Day 7-Readings

So yesterday I was able to catch up on some reading mostly on remote sensing projects in the Mediterranean region. Each article had a different focus and use different data sources, two of which were not sensors we had thought to use in this project. These are summaries for only two of them.

Estimating spectral separability of satellite derived parameters for burned areas mapping in the Calabria region by using SPOT-Vegetation data 2006

The study area for this article was Southern Italy and compared known burned areas to their image on remotely sensed data. They chose to do this because of the affect of wildfires in the Mediterranean region since a small fire can in fact have a big impact. Their overarching purpose was to see if remote sensing could help to evaluate the disturbance by fire of and testing fire models. They used 10-image composites from June till September 1998 received from the Vlaamse Instelling voor Technologisch Ondersock (VITO) Image Processing center which has free vegetation products. They then compared this to the Italian National Forestry Services record of fires for that time period. They eventually came up with different indices depending on the area within their study zone. They suggest that a better exploration would be to discrimination of areas depending on the land cover type, such as pasture v. forests, and that their processes could be applied using different sensors, such as MODIS-Terra.

An integrated spatial and spectral approach to the classification of Mediterranean land cover types: the SSC Method 2004

This explores a new way of classifying remotely sensed data by taking into account the idea that undefined pixels are most likely to be closely related to those nearby. It wants to use this principle to help define "open" types of land cover that do not have definite boundaries, such as shrub vegetation or vineyards. They used ENVI to do their project and relied on three main steps in their method.

1) Stratification: which was used to find "homogeneous" regions based on spatial and spectral data.
2) Classification of those homogeneous regions
3) Classification of the rest of the "heterogeneous" image

They used a lot of equations to defined exactly how "similar" mixed pixels were and whether or not they could be incorporated into a closer homogeneous area. Using these principles and equations they compared their remotely sensed classification to ground proofing classifications. They tested two regions, one most open land and the other mostly farmland. They found that this method classified open area 8% better than the regular method but that "closed" areas did not have an improved classification. In their acknowledgements they state that their methods are available on request which might be useful for the what we are doing. However, its in ENVI which I don't know how to use.