Document Type : Original Article

Authors

• Seyed Hasan Tabatabaii ¹
• Seyed Majid Mirlatifi ²
• Hosein Dehghanisanij ³
• Ashkan Shokri ⁴

¹ PhD Candidate in Water Engineering-Irrigation and Drainage , Department of Water Engineering and Management, Faculty of Agriculture, Tarbiat Modares University

² Dept. of Water Engineering and Management, Tarbiat Modares University, Tehran, Iran

³ Agricultural Engineering Research Institute, Karaj,

⁴ Scientific Researcher, Bureau of Meteorology, Canberra, Australia

Abstract

Extended Abstract
Abstract
The impact of the application of data assimilation (DA) on improving the accuracy of estimates of ET was investigated by applying two different methods of (DA) namely Ensemble Kalman Filter (EnKF) and Particle Filter (PF). The ET calculated using the SEBAL algorithm was used as the observations and the ET calculated according to the FAO 56 was considered as the model outputs. In order to ascertain the performance of these two aforementioned DA methods, the results obtained from three different approaches including a) application of DA using PF method, b) application of DA using EnKF method, and c) application of an open-loop simulation (OL) were compared with the results of another data assimilation system. In this alternative system, soil moisture contents were measured by a soil moisture sensing device (TDR) and the numerical solution of the Richards equation was utilized to calculate soil moisture content (BL). The results indicated that the PF and EnKF methods reduced the average bias error in the soil moisture estimations in the root zone layer by 7% and 9%, respectively, compared to the OL. Furthermore, the PF and EnKF methods were able to reduce the nRMSE of the soil moisture contents by 8% compared to the open-loop simulation (OL). These findings suggest that with the application of the DA, it is possible to improve the accuracy of the ET estimations and therefore, to enhance irrigation management.
Introduction
Data assimilation is a scientific method which integrates information from the actual measurements and the model predictions within a defined framework to enhance the accuracy of the estimations of variables or parameters under investigation. This process comprises of two phases called prediction and update. In the prediction phase, the model estimations are computed using the Monte Carlo simulation method. This process continues until observation data (measurements) becomes available. In the update phase, model estimations and observations are combined, taking into account the confidence level associated with each one of the data sources (observations, and model estimations), resulting in posteriori estimates (updated outcomes). This study focuses on improving the accuracy of the estimates of the soil moisture contents at root zone depth using the ET estimation model suggested for non-standard conditions by FAO 56 (Allen et al., 1998) through the utilization of two data assimilation methods namely the Ensemble Kalman Filter (EnKF) and the Particle Filter (PF). ET was calculated according to the Surface Energy Balance (SEBAL) algorithm using Landsat 8 satellite imagery as observation data for the data assimilation system.
Materials and Methods
In order to ascertain the effectiveness of the data assimilation methods applied, results obtained from a data assimilation system (referred to as BL) which was implemented in two sugar beet fields and two corn fields in the Jovein region were utilized. In the BL system, simulated soil moisture contents of the root zone layer obtained by numerically solving the Richards equation were combined with the soil moisture measurements taken at specific points in the fields using soil moisture sensors (TDR). 51 TDR access tubes were installed in the fields to measure soil moisture contents at various depths using Time Domain Reflectometry (TDR) sensors. Soil moisture measurements were recorded from Khordad to Aban 1399 (write in English calendar).
The essence of data assimilation methods lies in the amalgamation of homogeneous information about the studied phenomenon obtained through different mechanisms. In this study, the observations utilized included ET calculated based on the SEBAL algorithm. In the FAO-56 model, evapotranspiration and soil moisture content of the surface and root zone layers were computed. Since soil moisture contents of the surface and the root zone layers serve as the initial conditions for subsequent simulation steps, data assimilation was applied to the soil moisture content instead of the ET. To achieve this, ET obtained from the SEBAL algorithm was converted into soil moisture content and subsequently used in the data assimilation process.
Conclusion
The average standard deviation of the simulated soil moisture contents (σ) in the PF and EnKF approaches was 36% and 32% lower, respectively, compared to the open-loop (OL) approach. Throughout the growth period, PF and EnKF consistently resulted in lower σ compared to OL in the three fields which were irrigated by center-pivot irrigation systems. However, following each irrigation event in the field M, which was irrigated by a drip irrigation system, σ suddenly increased and became nearly equivalent to OL. This was attributed to the greater depth of irrigation water in this field as compared with the other three fields. On the average, the magnitude of σ change (σΔ), representing the reduction in σ before and after the update, was 0.032 and 0.037 for PF and EnKF, respectively. Consequently, the results suggested that data assimilation reduces the uncertainty of simulation results.
The research findings indicate that data assimilation significantly reduced the BIAS and nRMSE indices compared with the OL approach. The average BIAS for EnKF, PF, and OL was 0.018, 0.020, and 0.028, respectively, while the average nRMSE for the three methods was 17.3%, 17.5%, and 18.9%, respectively. In other words, the use of ET observations obtained from Landsat 8 satellite imagery and the SEBAL algorithm significantly improved the accuracy of the estimation of ET with the FAO-56 model. Therefore, the obtained results suggest that data assimilation can be employed to enhance the accuracy water consumption estimates and to improve irrigation management.

Keywords

• Ensemble Kalman Filter
• Particle Filter
• SEBAL
• Soil-Water-Plant Modeling

Main Subjects

• Water Resources and Consumption