I have been experimenting recently with some GIS freeware. My start point for GIS testing is always raster processing: raster operations are a common task in my research and teaching and consequently the first functions I evaluate are the raster tools. Recently I have been evaluating some of the most popular Open Source GIS packages and will compare them to Arc, the benchmark GIS (if for no other reason than it is the market leader). First let me set out a few test parameters:
-the GIS should be able to import and perfrom operations on common raster datasets,
-a range of tools for simple image processing operations and visualisation options should be available,
In addition of course there should be vector editing and analysis tools. These we will deal with at a later juncture.
So to the the my first observations. Quantum GIS (QGIS) at first seems easier to use than SAGA. The GUI feels more polished and more familiar. SAGA on the other hand feels more dated and the interface is less intuitive. Up to five tabs can be found in the Workspace, the info space at the bottom of the interface is congested: the Modules menu even more so! That said, if you use QGIS with third party plugins you will be faced with more icons than there are in the Hermitage in St Petersberg.
Opening a multispectral image should be easy: neither QGIS nor SAGA can open the USGS MTL file that accompanies Landsat-7 ETM+ data. Hence each band must be opened separately. In QGIS this means we must use Merge to create a multi-band file and false colour composite (FCC). The product of the Merge can be stretched by right clicking on the relevant layer after you have loaded it into the Layers menu.
In SAGA we use:
Modules>File>Grid>Import
to open the common rasters. Then we use:
Module>Grid>Visualisation>RGB Composite
to create an FCC. The window that opens allows us to select the Grid System (file), Red, Green, and Blue bands. The contrast enhancement can also be selected here: rescale to 0-255 is a linear stretch; alternatively you can chose user specified or a standard deviation based stretch, as in ArcGIS.
QGIS allows us to manipulate the visualisation of our FCC raster more easily and intuitive than SAGA. That is not to say SAGA is bad, just clumsy. The range of image processing tools in SAGA is far more comprehensive than those in QGIS. In QGIS I have tested the raster calculator which was easy to use: it is pretty much the only raster processing tool. In SAGA I tried several tools. I immediately encountered a memory exceedance error with a colour assignation process when using the whole Landsat-7 image. To avoid memory issues I subset the image using:
Modules>Shapes>Grid>Spatial Extent>Clip using polygon
This didn’t stop me getting an error when trying to produce a Fourier transform of a small single band subset….and with another Fourier algorithm, and with a classification routine….
So, SAGA disappoints. QGIS fails to excite and well GRASS wouldn’t even visualise a Landsat image without a fatal error (couldn’t find a library). ArcGIS in this case offers a much better experience, but given the price tag it should do. For image processing the ESA freeware packages BEAM and NEST, offer much better raster tools, but lack sophistication on the vector side. For me the next step is to try the vector modules of SAGA and QGIS (and possibly GRASS). I’ll probably have an other bash at some simple raster workflows too. Watch this space or comment if you want something specific tested.
geoNotes 