diff --git a/data/fields_jena.tif b/data/fields_jena.tif
index 839afe968b6e110aab2f8ea9dd44274c01ace912..1d5e88a54c96f855db61aa135e1f6c62dd486f26 100644
Binary files a/data/fields_jena.tif and b/data/fields_jena.tif differ
diff --git a/data/fields_thueringer_becken.tif b/data/fields_thueringer_becken.tif
new file mode 100644
index 0000000000000000000000000000000000000000..13215a12706d86af059dd631404182e54d94ad18
Binary files /dev/null and b/data/fields_thueringer_becken.tif differ
diff --git a/docs/gis.md b/docs/gis.md
new file mode 100644
index 0000000000000000000000000000000000000000..52ef3a804ecaa146021ae5f5a35f95d68f62c053
--- /dev/null
+++ b/docs/gis.md
@@ -0,0 +1,72 @@
+# GIS data in Persephone
+
+Persephone currently requires two separate map input files: one for land cover,
+the other for field geometry. This documents describe how to obtain and process
+the data needed for each of these.
+
+## Land cover maps
+
+Land cover maps for Germany at 10m resolution can be obtained from 
+[Mundialis](https://data.mundialis.de/geonetwork/srv/eng/catalog.search#/metadata/9246503f-6adf-460b-a31e-73a649182d07).
+These are generated annually from Sentinel data and comprise the following 
+land cover classes:
+
+```
+10: forest
+20: low vegetation
+30: water
+40: built-up
+50: bare soil
+60: agriculture
+```
+
+To create a Persephone map input file, you need to crop the national Mundialis
+map to the extent that you want to simulate (suggestion: approx. 10x10km is a 
+reasonable size).
+
+To do so, download the Mundialis map and import it into QGIS. Then create a new
+vector layer and create a rectangle feature to delimit the extent of your
+region. Then go to `Raster -> Extraction -> Clip Raster by Extent`. Select
+the Mundialis map as the input layer, set the clipping extent by choosing your
+region vector layer under `Calculate from Layer` and specify the output
+file name before clicking `Run`. This will generate a TIF file that you can
+pass to Persephone as the `landcovermap` parameter.
+
+## Field ID maps
+
+In addition to the land cover data explained above, Persephone also needs
+information about agricultural field boundaries in order to assign these to the
+farming agents. Unfortunately, getting this is rather more complicated.
+
+In the EU, every country runs a Land Parcel Information System (LPIS) to
+administer CAP payments. In Germany, this is called InVeKoS and is run by the
+Länder. For example, you can view and download the InVeKoS data for Thüringen
+[here](https://thueringenviewer.thueringen.de/thviewer/invekos.html).
+This gives you a vector layer which can be loaded into QGIS. However, it needs
+to be converted to a raster layer and cropped to your region extent before it
+can be used in Persephone.
+
+The first thing to do is to make sure that the vector layer has a numeric (!)
+field with a unique identifier for each field block (check the attribute table). 
+The Thüringen data has the FBI ("Feldblockident") field, but this is a string
+value and therefore not usable by the rasteriser. So, we set the vector layer to
+edit mode, open the field calculator, enter the information for a new field 
+(call it "FID" and set it to a 32-bit integer), and enter `@row_number` in the 
+expression field. Then save the layer and close the calculator.
+
+Secondly, you need to filter out all non-field/non-grassland plot types. (LPIS
+also has data on forests and various landscape elements that are not relevant to
+our use case.) Assuming you're working with the Thüringen InVeKoS data (other
+data sets may have a different structure), right-click on the layer name in
+QGIS' layer overview and click on "Filter...". Then, enter this expression in
+the query builder: `"BNK" = 'AL' OR "BNK" = 'GL'` and click "OK". This
+will select only field and grassland plots.
+
+Next, open the rasteriser (`Raster -> Conversion -> Rasterize`). Select your
+FID field as the "Field to use for a burn-in value", and your land cover map (as
+created above - this ensures the two layers match) as the output extent. Make sure 
+the "fixed value to burn" is "Not set". Then choose "Georeferenced units" as the 
+"Out raster size units" and set horizontal and vertical resolution to 10.0. In 
+the advanced parameters, set the output data type to `UInt32`. Finally, enter an 
+output file name and run.
+