Tuesday, 18 November 2014

An unusual application of a MaxEnt habitat suitability model

The MaxEnt software is often used by ecologists, and others for species habitat modeling based on environmental layers.

So some data I used from the 2011 UK census (England, Wales and Cornwall) was 1. those with a skill in the Welsh language (the full question was only asked of census respondents living in Wales) and 2. those self-describing as Cornish for national identity.

The data is converted from census output polygons, to dots randomly placed within the part of the output polygon below 300m altitude.

Although there is quite a lot of land above 300m in Wales, there is actually only one or two census output area polygons that entirely disappear when terrain above 300m is cut out. So if you're in Blaenavon, apologies for deleting you.

Using the environmental layers of elevation, slope (from Shuttle Radar Topography Mission) and distance from the coast, this is the output:

Notice that the habitat suitability for Welsh speakers is actually higher in areas such as the North York Moors, and North Devon than Ynys Môn.

Habitat suitability drops off further than 60km from the coast

Altitudes of 200m-300m appear to be most suitable for Welsh speakers according to the observations of the census data.

The Welsh speakers are not suited to living on flat terrain.

A range of coastal areas are suitable for resettlement of the Cornish in the event of for example,  unexpected reactivation of the igneous activity of the Cornubian batholith.

1 comment:

  1. Fest da. (Martesen ny wonn konvedhes an graph "KW coastdist" poran.)

    Funnily enough there's a slight parallel to this tongue-in-cheekness in an article by Andrew Symons in An Baner Kernewek #95 (Feb '99) "Stress & Intonation Patterns in Cornish Dialect", in part of which there are sketch maps showing areas of granite moorland, and, beside these areas, sketches overlain of zones of dense 'occurrednces' of place names which d'show effects of pre-occlusion. They correlate well. The idea is that, helped by roads, recorded language transfer maybe wasn't so much like a blot from the east spreading south-west through flat paper, but more like a flood gradually filling up low ground and later reaching the hilltops - people who kept the language longest lived in higher elevation areas of the west.

    If yours were not such an inventively off-the-wall application of MaxEnt, his conclusions would seems to tally significantly with your "CY elev graph".

    On the serious side, have any unexpected and useful correlations been found between census data and topography more generally ? Talk of Cornubian hot rocks getting too hot is a bit worrying. Byttegyns, dhe les dhyn y fia nebes enesow nowyth dhe Gernow.