REDD pilot project in Cameroon - method development and first results.

This paper describes a method for monitoring the implementation of the REDD (Reducing Emissions from Deforestation and Degradation) process for a pilot project in Cameroon. A method was developed and tested in a test area in the transition zone between tropical evergreen forest and savannah in Cameroon. The method developed involved a series of processing steps which can be divided into three groups: (1) pre-processing steps including cloud masking, geometric and radiometric adjustment as well as topographic normalization; (2) forest masking including segmentation, classification and manual correction and finally (3) land cover classification including spectral signature analysis and training. The last part showed the most crucial problem encountered in this test phase: due to shifting cultivation, it is difficult to differentiate cropland in the stage, where it is left to bush fallow and natural bush land and secondary forest. The method and results were compared to similar studies already performed in Bolivia and Vanuatu and the differences are discussed. The results gained in the Cameroon test area showed, that 529.786 km2 (equivalent to 3.45 % of the forested area) was deforested between 1990 and 2000, further 88.547 km2 (3.44 % of the forested area) between 2000 and 2005. Further investigations will include more detailed studies on the issue of degradation and the applicability of the method for the roll-out on the whole country of Cameroon as well as on other countries in the Congo region. 1 BACKGROUND & OBJECTIVES At the UNFCCC and Kyoto Protocol Conference of Parties (COP) meeting in Montreal, 2005, the governments of Papua New Guinea and Costa Rica, supported by Latin American and African countries, submitted a proposal for the consideration of reducing emissions from deforestation and degradation in developing countries (REDD) as a mechanism for the post-Kyoto reporting. Avoiding deforestation is considered to be a contribution to the reduction of green house gas (GHG) emissions. Parties agreed to a two-year process of evaluation of the issue by initiating REDD pilot projects to better understand the REDD process. Under the auspices of the European Space Agency (ESA) GMES Service Element on Forest Monitoring a pilot project was developed for REDD implementation and testing in Bolivia and Cameroon with user consultation and endorsement. The overall aim of the REDD pilot in Cameroon is to establish baseline projections of emissions caused by deforestation (using Earth Observation) combined with regional projections of degradation nested in a wall-to-wall approach. In order to assess deforestation and degradation at a national level a two tier remote sensing analysis to provide forest area maps and forest cover change maps at specific time periods of 1990-2000-2005 using high resolution data for the whole country was proposed. Several methodological studies for the REDD process are currently underway in tropical rainforest countries. An example of a study that also uses high resolution satellite data is the one being undertaken in the pacific island state of Vanuatu (Herold et al. 2007). In this study, the deforestation between 1990 and 2000 was assessed using Landsat, Spot and Aster data. The approach consists of several pre-processing steps such as co-registration, image mosaicking, water and cloud masking and image normalization. The main procedure is a wall-to-wall deforestation approach on a pixel basis with manual interaction. The second similar work has been done in Bolivia (Steininger et al. 2001; Killeen et al. 2007) also based on Landsat imagery. The time span was from the mid-1980ies until 2005 with a focus on the Santa Cruz department and thus was not a full country wall-to-wall mapping. Also, the mapping is restricted to the forest zone with a precipitation of more than 1000 mm per year. The procedure was a pixel-based mapping with a common pixel size of 60 m. A classification of different forest types and also other land cover classes was performed. Training and evaluation was based on field trips and aerial surveys. A study dealing with the topic of selective logging, a type of forest degradation has been carried out in Brazil (Souza et al., 2005). The authors used a contextual classification algorithm (CCA) to assess logging and fire-derived canopy damages.