Last week a colleague sent me an article published in Global Change Biology for comment. I have just got round to looking at it in detail. The abstract is available here (with full text if you have the institutional access.)
The authors claim have identified a correlation at the global scale between forest loss and flooding.
I have a lot of issues with the way the data has been analysed that I intend to discuss here. However one of the positive features of this article is that all the data that used is included as an appendix. This allows anyone to make their own mind up regarding the merits of the analysis.
I will make specific comments in a later post. But I first deal with different technical issue. Having setup a postgis database for Mexico and Central America I realised that I lacked a global country map to help visualise data of this sort. I tried importing the shapefiles that come with Arcview, but with no success. They apparently have errors in the topography. I also tried converting the worldhires data in the R mapdata package, but that also had issues with bad topography (apparently American Samoa was to blame). Finally I found this reply to a R help post from the prolifically helpful Roger Bivand
writeOGR(world_countries, “PG:dbname=mydb”, “world_countries”, “PostgreSQL”)
Bingo. In a couple of minutes I had a rather low resolution world countries map with ISO3166 two letter country codes ready for linkage to any data with a column that uses this international standard. It is now nicely tucked away in my postgis data base ready for use whenever needed. Very satisfying. Notice that I include this code as an example of how R connects to a postgis database once one has been set up. It obviously won’t work if you haven’t gone through the steps to establish your local data base. Here is how on Ubuntu. It is not something I would try under Windows,even though posgresql 8.2 apparently has now been ported to Windows.
I then had to reformat the data from the pdf copy of the article. This was slightly frustrating as it needed a bit of manual work in an open office spreadsheet. I then added the ISO3166 codes. This line (also suggested by Roger Bivand) helped a lot. It would be even better if editors and reviewers insisted on the use of standards when they are available.
The nice thing about a postgis data base is that you can use it for storing all your data, spatial or otherwise, in one convenient place. So using RODBC connectivity the floods data can be incorporated by
The only slightly fiddly aspect is that the saved table lacks a primary key. That can be resolved from R with few lines that run sql in postgresql.
odbcQuery(con,”ALTER TABLE floods ADD COLUMN index int4;”)
odbcQuery(con,”UPDATE floods SET index=cast(rownames AS integer);”)
odbcQuery(con,”ALTER TABLE floods ADD CONSTRAINT floods_pkey PRIMARY KEY(\”index\”);”)
odbcQuery(con,”ALTER TABLE floods DROP COLUMN rownames;”)
Then the tables can linked together as a temporary view from within R
sql<-“create view floods2 as select f.*, w.wkb_geometry from floods f,world_countries w where f.ISO3166=w.ISO3166”
The view can be removed from the database whe it is finished with by
odbcQuery(con,”drop view floods2″)
With the view in place Qgis can be used for the visualisation instead of R, which is really helpful for casual analysis. This convenient interoperability, drawing on the particular strengths of a range of different tools within a common framework (statistical environment, database, GIS visualisation), is one of the great plus points of Open Source software from the point of view of a user. However you do have to discover how to do it first. You don’t get a “wizard” to help.
R does produce very good publication quality maps, but there is rather more overhead involved than when using the simple point and click graphical user interface of QGis I am now finding it much easier to send vector layers strait to postgis and work from there.
The view can also be exported as a shapefile strait from QGis. As the wordpress site only allows me to upload files it thinks are documents I have placed it here with a false doc extension. Don’t try to open it, save the file to disk and then change the extension to .zip. floods2thisisreallyazipfile.doc