APPENDIX 1. Field Sites and Research Methods.

The research on which much of the data presented in this paper are based was carried out in three field research sites (see Fig. 1). Two are located in the estuary of the Amazon River: one on Marajó Island, focused on the region around the town of Ponta de Pedras in the Brazilian State of Pará, the second on the floodplains around the city of Macapá in the State of Amapá. The third is on the Ucayali and Aguaytia Rivers in the lowland Peruvian Amazon, focusing on rural sites near the town of Contamana along the Ucayali and the village of Nuevo Piura on the Aguaytia. Urban data were gathered in the following cities: Pucallpa and Yarinacocha in Peru, and Macapá, Santana, Belém and Ponta de Pedras in Brazil.

Marajó. The Marajó Island research site has been studied since 1989. Data reported on local patterns of resource management are based on systematic ethnographic surveys that began in 1990. Data on migration and land use histories are based on structured and semi-structured interviews and questionnaires applied to 129 sampled households. A strong emphasis was also placed on participant observation during this and later phases of fieldwork. All data collected were spatially referenced.

The changes reported in land use and forest composition in the Ponta de Pedras site are based on data gathered using vegetation inventories, land use histories, and experiments to measure fruit productivity under different intensities of forest management, as well as through analysis of remotely sensed images. Selection of areas for inventory was based two main factors: (1) years since management started and (2) management steps applied to the stand. Four adjacent plots (25 X 25 meters), and four randomly distributed subplots (5 X 2 meters and in some cases 10 X 2 meters) were distributed within the area. The total sampling area (2500 m2) was distributed according to spatial characteristics of açaí stands and the floodplain environment. However, in areas of unmanaged floodplain forest, the sampled area was doubled (5000 m2) to increase the representation of floristic and structural measurements. In these areas plots were sorted randomly into different sides of a 200-meter transect, and subplots were randomly distributed within each plot. In each plot, all tree species with dbh 10 cm were identified at the species level and measured for dbh, stem height (measured at the first major branch) and total height (measured at the top of the canopy). In the plots, all açaí stems with dbh ≥5 cm were measured for dbh and total height, and number of stems per clump assessed. In each subplot, all individuals, including grasses, were identified and counted, and individuals with dbh 2 cm were measured for dbh and total height. The identification of plant species was done in the field (for commonly known species) and samples were collected and later identified at EMBRAPA’s herbarium in Belém.

All sampled areas were geo-referenced and contributed to a multitemporal analysis of land cover change using aerial photography and satellite data. Land cover mapping and change detection was carried out for 1969, 1985, 1988, 1991, 1995, 2001. Intensively managed açaí agroforestry areas were mapped using Landsat TM, ETM, and IKONOS data coupled with intensive fieldwork and vegetation sampling. Transition matrices were used to estimate rates and direction of change between dates, including for areas of intensively managed açaí agroforestry. See Brondizio (2006, 2008) for more detailed descriptions of multi-temporal assessment of land cover change using remote sensing data and classification accuracy.

Experiments to measure açaí fruit productivity were carried out during the whole production season of 1994 and 1995. Experimental plots were located based on the producer’s indication of the site (with its respective history) and on the analysis of area boundaries and characteristics. At each site, a 25 X 25 meter plot was marked. A subplot of 10 X 10 meters was set up inside the plot and marked in the same way. Subplot location was based on a stratified random selection.

Amapá. In the Amapá floodplain data collection began in 1994. Changes in land use were studied in seven communities in the Mazagão watershed southwest of Macapá and in five communities in Ipixuna, northeast of the city. Demographic, economic, and land use surveys were carried out using semi-structured household interviews with 140 families in 1994, 1999, and 2006. Market surveys were conducted at various times over the last decade.

Permanent plots of one hectare were set up in twelve landholdings in each of the two floodplain areas, Mazagão and Ipixuna in 1999. Permanent plots of one hectare were also established in 12 unmanaged forests (6 in each site) surrounding the 24 selected landholdings. Floristic inventories of all trees ≥5cm dbh in 100% of the total area of the permanent plots were carried out in the 24 selected landholdings and 12 surrounding unmanaged forests. Floristic inventories continue to be carried out every three years in all the permanent plots, measuring growth to estimate mortality and recruitment, and collecting data on productivity of extracted products.

For estimates of changes in land cover we have been using greyscale areal photographs from November 1976, Landsat TM images from July 1986, August 1992 and October 2006. Land cover mapping and change detection was carried out for 1976, 1986, 1988 and 1992 and published in a master thesis (Pereira 1998).

Peru. Research on the specific areas and issues relevant to this article began in 1999. The data cited in Tables 2 and 3 were gathered through interviews with 47 farm households with rural landholdings around the town of Contamana and from data in cadastres located in the archives found in Contamana. Floristic inventories were also conducted in 20 of those landholdings in Contamana as well as nine landholdings in Nuevo Piura along the Aguaytia River. In each of the 29 selected landholdings one hectare of forest was sampled and all trees ≥5cm dbh were measured.

Urban data have been collected in the city of Pucallpa and neighboring Yarinacocha since 2005. Semi-structured interviews were carried out with residents of about 500 households in 42 of the city’s asentamientos humanos. Interviews focused on the migration history, employment history, and use of a variety of rural products including construction materials in housing. In Peru we have used Landsat TM images from July 1992 and August 2005 to build land cover maps to detect the impact of land-use changes, particularly timber extraction.