Data assimilation application in prediction of flowrate for a sustainable groundwater resource: Falaj Al-Khatmain, Oman

As an arid country, Oman's falaj flowrate directly affects the environmental, agricultural, and ecological characteristics of its regions. For this reason, it is vitally important for water decision managers, farmers, and governments to have an accurate prediction of falaj’s flowrate. The present research aims to develop a new method for flowrate of falaj modeling named particle filter-Bayesian network (PF-BN) based on data assimilation method and Bayesian Network. A falaj in Birkat Al-Mouz, Nizwa, Oman, named Al-Khatmain is selected as a case study. This falaj, which is listed as a UNESCO heritage, plays an important role in agriculture (mainly palms, bananas and mangoes). A weekly flow dataset was collected by the authors from 2021 to 2022. Results of PF-BN are compared with the conventional BN model. The models are all written in MATLAB. As a result, four input patterns were introduced to the models. In these patterns, the flow rate of previous weeks or months is taken into account. A total of four patterns including: M 1 , M 2 , M 3 and M 4 , are considered. According to results, data assimilation leads to more accurate responses, as three error criteria are calculated. There are three criteria here: RMSE, NRMSE, and NSE. In terms of efficiency indices, M 4 (PF-BN) is 0.87, 0.011, 0.93 and M 4 (BN) is 2.4, 0.0315, 0.89. According to this error values M 4 (PF-BN) performance is satisfactory.