Landscape dynamics and its effects on chemical composition of water in central Rondônia

Author(s)

Silvio Frosini de Barros Ferraz

• Document type: Doctoral Thesis
• Press: Piracicaba.
• Institution: Universidade de São Paulo (USP). Escola Superior de Agricultura Luiz de Queiroz (ESALA/BC)
• Defense date: 2004-12-20
• Advisor: Carlos Alberto Vettorazzi

Abstract

Central Rondônia is one of the most deforested regions in the Brazilian Amazon and presents areas at different stages of fragmentation, creating a gradient from Primary Forest to developed areas. This paper deals with assessment of landscape changes between 1984 and 2002 in a watershed located in the central region of Rondônia State, Brazil, due to a systematic deforestation and pasture introduction since the 1970s. Landsat TM/ETM+ images (one every two years) were classified, resulting in a time series of landuse/ land-cover maps. Landscape changes were evaluated using cross tabulation between years, transition rates, landscape metrics related to size, density, edge, shape, connectivity, configuration, and deforested patches distribution related to patch size and spatial proximity to roads and old pastures. Transition probability functions were fitted to the time series to predict land-use changes for the next ten years for three different scenarios. Using the land use maps, landscape structure was analyzed for 20 catchments by landscape metrics and proposed land use dynamics indices. Catchments were grouped by Cluster Analysis using land use dynamics data and landscape data for 2002. Current dynamics can be maintained in the region for ten years, but present-day land use changes cannot be sustained for more than 15 years. A more sustainable scenario for the region includes ceasing the deforestation process, implementing the "Permanent Preservation Area" along rivers and controlling the dynamics at balanced levels of transition. In catchment's landscape dynamics, results show that landscape structure variability is correlated to Iand use for the three considered classes showing the importance on considering matrix and coexistent classes' dynamics in landscape dynamics studies. Land use dynamics presented four principal components, and three of them represented 85% of variation and were correlated to proposed indices, which ones were considered good indicators for the region. The addition of historical and current data to the environmental condition classification of catchments improved the methodology. Regarding landscape influence on chemical composition of water assessment, results show that physical and human factors predict better (p>0.05) the electrical conductivity (R2= 0.88), total nitrogen (R2= 0.48), and pH (R2= 0.42), while landscape structure for 2002 could explain better total nitrogen (R2= 0.68), dissolved oxygen (R2= 0.56), chloride (R2= 0.51) and phosphorus (R2= 0.46). The study revealed the importance of an integrated analysis of factors like historical land use, physical conditions, and landscape structure as indicators of water quality in the region. (AU)