A landscape mosaics approach for characterizing swidden systems from a REDD+ perspective

This document presents a landscape mosaics approach for characterizing swidden systems from a REDD+ (Reducing Emissions from Deforestation and Forest Degradation) perspective. It addresses the challenge of spatial delineation and monitoring of swidden agriculture, often overlooked in land cover/use maps. The practical value lies in enabling a swidden-compatible monitoring, reporting, and verification (MRV) system for future REDD+ frameworks. The approach is tested in a case study in Viengkham District, Northern Laos.

Key Insights

Defining Swidden Systems

Swidden systems, also called shifting cultivation, are characterized by the alternation of cropping and fallow phases, yet their dynamics depend on cultivation techniques, market opportunities, and policies affecting forest and land uses (Nair, 1993; Sanchez, 1976).

Principle 1: Spatial Delineation of Swidden Systems

The lack of spatially explicit knowledge about the location and intensity of swidden cultivation results in a general failure of land cover/land use maps to capture this practice. Studies bypass methodological challenges and apply conventional methods of satellite image analysis in landscapes dominated by swidden agriculture. As a result of methodological gaps, areas under swidden are usually categorized as “Unstocked Forest”, “Unclassified” or even “Degraded Lands” on land cover maps (Padoch et al., 2007).

Principle 2: Delineating swidden landscapes in Northern Laos

Yamamoto, Oberthür, and Lefroy (2009) classified an eight year series of Landsat images into vegetated and bare areas. Using the created eight years land use history of every pixel different crop-fallow rotation cycles were identified. Using landscape mosaics Messerli, Heinimann, and Epprecht (2009) developed an approach to distinguish swidden landscapes from other basic land use types at the national level based on official national land cover data of Laos.

Principle 3: Understanding Swidden within REDD+

Driven by the negative perception that swidden degrades vegetation and contributes to deforestation, swidden agriculture is explicitly listed today as a main cause of the emissions related to this form of land cover change by many tropical countries (Griffiths, 2008). Consequently, swidden landscapes have become a central focus of actions against climate change.

Principle 4: The Mosaic Approach

While the land cover types of the single pixels change constantly from one year to the next, the overall composition of a landscape remains the same in a stable swidden system. Hence, we focused on the patterns that are created from the different land cover types and the way the patterns are repeated in similar combinations over larger areas.

Principle 5: Classification of change processes

Changes from one landscape type to another causing forest loss were classified as a type of deforestation. Particularly a change from Forest and wooded landscapes to Swidden landscapes was called Deforest-FWL-SWL (see Fig. 3), a change from Forest and wooded landscapes to Permanent agriculture landscapes was called Deforest-FWL-PAL, and a change from Swidden landscapes to Permanent agriculture landscapes was called Deforest-SWL-PAL. Generally, these changes not only imply loss in forest areas but they denote changes of the land use practices.

Key Statistics & Data

• Tropical deforestation caused close to 20% of global anthropogenic CO2 emissions in the 1990s and 12% in 2008 (Le Quere et al., 2009).
• Across the tropics more than 55% of new agricultural land was created at the expense of intact forests, and another 28% came from disturbed forests between 1980 and 2000 (Gibbs et al., 2010).
• In 2007 Swidden landscapes occupied 53% and Forest and wooded landscapes made up the remaining 47% of the area. The shares were 55%, respectively 45% in 2009.
• Half of the area (49.4%) was subject to landscape mosaic change from 2007-2009.
• From 1979 to 2009, the area under swidden increased from 29% to 55% (Fig. 8).

Methodology

The published land cover data used as basic data for developing the approach presented in this paper were created based on Landsat imagery using supervised classification and visual image interpretation (Kongay et al., 2010). The land cover classes include: ‘Dense Forest’ - natural undisturbed forest; ‘Open Forest’ - degraded primary forest due to timber extraction or regenerating after cultivation; ‘Shrub’ shifting cultivation fallow areas older than three years and up to 10-15 years; ‘Recent Fallow’- shifting cultivation fallow areas of up to three years of age; ‘Upland Crop’ - areas currently under cultivation; and finally ‘Paddy Rice’, ‘Residential Areas’ and ‘Clouds’.

In order to derive landscape mosaics from the basic land cover provided for our study area, we followed the general steps introduced by Messerli et al. (2009). We first analyzed spatial patterns of the original land cover data, without interpreting them topically. For every pixel we found out by what type of land cover it was surrounded. Using a moving window technique we computed the share of each land cover class in the neighborhood of each pixel and attributed this information to the pixel. “Neighborhood” was defined through the choice of the window size. Based on previous studies on the relation of accessibility and land cover change in Laos a 2 km window size was used (Heinimann, 2006).

Implications and Conclusions

The landscape mosaics approach facilitates the analysis of land use trends and the detection of degradation as well as reforestation. The authors conclude for their study area that (1) no pioneering swidden agriculture took place, neither between 2007 and 2009, nor between 1979 and 2009; (2) the swidden system is generally intensifying from long crop-fallow cycles to short crop-fallow cycles, (3) there is a potential problem of ‘carbon leakage’ as forest conservation within and near to the protected area most likely trigger degradation and deforestation outside the park boundaries. The landscape mosaics could be used for landscape change monitoring and carbon monitoring in a swidden compatible MRV-system of a future REDD+ framework.