12 of 12 themes carry at least one mapped attribute.

Counts MAPPED attributes only. A theme reading zero may still have data in the field inventory that is not yet mapped to a SEL or CCL attribute.

Components sum to 9,596,960 against the published total area 9,596,960; the figure is drawn only because they agree.

Inland water only. Territorial sea and the exclusive economic zone are claims, not area, and are listed in the setting table above.

All 3 components are present and sum to 581.3 cubic km/year. The source publishes no total for this family, so that sum is the components added up, not a figure checked against a published one.

Withdrawal, not consumption: water returned to the system after use is counted here. Not comparable with the renewable resource figure in the water table, which is a stock.

China in its neighbourhood: fourteen land neighbours, as many as any state in the world, across a frontier running from the Pamirs to the Tumen. All fourteen are named on the map and itemised, with each one's share of the boundary, in the setting table below.

The drainage divide is the country's other defining line. The east drains to the Pacific; a large interior has no outlet to any sea at all. Line width is mean discharge, so the Yangtze reads as the system it is rather than as one blue line among thousands.

HydroRIVERS is a hydrological network keyed on reach ids, not a gazetteer: it carries no river NAMES, so this map does not print any. The provinces are labelled instead.

The 31 province-level divisions of mainland China and their capitals. This is the frame every province-level figure in this book is drawn on.

22 provinces, 5 autonomous regions and 4 municipalities. Hong Kong, Macau and Taiwan are outside the 2020 census universe and are not shaded.

A statement about DATA SCOPE, not about sovereignty. HERA's 2020 census universe is the 31 mainland province-level divisions; Hong Kong, Macau and Taiwan are not in it, and every other figure in this book is drawn on the darker area. Several chapters say so in a sentence and no figure showed it.

Boundaries as geoBoundaries publishes them for each code (CHN, TWN, HKG, MAC). The book names its boundary publisher so a reader who disagrees with a line can take it up with the source rather than with HERA.

Counts as published. Size classes are OpenFactBook's own and are not a tonnage measure.

Shares as published; the listed sources total 100.0%. A source this country does not report is absent from the chart rather than drawn as zero.

Share of the 565,125 place names that carry a language code. Names without one are excluded from the denominator. A further 12 language(s) fall below the listing threshold.

Toponymic evidence, not speaker counts. A language may name a large share of the landscape while having few remaining speakers; the two measures are independent and are not comparable.

Shares as published; the listed groups total 99.9%.

National shares as published by the source. HERA holds no subnational religion surface for this country, so this cannot be broken down by region.

Shares as published; the listed groups total 100.0%.

National shares as published. Ethnic categories are the source's own and are not comparable between countries: a census that asks about race, one that asks about nationality and one that asks about mother tongue produce different answers to different questions.

Where minorities are the largest SHARE of the population. Read it with Figure 8-b: share and count do not name the same provinces.

Where minorities are the most PEOPLE. The pair is the point: the province with the highest minority share is not the province where most of China's minority citizens live.

Components sum to 100 against 100%; the figure is drawn only because they agree.

Shares of the total population as published. A share is not a count: applied to a population that is itself an estimate, these give an order of magnitude and not a headcount.

31 units

Equal-area projection (Albers), so unit areas are comparable. Boundaries: geoBoundaries gbOpen (CC BY 4.0, https://creativecommons.org/licenses/by/4.0/). Runfola et al. (2020), PLoS ONE 15(4): e0231866. Modified by Masego Inc: geometry simplified at 0.01° tolerance and coordinates rounded to 4 decimal places.

93.6% of China's people live on 43.4% of its land, southeast of a line Hu Huanyong drew in 1935. HERA re-derives the split from its own holdings rather than citing it: 406 county-level divisions northwest, 1,933 southeast.

Mean density 16.4/km2 northwest against 313.0/km2 southeast. Split on the great circle through Tengchong (98.50E, 25.02N) and Aihui (127.50E, 50.25N); the planar chord gives a materially different answer.

The same 2,339 counties as Figure 9-a, extruded by density and tilted, on the same class breaks and the same palette. The Hu Line is not a gradient - it is an escarpment.

Males per 100 females. The ramp diverges on China's own national ratio, not on 100: centring on parity puts 29 of 31 provinces on one side and wastes half the scale.

Ageing, at county resolution. Class breaks are the UN's ageing thresholds - 7% ageing, 14% aged, 20% super-aged - not quantiles, so the colours mean the same thing here as they would for any other country.

Absolute population. Read against Figure 9-a: density and headcount disagree at both ends - the emptiest provinces are the largest, and the densest are the smallest.

People per household, Heilongjiang at 2.22 to Tibet at 3.19. · 31 units

Equal-area projection (Albers), so unit areas are comparable. Boundaries: geoBoundaries gbOpen (CC BY 4.0, https://creativecommons.org/licenses/by/4.0/). Runfola et al. (2020), PLoS ONE 15(4): e0231866. Modified by Masego Inc: geometry simplified at 0.01° tolerance and coordinates rounded to 4 decimal places.

Youth, on the same frame and the same units as Figure 9-c. The two maps are close to photographic negatives of each other, which is what a demographic transition looks like on a map.

Rural households, Shanghai at 1.97 to Tibet at 4.03. Read with Figure 9-h: the same province can hold the country's largest rural household and one of its smallest urban ones. · 31 units

Equal-area projection (Albers), so unit areas are comparable. Boundaries: geoBoundaries gbOpen (CC BY 4.0, https://creativecommons.org/licenses/by/4.0/). Runfola et al. (2020), PLoS ONE 15(4): e0231866. Modified by Masego Inc: geometry simplified at 0.01° tolerance and coordinates rounded to 4 decimal places.

Urban households, Tibet at 2.06 to Jiangxi at 2.89. Tibet is the extreme of the pair in both directions -- 4.03 against 2.06 people per household, rural against urban -- which is a bigger gap than any between two provinces. · 31 units

Equal-area projection (Albers), so unit areas are comparable. Boundaries: geoBoundaries gbOpen (CC BY 4.0, https://creativecommons.org/licenses/by/4.0/). Runfola et al. (2020), PLoS ONE 15(4): e0231866. Modified by Masego Inc: geometry simplified at 0.01° tolerance and coordinates rounded to 4 decimal places.

China's urban centres, one circle each, area proportional to population and the largest named. This is the third tier of geography in this book - below the province, beside the county - and the only one that shows the coastal megaregions as objects rather than as shading.

Rural sex ratio, Henan at 100.54 to Shanghai at 130.93. Drawn on the same breaks and the same palette as Figure 9-l so the two can be compared directly.

Households, not people. Read against Figure 9-d: a province's rank by household count is not its rank by population, because household size is not constant across China. · 31 units

Equal-area projection (Albers), so unit areas are comparable. Boundaries: geoBoundaries gbOpen (CC BY 4.0, https://creativecommons.org/licenses/by/4.0/). Runfola et al. (2020), PLoS ONE 15(4): e0231866. Modified by Masego Inc: geometry simplified at 0.01° tolerance and coordinates rounded to 4 decimal places.

Urban sex ratio, Jilin at 95.56 to Tibet at 115.62. The pair says something neither map says alone: the RURAL ratio is the more extreme of the two, and its extremes are Shanghai and Beijing - whose census 'rural' population is a thin periphery of a megacity, not a countryside. Figure 12-b tests whether that is a national pattern; it is not (r=0.487 across all 31 provinces, 0.041 once those two are removed).

Figures 9-j and 9-l in one frame. 29 of 31 provinces sit above the diagonal — their countryside is more male-skewed than their cities — and the three furthest above it are all municipalities: Shanghai by 27.3 points, Beijing by 18.82, Chongqing by 14.03.

The exceptions are the interesting part. Guangdong (114.75 urban against 111.36 rural) and Tibet (115.62 urban against 107.83 rural) run the other way: their CITIES are the male-skewed half. For Guangdong that is the Pearl River Delta's manufacturing workforce, which is the same migration this book's other figures see from the sending end.

Children and elders per 100 people of working age, Beijing at 33.68 to Henan at 57.67. Figures 9-o and 9-p split this into its two halves, and the split is the point: the same dependency ratio can mean opposite things. · 31 units

Equal-area projection (Albers), so unit areas are comparable. Boundaries: geoBoundaries gbOpen (CC BY 4.0, https://creativecommons.org/licenses/by/4.0/). Runfola et al. (2020), PLoS ONE 15(4): e0231866. Modified by Masego Inc: geometry simplified at 0.01° tolerance and coordinates rounded to 4 decimal places.

Under-15s per 100 of working age, Shanghai at 13.27 to Guizhou at 37.31. · 31 units

Equal-area projection (Albers), so unit areas are comparable. Boundaries: geoBoundaries gbOpen (CC BY 4.0, https://creativecommons.org/licenses/by/4.0/). Runfola et al. (2020), PLoS ONE 15(4): e0231866. Modified by Masego Inc: geometry simplified at 0.01° tolerance and coordinates rounded to 4 decimal places.

Over-65s per 100 of working age, Tibet at 8.06 to Sichuan at 25.37. Guangdong and Liaoning carry dependency ratios 40.25 and 40.05 - the same burden - and nothing else in common: Guangdong's is 27.20 children against 13.04 elders, Liaoning's is 15.58 against 24.47. One needs schools, the other needs pensions. 9 of 31 provinces are already elder-driven. · 31 units

Equal-area projection (Albers), so unit areas are comparable. Boundaries: geoBoundaries gbOpen (CC BY 4.0, https://creativecommons.org/licenses/by/4.0/). Runfola et al. (2020), PLoS ONE 15(4): e0231866. Modified by Masego Inc: geometry simplified at 0.01° tolerance and coordinates rounded to 4 decimal places.

Children and over-65s per 100 people of working age, at county resolution. Figure 9-n draws the same measure by province and runs 34 to 58; the counties run 3 to 87. Averaging to the province is what removes the range.

The elders' half alone. By province it runs 8 to 25 (Figure 9-p); across counties it reaches 53.7 — more than half as many people over 65 as of working age, in the same country whose national figure is 20.3.

121 county-level divisions at 2,000 people per km2 or more hold 20.86% of China's people on 0.88% of its land. Figure 9-a says which half of the country is populated; this says how small the populated part actually gets.

5 urban centres pass 10 million on their own, holding 113,936,629 people between them. Everything below 2,000 is drawn in one flat grey rather than on a ramp: a gradient there would invite the reader to compare the empty parts of China to each other, which is not what this figure is about.

World Bank Open Data, 1960–2023.

National figure. World Bank indicators are country-level, so this cannot be broken down by admin unit.

World Bank Open Data, 1960–2023.

National figure. World Bank indicators are country-level, so this cannot be broken down by admin unit.

World Bank Open Data, 1982–2020.

National figure. World Bank indicators are country-level, so this cannot be broken down by admin unit.

University attainment, Guangxi at 4.6% to Beijing at 22.4%.

Postgraduate qualifications, Tibet at 0.22% to Beijing at 5.92%. The ratio between the top and bottom province is far wider here than for university attainment in Figure 11-a: advanced study concentrates where the institutions are. · 31 units

Equal-area projection (Albers), so unit areas are comparable. Boundaries: geoBoundaries gbOpen (CC BY 4.0, https://creativecommons.org/licenses/by/4.0/). Runfola et al. (2020), PLoS ONE 15(4): e0231866. Modified by Masego Inc: geometry simplified at 0.01° tolerance and coordinates rounded to 4 decimal places.

Illiteracy, Beijing at 0.89% to Tibet at 28.08%.

This is the census's published ILLITERACY rate: a higher value means more residents aged 15 and over who cannot read or write.

Rural illiteracy, Liaoning at 2.11% to Tibet at 36.64%. A higher value means more residents who cannot read or write; see Figure 11-c. · 31 units

Equal-area projection (Albers), so unit areas are comparable. Boundaries: geoBoundaries gbOpen (CC BY 4.0, https://creativecommons.org/licenses/by/4.0/). Runfola et al. (2020), PLoS ONE 15(4): e0231866. Modified by Masego Inc: geometry simplified at 0.01° tolerance and coordinates rounded to 4 decimal places.

Of the residents who never attended school, the share who are women: Tibet at 58.8% to Jiangsu at 75.5%. Women are the majority in every one of the 31 provinces; the map shows how large a majority.

Urban illiteracy, Liaoning at 0.52% to Tibet at 11.73%. The pair is the claim: Tibet reads 36.64 against 11.73 percent illiterate, rural against urban. The national figure conceals a gap that is larger than the gap between most countries. · 31 units

Equal-area projection (Albers), so unit areas are comparable. Boundaries: geoBoundaries gbOpen (CC BY 4.0, https://creativecommons.org/licenses/by/4.0/). Runfola et al. (2020), PLoS ONE 15(4): e0231866. Modified by Masego Inc: geometry simplified at 0.01° tolerance and coordinates rounded to 4 decimal places.

Senior secondary completion, Tibet at 7.4% to Shanghai at 19.4%. The narrowest spread of the three attainment maps -- the level China has come closest to universalising. · 31 units

Equal-area projection (Albers), so unit areas are comparable. Boundaries: geoBoundaries gbOpen (CC BY 4.0, https://creativecommons.org/licenses/by/4.0/). Runfola et al. (2020), PLoS ONE 15(4): e0231866. Modified by Masego Inc: geometry simplified at 0.01° tolerance and coordinates rounded to 4 decimal places.

Shares as published; the listed sectors total 100.0%. The source notes that figures may not reach 100% because some consumption is not allocated to a sector.

Top-level classes only. Arable land, permanent crops and permanent pasture are subsets of agricultural land and are listed in the table rather than charted beside their own parent.

Urbanisation by province. An earlier draft of this caption claimed the most urbanised provinces carry the most male-skewed rural populations. Measured, they do not: see Figure 12-b, which is why that figure exists.

One dot per province. Across all 31 the correlation is r=0.487, which looks like a relationship; drop Shanghai and Beijing and it is r=0.041, which is none. The effect is two municipalities, not a gradient — and an earlier draft of Figure 12-a's caption asserted the gradient.

Shanghai is 89.3% urban with a rural sex ratio of 130.93; Beijing 87.55% and 120.21. Tianjin is the control, and it is what kills the story: 84.7% urban — third highest in China — and a rural sex ratio of only 107.34, in the middle of the cloud. What Shanghai and Beijing share is not urbanisation but a census 'rural' category that describes a city's periphery.

Where China farms. 661 counties are more than 75% cropland and they hold 33.53% of the population; 23 counties have none at all.

Overture's `crop` class is LAND COVER, not the land USE a statistical office counts. It covers 21.35% of China's land; China's published arable share is about 12.7%. A satellite-derived crop class takes in orchards, managed and cultivated vegetation and agricultural mosaics that the arable definition excludes. Each county's cropland is ST_AREA(ST_UNION_AGG(ST_INTERSECTION(county, crop))). Clipping rather than centroid assignment, because a crop polygon spanning four counties otherwise lands entirely in one; union rather than SUM, because Overture's crop polygons overlap and summing them put one county at 52,662% of its own area.

One shared denominator and one shared unit, which is what lets these three sit on one axis. Mobile routinely exceeds 100: it counts subscriptions, not people.