Cornell University Website Template

Cornell University Unit Name 1 Unit Name 2

About IRIS	Geographic Information Systems	Resource Inventory	Remote Sensing	Staff/Faculty

Graduate Student Research

Participatory research data collection methods for accuracy assessment of land-cover maps

This project is part of John Sydenstricker-Neto’s (Dept. Development Sociology) dissertation research entitled Land Cover Changes and Social Organization in Brazilian Amazonia: A Temporal Spatial Analysis, 1986-1995/99. The primary goal of the dissertation is to unfold how historically grounded local social relations and specific conditions of natural resource systems have jointly shaped the ways in which settlers use their agricultural parcels and common-forest reserves. The project links socio-demographic information, Remote Sensing/GIS analysis, and qualitative data on local social organizations and institutions.

The objectives of Remote Sensing/GIS analysis were to: (1) determine land cover change in the recent colonization area for small-scale migrant farmers (1986-2000) in the municipalities of Machadinho D’Oeste and Vale do Anari, State of Rondonia, Brazil; (2) engage community stakeholders in the processes of mapping and assessing the accuracy of land cover (LC) maps; and (3) evaluate the relevance of LC maps (inventory) for understanding community-based land use dynamics in the study area. We were interested in learning if there would be increased efficiencies, quality, and ownership of the inventory and evaluation process by constructively engaging stakeholders in local communities and farmer associations.

Landsat data were used to create maps of LC conditions for 1986, 1994, and 1999. Images were obtained in July/August (dry season) and field data was collected during August 2000 with the assistance of nine local farm associations and approximately 100 independent farmers. At meetings with the associations hardcopy false color composites of imagery data with parcel boundaries were presented to individual landowners. Each individual provided historical and contemporary LU for known areas. Polygons were annotated and labeled on stable acetate for each cover type, corresponding to the seven-category classification scheme. Notes were taken during the interview process to indicate the dates of land clearing, cover type, and level of uncertainty expressed by the participants.

Approximately 1,000 polygons were field annotated and random samples were selected for classifier training and map validation. Overall accuracy for each year ranged between 85-95% (Kappa 0.52-0.78). LC changes were consistent with the trends observed in the study area and reported by others. The participatory process involving local farmers was crucial for achieving the objectives of the study. The specific protocol developed for data collection should be applicable in a wide range of cases and contexts.

The building of trust with the local stakeholders is important with contested issues such as deforestation in the topics. Systematic data collection among farmers (the primary land users) provided a valuable source of information based on their direct observation in the field and historical data not directly available through other sources. This procedure provided greater confidence for interpreting and understanding classification errors. Finally, the process itself empowered local farmers and provided a forum for discussing land use processes in the region, including challenges to alleviate poverty, increase agrosilvopastoral farming systems, arrest deforestation, and study its implications for developing more effective land use policies.

Including the local stakeholders in the research was very effective process for evaluating LC change in the region. For stakeholders and researchers, the mapping and reporting process fosters better understanding of the patterns and processes of environmental change in the study area. We foresee that participatory mapping projects such as the one used in this study have the potential to become an important planning device for regional-scale development in Brazil. With greater economic opportunities and stronger institutions at the local level, society is likely to improve the ability to identify and adopt more environmentally sound LU activities. For more information, contact John Sydenstricker-Neto at jms56@cornell.edu.

Resulting Publications:
Sydenstricker-Neto, J., A.W. Parmenter, A.W., and S. D. DeGloria (2003). Participatory rResearch data collection methods for accuracy assessment of land-cover maps. Luneta, R.S. and J.G.Lyon (Ed.) Remote Sensing and GIS Accuracy Assessment. CRC Press, Boca Raton, FL.