Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

# High-spatiotemporal-resolution mapping of global urban change from 1985 to 2015

## Abstract

High-resolution global maps of annual urban land coverage provide fundamental information of global environmental change and contribute to applications related to climate mitigation and urban planning for sustainable development. Here we map global annual urban dynamics from 1985 to 2015 at a 30 m resolution using numerous surface reflectance data from Landsat satellites. We find that global urban extent has expanded by 9,687 km2 per year. This rate is four times greater than previous reputable estimates from worldwide individual cities, suggesting an unprecedented rate of global urbanization. The rate of urban expansion is notably faster than that of population growth, indicating that the urban land area already exceeds what is needed to sustain population growth. Looking ahead, using these maps in conjunction with integrated assessment models can facilitate greater understanding of the complex environmental impacts of urbanization and help urban planners avoid natural hazards; for example, by limiting new development in flood risk zones.

## Main

Many global environmental problems are associated with cities1,2, whose residents currently account for more than 55% of the world population. This percentage is projected to increase to 68% by 2050 (ref. 3). Urban area composition is highly dynamic4,5; however, our understanding of how cities change in space and over time is limited by the lack of spatially and temporally comprehensive urban land cover data at a high resolution6,7. Development of this information lags behind that of state-of-the-art non-urban landcover change data (for example, at 30 m with an annual resolution showing global forest cover change8). Detailed high-resolution maps of urban change are crucial for studying greenhouse gas emissions2, urban heat island effects9, population exposure to extreme weather events and air pollution10, land-use change dynamics11, ecosystem carbon losses12, urban sustainability13 and development pathways14. Such maps would also help urban planners make informed decisions regarding urban expansion in areas exposed to natural hazards, such as landslides, earthquakes, tsunamis, floods and disease15.

The current understanding of urban growth is largely based on demographic (population) data rather than information describing the spatial and temporal patterns of urban land-cover change. At a global scale, satellite-based approaches are essential for mapping changes in urban extent, but higher-resolution satellite-derived global urban land-cover maps are available only for specific epochs, while annual global products have coarse spatial resolutions of 0.5 to 1.0 km7,16,17. The result is that existing datasets lack the spatial and temporal resolutions needed to fully capture urban changes, especially in areas undergoing rapid growth, such as China and India18. Prior work from a meta-analysis of reported satellite-based studies suggests a global urban growth rate of ~1,900 km2 yr−1 during the period between 1970 and 2000 (the total area growth during the period was ~58,000 km2)18. These rate calculations require reassessment because of the inherent uncertainty associated with statistics calculated from datasets produced with different algorithms and data sources, which is the case with the existing global urban maps that were the basis for these estimates.

To generate a consistent database of global annual urban dynamics (GAUD), we used Landsat imagery to map annual changes in urban extent at a 30 m resolution for the period 1985 to 2015 on the Google Earth Engine (GEE) platform. First, we defined the extent of urbanized land between 1985 and 2015 by fusing four available global urban-extent maps with similar spatial resolutions for 1985 and 2015 (that is, the Global Human Settlement Layer6,19, the Global Urban Footprint20, the Global Urban Land 7 and the Global Artificial Impervious Area16; Supplementary Table 1). We extracted cells from the two fusion maps that changed from non-urban in 1985 to urban in 2015. We then used an annual time series (1985–2015) of the normalized urban areas composite index (NUACI) to detect the year of urbanization and green recovery (vegetative regrowth or new plantings in built environments) for each pixel within this urbanized extent. Finally, we assessed the accuracy of all derived products over the past three decades (for details, refer to Supplementary Methods 13).

## Performance of the global annual urban database

Accuracy assessment demonstrates that the fused GAUD global urban extent maps are robust across different urban ecoregions and have relatively high mean accuracy. The mean kappa coefficient of the GAUD in 2015 is 0.57, as assessed using ~200,000 samples across a range of globally representative urban ecoregions (Supplementary Fig. 1). This accuracy is higher than that of existing global urban extent maps, with the exception of 1985, for which the GAUD and GHSL perform similarly (Extended Data Fig. 1). Our fused results also have smaller commission and omission errors compared with four other global urban products (Supplementary Figs. 2 and 3). Better accuracy is mainly attributed to (1) our GAUD data having urban regions that are consistent in different urban products and (2) inconsistent regions being reclassified using a locally adaptive random forest classifier. We also excluded intermediate changes such as conversions from non-urban to urban followed by subsequent recovery to a green area. The accuracies of maps detailing the year of urbanization are 76% (1985–2000) and 82% (2000–2015) for the humid regions at the global scale (accounting for more than 90% of global urban lands), on the basis of a globally distributed sample of 12,000 human-interpreted validation points (Extended Data Figs. 2 and 3 and Supplementary Fig. 4). The derived annual maps of green recovery have an overall accuracy of nearly 80% for the 1985–2015 period (Extended Data Fig. 4 and Supplementary Fig. 5). Considering the unequal distribution of Landsat observations across space and time and their uneven quality due to varying cloud cover, limited observations exist for earlier years (that is, before the 1990s) particularly over Africa and Middle Asia and throughout the global tropics (Supplementary Figs. 6 and 7), resulting in higher uncertainties of detected urban expansion and green recovery for these regions and years. Our global maps of urban dynamics (that is, GAUD) are freely available online (refer to Data Availability).

## Global urban expansion and green recovery

In analysing the GAUD data, we found that the global urban extent increased from 362,747 km2 to 653,354 km2 from 1985 to 2015, representing a net expansion of 80% and a significant average expansion rate of 9,687 km2 per year (P value < 0.01, Fig. 1 and Supplementary Table 2). This net increase is four times greater than that estimated by Seto et al.18 for the period 1970–2000. During the same period (1985–2015), data from the United Nations show that urban population, an essential driver of urban area expansion, increased by 52% (Supplementary Fig. 8)3. Given that the various Shared Socioeconomic Pathways21 project continued steady growth of global urban population during the coming decades (Supplementary Fig. 8), the expansion of urban areas is likely to continue (Fig. 1). However, the global growth rate of urban lands (80%) is considerably higher than the population growth rate (52%) over the past decades (Extended Data Fig 5, Supplementary Fig. 9 and Methods), suggesting that the growth in urban areas has substantially exceeded what was needed to accommodate increasing numbers of urban residents. Thus, much of the newly developed urban lands were not used for housing but for other purposes (for example, commercial and industrial districts). This unbalanced growth of urban lands relative to population growth has been more pronounced in developing regions (for example, China and India).

Approximately 69% of the newly developed urban areas are in Asia and North America, where urban areas increased by an average of 4,970 ± 319 km2 yr–1 and 2,358 ± 150 km2 yr–1, respectively. The urban expansion rates in Asia and North America substantially exceeded those of Europe (1,883 km2 yr–1). Urban growth rates were smaller in Africa, Australia and South America, averaging less than 1,000 km2 yr–1. In addition, an ‘area under the curve’ analysis (Methods) shows that the rates of urbanization in Asia, Africa and South America accelerated during the 30 yr period but began to slow in North America, Europe and Australia (Fig. 2). These changes in growth rates between these regions were driven primarily by the development patterns of large countries; for example, China and India in Asia, and the United States in North America (Extended Data Fig. 6 for a comparison of the three nations). This pattern is even more pronounced when examining changes in urban clusters (that is, administrative units of urban that consist of central cores with adjacent densely populated territory using a morphological approach22, with a total number of 1,692 city boundaries in the world; Supplementary Methods 1). Urbanization accelerated in developing countries (for example, China, India, Saudi Arabia and a few urban clusters in Africa) but decelerated in developed countries (for example, United States, Canada, Japan and European countries) (Fig. 2b). For example, in the Shanghai area of China, the average growth rate of urban lands is 49.5 km2 yr–1, with the rate nearly twice as fast in the 2005–2015 period relative to 1985–1995 (Fig. 2c). By contrast, in the Chicago area, one of the most urbanized regions in the United States23, the mean increase in urban area during 1985–2015 (a total increase of 640 km2) is 21.3 km2 yr–1, with the rate decreasing from 37.2 km2 yr–1 between 1985 and 1995 to 6.6 km2 yr–1 between 2005 and 2015 (Fig. 2c).

The United States, China and India, the three countries with the most cities, experienced different urbanization trajectories (Fig. 2b), which is clearly seen in the patterns of gains and losses in urban cover around urban clusters (Fig. 3). Urban expansion in China and India occurred mainly in some large cities (that is, those with more than 300,000 inhabitants) and their surrounding rural areas by encroaching upon agricultural lands24. Urban growth in China was concentrated primarily in the Beijing–Tianjin–Hebei region; the Yangtze River Delta, including Shanghai; and the Pearl River Delta, including Guangzhou and Shenzhen (Fig. 3a). The increases in these particular urban areas account for more than 60% of China’s total urban expansion. In India, the most notable urban growth also occurred in large cities, such as New Delhi and Mumbai, around which many small human settlements have emerged over the past few decades (Fig. 3e). The numbers of large cities with notable urban growth (red colours) in China and India are notably smaller than those in the United States (Fig. 3c), which include Chicago, Atlanta, Dallas, Boston, Huston, New York and Las Vegas. Rapid expansion around large cities poses great challenges to achieving sustainable development goals because of the large increases in demand for water, energy and food, particularly when these occur in regions with limited resources11,25.

Green recovery was observed in many countries, including the United States, China and India (Fig. 3 and Extended Data Fig. 4). Interestingly, China experienced the largest green recovery of all nations, with nearly twice the area compared with developed countries in North America and Europe. This larger green recovery in China probably results from the recent national urban renewal efforts26, which aim to replace the previous low, wide and disordered development with dense high-rise buildings and green landscapes.

## Urbanization impacts on land system

Urbanization often comes at the cost of other valuable land resources, such as agricultural and forest lands. The GAUD dataset enables such costs to be assessed with a degree of accuracy and comprehensiveness that was previously unachievable. Using an existing 300 m resolution land-cover dataset covering the 1992 to 2015 period provided by the European Space Agency Climate Change Initiative (ESA-CCI) (Methods), we found that the bulk (~70%) of urban growth since 1992 occurred at the expense of agricultural land, followed by grasslands (~12%) and forests (~9%) (Fig. 4). Urbanization consumed agricultural land at a significant rate of 61,567 km2 per decade (P value < 0.01, Fig. 4b), while grassland and forest were lost at rates of 10,246 and 7,624 km2 per decade, respectively. New urban land came primarily from agricultural land, which represented 65% of new urban lands in 1992, increasing to 71% in 2015 (Fig. 4c and Supplementary Fig. 10). By contrast, the proportion of grassland converted to cities declined during this period, indicating that urban expansion is increasingly occurring on other land-cover types (Fig. 4c and Supplementary Fig. 10).

Regionally, the dominant types of land converted to urban areas are as follows: (1) agricultural land in China, India, South Korea, Japan, Europe, Southeast Asia countries and the central United States; (2) forest land in northern Europe and the eastern United States; (3) other cover types in Saudi Arabia and the western United States (Fig. 4a). China and India, the world’s most populous countries and where urban populations are still increasing rapidly, are threatened by food insecurity24,27 that might be exacerbated by urban-driven agricultural land loss, which was 49,077 km2 in China and 12,082 km2 in India. Given that urbanization often encroaches on productive agricultural land24, urban development planning should integrate management of food systems to ensure food security and promote sustainability. For example, China has implemented a policy of retaining at least 120 million hectares of arable land (the ‘1.8-billion-mu red line’)28 to prevent further loss of agricultural land to urbanization.

## Summary

Our unique global dataset of annual urban extent can be used to inform a wide range of global environmental change analyses as well as sustainable development planning29. For example, the dataset can be used to improve our understanding of how urban areas affect carbon cycles, a negative anthropogenic impact that has not been studied thoroughly30,31 because of the lack of accurate and continuous global urban extent maps. The GAUD data can also be used to study how urban expansion drives global changes in energy and water use and in turn triggers changes in water and energy fluxes between the land and atmosphere25,32. GAUD may also aid research into the impacts of rapid urban expansion on habitat and biodiversity loss or the extent that health risks arise from urban heat island effects. In summary, this dataset provides a new capability for studying urban change processes at fine spatial and temporal resolutions across the entire planet and can thus be a key input into many sustainable development studies.

## Methods

We mapped global annual urban dynamics and green recovery (1985–2015) at a fine resolution of 30 m (a dataset we named GAUD) using existing global urban extent maps in the beginning (1985) and ending (2015) years and a time series data of NUACI, which is an index for distinguishing urban land cover from Landsat images7. The framework is presented in Supplementary Fig. 11. This framework includes two components: (1) the generation of two global urban extent maps (1985 and 2015) using a data fusion approach and (2) the mapping of annual dynamics of urban expansion and green recovery using a temporal segmentation approach. We defined ‘urban areas’ as pixels that are dominated by built elements (for example, buildings, roads, runways and so on), which is a definition commonly used by the remote-sensing community. We defined ‘urban expansion’ as an increase in urban area over time that occurs when previously non-urban covers are transformed to urban cover. Change information (that is, conversion to urban between 1985 and 2015) was extracted from annual time series data of NUACI. We defined ‘green recovery’ as built-up areas reverting to more vegetation-dominated states during the study period. Processes behind the green recovery include reconstruction of parks or vegetation planting on built-up areas.

### Global urban extent maps in 1985 and 2015

The quality of urban extent maps in 1985 and 2015 was crucial to detecting urbanized and green recovery years because most uncertainties caused by natural disturbances can be constrained by these two maps if accurate. Details of the datasets used and the data fusion approach can be found in Supplementary Table 1 and Supplementary Fig. 12, respectively. First, we identified potential urban areas by overlaying three global urban extent maps listed in Supplementary Table 1 representing each period (1985 and 2015). The union of these three maps provides the potential maximum urban extent in 2015. Pixels that were identified as urban in these three products were regarded as high-confidence urban areas; otherwise, they were regarded as potential urban areas. Second, within the potential urban areas, we implemented a random forest classifier that was trained on samples collected in high-confidence areas using features derived from the six optimal bands in Landsat imagery during the green season. Considering the heterogeneity of urban environments and climate conditions, we trained the random forest model and classified these potential urban areas within one-degree tiles at the global scale. Similarly, we implemented the same approach to derive the fused urban extent map in 1985 (for more details, refer to Supplementary Methods 1).

### Algorithms to detect urban expansion and green recovery

We detected urbanized and green recovery years using a temporal segmentation approach (Supplementary Fig. 13). First, we constructed an annual NUACI time series data for each urbanized pixel during the period 1985–2015 using the full Landsat archive for this period on the GEE platform. NUACI integrates remotely sensed indicators of vegetation, water and built-up area for characterizing regional patterns of urban extent change. We calculated annual time series data of NUACI, which integrates the Normalized Difference Vegetation Index (NDVI) for the presence of green vegetation, the Normalized Difference Water Index (NDWI) for the presence of water, and the Normalized Difference Built-up Index (NDBI) for the presence of built-up area (Equation (1)). All were calculated for the vegetation growing season. We then detected urbanized years for these changed urban pixels by applying a temporal segmentation approach from the annual NUACI time series data. As the magnitude of NUACI should increase substantially following an urbanization event, the corresponding change year can be detected using a regression approach: the year with the largest residual relative to the regression line between the NUACI value and its corresponding year. This approach is efficient for detecting abrupt changes caused by urban expansion. Third, we detected the year of green recovery (that is, built-up areas transitioned to a predominantly vegetated status with similar magnitudes of change during the urban expansion period) for all pixels that were already urbanized. The detection approach is similar to that for identifying urbanization events, except we only consider the temporal segment of the post-urbanization period (that is, after the urbanized year) (for more details, refer to Supplementary Methods 2). NUACI values for each pixel (i) were calculated as:

$$\begin{array}{l}{\mathrm{NUACI}}_i =\\ U_{{\mathrm{Mask}}} \times \left( {1 - \sqrt {\left( {{\mathrm{NDWI}}_i - a_{{\mathrm{NDWI}}}} \right)^2 + \left( {{\mathrm{NDVI}}_i - b_{{\mathrm{NDVI}}}} \right)^2 + \left( {{\mathrm{NDBI}}_i - c_{{\mathrm{NDBI}}}} \right)^2} } \right)\end{array}$$
(1)

where UMask is a mask of urban areas from the fused urban product (0, non urban; 1, urban); aNDWI, aNDVI and cNDBI are the mean values of the three indicators NDWI, NDVI and NDBI, respectively. Values of aNDWI, bNDVI and cNDBI were calculated from the pre-collected urban samples in 15 urban ecoregions (Supplementary Fig. 1).

### Data validation and performance

We evaluated maps of GAUD in 1985 and 2015 using a global reference dataset7. Records in this reference dataset are 6 × 6 km urban and non-urban blocks (200 in total) that were proportionally allocated in different urban ecoregions (Supplementary Fig. 1). Within each block, we implemented an object-based segmentation approach to derive different cover types (urban and non-urban) from Landsat images in the corresponding validation year (1985 or 2015). We randomly selected 500 validation samples of urban and non-urban in each block (200,000 samples) and interpreted their types with knowledge from Google Earth high-resolution images and other available datasets. We evaluated the fused urban extent maps in 1985 and 2015 using the kappa coefficient within each block (Extended Data Fig. 1). We assessed the accuracy of mapped change (urbanized and green recovery) years using a multi-stage validation approach. We first randomly selected 140 cities across the globe according to their total urban increments since 1985, urban sizes in 2015 and associated biophysical backgrounds. We grouped these cities into three categories on the basis of their low (48), medium (63) and high (29) rate of growth using a geometrical interval approach (an advanced grouping approach for skewed data). Within each city, we randomly placed 20, 30 and 40 sample units within the low-, medium- and high-growth cities, respectively, to identify years of change during the period of 1985–2000. Given that there are more urbanized pixels and available datasets (for example, Google Earth high-resolution images and Landsat observations) during the 2000–2015 period, we doubled the random sample size in each cluster to assess changes between 2000 and 2015. For each validation pixel (in total 4,010 for period 1985–2000 and 8,020 for period 2000–2015), we identified the urbanized and green recovery years on the basis of time series data from Google Earth high-resolution images, Landsat images and the raw time series of NUACI over the three decades in question (Supplementary Fig. 14). We assessed the mapped urbanized year using a one-year tolerance, crediting change identified within a year on either side of a given year as change occurring in that year to accommodate the possibility of the expansion process occurring at the beginning of the following year or end of the previous year33. Our validation results can be found in Extended Data Figs. 24 and Supplementary Figs. 4 and 5. For more details, please refer to Supplementary Methods 3.

### Comparisons on the growth rates of urban extent versus urban population

We collected historical population data of each country and of urban clusters in three primary countries (United States, China and India). The country-specific annual population data were derived from the World Bank Database (https://data.worldbank.org) spanning 1985 to 2015. We also collected gridded population data from the Gridded Population of World (GPW)34 database, which includes records from four years (2000, 2005, 2010 and 2015). These GPW population grids provide estimates of the number of people per 30 arcsec grid cell (approximately 1 km at the Equator) and are designed to be consistent with national censuses. At the country level, we divided the 30 yr period into three intervals—P1 (1985–1995), P2 (1995–2005) and P3 (2005–2015)—using the population data from the World Bank. At the cluster level, we divided the period 2000–2015 into three 5 yr intervals (2000–2005; 2005–2010; 2010–2015) using the GPW population data. Within each time interval, we calculated the relative growth rate of urban areas and population using the following two equations:

$${\mathrm{Rate}}_{\mathrm{u}} = (U_i - U_j)/U_j$$
$${\mathrm{Rate}}_{\mathrm{p}} = (P_i - P_j)/P_j$$
(3)

where Rateu and Ratep indicate growth rates of urban areas and population relative to the base year in each interval, respectively; Ui and Uj are urban areas in the beginning and ending years within an interval (that is, P1, P2 and P3), respectively. Similarly, Pi and Pj denote population data in different years.

### Analysis of urban growth patterns

Urban areas experienced different rates and timing of growth during the 30 yr study period. To distinguish these varying characteristics of urban growth, we used an ‘area under the curve’ approach. We first calculated the normalized annual growth rate by dividing the annual increase in urban area by the total growth of each city over the 30 yr to make the growth rates of urban areas of different extents comparable. We then plotted this normalized growth rate by year and then integrated the area between the resulting growth curve, and the line represented a steady linear growth. The resulting integral was either positive or negative. The results show three characteristic growth patterns (Supplementary Fig. 15). Positive values of the integral, which occur between the blue and grey lines, describe urban areas that experienced more growth in the earlier years but slower or no growth in the later years. Negative values of the integral (found between the red and grey lines) represent cities that grew slowly in the early years and then turned to more rapid growth in later years. The closer the absolute value of an integral was to 0, the more constant the annual growth rate was for a given urban area.

### Conversion source of urbanized lands

We evaluated the impact of urban expansion on different land systems by determining the type and area of non-urban cover types that were converted to new urban areas. We used the annual land-cover data from the ESA-CCI35 to calculate the area of agriculture, forest, grassland, wetland, and other covers (for example, bare land—rocks or bared soils) at different spatial scales (for example, urban cluster, regional, global) that were converted annually between 1992 and 2015 (the period of the ESA-CCI data) for each urban cluster. The results were presented in terms of absolute and relative conversions (measured as percentage of total converted areas).

### Reporting Summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.

## Data availability

Satellite-derived high-resolution global urban maps from 1985 to 2015, the validation samples used in this study, are freely available at https://doi.org/10.6084/m9.figshare.11513178.v1. Other ancillary datasets are available on request from X. Liu (liuxp3@mail.sysu.edu.cn).

## Code availability

The script used for preprocessing the Landsat time series data in GEE (a cloud-based computational platform) is freely available at https://doi.org/10.6084/m9.figshare.11513178.v1. Analysis scripts are available on request from X. Liu (liuxp3@mail.sysu.edu.cn).

## References

1. 1.

IPCC Special Report on Climate Change, Desertification, Land Degradation, Sustainable Land Management, Food Security, and Greenhouse Gas Fluxes in Terrestrial Ecosystems (SR2) (WPO and UNEP, 2017).

2. 2.

Seto, K. C. et al. in Climate Change 2014: Mitigation of Climate Change (eds Edenhofer, O. et al.) 927–1000 (Cambridge Univ. Press, 2014).

3. 3.

World Urbanization Prospects: The 2018 Revision (UN, 2019).

4. 4.

Sexton, J. O. et al. Urban growth of the Washington, D.C.–Baltimore, MD metropolitan region from 1984 to 2010 by annual, Landsat-based estimates of impervious cover. Remote Sens. Environ. 129, 42–53 (2013).

5. 5.

Seto, K. C. et al. Urban land teleconnections and sustainability. Proc. Natl Acad. Sci. USA 109, 7687–7692 (2012).

6. 6.

Pesaresi, M. et al. GHS Built-up Grid, Derived from Landsat, Multitemporal (1975, 1990, 2000, 2014) (European Commission, 2015).

7. 7.

Liu, X. et al. High-resolution multi-temporal mapping of global urban land using Landsat images based on the Google Earth Engine Platform. Remote Sens. Environ. 209, 227–239 (2018).

8. 8.

Hansen, M. C. et al. High-resolution global maps of 21st-century forest cover change. Science 342, 850–853 (2013).

9. 9.

Zhou, L. et al. Evidence for a significant urbanization effect on climate in China. Proc. Natl Acad. Sci. USA 101, 9540–9544 (2004).

10. 10.

IPCC Climate Change 2013: The Physical Science Basis (eds Stocker, T. F. et al.) (Cambridge Univ. Press, 2013).

11. 11.

Kiswanto, Tsuyuki, S., Mardiany & Sumaryono. Completing yearly land cover maps for accurately describing annual changes of tropical landscapes. Glob. Ecol. Conserv. 13, e00384 (2018)..

12. 12.

Foley, J. A. et al. Global consequences of land use. Science 309, 570–574 (2005).

13. 13.

Stokes, E. C. & Seto, K. Characterizing and measuring urban landscapes for sustainability. Environ. Res. Lett. 14, 045002 (2018).

14. 14.

Zhang, W., Villarini, G., Vecchi, G. A. & Smith, J. A. Urbanization exacerbated the rainfall and flooding caused by hurricane Harvey in Houston. Nature 563, 384–388 (2018).

15. 15.

Reducing Disaster Risk by Managing Urban Land Use: Guidance Notes for Planners (Asian Development Bank, 2016).

16. 16.

Gong, P. et al. Annual maps of global artificial impervious area (GAIA) between 1985 and 2018. Remote Sens. Environ. 236, 111510 (2020).

17. 17.

Zhou, Y., Li, X., Asrar, G. R., Smith, S. J. & Imhoff, M. A global record of annual urban dynamics (1992–2013) from nighttime lights. Remote Sens. Environ. 219, 206–220 (2018).

18. 18.

Seto, K. C., Fragkias, M., Güneralp, B. & Reilly, M. K. A meta-analysis of global urban land expansion. PloS ONE 6, e23777 (2011).

19. 19.

Pesaresi, M. et al. A global human settlement layer from optical HR/VHR RS data: concept and first results. IEEE J Sel. Top. Appl. Earth Obs. Remote Sens. 6, 2102–2131 (2013).

20. 20.

Esch, T. et al. Urban footprint processor—Fully automated processing chain generating settlement masks from global data of the TanDEM-X mission. IEEE Geosci. Remote Sens. Lett. 10, 1617–1621 (2013).

21. 21.

Riahi, K. et al. The Shared Socioeconomic Pathways and their energy, land use, and greenhouse gas emissions implications: an overview. Glob. Environ. Change 42, 153–168 (2017).

22. 22.

Taubenböck, H. et al. A new ranking of the world’s largest cities—do administrative units obscure morphological realities? Remote Sens. Environ. 232, 111353 (2019).

23. 23.

Homer, C. G. et al. Completion of the 2011 National Land Cover Database for the conterminous United States-Representing a decade of land cover change information. Photogramm. Eng. Remote Sens. 81, 345–354 (2015).

24. 24.

Seto, K. C. & Ramankutty, N. Hidden linkages between urbanization and food systems. Science 352, 943–945 (2016).

25. 25.

McDonald, R. I. et al. Water on an urban planet: urbanization and the reach of urban water infrastructure. Glob. Environ. Change 27, 96–105 (2014).

26. 26.

Fu, Y. C. et al. Characterizing the spatial pattern of annual urban growth by using time series Landsat imagery. Sci. Total Environ. 666, 274–284 (2019).

27. 27.

d’Amour, C. B. et al. Future urban land expansion and implications for global croplands. Proc. Natl Acad. Sci. USA 114, 8939–8944 (2017).

28. 28.

Wu, Y., Shan, L., Guo, Z. & Peng, Y. Cultivated land protection policies in China facing 2030: dynamic balance system versus basic farmland zoning. Habitat Int. 69, 126–138 (2017).

29. 29.

Stokes, E. C. & Seto, K. C. Principles for minimizing global land impacts of urbanization. Technol. Archit. Des. 3, 5–10 (2019).

30. 30.

Nangini, C. et al. A global dataset of CO2 emissions and ancillary data related to emissions for 343 cities. Sci. Data 6, 180280 (2019).

31. 31.

Xi, F. et al. Substantial global carbon uptake by cement carbonation. Nat. Geosci. 9, 880–883 (2016).

32. 32.

Zhang, G. J., Cai, M. & Hu, A. Energy consumption and the unexplained winter warming over northern Asia and North America. Nat. Clim. Change 3, 466–470 (2013).

33. 33.

Li, X. et al. A dataset of 30-meter annual vegetation phenology indicators (1985–2015) in urban areas of the conterminous United States. Earth Syst. Sci. Data 12, 357–371 (2019).

34. 34.

Doxsey-Whitfield, E. et al. Taking advantage of the improved availability of census data: a first look at the gridded population of the world, version 4. Pap. Appl. Geogr. 1, 226–234 (2015).

35. 35.

Bontemps, S. et al. Consistent global land cover maps for climate modelling communities: current achievements of the ESA’s land cover CCI. In Proc. ESA Living Planet Symposium 9–13 (ESA, 2013).

## Acknowledgements

This research was funded by the National Key R&D Program of China (grant no. 2017YFA0604404 and grant no. 2019YFA0607203). Z.Z. was supported by the start-up fund provided by Southern University of Science and Technology (grant no. G02296001). We thank the French ANR Convergence Institute CLAND project for support. We thank many students (for example, Z. Lin) for their days and nights validating our GAUD product via high-resolution satellite image interpretation. We also thank K. C. Seto and M. Hansen for their constructive comments on this paper.

## Author information

Authors

### Contributions

X. Liu, Xia Li and Z.Z. designed the research; Y.H. and X.X. performed analysis; X. Liu, Xuecao Li and Z.Z. wrote the draft; and all the authors contributed to the interpretation of the results and the writing of the paper.

### Corresponding authors

Correspondence to Xia Li or Zhenzhong Zeng.

## Ethics declarations

### Competing interests

The authors declare no competing interests.

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

## Extended data

### Extended Data Fig. 1 The box-plot of the kappa coefficient for fifteen urban ecoregions in 1985 (a) and 2015 (b).

GHSL: global human settlement layer; GAIA: Global Artificial Impervious Areas; GUL: global urban land; GUF: global urban footprint; GAUD: global annual urban dynamics. Note: kappa with negative value suggests the urban product has notable overestimation or underestimation in particular ecoregions.

### Extended Data Fig. 2 Overall accuracy of mapped urban land expansion years during periods of 1985–2000 and 2000–2015.

Cities used for validation were randomly selected globally with different sizes, biomes, and climate zones.

### Extended Data Fig. 3 Detected urban land expansion years compared with the Landsat-derived time series data for six representative cities.

a, Shanghai (China), b, Chicago (USA), c, Tianjin (China), d, Paris (France), e, Moscow (Russia), and f, Bangkok (Thailand).

### Extended Data Fig. 4 Detected green recovery years compared to the Landsat derived time series data for four representative cities.

a, Beijing (China), b, New York (USA), c, Tokyo (Japan), d, Baltimore (USA).

### Extended Data Fig. 5 Comparison of growth rates between urban extent and population.

Left: ratio of growth rates between urban area and population. Right: scatter plot of total growth rate versus urban growth rate for each country.

### Extended Data Fig. 6 Trend of urban expansion in China, India, and USA regarding the normalized urban area relative to 2015.

Grey lines represent each urban cluster while colored lines represent China, India or the USA.

## Supplementary information

### Supplementary Information

Supplementary Methods 1–3, Figs. 1–15 and Tables 1 and 2.

## Rights and permissions

Reprints and Permissions

Liu, X., Huang, Y., Xu, X. et al. High-spatiotemporal-resolution mapping of global urban change from 1985 to 2015. Nat Sustain 3, 564–570 (2020). https://doi.org/10.1038/s41893-020-0521-x

• Accepted:

• Published:

• Issue Date:

• ### The relationship between land surface temperature and artificial impervious surface fraction in 682 global cities: spatiotemporal variations and drivers

• Qiquan Yang
• , Xin Huang
• , Jie Yang
•  & Yue Liu

Environmental Research Letters (2021)

• ### Analyzing the Spatiotemporal Uncertainty in Urbanization Predictions

• Jairo Alejandro Gómez
• , ChengHe Guan
• , Pratyush Tripathy
• , Juan Carlos Duque
• , Santiago Passos
• , Michael Keith
•  & Jialin Liu

Remote Sensing (2021)

• ### Adaptive Evolution in Cities: Progress and Misconceptions

• Max R. Lambert
• , Kristien I. Brans
• , Simone Des Roches
• , Colin M. Donihue
•  & Sarah E. Diamond

Trends in Ecology & Evolution (2021)

• ### High-Resolution Maps of Material Stocks in Buildings and Infrastructures in Austria and Germany

• Helmut Haberl
• , Dominik Wiedenhofer
• , Franz Schug
• , David Frantz
• , Doris Virág
• , Christoph Plutzar
• , Karin Gruhler
• , Jakob Lederer
• , Georg Schiller
• , Tomer Fishman
• , Maud Lanau
• , Andreas Gattringer
• , Thomas Kemper
• , Gang Liu
• , Hiroki Tanikawa
• , Sebastian van der Linden
•  & Patrick Hostert

Environmental Science & Technology (2021)

• ### Crop Production Allocations for Saving Water and Improving Calorie Supply in China

• Yilin Liu
• , La Zhuo
• , Xi Yang
• , Xiangxiang Ji
• , Zhiwei Yue
• , Danyue Zhao
•  & Pute Wu

Frontiers in Sustainable Food Systems (2021)