The  influence  of  private  and  public   greenspace  on  short-­‐term  health  indicators                 Nicholas  Brunsdon    

GEOG420  Dissertation  2013   University  of  Canterbury    

  Abstract   The   contribution   of   contextual   factors   such   as   greenspace   exposure   and   accessibility   towards   individual   health   has   been   subject   to   an   increasing   amount   of   research   in   the   recent   years.   However,   significant   results   elsewhere   have   not   been   widely   replicated  in  New  Zealand.  This  study  looked  at  the  time  spent  undertaking  physical   activity   and   mental   health   symptoms   of   2,557   individuals   through   Auckland,   New   Zealand,   relating   it   to   a   range   of   measures   of   neighbourhood   greenspace.   Greenspace   exposure   and   accessibility   was   calculated   for   each   meshblock   in   Auckland   City   using   a   geographic   information   system   (GIS),   indicating   accessibility   to   usable   greenspace   through   network   distance;   the   area   of   all   public   greenspaces   through   a   Euclidean   buffer,   and   the   area   of   all   greenspaces   (including   private   gardens   through   a   Euclidean   buffer,   created   with   a   new   method   from   impervious   surface   data.   Regression   analysis   related   these   area   level   variables   to   individual   health   variables   from   the   New   Zealand   Health   Survey,   controlling   for   a   range   of   demographic   factors.   Vigorous   physical   activity   was   found   to   be   lowest   in   neighbourhoods   with   the   highest   access   to   usable   greenspace,   and   a   significant   relationship   was   found   between   private   greenspace   and   nervous   symptoms,   although   with   a   contradictory   direction   of   association.   No   significant   relationship   was   found   between   any   measures   of   greenspace   and   depressive   symptoms.   A   number   of   opportunities   to   investigate   the   contribution   of   greenspace   and   bluespace   on   health   are   recommended.   Understanding   how   urban   environments   influence  health  is  important  to  inform  urban  planning.  

2  

Nicholas  Brunsdon  

Table  of  Contents   Abstract  ……………………………………………………………………………………………………………….2   Table  of  Contents  …………………………………………………………………………………………………3   List  of  Figures  ……………………………………………………………………………………………………….4   List  of  Tables  ………………………………………………………………………………………………………..5   1.   Introduction  ....................................................................................................  6   2.   Background  .....................................................................................................  8   2.1   Theoretical  framework  ....................................................................................  8   2.2   Prior  studies  .....................................................................................................  9   2.3   Health  and  greenspace  methodologies  .........................................................  11   2.3.1   Health  Methodologies  ............................................................................  11   2.3.2   Greenspace  Methodologies  ....................................................................  12   3.   Research  Objectives  ......................................................................................  15   3.1   Research  Questions  .......................................................................................  15   4.   Methods  ........................................................................................................  16   4.1   Spatial  Analysis  ..............................................................................................  17   4.1.1   Public  greenspaces  .................................................................................  19   4.1.2   Usable  greenspaces  ................................................................................  20   4.1.1   Impervious  surfaces  ................................................................................  21   4.2   Health  analysis  ...............................................................................................  23   5.   Results  ..........................................................................................................  28   6.   Discussion  .....................................................................................................  36   6.1   Greenspace  Indicators  ...................................................................................  36   6.2   Socioeconomic  Deprivation  ...........................................................................  37   6.3   Physical  Activity  .............................................................................................  37   6.4   Mental  Health  ................................................................................................  38   6.5   Future  Research  .............................................................................................  39   6.6   Theoretical  Pathways  ....................................................................................  40   7.   Conclusion  .....................................................................................................  42   Acknowledgements  ……………………………………………………………………………………………  43   References  …………………………………………………………………………………………………………  44   Appendix  A  ………………………………………………………………………………………………………..  53  

3  

List  of  Figures   Figure  1:  -­‐  Theoretical  overview  of  linkages  between  greenspace  data,  greenspace   classifications,  theoretical  linkages  and  hypothesised  health  outcomes  ...........  16   Figure  2:  -­‐  Map  of  study  area  in  relation  to  Auckland  cities,  noting  the  exclusion  of   North  Shore,  Albany  and  Gulf  Harbour  islands.  Inset:  New  Zealand,  with   Auckland  region  indicated.  .................................................................................  17   Figure  3:  -­‐  Map  of  illustrative  meshblock  showing  the  relationship  between  the   meshblock  area,  where  NZHS  respondents  are  located,  and  the  PWC   representing  the  meshblock.  500m  and  2000m  buffers  are  measured  as  radii   from  the  PWC  .....................................................................................................  18   Figure  4:  -­‐  Map  of  illustrative  meshblock  showing  0-­‐500m  and  500-­‐2000m   (doughnut)  buffers  from  the  meshblock  PWC  and  how  all  public  greenspace   (usable  and  non-­‐usable)  within  each  buffer  is  identified  from  which  to  be   calculated  as  a  proportion  of  the  total  buffer  size.  ............................................  20   Figure  5:  -­‐  Map  of  illustrative  meshblock  showing  usable  greenspaces  and  how  their   accessibility  to  the  meshblock  is  appraised  –  by  the  average  travel  time  via  the   road  network  from  the  meshblock  PWC  to  the  nearest  five  usable  greenspaces  ............................................................................................................................  22   Figure  6:  -­‐  Map  of  illustrative  meshblock  showing  impervious  surfaces  (such  as  roads,   buildings  and  footpaths)  within  and  beyond  a  500m  buffer  of  the  meshblock.   23    

4  

Nicholas  Brunsdon  

List  of  Tables   Table  1:  -­‐  Summary  information  for  area-­‐level  variables  of  respondents  ..................  24   Table  2:  -­‐  Summary  information  for  2,557  adult  respondents  to  the  2011/12  NZHS   within  the  Auckland  study  area,  representing  a  population  of  810,159.  ...........  26   Table  3:  -­‐  Correlation  matrix  of  area  level  greenspace  and  deprivation  variables,   indicating  correlation  coefficient  (p-­‐value).  .......................................................  28   Table  4:  -­‐  Logistic  regression  of  the  relationship  between  time  spent  exercising   (dependent  variable)  and  individual  and  area  level  variables,  each  considered   separately.  ..........................................................................................................  29   Table  5:  -­‐  Logistic  regression  of  the  relationship  between  time  spent  exercising   (dependent  variable)  and  area  level  usable  greenspace  access,  controlling  for   age,  sex  and  area  deprivation.  ............................................................................  30   Table  6:  -­‐  Logistic  regression  of  the  relationship  between  mental  illness  symptoms   (dependent  variable)  and  individual  and  area  level  variables,  each  considered   separately.  ..........................................................................................................  31   Table  7:  -­‐  Logistic  regression  of  the  relationship  between  mental  illness  symptoms   (dependent  variable)  and  usable  greenspace,  controlling  for  prior  doctor   diagnosis.  ............................................................................................................  32   Table  8:  -­‐  Logistic  regression  of  the  relationship  between  mental  illness  symptoms   (dependent  variable)  and  public  greenspace,  controlling  for  prior  doctor   diagnosis.  ............................................................................................................  33   Table  9:  -­‐  Logistic  regression  of  the  relationship  between  mental  illness  symptoms   (dependent  variable)  and  all  greenspaces  (derived  from  impervious  surfaces),   controlling  for  doctor  diagnosis.  .........................................................................  34   Table  A1:  -­‐  Logistic  regression  of  the  relationship  between  prior  doctor  diagnosis  for   mental  illness  (dependent  variable)  and  individual  and  area  level  variables,  each   considered  separately.  .  ......................................................................................  56  

5  

1. Introduction   During   the   preceding   century   our   understanding   of   health   has   broadened   beyond   merely   the   absence   of   illness   to   include   the   salutogenic   influence   of   environments   towards   health.   For   highly   urbanised   populations,   scarce   areas   of   greenspace   can   have  an  important  contribution  towards  health.  The  scarcity  and  uneven  distribution   of   greenspace   throughout   urban   spaces   gives   rise   to   analysis   of   how   the   distribution   of   health   relates   to   that   of   greenspaces.   A   theoretical   framework   established   through   previous   empirical   work   proposes   that   greenspace   exposure   influences   health   through   three   key   avenues   –   as   a   therapeutic   landscape,   in   promoting   physical   activity   and   as   a   venue   for   informal   social   interactions   (Dillen,   Vries,   Groenewegen,  &  Spreeuwenberg,  2012).     Quantitative  research  in  many  other  developed  nations  has  provided  evidence  of  a   strong   positive   association   between   greenspace   exposure   and   accessibility   and   positive  health  outcomes  (Lachowycz  &  Jones,  2011);  however  this  relationship  has   not   been   replicated   widely   in   studies   of   the   New   Zealand   context   to   date   (Richardson,  Pearce,  Mitchell,  Day,  &  Kingham,  2010).  Prior  research  has  tended  to   estimate   population   exposure   by   measuring   exposure   to   public   greenspace   only,   without   including   private   gardens   (Lachowycz   &   Jones,   2011),   however   this   study   measures   greenspace   through   several   different   measures.   The   distance   from   the   population-­‐weighted   centroid   of   each   census   meshblock 1  to   the   nearest   public   usable   (able   to   be   visited)   greenspace   is   used   as   a   measure   of   greenspace   accessibility.  Similarly,  the  area  of  all  public  greenspace  (incorporating  usable  public   greenspace)   surrounding   the   meshblock   is   used   as   a   measure   of   greenspace   exposure.   The   proportion   of   impervious   surfaces2  in   each   meshblock   is   used   as   a   measure   of   exposure   to   all   forms   of   greenspace,   including   private   gardens,   which   have  not  been  quantified  in  relation  to  health  in  prior  studies.                                                                                                               1     Meshblocks   are   the   smallest   geographic   areas   that   Statistics   New   Zealand   (SNZ)   collects  data  in,  and  are  designed  to  include  approximately  110  individuals  (Statistics   New  Zealand,  2013).   2  Impervious   surfaces   are   artificial   surfaces   such   as   buildings,   roads,   footpaths   and   car  parks   6  

Nicholas  Brunsdon     Quantification   of   health   in   existing   work   tends   to   use   cross   sectional   data,   looking   across   society   at   a   single   point   in   time;   however   this   ignores   migration,   and   thus   the   contribution  that  a  previous  neighbourhoods  may  have  had  on  the  present  health  of   an  individual.  This  effect  is  exacerbated  by  the  use  of  chronic  health  conditions,  such   as  cardiovascular  disease,  to  measure  health  as  they  tend  to  be  strongly  influenced   by   environmental   exposures   and   lifestyle   factors   over   a   lifetime   (World   Health   Organization,   World   Heart   Federation,   &   World   Stroke   Organization,   2011).   Following   a   quantitative   methodology,   this   research   investigated   the   greenspace-­‐ health   relationship   through   two   short-­‐term   health   measures   -­‐   time   spent   undertaking   physical   activity,   as   a   proxy   for   individuals   physical   health,   and   mental   illness   symptoms;   both   measures   were   expected   to   relate   more   closely   to   an   individual’s   current   neighbourhood   environment   than   if   measured   by   a   chronic   medical   state.   This   is   based   upon   respondents   to   the   2011/12   NZ   Health   Survey   (NZHS),   focused   on   the   area   of   the   Auckland   Unitary   Authority   (‘supercity’)   which   represents   a   range   of   urban   environments   with   varying   demographic   makeup   and   population  densities.        It   was   anticipated   that   distinguishing   between   exposure   to   usable   and   all   greenspace  would  yield  further  insights  into  the  causal  pathways  of  greenspace  and   health,   as   the   physical   activity   and   social   interaction   pathways   should   only   be   affected  by  usable  greenspace,  whereas  the  therapeutic  landscape  pathway  should   be  affected  by  both.  Such  insights  could  be  useful  in  the  development  of  policy  and   urban  planning  to  improve  the  salutogenic  properties  of  future  communities.  

7  

2. Background   The   contribution   of   environmental   influences   on   health   was   developed   notably   by   Antonovsky   (1979)   who   introduced   the   notion   of   salutogenesis,   broadening   the   understanding  of  health  beyond  the  prevailing  pathogenic  view  to  encompass  both   positive   and   negative   influences   on   health.   This   notion   has   been   embraced   in   the   field   of   health   geography,   which   has   taken   a   closer   look   at   the   role   of   place   in   influencing   the   health   of   individuals,   commonly   dividing   such   influences   into   the   compositional   or   contextual   (Macintyre,   Maciver,   &   Sooman,   1993).   Compositional   effects  take  account  of  the  selection  of  individuals  in  a  neighbourhood,  for  instance   neighbourhoods   with   more   affluent   individuals;   whereas   contextual   factors   include   the   characteristics   of   the   neighbourhood   that   have   an   association   with   the   health   of   residents,   such   as   the   provision   of   public   amenities.   Both   are   often   intrinsically   related,   for   instance   affluent   individuals   may   be   attracted   to   areas   with   greater   public   amenities;   thus   separating   composition   from   context   is   an   on-­‐going   challenge   to   understanding   the   contribution   of   place   towards   health   (Vries,   Verheij,   Groenewegen,   &   Spreeuwenberg,   2003).   Investigation   of   the   spatial   relationship   between  greenspace  and  health  began  with  the  quantitative  study  of  Ulrich  (1984)   which  found  significant  differences  in  recovery  rates  for  gall-­‐bladder  removal  surgery   patients  between  those  in  rooms  with  a  view  of  a  stand  of  trees  and  those  looking   out  to  a  brick  wall.  A  significant  body  of  work  has  extended  this  notion  to  investigate   the   contribution   of   greenspace   in   urban   areas   on   the   health   of   urban   inhabitants   (Dillen  et  al.,  2012;  Lachowycz  &  Jones,  2011;  Vries  et  al.,  2003).      

2.1 Theoretical  framework   The   association   between   urban   greenspace   and   health   is   underpinned   through   a   set   of   three   theoretical   pathways   developed   in   an   extensive   body   of   quantitative   and   qualitative   studies.   Greenspace   is   proposed   to   act   as   a   therapeutic   landscape,   supporting  relaxation  and  recovery  from  stressful  activities  and  attention  fatigue;  it   is  proposed  that  this  is  mediated  by  visualisation  of  greenspace  (Hartig  et  al.,  2011).   As   visualisation   is   derived   through   passive   engagement,   it   is   not   dependent   on  

8  

Nicholas  Brunsdon   physical   access,   and   thus   views   of   inaccessible   greenspaces   such   as   neighbouring   private  gardens  may  contribute  towards  this  pathway  (Dillen  et  al.,  2012).  Secondly,   greenspace   is   theorised   to   promote   physical   activity   by   providing   a   venue   for   physical   activity   and   to   encourage   active   lifestyles   more   generally,   all   of   which   mitigate   mental   illnesses   and   improve   physical   health   with   lower   risk   of   heart   disease  and  obesity  related  conditions  (Vries  et  al.,  2011).  Thirdly,  it  is  proposed  that   access   to   greenspace   may   provide   a   venue   for   social   interaction   with   resultant   positive   impacts   on   mental   health   through   group   activities   and   social   encounters   (Maas,   van   Dillen,   Verheij,   &   Groenewegen,   2009;   Zhou   &   Rana,   2012).   Both   the   physical   activity   and   social   interaction   pathways   rely   on   active   engagement   with   greenspaces;   consequently   such   greenspaces   should   be   accessible,   publicly   accessible   and   have   sufficient   size   to   influence   health   through   these   pathways.   A   number  of  studies  have  established  the  significance  of  environmental  determinants,   particularly   the   accessibility   of   recreational   facilities   on   physical   activity   behaviours   (Wendel-­‐Vos,  Droomers,  Kremers,  Brug,  &  Van  Lenthe,  2007).    

2.2 Prior  studies   The   contribution   of   greenspace   to   health   outcomes   has   been   studied   extensively,   particularly   over   the   past   decade   following   improvement   in   computing   power   and   development   of   geographic   information   systems   (GIS).   Of   60   studies   investigating   greenspace,   68%   found   positive   or   weak   positive   relationships   with   obesity-­‐related   measures,  and  40%  found  a  positive  relationship  with  self-­‐reported  physical  activity   measures   (Lachowycz   &   Jones,   2011).   As   part   of   isolating   compositional   and   contextual  effects,  studies  tend  to  find  significant  interaction  in  this  relationship  with   sex   and   ethnicity   (Kerr,   Frank,   Sallis,   &   Chapman,   2007;   Pate   et   al.,   2008).   While   gross   measures   of   greenspace   availability   are   common,   studies   attempting   to   quantify   the   quality   of   greenspace   have   found   that   the   two   are   often   related,   and   that  the  quality  of  greenspaces  can  also  be  positively  correlated  with  health  (Dillen   et  al.,  2012).    

9  

In  New  Zealand,  only  a  handful  of  studies  attempted  to  quantify  the  health  impact  of   exposure  to  greenspace,  and  found  no  relationship  between  greenspace  access  and   cause-­‐specific   mortality   (Richardson,   Pearce,   Mitchell,   Day,   &   Kingham,   2010);   or   BMI,   sedentary   behaviour   or   physical   activity   (Witten,   Hiscock,   Pearce,   &   Blakely,   2008).   More   recently,   Richardson,   Pearce,   Mitchell,   &   Kingham   (2013)   found   lower   risks   of   mental   health   and   cardiovascular   disease   in   neighbourhoods   with   greater   areas  of  public  greenspace,  and  slightly  higher  levels  of  physical  activity.  It  has  been   suggested   that   universally   high   access   to   greenspaces,   particularly   private   gardens   results  in  low  variation  in  accessibility  or  exposure  measures,  and  this  explains  their   poor  explanatory  power  in  the  New  Zealand  context.  Private  gardens  in  New  Zealand   tend  to  be  larger  than  the  United  Kingdom  counterparts  (Loram,  Tratalos,  Warren,  &   Gaston,  2007;  Mathieu,  Freeman,  &  Aryal,  2007);  and  residents  in  three  out  of  four   New  Zealand  neighbourhoods  are  within  a  2.4  minute  car  trip  to  a  park  of  some  form   (Witten  et  al.,  2008).  Indeed,  Pearce,  Witten,  &  Bartie  (2006)  found  that  parks  were   the  most  accessible  of  all  community  facilities  in  their  New  Zealand  study.     Due   to   significant   variation   in   health   states   and   behaviours   across   society,   differences   in   age,   gender   and   income   must   be   considered   before   isolating   the   contribution  of  greenspace  on  health,  otherwise  compositional  factors,  such  as  the   demographics   of   a   neighbourhood   will   be   attributed   to   the   contextual   factors   of   greenspace.   The   prevalence   of   depression   is   found   to   be   higher   in   women   than   in   men,   and   across   both   genders   is   higher   at   early   and   later   stages   of   life   (Mirowsky,   1996).   There   is   non-­‐causal   evidence   of   a   higher   prevalence   amongst   individuals   of   lower   incomes   (Zimmerman   &   Katon,   2005).   Anxiety   disorders   are   slightly   more   prevalent  amongst  women,  however  age  and  gender  correlates  of  anxiety  disorders   are   weak   due   the   heterogeneous   nature   of   conditions   that   comprise   anxiety   disorders,   as   there   are   strong   trends   within   specific   anxiety   disorders   (Anthony   &   Stein,   2008;   Crawford   &   Henry,   2003).   Neighbourhood   socioeconomic   deprivation   (SED)   is   identified   as   a   correlate   of   depression   and   anxiety   disorder   prevalence   in   some   studies   (Kling,   Liebman,   &   Katz,   2007;   Lofors,   Ramírez-­‐León,   &   Sundquist,   2006;   Mair,   Roux,   &   Galea,   2008;   Stafford   &   Marmot,   2003).   Ethnicity   is   often   a  

10  

Nicholas  Brunsdon   significant  correlate  of  both,  although  this  is  often  captured  within  income  and  SED   measures,  and  is  ultimately  context  dependent  (Anthony  &  Stein,  2008).      

2.3 Health  and  greenspace  methodologies     Greenspace  and  health  studies  tend  to  take  a  cross  sectional  approach  –  seeking  to   understand  the  relationship  by  looking  at  a  cross  section  of  a  society  at  a  single  point   in   time.   Studies   will   often   quantify   health   with   long-­‐term   indicators   such   as   cardiovascular  disease  or  obesity,  and  attempt  to  associate  their  prevalence  with  an   individuals’  residence  at  that  point  in  time.  However,  this  approach  does  not  account   for  migration  and  the  effect  of  previous  residences  on  an  individuals’  present  health,   particularly   when   measuring   an   individuals’   health   state   through   the   presence   of   long   term   health   conditions   such   as   obesity   and   cardiovascular   disease.   Long   term   conditions  manifest  out  of  lifestyle  and  environmental  factors  over  a  lifetime,  thus  to   accurately   understand   their   prevalence   requires   an   understanding   of   all   of   individuals   residences   and   other   environmental   factors   throughout   their   lifetime   (World   Health   Organization   et   al.,   2011).   This   would   ideally   be   achieved   through   a   longitudinal  study,  however  these  are  rare  and  generally  expensive.      

2.3.1 Health  Methodologies   Studies   analysing   the   impact   of   greenspace   on   mental   health   tend   to   use   doctor-­‐ diagnosed   disorders,   however   this   is   subject   to   the   aforementioned   limitations   of   the  cross-­‐sectional  approach  -­‐  mental  illnesses  can  be  persistent  (Mann,  2005),  thus   mental   illness   can   be   a   function   of   previous   environments.   Short-­‐term   indicators   such  as  depressive  or  nervous  feelings  provide  an  alternative  to  this,  as  these  are  less   dependent   on   prior   mental   illnesses   and   more   likely   to   reflect   the   present   environment   experienced   by   an   individual,   whilst   also   consistently   indicate   clinical   mental  illness  (Solomon,  Haaga,  &  Arnow,  2001;  Weissman,  Sholomskas,  Pottenger,   Prusoff,   &   Locke,   1977).   Self-­‐reported   mental   illness   or   symptoms   also   have   lower   survey   collection   costs   and   are   commonly   used   instead   of   diagnosis   by   a   medical   professional  (Dillen  et  al.,  2012;  Gove  &  Geerken,  1977;  Spitzer,  Kroenke,  &  Williams,  

11  

1999).  A  small  number  of  studies  have  qualitatively  assessed  the  restorative  health   benefit  of  private  gardens,  particularly  for  mental  health  (Stigsdotter  &  Grahn,  2004;   Unruh,  2002),  however  these  often  fail  to  account  for  the  size  of  the  garden  or  the   neighbourhood  characteristics  such  as  accessibility  to  public  greenspace.     Levels   of   physical   activity   are   commonly   obtained   through   self-­‐reported   measures   (Lachowycz  &  Jones,  2011),  and  have  a  significant  variation  by  gender  and  age,  with   lower   activity   in   woman   compared   to   men,   and   declining   as   age   increase   (Trost,   Owen,   Bauman,   Sallis,   &   Brown,   2002).   There   is   some   evidence   of   higher   levels   of   physical   activity   amongst   individuals   of   higher   incomes   (Craig,   Russell,   Cameron,   &   Bauman,  2003).  Contextual  factors  were  found  to  explain  physical  activity  in  a  study   of  individuals  in  Perth,  Australia,  with  those  in  high  SED  neighbourhoods  less  likely  to   undertake  physical  activity  than  those  in  low  SED  areas,  in  spite  of  higher  access  to   recreational   facilities   in   high   SED   areas   (Giles-­‐Corti   &   Donovan,   2002).   Physical   activity   is   also   related   to   mental   illness,   with   significant   evidence   that   physical   activity   is   an   effective   treatment   for   depression   (Babyak   et   al.,   2000)   and   anxiety   disorders  (O’Connor,  Raglin,  &  Martinsen,  2000).    

2.3.2 Greenspace  Methodologies     Greenspace   is   typically   defined   to   include   parks,   reserves   and   other   public   open   spaces   within   an   urban   setting   (American   Planning   Association   &   National   Association   of   County   &   City   Health   Officials,   2013;   Lee   &   Maheswaran,   2011);   however   this   has   been   broadened   to   include   vegetation   in   a   streetscape   setting   (Vries,  Dillen,  Groenewegen,  &  Spreeuwenberg,  2013).  This  is  frequently  quantified   through  accessibility  or  exposure  measures,  with  some  studies  delving  beyond  into   the   usability   or   quality   of   greenspace.   Public   greenspaces   are   often   studied   due   to   the   ease   of   access   to   data   through   public   amenity   databases,   however   techniques   have   been   developed   with   remote   sensing   to   map   private   greenspaces.   Such   measures  applied  to  private  gardens  have  been  advanced  in  the  field  of  ecology,  but   are  limited  by  the  availability  and  cost  of  high  resolution  imagery,  and  have  not  been   widely  applied  to  health  (Mathieu  et  al.,  2007).     12  

Nicholas  Brunsdon   Greenspace   is   commonly   sub-­‐classified   into   usable   greenspace,  that   is,   public   spaces   that   not   only   provide   visual   amenity,   but   also   are   functional   for   the   purposes   of   physical  activity  or  social  interaction;  and  non-­‐usable  greenspace  from  which  benefit   is  largely  derived  through  visualisation  rather  than  physical  engagement.  A  range  of   criteria   have   been   applied   for   this   classification,   but   the   most   common   criteria   for   physical   activity   by   adults   is   that   a   contiguous   area   be   greater   than   2   hectares   (Coombes,  Jones,  &  Hillsdon,  2010;  Natural  England,  2009;  Stubbs,  2008).  As  physical   activity  in  a  greenspace  requires  active  engagement,  the  most  appropriate  measure   of   accessibility   is   based   on   road   network   distance   or   time,   as   travel   time   to   visit   a   greenspace   will   depend   on   the   connectivity   of   the   road   network   in   most   cases   (Comber,   Brunsdon,   &   Green,   2008;   Pearce   et   al.,   2006).   Studies   typically   use   the   network  travel  time  or  distance  to  the  nearest  single  greenspace,  or  the  number  of   greenspaces   within   a   specific   travel   time   as   an   indicator   of   access,   however   this   fails   to   take   account   of   other   preference   factors   in   the   decision   making   process   of   individuals   selecting   a   greenspace   to   exercise   in   (Lachowycz   &   Jones,   2011).   Indicators  of  the  distribution  of  greenspaces  are  most  commonly  used  to  predict  the   health   status   or   physical   activity   level   of   individuals   in   the   absence   of   specific   information  on  greenspace  preferences.       The  visualisation  of  greenspace  for  the  therapeutic  landscape  pathway  is  commonly   measured  by  exposure  rather  than  accessibility  of  greenspace,  as  visualisation  can  be   passive.   Given   that   this   benefit   does   not   require   physically   visiting   greenspace,   a   network   measure   of   accessibility   is   not   appropriate;   instead   a   buffer   approach   is   commonly  used.  A  buffer  is  a  circular  area  cast  around  a  point  by  a  fixed  distance,  an   example   of   which   can   be   seen   later   in   Figure   3.   Greenspace   studies   tend   to   derive   buffers   around   points   that   represent   individuals   or   neighbourhoods,   and   calculate   the  area  of  greenspace  contained  by  the  buffer  to  indicate  greenspace  exposure  for   those  individual(s)  (Lachowycz  &  Jones,  2011).  The  influence  of  greenspace  on  health   been  observed  with  buffers  with  radii  from  500  metres  to  3  kilometres,  although  this   does   not   take   into   account   terrain   and   other   obstructions   to   the   visualisation   of   greenspace   (Barbosa   et   al.,   2007;   Dadvand   et   al.,   2012;   Dillen   et   al.,   2012;   Richardson   et   al.,   2012;   Schipperijn,   Stigsdotter,   Randrup,   &   Troelsen,   2010;   13  

Villeneuve  et  al.,  2012).  Similarly,  the  influence  of  greenspace  on  house  prices  was   noted  to  extend  to  2km  (Mayor,  Lyons,  Duffy,  &  Tol,  2009).    

14  

Nicholas  Brunsdon  

3. Research  Objectives   The  contribution  of  urban  greenspace  to  health  outcomes  has  been  studied  widely   internationally,  and  to  a  limited  degree  in  New  Zealand.  However,  significant  results   overseas   have   not   been   widely   replicated   in   New   Zealand   to   date,   with   private   gardens   typically   ignored.   This   study   sought   to   adapt   existing   methodologies   to   incorporate   private   greenspace   into   an   assessment   of   the   impact   of   urban   greenspace   exposure   and   accessibility   on   health.   Furthermore,   this   study   will   appraise  health  through  short-­‐term  indicators,  instead  of  the  prevailing  approach  of   using  long-­‐term  indicators.     The  following  research  questions  were  devised  to  further  the  understanding  of  the   greenspace   and   health   relationship,   particularly   in   the   New   Zealand   context   and   through  the  investigation  of  private  greenspace.    

3.1 Research  Questions   Does   living   near   usable   greenspace   influence   the   time   spent   by   individuals   undertaking  physical  activities?     Does   the   exposure   or   accessibility   to   public   greenspace   in   urban   contexts   influence   short-­‐term  mental  health  indicators?     Does   the   exposure   to   private   greenspace   in   urban   contexts   influence   short-­‐term   mental  health  indicators?      

15  

4. Methods   This  study  follows  the  conceptual  model  illustrated  in  Figure  1  that  relates   greenspace  data  sources,  theoretical  pathways  and  health  outcomes,  with  analysis   structured  such  that  the  contribution  of  each  causal  pathway  can  be  quantified.   Spatial  analysis  of  the  greenspace  data  inputs  was  conducted  in  the  Spatial  and   Network  analysis  extension  of  ESRI  ArcMap,  which  then  fed  into  health  analysis  in   the  statistical  analysis  package  STATA.  Health  analysis  was  based  upon  the  2,557   respondents  to  the  New  Zealand  Health  Survey  (NZHS)  within  the  Auckland  study   area.      

  Figure  1:  -­‐  Theoretical  overview  of  linkages  between  greenspace  data,  greenspace  classifications,   theoretical  linkages  and  hypothesised  health  outcomes  

      16  

Nicholas  Brunsdon  

4.1 Spatial  Analysis   Spatial   analysis   was   restricted   to   the   territory   of   the   Auckland   Unitary   Authority   (‘Supercity’).   High-­‐resolution   impervious   surface   data   was   sourced   from   the   Auckland   Council,   however   this   was   not   available   for   meshblocks   in   the   North   Shore   and  Albany  wards,  and  thus  these  areas  were  excluded  from  analysis.  Likewise,  NZHS   data   was   not   collected   for   meshblocks   on   populated   islands   in   the   Waitemata   and   Gulf  wards,  so  the  study  area  was  reduced  further  to  the  extent  highlighted  in  Figure   2.  

  Figure  2:  -­‐  Map  of  study  area  in  relation  to  Auckland  cities,  noting  the  exclusion  of  North  Shore,   Albany  and  Gulf  Harbour  islands.  Inset:  New  Zealand,  with  Auckland  region  indicated.  

 

17  

Spatial  analysis  was  conducted  with  census  meshblocks,  as  data  from  the  NZHS  was   aggregated   to   this   level   to   preserve   the   anonymity   of   individuals.   With   no   indication   of   greenspace   preferences   and   aggregation   of   individual   respondents   to   the   meshblock   level,   the   influence   of   greenspace   on   individuals’   health   was   estimated   from   the   exposure   of   their   meshblock   to   greenspace.   Greenspace   exposure   was   assessed   from   a   population-­‐weighted   centroid   (PWC)   for   each   meshblock,   a   single   point   that   represents   a   meshblock   area   with   consideration   of   the   population   distribution   within   the   meshblock,   which   is   illustrated   in   Figure   3   for   an   illustrative   meshblock.   In   this   study   greenspace   was   assessed   in   three   senses   –   all   public,   public   usable  and  all  greenspace  (private  and  public,  from  impervious  surfaces).  

  Figure  3:  -­‐  Map  of  illustrative  meshblock  showing  the  relationship  between  the  meshblock  area,   where  NZHS  respondents  are  located,  and  the  PWC  representing  the  meshblock.  500m  and  2000m   buffers  are  measured  as  radii  from  the  PWC  

18  

Nicholas  Brunsdon  

4.1.1 Public  greenspaces   Public  greenspaces  were  located  from  the  dataset  developed  in  Pearce  et  al.  (2006)   and  Richardson  et  al.  (2010)  which  augmented  the  nationwide,  low  resolution  Land   Cover   Database   (LCDB2)   with   finer,   less   contiguous   datasets   from   Department   of   Conservation  (DoC)  and  Land  Information  New  Zealand.  This  included  natural  areas   such   as   parks,   beaches   and   fields,   although   excluded   aquatic   areas   such   as   lakes   and   the   sea,   as   these   are   not   considered   greenspaces.   The   dataset   classifies   greenspaces   into  two  types  based  on  the  way  in  which  individuals  are  theorised  to  benefit  from   them.  Visual  amenity  can  be  derived  through  passive  engagement  with  all  forms  of   greenspace;   usable   public   greenspaces   as   a   subset   of   public   greenspace   include   parks   where   individuals   can   derive   benefit   through   active   engagement,   such   as   through   physical   activity   or   social   interaction   within   the   space.   These   greenspaces   were   dissolved   such   that   areas   with   common   boundaries   were   joined   together,   which   included   of   9,052   distinct   greenspaces   (134,680   hectares)   within   the   study   area.     The   exposure   to   all   public   greenspace   (usable   and   non-­‐usable)   for   the   purposes   of   assessing   therapeutic   landscape   effects   was   assessed   in   terms   of   a   Euclidean   distance   buffer   of   0-­‐500   metres   and   500-­‐2000   metres   (effectively   a   doughnut),   as   the   visualisation   of   greenspace   need   not   depend   on   access   via   the   road   network   (Barbosa   et   al.,   2007;   Dadvand   et   al.,   2012;   Dillen   et   al.,   2012;   Richardson   et   al.,   2012;  Schipperijn  et  al.,  2010;  Villeneuve  et  al.,  2012).  Visualisation  will  be  affected   by   terrain   and   built   environment   obstructions,   however   this   was   not   able   be   appraised   in   this   study.   Euclidean   buffers   around   each   meshblock   PWC   were   generated  (Figure  4),  and  the  proportion  of  land  within  each  buffer  area  (excluding   the   ocean)   that   was   covered   by   public   greenspace   within   the   buffer   derived.   For   buffers   that   included   bodies   of   water,   the   greenspace   areas   were   given   as   a   proportion  of  the  buffer  area  that  comprised  land.    

19  

  Figure  4:  -­‐  Map  of  illustrative  meshblock  showing  0-­‐500m  and  500-­‐2000m  (doughnut)  buffers  from   the  meshblock  PWC  and  how  all  public  greenspace  (usable  and  non-­‐usable)  within  each  buffer  is   identified  from  which  to  be  calculated  as  a  proportion  of  the  total  buffer  size.  

 

4.1.2 Usable  greenspaces   Usable  greenspaces  are  a  subset  of  public  greenspaces,  defined  as  being  larger  than   2  hectares  (Coombes  et  al.,  2010;  Natural  England,  2009)  and  accessible  via  the  road   network  as  this  is  the  most  likely  connection  between  an  individuals’  residence  and   their   physical   use   of   greenspace.   The   spatial   resolution   of   the   public   greenspace   dataset  was  variable,  with  resolution  as  coarse  as  15  metres  in  some  instances.  To   accommodate   this,   greenspace   within   15   metres   of   the   road   network   was   determined   to   be   usable   and   located   on   the   road   network.   This   usable   subset   of  

20  

Nicholas  Brunsdon   public   greenspaces   comprised   1,499   usable   recreation   spaces   (38,260   hectares).   In   the   absence   of   specific   information   on   the   entranceways   to   greenspaces,   areas   of   usable  greenspace  within  15  metres  of  roads  were  assumed  to  be  entrances;  this  will   likely   overstate   the   accessibility,   as   it   does   not   incorporate   terrain   or   fencing   obstructions.   Without   knowing   how   respondents   decide   which   greenspace   to   visit,   a   typical   approach   is   to   assume   the   nearest   park   and   measure   the   distance   to   that;   however   this   approach   ignores   the   broader   influence   of   multiple   parks   in   their   vicinity.   Measuring   accessibility   through   the   average   distance   to   the   nearest   five   usable   greenspaces   provides   a   greater   indication   of   this   influence,   and   the   mean   distance   to   the   nearest   five   facilities   is   a   common   compromise   that   draws   from   a   slightly  wider  sphere  of  influence  without  being  distorted  by  distant  features  without   influence.   The   access   for   each   meshblock   to   the   nearest   usable   greenspace   was   appraised   from   the   population   weighted   centroid   (PWC)   to   the   nearest   five   usable   greenspaces   (Figure   5),   measured   in   terms   of   the   median   travel   time   in   minutes   along  the  road  network.    

4.1.1  Impervious  surfaces   Private   greenspaces   were   identified   using   a   novel   method   that   takes   impervious   surface   data   to   identify   spaces   that   are   not   greenspace,   and   inferring   that   the   remainder   of   spaces   are   likely   to   be   private   or   public   greenspace   of   some   form.   This   followed   a   similar   theoretical   approach   and   method   as   for   public   greenspace,   calculating   the   proportion   of   a   0-­‐500   metre   and   500-­‐2000   metre   buffer   around   each   meshblock   PWC   that   is   covered   by   impervious   surfaces,   the   inverse   of   which   is   proposed  as  a  measure  of  exposure  to  private  and  public  greenspaces  for  individuals   residing  in  the  meshblock.  Data  illustrating  impervious  surfaces  across  the  Auckland   Region  (excluding  North  Shore  and  Albany)  was  obtained  from  the  Auckland  Council.   Figure   6   shows   the   impervious   surface   layer   illustrating   roads,   building   rooftops,   paths   and   carparks   within   a   0-­‐500   metre   buffer.   This   follows   the   inherent   assumption   that   areas   outside   impervious   surfaces   are   some   form   of   greenspace,   and  that  neighbours  derive  a  therapeutic  landscape  effect  from  the  presence  of  their   neighbours   private   gardens.   Due   to   the   different   spatial   resolutions   of   greenspace  

21  

and  impervious  surfaces  data,  it  was  impractical  to  remove  areas  of  overlap  between   these   layers   without   losing   significant   data.   As   such,   impervious   surface   data   did   overlap   public   greenspaces   in   some   instances,   such   as   sealed   carparks   within   a   recreational  reserve  land  parcel.  However  this  overlap  was  minimal,  with  an  average   7.8%   of   public   greenspaces   covered   by   impervious   surfaces,   up   to   a   maximum   of   15.5%.      

  Figure  5:  -­‐  Map  of  illustrative  meshblock  showing  usable  greenspaces  and  how  their  accessibility  to   the  meshblock  is  appraised  –  by  the  average  travel  time  via  the  road  network  from  the  meshblock   PWC  to  the  nearest  five  usable  greenspaces  

   

22  

Nicholas  Brunsdon  

  Figure  6:  -­‐  Map  of  illustrative  meshblock  showing  impervious  surfaces  (such  as  roads,  buildings  and   footpaths)  within  and  beyond  a  500m  buffer  of  the  meshblock.  It  is  proposed  that  the  inverse  of  the   impervious  surfaces  represent  greenspaces  of  some  form  

 

4.2 Health  analysis   Health  analysis  was  based  upon  responses  to  the  NZHS,  which  is  administered  and   distributed  by  the  Ministry  of  Health  (MoH).  Greenspace  characteristics  derived  for   each  of  the  237  Auckland  region  census  meshblock  in  which  the  2011/12  NZHS   sampled  were  sent  to  the  MoH.  As  these  characteristics  were  calculated  to  several   decimal  places,  they  may  have  served  as  unique  identifiers  for  individual   meshblocks,  and  thus  could  have  potentially  been  used  to  locate  survey  respondents   and  breach  their  anonymity.  This  was  addressed  by  categorising  these  variables  into  

23  

quintiles  prior  to  sending  the  data  to  MoH,  which  are  summarized  in  Table  1   alongside  the  New  Zealand  Index  of  Deprivation  (Salmond,  Crampton,  &  Atkinson,   2007)  as  an  indicator  of  SED,  based  on  the  2006  national  census,  which  was   collapsed  from  deciles  into  quintiles  for  ease  of  interpretation.  MoH  then  attached   the  NZHS  data  of  all  respondents  in  these  Auckland  meshblock  to  meshblock   characteristics  before  meshblock  identifiers  were  removed  to  ensure  the   confidentiality  of  NZHS  respondents,  and  anonymised  data  was  returned  to  this   researcher.  Based  on  the  NZHS  sampling  methodology,  the  sample  of  2,557  adults  is   expected  to  be  representative  of  a  population  of  810,159  adults.     Table  1:  -­‐  Summary  information  for  area-­‐level  variables  of  respondents  

  n   %   NZ  Index  of  Deprivation   1  -­‐  Low   373   14.6   2   501   19.6   3   382   14.9   4   538   21.0   5  -­‐  High   763   39.8   Usable  greenspace  accessibility   1  -­‐  Low     96   3.8   2   384   15.0   3   542   21.2     4   1042   40.8   5  -­‐  High     493   19.3   Proportion  of  all  public  greenspace  within  500m  buffer  (usable  and  non-­‐usable)   1  -­‐  Low     675   26.4   2   663   25.9   3   622   24.3   4   368   14.4   5  -­‐  High     229   9.0   Proportion  of  all  public  greenspace  within  2000m  buffer  (usable  and  non-­‐usable)   1  -­‐  Low     765   29.9   2   808   31.6   3   555   21.7   4     262   10.2   5  -­‐  High     167   6.5    

24  

Nicholas  Brunsdon   Table  1  (cont.)  

  n   %   Proportion  of  all  greenspace  (from  impervious  surfaces)  within  0-­‐500m  buffer     1  -­‐  Low   487   19.0   2   499   19.5   3   545   21.3   4   474   18.5   5  -­‐  High     552   21.6   Proportion  of  all  greenspace  (from  impervious  surfaces)  within  500-­‐2000m  buffer   1  -­‐  Low   475   18.6   2   522   20.4   3   536   21.0   4   527   20.6   5  -­‐  High   497   19.4    

A  series  of  processing  steps  was  taken  with  the  raw  survey  data  to  enable  regression   analysis.     Continuous   variables   were   recoded   into   categorical   form   to   meet   the   requirements  of  logistic  regression,  and  all  variables  were  recoded  such  that  values   of   0   represented   the   reference   category   –   for   example   the   depressive   feelings   question  was  recoded  such  that  value  0  represented  no  symptoms  of  depression  and   values   1   to   4   represented   increasing   frequency   of   such   symptoms.   Survey   responses   indicating  the  average  time  spent  exercising  in  hours  and  minutes  and  the  frequency   of   such   exercise   in   the   past   seven   days   were   combined   to   estimate   the   total   time   spent  exercising  under  each  measure  in  the  past  seven  days,  and  were  then  recoded   into   tertiles   as   this   was   best   suited   to   their   distribution.   The   recoded   variables   of   individual  characteristics  of  NZHS  respondents  in  the  study  area  are  summarised  in   Table  2.    

25  

Table  2:  -­‐  Summary  information  for  2,557  adult  respondents  to  the  2011/12  NZHS  within  the   Auckland  study  area,  representing  a  population  of  810,159.  

  n   %   Sex   Female   1537   60.1   Male   1020   39.9   Age  (years)   15  –  24   341   13.3   25  –  34   522   20.4   35  –  44   556   21.7   45  –  54   407   15.9   55  –  64   331   12.9   Over  65   400   15.6   Depressive  feelings  in  past  4  weeks   Never   2048   80.2   A  little  of  the  time   295   11.5   Some  of  the  time   154   6.0   Most  of  the  time   46   1.8   All  of  the  time   12   0.5   Nervous  feelings  in  past  4  weeks   Never   1805   70.7   A  little  of  the  time   458   17.9   Some  of  the  time   223   8.7   Most  of  the  time   52   2.0   All  of  the  time     15   0.5   Ever  been  diagnosed  by  a  doctor  with  depression   No   2245   87.9   Yes   308   12.1     Ever  been  diagnosed  by  a  doctor  with  an  anxiety  doctor   No   2420   94.8   Yes   133   5.2   Time  spent  undertaking  moderate  physical  activity  in  past  week  (minutes)   0   1023   40.0   10  –  60   914   35.7   61  –  1020   620   24.2   Time  spent  undertaking  vigorous  physical  activity  in  past  week  (minutes)   0   1793   70.1   10  –  55   271   10.6   60  –  1200   493   19.3   Time  spent  brisk  walking  in  past  week  (minutes)   0   942   36.8   10  –  30   809   31.6   31  –  1200   806   31.5    

26  

Nicholas  Brunsdon   Table  2  (cont.)  

  n   All  time  spent  exercising  in  past  week  (minutes)   0   404   10  –  90   1045   91  –  1810   1108    

%   15.8   40.9   43.3  

An   initial   correlative   analysis   was   applied   to   the   area   level   variables   to   yield   correlation   coefficients   with   p-­‐values   for   meshblock   greenspace   measures   and   SED   (Table   3).   The   primary   analysis   was   undertaken   with   logistic   regressions   in   the   statistical   software   package   STATA   using   different   measures   of   time   spent   exercising   and  mental  illness  symptoms  as  the  dependent  variables.  This  produced  odds  ratios   that   were   relative   to   the   reference   category   alongside   95%   confidence   intervals   -­‐   intervals   that   included   the   value   1.00   are   statistically   insignificant.   For   each   of   the   dependent   variables   (time   spent   undertaking   various   forms   of   physical   activity   and   mental  illness  symptoms)  these  were  regressed  separately  against  each  explanatory   variable   through   a   logistic   regression   (Tables   4   and   6).   This   provided   a   basic   understanding   of   how   these   variables   related   to   the   dependent   variables   before   constructing   a   complete   model   that   used   the   explanatory   variables   simultaneously   (Tables  5,  7,  8  and  9).  A  model  for  each  dependent  variable  was  built  up  iteratively   by   adding   demographic   and   neighbourhood   SED   explanatory   variables,   seeking   the   model   with   the   greatest   explanatory   power   as   given   by   the   F   statistic.   From   the   selected  model  for  each  dependent  variable,  greenspace  variables  were  then  added   to   assess   the   explanatory   power   of   neighbourhood   greenspace.   The   NZ   Index   of   Deprivation,  of  which  average  meshblock  income  is  a  component,  was  found  to  be   more   significant   than   individual   income   as   a   control   variable,   as   not   all   NZHS   respondents   reported   their   income.   The   NZHS   sampling   methodology   aims   to   oversample   minority   groups   to   improve   analysis   of   small   populations;   as   a   result   the   raw  survey  data  is  not  representative  of  the  overall  population  (Ministry  of  Health,   2012).   To   overcome   this,   MoH   provided   weightings   that   were   incorporated   into   regression   analysis   to   modify   variances,   and   therefore   confidence   intervals,   such   that  the  actual  population  distribution  is  reflected.  

27  

5. Results   The   correlation   matrix   (Table   3)   indicates   the   covariance   of   each   of   the   area   level   variables–  greenspace  and  deprivation.  The  correlation  between  usable  greenspace   measures   and   all   public   greenspace   within   0-­‐500   metres   was   a   weak   0.3477,   and   diminished  with  the  500-­‐2000  metre  buffer  of  all  public  greenspace  with  0.1852.  All   public   greenspace   and   all   greenspace   (derived   from   impervious   surfaces)   were   moderately   correlated   with   a   coefficient   of   0.5890   within   the   0-­‐500   metre   buffers   and   0.5453   in   the   500-­‐2000   metre   buffers.   All   public   greenspace   within   0-­‐500   metre   and   500-­‐2000   metre   were   moderately   correlated   with   a   coefficient   of   0.4720;   similarly  the  two  buffer  areas  for  all  greenspace  (derived  from  impervious  surfaces)   were   correlated   by   a   factor   of   0.6259.   Neighbourhood   access   or   exposure   to   greenspace   was   weakly   correlated   with   socioeconomic   deprivation,   with   a   coefficient  of  0.1354  between  deprivation  and  usable  greenspace  access  and  weaker   correlations  for  the  all  and  all-­‐public  greenspace  measures.     Table  3:  -­‐  Correlation  matrix  of  area  level  greenspace  and  deprivation  variables,  indicating   correlation  coefficient  (p-­‐value).  

28  

Usable  public   greenspace   access  

All  public   greenspace   0-­‐500m  

All  public   greenspace   500-­‐2000m  

All  greenspace     0-­‐500m  

All  greenspace   500-­‐2000m  

NZ  Index  of   Deprivation   Usable  public   greenspace  access   All  public   greenspace   0-­‐500m   All  public   greenspace   500-­‐2000m   All  greenspace   0-­‐500m   All  greenspace   500-­‐2000m  

NZ  Index  of   Deprivation  

 

1.000  

 

 

 

 

 

0.1354   (0.0000)   -­‐0.0154   (0.4374)  

1.000  

 

 

 

 

0.3477   (0.0000)  

1.000  

 

 

 

0.0661   (0.0008)  

0.1852   (0.0000)  

0.4720   (0.0000)  

1.000  

 

 

-­‐0.0495   (-­‐.0123)   -­‐0.0023   (0.9085)  

-­‐0.0703   (0.0004)   -­‐0.1261   (0.0000)  

0.5890   (0.0000)   0.3102   (0.0000)  

0.4020   (0.0000)   0.5453   (0.0000)  

1.000  

 

0.6259   (0.0000)  

1.000  

Nicholas  Brunsdon   An  initial  regression  of  time  spent  undertaking  various  forms  of  physical  activity  with   age,   sex   and   neighbourhood   access   to   usable   greenspace   is   shown   in   Table   4,   indicating  the  underlying  relationship  with  the  physical  activity  before  a  model  was   built   to   control   for   each   of   these   factors   together.   Table   4   indicates   that   older   age   groups  generally  undertake  increasingly  less  physical  activity  -­‐  individuals  over  65  are   78%   less   likely   to   undertake   vigorous   physical   activity   than   those   aged   15   to   24.   Moderate  physical  activity  runs  counter  to  this,  with  individuals’  aged  25  –  64  equally   or   more   likely   to   than   the   reference   group.   Similarly,   males   are   more   likely   to   undertake   all   forms   of   physical   activity   than   females   –   from   80%   more   likely   to   undertake  vigorous  physical  activity  to  9%  more  likely  to  undertake  brisk  walking.     Table  4:  -­‐  Logistic  regression  of  the  relationship  between  time  spent  exercising  (dependent  variable)   and  individual  and  area  level  variables,  each  considered  separately.  Results  are  given  as  odds  ratios   (95%  confidence  intervals)  relative  to  the  reference  category  (odds  ratio=1.00).  

 

Brisk  walking  

Age   15  –  24   1.00   25  –  34   0.94  (0.61  –1.46)   35  –  44   0.82  (0.54  –1.23)   45  –  54   0.88  (0.58  –  1.34)   55  –  64   0.91  (0.59  –  1.41)   Over  65   0.44  (0.28  –  0.69)   Sex   Female   1.00   Male   1.09  (0.91  –1.32)   NZ  index  of  deprivation   1  -­‐  Low   1.00   2   0.71  (0.47  –  1.07)   3   0.63  (0.35  –  1.12)   4   0.54  (0.37  –  0.81)   5  -­‐  High   0.58  (0.40  –  0.84)   Usable  greenspace  accessibility   1  -­‐  Low   1.00   2   1.48  (0.59  –3.72)   3   1.56  (0.73  –3.33)   4   1.39  (0.62  –  3.12)   5  -­‐  High   1.57  (0.70  –3.50)    

Moderate   physical  activity  

Vigorous  physical   activity  

All  physical   activity  

1.00   1.06  (0.72  –  1.56)   1.00  (0.69  –  1.44)   1.07  (0.71  –  1.62)   1.13  (0.70  –  1.83)   0.61  (0.39  –  0.95)  

1.00   0.77  (0.53  –  1.11)   0.67  (0.46  –  0.98)   0.49  (0.31  –  0.78)   0.51  (0.33  –  0.78)   0.22  (0.14  –  0.36)  

1.00   1.19  (0.76  –  1.81)   1.10  (0.77  –  1.59)   1.18  (0.82  –  1.69)   1.10  (0.69  –  1.74)   0.50  (0.33  –0.77)  

1.00   1.39  (1.13  –  1.72)  

1.00   1.80  (1.41  –  2.30)  

1.00   1.79  (1.44  –  2.24)  

1.00   1.06  (0.69  –  1.62)   1.13  (0.68  –  1.88)   0.79  (0.54  –  1.17)   0.61  (0.42  –  0.86)  

1.00   0.61  (0.38  –  0.98)   0.66  (0.44  –  0.99)   0.52  (0.38  –  0.71)   0.52  (0.37  –  0.74)  

1.00   0.76  (0.50  –  1.15)   0.52  (0.35  –  0.76)   0.48  (0.36  –  0.64)   0.48  (0.32  –0.67)  

1.00   0.49  (0.19  –  1.23)   0.53  (0.22  –  1.24)   0.56  (0.24  –  1.31)   0.55  (0.24  –  1.28)  

1.00   0.44  (0.17  –  1.19)   0.59  (0.25  –  1.40)   0.56  (0.22  –  1.41)   0.33  (0.13  –  0.85)  

1.00   0.83  (0.30  –  2.32)   1.08  (0.40  –  2.90)   0.86  (0.31  –  2.36)   0.89  (0.32  –2.52)  

29  

The   complete   regression   models   of   Table   5   outline   the   relationship   between   neighbourhood  access  to  usable  greenspace  and  physical  activity,  with  age,  sex  and   neighbourhood   deprivation   controlled   for   when   significant.   Neighbourhood   SED   was   a  significant  descriptor  of  all  forms  of  physical  activity,  with  a  clear  pattern  of  lower   levels  of  physical  activity  in  more  deprived  areas.  A  significant  relationship  was  found   between   vigorous   physical   activity   and   access   to   usable   greenspace   in   the   quintile   with   highest   access,   however   this   was   not   significant   across   other   quintiles.   Individuals   residing   in   neighbourhoods   with   the   highest   accessibility   to   usable   greenspace   61%   less   likely   (odds   ratio   0.39)   to   undertake   vigorous   physical   activity   than   those   in   the   lowest   quintile   for   accessibility.   No   significant   relationship   or   discernible   trend   was   found   for   brisk   walking,   moderate   or   all   forms   of   physical   activity.       Table  5:  -­‐  Logistic  regression  of  the  relationship  between  time  spent  exercising  (dependent  variable)   and  area  level  usable  greenspace  access,  controlling  for  age,  sex  and  area  deprivation.  Results  are   given  as  odds  ratios  (95%  confidence  intervals)  relative  to  the  reference  category  (odds  ratio=1.00)  

 

Brisk  walking  

Usable  greenspace  accessibility   1  -­‐  Low   1.00   2   1.56  (0.56  –  4.37)   3   1.66  (0.68  –  4.07)   4   1.50  (0.57  –  3.93)   5  -­‐  High     1.80  (0.70  –  4.62)    

Moderate   physical  activity  

Vigorous  physical   activity  

All  physical   activity  

1.00   0.51  (0.21  –  1.27)   0.54  (0.23  –  1.31)   0.60  (0.25  –  1.44)   0.62  (0.26  –  1.48)  

1.00   0.50  (0.22  –  1.12)   0.65  (0.31  –  1.37)   0.66  (0.30  –  1.47)   0.39  (0.17  –  0.87)  

1.00   0.93  (0.31  –  2.74)   1.18  (0.41  –  3.42)   0.98  (0.33  –  2.85)   1.09  (0.36  –  3.27)  

 An   initial   regression   of   depressive   feelings   and   nervous   feelings   with   prior   doctor   diagnosis,  age,  sex  and  various  measures  of  neighbourhood  greenspace  is  shown  in   Table   6.   A   strong   correlation   between   depressive   feelings   and   previous   doctor   diagnosis   for   depression   is   evident,   as   is   nervous   feelings   and   previous   doctor   diagnosis   for   anxiety   disorders,   with   higher   prevalence   associated   with   prior   diagnosis.  Age,  sex  and  neighbourhood  SED  added  insignificant  explanatory  power  to   the   mental   illness   models,   however   did   present   some   statistically   insignificant   trends.   A   higher   prevalence   of   depressive   feelings   was   found   in   middle   age   groups   and   in   woman,   and   in   the   most   deprived   neighbourhoods.   Nervous   feelings   were  

30  

Nicholas  Brunsdon   more   prevalent   in   younger   age   groups   and   in   males,   and   had   no   discernible   relationship   with   neighbourhood   SED.   No   discernible   pattern   was   found   between   usable   greenspace,   all   public   greenspace   or   all   greenspace   (from   impervious   surfaces)  and  mental  illness  when  considered  alone.     Table  6:  -­‐  Logistic  regression  of  the  relationship  between  mental  illness  symptoms  (dependent   variable)  and  individual  and  area  level  variables,  each  considered  separately.  Results  are  given  as   odds  ratios  (95%  confidence  intervals)  relative  to  the  reference  category  (odds  ratio=1.00).  

  Depressive  feelings   Age     15  –  24   1.00   25  –  34   1.21  (0.76  –  1.93)   35  –  44   1.19  (0.80  –  1.78)   45  –  54   1.20  (0.74  –  1.94)   55  –  64   0.90  (0.56  –  1.46)   Over  65   1.04  (0.64  –  1.69)   Doctor  diagnosis   No   1.00   Yes   5.10  (3.48  –  7.46)   Sex   Female   1.00   Male   0.77  (0.60  –  0.99)   NZ  index  of  deprivation   1  -­‐  Low   1.00   2   1.94  (1.15  –  3.29)   3   1.44  (0.85  –  2.43)   4   1.82  (1.05  –  3.15)   5  -­‐  High   2.04  (1.18  –  3.52)   Usable  greenspace  accessibility   1  -­‐  Low   1.00   2   0.76  (0.45  –  1.30)   3   0.71  (0.38  –  1.33)   4   1.01  (0.56  –  1.83)   5  -­‐  High     0.98  (0.56  –  1.72)   Proportion  of  all  public  greenspace  within  0-­‐500m  buffer   1  -­‐  Low     1.00   2   1.03  (0.70  –  1.51)   3   0.73  (0.47  –  1.13)   4   0.78  (0.59  –  1.02)   5  -­‐  High     1.02  (0.58  –  1.79)  

Nervous  feelings   1.00   0.85  (0.57  –  1.29)   0.59  (0.43  –  0.82)   0.62  (0.43  –  0.89)   0.34  (0.21  –  0.54)   0.23  (0.15  –  0.37)   1.00   3.81  (2.43  –  5.97)   1.00   1.07  (0.86  –  1.33)   1.00   1.41  (0.95  –  2.10)   1.06  (0.69  –  1.64)   1.19  (0.71  –  2.00)   1.10  (0.70  –  1.72)   1.00   0.80  (0.50  –  1.21)   0.76  (0.40  –  1.41)   1.00  (0.62  –  1.61)   2.23  (0.80  –  6.24)   1.00   0.45  (0.17  –  1.17)   0.34  (0.13  –  0.87)   0.35  (0.13  –  0.94)   0.45  (0.16  –  1.25)  

31  

Table  6  (cont.)  

    Depressive  feelings   Nervous  feelings   Proportion  of  all  public  greenspace  within  500-­‐2000m  buffer   1  -­‐  Low     1.00   1.00   2   1.02  (0.64  –  1.64)   0.84  (0.56  –  1.24)   3   1.15  (0.74  –  1.80)   0.86  (0.55  –  1.33)   4   0.96  (0.57  –  1.61)   0.75  (0.38  –  1.47)   5  -­‐  High   0.99  (0.37  –  2.64)   0.52  (0.23  –  1.19)   Proportion  of  all  greenspace  (impervious  surfaces)  within  0-­‐500m  buffer   1  -­‐  Low     1.00   1.00   2   1.38  (0.84  –  2.29)   1.16  (0.72  –  1.88)   3   1.15  (0.52  –  2.55)   0.71  (0.39  –  1.27)   4   1.13  (0.66  –  1.94)   0.73  (0.42  –  1.28)   5  -­‐  High     1.12  (0.70  –  1.81)   1.14  (0.68  –  1.90)   Proportion  of  all  greenspace  (impervious  surfaces)  within  500-­‐2000m  buffer   1  -­‐  Low     1.00   1.00   2   1.05  (0.55  –  2.00)   0.78  (0.47  –  1.30)   3   0.69  (0.38  –  1.25)   0.48  (0.28  –  0.81)   4   0.91  (0.57  –  1.47)   0.75  (0.45  –  1.28)   5  -­‐  High     0.86  (0.55  –  1.33)   0.74  (0.44  –  1.27)     Table   7   outlines   the   regression   of   mental   illness   symptoms   with   neighbourhood   access  to  usable  greenspace  while  controlling  for  doctor  diagnosis.  Individuals  living   in   the   third   and   fourth   quintile   for   usable   greenspace   access   were   found   to   have   significantly   lower   incidence   of   nervous   feelings,   66%   less   likely   (odds   ratio   0.34)   than   those   residing   in   the   lowest   quintile   for   access;   however   there   was   no   significant  relationship  overall.  In  the  building  of  the  model,  individual  or  area  level   variables  established  in  other  studies  did  not  improve  the  explanatory  power  of  the   model  and  were  excluded.       Table  7:  -­‐  Logistic  regression  of  the  relationship  between  mental  illness  symptoms  (dependent   variable)  and  usable  greenspace,  controlling  for  prior  doctor  diagnosis.  Results  are  given  as  odds   ratios  (95%  confidence  intervals)  relative  to  the  reference  category  (odds  ratio=1.00).  

  Usable  greenspace  accessibility   1  -­‐  Low     2   3   4   5  -­‐  High    

32  

Depressive  feelings  

Nervous  feelings  

1.00   1.07  (0.72  –  1.58)   0.82  (0.51  –  1.32)   0.82  (0.57  –  1.18)   1.21  (0.64  –  2.29)  

1.00   0.43  (0.17  –  1.09)   0.34  (0.13  –  0.84)   0.34  (0.13  –  0.91)   0.46  (0.16  –  1.28)  

Nicholas  Brunsdon    

Table   8   outlines   the   regression   model   of   mental   illness   symptoms   and   all   public   greenspace   as   measured   in   0-­‐500   metre   and   500-­‐2000   metre   buffer   areas.   This   relationship  was  found  to  be  insignificant  for  depressive  symptoms  with  odds  ratios   higher   and   lower   than   1.00.   Nervous   feelings   were   insignificantly   correlated   with   neighbourhood   greenspace,   but   it   is   noteworthy   that   odds   ratios   for   greenspace   within   the   0-­‐500   metre   buffers   were   primarily   above   1.00,   and   odds   ratios   for   the   500-­‐2000   metre   buffers   were   consistently   below   1.00.   This   could   suggest   that   greenspace  within  the  different  buffer  areas  has  a  divergent  effect  on  mental  health,   however   no   conclusion   can   be   drawn   due   to   the   insignificance   of   these   results.   Prior   doctor  diagnosis  was  a  significant  predictor  of  present  mental  illness  symptoms.       Table  8:  -­‐  Logistic  regression  of  the  relationship  between  mental  illness  symptoms  (dependent   variable)  and  public  greenspace,  controlling  for  prior  doctor  diagnosis.  Results  are  given  as  odds   ratios  (95%  confidence  intervals)  relative  to  the  reference  category  (odds  ratio=1.00).  

  Depressive  feelings   Proportion  of  all  public  greenspace  within  0-­‐500m  buffer   1  -­‐  Low   1.00   2   0.84  (0.55  –  1.28)   3   1.07  (0.74  –  1.56)   4   1.06  (0.43  –  2.59)   5  -­‐  High   0.93  (0.39  –  2.22)   Proportion  of  all  public  greenspace  within  500-­‐2000m  buffer   1  -­‐  Low   1.00   2   1.08  (0.65  –  1.79)   3   1.17  (0.72  –  1.89)   4   1.06  (0.64  –  1.75)   5  -­‐  High   1.00  (0.35  –  2.85)    

Nervous  feelings   1.00   1.02  (0.65  –  1.60)   1.07  (0.67  –  1.70)   0.68  (0.32  –  1.42)   1.40  (0.68  –  2.89)   1.00   0.87  (0.58  –  1.29)   0.87  (0.54  –  1.39)   0.75  (0.38  –  1.48)   0.49  (0.21  –  1.15)  

Table   9   outlines   the   regression   model   of   mental   illness   symptoms   and   the   all   greenspace   measure   derived   from   impervious   surfaces   within   a   0-­‐500   metre   and   500-­‐2000   metre   buffer.   This   greenspace   measure   was   statistically   insignificant   predictor   of   depressive   symptoms.   Nervous   feelings   were   significantly   correlated   with   the   greenspace   measure   overall,   but   became   insignificant   when   analysed   separately   by   quintile   (Table   9).   The   statistically   insignificant   trend   throughout   these   quintiles   was   consistent   with   the   statistically   significant   result   for   greenspace   overall  

33  

(with   greenspace   quintiles   collapsed).   The   only   statistically   significant   results   for   nervous   feelings   and   all   greenspace   were   found   in   the   quintile   for   the   highest   proportion   of   greenspace.   Individuals   in   the   highest   quintile   for   greenspace   within   the   0-­‐500   metre   buffer   were   1.62   times   more   likely   to   have   experienced   nervous   feelings  in  the  past  four  weeks  than  those  in  the  lowest  quintile;  inversely  those  in   the  highest  quintile  for  greenspace  within  the  500-­‐2000  metre  buffer  had  only  0.55   time   the   odds   of   experiencing   nervous   feelings   than   those   in   the   lowest   quintile.   Although   no   other   quintiles   reported   a   statistically   significant   result   and   thus   are   inconclusive,   it   is   worthwhile   to   note   that   the   insignificant   quintiles   suggested   a   relationship   consistent   with   the   significant   quintiles   -­‐   a   high   proportion   of   greenspace   within   the   0-­‐500   metre   buffer   is   associated   with   higher   prevalence   of   nervous  feelings,  but  a  high  proportion  within  the  0-­‐2000  metre  buffer  is  associated   with   a   lower   prevalence   of   nervous   feelings.   Prior   doctor   diagnosis   remained   a   significant  and  strong  predictor  of  current  mental  illness  symptoms.     Table  9:  -­‐  Logistic  regression  of  the  relationship  between  mental  illness  symptoms  (dependent   variable)  and  all  greenspaces  (derived  from  impervious  surfaces),  controlling  for  doctor  diagnosis.   Results  are  given  as  odds  ratios  (95%  confidence  intervals)  relative  to  the  reference  category  (odds   ratio=1.00).  

  Depressive  feelings   Nervous  feelings   Proportion  of  all  greenspace  (from  impervious  surfaces)  within  0-­‐500m  buffer   1  -­‐  Low     1.00   1.00   2   1.64  (1.00  –  2.67)   1.37  (0.83  –  2.25)   3   1.49  (0.81  –  2.73)   0.92  (0.57  –  1.48)   4   1.57  (0.90  –  2.75)   1.01  (0.57  –  1.80)   5  -­‐  High   1.34  (0.74  –  2.46)   1.62  (1.04  –  2.55)   Proportion  of  all  greenspace  (from  impervious  surfaces)  within  500-­‐2000m  buffer   1  -­‐  Low     1.00   1.00   2   0.99  (0.50  –  1.95)   0.80  (0.50  –  1.29)   3   0.60  (0.35  –  1.05)   0.51  (0.30  –  0.88)   4   0.79  (0.49  –  1.29)   0.75  (0.45  –  1.25)   5  -­‐  High   0.71  (0.42  –  1.22)   0.55  (0.32  –  0.94)     As  the  study  focused  on  short  term  health  indicators,  regression  analysis  with  doctor   diagnosed   depression   and   anxiety   disorders   was   beyond   the   research   questions.   However,   this   analysis   was   conducted   to   compare   the   short-­‐term   indicators   with  

34  

Nicholas  Brunsdon   doctor   diagnosis   and   to   validate   the   study   relative   to   existing   works   that   tend   to   use   doctor   diagnosed   depression.   Models   of   usable   greenspace,   all   public   greenspaces   and  all  greenspaces  are  applied  to  doctor  diagnosed  depression  and  anxiety  disorder   in   Table   A1   in   Appendix   A.   In   these   models,   sex   was   a   significant   predictor   of   mental   illness,   and   the   insignificance   of   greenspace   measures   in   predicting   mental   illness   was  replicated.        

35  

6. Discussion   6.1 Greenspace  Indicators   The  three  different  measures  of  neighbourhood  greenspace  overlapped  each  other   by   definition.   Because   of   this,   it   was   important   to   understand   the   relationship   between  these  indicators  before  comparing  their  relationship  with  health  indicators.   A   degree   of   correlation   is   expected   due   to   the   definitional   overlap,   however   too   strong   a   correlation   would   suggest   that   indicators   are   too   similar   and   thus   not   adding   value.   This   was   especially   important   for   the   different   buffer   areas,   as   these   were   analysed   together   in   the   same   regression   model,   and   a   strong   correlation   between   variables   (multicolinearity)   can   distort   the   estimation   of   odds   ratios.   The   correlation   matrix   (Table   3)   describes   these   relationships,   none   of   which   were   particularly   strong,   thus   the   problem   of   multicolinearity   could   be   discounted.   Usable   greenspace   is   a   derived   from   all   public   greenspace,   which   is   a   subset   of   all   greenspace,   with   lower   values   representing   lower   travel   times,   better   access,   and   therefore   more   greenspace.   Accordingly   there   was   a   weak   correlation   between   usable   greenspace   and   all   public   greenspaces   within   0-­‐500   metre   buffer   (0.3477),   and   an   even   weaker   in   the   500-­‐2000   metre   buffer   (0.1852),   reflecting   that   as   the   usable  greenspace  measure  is  based  on  the  nearest  five  usable  public  greenspaces,   and   these   are   more   likely   to   be   located   within   the   area   of   the   0-­‐500   metre   buffer   rather   than   the   500-­‐2000   metre   buffer.   Similarly,   the   impervious   surfaces   measure   will  include  both  private  greenspace  and  the  public  greenspace  areas  included  in  the   public   greenspace   measure,   thus   impervious   surfaces   and   greenspace   are   strongly   related:  0.5890  in  the  0-­‐500  metre  buffer  and  0.5453  in  the  500-­‐2000  metre  buffer.   The   fact   that   the   correlation   is   not   perfect   affirms   that   the   impervious   surfaces   layer   is   measuring   the   distribution   of   not   just   public   greenspace,   but     also   private   gardens   not   captured   by   the   traditional   public   greenspace   measure.   Conjoining   neighbourhoods   are   likely   share   similar   characteristics,   and   thus   the   0-­‐500   metre   and  500-­‐2000  metre  buffers,  while  measuring  distinct  areas,  are  strongly  correlated   reflecting  the  similarity  of  adjacent  areas.  This  is  shown  in  the  correlation  of  0.4720   between  the  public  greenspace  buffers  and  0.6259  in  the  impervious  surface  buffers.  

36  

Nicholas  Brunsdon      

6.2 Socioeconomic  Deprivation   The  correlation  between  neighbourhood  greenspace  measures  and  SED  can  provide   an   insight   to   the   environmental   justice   implications   for   greenspace,   as   an   unequal   distribution   of   greenspace   could   contribute   towards   health   inequalities.   However,   in   this   case   the   correlations   were   weak   and   or   statistically   insignificant,   and   thus   no   inferences   can   be   made   about   the   access   or   exposure   of   greenspace   relative   to   socioeconomic   status.   Universally   high   access   to   greenspace   in   New   Zealand   (Witten   et   al.,   2008)   may   serve   to   obfuscate   this   relationship,   however   Field,   Witten,   Robinson,   &   Pledger   (2004)   did   note   higher   access   to   community   facilities   in   more   deprived  neighbourhoods  in  the  North  Shore  and  Waitakere  areas  of  Auckland  City.   The  socioeconomic  gradient  of  greenspace  exposure  could  be  analysed  further  by  a   measure   of   greenspace   quality,   and   measuring   how   this   varies   in   relation   to   neighbourhood  SED.      

6.3 Physical  Activity   A   key   premise   of   this   study   was   the   use   of   short-­‐term   health   indicators   as   an   alternative   to   the   prevailing   use   of   longer   term   indicators,   with   time   spent   undertaking  physical  activity  as  a  short  term  measure  of  physical  health.  Regression   models   including   age   or   sex   provided   results   consistent   with   prior   studies,   indicating   that  physical  activity  levels  are  highest  amongst  males  and  young  age  groups  (Trost   et   al.,   2002).   Statistically   significant   lower   levels   of   physical   activity   by   individuals   residing   in   socioeconomically   deprived   areas   are   also   consistent   with   prior   studies,   however   with   no   socioeconomic   gradient   for   greenspace   access,   it   is   difficult   to   explain  (Giles-­‐Corti  &  Donovan,  2002).  Contextual  differences  in  physical  activity  may   warrant   future   investigation,   perhaps   incorporating   the   distribution   of   a   broader   range   of   physical   activity   facilities   than   greenspace   generally,   such   as   gyms.   The   relationship   between   usable   greenspace   and   physical   activity   was   insignificant   and   indeterminate   for   brisk   walking,   moderate   and   all   physical   activity;   however   a   significant   and   strong   relationship   between   usable   greenspace   and   vigorous   physical  

37  

activity  is  difficult  to  reconcile.  It  is  difficult  to  assess  the  utility  of  physical  activity  as   a   short   term   physical   health   from   these   results,   given   that   its   insignificance   with   usable   greenspace   is   consistent   with   existing   studies   in   New   Zealand   regressing   long   term   physical   health   (e.g.   cardiovascular   disease)   with   greenspace.   Future   studies   could   investigate   the   use   of   physical   activity   further,   potentially   relating   it   to   self-­‐ reported  health  status  or  against  long-­‐term  physical  health  measures.      

6.4 Mental  Health   The   use   of   short   term   mental   health   measures   as   opposed   to   formally   diagnosed   mental   health   disorders   was   a   key   premise   of   this   study,   and   doctor   diagnosed   mental   disorders   were   used   as   a   control   variable   taking   into   account   the   persistence   of  mental  illnesses  (Mann,  2005).  It  was  expected  that  prior  diagnosis  would  increase   the   odds   of   experiencing   mental   disorder   symptoms,   and   this   was   affirmed   with   prior  doctor  diagnosis  of  depression  and  anxiety  disorders  as  a  significant  and  strong   explanatory   variable   in   all   models   of   depressive   and   nervous   feelings   respectively.   The   correlation   for   depression   was   much   stronger   than   with   nervous   feelings,   reflecting   that   nervous   feelings   are   only   a   component   of   the   suite   of   anxiety   disorders.   For   completeness   and   comparison   with   existing   studies   that   use   clinical   mental  illness  instead  of  short  term  indicators,  Table  A1  in  Appendix  A  indicates  that   the  greenspace  indicators  derived  in  this  study  are  also  generally  poor  indicators  of   clinical   depression   and   anxiety   disorders.   Some   established   demographic   trends   were   replicated,   with   a   statistically   significant   higher   prevalence   of   depressive   symptoms  amongst  those  in  residing  in  more  deprived  neighbourhoods  (Kling  et  al.,   2007;  Lofors  et  al.,  2006;  Mair  et  al.,  2008;  Stafford  &  Marmot,  2003),  and  amongst   women   (Mirowsky,   1996).   Anxiety   disorders   tend   not   to   have   discernable   aged   trends   due   to   the   heterogeneous   group   of   sub-­‐conditions   (Anthony   &   Stein,   2008;   Crawford   &   Henry,   2003).   However,   a   clear   trend   of   lower   prevalence   of   nervous   feelings  amongst  older  age  groups  was  found,  reflecting  that  nervous  feelings  are  a   specific   sub-­‐condition   of   anxiety   disorders   and   may   have   a   more   distinct   demographic   trend   than   the   group   as   a   whole.   None   of   the   greenspace   measures   had  a  significant  explanatory  power  for  the  prevalence  of  depressive  symptoms,  with   38  

Nicholas  Brunsdon   prior   diagnosis   for   depression   controlled   for.   This   result   is   consistent   with   existing   studies  of  public  greenspace  and  depression  in  New  Zealand,  and  the  replication  of   this  with  the  more  inclusive  greenspace  measure  derived  from  impervious  surfaces   provides   further   evidence   that   there   may   be   no   connection   in   the   New   Zealand   context  (Richardson  et  al.,  2013;  Richardson  et  al.,  2010;  Witten  et  al.,  2008).  It  had   been  expected  that  usable  greenspace  would  explain  variations  in  not  only  exercise   levels,   but   depressive   symptoms,   as   physical   activity   is   established   as   an   effective   treatment  for  depression  (Babyak  et  al.,  2000),  however  both  were  insignificant.      

6.5 Future  Research   The   usable   greenspace   and   all   public   greenspace   measures   had   no   significant   explanatory   power   for   the   prevalence   of   nervous   feelings.   The   measure   of   all   greenspace   derived   from   impervious   surfaces   had   a   significant   explanatory   power   for  the  prevalence  of  nervous  feelings,  however  with  in  opposite  directions  for  the   two  buffer  areas.  This  suggested  that  greater  areas  of  greenspace  within  500  metres   of   an   individual’s   meshblock   reduced   their   odds   of   experiencing   nervous   feelings,   but   greenspace   within   500-­‐2000   metres   increased   these   odds.   It   is   difficult   to   reconcile  the  implications  of  this  for  mental  health,  and  this  points  towards  the  need   for   comprehensive   sensitivity   analysis   of   the   buffer   size,   as   the   relationship   is   clearly   sensitive   to   the   size   of   the   buffer.   This   is   a   significant   area   warranting   further   investigation,   both   for   the   impervious   surfaces   measure   and   the   traditional   public   greenspace   measure.   Existing   studies   have   used   buffer   sizes   ranging   from   500   metres  to  3000  metres,  so  finer  analysis  within  and  even  beyond  these  bounds  may   be  insightful.  This  simple  proximity-­‐based  analysis  does  not  take  into  consideration   the   visualisation   of   greenspace   and   the   influence   of   terrain   and   the   built   environment   upon   it,   so   developments   in   the   field   of   visualisation   may   prove   useful.   A   unique   characteristic   of   the   New   Zealand   urban   environment   is   the   ubiquity   of   bluespace,   however   no   research   has   quantified   the   salutogenic   contribution   of   bluespace   in   New   Zealand   to   date   (Richardson   et   al.,   2010);   investigation   of   the   association  between  health  and  the  visualisation  of  greenspace  and  bluespace  may   prove   worthwhile.   The   health   benefits   of   private   gardens   have   been   established   39  

qualitatively  (Stigsdotter  &  Grahn,  2004;  Unruh,  2002),  thus  further  development  of   greenspace   measures   including   public   and   private   greenspace   separately   is   warranted.   The   impervious   surfaces   method   could   be   developed   further   with   the   addition   of   residential   property   data   to   enable   identification   of   the   subset   of   all   greenspaces  that  are  private  gardens,  and  thus  build  a  quantitative  understanding  of   the  health  benefits  associated  with  private  gardens.       The   usable   greenspace   measure   was   based   primarily   on   the   access   and   size   of   greenspaces,   rather   than   quality   or   facilities   present.   There   is   potential   for   a   more   comprehensive  usable  greenspace  indicator,  weighted  for  the  quality  and  quantity  of   exercise   facilities   including   both   indoor   and   outdoor   sporting   facilities,   and   may   provide  for  a  greater  understanding  for  the  influence  of  accessibility  of  greenspace   and   exercise   facilities   more   generally   on   physical   activity   levels.   The   social   interaction   pathway   between   greenspace   and   mental   health   could   be   informed   similarly   through   weighting   of   greenspace   by   patronage.   Linear   features   such   as   waterfronts,   beaches   and   walkways   and   cycleways   are   poorly   captured   in   analysis   focused   on   greenspace   as   an   area,   so   these   areas   may   form   an   important   role   in   the   exercise   facility   accessibility   relationship.   A   similar   study   by   (Field   et   al.,   2004)   analysed  the  distribution  of  such  facilities  and  community  facilities  more  broadly,  but   this  distribution  had  not  been  related  to  utilisation  or  physical  activity  levels  to  date.   Established  size  thresholds  for  greenspaces  to  be  considered  usable  are  based  on  the   requirements  of  adults,  however  children  may  be  able  undertake  vigorous  physical   activity   in   far   smaller   areas;   future   studies   could   incorporate   the   physical   activity   levels  of  children  in  relation  to  a  smaller  threshold  for  usable  greenspace.      

6.6 Theoretical  Pathways   With   few   statistically   significant   results,   no   conclusions   can   be   drawn   for   the   theoretical   pathways   identified   earlier.   Correlations   between   mental   illness   symptoms   and   usable   greenspace   could   have   validated   the   physical   activity   and   social  interaction  pathway  in  relation  to  mental  health.  The  physical  activity  pathway   could  have  been  validated  in  terms  of  physical  health  through  a  correlation  of  usable   40  

Nicholas  Brunsdon   greenspace   and   physical   activity.   The   therapeutic   landscape   pathway   could   have   been   validated   through   the   correlation   of   mental   illness   symptoms   and   public   greenspace;   similarly   a   correlation   between   mental   illness   symptoms   and   all   greenspace   but   not   all   public   greenspace   could   suggest   the   importance   of   private   gardens.  As  above,  further  investigation  is  needed  to  establish  these  pathways  in  the   New  Zealand  context.      

41  

7. Conclusion   This   study   found   no   widespread   conclusive   evidence   of   greenspace   exposure   or   accessibility  influencing  short-­‐term  physical  activity  or  mental  health  indicators.  The   significance   of   impervious   surfaces   as   an   indicator   for   private   greenspace,   and   its   significant  relationship  with  nervous  feelings  suggests  promise  for  future  studies  of   private   gardens,   although   no   conclusions   can   be   drawn   from   these   results.   Conflicting   relationships   from   greenspace   in   adjacent   buffer   areas   calls   for   further   sensitivity   analysis   of   buffer   size   and   its   influence   on   the   greenspace   and   health   relationship.   Neighbourhood   accessibility   to   usable   greenspace   was   found   to   be   a   significant   determinant   of   vigorous   physical   activity,   however   no   relationship   was   found   for   other   forms   of   physical   activity   and   thus   no   wider   conclusions   can   be   drawn.     The  application  of  coarse  greenspace  measures  has  not  been  successful  in  the  New   Zealand  context;  further  refinement  and  development  of  more  detailed  greenspace   measures   reflecting   visualisation   and   quality   may   provide   useful   insight   into   contribution   of   greenspace   to   health   locally   and   internationally.     The   contribution   of   bluespace   on   health   has   not   been   assessed   locally,   however   this   would   be   readily   achieved  through  an  adaption  of  greenspace  methods.  

42  

Nicholas  Brunsdon    

Acknowledgements     Special  thanks  are  due  to  my  supervisor  Professor  Simon  Kingham  and  associates  in   the  GeoHealth  laboratory  -­‐  Chris  Bowie,  Edward  Griffin,  Niamh  Donnellan  and  Daniel   Nutsford   -­‐   for   their   support   throughout   this   project,   from   inception   to   completion.   I’m  looking  forward  to  working  with  you  all  again  in  2014.     I   also   wish   to   thank   my   GEOG420   colleagues   for   their   support   and   positivity   throughout  the  year,  and  I  wish  them  all  the  best  in  their  promising  futures.    

43  

 

References   American  Planning  Association,  &  National  Association  of  County  &  City  Health   Officials.  (2013).  Public  health  terms  for  planners  and  planning  terms  for   public  health  professionals.  Washington,  D.C:  Centers  for  Disease  Control  and   Prevention.   Anthony,  M.,  &  Stein,  M.  (2008).  Oxford  Handbook  of  Anxiety  and  Related  Disorders.   Oxford  University  Press.   Antonovsky,  A.  (1979).  Health,  stress,  and  coping.  Jossey-­‐Bass.   Babyak,  M.,  Blumenthal,  J.  A.,  Herman,  S.,  Khatri,  P.,  Doraiswamy,  M.,  Moore,  K.,  …   Krishnan,  K.  R.  (2000).  Exercise  Treatment  for  Major  Depression:   Maintenance  of  Therapeutic  Benefit  at  10  Months.  Psychosomatic  Medicine,   62(5),  633–638.   Barbosa,  O.,  Tratalos,  J.  A.,  Armsworth,  P.  R.,  Davies,  R.  G.,  Fuller,  R.  A.,  Johnson,  P.,   &  Gaston,  K.  J.  (2007).  Who  benefits  from  access  to  green  space?  A  case   study  from  Sheffield,  UK.  Landscape  and  Urban  Planning,  83(2),  187–195.   Comber,  A.,  Brunsdon,  C.,  &  Green,  E.  (2008).  Using  a  GIS-­‐based  network  analysis  to   determine  urban  greenspace  accessibility  for  different  ethnic  and  religious   groups.  Landscape  and  Urban  Planning,  86(1),  103–114.   doi:10.1016/j.landurbplan.2008.01.002   Coombes,  E.,  Jones,  A.  P.,  &  Hillsdon,  M.  (2010).  The  relationship  of  physical  activity   and  overweight  to  objectively  measured  green  space  accessibility  and  use.   Social  Science  &  Medicine,  70(6),  816–822.   doi:10.1016/j.socscimed.2009.11.020   44  

Nicholas  Brunsdon   Craig,  C.,  Russell,  S.  J.,  Cameron,  C.,  &  Bauman,  A.  (2003).  Twenty-­‐year  trends  in   physical  activity  among  Canadian  adults.  Canadian  journal  of  public  health  =   Revue  canadienne  de  sante  publique,  95(1),  59–63.   Crawford,  J.  R.,  &  Henry,  J.  D.  (2003).  The  Depression  Anxiety  Stress  Scales  (DASS):   Normative  data  and  latent  structure  in  a  large  non-­‐clinical  sample.  British   Journal  of  Clinical  Psychology,  42(2),  111–131.   doi:10.1348/014466503321903544   Dadvand,  P.,  de  Nazelle,  A.,  Figueras,  F.,  Basagaña,  X.,  Su,  J.,  Amoly,  E.,  …   Nieuwenhuijsen,  M.  J.  (2012).  Green  space,  health  inequality  and  pregnancy.   Environment  International,  40,  110–115.  doi:10.1016/j.envint.2011.07.004   Dillen,  S.  M.  E.  van,  Vries,  S.  de,  Groenewegen,  P.  P.,  &  Spreeuwenberg,  P.  (2012).   Greenspace  in  urban  neighbourhoods  and  residents’  health:  adding  quality  to   quantity.  Journal  of  Epidemiology  and  Community  Health,  66(6),  e8–e8.   doi:10.1136/jech.2009.104695   Field,  A.,  Witten,  K.,  Robinson,  E.,  &  Pledger,  M.  (2004).  Who  gets  to  what?  Access  to   community  resources  in  two  New  Zealand  cities.  Urban  Policy  and  Research,   22(2),  189–205.  doi:10.1080/0811114042000229645   Giles-­‐Corti,  B.,  &  Donovan,  R.  J.  (2002).  Socioeconomic  Status  Differences  in   Recreational  Physical  Activity  Levels  and  Real  and  Perceived  Access  to  a   Supportive  Physical  Environment.  Preventive  Medicine,  35(6),  601–611.   doi:10.1006/pmed.2002.1115   Gove,  W.  R.,  &  Geerken,  M.  R.  (1977).  Response  Bias  in  Surveys  of  Mental  Health:  An   Empirical  Investigation.  American  Journal  of  Sociology,  82(6),  1289–1317.   doi:10.2307/2777936  

45  

Hartig,  T.,  Berg,  A.  E.  van  den,  Hagerhall,  C.  M.,  Tomalak,  M.,  Bauer,  N.,  Hansmann,   R.,  …  Waaseth,  G.  (2011).  Health  Benefits  of  Nature  Experience:   Psychological,  Social  and  Cultural  Processes.  In  K.  Nilsson,  M.  Sangster,  C.   Gallis,  T.  Hartig,  S.  de  Vries,  K.  Seeland,  &  J.  Schipperijn  (Eds.),  Forests,  Trees   and  Human  Health  (pp.  127–168).  Springer  Netherlands.  Retrieved  from   http://link.springer.com.ezproxy.canterbury.ac.nz/chapter/10.1007/978-­‐90-­‐ 481-­‐9806-­‐1_5   Kerr,  J.,  Frank,  L.,  Sallis,  J.  F.,  &  Chapman,  J.  (2007).  Urban  form  correlates  of   pedestrian  travel  in  youth:  Differences  by  gender,  race-­‐ethnicity  and   household  attributes.  Transportation  Research  Part  D:  Transport  and   Environment,  12(3),  177–182.  doi:10.1016/j.trd.2007.01.006   Kling,  J.  R.,  Liebman,  J.  B.,  &  Katz,  L.  F.  (2007).  Experimental  Analysis  of   Neighborhood  Effects.  Econometrica,  75(1),  83–119.  doi:10.1111/j.1468-­‐ 0262.2007.00733.x   Lachowycz,  K.,  &  Jones,  A.  P.  (2011).  Greenspace  and  obesity:  a  systematic  review  of   the  evidence.  Obesity  Reviews,  12(5),  e183–e189.  doi:10.1111/j.1467-­‐ 789X.2010.00827.x   Lee,  A.  C.  K.,  &  Maheswaran,  R.  (2011).  The  health  benefits  of  urban  green  spaces:  a   review  of  the  evidence.  Journal  of  Public  Health,  33(2),  212–222.   doi:10.1093/pubmed/fdq068   Lofors,  J.,  Ramírez-­‐León,  V.,  &  Sundquist,  K.  (2006).  Neighbourhood  income  and   anxiety:  a  study  based  on  random  samples  of  the  Swedish  population.  The   European  Journal  of  Public  Health,  16(6),  633–639.   doi:10.1093/eurpub/ckl026  

46  

Nicholas  Brunsdon   Loram,  A.,  Tratalos,  J.,  Warren,  P.  H.,  &  Gaston,  K.  J.  (2007).  Urban  domestic  gardens   (X):  the  extent  &  structure  of  the  resource  in  five  major  cities.  Landscape   Ecology,  22(4),  601–615.  doi:10.1007/s10980-­‐006-­‐9051-­‐9   Maas,  J.,  van  Dillen,  S.  M.  E.,  Verheij,  R.  A.,  &  Groenewegen,  P.  P.  (2009).  Social   contacts  as  a  possible  mechanism  behind  the  relation  between  green  space   and  health.  Health  &  Place,  15(2),  586–595.   doi:10.1016/j.healthplace.2008.09.006   Macintyre,  S.,  Maciver,  S.,  &  Sooman,  A.  (1993).  Area,  Class  and  Health:  Should  we   be  Focusing  on  Places  or  People?  Journal  of  Social  Policy,  22(02),  213–234.   doi:10.1017/S0047279400019310   Mair,  C.,  Roux,  A.  V.  D.,  &  Galea,  S.  (2008).  Are  neighbourhood  characteristics   associated  with  depressive  symptoms?  A  review  of  evidence.  Journal  of   Epidemiology  and  Community  Health,  62(11),  940–946.   doi:10.1136/jech.2007.066605   Mann,  J.  J.  (2005).  The  Medical  Management  of  Depression.  New  England  Journal  of   Medicine,  353(17),  1819–1834.  doi:10.1056/NEJMra050730   Mathieu,  R.,  Freeman,  C.,  &  Aryal,  J.  (2007).  Mapping  private  gardens  in  urban  areas   using  object-­‐oriented  techniques  and  very  high-­‐resolution  satellite  imagery.   Landscape  and  Urban  Planning,  81(3),  179–192.   doi:10.1016/j.landurbplan.2006.11.009   Mayor,  K.,  Lyons,  S.,  Duffy,  D.,  &  Tol,  R.  S.  J.  (2009).  A  hedonic  analysis  of  the  value  of   parks  and  green  spaces  in  the  Dublin  area  (No.  331).  ESRI  working  paper.   Retrieved  from  http://www.econstor.eu/handle/10419/50078  

47  

Ministry  of  Health.  (2012).  New  Zealand  health  survey  methodology  report  2012.   Wellington,  N.Z.:  Ministry  of  Health.  Retrieved  from   https://www.health.govt.nz/system/files/documents/publications/new-­‐ zealand-­‐health-­‐survey-­‐methodology-­‐report.pdf   Mirowsky,  J.  (1996).  Age  and  the  Gender  Gap  in  Depression.  Journal  of  Health  and   Social  Behavior,  37(4),  362–380.  doi:10.2307/2137263   Natural  England.  (2009).  Our  Natural  Health  Service  The  role  of  the  natural   environment  in  maintaining  healthy  lives.  Natural  England.   O’Connor,  P.  J.,  Raglin,  J.  S.,  &  Martinsen,  E.  W.  (2000).  Physical  activity,  anxiety  and   anxiety  disorders.  International  Journal  of  Sport  Psychology,  31(2),  136–155.   Pate,  R.  R.,  Colabianchi,  N.,  Porter,  D.,  Almeida,  M.  J.,  Lobelo,  F.,  &  Dowda,  M.  (2008).   Physical  Activity  and  Neighborhood  Resources  in  High  School  Girls.  American   Journal  of  Preventive  Medicine,  34(5),  413–419.   doi:10.1016/j.amepre.2007.12.026   Pearce,  J.,  Witten,  K.,  &  Bartie,  P.  (2006).  Neighbourhoods  and  health:  a  GIS   approach  to  measuring  community  resource  accessibility.  Journal  of   Epidemiology  and  Community  Health,  60(5),  389–395.   doi:10.1136/jech.2005.043281   Richardson,  E.,  Mitchell,  R.,  Hartig,  T.,  Vries,  S.,  Astell-­‐Burt,  T.,  &  Frumkin,  H.  (2012).   Green  cities  and  health:  a  question  of  scale?  Journal  of  Epidemiology  and   Community  Health,  66(2),  160–165.  doi:10.1136/jech.2011.137240   Richardson,  E.,  Pearce,  J.,  Mitchell,  R.,  Day,  P.,  &  Kingham,  S.  (2010).  The  association   between  green  space  and  cause-­‐specific  mortality  in  urban  New  Zealand:  an  

48  

Nicholas  Brunsdon   ecological  analysis  of  green  space  utility.  BMC  Public  Health,  10(1),  240.   doi:10.1186/1471-­‐2458-­‐10-­‐240   Richardson,  E.,  Pearce,  J.,  Mitchell,  R.,  &  Kingham,  S.  (2013).  Role  of  physical  activity   in  the  relationship  between  urban  green  space  and  health.  Public  health,   127(4),  318–324.  doi:10.1016/j.puhe.2013.01.004   Salmond,  C.  E.,  Crampton,  P.  R.,  &  Atkinson,  J.  (2007).  NZDep2006  Index  of   Deprivation  User’s  Manual.  Wellington:  Department  of  Public  Health   University  of  Otago  Wellington.   Schipperijn,  J.,  Stigsdotter,  U.  K.,  Randrup,  T.  B.,  &  Troelsen,  J.  (2010).  Influences  on   the  use  of  urban  green  space  –  A  case  study  in  Odense,  Denmark.  Urban   Forestry  &  Urban  Greening,  9(1),  25–32.  doi:10.1016/j.ufug.2009.09.002   Solomon,  A.,  Haaga,  D.,  &  Arnow,  B.  (2001).  Is  Clinical  Depression  Distinct  from   Subthreshold  Depressive  Symptoms?  A  Review  of  the  Continuity  Issue  in   Depression  Research.  The  Journal  of  Nervous  and  Mental  Disease,  189(8).   Retrieved  from   http://journals.lww.com/jonmd/Fulltext/2001/08000/Is_Clinical_Depression _Distinct_from_Subthreshold.2.aspx   Spitzer,  R.,  Kroenke,  K.,  &  Williams,  J.  (1999).  Validation  and  utility  of  a  self-­‐report   version  of  prime-­‐md:  The  phq  primary  care  study.  JAMA,  282(18),  1737– 1744.  doi:10.1001/jama.282.18.1737   Stafford,  M.,  &  Marmot,  M.  (2003).  Neighbourhood  deprivation  and  health:  does  it   affect  us  all  equally?  International  Journal  of  Epidemiology,  32(3),  357–366.   doi:10.1093/ije/dyg084  

49  

Statistics  New  Zealand.  (2013).  Geographic  hierarchy.  Retrieved  September  4,  2013,   from   http://www.stats.govt.nz/surveys_and_methods/methods/classifications-­‐ and-­‐standards/geographic-­‐hierarchy.aspx   Stigsdotter,  U.  A.,  &  Grahn,  P.  (2004).  A  garden  at  your  doorstep  may  reduce  stress  –   Private  gardens  as  restorative  environments  in  the  city.  In  International   Conference  on  Inclusive  Environments  “Open  space:  people  space”.  Edinburgh   (Scotland)  (pp.  27–29).  Retrieved  from   http://friskinaturen.org/media/terapihaven_a_garden_at_your_doorstep.pdf   Stubbs,  M.  (2008).  Natural  green  space  and  planning  policy:  devising  a  model  for  its   delivery  in  regional  spatial  strategies.  Landscape  research,  33(1),  119–139.   Trost,  S.  G.,  Owen,  N.,  Bauman,  A.  E.,  Sallis,  J.  F.,  &  Brown,  W.  (2002).  Correlates  of   adults’  participation  in  physical  activity:  review  and  update.  Medicine  &   Science  in  Sports  &  Exercise,  34(12),  1996–2001.  doi:10.1097/00005768-­‐ 200212000-­‐00020   Ulrich,  R.  S.  (1984).  View  through  a  Window  May  Influence  Recovery  from  Surgery.   Science,  224(4647),  420–421.  doi:10.2307/1692984   Unruh,  A.  M.  (2002).  The  meaning  of  gardens  and  gardening  in  daily  life:  a   comparison  between  gardeners  with  serious  health  problems  and  healthy   participants.  In  XXVI  International  Horticultural  Congress:  Expanding  Roles  for   Horticulture  in  Improving  Human  Well-­‐Being  and  Life  Quality  639  (pp.  67–73).   Retrieved  from  http://www.actahort.org/books/639/639_7.htm   Villeneuve,  P.  J.,  Jerrett,  M.,  G.  Su,  J.,  Burnett,  R.  T.,  Chen,  H.,  Wheeler,  A.  J.,  &   Goldberg,  M.  S.  (2012).  A  cohort  study  relating  urban  green  space  with  

50  

Nicholas  Brunsdon   mortality  in  Ontario,  Canada.  Environmental  Research,  115,  51–58.   doi:10.1016/j.envres.2012.03.003   Vries,  S.  de,  Claßen,  T.,  Eigenheer-­‐Hug,  S.-­‐M.,  Korpela,  K.,  Maas,  J.,  Mitchell,  R.,  &   Schantz,  P.  (2011).  Contributions  of  Natural  Environments  to  Physical  Activity.   In  K.  Nilsson,  M.  Sangster,  C.  Gallis,  T.  Hartig,  S.  de  Vries,  K.  Seeland,  &  J.   Schipperijn  (Eds.),  Forests,  Trees  and  Human  Health  (pp.  205–243).  Springer   Netherlands.  Retrieved  from   http://link.springer.com.ezproxy.canterbury.ac.nz/chapter/10.1007/978-­‐90-­‐ 481-­‐9806-­‐1_8   Vries,  S.  de,  Dillen,  S.  M.  van,  Groenewegen,  P.  P.,  &  Spreeuwenberg,  P.  (2013).   Streetscape  greenery  and  health:  Stress,  social  cohesion  and  physical  activity   as  mediators.  Social  Science  &  Medicine,  94,  26.   Vries,  S.  de,  Verheij,  R.  A.,  Groenewegen,  P.  P.,  &  Spreeuwenberg,  P.  (2003).  Natural   environments  -­‐-­‐  healthy  environments?  An  exploratory  analysis  of  the   relationship  between  greenspace  and  health.  Environment  and  Planning  A,   35(10),  1717  –  1731.  doi:10.1068/a35111   Weissman,  M.,  Sholomskas,  D.,  Pottenger,  M.,  Prusoff,  B.,  &  Locke,  B.  (1977).   Assessing  depressive  symptoms  in  five  psychiatric  populations:  A  validation   study.  American  Journal  of  Epidemiology,  106(3),  203–214.   Wendel-­‐Vos,  W.,  Droomers,  M.,  Kremers,  S.,  Brug,  J.,  &  Van  Lenthe,  F.  (2007).   Potential  environmental  determinants  of  physical  activity  in  adults:  a   systematic  review.  Obesity  Reviews,  8(5),  425–440.  doi:10.1111/j.1467-­‐ 789X.2007.00370.x  

51  

Witten,  K.,  Hiscock,  R.,  Pearce,  J.,  &  Blakely,  T.  (2008).  Neighbourhood  access  to  open   spaces  and  the  physical  activity  of  residents:  A  national  study.  Preventive   Medicine,  47(3),  299–303.  doi:10.1016/j.ypmed.2008.04.010   World  Health  Organization,  World  Heart  Federation,  &  World  Stroke  Organization.   (2011).  Global  atlas  on  cardiovascular  disease  prevention  and  control.   Geneva:  World  Health  Organization  in  collaboration  with  the  World  Heart   Federation  and  the  World  Stroke  Organization.   Zhou,  X.,  &  Rana,  M.  M.  P.  (2012).  Social  benefits  of  urban  green  space:  A  conceptual   framework  of  valuation  and  accessibility  measurements.  Management  of   Environmental  Quality:  An  International  Journal,  23(2),  173–189.   doi:10.1108/14777831211204921   Zimmerman,  F.  J.,  &  Katon,  W.  (2005).  Socioeconomic  status,  depression  disparities,   and  financial  strain:  what  lies  behind  the  income-­‐depression  relationship?   Health  Economics,  14(12),  1197–1215.  doi:10.1002/hec.1011    

52  

Nicholas  Brunsdon  

Appendix  A   Table  A1:  -­‐  Logistic  regression  of  the  relationship  between  prior  doctor  diagnosis  for   mental  illness  (dependent  variable)  and  individual  and  area  level  variables,  each   considered  separately.  Results  are  given  as  odds  ratios  (95%  confidence  intervals)  relative   to  the  reference  category  (odds  ratio=1.0).  

  Depression   Anxiety  disorders   Age     15  –  24   1.00   1.00   25  –  34   3.96  (1.93  –  8.14)   4.63  (1.90  –  11.26)   35  –  44   4.71  (2.40  –  9.25)   4.24  (1.63  –  11.07)   45  –  54   5.31  (2.63  –  10.71)   2.28  (0.83  –  6.25)   55  –  64   5.51  (2.74  –  11.11)   2.25  (0.73  –  6.97)   Over  65   4.46  (2.30  –  8.67)   1.94  (0.68  –  5.53)   Sex   Female   1.00   1.00   Male   0.54  (0.41  –  0.70)   0.76  (0.46  –  1.26)   NZ  index  of  deprivation   1  –  Least  deprived   1.00   1.00   2   1.23  (0.76  –  1.99)   1.67  (0.77  –  3.65)   3   0.79  (0.38  –  1.64)   1.15  (0.54  –  2.44)   4   0.88  (0.51  –  1.52)   0.79  (0.33  –  1.89)   5  –  Most  deprived   0.80  (0.45  –  1.42)   0.88  (0.35  –  2.26)   Usable  greenspace  accessibility   1  –  Low     1.00   1.00   2   0.73  (0.35  –  1.53)   0.89  (0.29  –  2.74)   3   0.57  (0.30  –  1.06)   0.81  (0.27  –  2.44)   4   0.75  (0.40  –  1.40)   0.98  (0.38  –  2.56)   5    -­‐  High     0.47  (0.24  –  0.91)   0.51  (0.16  –  1.66)   Proportion  of  all  public  greenspace  within  0-­‐500m  buffer   1  –  Low     1.00   1.00   2   0.90  (0.55  –  1.46)   0.60  (0.29  –  1.26)   3   0.85  (0.55  –  1.31)   0.58  (0.33  –  1.02)   4   0.98  (0.46  –  2.08)   0.46  (0.23  –  0.91)   5  –  High     1.53  (0.92  –  2.55)   0.96  (0.39  –  2.38)   Proportion  of  all  public  greenspace  within  500-­‐2000m  buffer   1  –  Low     1.00   1.00   2   0.88  (0.55  –  1.41)   0.54  (0.29  –  1.02)   3   1.03  (0.46  –  2.33)   0.65  (0.23  –  1.83)   4   0.60  (0.27  –  1.30)   0.64  (0.23  –  1.77)   5  –  High     1.40  (0.75  –  2.61)   1.48  (0.70  –  3.12)  

53  

 

Table  A1  (cont.)  

  Depression   Anxiety  disorders   Proportion  of  all  greenspace  (from  impervious  surfaces)  within  0-­‐500m  buffer   1  –  Low   1.00   1.00   2   0.92  (0.50  –  1.71)   0.88  (0.49  –  1.57)   3   0.73  (0.38  –  1.38)   0.47  (0.22  –  0.97)   4   0.70  (0.38  –  1.29)   0.39  (0.16  –  1.00)   5  –  High   1.35  (0.72  –  2.53)   1.05  (0.55  –  2.01)   Proportion  of  all  greenspace  (from  impervious  surfaces)  within  500-­‐2000m  buffer   1  –  Low   1.00   1.00   2   0.79  (0.45  –  1.38)   0.60  (0.29  –  1.25)   3   0.72  (0.39  –  1.33)   0.34  (0.14  –  0.80)   4   0.66  (0.38  –  1.15)   0.64  (0.32  –  1.28)   5  –  High   1.35  (0.74  –  2.47)   1.10  (0.54  –  2.12)        

54  

The influence of private and public greenspace on short ... - Brunsdon

field of health geography, which has taken a closer look at the role of place in influencing .... of individuals in Perth, Australia, with those in high SED neighbourhoods less likely to ..... statistical software package STATA using different measures of time spent exercising ...... Our Natural Health Service The role of the natural.

3MB Sizes 2 Downloads 320 Views

Recommend Documents

the impact of private schools on public school ...
SCHOOL STUDENTS AS THE BEST AND BRIGHTEST TRANSFER TO PRIVATE ... The State of Ohio also created an Educational Management Information System ... the percent of persons in the district over the age of 25 who where college.

the impact of private schools on public school ...
SCHOOL STUDENTS AS THE BEST AND BRIGHTEST TRANSFER TO PRIVATE SCHOOLS. VOUCHER AND TAX CREDIT SUPPORTERS. REJECT THIS ...

The influence of landscape configuration and environment on ...
cal distance due to migration–genetic drift equilibrium in organisms with restricted dispersal (i.e. isolation by distance, IBD ... migration pathways for species with cryptic life-history traits or for which dispersal movements are difficult to ..

Public-Private Partnerships and Information ...
Department, and Rajiv Internet Villages, Rural Eseva and RSDP staff and entrepreneurs for .... an explicit development agenda in addition to a business orientation. ..... government programs, issuance of certificates, and access to. Government ...

Callogenesis and the influence of IBA on rhizogenesis ...
compared to Control. The cultivars had good callogenesis of 43.6/ 95.4% (5>8>2>0 g/l-1 IBA), whereas ... analysis and performance through Summary statistics.

Public and Private Administration.pdf
Download. Connect more apps... Try one of the apps below to open or edit this item. Public and Private Administration.pdf. Public and Private Administration.pdf.

On the Influence of Sensor Morphology on Vergence
present an information-theoretic analysis quantifying the statistical regu- .... the data. Originally, transfer entropy was introduced to identify the directed flow or.

National Policy on Public-Private Partnership for Health in Nigeria ...
National Policy on Public-Private Partnership for Health in Nigeria 2006.pdf. National Policy on Public-Private Partnership for Health in Nigeria 2006.pdf. Open.

Study on the influence of dryland technologies on ...
Abstract : A field experiment was conducted during the North East monsoon season ... Keywords: Sowing time, Land management, Seed hardening and Maize ...

Implementation of Cross Platorm Using Public and Private WEBOS.pdf ...
There was a problem previewing this document. Retrying... Download. Connect more apps... Try one of the apps below to open or edit this item. Implementation ...

Learning from Private and Public Observations of ...
Aug 24, 2011 - 6495; Fax: (310) 825-9528. 1 .... free-riding problem: given that other agents' actions generate observable .... In an online Addendum, available.

doc-Disclosure of Project and Contract Information in Public-Private ...
Retrying... doc-Disclosure of Project and Contract Information in Public-Private Partnerships in Nigeria..pdf. doc-Disclosure of Project and Contract Information in ...

doc-National Policy on Public Private Partnership.pdf
There was a problem previewing this document. Retrying... Download. Connect more apps... Try one of the apps below to open or edit this item. doc-National ...

The importance of topography and climate on short ...
account when developing reclamation management plans in Mediterranean ecosystems, and an ..... All statistical analyses were implemented in the R software.

National Policy on Public-Private Partnership for Health in Nigeria ...
National Policy on Public-Private Partnership for Health in Nigeria 2006.pdf. National Policy on Public-Private Partnership for Health in Nigeria 2006.pdf. Open.

Influence of different levels of spacing and manuring on growth ...
Page 1 of 8. 1. Influence of different levels of spacing and manuring on growth, yield and. quality of Alpinia calcarata (Linn.) Willd. Baby P Skaria, PP Joy, Samuel Mathew and J Thomas. 2006. Kerala Agricultural University, Aromatic and Medicinal Pl