Applying RegMAS to a new Region

While RegMAS ships with a default region that can be used for simulation purposes, most likely users want to run simulations over their own case-study regions.

In order to be modelled with RegMAS a region must be available with the following set of data:

Quantitative regional data

Aggregated data of the region, normally available from the Census.

Individual farmer detailed data

Individual farmers are used in the model as ''bricks`` to build a simulation region and the crucial information here become the individual farms production factors (capitals, humans, land). In order to obtain satisfactory congruence between the real and the simulated region, a basket in the magnitude of tens of farmers data is often necessary¹¹.

Technical coefficients and prices

Technical coefficients are needed to link the activities pool with the resource pool, while prices are needed for both products and factors

Land use map

As RegMAS is fully spatial explicit, it requires a detailed map of land uses (In Europe this is available from the Corine Land Cover project¹²).

The basic idea is to reproduce in the model a simulation region composed only of ``typical'' farms, but of which aggregate values are as close as possible to the real one.

To each farm in the region of which detailed data is available (e.g. because the farm is member of the FADN network) is associated a coefficient. A 0-coefficient means that the farm is not selected, while a non-0 coefficient implies that the farm becomes one of the typical farms of our simulation region. The key point is to find these scaling coefficients that minimise the difference between the simulation region and the real one (Eq. 4).

$$\min \sum_{k=1}^{K}{( \frac{\sum_{n=1}^{N}{(FADN_{n,k}*UC_{n} )}}... ... \hspace*{0.3cm} sub \hspace*{0.3cm} UC_{n} \geqslant 0 \hspace*{5mm} \forall n$$ (4)

Where:

Indeces:	Variables:
$n=\{1...N\}$ Individual farms	$FADN_{n,k}$ FADN data
$k=\{1...K\}$ Characteristics	$REGIO_{k}$ Regional aggregated data
	$UC_{n}$ ``upscaling'' coefficient

This procedure is called "upscaling" and it is well documented in Kellermann et al. (2007), while a practical implementation is discussed in Sahrbacher et al. (2005)¹³.

The upscaling can be conveniently obtained using the quadratic solver in Excel, like shown on Fig 6 ¹⁴

Figure 6: Regional upscaling using Excel