New high resolution global forest change map 1

This week saw the publication of what may turn out to be one of the most influential products derived from processing Landsat imagery. Mathew Hansen of Maryland University and colleagues, including researchers working for Google, have pieced together a fascinating map of global deforestation over the last decade using the 28.5m pixel resolution imagery.
http://www.sciencemag.org/content/342/6160/850
As far as I am aware this is the first time that a consistent methodology has been applied across the complete set of global imagery. It is certainly the first time the results have been made public through such a nice portal.
http://earthenginepartners.appspot.com/science-2013-global-forest
The simplicity of the Google app and the speed at which the imagery is delivered is impressive. This will lead to much greater public awareness of deforestation. I personally hope that the conservation community will focus not only on the still depressing picture in the notorious hotspots of deforestation, but also find some more positive messages. Rates of deforestation in many areas have slowed. Southern Mexico is one of these, as the article based on our own classification of Landsat data from the 1990 to 2004 points out.
http://www.plosone.org/article/info:doi/10.1371/journal.pone.0042309#s5

Patterns of deforestation in Central America

The most noticeable regional pattern, that becomes immediately apparent as you zoom in on Central America and Southern Mexico, is the extensive deforestation in the southern Peten of Guatemala and the Atlantico Sur region of eastern Nicaragua. In the case of Guatemala the deforestation mainly stops at the border with Mexico, with the exception of the comparably rapid deforestation in the Marques de Comillas region. At this scale there does not appear to be much deforestation in Mexico at all. I will discuss patterns in Mexico in more detail in a later post.

In both these hotspots deforestation is following the traditional model for the region. Settlers first clear the forest in order to plant subsistence crops, usually maize, in relatively small clearings of around a hectare in size. In the process they tend to degrade the surrounding forest as a result of logging and uncontrolled fire that extends into the forest when fields are burnt. Fires was particularly important in the Peten in 1998, which was an exceptionally dry year in the region. There were further intense fires in 1999 that burned fuel that had accumulated as a result of the tree deaths in 1998.

modis

Fires in the Peten in 1998 detected by Modis imagery.
(http://www.nasa.gov/vision/earth/environment/16may_biocorridors.html)

An issue with the  recent classification that I noticed is that the percentage forest cover maps for the baseline year 2000 tended to show more forest than our classification. Areas that our analysis show to have been already heavily degraded prior to 2000 are classed as mainly undisturbed. For example, we classified a lot of the secondary growth that comprises a shifting mosaic due to traditional long fallow milpa farming in  the Yucatan as non forest, while Hansen’s map shows these areas as forest undergoing frequent disturbance (see footnote).

So,  Hansen’s map may show some loss in the last decade that could have in fact occurred previously. Nevertheless, there will always be issues of interpretation for areas of secondary regrowth. Comparing the results with our analysis for Marques de Comillas ( http://files.figshare.com/313034/File_S1.zip) where the deforestation represents conversion of forest to pasture show that the results are comparable. Our layer shows deforestation prior to 2000 as red pixels and 2000 to 2006 as pink pixels. The new classification shows recent deforestation in red. The areas we found to be deforested during the 90s are typically black, visually demonstrating the consistency between the two layers. Click on the image below to activate an animation between the two layers.

Deforestation in the Marques de Comillas region of Southern Mexico

Marquez

Deforestation in Marques de Comillas. Our classification shows deforestation between 2000 and 2006 as pink pixels. These coincide with the areas detected by Hansen et al.

Patterns in the Petén

defor2

Deforestation in the Southern Peten

defor3

Deforestation in the Southern Peten. Note the occurrence of some regular geometric shapes. This suggests consolidation into formally recognised land use units.

In both the Peten and Nicaragua the recent deforestation is a continuation of a process that begun several decades ago. In both regions the land is not well suited to permanent production of maize on a commercial basis. However cattle farming is a profitable, if rather inefficient use of land. In the Peten the comparatively wealthy cattle ranchers traditionally used violence and coercion to drive the initial settlers from the land that they had originally cleared. Governance in Guatemala is still an issue, so I assume that this practice continues. The presence of geometrical shapes on the landscape suggests some regularisation of the process that may be the result of formally recognised land tenure. Regularisation of land use tends to take place several decades after the initiation of deforestation.

So, some of the processes taking place on the ground can be inferred from the shapes making up the landscape pattern. It would be a very interesting challenge to attempt to use automated pattern recognition and object identification in this context.
In contrast to the Peten, regular shapes are still much less noticeable in the Atlantic region of Nicaragua. This suggests that the land is still being cleared by the original settlers. Population density remains low in this area of Nicaragua, which can be regarded as a genuine “agricultural frontier”.

defor11

Deforestation in the Atlantic region of Nicaragua. Note the irregular pattern that suggests ongoing clearance using fires and machetes.

The sustainability of land use following deforestation, particularly in Nicaragua, is questionable. Much of the converted land there appears to be undergoing erosion.

The soils and topography of the Southern Peten are comparable to the lowlands of Chiapas, Tabasco and Southern Veracruz. Large areas of forest in Southern Mexico were converted to pasture decades, even centuries, ago. Much of this land has remained reasonably productive, providing that it is carefully managed.

——————————————————-

Comparison of the pattern in the Yucatan. Our classification represented areas of secondary growth as “non forest” in 2000, while Hansen’s classification shows a rapid dynamic with many pixels changing state. Maize is produced in this area using tradional slash and burn with a rotation period of around 12 years. The fields quickly regain cover through resprouting from permanent rootstocks.

Hansen’s map (left) shows blue, pink and red pixels representing change in areas that our map show as deforested (right). It would be interesting to look at this in more detail by overlaying the layers formally.

yucatan2

Hansen et al’s classification of a part of the Yucatan. Note the presence of pink pixels that have undergone both deforestation and regrowth in the 12 year period.

yucatan3