Annual Report 2 to the David and Lucile Packard Foundation

Integrating Dynamics of Human Resource Use and Their Effects on Rainforests in Madagascar: Linking Landscape Ecology, Cultural Anthropology, Behavioral Ecology, and Applied Mathematics for a Science-Based Strategy against Deforestation in Madagascar.

Summary of Year 2 (July 2003- June 2004) Activities:

Socioeconomic Field Research Program (SFRP)
In the first year of the project SFRP established a research protocol that would meet the project goals; this protocol included household surveys and village land mapping. In developing this protocol various demographic, socioeconomic and cultural factors were identified that likely influence household-level decisions regarding land and forest resource use. The demographic, socioeconomic and cultural factors of interest include, but are not limited to, household wealth, age structure, dependency ratio, kin network, economic orientation (i.e. subsistence vs market economy), and residency (i.e. how long in the area). While these endogenous factors will shape an individual's - a household's -perception of the relative risk inherent in any decision regarding resource use, they will be played upon by more exogenous factors that define the individual's socioecological landscape, namely land availability, exposure to markets, exposure to protected area authority, type and extent of community-based resource use monitoring. By identifying behavioral mechanisms that underlie household-level decisions regarding mode of agricultural production, human land use practices and resultant environmental impacts can be more accurately predicted.

From June through December 2003 the SFRP, an interdisciplinary team of Malagasy researchers coordinated by Dr. Christopher Holmes conducted a first survey round of 12 target villages, with results described below.

Village Demographics - Understanding the causes of deforestation in Madagascar requires an understanding of the economic and sociocultural factors that lead to alternate patterns of land use and the dynamic relationship between landscape change, land use decisions, and population growth. Around RNP, apparent increasing population densities and shortening fallow periods perpetuate an unsustainable cycle of land use that ultimately results in households having to decide between intensifying agricultural effort through the cultivation of wetland paddy rice -if land is available, or opening new forest land to tavy - which is illegal. In 2003, the average size of target villages ranged from 16 households in the northern village of Morondava to 74 households in the western village of Anjamba (data collected by a Norwegian team in 2003). Average household sizes ranged from 4.42 in the northern village of Tsinjorano to 7.19 in the western village of Sahateja. Average residency period ranged from 14.2 years in Sahateja to 39 years in Tsinjorano.

Educational Investment - Assessment of educational investment revealed considerable variation among the 12 villages; the northern zone village of Ampozasaha reporting the lowest adult-level of education with 11% and the western village of Vohiparara reporting the highest level at 72%. The investment in child education varied substantially across the four Park zones with the southern zone reporting the lowest child school attendance and the northern and southern zones reporting lower child school attendance than the central and western zones.

Rice Production - Reported total household-level rice production varied among the 12 target villages with the northern village of Torotosy reporting the greatest average household-level production at 652.7 kilos and Ampozasaha reporting the lowest at 102 kilos. When aggregated at the zonal level the central zone reported the greatest average production at 387.9 kilos per household and the northern zone reported the lowest average production at 234.6 kilos per household.

The relative contribution of tavy rice and paddy rice to total household production also varied among the 12 villages with, on average 78% of household rice being produced from paddy and 22% being produced from tavy. When aggregated at the zonal level the western zone reported the greatest average paddy production at the household level (317.5 kilos per household). The central zone reported the greatest household-level tavy production (114.5 kilos per household).

When considering the forms of rice consumption at the zonal level, it is the southern zone that reports the most equitable consumption patterns (even distribution between 'market', 'home', and 'other') while close to 80% or greater of the rice produced in the northern central and western zones goes to the households

Considerations - These preliminary results suggest a dynamic ethno-geographic system in which multiple factors influence land-use practices, resource-use decisions and conservation attitudes. We see, for example, that it is the southern villages, which 1) are the furthest from main roads, 2) have the largest average houshold size, 3) have the lowest investment in education, and 4) the lowest recognition of personal benefit received from RNP that are the least likely to support the protection that define Ranomafana National Park. If we consider an interest in seeing the park degazetted as a reflection of one's lack of willingness to forego the short-term gain of resource extraction - clearing forest for cultivation - for the long-tern benefit of resource protection then the southern zone of the park deserves focused attention. Relatedly, it is the western zone, which expressed little change in land use since the park was inaugurated that was most supportive of the park's protection. Such variations existing within the same 'general' population underscores the need to explore land and resource use decisions at the household level. Future stages of this research will model decision-making to predict the breakpoint at which individuals switch between modes of agricultural production.

Village Land mapping - Important to proper assessment of variation in land use practices among the communities surrounding Ranomafana National Park is understanding systems of i) land tenure, ii) land inheritance, and iii) cooperation with and among the households that define these communities. Towards improving our understanding of these systems, in February of 2004 the SFRP initiated its land mapping protocol in the central zone village of Sambivinany.

The protocol involves the i) physical mapping of all plots of land claimed by each of the 47 households in the village and ii) the quantification of several parameters relating the past, current and future activities on each plot (see Appendix 3 for sample data sheets) .
The data collected, and information gained, during household land mapping compliments the report data collected during the socioeconomic surveys of the village households carried out during August-December 2003. In the village of Sambivinany SFRP successfully mapped 274 land plots.

Thus far the mapping has revealed some very interesting results concerning the sharing of and inheritance patterns of claimed land. It seems that 'cooperative' land use practices are breaking down and relative social status in the community is greatly important - that is who you are related to and where do you stand when it comes to inheriting land from a close relative. Such observations are critical when we consider that currently greater than 50% of the local population is 15 years of age or younger; suggesting that any pressure currently being placed on land and other resources will likely only increase in future.

Back to Top

Biological Field Research Program (BFRP)
The objectives of BFRP are to measure variation in plant and animal populations, and to examine how these are related to shifts in human activities and forest cover. The general protocol is to survey along a disturbance gradient from forest edge to intact interior. Surveys are conducted in habitats adjacent to SFRP villages. BFRP fieldwork in the second year included expeditions to six sites in the periphery of Ranomafana National Park, led by Dr. Steig Johnson.

Survey sites are are shown in the map. We collected data on habitat structure (including direct evidence of human disturbance), climate, primate populations (both day-active and nocturnal species), rodent and insectivore communities, chameleons, and bird populations. Moreover, transects, botanical plots, trapping grids and other habitat features were mapped and georeferenced for future spatial analyses. In addition, we established collaborations with University of Helsinki researchers (Ilkka Hanski and Johanna Rainio) to collect beetles (Scarabidae and Carabidae) on our transects. Data from these groups will widen our taxonomic approach, providing information on disturbance impacts on highly diverse insect communities that may vary on much smaller spatial scales than the vertebrate groups in our sample.

Habitat Structure - Habitat structure was compared across and within sites. Seven plots (1600 m2 each) were surveyed at each of the six sites, yielding a total sample of 6,283 trees. Although data on many structural and disturbance variables were collected, initial analyses have indicated that a smaller subset are more variable naturally or more indicative of anthropogenic disturbance. These include: tree diameter, tree height, canopy openness, percentage of herbaceous ground cover, trees cut down, and the presence of trails.

Both natural variation in microhabitat and human disturbance may impact forest structure across the study sites. We examined the effect of several of these predictors on those structural characteristics that varied significantly across the Ranomafana National Park periphery. We found that both elevation and distance to the forest edge are significant predictors of tree diameters and heights. Altitude also influences the percentage of herbaceous ground cover. Interestingly, neither trails nor trees cut were significant predictors of forest structure.

Birds - Bird populations were sampled using point-count protocols. Birds were recorded at 21 points (every 200 m) along each transect. A total of 9,138 individuals from 68 species were observed across sites in the Ranomafana periphery. While populations vary across sites, it is difficult to discern an edge effect from comparisons of interior and edge transects. More discrete distance categories, as well as additional analyses of individual species (or higher taxonomic groupings) or guilds, will help elucidate the impact of edge, disturbance, and natural microhabitat variation on bird populations at Ranomafana.

Chameleons - Chameleons were surveyed using line-transect techniques. Sub-transects were established at 200-m intervals to distances of 2400-2800 m fromthe forest edge along the main transects. Eleven species from two genera (Calumma and Brookesia) have been recorded to date, and 348 individuals have been counted. The relationship of chameleon richness and abundance relative to microhabitat and disturbance will be investigated with additional sampling in the northern and southern zones.

Small Mammals - Rodent and insectivore populations were sampled using trapping grids and pitfall traplines, respectively. We trapped rodents (native forest rats and invasive black rats) in six 40 x 40 m grids from the forest edge to 1600 m into the forest interior. At each study site, animals were sampled during 1800 trap-nights (number of traps x the number sampling nights). Insectivores (tenrecs and invasive shrews) were caught in 3 traplines (with 11 traps each) in both edge and interior locations for a total of 396 trap-nights per site. A total of 207 rodents from 7 species (genera Eliurus, Nesomys, and Rattus) were captured, marked, and released across sites; 207 insectivores from 17 species (genera Hemicentetes, Microgale, Oryzorictes, Setifer, Suncus, and Tenrec) were also recorded.

The rodent community at in our study sites shifts along a disturbance gradient. There is an increase in the abundance of invasive black rats (Rattus rattus) in edge habitats across our study sites; in contrast, there was a tendency for both richness and abundance of native species to decrease closer to human-disturbed areas. This supports previous research in Ranomafana and elsewhere in Madagascar. Whether this division represents competition between endemic and introduced species or divergent habitat preferences in these groups is a question that warrants further investigation upon the completion of sampling across the four park zones (during Y3). The insectivores showed no clear patterns of variation among sites or between edge and interior habitats. However, with additional and repeated sites, a clearer analysis of species turnover related to disturbance will be possible.

Primates - Primate groups were sampled at all sites using line-transect techniques. At each site, lemurs were surveyed along the 4 km main transect, which was divided into two 2-km segments. Each segment was monitored 21-24 times during daylight hours and 5-10 times at night. We found that species richness did not differ substantially in the Ranomafana periphery. There were 3-5 day-active and 3-4 nocturnal species present across sites. These counts do not include the two most rare diurnal species (Hapalemur aureus and Hapalemur simus) and the most rare nocturnal species (Daubentonia madagascariensis). These species were simply too scarce or too patchily distributed to be detected with transect techniques. Richness also did not vary greatly between the edge and interior transect among diurnal species. However, there was a tendency for more nocturnal species in the interior forest. This may be due to the smaller home ranges of these species; their smaller territories may not be productive enough if they contain areas of greater human encroachment.

There was significant variation in total diurnal species abundance across the study sites. However, unexpectedly, there were no differences across species; thus, when individuals were rare in one or more site, they were common in others. Comparing each individual species' abundance across the study transects, only two species had substantially different numbers across sites. These were Varecia variegata and Eulemur fulvus. Both species are frugivorous and have the largest ranges among these lemur species. V. variegata is also considered highly sensitive to environmental degradation. This pattern was driven by both differences in abundance and both species' absence from some sites. Comparing segments within transects, we also found significant differences only in these two species: V. variegata was more abundant in the edge overall and at Mangevo, and E. fulvus was more common in the interior (but only at the Vohiparara site).

Using those variables that influenced forest structure - distance from edge and elevation - we examined what factors predicted primate abundance. Multiple regression models showed significant effects only for elevation - and only for V. variegata. This species declined with increasing altitude. Therefore, V. variegata was likely found in the edges mainly because these were typically lower in elevation. These results underscore the vulnerability of this endangered lemur species. It prefers lowland habitats, likely due to generally higher productivity in fruit resources. However, it is these areas that are far more likely to be affected by human encroachment, including complete conversion to uninhabitable agricultural landscapes.

Spatial Modeling Research Program (SMRP)
The spatial modeling team spent much of the year assembling data to use in analyses of deforestation and to support future modeling efforts. We undertook an exhaustive search of Landsat imagery available through various government archives and acquired all imagery of sufficient quality. We purchased six scenes covering RNP, adding to several we had acquired previously. We now have coverage over this area extending from 1989 through the middle of 2002, though the majority of those scenes are concentrated in the period after 1999. Unfortunately, Madagascar is located in a part of the world where clouds cover the earth's surface much of the year, thereby reducing the availability of useable imagery. Additionally, the malfunction of the Landsat 7 sensor that occurred in 2003 inhibits our ability to acquire more up-to-date scenes over RNP.

During the year we also acquired critical GIS files of the park boundaries and collected locational data on the 12 villages that are the focus of the SFRP. The combination of these data allowed us to start processing Landsat data by keying in on the area of focus: RNP and the surrounding landscape.

We have completed a preliminary classification of the 1989 and 2002 imagery with five cover classes: shadow (often stemming from clouds or the terrain), forest, water, non-forest, and cloud. Visual inspection of the forest demonstrates that forest margins have retreated during this 12-year period and fragments outside the main forested corridor have shrunk or disappeared in some cases.

Landsat imagery has a pixel resolution of 30m, making it highly suitable for investigating deforestation. We traveled to Madagascar for two weeks in June 2004 to examine the landscape first hand. One objective was to familiarize ourselves with the land cover and land uses so image classification schemes could be refined. Ultimately, we would like to be able to refine further the classification schemes so as to be able to discriminate among more types of land cover, such as different agricultural practices (e.g., tavy vs. paddy) and different states of forest disturbance. One important constituent of this refinement task is a digital elevation (terrain) model (DEM).

Before leaving Madagascar, we met with personnel from FTM and contracted with them to digitize terrain maps in their archives. This product, which we expect to take delivery of shortly, should represent 30m contours across the land surface, and may be sufficient to meet our DEM needs.

Return to Packard Project

Back to Top