Linking remotely sensed Urban Green Space (UGS) distribution patterns and Socio-Economic Status (SES) - A multi-scale probabilistic analysis based in Mumbai, India

Figures & data

Figure 1. Geographical location of Mumbai, and map of census sections in Mumbai.Note: Numbers on the map represent the ID of census sections; Insets A and B provide magnified views of census sections.

Figure 2. Methodological framework adopted for the study.Note: The numbers represent the ID of the census section; Insets A and B provide magnified views of census sections.

Table 1. Description of the census variables used for computing SESI.

Variable	Description	Mean	Std.Dev.	Minimum	Maximum
Households owning a car/jeep/van (Car)	% of households	15.17	11.41	1.40	53.80
Households availing banking services (Bank)	% of households	88.11	7.27	66.20	98.90
Households using kerosene as the main source of lighting (Light_Kero)	% of households	1.88	1.56	0.10	7.60
Households using kerosene as fuel for cooking (Fuel_Kero)	% of households	13.52	9.45	2.20	39.50
Households owned (House)	% of households	71.41	10.96	38.40	90.50
Female working population (FWP)	% of female population	19.84	5.23	9.06	36.73
Total Working Population (TWP)	% of population	41.99	4.43	35.66	71.09
Literacy Rate (LR)	% of population	82.15	4.70	69.24	90.66
Average Household Size (AHS)	Number of persons per household	4.46	0.32	3.85	5.40
Sex Ratio (SR)	Number of females per 1000 males	856	98.01	329	1022
Special Class Population (SCP)	% of population	6.39	4.32	0.32	21.39

Short form of the variables are given within brackets.

Table 2. Description of the urban green space distribution indictors used in the study.

Domain	Indicator	Formula	Description	Units and Range
Quantity	Percentage of UGS area (PUGS)	PUGS = $\frac{A_{UGS}}{A_{CS}}$ * 100	where $A_{UGS}$ is the area of UGS in the census section, and $A_{CS}$is the area of census section.	Percent; 0 ≤ PUGS ≤ 100
	UGS area per inhabitant (UPI)	UPI = $\frac{A_{UGS}}{P_{CS}}$	where $A_{UGS}$ is the area of UGS in the census section, and $P_{CS}$is the population of census section.	m^2 per person; UPI ≥ 0
Quality	Mean area of UGS patches* (Area_MN)	Area_MN = $\frac{A_{UGS}}{n}$	where $A_{UGS}$ is the area of UGS in the census section, n is the number of UGS patches in the census section.	ha; Area_MN ≥ 0
	Urban green space patch density* (UPD)	UPD = $\frac{n}{A_{CS}}$ * 100	where n is the number of UGS patches in the census section, and $A_{CS}$ is the area (in ha) of that census section.	number per 100 ha; PD ≥ 0
	Mean Euclidean nearest neighbour distance of UGS patches* (ENN_MN)	ENN_MN = $\frac{{\sum }_{i=1}^n{d}_i}{n}$	where $d_i$ is the distance between an UGS patch i and its nearest neighbouring patch in the census section, n is the number of UGS patches in the census section.	m; ENN_MN ≥ 0
Accessibility	Area weighted mean fractal dimension index* (FRAC_AM)	FRAC_AM = $\sum _{i=1}^{i=n}\left[\left(\frac{a_i}{A_{UGS}}\right)\left(\frac{2ln0.25p_i}{ln{a}_i}\right)\right]$	where $a_i$ and $p_i$ denote the area and perimeter of an UGS patch i respectively, n is the number of UGS patches in the census section, $A_{UGS}$ is the area of UGS in the census section.	Unitless; 1 ≤ FRAC_AM ≤ 2 FRAC_AM = 1 if the patches are square-shaped. FRAC_AM = 2 if the patches are highly convoluted.
	Aggregation index* (AI)	AI = $\left[\frac{J}{maxJ}\right]\ast 100$	where J is the number of joins between pixels of UGS patches in the census section based on single-count method, max J is the maximum possible value of J.	Percent; 0 ≤ AI ≤ 100 AI = 0 if the patches are maximally disaggregated. AI = 100 if the patches are maximally compact.

*Source: McGarigal (2015).

Table 3. Descriptive statistics of the socio-economic status clusters.

SES Cluster	N	MedianSESI	Std.Dev.	MinimumSESI	MaximumSESI
Very Low SES (VLS)	13	19.08	7.86	0	23.79
Low SES (LS)	12	33.72	3.32	27.74	38.93
Moderately Low SES (MLS)	19	44.96	3.24	39.36	49.56
Moderately High SES (MLS)	18	56.91	3.70	51.32	62.28
High SES (HS)	15	70.61	3.72	65.79	78.07
Very High SES (VHS)	11	88.05	6.17	80.26	100
Total	88	50.44	22.64	0	100

Figure 3. Socio-economic status clusters of census sections in Mumbai.

Figure 4. Map of urban green spaces in Mumbai extracted from satellite images of 5m spatial resolution.

Table 4. Urban green space distribution indicators of census section 1 at different spatial resolutions.

SpatialResolution (m)	PUGS	UPI	Area_MN	UPD	ENN_MN	FRAC_AM	AI
5	62.85	21.51	0.77	80.83	14.57	1.39	92.07
15	66.16	22.65	4.19	15.78	36.14	1.33	86.15
30	71.95	24.65	32.58	2.21	71.66	1.28	84.92

Figure 5. Scene covering Horniman Circle Garden, Mumbai.a) Google Earth image; b) Standard false colour composite of LISS IV image (5m); c) UGS map at 5m resolution; d) UGS map at 15m resolution, and e) UGS map at 30m resolution. Note: In panels (c) to (e), the urban green spaces are shown in green.

Table 5. Multinomial Logistic Regression model for socio-economic status and the urban green spaces indicators derived at 5m spatial resolution.

SES Category	Independent Variable	Value	Odds Ratio (OR)	ORLower Bound (95%)	ORUpper Bound(95%)
LS	Intercept	−60.043***
	UPI	0.000	1.000	0.997	1.002
	UPD	0.032***	1.033	1.020	1.045
	FRAC_AM	14.688***	2.392 e + 06	100.456	5.698 e + 10
	AI	0.406***	1.501	1.275	1.767
MLS	Intercept	−1.340
	UPI	−0.013	0.987	0.960	1.016
	UPD	−0.009	0.990	0.978	1.004
	FRAC_AM	3.062	21.364	0.021	2.152 e + 04
	AI	−0.004	0.996	0.909	1.092
MHS	Intercept	−3.406
	UPI	−0.007	0.992	0.969	1.017
	UPD	−0.001	0.999	0.988	1.011
	FRAC_AM	6.247	516.411	0.443	6.013 e + 05
	AI	−0.046	0.955	0.873	1.046
HS	Intercept	−16.822***
	UPI	−0.096	0.909	0.818	1.009
	UPD	0.005	1.005	0.992	1.018
	FRAC_AM	12.571***	2.882 e + 05	340.544	2.439 e + 08
	AI	0.015	1.015	0.877	1.174
VHS	Intercept	−22.336***
	UPI	−0.003	0.997	0.985	1.008
	UPD	0.001	1.001	0.988	1.015
	FRAC_AM	24.906***	6.553 e + 10	2.672 e + 08	1.607 e + 13
	AI	−0.118	0.888	0.779	1.013

Reference Category: VLSAIC: 320.35

**significant at p < 0.05***significant at p < 0.01

Table 6. Multinomial Logistic Regression model for socio-economic status and the urban green spaces indicators derived at 15m spatial resolution.

SES Category	Independent Variable	Value	Odds Ratio (OR)	OR Lower Bound (95%)	OR Upper Bound (95%)
LS	Intercept	−1.668
	PUGS	0.049	1.050	0.997	1.106
	Area_MN	−0.165	0.848	0.658	1.093
MLS	Intercept	0.043
	PUGS	0.017	1.017	0.974	1.062
	Area_MN	−0.139	0.870	0.678	1.118
MHS	Intercept	−0.105
	PUGS	0.024	1.025	0.975	1.077
	Area_MN	−0.320	0.726	0.411	1.282
HS	Intercept	−0.507
	PUGS	0.068	1.070	0.992	1.156
	Area_MN	−1.991	0.136	0.015	1.249
VHS	Intercept	−2.669**
	PUGS	0.095**	1.100	1.011	1.197
	Area_MN	−0.953	0.385	0.110	1.348

Reference Category: VLSAIC: 319.174

**significant at p < 0.05***significant at p < 0.01

Table 7. Multinomial Logistic Regression model for socio-economic status and the urban green spaces indicators derived at 30m spatial resolution.

SES Category	Independent Variable	Value	Odds Ratio (OR)	OR Lower Bound (95%)	OR Upper Bound (95%)
LS	Intercept	−6.126
	UPI	0.000	1.000	0.996	1.002
	UPD	0.091	1.095	0.971	1.234
	ENN_MN	−0.020	0.980	0.894	1.075
	AI	0.083	1.086	0.994	1.187
MLS	Intercept	−10.514**
	UPI	−0.007	0.993	0.977	1.010
	UPD	0.057	1.059	0.955	1.175
	ENN_MN	0.033	1.033	0.999	1.069
	AI	0.104***	1.110	1.028	1.198
MHS	Intercept	−0.135
	UPI	−0.006	0.994	0.969	1.018
	UPD	0.024	1.024	0.933	1.124
	ENN_MN	−0.002	0.998	0.963	1.034
	AI	0.005	1.005	0.942	1.071
HS	Intercept	7.949
	UPI	−0.110	0.896	0.796	1.008
	UPD	−0.050	0.951	0.859	1.053
	ENN_MN	−0.049	0.952	0.904	1.003
	AI	−0.025	0.975	0.905	1.050
VHS	Intercept	8.704
	UPI	−0.005	0.994	0.980	1.009
	UPD	−0.111	0.895	0.797	1.005
	ENN_MN	−0.044	0.957	0.908	1.009
	AI	−0.050	0.951	0.881	1.026

Reference Category: VLSAIC: 309.001

**significant at p < 0.05***significant at p < 0.01

McGarigal, K. 2015. “FRAGSTATS Help.” Accessed 19 July 2017. http://www.umass.edu/landeco/research/fragstats/documents/fragstats.help.4.2.pdf

 Google Scholar