Training data selection (SCP)

Defining land use/cover classes

Before starting to map land cover classes using Sentinel-2 satellite images we need clear definitions and a classification scheme. An example for a hierarchical land use and land cover (LUC) classification scheme is the European Urban Atlas. The scheme defines 5 meta classes where the class 1. Artificial surfaces has many sub classes as shown in figure A.

Figure A:LUC classification scheme of the European Urban Atlas

Figure B: LUC classes for Göttingen

The specification of other meta classes is not as detailed. If we are more interested in forest and open area classes we need to adapt and modify this scheme. On the lowest level not all classes defined in the European Urban Atlas do also appear in the surroundings of Göttingen (figure B). In addition we need to consider the phenolgical development of vegetation at specific acquisiton dates and to specify more classes which can possibly be identified in multispectral feature space of the satellite image. An example of an adapted simplified scheme is shown in the table below.

MC ID	MC info	C ID	C info	RGB code
1	Artificial surfaces	11	Built-up	230-77-0
2	Agricultural	21	Arable land	255-255-168
2	Agricultural	22	Bare soil	253-191-111
2	Agricultural	23	Pasture	230-230-77
3	Natural and semi-natural areas	31	Broad leaved tree cover	128-255-0
3	Natural and semi-natural areas	32	Coniferous tree cover	128-255-0
4	Wetlands	41	Reed	128-255-0
5	Water	51	Water	128-242-230
9	Miscellaneous	91	No data (cloud)	255-255-255
9	Miscellaneous	92	No data (shadow)	0-0-0
9	Miscellaneous	93	No data (missing imagery)	0-0-0

Creating systematic sampling grid

• Use a processing toolbox model downloaded from here urban_atlas_systematic_sample.model (Right click, Save as ..). Note, save the file with the extension *.model.
• Open the Processing --> Toolbox and Models --> Tools --> Add model from file. Load the previously downloaded model.
• The model should appear in the Models tab.
• Double click to open the model.
• Specify an original European Urban Vectors layer for Göttingen: geodata_lab01\vector\DE021L1_GOTTINGEN\Subset-Goe_DE021L1_GOTTINGEN_UA2012_UTM32N.shp or download data from other European cities.
• Project --> Project Properties --> CRS. Check that the Project spatial reference system is set to EPSG:32632.

• Click OK

Stratified random sampling

• Vector --> Research Tools --> Random selection within subsets
• The input layer is SCP_systematic_points
• ID Field containing the code for the strata is MC_ID
• Method is equal Number of selected features in each stratum.
• Number of selected features type 25.

The module selects a subset of the layer SCP_systematic_points. The selected points need to be saved as an ESRI shapefile. Right click on the layer SCP_systematic_points, Save as.... Make sure to check the box: 'Save selected only..as shown below.

Interpretation of selected points

Prepare main QGIS map viewer:

• Install the Openlayers and the Go2nextFeature plugin (if not yet done) Plugins --> Manage and Install Plugins --> Install.
• Load the multiband Sentinel-2 satellite image as surface reflectance (Converting DN to reflectance how see Land Cover/Use Classification using the Semi-Automatic Classification Plugin for QGIS).
• Load the vector file with stratified sample points stratified_sample_25.shp.

• Load Google maps as background WMS layer. Web --> OpenLayers Plugin --> Google Maps --> Google Satellite.
• In Layers panel drag the GCP vector file on top of Google Satellite.