Set of exercises that offers training on different techniques to map forests and its degradation. Training is
based on working with the IMPACT toolbox, processing Landsat and Sentinel data. Each of the exercises is shown
on real case in particular area in Eastern Africa. Different types of landscapes are reviewed, and a series of
tools and methods available in IMPACT are tested.
For each exercise the necessary data are available for download. The data exercises for 2., 3. and 4. are
provided also in lighter version, where some heavy files are missing (details under each exercise).
Should you have problems downloading / accessing the data - please contact: Andrea MARELLI
Should you have comments on the exercises or wish to propose other exercises - please contact: Hugh EVA
Overview and testing of all the features that IMPACT tool offers.
Example on the area of Ogo Mountains in Somalia. Processing and mosaicking three Landsat scenes from the year 2000.
Subsequently producing a pixel classification, reviewing the results and recoding the classification (forest / non-forest).
In the end a vector segmentation should be created.
Light version of the dataset contains only one image which is already process and is covering only one part of Ogo area.
The provided processed image can be clipped by Ogo shape file and then the exercise can be continued from the step g.)
Case study over the Solio area in Kenya. Sentinel-2 data are used to map the proportions of grassland and woodland in
Solio protected area. Different types of classifications (Automatic, K-means, NDVI_cluster, PCA_cluster, UNMIX cluster)
are tested in order to choose the most precise one, which should be used to create a vector segmentation.
Light version of this dataset contains all necessary images to run all the steps of the exercise.
The only difference is that the Landsat images are already processed and clipped for the Solio area
Example in Mwingi region in Kenya. Estimation of the areas of forest removed or degraded in two periods
of time, using Landsat and Sentinel 2 images. Subsequently corresponding amount of carbon released is counted
using the carbEF module.
Light version of the dataset include everything except General data folder which is
needed only for very last part of this exercise. General data folder contains activity and biomass maps
for Kenya, Sudan, Ethiopia, Mozambique, Rwanda and Uganda
Case study on Nandi forest in Kenya. A random sample of boxes (fishnet) over the area is used to estimate
degradation from 2001 to 2015. Subsequently a vector legend is created to identify changes (deforestation and degradation).
In the end the statistic tool is used to calculate the changes that have occurred.