Physical versus economic water footprints in crop production : a case 1 study for China 2

10 A core goal of sustainable agricultural water resources management is to implement lower water footprint (WF), i.e., higher 11 water productivity, while maximising economic benefits in crop production. However, previous studies mostly focused on 12 crop water productivity from a single physical perspective. Little attention is paid to synergies and trade-offs between water 13 consumption and economic value creation of crop production. Distinguishing between blue and green water composition, grain 14 and cash crops, and irrigation and rainfed production mode in China, this study calculates the production-based WF (PWF) 15 and derives the economic value-based WF (EWF) of 14 major crops in 31 provinces for each year over 2001-2016. The 16 synergy evaluation index (SI) of PWF and EWF is proposed to evaluate quantitatively the synergies and trade-offs between 17 the two. Results show that both the PWF and EWF of most considered crops in China decreased with the increase of crop yield 18 and prices. The high (low) values of both PWF and EWF of grain crop tended to obvious cluster in space and there existed a 19 huge difference between blue and green water in economic value creation. Moreover, the SI revealed a serious incongruity 20 between PWFs and EWFs both in grain and cash crops. Negative SI values occurred mostly in northwest China for grain crops, 21 and overall more often and with lower values for cash crops. Unreasonable regional planting structure and crop prices resulted 22 in this incongruity, suggesting the need to promote regional coordinated development to adjust the planting structure according 23 to local conditions and to regulate crop prices rationally. 24

Nevertheless, the above studies lacked a complete temporal and spatial evolution analysis of the WF from the economic 51 perspective. More importantly, the above studies did not involve the study of WF coordination in different aspects thus ignored 52 the synergies and trade-offs between water consumption and economic value creation during crop production in WF 53 assessment, which is undoubtedly of great significance.  Sun et al. (2013) found that the WF of crop depended on agricultural management rather than on 62 regional climate differences; Zhuo et al. (2016a) showed that China's domestic food trade was determined by the economy and 63 government policies, not by regional differences in water endowments; Wang et al. (2019) showed possibility and importance 64 https://doi.org/10.5194/hess-2020-157 Preprint. Discussion started: 30 April 2020 c Author(s) 2020. CC BY 4.0 License.
of accounting for developments of water-saving techniques in largescale crop WF estimations. However, most of these studies 65 focused on quantifying WF from a single physical perspective. To our knowledge, there is no study yet to provide clear insights 66 into the economic benefits of water use. 67 To fill the above research gap, the current study objective is, taking China over 2001-2016 as the study case, to explore the 68 relationship between water resource consumption and economic value creation of intra-national scale crop production, and to 69 propose a synergy evaluation index (SI) of PWF and EWF. First, the blue and green PWF (PWFb, PWFg) of 14 major crops 70 (winter wheat, spring wheat, spring maize, summer maize, rice, soybean, cotton, groundnut, rapeseed, sugar beet, sugarcane, 71 citrus, apple, and tobacco) is calculated annually in 31 provinces, and the corresponding EWF is derived. Second, crops are 72 distinguished between grain and cash crops, with Mann-Kendall trend test and spatial autocorrelation analysis method for 73 evaluation of the temporal and spatial evolution characteristics of PWF and EWF. Finally, the synergy evaluation index (SI) 74 is constructed to reveal the synergies and trade-offs of crop water productivity and its economic value from the WF perspective. 75 2 Method and data 76

Calculation of production-based water footprint (PWF) 77
The PWF (m 3 kg -1 ) consists of the blue PWF (PWFb, m 3 kg -1 ) and the green PWF (PWFg, m 3 kg -1 ), which are respectively 78 calculated from the daily green (ETg where Sb[t] and Sg[t] (mm) respectively represent the blue and green soil water content at the end of day t. According to Siebert 104 and Döll (2010), the maximum soil moisture of rainfed fallow land two years before planting is taken as the initial soil moisture 105 for simulating. At the same time, the initial soil water during the growing period is set as green water (Zhuo et al., 2016b). 106

Calculation of economic value-based water footprint (EWF) 107
Following Hoekstra et al. (2011), the EWF (m 3 USD -1 ) of crop production represents the water consumption per unit of 108 where PWF (m 3 kg -1 ) the production-based WF, and UP (USD kg -1 ) the crop unit price. The EWF is numerically equal to the 111 inverse of the economic water productivity. Considering the PWF and the EWF together provides a clear and intuitive 112 measurement to analyse the synergy relationship between water consumption of crop production and economic value creation. 113 To eliminate the influence of inflation, we use the consumer price index (CPI) to calculate the inflation rate of China based on 114 https://doi.org/10.5194/hess-2020-157 Preprint. Referring to Chouchane et al. (2015), when calculating the blue and green EWF, we distinguish between irrigation and rainfed 117 agricultural modes. In rainfed agriculture, the green EWF (EWFg,rf) is obtained by dividing the green water consumption per 118 unit yield under rainfed conditions by the unit price of crops, as shown in Eq. (9). Compared to rainfed agriculture, the ratio 119 of crop yield increment under full irrigation is obtained by AquaCrop model. We use it to distinguish the blue and green EWF 120 in irrigation agriculture (EWFb,ir, EWFg,ir), as shown in Eqs. (10) -(12): 121 , , where CWUg,rf (m 3 ha -1 ) represents the consumption of green water per unit area in rainfed agriculture; CWUb,ir (m 3 ha -1 ) and 126 CWUg,ir (m 3 ha -1 ) represent the consumption per unit area in irrigation agriculture of blue and green water, respectively; α is 127 the ratio of crop yield increment under full irrigation obtained by AquaCrop model; YRF(kg ha -1 ) and YIR(kg ha -1 ) represent 128 the actual crop yield under the rainfed and irrigation mode, respectively; Yrf and Yir represent the model simulated yield under 129 the rainfed and irrigation mode, respectively. The EWFg,rf represents the amount of green water consumption per economic 130 benefit unit in rainfed agriculture (also refers to the amount of green water input for each additional economic benefit unit); 131 EWFb,ir (EWFg,ir) refers to the additional amount of blue (green) water for each additional unit economic benefit under the 132 same green (blue) water input in irrigation agriculture. 133

Spatial and temporal evolution of WFs 134
The Mann-Kendall (M-K) trend test (Mann, 1945;Kendall, 1975 Using two-tailed test, when the absolute value of Zc exceeds 1.96 and 2.58, it means that the significance test of 95% and 99% 150 has been passed, respectively. The positive Zc indicates an upward trend, while a negative value means a downward trend. The 151 first law of geography states that everything is related, and things close to each other are more relevant (Tobler, 1970  with adjacent provinces; high-low (H-L) and low-high (L-H) mean that the level (high or low) of WF in this province is 167 opposite to adjacent provinces. The analysis of spatial autocorrelation can be realised by GeoDa. 168

The synergy evaluation index (SI) of PWF and EWF 169
To conduct a comprehensive assessment of WF from a physical and economic perspective, we compared provincial PWF and 170 EWF with the respective average at the national level, and the sum of the ratios of the differences to the ranges is the 171 synergy evaluation index (SI) of this province. The SI is calculated as follows: 172 , , , where SIi,j,c is the synergy evaluation index of PWF and EWF of crop c at province i in year j, PWF j,c (m 3 kg -1 ) and EWF j,c (m 3 174 USD -1 ) are the averages at the national level in year j. Obviously, -2 ≤ SIi,j,c ≤ 2. When the PWF and EWF in a region are both 175 lower than the respective average at the national level, the SI of the region must be positive; when the PWF and EWF in a 176 region are both higher than the respective average at the national level, the SI of the region must be negative. When one is 177 higher, and the other is lower than the corresponding average, the SI may be positive or negative, depending on the difference 178 between the provincial value and the national average. The greater the SI, the more advantageous the region is in terms of 179 water resource consumption and economic value creation in crop production (less water consumption per yield and higher 180 economic benefits per water consumption unit). On the contrary, a low SI indicates that the contradiction between water 181 resource consumption and economic value creation in crop production is sharp (more water consumption per yield but lower 182 economic benefits per unit water consumption). 183

Data 184
The planting area and yield data of each province was obtained from NBSC (2019). The provincial price data of crops were 185 of WF, the EWFg,ir was the lowest (mean 2.79 m 3 USD -1 ), EWFb,ir was the highest (mean 23.68 m 3 USD -1 ), and EWFg,rf (mean 276 5.60 m 3 USD -1 ) was close to the average EWF (5.41 m 3 USD -1 ) in irrigation and rainfed production mode. This suggests that 277 more water was required per additional benefit unit under irrigation than under rainfed mode, whereas in the rainfed agriculture, 278 compared with blue water, increasing the input of green water may result in more economic benefits. Therefore, utilisation 279 efficiency of green water resource for grain crops should be improved. 280 Concerning cash crop, the EWFb,ir decreased by 62% from 14.39 m 3 USD -1 to 5.47 m 3 USD -1 . Compared to grain crop, the 281 difference between the EWFg,ir and EWFg,rf was smaller, with average values of 2.60 m 3 USD -1 and 2.29 m 3 USD -1 , respectively. 282 In addition, compared to grain crop, the EWF of cash crop was lower, which indicated that cash crop production could get 283 more economic benefits per water consumption unit. Besides, increasing the input of green water resource could obtain higher 284 economic benefits, and the rainfed production model had greater economic potential. EWFg,rf of cash crops decreased most significantly. As for the EWFb,ir, the downward trend of cash crops was more significant, 300 compared to that of grain crops. The M-K test results in Table 7  Concerning the composition of blue and green water for grain crop, the EWFb,ir in north-western China was lower, while the 319 EWFg,ir and EWFg,rf were higher. In contrast, the EWFb,ir in southern China was higher, while the EWFg,ir and EWFg,rf 320 were lower. Specifically, in the northwest region, Ningxia had the highest EWFg,ir and EWFg,rf (mean 5.57 m 3 USD -1 and 8.16 321 m 3 USD -1 , respectively), while the EWFb,ir was only 7.08 m 3 USD -1 , far lower than the national average (23.68 m 3 USD -1 ).
Instead, the EWFg,ir and EWFg,rf in Yunnan were close to the national average level (2.79 m 3 USD -1 and 5.60 m 3 USD -1 ), and 323 EWFb,ir was the highest (50.55m 3 USD -1 ). The EWF of cash crop had no obvious spatial clustered phenomenon, decreasing 324 significantly over time in 31 provinces, which was consistent with the spatial evolution characteristics of the corresponding 325 PWF previously discussed (Fig. 6b). EWF in both grain and cash crops (see Fig. 9), from the perspective of planting structure (see Fig. 10). In terms of grain crop, 372 the PWF and EWF in 9 provinces (Shaanxi, Gansu, Shanxi, Tianjin, Inner Mongolia, Qinghai, Hebei, Xinjiang, and Ningxia) 373 were significantly higher than the national average level; the PWF and EWF in Fujian, Guangdong, Hunan, Hubei and Jiangxi 374 were significantly lower than the national average. Shaanxi province had the highest PWF in China (1.23 m 3 kg -1 ), and the 375 second highest EWF (7.48 m 3 USD -1 ). In Shaanxi province, winter wheat and spring maize with high water consumption and 376 low yield accounted for more than 90% of the total sown area of grain crops, with yields lower than the national averages by 377 24% and 26%, respectively. Moreover, the price of wheat in Shaanxi province (0.17 USD kg -1 ) was lower than the national 378 average (0.19 USD kg -1 ). The reasons for high water consumption per unit of grain production coupled with poor economic 379 benefits in Shaanxi province can be attributed to the above two points. In contrast, in Jiangxi province, where rice, which has 380 low water consumption intensity, is the main grain crop (rice accounting for 95% of the grain crops), PWF and EWF were 381 0.77 m 3 kg -1 and 3.63 m 3 USD -1 , well below the national averages (0.93 m 3 kg -1 , 5.04 m 3 USD -1 ). 382 As for cash crop, the PWF and EWF in 15 provinces, including Tianjin, Jilin and Jiangxi, were significantly higher than the 383 national average values, while the PWF and EWF of the five provinces represented by Shanxi were lower than the national 384 average level. The PWF of Tianjin was 1.92 m 3 kg -1 , the highest in China, and the EWF was 3.26 m 3 USD -1 , the fifth highest 385 in China, which was significantly higher than the national average (2.05 m 3 USD -1 ). It can be seen from Fig. 24 that cotton  386 accounted for the largest proportion (70%) in the planting structure of cash crops in Tianjin. Cotton consumed the most water 387 per yield unit of cash crops, while the price unit of cotton in Tianjin was the second lowest in China (1.11 USD kg -1 ), which 388 did not reflect the advantage of cotton as a high-value crop. Large-scale planting of water-intensive crops sold at a low price led to high water consumption per yield but poor economic benefits in cash crop production in Tianjin. Jiangxi province 390 showed the highest EWF in China (3.86 m 3 USD -1 ), and a PWF (0.96 m 3 kg -1 ) which was also higher than the national average 391 (0.46 m 3 kg -1 ). Figure 10b shows that citrus (planting area accounting for 29% of cash crops) and rapeseed (planting area 392 accounting for 48% of cash crops) are the main cash crops in Jiangxi. However, the price unit of citrus in Jiangxi was the third 393 lowest (0.17 USD kg -1 , only 62% of the national average), and the yield of rapeseed was also the third lowest (1.34 t ha -1 , 32% 394 lower than the national average). The low selling price and yield per unit area explain the poor economic benefits per water 395 consumption unit in cash crop production in Jiangxi. In contrast to the situation of Tianjin and Jiangxi, the main cash crop in 396 Shanxi was apple (planting area accounting for 87% of cash crops), with low water consumption intensity and a yield which 397 was the second highest in China (28.5 t ha -1 ), 1.5 times larger than the national average (18.8 t ha -1 ). Therefore, the large-scale 398 planting of low water consumption crops with high level of crop yield contributed to the higher economic benefits per water 399 consumption unit displayed in cash crop production in Shanxi. The goal of WF regulation is to reduce its magnitude to a sustainable level (Hoekstra, 2013), but the contradictions faced 408 during implementing sustainable development are rarely encountered in a single dimension. However, previous research has 409 most commonly adopted a single perspective approach to WF analysis. Based on the temporal and spatial evolution of PWF 410 and EWF, the synergy evaluation index (SI) is constructed to achieve a more comprehensive assessment in this study. This 411 approach has led to some differences in the results of WF compared to previous research. 412 Table 7 compares the PWF results of crops production between the current study and previous ones. Differently from 413 Mekonnen and Hoekstra (2011) and Zhuo et al. (2016a), this study distinguishes between wheat and maize varieties when 414 calculating the WF, despite China's wheat production is mainly of winter wheat (accounting for 95% in 2016). Due to the 415 differences of varieties, water consumption intensity and planting conditions, it is necessary to distinguish between crops in 416 the provinces where spring wheat is the main crop. In addition, due to the differences in model selection and parameters, the 417 calculation results will also be different. study is close to that of previous studies, which shows the rationality of the calculated results. 422 Table 8 compares the EWF of this study with previously calculated results of the economic water productivity. Since the 423 economic water productivity is numerically equal to the reciprocal of the EWF, the previous results are expressed in the form 424 of EWF for comparison. The results for wheat production show that, although the average EWF is close, differences in crop 425 varieties, planting environment, and climate condition result in huge differences in EWFb,ir under the same production mode. 426 Therefore, specific problems should be investigated separately. Selection and adjustment of production mode should be made 427 according to local conditions to promote coordinated development. 428 From the results of the multi-perspective analysis conducted in this study, we found that with the increase of yield unit and 429 price unit, the PWF and EWF of crop production both showed a decreasing trend, and the EWF decreased more significantly 430 compared with the PWF. The change of WF of cash crops was more obvious than that of grain crops. In terms of the spatial 431 pattern, compared with cash crops, WF of grain crops had a more significant spatial correlation, and the spatial distribution of 432 PWF was similar to that of EWF. H-H areas mainly gathered in north-western China, while L-L areas in south-eastern coastal 433 provinces. The average Moran's I of EWF (0.482) was higher than that of PWF (0.263). 434 Moreover, the SI results showed that the economic benefits of blue water and green water differed greatly. As for grain 435 production at the national level, the EWFb,ir (mean 23.68 m 3 USD -1 ) was much higher than the EWFg,ir (mean 2.79 m 3 USD -1 ), 436 and the EWFg,rf (mean 5.60 m 3 USD -1 ) was the closest to the average EWF in irrigation and rainfed agriculture (mean 5.41 m 3 437 USD -1 ). Compared with grain crops, the difference between EWFg,ir and EWFg,rf of cash crops was smaller, with average values 438 of 2.60 m 3 USD -1 and 2.29 m 3 USD -1 , respectively. Moreover, the EWF of cash crops was lower than that of grain crops. It 439 was more cost-effective to increase the input of green water than that of blue water during crop production. In north-western 440 China, the EWFb,ir was lower, while the EWFg,ir and EWFg,rf were higher; on the contrary, in southern China, the EWFb,ir was 441 higher, while the EWFg,ir and EWFg,rf were lower. Therefore, the utilisation efficiency of green water resources should be 442 improved, rainwater collection and storage should be developed, and the proportion of green water in the acquisition of 443 irrigation water should be increased. As for northern China, green water (rain water) should be converted into blue water 444 (irrigation water) as far as possible, so as to reduce blue water consumption while ensuring and increasing economic 445 benefits. As for southern China, rainfed agriculture should be chosen as far as possible. The necessary way to alleviate the 446 contradiction between water resource consumption and economic value creation is to adjust the agricultural production mode 447 and the irrigation method according to local conditions. 448 There was a serious incongruity between water consumption for crop production and economic value creation both in grain 449 and cash crops. In terms of grain production, the water consumption per yield was large, but the economic benefit per water 450 consumption unit was poor in the northwest region, while the opposite was true in the southeast coastal region. Over time, the 451 contradiction has not been alleviated, showing a relatively stable spatial pattern. Through analysis, this study shows that the 452 unreasonable regional planting structure and crop price may be the direct cause of the incongruity between water resource 453 consumption and economic value creation for crop production in China. Therefore, the government should adjust the planting 454 https://doi.org/10.5194/hess-2020-157 Preprint. Discussion started: 30 April 2020 c Author(s) 2020. CC BY 4.0 License. structure appropriately according to local conditions, reduce the crops requiring high water consumption and generating poor 455 economic benefits in non-main producing areas, and regulate crop prices rationally, to balance the economic benefits of the 456 water-intensive crops in different regions. 457 The study reveals the synergies and trade-offs of crop water productivity and its economic value from the perspective of WF. of crop price unit data in the selection of research objects, it is still representative because the crops selected in this paper 467 accounts for more than 85% of the national crop production. As for the study perspective, this article focuses on contradictions 468 between water consumption and economic value creation in crop production. In fact, the ecological impacts on the environment 469 cannot be ignored. Therefore, further research is expected to tackle this limitation by including the ecological impacts on the 470 environment in a more comprehensive assessment. 471 472 Table 7. Comparison between production-based water footprint (PWF) of crops production in mainland China in the current 473 study and previous studies. 474 475 Table 8. Comparison between economic value-based water footprint (EWF) in the current results and previous studies. 476 477

Conclusions 478
Based on temporal and spatial evolution analysis of WF of China's crop production from a physical and economical perspective, 479 this study makes a comprehensive assessment by constructing a SI between PWF and EWF, and reveals the synergies and 480 trade-offs of crop water productivity and its economic value. Results show that: 481 (1) With the increase of yield unit and price unit, the PWF and EWF of crop production both showed a decreasing trend, and 482 the EWF decreased more significantly. The change of WF of cash crops was more obvious than that of grain crops. 483 (2) Compared to cash crops, WF of grain crops had a more significant spatial correlation, and the spatial distribution of PWF 484 was similar to that of EWF. H-H areas mainly gathered in north-western China, while L-L areas in southeast coastal provinces. 485 The average Moran's I of EWF (0.482) was higher than that of PWF (0.263).