Enabling Change Mapping in Eastern Africa

Enabling Local Monitoring of Landscape Change Across Eastern Africa

LARSE Investigators: Z Yang, WB Cohen

Storage of atmospheric carbon is an important ecosystem service of healthy forests and woodlands because it mitigates the effects of anthropogenic greenhouse gas emissions. International reporting of this service places a premium on the specificity and precision of monitoring data used to estimate carbon storage or emission. An inventory of land cover change is a critical component of most national-level accounting systems, and the Landsat series of satellites is a uniquely positioned to provide this land cover change “activity data.” In Eastern Africa, there are already high-quality Landsat-based cover maps for 2 or 3 points in time. However, these maps do not provide the annual land cover change information needed for higher-tier IPCC reporting, and land cover changes inferred from independent maps at different dates cannot easily be assigned a level of uncertainty.

We will work with the SERVIR Hub institution (a joint venture between NASA and the U.S. Agency for International Development), opened in 2008 in the Regional Centre for Mapping of Resources for Development (RCMRD) to develop innovative and well-targeted maps to meet the accounting needs of RCMRD’s member states. Our strategy is to adapt state-of-the-art land cover change detection approaches to conditions and classes relevant for East Africa. Several different change detection algorithms will be run in parallel, and locally developed reference data will be used to calibrate a model that optimally integrates the output maps from each separate algorithm. This strategy was developed under an inter-agency, national change mapping project in the United States: the Landscape Change Monitoring System (LCMS). LCMS pilot research strongly suggests the value of integrating an ensemble of different approaches to map cover change. Cloud computing has only recently made this data- and processing-intensive approach feasible, and the fact that images no longer have to be downloaded and stored locally circumvents traditional access barriers in bandwidth-limited locations.

At the end of Year 1, statistical estimates (with uncertainties) of land cover change going back to 1999 will be produced for seven countries: Kenya, Malawi, Ethiopia, Zambia, Tanzania, Rwanda, and Uganda. These estimates will derive from a designed sample of land cover change observations developed through expert interpretation of the Landsat historical archive and available high-resolution imagery. A tool called “TimeSync,” which was developed by the project team, will be used to deliver the needed Landsat imagery and facilitate these observations. During Years 2 and 3, RCMRD will create, with Project support, annual maps of land cover change for the seven target countries using the above ensemble approach. Calibration and validation of these ensemble maps will be supported by the same sample of Landsat-based reference data supporting the statistical estimates delivered in Year 1. Engagement with RCMRD and the seven member countries in the areas of reference data collection and product delivery will be coordinated through the SilvaCarbon program to enhance the efficiency of technical transfer. Methods implemented here are globally scalable, and could easily be extended within and beyond the East Africa Hub.

Collaborators: Sean P. Healey (PI –USFS RMRS); Sylvia Wilson (USGS SilvaCarbon Program); Phoebe Oduor & Edward Ouko (RCMRD)

Funding: NASA