Creating consistency in British Census Space a Nigel Walford and Kelly Hayles Centre for Earth and Environmental Science Research, School of Earth Sciences and Geography GISRUK06, 5-7 April 2006
Presentation Outline Research Aim Issues and previous research Data Sources Methods Areal Interpolation Dasymetric Mapping Discussion Conclusion a
Research Aim To develop and implement a methodology for producing consistent small geographical units with associated estimates of key census counts spanning at least the British Population Censuses. Objectives Assess alternative GIS-based methodologies for deriving this information from existing data sources Examine alternative ways of visualising demographic and socio-economic changes between censuses. a
Issues Previous work Change files Census tracts (England and Wales) Reformatting SAS (Scotland) LCT – Linking Censuses Through time large spatial units a
Issues cont’d CATT – Consistent Areas Through Time (Daniel Exeter) LUTs Scotland only a
Issues cont’d Census comparability Definitions change Questions change Population base changes Geographies No digital boundaries existed of Tracts or 1981 EDs prior to Walford (2005) research Changes occurring each census year for enumeration purposes a
Data Sources a Census Counts CDU, MIMAS 142 counts for 1981 (~130,000 EDs) 304 counts for 1991 (~148,000 ED/OAs) 454 counts for 2001 (~218,000 OAs) Extracted counts and joined to DBFs due to Casweb download limits DBFs Edina ® UKBORDERS Downloaded as whole countries for England, Wales and Scotland Either as EDs or OAs Split into GOR and joined to count tables 2001 National Outline Edina ® UKBORDERS England, Wales and Scotland DBFs extracted and intersected with Strategi® depict populated non-populated areas
Data Sources cont’d a 1981 ED DBFs Walford (2005) England, Wales and Scotland EDs were joined to counts from CDU Census Tract DBFs Walford (2005) Available for England and Wales, identical over 2 years or amalgamations of two or more EDs from one or both years Census Tract Definitions Data Archive Definitions comprised 490 counts linking for England and Wales only Strategi® Edina ® Digimap ® Vector data derived from 1:250,000 topographic base. Urban, water, forest features extracted using feature codes ‘Other’ by polygon overlay
Data Sources cont’d Census Tracts DBFs DBFs a Census Tract Counts
Data Sources cont’d Strategi® England National Outline Polygonised Strategi® a
Methods Areal Interpolation Given data on one source zone, determines values based on another target system Splits one boundary based on another and proportions the counts based on area Allows aggregation back to another geographical level. Disadvantages: Assumes uniform distribution over ED/OA Does not account for areas of zero population. a
Areal Interpolation cont’d a 1981 EDs East Anglia 2001 OAs
Areal Interpolation cont’d 7126 (0.39%) a 1981 EDs which now have been proportioned to the 2001 OA boundaries ED ED_AI LOSS 861 (0.047%)
Dasymetric Mapping Using a-priori based weights to proportion counts by land use Water and forest = 0 Urban = 0.9 Other = 0.1 Weight populated EDs where both ‘urban’ and ‘other’ exist Weight all forest and water land use (x 0) Aggregate based on ‘FEATLAB’ – land use and ED (value not sum) Uses areal interpolation once weighting has been undertaken – disaggregate and aggregate on new 2001 lables More accurate analysis as accounts for areas with no population. a
Dasymetric Mapping cont’d a Intersection of Land use with 1981 ED DBFs
Dasymetric Mapping cont’d a 3. Weight selected query for all fields by the ‘rate_luse’ field 4. Select non populated land uses and weight (x 0) 2. Join the frequencies to the MapInfo Table and perform a SQL Query to select EDs where both ‘urban’ and ‘other’ are > 0 1. Create frequencies in SPSS of all land use types
Discussion cont’d Total: 100 Other (100) Urban (100) Total: 100 Other (10) Urban (90) Total: 100 Other (10) Urban (90) Water (0) Forest (0) a Urban (100) Water (0) Total: 100
Dasymetric Mapping cont’d Aggregate data (sum) where 2001 OAs contain the 1981 EDs. a ED ED_DM Loss 380 (0.02%) 380 (0.02%) 5273 (0.29%) Loss in total population of 380 using DM, whilst 861 missing in AI. Less people in external boundary in DM than AI
Loss of population in outer boundary Discrepancies in internal boundaries Discussion a 1981 and and 2001 (clipped boundaries)
Discussion cont’d ED ED_AI Loss 861 (0.047%) aED ED_DM Loss 380 (0.021%) Areal Interpolation Dasymetric Mapping
Discussion cont’d Complexity of processing Computationally intensive Large number of files Aggregation, then disaggregation and aggregation again a
Conclusion Issues with both techniques and boundaries not aligning Issues with assigning weights for land uses – what is best weight to use Land uses – assumption that they are the same for both years, not always the case Dasymetric mapping complex process, areal interpolation more simplistic to process a
Conclusion Is Dasymetric Mapping a valid technique? Is Dasymetric Mapping using land use weighting more superior than Areal Interpolation? a
References Walford, N (2005) Creating historical and contemporary small-area geography in Britain: The creation of digital boundary data for 1971 and 1981 census units, International Journal of GIS, Vol.19, No.7 August 2005, a
Copyright Statements Source: 2001 Census Output Area Boundaries. Crown copyright Crown copyright material is reproduced with the permission of the Controller of HMSO. This work is based on data provided with the support of the ESRC and JISC and uses boundary material which is copyright of the Crown and the ED-LINE Consortium. © Crown Copyright/database right 20(yy). An Ordnance Survey/EDINA supplied service. a