Perceiving progress and imbalance of environmental SDG indicators in China using big data

ABSTRACT

In the past decades, the environment has changed drastically in China with rapid economic development. Lack of data, especially data with spatiotemporal information, has however hindered evaluating progress of environmental indicators in the global Sustainable Development Goal (SDG) framework. Here, we evaluate and explore geospatial information of 20 environmental SDG indicators from six SDGs in China from 2010 to 2020 using big data. We show China has improved rapidly in most of the 20 indicators, except for indicator 13.2.2 (CO₂ emission). 63.5% of China’s land mass showed improved states for the SDG indicators. By 2020, four of the indicators were found to have achieved their 2030 targets. It is predicted from the existing data that 10 of the 20 indicators can achieve their targets smoothly by 2030, but two indicators will lag behind. We quantitatively show China’s progress towards the environmental SDG indicators, and put forward timelines and suggestions for policy makers.

KEYWORDS:

• Big earth data
• sustainable development goals
• spatial information
• environment

1. Introduction

The Sustainable Development Goals (SDGs) are at the core of ‘Transforming our world: the 2030 Agenda for Sustainable Development’ adopted by the 193 Member States of the United Nations (UN) in September 2015. These SDGs represent the commitment to address tri-dimensional social, economic, and environmental issues facing humankind to achieve sustainable development. The achievement of the SDGs requires monitoring and evaluation (Castro, Fernández, and Colsa 2021). Nearly half of the time has passed since the Agenda was adopted in 2015 yet gaps remain in measuring and monitoring the progress of SDG indicators at national and global levels (Schmidt-Traub et al. 2017; Xu 2020; Mastrángelo 2019; Huan et al. 2021) About 41% of the total 230 + indicators have no available data as at December 2022 (IAEG-SDG 2022) Furthermore, even countries with available data only have a small portion of data over time and space (UN 2021).

In the current SDG framework, China lacks an objective evaluation of progress and gaps. The rapid urbanization and industrialization of China over the past decades have generally promoted economic and social issues, but also have put great pressure on the environment. Meanwhile, China has taken positive measures in recent years to restore and improve the environment (Lu 2020; Wang et al. 2021b). Therefore, nature and the built environment may have undergone drastic changes and need frequent observation and evaluation. Most of the existing studies have examined environmental changes in China using local indicators and standards (Xu 2020; Wang et al. 2021b; Wang et al. 2020a; Jiang et al., 2022) but it is difficult to judge the distance to the targets of the 2030 Agenda. Therefore, there is an urgent need to objectively evaluate China’s environmental SDG indicators to aid scientific research and decision-makers.

Methodologies for data collection for the SDG indicators are developed and implemented by national statistical agencies, UN, and other organizations across the world (Huan et al. 2021; Mulholland, Dimitrova, and Hametner 2018) However, their implementation and use also face several important challenges. The main problem is that the process is time-consuming and updates are infrequent (Sachs 2020; Burke et al. 2021) The integration of geospatial and statistical information is also urgently needed for policymakers and the public to understand and respond to spatial imbalances and changes over different time horizons (UN 2020; UN 2021; Scown 2020).

Satellite observations have great potential to provide national and global scale environmental data for the SDG indicators (Burke et al. 2021; GEO et al. 2017; Anderson 2017; ESA 2018; Estoque 2020; Persello et al. 2022). It provides repeatable, long-term, and low-cost data. Satellite observations could contribute significantly to monitoring six SDGs (UN 2021), namely SDG 2 Zero hunger (Whitcraft et al. 2019), SDG 6 Clean water and sanitation (Mariathasan, Bezuidenhoudt, and Olympio 2019; Mulligan et al. 2020), SDG 11 Sustainable cities and communities (Aquilino et al. 2020), SDG 13 Climate action (Boesch et al. 2021), SDG 14 Life below water (Wang 2021a) and SDG 15 Life on land (Prince 2019; Mondal, McDermid, and Qadir 2020). In most cases, satellite data need other auxiliary statistical and basic geographic data, and often field observations for verification, to be translated into indicators (Whitcraft et al. 2019; Andries et al. 2019; Kavvada 2020; Chen et al. 2020).

This study is a novel attempt at evaluating the progress of the environmental SDG indicators using big data at the national scale. We combine data (Table 1 and Table S1) from satellite observations, field surveys, statistics, navigation, and basic geography (i.e. terrain, population density, and boundary), collectively known as Big Earth Data (Guo 2018; Guo 2019; Guo et al. 2021). Big Earth Data was employed to evaluate SDG indicators in the annual reports of Big Earth Data in support of Sustainable Development Goals since 2019, focusing on 6 SDGs including SDG 2, 6, 11, 13, 14, 15. However, in these reports indicators are usually calculated and analyzed one by one, and here it is a new attempt to integrate different indicators together with spatial–temporal information. The study makes three contributions: first, the values, states, and trends of environmental SDG indicators were newly evaluated or updated. Second, compared with traditional national or provincial level statistics, grid cell information was provided to better understand the nature and characteristics of the SDG indicators. Third, the state of the SDG indicators in 2030 was predicted using current data, which provides an opportunity to adjust work and policy according to real-time progress.

2. Data and method

The methods, data sources, standards, and results for all 20 environmental SDG indicators are displayed in Tables 1 and 2. These data sources include remote sensing, statistics, surveying, and field observation. In most cases, we use different data sources to make cross-validations. The employed methods are mainly developed from publications or UN’s definition of the indicators to ensure data production quality. Selection, normalization, and spatialization of the indicator data are presented below.

Table 1. Data and methods of the evaluated indicators.

SDG	Indicators	Evaluated items in this work	Data	Method	Resolution	Quality assurance
SDG 2	2.4.1	Efficiency in terms of production, irrigation, and excess fertilizer	1. Remote sensing products including the 1:100,000 National Land use/cover; vegetation index data; 2. Statistical data including national crop harvest area and yield, irrigation area, and fertilizer application; 3. Field survey data including pollution census and agricultural census data.	By integrating remote sensing, spatial allocation models, we estimated three sub-indicators of SDG 2.4.1for China from 2010 to 2015.	30 m	The method and part of data come from the reference (Zuo et al. 2018). In this work we extended the data from 2010 to 2015 using the method.
SDG 6	6.3.2	Proportions of lakes with water clarity level I and II	1. Land reflectance, land surface temperature, and Normalized Difference Vegetation Index (NDVI) during 2000-2019. 2. Precipitation data of the Tropical Rainfall Measuring Mission (TRMM) during 2000-2019; 3. In situ observation data in the 2236 lakes.	We developed a remote sensing algorithm of water transparency (Secchi Disk Depth, SDD) applicable to the land reflectance of MODIS (MOD09GA):	500 m	The method and data come from the reference Liu, Duan, and Loiselle 2020), and here we extended the data from 2018 to 2019 using the method. 2236 in situ SDD of China’s lakes were used for validation.
	6.4.1	Water-use efficiency	1. Water use of 30 provinces in 2015 and 2018 from the China Water Resources Bulletin. 2. Value added by industry and the cultivated and irrigated land of 30 provinces in 2015 and 2018 from the China Statistical Yearbook.	The water-use efficiency algorithm was calculated as follows: WUE = A_we × P_A + M_we × P_M + S_we × P_S where WUE is the water-use efficiency; Awe is the irrigated agricultural water-use efficiency; Mwe is the industrial water-use efficiency; Swe is the services water-use efficiency; PA is the proportion of water used by the agricultural sector over the total use; PM is the proportion of water used by the industrial sector over the total use; and PS is the proportion of water used by the service sector over the total use.	Provincial	Using the suggested method by UN (GEMI 2019)
	6.5.1	Degree of Integrated Water Resources Management (IWRM)	Survey data, from 2015-2017 and 2018-2020 China Integrated Water Resources Management Assessment Questionnaire. Data on Chinese provincial boundary cross-section water volume and quality, data on water intake, drinking water and water function zone, from 2015 to 2020, from China Statistical Yearbook.	Using the questionnaire survey methodology and spatial analysis technology, a semi-quantitative assessment of the IWRM level in China was made.	Provincial	Using the suggested method by UNEP (UNEP 2020)
	6.6.1	Areas of vegetated wetlands	1. Landsat TM/ETM+/OLI images in 2010, 2015, and 2020 and ZY-3 images 2. DEM, vector data of administrative divisions, 1:1,000,000 vegetation type map, climatic zones map, and global water distribution maps 3. Ground survey samples, governmental statistics, and monitoring data	We classified vegetated wetlands with a Hybrid Object-oriented and Hierarchical Classification (HOHC) approach.	30 m	The overall accuracy of classification, as validated against over 5,000 ground survey samples (Mao, Wang, and Du 2020) is above 92%.
SDG 11	11.2.1	Proportion of population that has convenient access to public transport	1. Public transportation network) data in China, 2015, 2018 and 2020. 2. Land use data with 100 m resolution in China, 2015, 2018 and 2020. 3. 1% national population sample survey data, 2015 and census data, 2020.	We first redistributed the resident population from an administrative division by the 1 km grid using a random forest spatial downscaling model. Next, we extracted the public transportation stations with spatial attributes from China’s public transportation network data, and a 500 m buffer zone was created for each of the stations. We superimposed this buffer layer on the above-mentioned gridded population data to calculate the population within the buffers by gender and age group. Finally, we superimposed those layers of the population with convenient access to public transportation on the spatial data of urban areas extracted from land-use data.	100 m	Using the suggested method by UN-Habitat (https://www.local2030.org/library/60/SDG-Goal-11-Monitoring-Framework-A-guide-to-assist-national-and-local-governments-to-monitor-and-report-on-SDG-goal-11-indicators.pdf)
	11.3.1	Ratio of Land Consumption Rate to Population Growth Rate (LCRPGR)	1. Population data of 433 Chinese cities with over 300 000 urban inhabitants from 2010 to 2018 (UN 2018). 2. Urban impervious surface products with a resolution of 30 m from 2010 to 2018.	Urban impervious surface extracted from remote sensing was converted to built-up area to calculate LCRPGR which is the ratio of Land Consumption Rate to Population Growth Rate following UN terms (UN-Habitat 2018).	30 m	The overall accuracy of unban land mapping is 88.03% (Sun, Liu, and Yu 2019).
	11.4.1	The per capita expenditure of nature protected areas	1. Statistics of revenue, expenditure, area, and visitor volume of 51 natural protected sites in China from 2006 to 2017 (from the website of the Ministry of Housing and Urban-Rural Development of the People’s Republic of China), as well as field surveys 2. Landsat 5 and Landsat 8 satellite images from 2006 to 2017.	We took total expenditure per capita and per unit area to measure the sustainable development of demonstration sites.	Provincial and 30 m	Data come from National Statistics. The method was proposed by Wang et al. (2018) and used here
	11.5.1	Numbers of deaths, missing persons and affected persons per 100,000 population	1. Annual statistics on disaster losses since 2010, including affected population and death toll at prefectural level. 2. Annual statistics on total population since 2010, at prefectural level.	We used both statistical data and spatial data to present multi-dimensional monitoring results of indicator 11.5.1. The formulas used in this study are as follows: $\mathrm{AP}={\displaystyle \frac{\mathrm{Number}\mathrm{of}\mathrm{affected}\mathrm{population}\ast 100000}{\mathrm{Total}\mathrm{population}\mathrm{of}\mathrm{the}\mathrm{year}}}$ $\mathrm{DMP}={\displaystyle \frac{\mathrm{Number}\mathrm{of}\mathrm{death}\mathrm{\&}\mathrm{missing}\mathrm{population}\ast 100000}{\mathrm{Total}\mathrm{population}\mathrm{of}\mathrm{the}\mathrm{year}}}$ Where AP is the affected population, DMP is the death/missing population.	Prefectural	Data come from National Statistics
	11.5.2	Direct economic loss in relation to global GDP attributed to disasters	1. Annual statistics on direct economic losses at prefectural level. 2. Annual statistics on total population, gross Domestic product (GDP), gross regional product (GRP) since 2010, at prefectural level.	On national scale, the formulas used in this study are as follows: $\mathrm{Ratio}={\displaystyle \frac{\mathrm{Direct}\mathrm{economic}\mathrm{loss}}{\mathrm{GDP}\mathrm{of}\mathrm{the}\mathrm{year}}\times 100\text{\%}}$ on prefectural scale, the economic loss is changed to local loss, and the GDP is changed to local GRP.	Prefectural	Data come from National Statistics
	11.6.2	Mean ground-level PM2.5 concentration	1. Aerosol optical thickness and Vegetation index data from satellite. 2. The meteorological parameters from China’s National Meteorological Administration. 3. Ground monitoring data of national control distribution points of China Environmental Monitoring Station. 4. The gridded Population density data with 1 × 1 km resolution (CIESIN 2020)	We calculated the average annual concentration of PM2.5 in China from 2010 to 2020 according to the population weight.	1 km	Cross validation of the predicted vs. ground observed PM2.5 concentrations show that R^2= 0.78 (Zeng, Tao, and Chen 2020).
	11.7.1	The share of open public space area in built-up areas	1. Navigation data for China acquired in 2015 and 2018, which includes public green spaces, public squares, and roads at all levels. 2. Land use data for China in 2015 and 2018 with a 100-meter resolution. Data was obtained from China’s Land use Status Remote Sensing Monitoring Database, CAS.	Open public green spaces and squares are extracted within defined urban boundaries.	100 m	The method was published in (Wang et al. 2020b), and in this work we extended the data from 2015 to 2018 using the method.
SDG 13	13.1.1	Numbers of deaths, missing persons per 100,000 population	1. Annual statistics on disaster losses since 2010, including affected population and death toll at prefectural level. 2. Annual statistics on total population since 2010, at prefectural level.	This study used both national data and prefectural data to present multi-dimensional monitoring results of indicator 11.5.1.	Prefectural	Indicators are calculated using national statistics
	13.2.2	CO2 emissions per capita	1. China’s carbon emission data (2011-2018) from World Bank (WB); Emissions Database for Global Atmospheric Research (EDGAR) version 6.0 (2010-2020) (Crippa et al. 2021), Carbon Emission Account & Datasets (CEADs) emission inventory (2010-2019). 2. Open-Data Inventory for Anthropogenic Carbon dioxide (ODIAC) products (Oda and Maksyutov 2015) in 2010, 2015 and 2019 with 1 × 1 km resolution are used. 3. The gridded Population density data with 1 × 1 km resolution (CIESIN 2020)	We used the CO2 emission data of China from WB, EDGAR and CEADs. The three sets of data differ slightly, but the temporal trend of the values is similar. The data of WB was used for national scale. On local scale, we used the ODIAC data which provided 1 × 1 km resolution CO2 emission in 2010, 2015 and 2019. It was divided with population density data to calculate the local CO2 emission per capita.	Provincial	The changing trends during 2010-2020 of three datasets agree well with each other.
SDG 14	14.1.1	The average abundance of floating debris;	1. Floating marine debris data primarily from the Bulletin on the State of China’s Marine Ecological Environment 2018 (Ministry of Ecology and Environment of the People’s Republic of China, 2019). 2. The published data (Zhou et al. 2016)	We mapped the floating debris in China’s coastal waters using the data of 2018 from marine bulletins.	Point Survey	Data from National Statistics and published results.
	14.2.1	Areas of mangrove forests	1. Sentinel-2 images covering mangrove regions in China in 2015, 2018, and 2020. 2. Digital Elevation Model, 1:1,000,000 vegetation type map, coastline vector data. 3. Field samples, governmental statistics and monitoring data.	An object-oriented method and random forest classifier were used.	10 m	Upon validation against over 5000 field samples (Jia, Wang, and Zhang 2018), the classification accuracy of the mangrove forest datasets was about 95%.
	14.7.1	Sustainable fisheries as a proportion of GDP	1. Statistical bulletin of China’s marine economy (2010-2020). 2. Sentinel-2 satellite images (10 m resolution) in 2015 2018 and 2020. 3. global water distribution products (2015, 2018 and 2020), and Chinese coastlines data. 4. Field survey samples, industrial statistics and monitoring data.	Marine fishery as a proportion of GDP (MFP) was calculated by dividing GDP from marine fisheries with total GDP.	10 m	The method comes from the reference (Mao, Wang, and Du 2020). Upon validation against more than 6,000 field survey samples, the overall classification accuracy for datasets was above 95%.
SDG 15	15.3.1	The proportion of land that is net degraded over total land area	1. Land cover imagery from 2000 and 2018 with a 300-meter spatial resolution provided by the European Space Agency (ESA). 2. MODIS Enhanced Vegetation Index (2000-2018) with a 250 m spatial resolution. 3. Soil Carbon (0-30 cm) data with a 250 m spatial resolution provided by International Soil Reference and Information Centre (ISRIC).	Three sub-indicators, including land cover, productivity, and soil carbon content, were selected for analysis and calculated at global scales.	300 m	Analysis was performed following the Good Practice Guidance for Assessing SDG Indicator 15.3.1 issued by the United Nations Convention to Combat Desertification (UNCCD) (Neil, Jacqueline, and Glenn 2019).
	15.4.2	Mountain Green Cover Index (MGCI)	1. Landsat-8 OLI surface reflectance data at 30 m spatial resolution (2015-2020). 2. ASTER GDEM Version 2 Data at 30 m. 3. UNEP-WCMC 500 m mountain classification data. 4. FROM-GLC global sample sets for land cover classification in 2015.	We monitored MGCI spatiotemporal dynamics in the years of 2015 and 2020.	30 m	For the published data in 2017 using this method, the user’s and producer’s accuracy are 97.02% and 97.11%, respectively (Bian et al. 2020).
	15.5.1	Red List Index	The first assessment of the red lists for higher plants was gathered from China Species Red List, Vol. 1: Red List. The second assessment was obtained from the China Biodiversity Red List—Higher Plants with some minor revisions from the special issue of Higher Plants of China published in Biodiversity Science. The two assessments of terrestrial mammals and birds were obtained from China Species Red List, Vol. 2: Vertebrates and the Chinese Red List of Vertebrates.	The data were derived according to the IUCN classification criteria for threatened factors. A total of 3,948 higher plants, 568 terrestrial mammals, and 1,213 birds were used in the calculation of the Red List Index.	Surveying	Data from National Statistics. Part of the work about RLI of plants was published (Chen et al. 2020).

Table 2. Results and standards of the evaluated indicators.

SDG	Indicators	Evaluated values and units	Results of the evaluated items in China		Bounds			China’s state from other sources accessed in March 2022: value (year)
			Year	Value	Upper	Lower	Justification basis	UN	SDG tracker	SDSN value
SDG 2	2.4.1	Efficiency in terms of cereal yield (CY, kg/hec), crop kilocalorie production (CKP, 10^9 M cal), irrigation water intensity (IWI, m^3/M cal) and excess fertilizer application (kg/M cal)	2010	CY: 5526; CKP: 1.9;IWI: 79; Excess N:6.1;Excess P:1.5	CY = 8000;Excess N = 0	CY = 800; Excess N = 14	World Bank; SDSN (Sachs et al. 2021)	NA	NA	Sustainable Nitrogen Management Index (SNMI) = 0.7 (2015)
			2015	CY: 5896; CKP: 2.1;IWI: 78; Excess N:5.8;Excess P:1.6
			2018	CY: 6081
SDG 6	6.3.2	Proportions of lakes with water clarity level I and II	2010	60.1%	100%	0	Percentage data	NA	NA	NA
			2015	70%
			2019	69.1%
	6.4.1	Water-use efficiency (RMB/m^3)	2015	116.7	341 RMB/m^3	0	World Bank (Values of high income countries in 2015)	155.4 RMB/m^3 (2018)^1	115.3 RMB/m^3 (2017)	NA
			2018	143.3
	6.5.1	Degree of Integrated Water Resources Management (IWRM)	2015/17	75	100	0	Percentage data	NA	NA	NA
			2018/20	79
	6.6.1	Areas of China’s vegetated wetlands (10^5×km^2)	2010	1.71	Increase 5%	Reduce 20% compared to 2015	Wetlands International (Wetlands International 2021)	NA	0.66% (2017)	NA
			2015	1.64
			2020	1.63
SDG 11	11.2.1	Proportion of population that has convenient access to public transport	2015	64.3%	100	0	Percentage data	NA	NA	NA
			2018	80.6%
			2020	90.2%
	11.3.1	Ratio of Land Consumption Rate to Population Growth Rate (LCRPGR)	2010-13	1.57	0.6	2.5	UN-Habitat (UN-Habitat 2018);(Estoque et al. 2021)	NA	NA	NA
			2013-15	1.63
			2015-18	1.36
			2018-20	1.21
	11.4.1	The per capita annual expenditure on natural protected areas (PEP, RMB/person)	2006	16	Per capita 30 PPP^2 USD (126 RMB)	0	UNESCO (UNESCO-UIS 2021)	NA	NA	NA
			2010	24
			2015	38
			2017	51
	11.5.1	Numbers of deaths, missing persons (DMP) and affected persons (AP) per 100,000 population	2010	31200 (AP); 0.60 (DMP)	- 60%	0 (No reduction)	Sendai Framework for Disaster Risk Reduction 2015-2030	NA	DMP = 0.06 AP = 453 (2019)	NA
			2015	13500(AP); 0.07 (DMP)
			2020	9800 (AP); 0.04 (DMP)
	11.5.2	Ratio of direct economic loss to GDP attributed to disasters	2010	1.3%	−50% (Reduce direct disaster economic loss)	0 (No reduction)	Sendai Framework for Disaster Risk Reduction 2015- 2030	NA	NA	NA
			2015	0.37%
			2020	0.36%
	11.6.2	Annual mean ground-level PM2.5 concentration (μg/m^3)	2010	49.1	10	87	World Health Organization (WHO 2021) and SDSN (Sachs et al. 2021)	48.8 (2016)	52.66 (2017)	48.6 (2019)
			2015	51.2
			2019	38.1
	11.7.1	The share of open public space area in built-up areas	2015	17.98%	30%	11.2%	UN Habitat (UN Habitat 2018)	NA	NA	NA
			2018	19.5%
SDG 13	13.1.1	Numbers of deaths, missing persons and affected persons per 100,000 population	2010	3.2 (AP); 0.6 (DMP)	Substantial reduction by 2030 compared to 2005- 2015	No reduction	Sendai Framework for Disaster Risk Reduction 2015-2030	NA	DMP = 0.06 AP = 453 (2019)	NA
			2015	1.3 (AP); 0.7 (DMP)
			2020	1.0 (AP); 0.037 (DMP)
	13.2.2	Total CO2 emissions per year from WB, CEADs and EDGAR mt), per capita CO2 emissions (PC) (t/capita)	2010	6.8 (EDGAR);6.3 (CEADs)	0 (PC)	20 (PC)	SDSN (Sachs et al. 2021)	12300 mt in total (2014)	NA	PC = 7.1 (2019)
			2011	6.9 (WB); 7.4 (EDGAR)
			2015	7.1 (WB); 7.7 (EDGAR) 7.1 (CEADs)
			2018	7.4 (WB); 8.0 (EDGAR)7.3 (CEADs)
			2020	8.2 (EDGAR)
SDG 14	14.1.1	The average abundance of floating debris (pieces/km^2)	2010	3207	0	7700	Research Reference (Agamuthuet al. 2019)	NA	NA	NA
			2018	2379
	14.2.1	Areas of mangrove forests in China (×10^3 ha)	2015	23.2	Increase 20% area	Decrease 10% area	Global Mangrove Alliance (Waltham et al. 2020)	NA	Increase 0.86% to 2000 (2016)	NA
			2018	25.0
			2020	27.1
	14.7.1	Marine fishery as a proportion of GDP (MFP); Areas of aquaculture ponds in China’s coastal waters (AAP); (×10^4 km^2)	2010	0.68% (MFP)	0.1% (MFP)	8% (MFP)	FAO (Cai, Huang, and Leung 2019; FAO, 2020)	NA	NA	NA
			2015	0.63% (MFP); 1.07 (AAP)
			2018	0.52% (MFP); 0.94 (AAP)
			2020	0.46% (MFP); 0.90 (AAP)
SDG 15	15.3.1	The proportion of land that is net degraded over total land area (×10^4 km^2)	2000-15	198	0	100%	SDGs definition to achieve a land degradation-neutral world	NA	Land that is degraded = 27% (2015)	NA
			2015-18	114
	15.4.2	Mountain Green Cover Index of China	2015	82.43%	100%	0	Percentage data and FAO	84% (2018)	84.56% (2018)	NA
			2020	82.05%
	15.5.1	Red List Index of China	2004	0.62	1	0.6	Percentage data and SDSN(Sachs et al. 2021)	0.73 (2021, from model)	0.73 (2021, from model)	0.7 (2020)

2.1. Indicator selection

From the global indicator framework, two factors were considered in the selection of indicators for the environment. Firstly, indicators should effectively reflect the changes in the environment, covering the natural environment such as water, air, soil, and biology, as well as the built environment such as agriculture, residence, and transportation. There is also climate change, which has important impacts on the whole environment. Secondly, the data available for the indicators should have abundant spatial and temporal information. Many environmental indicators also happen to be the areas where satellite observation-based Big Earth Data can play an important role in data acquisition. Therefore, we try to select indicators that can be observed by satellite.

Using Big Earth Data, 20 SDG indicators that are closely related to the natural and built environment were selected and evaluated within the 2010–2020 period (Table 2). We matched the evaluation standards of the SDG indicators with their 2030 targets (Methods), which gave a more objective evaluation of the state and distance to each SDG. The score for each SDG indicator was normalized based on thresholds ranging from 0 (worst performance) to 100 (fully achieved) (Xu 2020; Sachs et al. 2021). An SDG indicator was classified as being very close to or having achieved the 2030 target if the score was no less than 90 (Table 3).

Table 3. Progress of the evaluated 20 environmental SDG indicators in China during the 2010-2020 period.

2.2. Normalization of the selected indicators

To make the data comparable across indicators, each indicator was rescaled from 0 to 100 with 0 denoting the worst performance and 100 describing the optimal performance. The equation for normalization is given as: $x^{\mathrm{\prime }}={\displaystyle \frac{x-min\left(x\right)}{max\left(x\right)-min\left(x\right)}\times 100}$where x’ is the normalized score, x is the value of the indicators, min(x) is the lower bound, and max(x) is the upper bound.

The bound settings refer to the following standards. (1) Global indicator framework has clear targets and definitions, such as zero growth in land degradation; the Sendai Framework for Disaster Reduction 2015-2030; (2) The targets were suggested by UN or international agencies, such as PM_2.5 concentration, wetland change, etc. (3) The limits of 0 and 100 were defined for some of the indicators measured by percentages. (4) The target was defined using the average value of a few top countries (such as top 5%) in the indicator ranking (excluding the extreme value of 2.5%) that was taken from international agencies’ databases or the Sustainable Development Solutions Network (Sachs et al. 2021). More details for the bounds and sources are listed in Table S2.

2.3. Definition of the states and trends of the indicators

An indicator was classified as being very close to or having achieved the 2030 target (score ≥ 90), showing a small gap to the targets (90 > score ≥ 70), facing a big gap to the targets (70 > score ≥ 50), and facing a huge gap the targets (score < 50).

Two periods were considered to monitor the change in indicators: 2010-2015 and 2015-2020. The results in Table 1 for the change in indicators were divided according to three situations, including rising values (improvement), stable values, and declining values (deterioration). Among them, indicator results were labeled as ‘improving’ if the score of the indicator increased greater than 2. Moreover, the indicator’s progress was defined as ‘deteriorating’ if the decrease was greater than 2, and ‘stable’ if the range was between ±2.

2.4. Generation of grid cell map of the indicators

The grid cell score maps were generated based on the following steps. (1) The indicators acquired from satellite imagery with spatial information for thematic maps were normalized and resampled to 1 km resolution. (2) The indicators acquired from statistics or survey data included indicators 6.4.1 (water use efficiency), 6.5.1 (water resource management), 11.5.1 (loss in disasters), and 11.5.2 (economic loss in disaster). These indicators were examined at the urban or provincial scale and projected to a 1 km resolution map of the province or city. (3) Indicators that did not have sufficient data in 2010, 2015, or 2020 were linearly interpolated from data in the neighboring period. (4) All of the available indicator score maps were overlaid together, and the value of each grid was the average score of the corresponding indicator. The weighted value for each indicator was based on the density of population, considering the relationship of different indicators with human activity. Indicators 14.1.1 and 15.5.1 were not included in the final map because the spatial information was only available for one year.

2.5 Prediction for 2030

Whether the 2030 Agenda can be achieved is a matter of common concern (UN 2022). We predicted 2030 values for the indicators. The prediction for 2030 was divided into three categories according to different indicators. First, the four indicators that had already achieved their targets of 2030 in 2020 and had a consistent trend were not subjected to predictions for 2030; Second, there was an attempt to use different curves or lines to find the best fit and prediction expression for the indicators with sufficient annual observation data for the 2010–2020 period. In the third category, the indicators with only two or three observations were predicted using a linear method.

3. Results

3.1. State and progress of the 20 environmental SDG indicators

Most of the measured 20 SDG indicators improved during the 2010-2020 period in China albeit at different rates (Table 3). Indicator 11.3.1 (land use efficiency) decreases during 2010–2015 due to fast spatial expansion coupled with slowdown in population in urban areas. There has been a notable decrease in the areal extent of new urban land since 2015, reversing the trend of the indicator. Indicator 13.2.2 (CO₂ emission) first declined and then remained stable during this period. Especially from 2015 to 2020, in the first five years after the implementation of the 2030 Agenda, none of the 20 SDG indicators deteriorated.

By 2020, four of the SDG indicators, including indicator 11.2.1 (public transport access), 11.5.2 (economic loss from natural disasters), 15.3.1 (land degradation), and 15.4.2 (Mountain Green Cover Index (MGCI)), were found to be very close to or had achieved their 2030 targets value. Although the average MGCI in China was 82.05% in 2020, the indicator’s target was still considered close or achieved. This is because the values in the western provinces of China were much lower than other provinces since a large proportion of mountains were higher than 4000 m above sea level and many were covered by glaciers and snow. Therefore, based on our results and spatial information, we believe that the value of 82.05% approaches the limit of the indicator’s value for China. In addition, indicators 11.5.1 and 13.1.1 (human loss from natural disasters) also reached their 2030 targets based on their 2020 values. However, the realization standards for these two SDG indicators need to measure the average values during the period 2020–2030 according to the Sendai Framework for Disaster Reduction (UNDRR 2015) because the annual values for the indicators often fluctuate greatly. Therefore, they were only considered to have temporarily realized their 2030 targets in 2020 in this study.

As of 2020, three SDG indicators, i.e. indicators 6.4.1 (water-use efficiency), 11.4.1 (per capita expenditure on heritage), and 15.5.1 (Red List Index (RLI)) have constantly improved, but there were still huge gaps to reach their 2030 targets.

3.2. Spatial imbalance of SDG indicators

Spatial information helps to understand natural differences (Liu 2021). The SDG indicators measured by different data sources have different resolutions (Table 1), which introduces data interoperability challenges. In order to improve the presentation and evaluation, the 20 measured SDG indicators were integrated into a score and normalized (Figure 1a) to a unified 1 km resolution map for the years 2010, 2015, and 2020.

Figure 1. Spatial differences and score maps of China. a, sketch map of the method of integrating different SDG indicator’s value into a score. b-d, score map in 2010 (b), 2015 (c) and 2020 (d).

It can be observed from Figure 1b-d that in northwest China the scores were generally very low in 2010, 2015, and 2020. In particular, for the areas in the rectangle, there were mainly desert or Gobi Desert areas, which were the main sources of sand and dust with very high concentrations of PM₁₀ and PM_2.5. In addition, there were also arid high-altitude mountains with very low MGCI values.

In China’s middle and east (the three ellipses), the improvement of scores in the period 2015–2020 was more pronounced than in the period 2010–2015. The challenge of CO₂ emissions increased in the highly populated regions within the three ellipses, but the per capita CO₂ emissions may not increase due to the growing population in urban regions. Meanwhile, urban infrastructure in terms of public transport, disaster resistance, and land/water use efficiency improved, and PM_2.5 concentration decreased, so the scores in these eastern regions improved significantly. The remaining scattered red points in the ellipse in Figure 1(d) were mainly lakes, which means that the water clarity in these regions improved at a slow rate and challenges remain to reach the target corresponding to the indicator.

Changes in the scores during the 20102020 period are shown in Figure 2. In 63.5% of China’s land area, the score improved, for 26% it was stable, and in the remaining 10.5%, the score deteriorated (Figure 2a). To better interpret the results, the score map is combined with a land cover classification map (Liu and Zhang 2019) (Figure 2b). Land cover is roughly classified into five area types: built up area (city and town), cropland, vegetation (forest and grassland), water (water and wetland), and extreme environment area (desert, Gobi desert, glacier, and multi-year snow). Based on the combined maps, scores improved for 77.3% of the built up area, 71.3% of the water area, 68.9% of the vegetation area, 67.9% of the cropland areas, and 48.1% of the extreme environment area. Scores deteriorated less than 10% for built up area, water area, vegetation area, and cropland area, respectively, but deteriorated for 20.75% of the extreme environment areas. The remaining areas were stable. Even if extreme environment areas were excluded, it can be seen that the improvement in north China still lags behind that in the south (Figure 2b). There were more improvements of built up land, cropland, and vegetation areas, but less improvement in water areas in south China than in the north.

Figure 2. Score change maps during the 2010–2020 period on grid and provincial level. a, Score changes map on grid level. b, Score changes on different land cover types including built up land, cropland, vegetation, water, and extreme environment area. Score changes for extreme environment area are excluded in the map. c, Score ranks for the 31 provinces (Hong Kong, Macao, and Taiwan are not included because of incomplete data).

We ranked 31 provinces (including the four province-level municipalities of Beijing, Shanghai, Tianjin, and Chongqing. Hong Kong, Macao, and Taiwan are not included because of incomplete data) based on their averaged SDG indicator scores (Figure 2c). The extreme environment area was excluded in the calculation of the score for each province. The three eastern provinces Beijing, Zhejiang, and Guangdong ranked in the top. Beijing and Zhejiang scored more than 80 in 2020, which means that the average scores of these provinces are not far from achieving their 2030 targets on the whole for the 20 environmental SDG indicators. The three northern provinces of Jilin, Inner Mongolia, and Xinjiang scored lower than 60 and were ranked as the furthest in achieving their 2030 targets. Inner Mongolia and Xinjiang lagged behind while scores for indicators 6.4.1 (water-use efficiency), 11.7.1 (public space), and 13.2.2 (per capita CO₂ emission) were very low. Jilin scored low on indicators 6.4.1 and 11.7.1, but higher on indicator 13.2.2 than the other two provinces. Jilin was also the province that showed the slowest improvement during the 2010-2020 period, which caused the province to lag behind in 2020, although it was ahead of more than half of the other provinces in 2015. Most other provinces scored in the 60–80 range in 2020 and made significant improvements during the 2010–2020 period.

3.3. Predicted state in 2030

We have predicted 2030 values for the 20 environmental SDG indicators based on a trend analysis of existing monitoring data (Figure 3).

Figure 3. Trend analysis of the observed environmental indicators during the 2010–2020 period and the predictions for 2030. (a) trend analysis and prediction of indicators which have annual measurements; (b) linear analysis of indicators which only have 2 or 3 observations. The arrows in (a) and (b) represent indicator trends, and the translucent areas in (a) are the prediction zones.

The predicted states of the eight indicators were compared with predictions from other literature or national plans (Table 4). The predicted data included the GDP, population, and CO₂ emissions in 2030 obtained from scientific research. The national plans included long-term planning for agriculture, land use, and environmental protection for 2030, and they contained future goals and government action projects, so they were also good references for our results. The results show that the average absolute score deviation between our work and other predictions was 2.6, which means they agreed very closely. The most significant difference is PM_2.5 concentration, and our prediction is more optimistic (the score is 7 points higher) than the government’s planning.

Table 4. Comparison of our results with other predictions and the national plans for 2030.

Indicators	Predicted value for 2030	Source of previous prediction	Score of previous prediction	Predicted score in our work
2.4.1	Water consumption per hectare of farmland irrigation will be reduced by 10%, the yield will be increased 10% per hectare compared to 2020.	National high standard farmland construction plan 2021-2030 https://www.ndrc.gov.cn/fggz/fzzlgh/gjjzxgh/ 202111/t20211102_1302810_ext.html	80	80
6.3.2	proportion of clean water in main basin will be increased to 75%	Outline of national land planning (2016-2030) http://www.gov.cn/zhengce/content/ 2017-02/04/content_5165309.htm	75	80
6.4.1	the water use efficiency will be increased to 267 RMB/m^3	Outline of national land planning (2016-2030) World economic league table 2022 (CEBR 2021)	78	75
6.6.1	The area of wetland will increase 3.75% compared to 2015	Outline of national land planning (2016-2030)	95	95
11.3.1	The urban area will increase 14.3% during 2020-2030, and the urban population will increase 19.1%.	National population development plan (2016-2030) http://www.gov.cn/zhengce/content/ 2017-01/25/content_5163309.htm; Outline of national land planning (2016-2030)	88	92
11.6.2	The average urban PM 2.5 concentration should be lower than 35μg/m^3	Ambient air quality standards (GB3095-2012) http://www.syq.gov.cn/syq_doc/ uploadfile/1448933137210.pdf	68	75
13.2.2	The national per capita emissions is predicted to be about 10 tCO2 yr^−1	Wang et al. 2019	50	60
	The national per capita emissions is predicted to be about 6.7 tCO2 yr^−1	Sun, Liu, and Yu 2019	67
14.2.1	Areas of mangrove forests in China will increase 31%	Special action plan for mangrove protection and restoration (2020-2025) (in Chinese) http://www.mnr.gov.cn/dt/ywbb/ 202008/t20200828_2544809.html	100	100

3.4. Suggestions for decision-making

Based on the calculations and the predictions discussed, the average score values of the 20 evaluated indicators for 2015, 2020, and 2030 are respectively: 63.5, 74.3, and 85.6 (Figure 4), respectively. At the current rate, 10 of the 20 environmental SDG indicators are expected to achieve their 2030 targets. There are two indicators that have a huge gap to their 2030 targets, indicator 13.2.2 (per capita CO₂ emissions), and indicator 15.5.1 RLI (score around 60). China has pledged to reach peak CO₂ emissions before 2030, which means that indicator 13.2.2 will not improve much before that time. Global biodiversity is facing serious threats due to climate change and human activity (Cowie, Bouchet, and Fontaine 2022), and indicator 15.5.1 will be difficult to improve greatly in a short period. Above all, these two indicators will lag behind in 2030 and will need long-term attention from policy makers.

Figure 4. Observed and predicted states of the 20 indicators. a,b, observed states of indicators in 2015 (a) and 2020 (b). c, predicted state in 2030. The red line in (a-c) is the average score for the year.

Aside from the ten SDG indicators that are likely to reach their 2030 targets and the two indicators that are lagging behind (13.2.2 and 15.5.1), there are two SDG indicators close to achieving their 2030 targets based on our predictions: indicator 14.1.1 (floating debris) will score 86, and indicator 6.5.1 (water resource management) will score 87. Some non-governmental organizations (NGOs) in China have established a net plastic action network in 2019 to achieve zero marine plastic waste emissions by 2030. The progress of indicator 14.1.1 may be accelerated and achieved if this action can be effectively implemented. The other six indicators, including indicators 2.4.1 (sustainable agriculture), 6.3.2 (water clarity), 6.4.1 (water use efficiency), 11.6.2 (PM 2.5 concentration), 11.4.1 (per capita expenditure on heritage) and 11.7.1 (open public space), will feature scores ranging between 70 and 80 in 2030 based on our predictions. These indicators require special attention from decision-makers since they are challenging but not impossible to achieve if more ambitious policies are developed.

It should be noted that our predictions were only based on indicator trends from the 2010-2020 period, and other factors including the COVID-19 pandemic, climate change, and policy changes were not considered. The COVID-19 pandemic may have temporary effects on economic development and GHG emissions but will not change long-term trends (UNEP 2020). Climate change is a real long-term challenge to human sustainable development (Cheng 2020; WEF 2021). It will bring higher uncertainties to indicators 11.5.1 (loss from natural disasters) and 11.5.2 (economic loss from natural disasters), which are directly related to natural disasters, and will gradually influence many other indicators (Nerini 2019).

The 20 indicators were divided into four groups according to their progress to better interpret the observed and predicted results, and different suggestions and timelines for action were provided in Figure 5. The first group (in the red box of Figure5) consists of the indicators that will lag far behind and have little chance of achieving their 2030 targets but have a chance to achieve them during the 2040-2050 period based on our predictions. The second group (in the orange box) consists of indicators that face great challenges in achieving their 2030 targets. The third group (in the yellow box) represents indicators that only have a small distance left and are very likely to achieve their 2030 targets if a little more effort is made. The fourth group (in the green box) represents indicators that will smoothly achieve their 2030 targets if they maintain their current progress. The classification of different groups can help decision-makers know which indicators need urgent action and which need long-term planning.

Figure 5. Suggestions to policymakers for different states of the indicators. The rectangles in red, orange, yellow, and green represent the four groups of the indicators and indicate their distance to 2030 targets from far to near.

4. Discussions

It is very important to obtain SDG indicator data that has temporal and spatial information. On the one hand, these data can help us recognize the spatial imbalance in sustainable development. On the other hand, it helps to recognize the inconsistent steps that hamper progress. This also helps us gain a better understanding of the factors that limit the reliability of some of the indicators, such as the mountain green cover index mentioned earlier that could be biased by high-mountain areas. Another example of an SDG indicator which needs deep understanding is indicator 14.7.1 (sustainable fisheries). The value of this indicator has decreased in recent years, but it is not easy to judge whether it is a good or bad trend in the case of China, which is the world’s largest fish producer. Using satellite observations, the area of aquaculture ponds in coastal waters was found to decrease from 1.07 × 10⁴ km² in 2015 to 0.94 × 10⁴ km² in 2018, and then further decrease to 0.9 × 10⁴ km² in 2020 (Table 2). The decrease in the area and density of aquaculture ponds was the result of offshore protection policies, which are beneficial for improving the marine environment. Therefore, we considered the decrease in proportion as a good and sustainable trend for indicator 14.7.1 from the perspective of environmental protection. However, the applicability of our standard elsewhere requires further testing, particularly in the context of small islands and developing countries that rely on fishing.

Another problem when evaluating the SDG indicators is how to set the 2030 targets. The setting of targets will directly influence judgements of the indicators’ status and may bring uncertainties for evaluation. Only a few of the selected 20 SDG indicators have quantifiable targets in the global indicator framework. In this work, many indicators refer to recommendations by international organizations or to the state of a few countries at the top of the indicator ranking in 2015 (Sachs et al. 2021; OECD 2022). Some of the indicators that use top country values do not have strict 2030 targets but only consist of relative evaluations between countries. It is difficult for most developing countries, not just China, to reach the level of developed or high-income countries for indicators involving money or expenditures, such as SDG indicators 6.4.1 (water use efficiency) and 11.4.1 (per capita expenditure on heritage). Therefore, using a change rate of the country itself for these indicators, such as setting a target of doubling efficiency or investment in 2030 compared to 2015, may be fairer for developing countries. However, it may then become too strict for developed countries where indicator values are already very high. Unfortunately, there is no such research or reference that can be used to make a widely accepted target at present. Another example of an indicator that needs a better 2030 target is indicator 11.3.1 (land use efficiency). This indicator promotes a well-planned compact urban environment, where lower values indicate a higher land use efficiency that is more friendly to the natural environment, and the 2030 target for the indicator is suggested to be lower than one (UN-Habitat 2018; Estoque et al. 2021). However, cities that are too compact can also produce problems for the living environment, so it is difficult to judge sustainable urbanization using only this indicator (UN-Habitat 2018).

The uncertainty for the employed method and data includes three factors. The first factor is the evaluation accuracy of each indicator. We try to use the methods recognized by UN, or the published papers with given accuracy to reduce the uncertainty. The second factor is the comparison of different provinces. The comparison among different provinces adopts the commonly used method of average weight of each indicator in this work, but we also found that the average weight is not fair for provinces of different types and characteristics. At the provincial level, take the highly urbanized coastal Shanghai and the highly natural inland Tibet as an example, how to compare their sustainable development level is a big challenge. This is also a problem for comparison among different countries worldwide. Using the spatial information, the land cover ratio can be considered as a weight in SDGs scoring and sorting in future work. The third factor comes from the interpolation and prediction on temporal scale. Some of the indicators are observed every five years, so there are not many available observation data during 2010-2020. The limited frequency of observation may lead to bias in interpolation and prediction, especially the indicators which may change severely.

Above all, as the year 2030 approaches, we need some adjustments for the SDG framework. Based on our work, we have two suggestions for future work. (1) Use more big data in SDG progress evaluation. Big data can not only be used in indicator state assessment but also provide much more detailed spatial–temporal information than statistics for SDG evaluation, which is very beneficial for decision making. (2) Set more reasonable and quantitative targets for each SDG indicator, taking into full consideration of the country-specific status (developed, developing) and conditions is urgently needed to judge whether the goals of the 2030 Agenda could be achieved.

5. Summary

At present, the data used for SDG indicator evaluation is still very scarce and in urgent need, and Big Earth Data is an effective means to solve this problem. This study evaluated the changing trend and mapped the spatial imbalance of 20 environmental SDG indicators in China using Big Earth Data, and provided suggestions on how to achieve the targets of the 2030 Agenda step by step for decision-makers.

Using the data of SDG indicators calculated from remote sensing, navigation, statistics, surveying et al. it is found that China’s environmental SDG indicators have made great progress in the first trimester of the 2030 Agenda (2015-2020), but still face great challenges. Four SDG indicators have already achieved their targets by 2020. From the current state and recent changing trend, it is also predicted the state of the indicators in 2030. The indicators are divided into four groups, including two indicators being lagged far behind, six indicators facing great challenge, two indicators being close to achieve the 2030 targets, and 10 indicators that will achieve the targets. Different suggestions and timelines for action were provided for the four types.

The spatiotemporal information of Big Earth Data has unique advantages in SDG index evaluation, imbalance, and change trend analysis, and will play a greater role in the future. Lessons learned from the mapping, evaluation, and prediction of SDG progress in China can be used and extended for the benefit of SDG implementation knowledge and practice in other countries in the world.

Author contributions

H.G. and J.L. designed and led the study. L.H., Y.C., Z.S., X.L., S.L., L.Z., and F.W worked on the data acquisition, processing and analysis. L.H and Y.C designed the figures. I.N., Z.X., Q.L., L.L., M.W., Y. C. contributed to interpretation of the results. All of the authors led the writing.

Competing interests

The authors declare no competing interests.

Acknowledgments

This work is supported by the Strategic Priority Research Program of the Chinese Academy of Sciences (grant number XDA19090000 and XDA19090122)

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The data that support the findings of this study are available from the corresponding author upon request.

Correction Statement

This article has been corrected with minor changes. These changes do not impact the academic content of the article.

Additional information

Funding

This work was supported by Key scientific and technological research projects in the Xinjiang Production and Construction Corps: [Grant Number 2023AB074]; the Strategic Priority Research Program of the Chinese Academy of Sciences: [Grant Number XDA19090122].

Notes

1 1 USD≈6.2 RMB in 2017, 1 USD≈6.7 RMB in 2017, 1 USD≈6.6 RMB in 2018.

2 1 PPP USD≈4.2 RMB in 2017.