Frescalo

Frescalo stands for "FREquency SCAling LOcal" and was developed by Mark Hill, former head of the Biological Records Centre when it was at Monks Wood. (See: Hill, M.O. 2011. Local frequency as a key to interpreting species occurrence data when recording effort is not known. Methods in Ecology and Evolution, 3, 195-205.)

It attempts to adjust ("rescale") the relative frequencies at which species were observed in local area ("neighbourhoods") according to the amount of recording effort that was expended in each neighbourhood. Neighbourhoods are defined as a set of locations that are both physically close and environmentally similar. Fo a target location, each location in its neighbourhood is assigned a weight calculated using a combination of its proximity and its environmental similarity. The target location has a weight of 1 and each locality in its neighbourhood has a weights which is less than 1 and declines towards zero as the localities get further away and less similar.

In our case, the localities are hectads (10 x 10km squares of the Ordnance Survey National Grid for Great Britain).Neighbourhoods consist of around 50 hectads with the environmental similarity component being calculated by comparing the topography (mean altitude, slope and aspect), weather (various ecologically relevant measures of temperature and rainfall) and land cover (areas of woodland, grassland, etc.) of these hectads.

The following map illustrates a neighbourhood. The target square on the coast of South Wales is shown in red. Its approx. 50 neighbouring hectads are shown in green with their weight represented by the area of the circle. As you can see, the neighbouring squares are considered to be mainly coastal and, as they get further inland or further away, their weights decline. Some relatively nearby coastal squares are not included in the neighbourhood - presumably because the environments they provide are rather different.

Visualisation of a neighbourhood
Visualisation of the weights assigned to hectads in a neighbourhood.

The basic input data for the analysis is unique combinations of hectad, species and year since 1980 from the Hoverfly Recording Scheme database. It looks like this:

NF75,Episyrphus balteatus,2007
NF75,Eristalis intricarius,2016
NF75,Eristalis pertinax,2016
NF75,Eristalis pertinax,2017
NF76,Anasimyia lineata,1989
NF76,Episyrphus balteatus,1989
NF76,Eristalis intricarius,1989
NF76,Eristalis intricarius,1997
NF76,Eristalis intricarius,2014

For each hectad, records that fall within its neighbourhood are accumulated with the contribution of that record being the weight assigned to its hectad. What we end up with is a weighted sum of the number of records of each species which can be turned into a weighted frequency by dividing by the total sum of weights for all the record from that neighbourhood. If we now rank the species by their weighted frequency, we can plot the weighted frequency against rank and get a form of species discovery curve.

Weighted frequency vs rank examples
Examples of weighted relative frequency vs rank for some hectads

It is clear some localities are better recorded than others. They have received more recording effort and more species have been found. The analysis fits an average species discovery curve across all the available data and then adjust the curve in each neighbourhood towards that average curve shape by rescaling the frequencies of species up or down. There is a basic assumption here that the species discovery curve is consistent across time and space for the particular taxonomic group.

These rescaled frequencies can then be averaged across the hectads to give a measure of the relative frequency of a species in a given year corrected for the recording effort (termed the "TFactor"). This can be plotted against year to investigate how the relative frequency of the species has changed, and hence estimate a trend. It can also be average, for a particular hectad across years and plotted as a map to show how the relative frequency varies spatially.

Further information: