Growth Potential of Towns in the Western Cape
Total Page:16
File Type:pdf, Size:1020Kb
Growth Potential of Towns in the Western Cape WESTERN CAPE SPATIAL INFORMATION FORUM 14 November 2013 Growth Potential Study (GPS) of Towns IN A NUTSHELL PURPOSE? Purpose of the GPS is not to identify where growth (e.g. economic, population and physical) should occur, but rather where it is likely to occur (in the absence of significant interventions). HOW? Use quantitative data (measurements) to model the growth preconditions and innovation potential. BACKGROUND • 2004: GPS1 • Van der Merwe, Zietsman, Ferreira, Davids • 2010: GPS2 • Van Niekerk, Donaldson, Du Plessis, Spocter • 2012/13: GPS3 • Van Niekerk, Donaldson, Du Plessis, Spocter, Ferreira, Loots GPS3: PROJECT PLAN 1. Functional region mapping 2. Qualitative analysis 3. Public participation 4. Public sector priorities alignment 5. Quantitative analysis 6. Draft Report 7. Public comment GPS3: PROJECT PLAN 1. Functional region mapping 2. Qualitative analysis 3. Public participation 4. Public sector priorities alignment 5. Quantitative analysis 6. Draft Report 7. Public comment QUANTITATIVE ANALYSIS QUANTITATIVE ANALYSIS PROCEDURE 1. Create conceptual framework for estimating growth potential 2. Identify indicators that address growth potential concepts 3. Collect and manipulate data 4. Carry out statistical analyses to reduce data duplication 5. Carry out spatial analyses 6. Perform sensitivity analysis 7. Present and interpret the results DATA COLLECTION & MANIPULATION • Needed to collect data for all local municipalities and 131 settlements (as defined in GPS1) • Thiessen (Voronoi) polygons were used to define each settlement extent (sphere of influence) because: • A settlement’s function extends beyond its urban boundary • Many indicators are regional in nature (i.e. not only applicable for urban extent of town) THIESSEN POLYGONS 101 • By definition, any location within a Thiessen polygon is closer to its associated point than any other point input feature • Not ideal for representing the influence spheres of settlements of different sizes • A new method was needed to create Thiessen polygons that better represent a settlement’s influence sphere WEIGHTED THIESSEN POLYGON PROCEDURE 1. For each settlement 1. Calculate raster of Euclidian distance (ED) between towns 2. Divide ED by settlement’s population and store resulting weighted distance raster 2. Use CellStatistics (Min) on all the rasters to identify areas of minimum weighted distance 3. Convert result to single part polygons and eliminate slivers 4. Spatial join attributes of input points to resulting polygons Procedure implemented in ArcGIS using Python scripting STRUCTURAL FRAMEWORK Growth Potential Socio-economic needs ECONOMIC DATA COLLECTED Municipal Settlement # Indicator name level level 1 Average per capita income 2011 (Rands) [+] 2 Tourism potential 2008 [+] 3 % Growth of economically active population 2001 - 2011 [+] 4 Gross Value Added 2011 (current prices R million) [+] 5 Medium term annual % GVA growth 2001 to 2011[+] 6 Diversity of economic activities (2011) [+] 7 Change in diversity index (2001-2011) [+] 8 Distance to Port Elizabeth, Cape Town and 6 leader towns [-] 9 Total personal income 2011 (Rands million) [+] 10 % Growth in highly skilled labour 2001 - 2011 [+] 11 Value of property transactions 2010 [+] 12 Property tax revenue 2010 [+] 13 Number of formal retail outlets and service sector businesses 2010 [+] 14 Number of formal retail outlets and service sector businesses per person 2010 [+] PHYSICAL DATA COLLECTED Municipal Settlement # Indicator name level level 1 Mean annual precipitation [+] 2 Projected short term (2020) surplus/shortfalls of peak summer GAADD considering internal reticulation storage 2011 (mcm/a) [+] 3 Groundwater availability 2011 (mcm/a) [+] 4 Groundwater quality 2011 [-] 5 Potential evaporation (mm) [-] 6 Grazing capacity [+] 7 % Area cultivated 2012 [+] 8 Growth in % area cultivated (2007 - 2012) [+] 9 Size and status of unexploited minerals 2010 [+] INFRASTRUCTURE DATA COLLECTED Municipal Settlement # Indicator name level level 1 % households with access to the Internet 2011 [+] 2 Distance to nearest scheduled airport [-] 3 Distance to nearest commercial harbour [-] 4 Distance to nearest small harbour and slipways [-] 5 Access to main and national roads [+] 6 Access to railways [+] 7 % households with access to cellphone 2011 [+] 8 % households with access to sanitation (flush) 2011 [+] 9 % households with access to water (in house) 2011 [+] 10 % households with access to electricity (lighting) 2011 [+] 11 % households with access to waste removal 2011 [+] 12 WWTW spare capacity per person 2011 (l/day/pop) [+] 13 State of WWTW infrastructure 2011 [+] 14 % households with access to the Internet 2011 [+] INSTITUTIONAL DATA COLLECTED Municipal Settlement # Indicator name level level 1 Management experience and capacity 2010 [+] 2 Qualified audits 2012 [+] 3 Infrastructure backlog reduction 2010 [+] 4 Staff per cap ratio 2010 [-] 5 % Posts filled 2010 [+] 6 % Crime (all) occurrences change 2009 - 2012 [-] 7 Crime (all) occurrences (09 - 12) per 100 000 population [-] 8 Small business support 2010 [+] 9 Voter turnout 2010 [+] 10 Amenities 2010 [+] 11 Social service organisations 2010 [+] INSTITUTIONAL DATA COLLECTED Municipal Settlement # Indicator name level level 1 Average per capita income 2011 (Rands) [+] 2 % change in economic empowerment 2001 - 2011 [+] 3 % Unemployment 2011 [-] 4 Human Development Index 2005 [+] 5 Matric pass rate 2012 (%) [+] 6 % 20 - 65 year olds with at least grade 12 and higher [+] 7 Ratio non-economically active population age 2011 [-] ANALYSIS • GPS1 and GPS2 used slightly different methods to analyse the spatial data • Both approaches had its strengths and weaknesses • Question: Which is the best method? • GPS3 considers 16 different methods (models/approaches). Models differ in terms of: 1. Redundancy removal method 2. Indicator weighting scheme 3. Scaling (normalization) method 4. Classification method REDUNDANCY REMOVAL METHODS • Aim: to remove data redundancy (“double- counting”) within each index • Procedures: • Principle component analyses • Only keep one indicator per principle component • Radical reduction (from 84 to 18 @ settlement level & 16 @ municipal level) • Bivariate correlation analyses • Only remove indicators that have R2 > 0.7 • Less radical (from 84 to 46 @ settlement level & 52 @ municipal level) WEIGHTING SCHEMES Three approaches 1. All indicators are assigned equal weights (equally important) 2. Only select the one representative indicator per principle component and assign a weight according to the contribution of the component towards the variation within index 3. Use all indicators and assign equal weights to indicators that are associated with each principle component. The weight is determined by how much the component contributes towards the variation within an index SCALING (NORMALIZATION) METHODS Two methods 1. Linear scaling: Scales values linearly to the same minimum (0) and maximum (100) values 2. Z-score: Ensures that the mean of an indicator is zero, and the standard deviation one CLASSIFICATION METHODS Two methods 1. Quintiles: Allocates equal number of cases (settlements) to each class 2. Natural breaks (Jenk’s) Groups cases with similar values into distinct classes CLASS VALUE Very Low 1 Low 2 Medium 3 High 4 Very High 5 MODEL REDUNDANCY REMOVAL WEIGHTS NORMALIZATION METHOD CLASSIFCIATION METHOD AELN CORRELATION ANALYSIS NO LINEAR NATURAL BREAKS AEZN CORRELATION ANALYSIS NO Z-SCORE NATURAL BREAKS AULN CORRELATION ANALYSIS YES LINEAR NATURAL BREAKS AUZN PCA YES Z-SCORE NATURAL BREAKS CELN PCA NO LINEAR NATURAL BREAKS CEZN PCA NO Z-SCORE NATURAL BREAKS CULN PCA YES LINEAR NATURAL BREAKS CUZN PCA YES Z-SCORE NATURAL BREAKS AELQ CORRELATION ANALYSIS NO LINEAR QUINTILES AEZQ CORRELATION ANALYSIS NO Z-SCORE QUINTILES AULQ CORRELATION ANALYSIS YES LINEAR QUINTILES AUZQ CORRELATION ANALYSIS YES Z-SCORE QUINTILES CELQ PCA NO LINEAR QUINTILES CEZQ PCA NO Z-SCORE QUINTILES CULQ PCA YES LINEAR QUINTILES CUZQ PCA YES Z-SCORE QUINTILES IMPLEMENTATION • Automated analysis environment (ArcView GIS & Avenue scripting) • Automation essential • Indicator sets changed as new data became available • Underlying spatial data sets were frequently updated • Several methods were tested (16 models) • Reporting needs changed Sensitivity Analysis Results Settlement and Municipal level Town MEAN SD AELN AEZN AULN AUZN CELN CEZN CULN CUZN AELQ AEZQ AULQ AUZQ CELQ CEZQ CULQ CUZQ George/Heroldsbaai 5 0 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 Knysna/Brenton-on-sea 5 0 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 Mosselbaai/Groot Brak 5 0 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 Paarl 5 0 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 Stellenbosch/Jamestown 5 0 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 Betty's Bay/Pringle Bay 4.94 0.25 5 5 5 5 5 5 5 4 5 5 5 5 5 5 5 5 Hermanus/Onrus/Hawston 4.81 0.4 5 4 5 5 5 4 5 4 5 5 5 5 5 5 5 5 Malmesbury 4.63 0.5 5 4 4 4 4 4 5 5 5 5 5 5 5 5 5 5 Vredenburg 4.5 0.52 5 4 4 4 4 4 4 4 5 5 5 5 5 5 5 5 Moorreesburg 4.25 0.45 4 4 4 4 4 4 4 4 5 4 4 4 5 4 5 5 Caledon 4.31 0.48 4 4 4 4 4 4 5 4 5 4 4 CLASS4 5 4 VALUE5 5 Very Low 1 Kalbaskraal 4.44 0.51 4 4 4 4 4 4 4 4 5 5 5 5Low 5 5 25 4 Plett/Kranshoek/Wittedrif/Keurboo Medium 3 4.44 0.51 4 4 4 4 4 4 4 4 5 5 5 5 5 5 4 5 ms High 4 Pniel/Kylemore 4.56 0.51 4 4 5 4 4 4 4 4 5 5 5 Very5 High5 5 5 5 SENSITIVITY ANALYSIS RESULTS • There is relatively little variation between methodologies • No cases with SD