Msitu iko wapi is Kiswahili and means ‘Where is the forest?’ It fits well to introduce my upcoming series as I am currently working on a project to practice my abilities of utilizing satellite data. I expect to publish several posts covering the project in the coming weeks.
I will be using Landsat 5 and Landsat 8 data to evaluate land change issues in the Subsaharan Africa and mainly focus on plugins that you can find in the official plugin repository. Within my study program, we were taught how to process the omnipresent Landsat data with ENVI, ArcMap and Erdas, but I am keen to learn about the capabilities of the open source alternatives.
Main targets for the upcoming series
- Quantify land use change of primary forest in the period from 1985 to 2015,
- learn and discuss the open source tools needed for the series,
- identify future projects that produce interesting outputs and increase my technical skills.
The study area will be set in southwest Kenya and focus on three primary forests. The Kakamega Forest and the South- and North Nandi Forest. These forests are some of the few natural habitats that are mostly spared from the immense population pressure in the equatorial belt of Kenya and provide ecosystem services for the region and record a high diversity of tree and animal species. Whereas the Kakamega complex was given national forest reserve status in 1985, the South- and North Nandi regions are endangered by the possible erection of multipurpose dams and general issues of deforestation for coal production and resource extraction. I chose this study area because I think it will be challenging to find out, whether the given data quality and tools can suffice my plans to map land use changes. Adjacent to the primary forests, we can find secondary forest structures and especially quite heterogeneously spread plots of smallholder farmers. Besides that, I have been in Kakamega two times, which must be enough to compensate the currently impossible ground truth methodologies.
Above, you can see two RGB composites of the study area. The scene from 1985 was created with Landsat 5 data, whereas the recent scene from January 2015 is based on Landsat 8 data. You can already see the advantages of the powerful Landsat 8 setup: The panchromatic band allows us to pan sharpen the data to increase the geometric resolution from 30 to 15 meters. Even the RGB composites show that there have been definitely some changes in forest cover since 1985. The South Nandi Forest seems to have decreased, whereas the Kakamega Forest is more dense.
Below you can browse a slide show of pictures from 2013 I took in Kakamega Forest to get a feeling for the environment in the study area.
Most of my methodology for this series is still in development. Generally, I will use QGIS tools to perform maximum likelihood classifications of Landsat 5 and Landsat 8 data to extract land cover classes for the study area. I expect the most challenging process to be the validation of the data and the comparison of my results. As I am quite new to processing satellite data and only doing this in my spare time, I cannot guarantee scientifically reliable methodologies and results. I will definitely not be able to avoid mistakes, but I will keep my posts updated and fix major mistakes as soon as I realize or get a hint on these.
I’m looking forward to work on the series!