swisslandstats-geopy: Python tools for the land statistics datasets from the Swiss Federal Statistical Office

The Swiss Land Statistics inventory by the Swiss Federal Statistical Office (SFSO) (Swiss Federal Statistical Office, 2017) provides land use/land cover (LULC) datasets at the Swiss national extent for a sequence of four survey periods in 1979/85, 1992/97, 2004/09 and 2013/18. The data is stored in a relational database format, where each row corresponds to one of the hectometric pixels that configure the Swiss territory, and features three groups of columns:

• Firstly, the E and N columns denote the coordinates of the pixel's centroid in the LV95 coordinate reference system (or alternatively X and Y in LV03).• Secondly, the FJ85, FJ97, FJ09 and FJ18 columns denote the exact years when the observations for each of the four survey periods were taken.For instance, the first dataset was produced between 1979 and 1985.Accordingly, for each row/pixel, the FJ85 column will denote the exact year where its LULC category attribution was made (it can be any year within the 1979/85 period, depending on the part of Switzerland).• Thirdly, the LULC data is provided in three different nomenclatures: the standard nomenclature, which feature 72 categories that combine land use and land cover information; the land cover nomenclature and the land use nomenclature.Accordingly, the LULC information for each pixel is stored in columns of the form LC85_27, where LC denotes the land cover nomenclature, 85 the survey period 1979/85 and 27 the number of categories considered.
The inventory for each nomenclature can be downloaded as a comma-separated value (CSV) file.For instance, the standard nomenclature aggregated to 17 categories can be download freely, and is of the form: • Read CSV files from the SFSO into LandDataFrame objects, which extend the conventional pandas DataFrame with additional attributes that store the coordinate reference system (CRS) and pixel resolution.
On the other hand, the settings module of swisslandstats-geopy allows changing the CRS and pixel resolution so that the library might also be used with similarly  (Ballin, Barcaroli, Masselli, & Scarnǿ, 2018) could be instantiated as LandDataFrames, the data could not be converted to NumPy arrays, likely because of the CRS and how their grid is sampled.Further exploration of the characteristics that a table-like raster dataset must fulfill in order to be seamlessly procesed within swisslandstats-geopy could signficantly enhance the reusability of the library.

Figure 1 :
Figure 1: Plot of a categorical LULC column as a raster image.
table-like raster datasets.For instance, a dataset of phenology in the Eastern US and Canadian forests for the 1984-2013 period (Melaas, Friedl, & Sulla-Menashe, 2018) has been processed seamlessly into a LandDataFrame instance.Nevertheless, although other table-like raster datasets such as the European land use/cover area frame statistical survey (LUCAS)