Methodology for mapping the distribution of small farms in Europe

For the purpose of the project, small farms were defined for statistical purposes as those with less than 5 hectares or less than 8 Economic Size Units (ESUs).

The project used also a participatory approach that ended up with a more complete, not statistical, definition depending on the position of the farm in the food chain/system. The approach was used to select small farms to be included in the SALSA survey, and it was flexible to consider what in the context of each region and in relation to its farm structures, was relevant to be classified as a small farm, and was accepted as such. 

Data collection at the NUTS 3 level or all countries except for Germany (NUTS 2 level due to data availability) combines assorted datasets from EUROSTAT and national statistic offices, as well as, data obtained from key experts in different countries, which was used to calculate another set of indicators.

The threshold defined in SALSA’s conceptual framework to classify small farms was based on two criteria:

• physical size (farms with less than 5 ha of Utilised Agricultural Area, UAA),
• economic size (farms with less than 8 Economic Size Units, ESU1) of Standard Gross Margin (SGM).

While UAA farm size allows one to assess the distribution of the physical size of the farms, the SGM provides a measure of a holding's business size and is representative of the level of profit that could be expected on the average farm under 'standard' conditions (e.g. with the average level of rainfall expected in a region).

The NUTS 3 level was selected as it is the most detailed administrative level for which the data that are required could be found for all European countries.

 Most of the data comes from the 2010 agricultural censuses and EUROSTAT: Farm Structure Survey (2007-2010) although there are a few exceptions for which the data are less up-to-date (Croatia and Austria).

The data was used to calculate a large set of structural and economic indicators. To measure the physical size, indicators included the percentage of all holdings below the thresholds (5 ha of UAA for the physical size; 8 ESU for the economic size). Farm density was computed based on the ratio between the total number of holdings and the UAA in each region. Small farm density was based on the ratio between the number of small farms with less than 5 ha and the UAA covered by these small farms (the higher the density of the farms, the higher the prevalence of small farms). The mean size of total farms and the mean size of small farms was also determined. The indicators were used to assess each indicator's relative distribution of small farms in each region through Standard Scores (also known as z-values, z-scores or normal scores).

To measure the economic size, indicators included Labour Force and Family Labour Force, as they can be used as a proxy for the importance of small farms in a region. Labour Force is expressed in persons and in Annual Work Units (AWU) where one AWU corresponds to the work performed by one person occupied on a full-time basis. The Family Labour Force of the agricultural holding refers to persons who carry out farm work on the holding and are classified either as a holder or the members of the holder's family.

To identify reference regions, statistical analysis using, first, cluster analysis and then correlation analysis was performed.

Cluster analysis was implemented in two steps:

• The first clustering approach using k-means, considered different combinations between a number of pre-indentified, non-correlated variables (indicators). As a result, more than 30 maps indicating the distribution of small farms (considering both physical farm size as well as economic dimensions) within Europe were produced.
• Given the large number of variables and emerging clusters as well as in order to improve the comprehensibility of results, a second cluster analysis using k-means was performed considering only five selected indicators:

• the share of UAA in each region;
• the share of UAA covered by small farms (defined as farms with less than 5 ha);
• the percentage of the region occupied by small farms, defined as farms with less than 8 ESU of SGM;
• the percentage of the number of holdings with less than 8 ESU of SGM;
• the density of small farms (number/ha) defined as farms with less than 5 ha.

Correlation analysis was used to reduce the number of variables employed throughout and identify reference regions. To this end, the cluster analysis was combined with the EDORA (European Development Opportunities for Rural Areas) structural types, in order to select 25 reference regions. The combination of both typologies allowed to highlight different aspects of differentiation in Europe at the regional level. In this way the project ensured that the selected reference regions cover higher levels of diversity than if only one typology was used.

3-5 regions were selected by project partners and experts as reference regions to illustrate the diversity of small farm situations across Europe. This approach facilitated the:

• Differentiation of the Scottish uplands (croft regions) from the lowlands or the sheep farming systems of northwestern Ireland, which follow a pluriactivity tradition;
• Identification of the three main farm structures in Poland;
• Separation of the plots subjected to intense afforestation in Portugal, but where the small agricultural plots remain;
• Differentiation of mountainous areas or Apennine areas in Italy;
• Differentiation of the southern and Alpine areas from the rest of the regions in France.

The estimated distribution of small farms in each reference region involved the implementation of two main stages using Sentinel-2A images:

i) build an agriculture and non-agriculture mask to exclude for the subsequent analysis all the non-agricultural lands existing in each reference region (land with agriculture-like land cover),

ii) estimation of a surface map presenting the probability of small farms presence in a square grid with 250 x 250 m size.

To produce an estimate about the extent of small farms over each reference region, a probabilistic model is developed. This is probably the first remote sensing-based small farm distribution map developed by using Sentinel-2A imagery. The results highlight the fact that not all the agricultural areas (used and unused) are being considered in the official agricultural statistics.

a) Predominantly agricultural regions:

Cluster 1. Extremely high number of small farms with very low incomes. These are the core regions for small farms in Europe.

Cluster 2. Regions with few small farms, which are relatively small and have medium incomes. These are predominantly large-scale farming regions, where small farms (both in terms of area and economic size) occupy just a small part of the farming area.

Cluster 3. Regions with a low proportion of small farms, which are close to the upper size threshold and have high incomes. These are large-scale, specialised and market-oriented farming regions.

b) Regions with a balanced distribution between agriculture and other land uses:

Cluster 4. Regions with a low proportion of small farms, which are relatively small and have low incomes.

c) Regions with little agricultural land surface:

Cluster 5. Small farms exist in large numbers, which are extremely small and have low incomes. These regions are either dominated by forestry or are primarily urban.

Cluster 6. Small parts of the region are occupied by small farms, which are close to the upper size threshold and have a medium income. These are mostly regions dominated by forests, which contain the lowest proportion of agricultural land in Europe.