Supporting Information: Reynolds et al. 2024. Socio-economic, not ecological, heterogeneity explains the abundance of urban pollinators across multiple spatial scales in a developing world city.

Appendix S1

Figure S1. Location of wards used in the study based on property price (filled circles) and urban cover categories (coloured polygons).

Appendix S2

Figure S2.1. Location of the 347 individual bee hotels overlaid on the ward polygons which are shaded by annual household income and proportion of urban land cover.

Appendix S3

Table S3. List of original and reclassified land cover classes taken from the South African National Land Cover (SANLC) dataset.

SANLC class	Reclassified category
unclassed	Other
contiguous (indigenous) forest	Natural vegetation
contiguous low forest & thicket	Natural vegetation
dense forest & woodland	Natural vegetation
open woodland	Natural vegetation
contiguous & dense plantation forest	Other
open & sparse plantation forest	Other
temporary unplanted (clear-felled) plantation forest	Other
low shrubland (other)	Natural vegetation
low shrubland (fynbos)	Natural vegetation
low shrubland (succulent karoo)	Natural vegetation
low shrubland (nama karoo)	Natural vegetation
sparsely wooded grassland	Natural vegetation
natural grassland	Natural vegetation
natural rivers	Water
natural estuaries & lagoons	Water
natural ocean & coastal	Water
natural lakes	Water
natural pans (flooded @ observation times)	Water
artificial dams (including canals)	Water
artificial sewage ponds	Water
artificial flooded mine pits	Water
herbaceous wetlands (currently mapped)	Water
herbaceous wetlands (previously mapped)	Water
mangrove wetlands	Water
natural rock surfaces	Other
dry pans	Other
eroded lands	Other
sand dunes (terrestrial)	Other
coastal sand & dunes	Other
bare riverbed material	Other
other bare	Other
cultivated commercial permanent orchards	Other
cultivated commercial permanent vines	Other
cultivated commercial sugarcane pivot irrigated	Other
cultivated commercial permanent pineapples	Other
cultivated commercial sugarcane non-pivot	Other
cultivated emerging farmer sugarcane non-pivot	Other
commercial annual crops pivot irrigated	Other
commercial annual crops non-pivot irrigated	Other
commercial annual crops rain-fed / dryland	Other
subsistence / small-scale annual crops	Other
fallow land & old fields (trees)	Natural vegetation
fallow land & old fields (bush)	Natural vegetation
fallow land & old fields (grass)	Natural vegetation
fallow land & old fields (bare)	Natural vegetation
fallow land & old fields (low shrub)	Natural vegetation
residential formal (tree)	Urban vegetation
residential formal (bush)	Urban vegetation
residential formal (low veg / grass)	Urban vegetation
residential formal (bare)	Urban vegetation
residential informal (tree)	Urban vegetation
residential informal (bush)	Urban vegetation
residential informal (low veg / grass)	Urban vegetation
residential informal (bare)	Urban vegetation
village scattered (bare & low veg/ grass combo)	Urban vegetation
village dense (bare & low veg / grass combo)	Urban vegetation
smallholdings (tree)	Natural vegetation
smallholdings (bush)	Natural vegetation
smallholdings (low veg / grass)	Natural vegetation
smallholdings (bare)	Natural vegetation
urban recreational fields (tree)	Urban vegetation
urban recreational fields (bush)	Urban vegetation
urban recreational fields (grass)	Urban vegetation
urban recreational fields (bare)	Urban vegetation
commercial	Urban
industrial	Urban
roads & rails (major linear)	Urban
mines: surface infrastructure	Urban
mines: extraction pits, quarries	Other
mines: salt mines	Other
mine: tailings and resource dumps	Other
land-fills	Other
fallow land & old fields (wetlands)	Water

Appendix S4

Figure S4. Line plots of weekly average (+- SD) maximum and minimum temperatures at each bee hotel for the 12 week study duration. Panel numbers indicate bee hotel ID.

Table S4. Regression parameters from the analysis of the relationship between bee abundance, and minimum and maximum temperature.

Characteristic	Beta	95% CI1	p-value
avg_min	-0.02	-0.14, 0.09	0.7
avg_max	0.05	-0.03, 0.12	0.3
1 CI = Confidence Interval

Appendix S5

300 m

Table S5.1. Regression parameters for each of the candidate models used to test five competing hypotheses regarding the drivers of bee abundance at a 300 m spatial scale.

Characteristic	Beta	95% CI1	p-value
Model 2
AHI	0.14	0.05, 0.23	0.002
Model 3
Urban veg cover	0.09	-0.01, 0.18	0.074
NDVI	0.09	-0.01, 0.19	0.092
Model 5
Urban veg cover	0.07	-0.02, 0.17	0.14
NDVI	0.04	-0.06, 0.15	0.4
AHI	0.10	0.01, 0.20	0.035
Urban veg cover x AHI	0.00	-0.08, 0.08	>0.9
NDVI x AHI	-0.02	-0.11, 0.08	0.7
Model 6
Urban veg PD	-0.11	-0.24, 0.02	0.10
Urban veg cohesion	-0.10	-0.35, 0.15	0.4
AHI	0.15	0.06, 0.24	0.001
Urban veg PD x AHI	0.01	-0.08, 0.10	0.9
Urban veg cohesion	-0.04	-0.17, 0.10	0.6
Model 4
Urban veg PD	-0.05	-0.15, 0.06	0.4
Urban veg cohesion	0.02	-0.10, 0.15	0.7
1 CI = Confidence Interval

2000 m

Table S5.2. Regression parameters for each of the candidate models used to test five competing hypotheses regarding the drivers of bee abundance at a 2000 m spatial scale.

Characteristic	Beta	95% CI1	p-value
Model 2
AHI	0.14	0.05, 0.23	0.002
Model 6
Urban veg PD	-0.09	-0.24, 0.06	0.2
Urban veg cohesion	-0.10	-0.39, 0.19	0.5
AHI	0.15	0.06, 0.25	0.002
Urban veg PD x AHI	0.03	-0.06, 0.11	0.5
Urban veg cohesion	-0.02	-0.13, 0.09	0.7
Model 5
Urban veg cover	0.06	-0.04, 0.15	0.3
NDVI	-0.05	-0.17, 0.08	0.5
AHI	0.14	0.03, 0.25	0.010
Urban veg cover x AHI	0.00	-0.07, 0.07	>0.9
NDVI x AHI	-0.01	-0.10, 0.09	0.9
Model 3
Urban veg cover	0.06	-0.04, 0.15	0.3
NDVI	0.03	-0.08, 0.14	0.6
Model 4
Urban veg PD	0.00	-0.09, 0.09	>0.9
Urban veg cohesion	0.01	-0.08, 0.10	0.8
1 CI = Confidence Interval

5000 m

Table S5.3. Regression parameters for each of the candidate models used to test five competing hypotheses regarding the drivers of bee abundance at a 5000 m spatial scale.

Characteristic	Beta	95% CI1	p-value
Model 2
AHI	0.15	0.06, 0.23	<0.001
Model 5
Urban veg cover	0.04	-0.06, 0.14	0.4
NDVI	-0.06	-0.21, 0.09	0.4
AHI	0.17	0.05, 0.29	0.007
Urban veg cover x AHI	0.02	-0.05, 0.10	0.5
NDVI x AHI	-0.06	-0.15, 0.03	0.2
Model 6
Urban veg PD	0.00	-0.13, 0.13	>0.9
Urban veg cohesion	-0.03	-0.30, 0.23	0.8
AHI	0.17	0.07, 0.27	<0.001
Urban veg PD x AHI	-0.03	-0.12, 0.06	0.5
Urban veg cohesion	0.00	-0.16, 0.16	>0.9
Model 3
Urban veg cover	0.02	-0.08, 0.12	0.7
NDVI	0.09	-0.02, 0.20	0.091
Model 4
Urban veg PD	0.04	-0.05, 0.13	0.3
Urban veg cohesion	0.01	-0.07, 0.09	0.8
1 CI = Confidence Interval

Appendix S6

Table S6.1. Details of the explanatory variables used in our secondary bee abundance analysis.

Explanatory variable	Abbreviation	Units	Ecological context
Median annual household income	AHI	South African Rands	Socio-economic status
Mean Normalised Difference Vegetation Index (NDVI)	NDVI	Range between 0 - 1	Landscape productivity
Mean tree cover during 2015	Tree cover	%	Vegetation structure and available nesting sites
Patch density of urban vegetation cover	Urban veg PD	Patches per 100 ha	Configuration of urban vegetation habitat patches - index of fragmentation
Patch density of natural vegetation cover	Natural veg PD	Patches per 100 ha	Configuration of natural vegetation habitat patches - index of fragmentation
Patch density of wetland cover	Water PD	Patches per 100 ha	Configuration wetland habitat patches - index of fragmentation
Proportion coverage of urban vegetation cover	Urban veg cover	%	Composition of urban vegetation - index of quantity
Proportion coverage of natural vegetation cover	Natural veg cover	%	Composition of natural vegetation - index of quantity
Proportion coverage of wetland cover	Wetland cover	%	Composition of wetlands - index of quantity
Patch Cohesion Index for urban vegetation cover	Urban veg cohesion	%	Configuration urban vegetation habitat patches - index of connectivity
Patch Cohesion Index for natural vegetation cover	Natural veg cohesion	%	Configuration natural vegetation habitat patches - index of connectivity
Patch Cohesion Index for wetland cover	Wetland cohesion	%	Configuration wetland habitat patches - index of connectivity

300 m

Table S6.2. Results of the AIC model selection procedure for our secondary linear modelling analysis which tested our five original competing hypotheses as well as the extra ecological covariates of tree cover, natural vegetation cover, wetland cover, and associate landscape metrics from these additional land cover categories at the 300 m spatial scale. K, number of model parameters; AICWt, AIC weight, and Cum.Wt, cumulative AIC weight. * Indicates an interaction between covariates.

Model	Model number	AIC	K	Delta_AIC	AICWt	Cum.Wt
abundance ~ AHI	2	2,582	4	0.00	0.46	0.46
abundance ~ NDVI + Tree cover + NDVI x Tree cover	10	2,584	6	2.73	0.12	0.58
abundance ~ Urban veg cover + NDVI	3	2,584	5	2.98	0.10	0.68
abundance ~ Tree cover	9	2,585	4	3.46	0.08	0.77
abundance ~ Urban veg cover + NDVI + Urban veg cover x AHI + NDVI x AHI	5	2,585	8	3.90	0.07	0.83
abundance ~ NDVI	8	2,586	4	4.06	0.06	0.89
abundance ~ Urban PD + Urban veg cohesion + Urban PD x AHI + Urban veg cohesion x AHI	6	2,587	8	5.03	0.04	0.93
abundance ~ Natural veg PD + Natural veg cohesion	12	2,587	5	5.41	0.03	0.96
abundance ~ 1	1	2,589	3	7.23	0.01	0.97
abundance ~ Natural veg cover + NDVI + Natural veg cover x NDVI	11	2,589	6	7.63	0.01	0.98
abundance ~ Natural veg cover	7	2,590	4	8.45	0.01	0.99
abundance ~ Wetland cover	13	2,591	4	9.12	0.00	0.99
abundance ~ Urban PD + Urban veg cohesion	4	2,591	5	9.42	0.00	1.00
abundance ~ Wetland PD + Wetland cohesion	14	2,592	5	10.96	0.00	1.00

Table S6.3. Regression parameters for secondary analysis candidate models used to test the drivers of bee abundance at a 300 m spatial scale.

Characteristic	Beta	95% CI1	p-value
Model 2
AHI	0.14	0.05, 0.23	0.002
Model 10
NDVI	0.08	-0.04, 0.19	0.2
Tree cover	0.02	-0.09, 0.14	0.7
NDVI x Tree cover	-0.07	-0.14, 0.00	0.057
Model 3
Urban veg cover	0.09	-0.01, 0.18	0.074
NDVI	0.09	-0.01, 0.19	0.092
Model 9
Tree cover	0.11	0.02, 0.20	0.014
Model 5
Urban veg cover	0.07	-0.02, 0.17	0.14
NDVI	0.04	-0.06, 0.15	0.4
AHI	0.10	0.01, 0.20	0.035
Urban veg cover x AHI	0.00	-0.08, 0.08	>0.9
NDVI x AHI	-0.02	-0.11, 0.08	0.7
Model 8
NDVI	0.11	0.02, 0.21	0.023
Model 6
Urban veg PD	-0.11	-0.24, 0.02	0.10
Urban veg cohesion	-0.10	-0.35, 0.15	0.4
AHI	0.15	0.06, 0.24	0.001
Urban veg PD x AHI	0.01	-0.08, 0.10	0.9
Urban veg cohesion	-0.04	-0.17, 0.10	0.6
Model 12
Natural veg PD	-0.10	-0.19, -0.02	0.020
Natural veg cohesion	-0.02	-0.10, 0.07	0.7
Model 11
Natural veg cover	-0.03	-0.11, 0.06	0.5
NDVI	0.11	0.01, 0.21	0.033
Natural veg cover x NDVI	0.00	-0.12, 0.11	>0.9
Model 7
Natural veg cover	-0.04	-0.12, 0.05	0.4
Model 13
Wetland cover	0.02	-0.07, 0.10	0.7
Model 4
Urban veg PD	-0.05	-0.15, 0.06	0.4
Urban veg cohesion	0.02	-0.10, 0.15	0.7
Model 14
Wetland PD	0.01	-0.14, 0.15	>0.9
Wetland cohesion	0.02	-0.13, 0.16	0.8
1 CI = Confidence Interval

2000 m

Table S6.4. Results of the AIC model selection procedure for our secondary linear modelling analysis which tested our five original competing hypotheses as well as the extra ecological covariates of tree cover, natural vegetation cover, wetland cover, and associate landscape metrics from these additional land cover categories at the 2000 m spatial scale. K, number of model parameters; AICWt, AIC weight, and Cum.Wt, cumulative AIC weight. * Indicates an interaction between covariates.

Model	Model number	AIC	K	Delta_AIC	AICWt	Cum.Wt
abundance ~ AHI	2	2,581	4	0.00	0.75	0.75
abundance ~ NDVI + Tree cover + NDVI x Tree cover	10	2,586	6	4.50	0.08	0.83
abundance ~ Urban PD + Urban veg cohesion + Urban PD x AHI + Urban veg cohesion x AHI	6	2,587	8	6.17	0.03	0.87
abundance ~ Urban veg cover + NDVI + Urban veg cover x AHI + NDVI x AHI	5	2,588	8	6.60	0.03	0.89
abundance ~ Wetland PD + Wetland cohesion	14	2,589	5	7.36	0.02	0.91
abundance ~ 1	1	2,589	3	7.45	0.02	0.93
abundance ~ NDVI	8	2,589	4	7.97	0.01	0.94
abundance ~ Tree cover	9	2,589	4	8.13	0.01	0.96
abundance ~ Wetland cover	13	2,590	4	8.39	0.01	0.97
abundance ~ Natural veg PD + Natural veg cohesion	12	2,590	5	8.42	0.01	0.98
abundance ~ Urban veg cover + NDVI	3	2,590	5	8.75	0.01	0.99
abundance ~ Natural veg cover	7	2,591	4	9.37	0.01	1.00
abundance ~ Urban PD + Urban veg cohesion	4	2,593	5	11.37	0.00	1.00
abundance ~ Natural veg cover + NDVI + Natural veg cover x NDVI	11	2,593	6	11.95	0.00	1.00

Table S6.5. Regression parameters for secondary analysis candidate models used to test the drivers of bee abundance at a 2000 m spatial scale.

Characteristic	Beta	95% CI1	p-value
Model 2
AHI	0.14	0.05, 0.23	0.002
Model 10
NDVI	0.01	-0.12, 0.15	0.9
Tree cover	-0.04	-0.18, 0.10	0.5
NDVI x Tree cover	-0.08	-0.14, -0.03	0.005
Model 6
Urban veg PD	-0.09	-0.24, 0.06	0.2
Urban veg cohesion	-0.10	-0.39, 0.19	0.5
AHI	0.15	0.06, 0.25	0.002
Urban veg PD x AHI	0.03	-0.06, 0.11	0.5
Urban veg cohesion	-0.02	-0.13, 0.09	0.7
Model 5
Urban veg cover	0.06	-0.04, 0.15	0.3
NDVI	-0.05	-0.17, 0.08	0.5
AHI	0.14	0.03, 0.25	0.010
Urban veg cover x AHI	0.00	-0.07, 0.07	>0.9
NDVI x AHI	-0.01	-0.10, 0.09	0.9
Model 14
Wetland PD	0.08	0.00, 0.17	0.064
Wetland cohesion	-0.06	-0.18, 0.05	0.3
Model 8
NDVI	0.06	-0.04, 0.16	0.2
Model 9
Tree cover	0.05	-0.04, 0.14	0.2
Model 13
Wetland cover	-0.05	-0.14, 0.04	0.3
Model 12
Natural veg PD	-0.05	-0.16, 0.05	0.3
Natural veg cohesion	-0.08	-0.18, 0.01	0.083
Model 3
Urban veg cover	0.06	-0.04, 0.15	0.3
NDVI	0.03	-0.08, 0.14	0.6
Model 7
Natural veg cover	-0.01	-0.09, 0.07	0.8
Model 4
Urban veg PD	0.00	-0.09, 0.09	>0.9
Urban veg cohesion	0.01	-0.08, 0.10	0.8
Model 11
Natural veg cover	0.00	-0.08, 0.08	>0.9
NDVI	0.06	-0.04, 0.16	0.2
Natural veg cover x NDVI	0.01	-0.09, 0.10	>0.9
1 CI = Confidence Interval

5000 m

Table S6.6. Results of the AIC model selection procedure for our secondary linear modelling analysis which tested our five original competing hypotheses as well as the extra ecological covariates of tree cover, natural vegetation cover, wetland cover, and associate landscape metrics from these additional land cover categories at the 5000 m spatial scale. K, number of model parameters; AICWt, AIC weight, and Cum.Wt, cumulative AIC weight. * Indicates an interaction between covariates.

Model	Model number	AIC	K	Delta_AIC	AICWt	Cum.Wt
abundance ~ AHI	2	2,580	4	0.00	0.81	0.81
abundance ~ Urban veg cover + NDVI + Urban veg cover x AHI + NDVI x AHI	5	2,585	8	5.68	0.05	0.86
abundance ~ Urban PD + Urban veg cohesion + Urban PD x AHI + Urban veg cohesion x AHI	6	2,586	8	6.16	0.04	0.90
abundance ~ NDVI	8	2,586	4	6.16	0.04	0.93
abundance ~ Urban veg cover + NDVI	3	2,588	5	8.04	0.01	0.95
abundance ~ NDVI + Tree cover + NDVI x Tree cover	10	2,588	6	8.20	0.01	0.96
abundance ~ 1	1	2,589	3	9.08	0.01	0.97
abundance ~ Tree cover	9	2,589	4	9.52	0.01	0.98
abundance ~ Natural veg cover + NDVI + Natural veg cover x NDVI	11	2,589	6	9.56	0.01	0.99
abundance ~ Wetland cover	13	2,589	4	9.63	0.01	0.99
abundance ~ Natural veg cover	7	2,591	4	11.02	0.00	1.00
abundance ~ Urban PD + Urban veg cohesion	4	2,592	5	12.10	0.00	1.00
abundance ~ Natural veg PD + Natural veg cohesion	12	2,592	5	12.51	0.00	1.00
abundance ~ Wetland PD + Wetland cohesion	14	2,593	5	13.04	0.00	1.00

Table S6.7. Regression parameters for secondary analysis candidate models used to test the drivers of bee abundance at a 5000 m spatial scale.

Characteristic	Beta	95% CI1	p-value
Model 2
AHI	0.15	0.06, 0.23	<0.001
Model 5
Urban veg cover	0.04	-0.06, 0.14	0.4
NDVI	-0.06	-0.21, 0.09	0.4
AHI	0.17	0.05, 0.29	0.007
Urban veg cover x AHI	0.02	-0.05, 0.10	0.5
NDVI x AHI	-0.06	-0.15, 0.03	0.2
Model 6
Urban veg PD	0.00	-0.13, 0.13	>0.9
Urban veg cohesion	-0.03	-0.30, 0.23	0.8
AHI	0.17	0.07, 0.27	<0.001
Urban veg PD x AHI	-0.03	-0.12, 0.06	0.5
Urban veg cohesion	0.00	-0.16, 0.16	>0.9
Model 8
NDVI	0.11	0.01, 0.20	0.024
Model 3
Urban veg cover	0.02	-0.08, 0.12	0.7
NDVI	0.09	-0.02, 0.20	0.091
Model 10
NDVI	0.11	-0.05, 0.27	0.2
Tree cover	-0.06	-0.20, 0.09	0.4
NDVI x Tree cover	-0.03	-0.09, 0.02	0.2
Model 9
Tree cover	0.06	-0.03, 0.15	0.2
Model 11
Natural veg cover	0.03	-0.06, 0.11	0.5
NDVI	0.11	0.01, 0.20	0.024
Natural veg cover x NDVI	0.02	-0.07, 0.11	0.7
Model 13
Wetland cover	-0.06	-0.14, 0.03	0.2
Model 7
Natural veg cover	0.01	-0.07, 0.09	0.8
Model 4
Urban veg PD	0.04	-0.05, 0.13	0.3
Urban veg cohesion	0.01	-0.07, 0.09	0.8
Model 12
Natural veg PD	0.05	-0.08, 0.17	0.5
Natural veg cohesion	0.04	-0.08, 0.16	0.5
Model 14
Wetland PD	0.01	-0.08, 0.10	0.8
Wetland cohesion	0.00	-0.10, 0.09	>0.9
1 CI = Confidence Interval

Appendix S7

Analysis of nursery locations

Nursery density

2000 m

Table S7.1. Regression parameters of the model used to test the relationship between density of garden nurseries and annual household income (AHI) at a 2000 m spatial scale.

Characteristic	Beta	95% CI1	p-value
AHI	0.27	0.00, 0.55	0.053
1 CI = Confidence Interval

Figure S7.1. Scatterplot of annual household income (AHI) and nursery density at the 2000 m scale with regression lines and 95% confidence intervals. AHI has been log transformed.

5000 m

Table S7.2.Regression parameters of the model used to test the relationship between density of garden nurseries and annual household income (AHI) at a 5000 m spatial scale.

Characteristic	Beta	95% CI1	p-value
AHI	1.8	0.91, 2.7	<0.001
1 CI = Confidence Interval

Figure S7.2. Scatterplot of annual household income (AHI) and nursery density at the 5000 m scale with regression lines and 95% confidence intervals. AHI has been log transformed.

Distance to nursery

300 m

Table S7.3. Regression parameters of the model used to test the relationship between distance to nursery and annual household income (AHI) at a 300 m spatial scale.

Characteristic	Beta	95% CI1	p-value
AHI	-0.64	-0.90, -0.39	<0.001
1 CI = Confidence Interval

Figure S7.3. Scatterplot of annual household income (AHI) and distance to nursery at the 300 m scale with regression lines and 95% confidence intervals. AHI has been log transformed.

2000 m

Table S7.4. Regression parameters of the model used to test the relationship between distance to nursery and annual household income (AHI) at a 2000 m spatial scale.

Characteristic	Beta	95% CI1	p-value
AHI	-0.85	-1.1, -0.58	<0.001
1 CI = Confidence Interval

Figure S7.4. Scatterplot of annual household income (AHI) and distance to nursery at the 2000 m scale with regression lines and 95% confidence intervals. AHI has been log transformed.

5000 m

Table S7.5. Regression parameters of the model used to test the relationship between distance to nursery and annual household income (AHI) at a 5000 m spatial scale.

Characteristic	Beta	95% CI1	p-value
AHI	-0.91	-1.2, -0.60	<0.001
1 CI = Confidence Interval

Figure S7.5. Scatterplot of annual household income (AHI) and distance to nursery at the 5000 m scale with regression lines and 95% confidence intervals. AHI has been log transformed.