Using a Grey Niche Model to Predict the Water Consumption in 31 Regions of China

Round 1

Reviewer 1 Report

This paper does not highlight with enough references in the Introduction the problem it is trying to address.
There is not a Background (not even part of the Introduction) section to show how previous studies have been developed globally to foster water conservation at the industry and residential levels.
It requires a systematic literature review based on:
Water Demand
Modeling and Forecasting Water Demand
Utilities Policies on Water Demand

The method section is poorly explained and does not provide context to the application.

I found the content unsalvageable for publication

 

Author Response

1. This paper does not highlight with enough references in the Introduction the problem it is trying to address.

There is not a Background (not even part of the Introduction) section to show how previous studies have been developed globally to foster water conservation at the industry and residential levels.

It requires a systematic literature review based on:

Water Demand

Modeling and Forecasting Water Demand

Utilities Policies on Water Demand

The Introduction has been revised in its entirety to add a background description showing global forecasts for water usage efforts and a systematic review of the literature.

Global water shortage is a difficult problem that human beings need to face together. At present, land freshwater assets account for solely 6% of the international water assets, of which 77.2% are allotted in Antarctica, 22.4% are distributed in the depths of the earth which are extremely tough to develop, and solely 0.4% of freshwater can be used for human life. Efficient conservation and utilization of water resources is of high-quality value to mankind, and water resources are also the basic guarantee of economic development and construction. Scientific prediction of water consumption is the premise and foundation of making water resources improvement and utilization plan, which is of great significance and value to realize the rational allocation of water resources and coordinate the improvement of social and economic sectors [1].

The prediction of water consumption has been carried out all over the world, and the selection of water consumption prediction methods has an important influence on the prediction results [2-4]. Most scholars use traditional forecasting models for forecasting. For example, Zhu Bo et al. applied ARIMA entity model to Hefei water consumption coding sequence to forecast and analyze the water consumption of Hefei in the next two years [5]. Leon Lee P. et al. used soft computing technology to predict water consumption in Tobago and other areas [6]. Li Xuan et al. took Jining as a typical city to analyze the trend of water consumption structure, and found that its water consumption decreased annually and the water consumption structure developed in a stable direction [7]. However, the change of water consumption is nonlinear, and the above methods have some limitations in practical application. At present, the deep learning and integration method has a good application in water consumption forecasting. For example, Guancheng Guo et al. studied the application possible of deep mastering in momentary water consumption forecasting, and established a gated recursive unit community model to forecast temporary water consumption [8]. In order to efficaciously deal with the weekly and annual seasonal integrated models that influence the data, a program of developing integrated neural network model is put forward for short-term prediction of water consumption of small-scale water furnish device [9]. Moreover, BP neural network is widely used in water resources prediction because of its strong self-learning ability and generalization ability [10]. Although deep learning can solve the problem of prediction accuracy, it needs a large amount of data to support its prediction, and it can't consider the influencing factors of water consumption.

For utilities on water consumption, accurate estimation usually needs to solve the problems of multi-measurement, mixed model and space-time, and single model has strong limitations in engineering application. Julia K. Ambrosio et al. established an accurate water consumption prediction model through the integrated framework of committee machine, that is, single machine learning model, and two water demand data sets from franca, Brazil are analyzed, which proves that their response is better than any single component model [11]. Then, a superposition model based on four models was applied to the real consumption data in the UK, and it was found that it was superior to other water consumption forecasting techniques [12]. Salah L. Z. et al. put forward a new method of forecasting monthly water consumption based on various weather conditions, which uses three methods, including discrete wavelet analysis, to predict water consumption under variable scenarios [13]. From the above, it has been proved that the model composed of multiple models has strong application value in the research of water consumption forecasting.

Recently, the shortage of water resources has become a significant problem in China, it has the risk of restricting the process of urbanization in China [14]. The development of China's urban economy has brought about obvious changes in industrial structure and water consumption [15]. Scientific application of industrial structure change to analyze future water consumption is the premise and foundation of water resources management, and it is of great significance to solve the contradiction of water resources and the sustainable development of cities. Jiao Shixing and others used niche theory to analyze and conclude that industrial structure will affect water consumption [16]. Taking industrial structure as a factor to discuss the change of water consumption is also crucial [17-20]. However, there are few water resources data in China. When the abovementioned model is used alone for forecasting, incomplete data will lead to a big deviation in the forecasting results. As an important part of grey system theory, grey prediction model can correctly describe and effectively monitor the system by establishing differential equations to extract valuable information from some known information. The characteristics of grey prediction model make up for the shortcomings of incomplete water consumption information, so it is widely used [21-23]. ZhengranQiao et al. established fractional cumulative grey forecasting model (FGM(1,1)) to forecast the water consumption in various regions of China, and the results showed that the total water consumption had little change [24]. The traditional single variable grey forecasting model only considers the change of water consumption when forecasting water resources. However, the change of things is influenced by many factors, so grey multivariable is put forward and widely used. Xiangmei Meng et al. put the adjacent accumulation operator into the grey multivariate convolution model to foretell the annual water consumption in various regions of China under different GDP and population growth rates [25]. Grey multivariate model is not only applicable to the foretelling of water consumption with little data, but also can take into account the factors affecting water resources. Subsequently, Zhicun Xu et al. proposed a non-equidistant GM(1,1) model (CGM(1,1)), which has better fitting degree in prediction [26]. The CGM(1,1) model has higher prediction accuracy for non-equidistant influencing factors, and it is used as the basic model of the prediction model. Combining the first niche theory with the grey prediction theory to predict water consumption can not only consider the industrial change as the influencing factor, but also solve the problem of little water consumption data in China in recent years.

In order to solve the shortcomings of traditional model, such as low prediction accuracy, large data requirement of deep learning model, small consideration range of single model, etc. In this paper, the combination of niche theory and grey prediction method is used to analyze the change of China's industrial structure and its impact on China's water consumption. Firstly, the grey prediction model CGM(1,1) takes into account the factors affecting water consumption. It is concluded from the above that the economic development in China has caused obvious changes in industrial structure and water consumption, so the three major industries are used as influencing factors to predict water consumption. Secondly, the niche theory can calculate the three major industries in China, get the niche of each industry, and directly observe the change of regional industrial structure and its strength and weakness. Finally, taking niche as the weighting coefficient, the water consumption under the three major industries in each region is calculated, and the water consumption under the simultaneous influence of the three major industries in each region in the future is obtained. The water consumption forecasting model based on industrial niche (W-CGM(1,1)) put forward in this paper not only has the advantages of grey forecasting theory with less data and high accuracy. It also shows the change of industrial structure directly and calculates it as a part of the forecast, which makes the forecast result more accurate. The calculated data results can provide a basis for all parts of China to formulate water-saving measures under the change of industrial structure.

References

1. Zitian, C.; Heng, X. Combination prediction of water use structure in Hebei Province based on component data. Practice and Understanding of Mathematics, 2019, 49(15): 192-198.
2. Shaofeng, J.; Shifeng, Z.; Hong, Y.; Jun, X. The relationship between industrial water use and economic development: the water Kuznets curve. Journal of Natural Resources, 2004(03):279-284.
3. Qifeng, Z.; Xiaosi, T.; Peng, Z. Prediction and Analysis of Economic Development on Water Resources Demand. Journal of Hubei Second Normal University, 2020, 37(02): 56-59.
4. Ge, W.; Changlai, X.; Zhiwei, Q.; Xiujuan, L.; Fanao, M.; Ying, S. Water Resource Carrying Capacity Based on Water Demand Prediction in Chang-Ji Economic Circle. Water, 2020, 13(1).
5. Bo, Z.; Yezheng, L.; Feifei, Z. Annual Water Consumption Forecast of Hefei Based on ARIMA Model[J]. Academic Journal of Computing & Information Science, 2021, 4.0(3.0).
6. Leon, L. P.; Chaplot, B.; Solomon, A. Water consumption forecasting using soft computing – a case study, Trinidad and Tobago. Water Supply, 2020, 20(8).
7. Xuan,Li.; Longcang, S.; Chengpeng, L.; Zhonghua, Z.; Jiang, G. Analysis and prediction of water use structure in Jining City. Hydropower and Energy Science, 2017,35(06):26-29+146.
8. Guancheng, G.; Shuming, L.; Yipeng, W.; Junyu, L.; Ren, Z.; Xiaoyun, Z. Short-Term Water Demand Forecast Based on Deep Learning Method. Journal of Water Resources Planning and Management, 2018, 144(12).
9. Zanfei, A.; Menapace, A.; Granata, F.; Gargano, R.; Frisinghelli, M.; Righetti, M. An Ensemble Neural Network Model to Forecast Drinking Water Consumption. Journal of Water Resources Planning and Management, 2022, 148(5).
10. Chuang, M.; Daiqi, Z.; Ye, Z. Prediction method of water resources demand based on BP neural network based on improved whale algorithm. Computer Science, 2020, 47(S2): 486-490.
11. Julia, K. A.; Bruno, M. B.; Manuel, H.; Edevar, L.; Lubienska, R.; Joaquín, I. Committee Machines for Hourly Water Demand Forecasting in Water Supply Systems. Mathematical Problems in Engineering, 2019, 2019.
12. Xenochristou, M.; Kapelan, Z. An ensemble stacked model with bias correction for improved water demand forecasting. Urban Water Journal, 2020, 17(3).
13. Salah, L. Z.; Khalid, H.; Saleem, E.; Nabeel, S. S. A.; Hussein, A.; Sadik, K. G. A novel methodology to predict monthly municipal water demand based on weather variables scenario. Journal of King Saud University - Engineering Sciences, 2020.
14. Liping, Z.; Jun, X.; Zhifang, H. Analysis of China's water resources status and water resources security. Yangtze River Basin Resources and Environment, 2009, 18(02): 116-120.
15. Haozhe, Y.; Lijuan, L.; Jiuyi, L. Research on the construction of risk assessment model for water resources carrying capacity in Beijing-Tianjin-Hebei. Geographical Research, 2021, 40(09): 2623-2637.
16. Xingjiao, S.; Yunli, Q.; Zhouwang, A.; Yaqi, L.; Xingyin, Y.; Qinzhao, R.; Jianwei Z. Dynamic evolution analysis of water use structure and industrial structure in Anyang City based on ecological niche. Water Resources Protection, 2021, 37(01):79-85+109.
17. Gang, W.; Songling, Z.; Pengyun, Z.; Qingcheng, C. Discussion on the definition of niche and the improvement of the measurement formula of niche overlap. Journal of Ecology, 1984(02):119-127.
18. Ocampo, C. B.; Guzmán, R. L.; Moreno, M.; Castro, M. D. M.; Valderrama; Ardila, C.; Alexander, N. Integration of Phlebotomine Ecological Niche Modelling, and Mapping of Cutaneous Leishmaniasis Surveillance Data, to Identify Areas at Risk of Under-Estimation.. Acta tropica, 2021, 224.
19. Kolanowska, M.; Rewicz, A.; Baranow, P. Ecological niche modeling of the pantropical orchid Polystachyaconcreta (Or-chidaceae) and its response to climate change. Scientific reports, 2020, 10(1).
20. Zhihong, H.; Zengchuan, D.; Tao, Z.; Wei, X.; Xinxin, C.; Xiaoqi, G. Eco-friendly reservoir group scheduling model. Journal of Hohai University, 2020, 48( 03):202-208.
21. Xiaoshen, Z.; Haibo, W.; Wenchuan, W. Water resources prediction based on improved grey model. People's Yellow River, 2011,33(10):46-47.
22. HaiNi, D.; ShuHong, M.; Bing, S. Prediction of urban domestic water demand based on grey GM(1,N) self-memory model. Hydropower and Energy Science, 2010, 28(12): 19-20.
23. Pengcheng, Z.; Weijiang, Z.; Jun, L.; Jianning, S. Prediction of water consumption per 10,000 yuan GDP in Ningxia based on grey system theory. Anhui Agricultural Science, 2011, 39(01): 384-385.
24. Zhengran, Q.; Lifeng, W.; Zizhao, Y. Prediction of Water Consumption in 31 Provinces of China Based on FGM(1,1) Model. Clean, 2022.
25. Xiangmei, M.; Leping, T.; Chen, Y.; Lifeng, W. Forecast of annual water consumption in 31 regions of China considering GDP and population. Sustainable Production and Consumption, 2021, 27(prepublish).
26. Zhicun, X.; Lianyi, L.; Lifeng, W. Forecasting the carbon dioxide emissions in 53 countries and regions using a non-equigap grey model. Environmental science and pollution research international, 2020.

2. The method section is poorly explained and does not provide context to the application.

The method section of the article has been rewritten to explain the application's notation.

2. Materials and Methods

2.1. Regions and data

This paper will conduct research on 31 regions of China, as shown in Table 1. The detailed geographic locations of China’s 31 provinces are shown in Figure 1. While China’s international status has improved, the tertiary industry has also reached a new era of rapid envelopment. The proportion of China’s tertiary industry in the national economy has continued to increase (http://www.stats.gov.cn/ztjc/ztfx/jwxlfxbg/200205/t2002 0530_35912.html). What follows is the shortage of water supply. In addition, the developments of industries in different regions are different, and there are also many differences in the demand for water resources. Therefore, it is essential to conduct industry dynamic analysis and water consumption forecasts across China.

Table 1. Study regions.

1	Beijing	12	Anhui	23	Sichuan
2	Tianjin	13	Fujian	24	Guizhou
3	Hebei	14	Jiangxi	25	Yunnan
4	Shanxi	15	Shandong	26	Tibet
5	Inner Mongolia	16	Henan	27	Shaanxi
6	Liaoning	17	Hubei	28	Gansu
7	Jilin	18	Hunan	29	Qinghai
8	Heilongjiang	19	Guangdong	30	Ningxia
9	Shanghai	20	Guangxi	31	Xinjiang
10	Jiangsu	21	Hainan
11	Zhejiang	22	Chongqing

Figure 1. Geographical location.

So that analyze the dynamic changes of China's industry and its impact on the total regional water consumption, this paper selects the industrial added value of various regions in China as the influencing factors to predict the regional water consumption as shown in Table 2. The data in this paper are the industrial added value and water consumption from 2014 to 2019, and the source is the China Statistical Yearbook. Due to the extremely limited number of samples, grey theory can be used for analysis. On this basis, the niche theory is used to analyze the dynamic change process of industrial structure. Combining niche theory and grey prediction theory to predict water consumption in China.

Table 2. Type of data.

Influencing factors

The added value of the primary industry

The added value of the secondary industry

The added value of the tertiary industry

2.2. Predictive model

2.2.1. CGM(1,1) model

Let be the original sequence and be the influencing factor sequence. If the gap is , is called a non-equigap sequence. is the order cumulative sequence of

,

(1)

where

,

(2)

the range of is . The specific value of is obtained by the particle swarm optimization algorithm (PSO), which is based on the smallest prediction error.

The contiguous mean generation sequence of the non-equigap sequence is ,where

.

(3)

The mean value formula of CGM(1,1) is

.

(4)

Its whitening differential equation is

.

(5)

The development coefficient is and grey action is . The least square estimation of CGM(1,1) model satisfies

,

(6)

where

.

(7)

The initial condition of the differential equation in Eq.(4) is：

.	(8)
,	(9)

is time responsive.

Therefore, the reduction sequence of can be obtained as

.

(10)

In gray systems, Eq.(10) is usually used to measure model stability

.

(11)

The prediction model generally uses the MAPE value as the evaluation standard (Table 3.). It is generally believed that the model has high predictive power when the MAPE is lower than 10%. The MAPE of the foretelling results of different models for the same set of data can be compared, and the model with better prediction effect can be selected.

Table 3. Test criteria for the forecasting model.

MAPE	Prediction accuracy
<10%	Higher
10%-20%	Better
20%-50%	Reasonable
>50%	Weaker

2.2.2. Niche model of industrial structure

Industrial structure niche reflects the status and role of different types of industries in the regions [16]. The calculation formula of the industrial structure niche model is [17].

,

(12)

Where: is the niche of the similar industry, ,The larger the , the larger the value-added of the type industry, and the larger the contribution to GDP growth. is the added value of the type industry. is the dimensional conversion coefficient, which is related to the time scale, and the value is 1 when calculating. is the difference between the research year and the initial year type industry added value. respectively represent the added value of China’s industry is the mean value of the type industry niche.

,

(13)

where: is the utilization rate of water consumption by the type industry.

2.2.3. Grey niche model(W-CGM(1,1) model)

The niche model of industrial structure shows the changing trend of the industrial structure, and the CGM(1,1) model predicts the water consumption of each region under the influence of the industry. Using the industrial niche as the weight to weight the water consumption prediction results to obtain the final regional water consumption, that is, the W-CGM(1,1) model

.

(14)

Where,is the regional water consumption, is the regional water consumption under the influence factors of different industries, is the utilization rate of water consumption by the type industry.

2.3. Calculation process

These standardized data are used to predict water consumption in 31 regions and cities in China. The prediction result is shown in Figure 2.

Figure 2. The forecasting process of water consumption.

Taking the added value of the primary industry as an explanatory variable, the initial sequence is . Water consumption is the explained variable, and the initial sequence is .

Step 1 The is calculated

.

Step 2The optimal order is obtained by the particle swarm algorithm (PSO), and then the cumulative sequence is obtained

.

Step 3 The non-equidly spaced generation sequence is

.

Step 4 The expression of and is

.

The least square method calculates the parameters and

.

Step 5 Put the calculated parameters into Eq.(8)

.

Step 6 The cumulative reduction sequence is

.

Putting the calculation result into Eq.(10), the MAPE value is 1.23%.

Step 7 Calculate the added value of Beijing’s secondary industry from 2020 to 2025

.

Step 8 The time response sequence of Beijing in the next 6 years is

.

Step 9 Forecast value of water consumption in Beijing from 2020 to 2025 is

.

Through calculation, the water consumption of Beijing from 2020 to 2025 under the influence of the tertiary industries is shown in Table 4.

Table 4. Beijing’s water consumption under the influence of different industries from 2020 to 2025.

Year	Water consumption(BCM)	Water consumption(BCM)	Water consumption(BCM)
2020	41.03	41.26	42.06
2021	41.63	41.96	42.86
2022	42.19	42.70	43.65
2023	42.72	43.48	44.44
2024	43.22	44.32	45.23
2025	43.70	45.22	46.00

The industrial niche of Beijing from 2014 to 2019 is calculated by Eq.(12), and show the results in Table 5.

Year	First industry niche	Second industry niche	Tertiary industry niche
2014	0.007449	0.213064	0.779488
2015	0.005677	0.183924	0.742222
2016	0.004325	0.164773	0.686324
2017	0.00347	0.153513	0.650384
2018	0.003019	0.143672	0.624628
2019	0.002301	0.115662	0.597884
	0.005164	0.191785	0.803051

Table 5. Beijing’s industrial niche from 2014 to 2019.

Use the data in Table 5 to calculate Beijing’s water consumption in the next 6 years, and the effects are shown in Table 6. The forecasting results obtained by the W-CGM(1,1) model were compared with other models with better prediction effects. The traditional gray prediction model (GM(1,1)) and the gray prediction model (FGM(1,1)) after fractional accumulation are used here. The results show that the MAPE of the W-CGM(1,1) model is smaller, and the smaller error means higher prediction accuracy. The prediction results of the GM(1,1) model and the FGM(1,1) model are consistent, indicating that the improvement of the model only in terms of improving the cumulative calculation is not helpful for this research.

The W-CGM(1,1) model proposed in this paper first uses a small amount of data collected to predict water consumption according to the effects of different industrial factors. Secondly, the industry niche model is used to measure the three major industries in the region, and to show the changes in the industrial structure from 2014 to 2019. Finally, using the industrial niche as the weight to weight the predicted water consumption under the action of industrial factors, the regional water consumption in the next 6 years is obtained. The W-CGM(1,1) model solves practical problems such as less water consumption data, small data changes, and the role of influencing factors. The following will directly use the W-CGM(1,1) model to analyze and predict the industrial structure changes and water consumption in 31 regions of China.Table 6. Water consumption in Beijing from 2020 to 2025.

Year	Water consumption(BCM)
	Actual value	Fitted value
		W-CGM(1,1)	GM(1,1)	FGM(1,1)
2020	37.5	37.50	37.50	37.50
2021	38.2	38.04	38.00	38.00
2022	38.8	38.77	38.74	38.74
2023	39.5	39.44	39.48	39.48
2024	39.3	40.06	40.25	40.25
2025	41.7	41.20	41.02	41.02
MAPE		0.0104%	0.9502%	0.9502%
2020		41.90	41.81	41.81
2021		42.68	42.62	42.62
2022		43.46	43.44	43.44
2023		44.25	44.28	44.28
2024		45.04	45.14	45.14
2025		45.84	46.01	46.01

Author Response File: Author Response.pdf

Reviewer 2 Report

Dear authors,

The manuscript proposes a water consumption forecasting model for some regions in China. My major concerns are:

• Some parts of the manuscript are unclear. You should consistently revise the manuscript to make it more readable, putting more attention in the form and in the use of English. Also, you have to explain better your methodology. For instance, why and how you use the PSO in this methodology is never mentioned in the text.
• The introduction does not reflect the current state-of-art in water consumption forecasting. I suggest you to expand it, including the more recent and meaningful studies in this field.
• I still do not understand the novelty of you work. Why should a reader use such methods? Could a linear regression model fit as well? The lack of a benchmark is an issue.
• Which are the major differences with this work: https://doi.org/10.1002/clen.202200052 ? Also, are the datasets identical? Some parts are really similar, and I am not sure if such work is enough novel at this point.

Afterwards, I will provide my comments to the authors trying to help them improving the manuscript.

1. The abstract is unclear. You should explain better and more concisely what is in the manuscript and what is the scope of you work. Lines [10-13] are a bit confusing. You should highlight better which is the aim of the manuscript. In the keywords, the use of ’31 regions’ is a bit useless; I suggest something like ‘multiple case studies’ or something like that.

 

2. The introduction in not deep enough. It starts with an overview of problems related to the Chinese context, and it is followed by a short revision about Niche theory and water consumption forecasting. Unfortunately, the revision does not provide the more recent works in the field of water consumption forecasting and does not discuss about the trends of the research in this field. For instance, I am talking about the state-of-art techniques that belong to the world of deep learning and ensemble approaches. I suggest the authors to look at the below studies:
   • https://doi.org/10.1061/(ASCE)WR.1943-5452.0001540
   • https://doi.org/10.1155/2019/9765468
   • https://doi.org/10.1080/1573062X.2020.1758164
   • https://doi.org/10.1061/(ASCE)WR.1943-5452.0000992

From my perspective, such works have to be cited in a nowadays water consumption forecasting paper, at least to provide the context to the reader. The authors, could also explain that, despite the current state-of-art is more focused on deep learning techniques, they had to adopt another methodology due to the lack of data. I also suggest to dedicate at least another paragraph in the introduction to water consumption forecasting, because lines [37-38] are not correct (because the state-of-art is different) and certainly not enough.

At the end of introduction, you should highlight more why your model is interesting, and what is the novelty behind such work.

• Line 35 and Line 41: keep the citation format uniform
• Line 42 and Line 45: introduce an acronym before using it

 

 

3. Material and Methods

• Concerning sect 2.1, the section could be clearer. For instance, I would like to see some insight of such data, like a small figure of them in this section. Also, what do you mean for ‘economy’ at line 65? Which kind of data are this latter? I would expect at least some tables or figures that show the kind of data you are using.
  • Lines 57-58 you repeat different 3 times
  • Line 67: do you mean ‘adopted’?

 

• Concerning sect 2.2, this section is unclear to me. The authors should explain better the model, and not pretending the reader to understand what are all the symbols adopted. This section must be revised, adding an introduction where every symbol and operation is adequately explained.
  • Line 87: typo Table2is (maybe Table 2 show …)
  • Table 2: who said that? I totally disagree. The prediction accuracy depends on the case study and on the aim. Such table is incorrect.

 

• Concerning sect 2.2.2, this time you explained all the symbols. I do not understand why you did not do it also before. Please, pay attention to formatting uniformly all the manuscript (lines 94-101)
• Lines 102 with formula (13) are not explained. The authors must pay more attention on the manuscript.
• Concerning sect 2.3, this section is full of calculations, and I believe that should be moved to supplementary materials. In fact, I believe that this does not provide additional contents to the section.

The authors could maintain the figure, and just explain better the methodology. Also, this section is confusing, since at lines 130-134 the authors show some results through tables 4 and 5. This kind of material should not belong to this section. Furthermore, another proof of what I am saying is that you talk about a FGM model, which was never mentioned in the method explanation. The authors should put more effort on explaining their methods, which are totally not clear from their paragraph.

 

At the end of the methodology, the reason why you are using a PSO is unknown to me. I believe that everyone is nowadays aware on what a PSO is and how it works. However, since you used it, I believe that is a good practice to explain, at least with a small subsection, how it works. In this section, you could highlight better which is the purpose of such algorithm in your methodology, and how you used it.

 

4. Results and discussion

In this section, you should also highlight why your model is working well, and not just comment you prediction. I guess, that you should at least employ a benchmark, to compare performances. Also, I am expecting some numbers/metrics to understand why your approach is interesting for the scientific community.

 

5. Conclusions

As for the previous part of the manuscript, also in the conclusion the novelties of your work are not highlighted. Basically, at lines [209-229] you keep discussing about Chinese water resource usage and future trends. I think you should rewrite this section, talking about your methodology and about the approach you used and your results.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 3 Report

This study presents prediction of the water consumption in China's 31 regions  using grey niche model. The proposed model is combined the niche theory with the grey prediction method. The paper is well written, but some minor revisions are needed. These revisions are:

1) Results of prediction performance for proposed prediction model should be given in abstract.

2) In introduction section, a few sentences about the importance of niche theory are necessary for the transition from paragraph 1 to paragraph 2. Otherwise, it may be difficult for the reader to understand the subject. Before we talk about this theory, it might be better to highlight the importance of prediction/forecasting and why the grey prediction model is needed.

3) The introduction should be expanded with a summary of the literature in which grey prediction model has been used in recent years.

4) In Table 3, 4 and 5, units of values should be given. For example; m³ for water consumption. 

5) MAPE values should be given in four digits. In Table 5, Mape results are the same for GM(1,1) and FGM(1,1).

6) In Figure 3, 4 and 5, units of water consumption should be given. Additionally, these figures might look better if the results were plotted for the years 2014-2025.

7) In conclusion section, advantages of the proposed model for this study should be mentioned.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The manuscript presents fundamental flaws that prevent me to accept it. 

The main questions are:

What is the contribution of this work to the body of knowledge of forecasting water demand?

How good is the proposed method with respect to other published data sets?

It is hard to understand the document and the storytelling as the technical writing has serious problems.

Not suitable for publication

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Dear authors,

I appreciated and noticed that you made a significant effort for improving your manuscript. Still, I believe that you should improve the English asking for a professional proofreading, since it is evident that the manuscript is demanding for a revision of the language. My comments are listed below:

• Abstract and Introduction have improved, and the authors followed my advices. However, your claim at lines 14-15 is not acceptable, since you are not making a comparison with traditional water consumption prediction models as we discussed in the previous revision. Please remove this sentence.
• Lines 153-154 you say:’ The specific value of r is obtained by the particle swarm optimization algorithm (PSO), which is based on the smallest prediction error.’. This sentence does not make sense. Probably you would say something like ‘The specific value of r is obtained by the particle swarm optimization algorithm (PSO), which aims at finding the optimal value that minimize the prediction error’ or something like that.
• At line 194 you say ‘The prediction results are shown in Figure 2’, which are not. You probably would like to link the diagram of your procedure in Figure 2.
• I still would like you to improve the section 2.2.1. You don’t have to expect that the reader understands what you are doing. For example, I guess that x is the water consumption. You could state this at the beginning of this section, explaining that during this problem x represents the water consumption.

Anyway, I believe that the study has improved compared to the previous version. But I still feel that its main issue is the language.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 3

Reviewer 1 Report

The revised article presents multiple flaws that are not salvageable to consider for publication

Author Response

1.Please condense the literature review in the introduction of the first part, and improve the contribution and goals.

The literature is re-organized and condensed to strengthen the expression of the contribution and goals of this paper. Water consumption forecasting research is of great significance to water resources. Firstly, it is analyzed that different water consumption forecasting methods have a great influence on the water consumption forecasting structure. Secondly, it is found that the combined prediction model can improve the prediction accuracy of water consumption. Third, China's regional economic development is closely related to water consumption. Due to the small amount of available data, a grey prediction model was adopted. Finally, a multivariate grey combination model is proposed to better analyze the impact of industrial structure on water consumption.

Global water shortage is a difficult problem that human beings need to face together. Terrestrial freshwater assets account for only 6% of international water assets. Among them, 77.2% are divided in Antarctica, and 22.4% are distributed in the deep earth, which is difficult to develop. Only 0.4% of freshwater is available for human life. Efficient conservation and use of water is of high-quality value to humanity, and water is also the basic guarantee of economic development and construction. Scientific fore-casting of water consumption is the premise and basis for drafting water improvement and use plans. This is of great significance and value for realizing the rational share of water and coordinating improving social and economic[1].

Predicting water consumption has been carried out all over the world, and select water consumption prediction methods have an important influence on the predicted results [2-4]. Most scholars use traditional forecasting models for forecasting. For example, Zhu Bo and others applied ARIMA entity model to Hefei water consumption coding sequence to forecast and analyze the water consumption of Hefei in the next two years [5]. Leon Lee P. used soft computing technology to predict water consumption in Tobago and other areas [6]. Li Xuan took Jining as a typical city to analyze the trend of water consumption, and found that its water consumption decreased yearly and the water consumption developed in a stable direction [7]. However, the change of water consumption is nonlinear, and methods have some limits in sensible application. The deepest learning and integration method has a good application in water consumption forecasting. For example, Guancheng Guo studied the application possibilities of deep mastering in momentary water consumption forecasting, and settled a gated recursive unit community model to forecast temporary water consumption [8]. To efficaciously deal with the weekly and annual seasonal integrated models that influence the data, a program of developing an integrated neural network model forward for short-term prediction of water consumption of small-scale water supply [9]. However, BP neural network for water quality prediction because of its strong self-learning skill and generalization skill [10]. Although deep learning can solve the problem of prediction accuracy, it needs many data to support its prediction, and it can't consider the influencing causes of water consumption.

For water consumption, accurate estimation usually needs to solve the problems of multi-measurement, mixed model and spacetime, and single model has strong limits in engineering application. Julia K.A. builds an accurate water consumption prediction model through an ensemble of committee machines, a single machine learning model. Two water demand datasets from Franca, Brazil, marching that their responses were better than any single-component model [11]. Then, a superposition model based on four models was applied to the real consumption data in the UK, and it was superior to other water consumption forecasting techniques [12]. Salah L. Z. puts forward a new method of forecasting monthly water consumption based on various weather climates, which uses three methods, including discrete wavelet analysis, to predict water consumption under variable scenarios [13]. It has the model composed of multiple models has strong application value in the research of water consumption forecasting.

Recently, the shortage of water has become a significant problem in China, it has the risk of restricting the process of urbanization in China [14]. Under the condition of limited water, it is necessary to keep the continuous growth of economic scale. Besides strengthening water management, promoting water-saving technologies and appliances and other water-saving measures. It is necessary to study how to reduce water consumption according to changes in economic and industrial structure. This will reduce the impact of short water supply on the national economy. Developing China's urban economy has brought about the obvious changes in industrial structure and water consumption [15]. Scientific application of industrial structure change to analyze future water consumption is the premise and foundation of water management, and it is of great significance to solve contradict water and the sustainable development of cities. Jiao Shixing and others used niche theory to analyze and assume that industrial structure will affect water consumption [16]. Taking industrial structure as a reason to discuss the change of water consumption is also important [17-20]. However, there are few water data in China. When the model is used alone for forecasting, incomplete data will lead to a big deviation in the forecasting results. As an important part of gray theory, the gray prediction model can correctly describe and effectively note the system by proving differential equations to extract valuable information from some known information. The characteristics of the gray prediction model make up for the lack of incomplete water consumption information, so it is widely used [21-23]. ZhengranQiao settled fractional cumulative gray forecasting model (FGM(1,1)) to forecast the water consumption in various regions of China, and the results showed the total water consumption had little change [24]. The traditional single variable gray forecasting model only considers the change of water consumption when forecasting water. However, the change of things is influenced by many reasons, so gray multivariable is in widespread use. Xiangmei Meng puts the near collection operator into the gray multivariate convolution model to foretell the annual water consumption in various regions of China under different GDP and population growth rates [25]. The gray multivariate model is not only applicable to the foretelling of water consumption with little data, but also can consider the reasons affecting water. Afterwards, Zhicun Xu proposed a non-equidistant GM(1,1) model (CGM(1,1)), which has better fitting degree in prediction [26]. The CGM(1,1) model has higher prediction accuracy for non-equidistant influencing reasons, and it is used as the basic model of the prediction model. Combining the first niche theory with the gray prediction theory to predict water consumption cannot only consider the industrial structure change as the influencing reasons, but also solve the problem of little water consumption data in China in recent years.

To solve the faults of traditional model, such as low prediction accuracy, large data need of a deep learning model, small consideration range of a single model. In this paper, combine niche theory and gray prediction method is used to analyze the change of China's industrial structure and its impact on China's water consumption. First, the gray prediction model CGM(1,1) considers the reasons affecting water consumption. It is concluded from the economic development in China has caused obvious changes in industrial structure and water consumption, so the three major industries are used as influencing reasons to predict water consumption. Second, the niche theory can calculate the three major industries in China, get the niche of each industry, and directly view the change of regional industry structure and its strength and weakness. Finally, using the niche coefficient as the weighting coefficient, the water consumption under the three major industries structure in each region is calculated. Exist the water consumption under the simultaneous influence of the three major industries in each region in the future. The water consumption forecasting model based on industrial niche (W-CGM(1,1)) put forward in this paper not only has the advantages of gray forecasting theory with less data and high accuracy. It also shows the change of industrial structure directly and calculates it as a part of the forecast, which makes the forecast result more accurate. The calculated data results can provide a basis for all parts of China to develop water-saving measures under the change of industrial structure.

2.Please describe the keywords accurately.

Strengthen the description of keywords.

Keywords: Niche; gray multivariate prediction; industrial structure; water consumption; multiple case studies

3.The suggestion is too general, please mention in the conclusion how to reduce water consumption in the process of industrial change.

Two measures are proposed to reduce water consumption in the process of industrial change.

Through industrial structure adjustment, two suggestions are put forward in best share of incoming water: (1) Continue industrial adjustment and gradually show a water-saving industrial. To achieve economic development under the existing water, modern service industries such as the financial industry should be vigorously developed. Gradually strengthen the "three-two-one" industrial structure, and adjust in the direction that is conducive to economic development and water conservation. (2) Timely adjust the regional industrial layout and economic development planning. Most of China's cities have relocated their urban centers to the outside world, focusing on developing of tertiary industries in urban centers to reduce urban water consumption. At the same time, the overall industrial layout is reintegrated nationally. The shortage of water, environmental degradation and other modernization problems. It is difficult to rely on a single area to solve it on its own. It is necessary to coordinate the entire Chinese region to consider allocates industries, population and resources, to promote China to form an economic circle with overall coordinated development.

4.The conclusion is short, please strengthen it.

The conclusions are expanded to include specific water-saving measures.

This is the first paper that combines niche theory and gray theory to calculate water consumption in various regions of China under changes in industrial structure. This paper uses W-CGM(1,1) model to analyze industrial changes in China from 2014 to 2019 and forecast the change of water consumption in 31 regions of China in the next 6 years. The W-CGM(1,1) model makes good use of a few industries and water consumption data collected, and predicts water consumption as an influencing reason. And calculate the industrial niche shows the industrial structure changes in China from 2014 to 2019 in 31 regions. According to the weights of industry niche, the water consumption under various reasons is weighted and the water consumption in the next 6 years is finally obtained. The result is not just a single data on the amount of water used in the future, but water used under industrial action, which is not achieved by other single and combined models.

Under the situation of limited water, the adjustment of industrial structure is conducive to the recognition of sustainable financial development. According to the calculation results of the W-CGM(1,1) model, most regions in my country are undergoing or have completed transform industrial structure. There is no significant change in water consumption in the next 6 years compared with the past. The booming Chinese industry has caused a serious waste of water resources. Resulting stift industrial water-saving measures have slowed or reduced the growth of industrial water consumption. And China's service industry progress is the sure result of modernization, which will inevitably increase water consumption. But the outbreak of COVID-19 has made growth in China's tertiary sector even slower. Therefore, there is no crucial increment in water consumption under the influence of the three major industries. Combined with China's purpose of establishing a water-saving country and research conclusions. Through industrial structure adjustment, two suggestions are put forward in best share of incoming water: (1) Continue industrial adjustment and gradually show a wa-ter-saving industrial. To achieve economic development under the existing water, modern service industries such as the financial industry should be vigorously developed. Gradually strengthen the "three-two-one" industrial structure, and adjust in the direction that is conducive to economic development and water conservation. (2) Timely adjust the regional industrial layout and economic development planning. Most of China's cities have relocated their urban centers to the outside world, focusing on developing of tertiary industries in urban centers to reduce urban water consumption. At the same time, the overall industrial layout is reintegrated nationally. The shortage of water, environmental degradation and other modernization problems. It is difficult to rely on a single area to solve it on its own. It is necessary to coordinate the entire Chinese region to consider allocates industries, population and resources, to promote China to form an economic circle with overall coordinated development.

5.Please rewrite the abstract and add the key research questions, methods and results, and highlight the superiority of the method.

The abstract has been rewritten. Specifically from the research questions, methods and conclusions of this paper. Demonstrate the superiority of the research method.

Regional development brings significant changes in industrial structure and water consumption. Research the trend of water consumption by changes in industrial structure to promote water conservation. The gray niche model describes the industrial changes in China and analyzes the water consumption of different leading industries. Use data from 2014 to 2019 and take the economy as the influencing reason and the industrial niche as the weight, the water consumption is predicted. While improving the forecasting accuracy, the water consumption forecasting has been strengthened. The calculation results show the regional industry is undergoing transformation, and the tertiary industry is rising in the national economy. The successful implementation of industrial water-saving measures has kept the water consumption of industrially developed cities stable. The rapid development of the tertiary industry will increase water consumption. Incorporating changes in industrial structure into water use analysis allows the Chinese government to draft water conservation policies for various industries.

6.It is better to add relevant policies and some suggestions on the change of industrial structure, which is the main point of the study. So that researchers can provide some innovative suggestions and policies.

Several suggestions are given for the industrial structure adjustment policy, so as to facilitate the followup researchers to provide innovative suggestions.

Under the situation of limited water, the adjustment of industrial structure is conducive to the recognition of sustainable financial development. According to the calculation results of the W-CGM(1,1) model, most regions in my country are undergoing or have completed transform industrial structure. There is no significant change in water consumption in the next 6 years compared with the past. The booming Chinese industry has caused a serious waste of water resources. Resulting stift industrial water-saving measures have slowed or reduced the growth of industrial water consumption. And China's service industry progress is the sure result of modernization, which will inevitably increase water consumption. But the outbreak of COVID-19 has made growth in China's tertiary sector even slower. Therefore, there is no crucial increment in water consumption under the influence of the three major industries. Combined with China's purpose of establishing a water-saving country and research conclusions. Through industrial structure adjustment, two suggestions are put forward in best share of incoming water: (1) Continue industrial adjustment and gradually show a water-saving industrial. To achieve economic development under the existing water, modern service industries such as the financial industry should be vigorously developed. Gradually strengthen the "three-two-one" industrial structure, and adjust in the direction that is conducive to economic development and water conservation. (2) Timely adjust the regional industrial layout and economic development planning. Most of China's cities have relocated their urban centers to the outside world, focusing on developing of tertiary industries in urban centers to reduce urban water consumption. At the same time, the overall industrial layout is reintegrated nationally. The shortage of water, environmental degradation and other modernization problems. It is difficult to rely on a single area to solve it on its own. It is necessary to coordinate the entire Chinese region to consider allocates industries, population and resources, to promote China to form an economic circle with overall coordinated development.

Author Response File: Author Response.docx

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.