiles that correspond to the Nyungwe Forest area. By doing this, I was able to determine that there is some normal variation in the categorization of pixels from year to year. I found the following:


    • 2017 – 2.96% not Trees

    • 2018 – 3.60% not Trees

    • 2019 – 3.25% not Trees

    • 2020 – 3.22% not Trees

    • 2021 – 3.33% not Trees

    • 2022 – 3.33% not Trees

This suggests that there is normal variation from year to year.

All in all, this means that the conservation efforts for the Nyungwe Forest area is largely successful.

Leave a Reply

Your email address will not be published. Required fields are marked *