A first look at historical land cover changes over the Sahel

By Niels Souverijns 17 June 2020
For those who already know us or follow us for quite some time it is no big surprise, we have a big interest and expertise in global land cover monitoring. Over the past years, our land cover team has released several global land cover products. Think about the global annual land cover layers released last year or the annual updated maps for Africa (2015-2019). These maps are extremely useful and broad applicable, but there is however also a large interest in consistent land cover change maps spanning longer time periods in order to meet international requirements for environmental reporting. Let me show you how we are using the Landsat satellite archive in combination with our land cover classification workflow to create consistent historical maps and discover some first results of the Sahel region.


Since the release of the Copernicus Global Land Cover Maps for 2015, the land cover team did not rest on its laurels. Apart from static maps for one time frame, there is a community-wide interest in land cover change maps. Within the Copernicus global land service, we are currently finalizing the production of the global land cover change products from 2015 onward at a resolution of 100 m, which will be released by mid-2020.

Next to near real-time land cover (change) mapping, registration of historical land cover changes and dynamics are of high interest to policy-makers and scientists. Apart from better understanding the ecological impact of anthropogenic and climatological impacts, land use changes are considered a vital driver of global environmental change. The Sahel for example has experienced large-scale transitions in land cover over the last decades. A good inventory of these changes is needed to determine its drivers and turning points within this ecosystem. This is the main goal of the UTURN project.


Up to now, most of our land cover maps have been produced using PROBA-V satellite data. But when we want to produce maps before and after the PROBA-V era, we have to consult other satellite data sources. That’s why we are adopting our land cover classification workflow to other sensors as well in order to achieve land cover maps for the past, present and future.


Sensor-generic workflow developed by VITO remote sensing to achieve land cover maps at different spatial resolutions for different time periods. Details about the Copernicus land cover classification workflow can be found in Buchhorn et al. (2020).

Landsat provides a continuous mapping record since 1972. By feeding this satellite data into our land classification algorithm, we can produce 30 m resolution land cover maps for the past.
The large data record of Landsat is available on the Google servers and has been partly preprocessed using Google Earth Engine. Colleague Marcel Buchhorn already shared some examples on how to use the Copernicus land cover maps in Google Earth Engine so you can use and process the maps to your specific needs.

Despite the available Landsat record since the seventies, a large number of data gaps is present in earliest Landsat missions, compelling us to use data composites for these time periods (White et al., 2014). This approach allowed us to create independent land cover and fraction cover maps of the Sahel for the period 1984-2015. The Hidden Markov Model is subsequently applied on the full time series of land cover maps, in order to achieve temporally consistent and coherent land cover products, and erase illogical or spurious land cover changes (Abercrombie & Friedl, 2016).

Nigeria 2015 land cover map

Discrete land cover classification at the Nigeria-Niger border for 2015.

Case studies over the Sahel

The Sahel is a region that has experienced substantial land cover changes over the last 30 years. One of the main anthropogenic influences over the Sahel is the expansion of cropland since the turn of the century. These human-induced changes are nicely captured in e.g. the resulting maps below.
Next to large expansions in cropland near the Nigeria-Niger border we also observe an intensification within existing agricultural areas. Near the cities in the north of Nigeria, in 2000, patches of grassland and shrubs were observed, whereas in 2015, these are almost completely converted to cropland.

Discrete land cover classification (upper) and crop cover fraction (bottom) at the Nigeria-Niger border for 2000 (left) and 2015 (right). Higher saturation indicates higher crop cover fractions.

Next to anthropogenic influences, the Sahel region experienced a long period of drought in the sixties and seventies. This lead to various claims of desertification of the Sahel. Recent research however showed a greening of these parts of the Sahel induced by increased precipitation numbers since the early nineties. The example below for Mali shows the expansion of herbaceous vegetation to the northern parts of the Sahel since 1988, confirming this greening. Apart from climatological influences related to increases in precipitation numbers, also grazing policies implemented by the local government may have contributed to the increase in herbaceous vegetation in the northern parts of the Sahel.

Discrete land cover classification (upper) and herbaceous vegetation cover fraction (bottom) in Mali for 1988 (left) and 2015 (right). Higher saturation of yellow indicates higher herbaceous vegetation cover fractions.

Next up? Annual land cover and fraction cover maps

These two examples showcase the capability of our land cover workflow to detect changes in land cover. In the next phase, we will focus on providing annual historical maps for the entire Sahel region for the period 2010-2015 and five year intervals for periods before 2010. These will be used to identify ecosystem tipping points within the UTURN project. An example of these historical maps for Sudan is given below.

UTURN Sudan discrete land cover classification

The resulting land cover and cover fraction maps will be published on Google Earth Engine and Zenodo by the end of summer, as well as a scientific paper presenting the methodology and full maps. But first we’re looking forward to the release of the Copernicus annual global land cover products at 100 m resolution for the period 2015-2020 to be published in a couple of weeks. Subscribe to our remote sensing newsletter and follow us on https://twitter.com/VITO_RS_ to stay informed.

Like this article? Share it on
Niels Souverijns
An article by
Niels Souverijns
R&D Professional

Related posts

The Copernicus Data Space Ecosystem, a game changer in the world of EO
  • EO Data ,
  • Copernicus ,
  • Terrascope ,
  • Ecosystem services ,
  • Copernicus Data Space Ecosystem

The Copernicus Data Space Ecosystem, a game changer in the world of EO

By Dennis Clarijs 26.01.2024

On December 16th, 2022, the new Copernicus Data Space Ecosystem project kicked off with the stringent ambition to become the public go-to platform and..

Lees meer
EXPLORE-VN: water quality information for Vietnam's inland & coastal waters
  • Copernicus ,
  • Sentinel ,
  • water quality monitoring ,
  • EOData

EXPLORE-VN: water quality information for Vietnam's inland & coastal waters

By Ils Reusen 07.11.2023

In an ever-changing environment where the consequences of climate change are increasing significantly, the use of Copernicus satellite data becomes even..

Lees meer
Simplify your processing with the WorldCover annual composites
  • Sentinel-2 ,
  • land cover ,
  • global land cover ,
  • Sentinel-1 ,
  • worldcover

Simplify your processing with the WorldCover annual composites

By Daniele Zanaga 21.06.2023

During the development of ESA WorldCover 2020 and 2021, the first global land cover maps at 10 meters resolution based on Sentinel-1 and Sentinel-2, we..

Lees meer