This research project plans to integrate data from new radar platforms (ALOS/PALSAR and RADARSAT2) with those from existing optical sensors (TM/ETM+, ASTER) to enable land cover classification and assessment of land cover dynamics in areas of constant cloud cover. The result will be timely and accurate assessment of land cover change in areas that typically experience constant cloud cover and therefore make traditional methods of remote sensing impossible.
Professors Moran and Lu from Indiana University, along with international collaborators at the Instituto Nacional de Pesquisas Espaciais (INPE) contend that the multiple polarizations of ALOS/PALSAR and RADARSAT2, fused with analysis of optical data, will provide better and timelier vegetation classification accuracy, and will improve the temporal and spatial analysis of land use and land cover in areas characterized by persistent cloud cover. Two study areas will be the focus of this study, Altamira and Santarem, both in the Brazilian Amazon state of Para. This will enable the analysis of data collected for several years by this research team using a nested-georeferenced approach including soil analyses, vegetation stand structure and composition, land use histories, institutional analyses, demography and decision-making of hundreds of households, and land cover classification using multi-temporal remote sensing data since the 1970s (MSS, TM/ETM+, and IKONOS/QuickBird,) thereby reducing costs and to achieve results much more quickly.
The PIs have obtained recent images of RADARSAT2 and ALOS/PALSAR for both study areas, as well as IKONOS and Quickbird images. These data will be fused with archived data for the study areas from Landsat MSS, TM and ETM+ (and ASTER) to advance the state of knowledge and methods in land cover assessment applicable across any part of the world where cloud-cover constrains seasonal observations of changes in land cover. Separatibility analysis will be used to examine the capability of the data fused images, which are developed with different data fusion techniques, in distinguishing the land cover classes of interest based on the training sample plots from extensive field work and the intensive use of the IKONOS and Quickbird very high resolution data. The results of this study will solve a major obstacle currently faced by scientists in quantifying land cover change in areas with persistent cloud cover. Much of the humid tropical world is covered by clouds for most of the year, and it can be sometimes years between dates when it is possible to obtain cloud-free images for monitoring changing forest conditions. This prevents timely responses to deforestation events and understanding of land cover changes over time. This project addresses an urgent need for timely analysis of land cover changes in the Brazilian Amazon, and in humid tropical regions more broadly.
