City water stress and industrial water-saving potential in stringent management of China

China’s industrial water withdrawal soared in the last decades and remained high. Stringent water management policies were set to save water through improving industrial withdrawal efficiency by 20% 35 between 2015 and 2020. Although China has a nation-wide water scarcity, scarcity at city-level has not 36 been fully explored. Thus, it is meaningful to use sectoral data to investigate industrial water saving 37 potential and implication for alleviating scarcity. Here, we account for water withdrawal and scarcity in 38 272 prefectural cities, using a 2015 data benchmark. The top 10% of low-efficiency sectors occupied 46% 39 water use. In scenario analysis of 41 sectors across 146 water scarce cities, we assume a convergence of 40 below-average efficiencies to the national sector-average. Results reveal overall efficiency could be 41 increased by 20%, with 18.9 km 3 ( ± 3.2%) water savings, equivalent to annual water demand of Australia 42 or Hebei province in China. A minority of sectors (13%) could contribute to most (43%) water savings 43 whilst minimizing economic perturbations. In contrast, implementing water efficiency measures in the 44 majority of sectors would result in significant economic disruption to achieve identical savings. Water 45 efficiency improvements should be targeted towards this minority of sectors: cloth(ing) supply-chain, chemical manufacturing, and electricity and heat supply.

Freshwater is an essential and global resource 1 . Over the last 50 years, China's industrial water 50 withdrawal increased in 90% of its cities 2 , and has remained at a high level above 126 km 3 /yr from 2013 51 to 2018 3 largely due to low water-use efficiency. China used to have transnationally low efficiency partly 52 owing to mis-management 4-7 , specifically poor sectoral controls and water-saving initiatives 8 . China's 53 response to this was to legislate for industrial water withdrawals through the so-called stringent water 54 resources management system ("Three-Redline" regulations), introduced by the Chinese State Council 55 in 2011 9 , and aimed at saving water through improving industrial withdrawal per value-added by 20% 56 between 2015 and 2020. More recently, China established national water-saving demonstration (sponge) 57 cities, but specific control on both industrial water withdrawal intensities and volumes still remains poor 10 .

58
Although nation-wide China is deficient in water 11 , with a wicked problem between water demand 59 and availability 4,12 , city-level water scarcity has not been fully explored 13 . The science of water scarcity 60 assessment has developed for the past 30 years and, as more spatial geo-data have been available, studies 61 have adopted more integrated and multi-faceted approaches typically based on spatial resolution in grid 62 units at the river basin scale 14,15 or global levels [16][17][18] , rather than at administrative/territory based units 63 such as the city level. There is only a single city-level based study in 2005 from the Ministry of Water 64 Resources in China, which is not widely available to the public 19 . Thus far, to the best of our knowledge, 65 an appraisal of cities and their water scarcity status is unavailable. In terms of measuring scarcity, the 66 criticality ratio (water withdrawal to annual renewable freshwater) is a simple and classical indicator of 67 blue water and quantitative scarcity 20,21 . It has thus far been applied at the provincial level 16,22-24 , but not 68 at the city level due to data limitations 7 .

69
Water scarcity is typically exacerbated by unsustainable levels of water withdrawal; hence, society 70 ought to be well placed to mitigate it by improving water use efficiency, especially by reducing water 71 withdrawal intensities. Many studies have focused on agricultural intensification 25,26 in relation to better 72 water management in land use 27 and irrigation 28 . However, due to lack of measured efficiency data, there 73 remains a dearth of research especially from an industrial and sectoral perspective 29 , to explore water 74 saving potential and implication on scarcity alleviation 30 at the city level.

75
We first accounted for datasets on water withdrawal for 41 industrial sectors in 272 prefecture-level 76 cities (88% of China's population), and water scarcity for all cities (343) in 2015, based on a point-affected city type, and detected water scarcity and differences amongst these city-groups. Finally, in 80 scenario analysis we assumed a convergence of below-average efficiencies to the national sector-average, 81 to explore water saving potential amongst 41 industrial sectors and implication on water stress of Chinese 82 cities under the constraint of the 20%-intensity-reduction. For key sectors and cities, our results help to 83 identify priorities and optimize efforts for improving water use efficiency and facilitate more effective 84 water management through enabling distinctive saving strategies.

119
We built up datasets using a general accounting framework for Chinese cities, as developed for previous 120 work 31,32 . Drawn on the datasets, Fig. 1a represents a map of total water withdrawal at the city level.
over-100% as extreme water scarcity stress, signifying that annual water withdrawal exceeds renewable 124 water resources 13 .

125
Overall, 146 of 272 cities (55% of population) were found to be under water scarce conditions, a 126 result consistent with previous studies 13 . These cities are represented by darker colors in Fig. 1(c):

127
Guangzhou and Shenzhen (south), Shanghai, Suzhou, and Yancheng (east), Harbin (north), and Hotan 128 (west). Notably, in contrast to an earlier study 13 , we also identified some severe water-scarce areas in 129 south China: Shenzhen (south; 108%) and Foshan (southeast; 107%). Water scarcity in China is known 130 to already be serious, thus caution should be exercised when interpreting the south expansion of scarcity.

131
Sixty-nine Chinese cities (25%) were found to be under extreme water scarcity. These cities   When constrained by severe water scarcity, one might expect industries in water scarce cities to adopt 150 water saving technologies, hence their industrial water withdrawal intensities should be lower than 151 comparable industries in water sufficient areas. In other words, water scarcity should force local such as Qiqihar (north), Yingkou (east), Wuhai (west) and Puyang (central), had water intensities which 154 were much higher than in cities abundant in water resources. Although China has set intensity reduction 155 targets in stringent management since 2011, reducing intensities of sectors in water-scarce cities should 156 therefore be prioritized. Awareness of industrial water savings should be given greater focus in these 157 sectors in water scarce cities to prevent the situation to get worse. For example, cities such as Wuhai,

158
Hegang, Puyang, and Qitaihe, had water intensities which were still high, yet they were all included in 159 the 69 cities known to be over-exploiting resources, as released by the Chinese government in 2018 41 .

167
A disproportionately small fraction of sectors at the city level contributed to large industrial water 168 withdrawals. Thus sectors of low-efficiencies across cities should be well targeted to save water. We 169 ranked a total of 41×272=11,152 city-sector combinations by order of water intensity from low to high 170 and then calculated share of cumulative water withdrawal accordingly. We depicted these shares relative 171 to shares of cumulative numbers of sectors and obtained a water-withdrawal Lorenz curve (Fig. 2b). The 172 curve indicates that the top 10% of high-intensity sectors account for 46% of water withdrawal, as a 173 disproportionate fraction. Such high-intensity water users were mostly found in small and developing 174 cities, with representative industries such as papermaking and product manufacturing in Chenzhou 175 (central), Lincang (southwest) and Qiqihar (northeast); liquor, beverage and tea manufacturing in We compared water scarcity occurrence amongst different city-groups. The most-severely affected 179 were found in the high-tech group (Fig. 3); 38 cities over the 40% criticality-ratio (water scarce) and 20 180 above 100% (extremely scarce). These are the highest in their corresponding tier, indicating economic 181 growth limitations subject to water resources constraints. Notably, population in high-tech cities 182 accounts for 33% of the total, and are commonly affected from severe water scarcity. Heavy-and light-183 manufacturing cities were also ranked, following high-tech cities. These water scarce cities with sectors 184 of low water withdrawal efficiencies should be targeted.

194
For scenario analysis in individual of 41 industrial sectors, we substituted above-average water 195 intensities with average ones, by assuming technical progress in water use efficiency. Scenario A was 196 for all 272 cities and B was for the 146 water-stressed cities. Water saving strategies are more stringent 197 in A than B. If water withdrawal intensity of a sector in a city was lower than the national sector-average,

198
we left water intensity as it was. This would help maintain a stable technological and economic structure 199 whilst improving efficiency; If intensity of a sector was higher than the national sector-average, but it 200 occurred in a city with no water stress (criticality ratio less than 40%), we did not substitute it either; Only for sectors that both had above-average intensities and were located in water-stressed cities, we did

263
We also decomposed structure of the important 13% sector fraction into different cities and groups, 264 and identified four sectors (Fig. 4(d)) which contributed to half of total water savings; cloth(ing) supply 265 chain, chemical material and product manufacturing, and electricity and hot water supply.  Most severely scarce city-groups were effectively pinned down, such as high-tech, heavy-and light-271 manufacturing cities. These city-groups basically hold the top three places for efficiency improvements.
individual city-groups were also checked and consistent (upon request). Thus, we were able to reliably 275 and robustly validate discussion on substitution.

276
Of course, realization of water intensity reductions is likely to be different 29 from our rather crude 277 scenario analyses; technologies between sectors and cities vary, and we must consider institutional as

288
In summary, we have reported water withdrawal and scarcity accounting for 272 Chinese cities, 289 using a 2015 data benchmark. The top 10% of low-efficiency sectors made up 46% industrial water use.

290
In scenario analysis of 41 sectors across 146 water-scarce cities, through efficiency improvements by 20% 291 and satisfying the stringent management policy, 18.9 km 3 (±3.2%) water saving would be realized.

292
Yet, here we recommend water saving potential in a handful of sectors, as these sectors identified 293 to contribute to half of total water savings amongst 41 sectors. Focusing on these sectors makes sense in 294 terms of producing water saving returns, whilst minimizing potential economic disruption across the 295 industrial base. China may therefore target key sectors and cities in stringent water management, rather 296 than requiring all industries and cities to be involved in water saving.

299
City-level industrial water withdrawal data sources. Industrial total water withdrawal and water-300 withdrawal per value added were compiled from water resources bulletins at provincial and city levels.

301
Industrial water withdrawal is a newly withdrawn water amount 3 . This variable may depict pressure on 302 available water resources from domestic economic activities more accurately since it excludes reused 303 water.

304
Industrial water withdrawal intensities for individual sectors in each city were derived from the

348
where i represents a city (one ratio number for one city); water withdrawal was the total amount from 349 including farming, forestry, animal husbandry, fisheries, industry, construction, service, household, and 350 ecosystem and environment preservation; and water availability included surface water and groundwater.

351
There are mainly three indicators in the current study: net runoff, natural streamflow, and natural 352 streamflow minus consumptive use from upstream human activities 13 . We adopted the natural- availability) was 2.8% (0.9%) more than, but close to, its average values through multiple years (1957-with water quality and biodiversity 58 . The higher the ratio is, the more stress is placed on available water 359 resources from withdrawal, and the greater the probability of water scarcity occurrence 35 .

360
In addition to Fig. 3, we further found there appeared to be discrepancies in criticality ratio in 361 different city-types, indicating frequency and severity of water scarcity occurrence, referring to 362 Veldkamp et al. (2016) 59 . For energy production cities (Appendix Fig. 2), frequency seemed relatively 363 higher, but not as severe when compared to heavy manufacturing group. Trendline curve peaked at 50%, 364 exceeding the 40% definition for water scarcity. In other words, most cities appeared to be distributed to 365 the right of scarcity threshold. Reassuringly, there appeared to be relatively few instances of cities 366 occurring in the extreme scarcity region (i.e. >100%).

367
In contrast, heavy manufacturing cities had lower frequencies of water scarcity occurrence, but once 368 over the 40% threshold it tended to be more severe.

383
Uncertainty analysis. We also clustered cities based on economic shares of GDP for primary, secondary 384 and tertiary industries, then classified cities into three groups for sensitivity analysis. We found only 385 minor differences between ratios of cities at individual water scarcity levels, from the groups using 25%; whilst for service-based cities they were 67% and 35%. Although clusters were based on different 389 indexes, we found no significant differences in water-scarcity distribution and status. We also verified 390 water withdrawal per GDP of agriculture-based cities of 211m 3 per 10 4 Yuan, which was close to the 391 magnitude of representative agriculture province such as Heilongjiang at 210 in 2015 60 . Finally, for 392 individual city groups we validated median and average criticality ratios and water intensities; these 393 results as well as significance tests for our group classification are available upon request.

394
Besides, we may over-estimate criticality ratio, considering water withdrawal statistics do include 395 those from reservoirs and upstream rivers, while water availability data do not include these parts. We

396
were unable to incorporate these data into water availability generally due to statistical incongruence 397 between cities. Thus, our results could suffer from an upward bias in some cities. In future, we will  to reduced precipitation in dry years. This further work will not only reduce uncertainty of water scarcity 406 status, but also explore temporal insights into understanding of water scarcity and allow for more time-407 series and statistical-significance testing.

408
Water quality-induced scarcity 16,63,64 has not been included in this paper due to lack of data for water 409 temperature and salinity, nutrient and other pollutants. Besides, the extent to which water savings could 410 be driven by water stress needs quantitative analysis.

411
At this stage our study is also limited by data availability for agriculture; we do not find sufficient 412 irrigation efficiency data for subdivided crops or lands in individual cities, in order to project water saving 413 potential for agriculture. For industrial sectors, it is better to use value-added to substitute output to assess 414 efficiency, especially when such sectoral value-added data will be accessible in the future.

415
Finally, we only considered direct water savings for isolated sectors. It is only partially feasible to 416 assume a smooth knowledge transfer of water efficiency experience from wealthier cities to poorer ones,