Readme


Analysis of Evolution of Roads and Timber Trails in Study Area

During the course of the recent months, it was determined by axis 1 members of the LUA-IAM project that to understand how institutions  and territorial units are emerging, and impacts on land-cover, it is important to understand how and where the largely illegal timber extraction activities took place.  In some cases on the Amazonian frontier the opening of trails into the forest for timber exploration led farmer and rancher expansion into the frontier, in other cases they led the creation of settlements, and in others the followed the creation of settlements into previously inaccessible lands for extraction within and aground the newly created settlement.  At times these were conducted in collaboration with the colonist farmers, and other times the forest of the smallholders or colonists were heavily harvested without collaboration or compensation.   To begin an analysis of the relation of creation of territorial units, timber extraction and land-cover change it was determined a map of the evolution of roads and timber trails was needed.

Timber trails are notoriously famous for the inability to detect them in remotely sensed products.  This is because they are often narrow and don~t open the canopy enough to expose the bare soil below, so moderate resolution images as the Landsat (30meter) images in our time-series are difficult, and even more so in the PRODES 90 meter products.  Yet, to conduct an analysis of the evolution of roads we created maps of roads and trails in each date in our time-series of Landsat images from 1975 to 2010.

To create roads based on visual interpretation of the Landsat images we utilized a shapefile of roads and trails in the state of Pará created by Imazon, where they used here resolution (20 meter) SPOT images and some field visits.  We overlaid this shapefile over each date in our images, and digitized those areas where there were visible roads or patterns indicating a trail.

To begin, we first complete the radiometric calibration of all images in our database.  This processing step is necessary to classify images, as it converts at sensor irradiance to at surface reflectance values.  It also sharpens the image making detection of roads or evidence of them easier.  We also displayed the image in different band combination other  that enhance forest/bare soil areas such as RGB:453 and RGB:666.  In some cases image enhancement methodologies were attempted such as Tassled Cap transformation and edge enhancement filters. 

My method for choosing to digitize a road were:

1)	Digitize a road whenever  a road is visible in both the image and in the overlaid Imazon road shapefile.
2)	Digitize a road if a road was obviously visible in an image, even if doesn’t exist in the Imazon road shapefile.  Assume the canopy grew over the trail between the time the image was collected and 2010 when the Imazon product was created.
3)	Digitize a road when a road exists in the Imazon road shapefile between a river or road and a disturbed area requiring access of a trail, even though none is visible.  Particurly digitize where geometric patterns exist along the road in the Imazon road shapefile, giving further evidence a trail exists but the canopy is covering the bare soil below making unobservable.
4)	Do not digitize a road that exists in the Imazon road shapefile if a) not observable in any image in the time-series or no geometric patterns giving evidence of a road.
5)	Do not digitize a road to connect the end of a visible road and a geometric pattern (small field or pasture) beyond it.  There could be a trail, or there could be only a foot path of a small colonist opening new land