Prediction of Blue Water Footprint Accounting for Water Treatment Plants in Kuantan River Basin

Water treatment plants (WTPs) in Kuantan river basin abstracts water from the blue water source, which is the Kuantan river. Therefore, by accounting the blue water footprint (WFb), the overall water consumption for all five WTPs namely; Sungai Lembing, Bukit Sagu, Panching, Semambu, and Bukit Ubi can be obtained. In order to predict the value, Backpropagation method is the best method to be used due to the historical data obtained from the WFb accounting for all five WTPs above. The objective of this study is to predict the overall blue water consumption for water treatment plants located along Kuantan river basin using Backpropagation method in artificial neural network. In this study, WFb has been accounted throughout all water treatment plants by using reference from water footprint manual. Then, the WFb will undergo a series of testing using application in MATLAB software in order to predict the future value based on historical data from 2015 until 2016. As a result, the total WFb accounting obtained was 190,543,378.2 m 3 /day, while the total maximum capacity of the WTPs was 189,654,000 m 3 /day. Hence, the prediction value that kept increasing will not be able to cater the future demand due to unstoppable urbanization.


Introduction
In Malaysia, most water sources come from blue water resources, which are either groundwater or surface water, and mostly are from the river [1]. Water resources are involved in the abstraction process; water abstraction is increasing remarkably and globally due to increasing demand of water and therefore, this will contribute to the imbalanced ecosystem and its function [2]. Moreover, the sustainability of water resources is believed to be dependent on the availability of water resources from the intake [3]. Water resources are part of the ecosystem and the sustainability of ecosystem is defined as when the ecosystem is able to meet human needs without causing any harm to the environment [4]. Due to that, there are many studies done in terms of environmental sustainability of water supply. Among them are the effects of human intervention and climate change to water resources [5], the effects of climate change to the water resources in Yellow river basin in China [6], sustainable urban water resources management by considering the Life Cycle Assessment (LCA) in terms of uncertain water utilization [7], and a modelling study of future water resources for food production in the involved river basin in South Asia [8]. Currently, the water intake, water discharge and all volumes of water along the conventional Water Treatment Plants (WTPs) are recorded. However, the type of water involved in the process is not well defined. In Water Footprint (WF), water is categorized into three types, namely; Blue WF, Green WF and Grey WF. As water resources in Malaysia are dependent on surface water sources such as rivers and reservoirs [9], some places in Malaysia are dependent on groundwater resources but it is not Malaysia's main water source [1]. Thus, better management of water resources is essential to ensure the continuous presence of clean water. Since the use of water has created good and bad effects, thus the WF concept is introduced to the public, especially to promote awareness. More significantly, it can be said that WF assessment is a holistic approach to an efficient use of the resource as water is an imperative resource for human and ecosystem health. Generally, Water Footprint (WF) is defined as the amount of water used to produce a product or service in a country [10]. Several studies have revealed that most researches in WF has also emphasized the use of water in wastewater treatment plant [11], citrus production [12], olive growing system [13], iron and steel industry [14], potato production [15], energy production and supply [16], winemaking industry [17], transport fuels [18], a pair of jeans [19], ethanol production [20], paddy rice system [21], coffee and tea consumption [22], farm animal products [23], humanity [24], tourism in Spain [25], food waste [26], bioenergy [27], crops and derived crops products [28], and for the service sector of water scarce in gaming industry of Macao [29]. As stated, most of the researches were conducted on product-basis and according to Hoekstra, by using WF approach, researcher would be able to assess sustainable water allocation [3]. Previously, several reports have shown that the ANN application is useful in hydrology field, especially in forecasting and predicting parameters [30]. Applications of ANN by researchers in analyzing water-based cases are quite enormous and directly related to one of our biggest challenges in managing the water issues, that is the water quality. ANN applications are playing a major role in predicting water quality parameters [31] and in the prediction of the groundwater recovery cost for drinking use based on the quality of water resources. One of the interesting findings is the development of prediction modelling by using ANN to predict the monthly values of two parameters for water quality of Delaware River, Pennsylvania [32]. Other than that, ANNs can also predict solar radiation accurately when compared with conventional methods [33]. Moreover, a research is done to create a model that allows the prediction of the flow of "Tomebamba" river at any specific day of a year [34], and to predict a variety of ocean water quality parameters [35] by using ANN. In addition, ANN was used in predicting the water quality of polluted aquifer [36]. Commonly, in time series prediction, which is basically describing a future value based on the current and historical data, is called as the backpropagation method. Recently, ANN has been massively used by many researchers due to its less-complex process, such as in predicting the NOx emission of diesel engine by improving the linear and nonlinear auto-regressive model [37]. Other than that, realtime damage detection can also be analyzed using time varying auto-regressive model and recursive principal components [38]. Backpropagation (BP), or also known as propagation of error is one of the methods with the ability to teach artificial neural network to perform tasks that are instructed to them. It was initially portrayed by Arthur E. Bryson and Yu-Chi Ho in 1969, and on 1986, this method was recognized by David E. Rumelhart, Geoffrey E. Hinton and Ronald J. Williams through their work, thus becoming popular in ANN research [39]. Hence, this method will be used in time series tool of prediction in ANN. However, before uploading a historical data, a problem need to be selected based on what prediction is to be analyzed. Hence, this study aims to predict the overall blue water consumption for water treatment plants located in Kuantan river basin using backpropagation method in artificial neural network based on overall WFb accounting.

Methodology
Firstly, total blue water consumption is calculated by using water footprint manual to get the total WFb for each WTP from year 2015 to 2017. Secondly, WFb for each WTP is predicted for the next three years' trend in order to know the rate of changes.

Blue Water Footprint Accounting
Blue water footprint (WFb) is the amount of surface water and groundwater required (evaporated or directly used) in each stage of the water treatment process. The method of blue water footprint calculation is expressed in Equation (1): (1) Definition 2.1: ET 0 is reference evapotranspiration in m 3 /day, p is mean daily percentage of annual daytime hours, T is mean daily temperature (°C), rainfall is amount of precipitation within the WTP area, and Area is open tank area that is exposed to precipitation and evaporation.

Definition 2.2:
ET 0 is the rate of blue water evaporation calculated using Blaney-Criddle method. The formula of Blaney-Criddle method are expressed as in Eq. (2). (2) The formula to calculate T 0 mean is expressed in Eq. (3). Meanwhile, Table 1 is used to determine the value of p. To be able to determine the p value, it is essential to know the approximate latitude of the area and the number of degrees north or south of the equator. In this study, all WTPs are located in Latitude 3.0.
Since the latitude of intake and plant station are not stated in the table of mean daily percentage of annual daytime hours for different latitude, thus, the interpolation must be done by using Eq. (3). After all total WFb for all WTPs involved have been accounted for, those values will be compared with the capacity of all WTPs in order to know whether the actual amount of water used in the process is still under the capacity of WTP or not. This method must be made before predicting the trend for the future.

Normalization of Data
The extreme value of data set in this study were normalized by using min-max method. Min-max normalization performs a linear transformation on the original data. This method helps to reduce the effects of outliers, i.e. extreme values in the input as well as output data (total blue water footprint), and makes the scaled data easier to be modelled in ANN. The formula for data normalization is: (4)

Definition 2.3:
Vn is the value of normalization, in order to get the normalized value, current value needs to be subtracted by minimum value of the data, then, the data is divided by maximum value minus the minimum value of the data set.

+
In this study, predicted data will be obtained by undergoing a series of historical data from 2015 until 2016 for all WTPs. This method is named backpropagation method, total WFb accounts for three years period undergo a series of testing after all the data have been taken and the amount of daily WF for three years have been accounted, the value will be listed and used as training elements in order to get the predicted value of WF in the future. Neural Fitting Application in ANN will be used because the result that will be obtained is a numeric value. A two-layer feed-forward network with sigmoid hidden neurons and linear output neurons (fitnet) can fit multi-dimensional mapping problems arbitrarily well, given the consistent data and enough neurons in its hidden layer. The network will be trained with scaled-conjugate backpropagation algorithm because this algorithm requires less memory. Training automatically stops when generalization stops improving, as indicated by an increase in the mean squared error of the validation samples. The WF value for two years period will be inserted in this application to obtain the prediction value. The Neural Fitting app helps select the data, create and train a network, and evaluate its performance using mean squared error and regression analysis. The input variables are presented in Table 2

Results and Analysis
The 3.1 sub-topic will be presenting the total blue water footprint WFb in all WTPs in Kuantan river basin. Meanwhile, sub-topic 3.2 will be presenting the prediction of all WFb in Kuantan River basin.  As seen in Fig. 2

Panching Water Treatment Plant
Kuantan River has become the source of water intake for Panching Water Treatment Plant. It is the second bigger WTP in Kuantan district and located at 3.5020, 103.1238, where the areas of water supply are both for Penur and Kualan Kuantan. The capacity of 7,000 m 3 /hour and expected to be 6.13 x 10 7 m 3 /year will be able to sufficiently provide 389,000 people with treated water including the residential area and as well as 5600 ha of industrial area.  Therefore, both values were still under the capacity, which is 6.13 x 10 7 m 3 /year.

Semambu Water Treatment Plant
Sub-districts of Sungai Karang and Beserah has received treated water from the Semambu Water Treatment Plant. Although located at 3.521, 103.2016 and 18 km away from the intake location, it is recognized as the biggest WTP's capacity with 12,000 m 3 /hour and will increase to 1.05 x 10 8 m 3 /year. The average population of 92,800 occupied in areas of 30,300 ha benefited from this water treatment plant. Residential area of Kotasas and industrial park of Semambu and Gebeng has become the major receiver of treated water.

Bukit Ubi Water Treatment Plant
The treated water supply for commercial area in Kuala Kuantan is being covered by the one and only, Bukit Ubi Water Treatment Plant. Located at 3.4920, 103.1973 at the centre of the town area, the maximum area of water distributed to that particular commer-cial area is 1500 m 3 /hour and with an exponential growth of 1.  The capacity of this WTP is 1.31 x 10 7 m 3 /year; therefore, the value of WFb recorded in 2016 was exceeded the capacity value. Rainfall amount was the main factor of the increment however, water demand due to urbanization is also a consideration factor due to the area of supply within the town area is occupied with commercial buildings and offices.

Comparison between Total Blue Water Footprint and WTPs Capacity in Kuantan River Basin
In this section, total WFb is analyzed based on the overall capacity of all WTPs located in Kuantan river basin and abstracted water from Kuantan river. Therefore, Fig. 7 [40]. This can also be adapted in Malaysia especially to face future water stress due to urbanization.

WFb Prediction for Kuantan River Basin
For the purpose of finding the optimal architecture of the ANN Table  3 shows the RSME values for training and testing data as a function of the number of hidden layers.  Fig. 9 represents the comparison between calculated data and prediction trend value based on normalized data of total WFb based on hidden neuron chosen for training data in ANN. As seen in Fig. 9, most of WFb is predicted to decrease from the current trend in Sungai Lembing Water Treatment Plant. The capacity for this WTP is 2.19 x 10 6 m 3 /year; therefore, based on this prediction, the need of opening new treatment plant can be on hold for Sungai Lembing because the predicted and current trends were still under its capacity. Meanwhile, in Bukit Sagu Water Treatment Plant, most of prediction trend are increasing, as shown in Fig. 10. However, as this WTP has yet to exceed its capacity, therefore, the increment in prediction will not affect the main function of this WTP, which is to supply water to FELDA Bukit Sagu. In Panching Water Treatment Plant, prediction value of WFb showed that an increasing trend from current trend then decreasing. This will not affect the function of this WTP as the capacity of this WTP is still broad and can cater more water for consumer in the future. Despite that, Panching Water Treatment Plant is new compared to other WTPs and is expected to be able to cater the demands for more years to come. As shown in Fig. 12, although the current value has exceeded the capacity of Semambu Water Treatment Plant, but predicted value showed trend that the amount of total WFb will decrease below than its capacity. This is after certain times of training in the ANN, variables that had contributed to the amount of total WFb have been normalized in the process and the result is shown, as in Fig. 12.

Conclusion
In summary, the determination of blue water footprint for water treatment process in Kuantan river basin can be determined by calculating the blue water footprint (WFb) at each WTP that is involved in this study through the design of water treatment process while considering certain parameters. The dependencies of the parameters such as water intake, rainfall intensity towards the uncertainty in many aspects like weather shall be specifically addressed in the accounting of blue water footprint. Thus, WFb has been calculated and the highest is recorded in 2016, that is at 190,543,378.22 m 3 /year. Therefore, WFb in Kuantan river basin is increasing by the year due to the increase in local population that subsequently leads to urbanization and also uncertainties of monsoonal changes (rainfall intensity).
After comparing total WFb of Kuantan river basin with the capacity of all WTPs within the same river basin, it showed that some of the WTPs were not be able to stay below the capacity such as in Sungai Lembing Water Treatment Plant, Semambu Water Treatment Plant and Bukit Ubi Water Treatment Plant . Hence, mitigation planning to cater the need of water and to ensure the amount of water intake will be sustainable for more years to come must be implemented. Although those WTPs are currently still able to supply treated water to consumers, however, the actual amount of water involved in the process as accounted by using water footprint approach has shown negative impact. This is because water footprint accounting is the most accurate approach in calculating the total amount of water utilized in process in producing goods and services [41]. Next objective is to predict the sustainability of water supply treatment process by using series of modelling called Artificial Neural Network (ANN)an Artificial Intelligence (AI) application tools in MATLAB-a software to develop patterns of changes in water footprint in the future with regards to some factors that are to be considered. Prediction values were found to have increased in certain WTPs, especially in Panching Water Treatment Plant as Panching Water Treatment Plant recorded an increasing data in each month throughout the year due to increasing wateruse activities, especially urbanization process that use Panching as the source of supply. Since Panching Water Treatment Plant capacity is broad, there are also much planning in taking this WTP to supply for new area. Therefore, it is important to consider new intake rather than clinging to the Kuantan river basin.

Recommendation
Generally, these findings provide a notable implications for better understanding towards maintaining sustainable resources to the future generations of this country. Benefits of this study shall be extended to the next research findings, where some improvements can and should be taken up in order to produce better and accurate results. For the WFb accounting, the development of technologies in this era should be fully implemented, especially when it comes to the centralization of data among local authorities. A manual or way of collecting data nowadays is partially relevant to the current technologies that exist. Therefore, a better management would be the best way to cater the problem while this study is being conducted.