Tracking Progress Toward Urban Nature Targets Using Landcover and Vegetation Indices: A Global Study for the 96 C40 Cities

Abstract

Access to urban natural space, including blue and greenspace, is associated with improved health. In 2021, the C40 Cities Climate Leadership Group set 2030 Urban Nature Declaration (UND) targets: "Quality Total Cover" (30% green area within each city) and "Equitable Spatial Distribution" (70% of the population living close to natural space). We evaluate progress toward these targets in the 96 C40 cities using globally available, high-resolution data sets for landcover and normalized difference vegetation index (NDVI). We use the European Space Agency (ESA)'s WorldCover data set to define greenspace with discrete landcover categories and ESA's Sentinel-2A to calculate NDVI, adding the "open water" landcover category to characterize total natural space. We compare 2020 levels of urban green and natural space to the two UND targets and predict the city-specific NDVI level consistent with the UND targets using linear regressions. The 96-city mean NDVI was 0.538 (range: 0.148, 0.739). Most (80%) cities meet the Quality Total Cover target, and nearly half (47%) meet the Equitable Spatial Distribution target. Landcover-measured greenspace and total natural space were strong (mean R ² = 0.826) and moderate (mean R ² = 0.597) predictors of NDVI and our NDVI-based natural space proximity measure, respectively. The 96-city mean predicted NDVI value of meeting the UND targets was 0.478 (range: 0.352-0.565) for Quality Total Cover and 0.660 (range: 0.498-0.767) for Equitable Spatial Distribution. Our translation of the area- and access-based metrics common in urban natural space targets into the NDVI metric used in epidemiology allows for quantifying the health benefits of achieving such targets.

Keywords: 0230 impacts of climate change: human health; 1640 remote sensing.

Figure 1

Flowchart of methods used to evaluate whether cities meet the two Urban Nature Declaration targets and to convert the targets to the normalized difference vegetation index scale. The colors indicate the analytical steps and spatial resolution of the data.

Figure 2

Distribution of maximum 2020 normalized difference vegetation index (NDVI) values for each 100 m pixel in C40 cities within each world region. Quartiles of NDVI are indicated by dashed vertical lines. These distributions do not include blue space.

Figure 3

Green and natural space across C40 cities by region in 2020, quantified using metrics comparable to the Quality Total Cover (panel a) and Equitable Spatial Distribution (panel b) Urban Natural Declaration targets. The scatter points represent cities and colors correspond to the region colors in Figure 2. The vertical lines in panel a mark the Quality Total Cover minimum goal range (0.30–0.40 of the urban area is greenspace) while the vertical line in panel (b) represents the Equitable Spatial Distribution target (0.70 of population has access to blue or greenspace within a 15‐min walk).

Figure 4

Fit statistics and predicted normalized difference vegetation index (NDVI) values for the regression models used to convert the Quality Total Cover target to the NDVI scale. Each dot represents a city. Panels (a) and (b) show the model adjusted R ² and root mean square error by region, respectively. Panel (c) shows the predicted NDVI value where the proportion of green area is 0.3, aligned with the lower minimum threshold proportion of greenspace in the Quality Total Cover target.

Figure 5

Fit statistics and predicted normalized difference vegetation index (NDVI) for the regression models used to convert the Equitable Spatial Distribution target to the NDVI scale. Each dot represents a city. Panels (a) and (b) show the model fit statistics by region. Panel a shows the adjusted R ² value, while Panel (b) shows the root mean square error. Panel (c) shows the predicted natural space NDVI value where 0.70 of the area, and thus population, has access to sufficient nearby natural space, aligned with the Equitable Spatial Distribution target. Models with poor fit (R ² less than 0.50) are shown with smaller dots.