Deforestation and floods 4: PostGIS raster layers?

One element struck me when looking at the original data compiled by Dartmouth college on flood  events. The data base contained a table headed “cause”. This was full of text such as “rain”, “heavy rain”, “torrential rain”. The orange points on the map above are the centres of recorded flood events in thals three years. Note VillaHermosa in Tabasco. The coloured layer is shaded by rainfall intensity (five bands ranging from below 600mm to 2500 mm +).

So, the headline would be that “scientists conclude that flooding is caused by ….. rain!”

This is not in fact as trivial as it seems. Devastating floods can occur in regions where heavy rain is uncommon, but there must be a clear correlation between flood frequency and the sheer amount of annual rainfall a region experiences. I am constantly surprised by how quickly the vegetation in Mexico changes from dry forest to exuberant tropical forest. The transition zone can be a matter of kilometers on the road north from Chiapas to Tabasco. The flooding in Villa Hermosa in 2007 was mainly the result of rainfall falling on the state of Tabasco itself (and a small part of the extreme north of Chiapas). During the dramatic events a very light drizzle was falling in San Cristóbal. The lowlands of Tabasco are almost completely deforested, but the mountains of northern Chiapas are covered in a mixture of secondary forest and cattle pastures. Rains in late October fall on already saturated soil profiles. Vegetation cover could only have played a marginal role in reducing the amount of water flowing out of the highland watersheds in this particular case.

This provided me with an excuse to experiment with ways to move raster layers into POSTGIS. The POSTGIS data base structure is clearly not designed for raster layers, but some quite good results can be obtained simply by moving a fairly coarse raster layer, such as an interpolated rainfall coverage, into POSTGIS as points. This is similar to a SpatialPixelsDataFrame in R. In fact the easy way to set up the coverage is to move the data from a sp object in R to POSTGIS by converting the coordinates into a geometry. When viewed at some scales an interference pattern results, but the coverage can be very useful because clicking the information tool on any point on the map results in information on the pattern of rainfall over twelve months, although the visualzation can use the total annual rainfall.

The first image used Udig. The lower image uses Qgis. Again although I have not put a clear legend in this very quick informal treatment, the rainfall ranges from 800mm per year (bright red)to 4000 mm (dark blue)

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s