We explore methodologies that allow conclusions to be drawn from the large Poverty Environment Network (PEN) dataset. First, we characterize the diverse parts of the tropics in terms of factors that influence forest resources, access and livelihoods. Secondly, for the conclusions drawn from the site-based analysis to be useful for roader policy recommendations, we need to know the extrapolation domains. We compared the characteristics of landscapes where PEN studies took place with overall tropical landscapes, and those of PEN villages with 'random' villages. Both methods rely on variables derived from global data sets using spatial analysis. Thirdly, we study the relationships of livelihoods and forests using multilevel regression analysis. Our study suggests that for global comparative analysis, it is necessary to identify the overall variation of the system of interest, to define the extrapolation domain of the samples/study sites, and to address relationships that by nature involve multiple scale processes. Available global data set, advances in spatial techniques and relatively cheap computer storage and computational power allow such analysis to be done, adding value through global comparative analysis of the interesting site-level findings.
Seminar 13 Mar 13 - Session 4 - Who drives deforestation in Kalimantan by DGa...
Seminar13 Mar 2013 - Sesion 1 - Analysis of forest-livelihoods nexus global data set by SDewi
1. Analysis of forest-livelihoods nexus:
how can global data set help?
Sonya Dewi, Brian Belcher,
Atie Puntodewo
“Tree cover transitions & investment in multicolored economy”
One Day Seminar, March 13 2013, Bogor
2. Outline
• PEN study and dataset
• Characterization of the diverse parts of the
tropics
• Extrapolation domain of large scale,
comparative studies
• Multilevel analysis of relationships of
livelihoods and forests
4. The PENis a…set
PEN data
• Large (360 villages, 10,000+ households)
• pan-tropical (25 countries, 3 regions)
• collection of detailed and (intended) high-quality
data by
• 38 PhD student partners on the
• poverty-forest (environment) nexus at the household
level,
Aim: produce the most comprehensive (breadth and
depth) analysis of poverty-forest links
6. Global dataset
Spatial analysis of global maps clipped for the tropics only:
• Global land cover: JRC, 2006. The Global Land Cover 2006
• Ecoregion: WWF, 2005. WWF Terrestrial Ecoregions
• Population density: CIESIN, 2005. Estimated Population Density
2005 from Gridded population of the World (GPW) version 2
• Settlement locations: World Gazeteer – population figures for cities,
places, regions, countries (http://world-gazeteer.com/)
• Roads: DMA, 2006. Digital Chart of the World, Roads
• Protected areas: UNEP, 2010. World Database on Protected Areas
(WDPA)
• Elevation: GTOPO30
• Watersheds: WWF Conservation Science Program, 2009.
Hydrological basins derived from HydroSHEDS.
7. Ecosystem
Scale 1:10,000,000
Source: WWF, 2005. WWF Terrestrial
Ecoregions
17. - Multi-level
o Hh characteristics
o Resource base
o Access to market
o Access to
resources
o …
- Policies should
address multiple-
level issues
18. Coeff Signif. Coeff Signif.
Total income (ln) Watershed-level variables
Intercept 0.805 Dry broadleaved forest
Household-level variables compared to Moist broadleaved
Members -0.159 ** forest -0.356 **
Age of head -0.003 ** Grassland, savanna, shrubland -0.747 **
Number of adults eq 0.162 ** Coniferous forest 0.733 **
Female headed -0.235 ** Montane grassland -0.737 *
Percent of forest land managed -0.001 Desert and xeric shrubland -1.240 **
Percent of agricultural land Distance to core forest 0.154 **
managed -0.04 % Core forest 1.125 **
Total land (ln ha) 0.183 ** Mean Population dens 0.632 **
Herfindahl index (diversity of FT x dry broadleaved forest -0.077
source of income) ** FT x grassland, savanna,
Village-level variables shrubland -0.217 **
Road density 0.443 ** FT x coniferous forest -0.275 **
Population density 2.84 ** FT x montane grassland -0.474 **
Road dens x Population dens -0.298 ** FT x Desert and xeric shrubland -0.133
Distance to Protected Areas 0.079 ** Village x WS-level
Sub-montane compared to Population density -0.197 **
lowland -0.227 **
Montane compared to lowland -0.018 **
Sub-alpine compared to lowland -0.675 **
Alpine compared to lowland -0.445 **
19. Global dataset can help …
• Providing context to case studies and
comparative studies at different scales
• Finding the sampling frame and population
• Analysis of typologies; finding extrapolation
domain
• Generating data for multiple and cross-scale
analysis, e.g., with multiple level regression
analysis