Although the expansion of the fluidised bed is a key parameter for defining the optimal working condition of a pressurised porous media filter for microirrigation operating in backwashing mode, there are no ready to use tools available at the user level to calculate it. Complex machine learning models have previously been applied to predict the height of the expanded bed, but simpler methods are required to promote the practical use of bed expansion calculations. Therefore, the capabilities of simpler and easier to both implement and generalise multiple linear regression (MLR) models to forecast both the height of the fluidised bed and the backwashing filter pressure drop were investigated, with input variables being the porous medium type, the initial packed bed height, the underdrain design, and the superficial velocity. In addition, semi-analytical equations and a simplified computational fluid dynamics (CFD) model for calculating the expanded bed height as a function of porous medium properties were tested. The exponential model provided the best fit ( R^2 = 0.971) for the pressure drop, while the power model the best fit ( R^2 = 0.954) for the bed expansion. A simple semi-analytical expression reproduced the observed trend, but less efficiently than the MLR models. In terms of energy consumption, the results highlighted the relevance of improving the underdrain design, as well as avoiding operation at the point of maximum hydrodynamic shear stress, despite being the optimum condition for particle removal. Bed expansion values of 30–40

The location and evolution of the pressure losses across sand media filters commonly used in microirrigation have not been extensively studied. However, understanding these patterns may allow identifying feasible redesign strategies that reduce energy consumption and, therefore, enhance the sustainability of this irrigation equipment. The pressure loss in the main regions (diffuser, upper and lower bed media, underdrain and collector) of three media filters with different underdrain types (collector arms, inserted domes and porous media) was measured in filtration operation with two sand bed heights (0.2 and 0.3 m) and two filtration velocities (30 and 60 m/h). Each combination of filter, bed height and filtration velocity was tested for 250 h using reclaimed effluents. Higher pressure losses were observed at the greatest filtration velocity and the deepest media bed. Overall, the highest pressure loss was observed in the sand bed, with the diffuser showing the second largest loss. The evolution of pressure loss throughout each filtration cycle showed a progressive increase across the media, attributed to the accumulation of retained solids. New designs of diffusers are advisable to reduce pressure loss in both the inserted domes and porous underdrain filters, while for the arm collector filters a redesign of the underdrain should be prioritized.

Suspended sediment load (Qs) is an important parameter in the analysis of hydrological processes and management of water resources. Direct methods of measuring Qs are costly and require precise instruments, which makes their application limited, especially in remote regions. Indirect methods, on the other hand, discover the relationships between river hydrological parameters and Qs. Machine learning-based models are among the empirical data mining approaches that have been employed for the prediction of Qs under various conditions. Ensemble models, e.g., XGBoost (Python 3.12.3 with XGBoost version 3.1.0), are among the widely used machine learning approaches in the hydrologic context. A challenging step in establishing such models is conducting suitable hyperparameter tuning. A modeling study is reported here that combines the metaheuristic red fox algorithm (RFO) with XGBoost to improve Qs prediction. Daily observations of 21 years from Illinois State, USA (12 rivers), were used to assess the proposed methodology. Hydrologic data, including water stage, temperature, sediment concentration and river water flowrate were used as input variables when defining two input configurations. The obtained results reveal that the proposed RFO-XGBoost model outperformed the standalone XGBoost model in all the studied sites for both input configurations. However, the performance improvement percentage fluctuated among the sites. It was found that the model improvement was primarily affected by river hydrologic characteristics. A SHAP analysis revealed river flowrate as the most empirically influential input parameter in the model's predictions of Qs. Uncertainty analysis through the Monte Carlo simulations further confirmed the proposed model's enhanced performance and robustness.

Accurate modeling of filter head losses is essential for the optimal design and management of microirrigation systems, particularly when treating reclaimed water. In this study, Gene Expression Programming (GEP) was applied—for the first time in the literature—to develop predictive models for head losses in disc, screen, and sand filters operating with effluents. A k-fold cross-validation procedure was adopted to rigorously evaluate model performance under both direct learning (DL) and transfer learning (TL) scenarios. GEP models achieved Scatter Index (SI) values ranging from 0.049 to 0.071 for direct learning, consistently outperforming existing empirical equations when a fair comparison using independent test sets was conducted. Under transfer learning, screen filter-derived models demonstrated superior transferability to both sand and disc filters (SI 0.058–0.101), while combining disc and sand filter data improved predictions for screen filters (SI 0.064–0.069). These results provide a novel, data-driven framework for filter head loss estimation that can support microirrigation system design when filter-specific training data are unavailable.

Rice is cultivated on approximately 1,000,000 ha in the Mediterranean area, with production concentrated in Egypt, Italy, Türkiye, Spain, Grece and Portugal. In these areas, rice is traditionally established by wet seeding and cultivated under continuous flooding (WFL), which requires larger volumes of water compared to other irrigation practices. The aim of this study is to benchmark irrigation methods alternative to WFL across sites representative of the rice agro-ecosystems producing areas of 5 of the main rice-producing countries. For each site, WFL and one or more alternative methods, selected and adapted to site-specific conditions, were implemented and monitored for at least two years. The alternative methods included: alternate wetting and drying (AWD), dry seeding and delayed flooding (DFL), water input/output reduction (WIR), hybrid irrigation (HYBRID), sprinkler irrigation (SPRINKLER), surface drip irrigation (DRIP), and subsurface drip irrigation (SDI). The results suggest that AWD, DFL and WIR, which are easy-to-implement flooding techniques, increase water productivity (WP) and preserve yield production. Both SPRINKLER and HYBRID showed a higher increase in WP (by about 50

Rice is the world's most important food crop, as it is a staple food for more than half of the world's population, and the global demand for rice is expected to increase. More than 1,000,000 hectares in the Mediterranean basin are devoted to rice cultivation. The most important producing countries are Italy (IT) and Spain (SP) in Europe (over 310,000 ha), and Egypt (EG) and Turkey (TR) among non-EU countries (over 600,000 ha). In the Mediterranean region, rice production is of great socio-economic and environmental importance, as rice is often a crucial product for internal consumption and export, especially in Egypt, where it is considered strategic for food security. Despite of this, the peculiar flooding conditions in which rice is traditionally grown lead to the use of huge water volumes, as well as to the potential release of greenhouse gases and pesticides into the environment. For this reason, the introduction of water-saving irrigation strategies could reduce water consumption and decrease the harmful environmental impacts associated with rice flooding, while maintaining yield and rice grain quality. In the context of the MEDWATERICE project (https://www.medwaterice.org/; PRIMA-2018), alternative irrigation methods to WFL were tested in case studies implemented in five Mediterranean countries (Italy, Spain, Portugal, Turkey, Egypt). Irrigation strategies for each CS were selected with the support of local Stake-Holder groups and applied in experimental fields measuring/estimating all the water balance terms on a daily basis. Wet seeding and alternate wetting and drying (AWD), dry seeding and delayed flooding (DFL), reduction of inlet/outlet discharges (WIR), a better control of ponding water level through automated gates (DFL-aut), hybrid irrigation (HYBRID), sprinkler irrigation (SPRINKLER), surface drip (DRIP) and subsurface drip irrigation (SDI) were implemented for at least two years in the period 2019-2021 alongside the traditional WFL, to investigate their environmental and economic sustainability and social acceptability.

The filtration capacity of media filters, which are widely used in drip irrigation systems to prevent emitter clogging, must be periodically restored by backwashing, which fluidizes the media bed and removes those trapped particles. Bed expansion (BE) and pressure drop (PD) are the key parameters for assessing the hydraulic performance of backwashing, but the available equations and models frequently fall short of their prediction. An experiment with three medium types, four filter underdrain designs, two bed heights and different backwashing superficial velocities as input variables was conducted to measure both BE and PD. A dataset of 705 backwashing runs was obtained and with 80 % of data for training and 20 % for testing, a machine learning-based model that uses Artificial Neural Networks (ANN) to predict both BE and PD was developed and compared with the Ridge, Elastic-net, and Lasso regression models. With coefficients of determination of 0.9932 and 0.9988 for BE and PD, respectively, the results demonstrated that the ANN model not only ranked the importance of the input variables and showed strong agreement with experimental data but also attained superior predictive accuracy regarding the Lasso, Elastic-net, and Ridge models. This study presents a novel and optimized approach for predicting bed expansion and pressure drop, enhancing the reliability of media filter backwashing performance assessments in smart irrigation systems.

Making effective decisions about scaling up on-farm irrigation practices to the district level requires a comprehensive assessment of irrigation management at the farm level. In this context, a bucket-type water mass balance model was developed, calibrated, and validated over five irrigation seasons on a 121-hectare rice farm located in the lower Ter River valley (north-east Spain), to assess the water use efficiency and the impact of different irrigation practices on water savings. The model was implemented considering the spatial variability of the soils within the farm. It showed a satisfactory performance in both the calibration (2020, 2021, 2022) and validation (2023, 2024) cropping seasons, with NSE values greater than 0.50, PBIAS lower than ±20%, and RSR lower than 0.70. After model validation, the simulation of alternative water management practices revealed that the 10-day fixed-turn irrigation reduced irrigation water use by 30% compared to the traditional water management, although it may negatively impact rice yield. Simulations of an early irrigation cut-off at the end of the season and dry seeding with delayed flooding accounted for 17% and 15% irrigation water savings, respectively. The implementation of the no-runoff practice only accounted for a 6% reduction in water use. The water-saving potential of the simulated strategies was mainly driven by shortening the flooded period of rice paddies, thus demonstrating that managing the ponding water level is critical to diminishing water use in rice irrigation.

Drip irrigation is a widely spreading technology, mainly due to its high water-use efficiency. This technique requires a filtration process that exhibits cyclical behavior where both filtration and backwashing modes repeat. In filtration, pressure increases with time due to the particle retention up to a preset value. In backwashing, the flow is reversed to clean the filter. Different design strategies to reduce energy and water consumption have been proposed, but their practical effects are not yet clear. Here, a global analysis method based on the classification of the time evolution of the pressure curve in filtration mode was developed. Energy and water use efficiency indices were defined and evaluated under different scenarios. More design options can be undertaken to reduce the consumption of energy than of water. The decrease in the pressure drop for clean filter conditions arose as the best option to increase energy efficiency (in a realistic scenario, a reduction of 20% in the pressure drop with tap water resulted in a reduction of 7.6% in the energy consumption per volume of filtered water). Precise backwashing times and flow rates were essential to improve water use efficiency (e.g., doubling the backwashing time led to a 4.5% decrease in water use efficiency).

Sand media filters are especially recommended to prevent emitter clogging with loaded irrigation waters, but their performances rely on backwashing. Despite backwashing being a basic procedure needed to restore the initial filtration capacity, there is a lack of information about the solid removal efficiency along the media bed depth. An experimental filter with a 200 mm silica sand bed height was used to assess the effect of two operation velocities (30/45 and 60/75 (filtration/backwashing) m h−1) and two clogging particles (inorganic sand dust and organic from a reclaimed effluent) on the efficiency of backwashing for removing the total suspended solids retained in different media bed slices. The average solid removal backwashing efficiency was greater with organic particles (78%) than with inorganic ones (64%), reaching its maximum at a 5–15 mm bed depth. A higher operation velocity increased the solid removal efficiency by 16%, using organic particles, but no significant differences were observed with inorganic particles. The removal efficiencies across the media bed were more uniform with organic particles (63–89%) than with inorganic (40–85%), which makes it not advisable to reduce the media height when reclaimed effluents are used. This study may contribute to future improvements in sand media filter design and management.

Pressurised sand filters used in drip irrigation need periodic backwashing to flush the contaminant particles out of the porous media. This process consumes high amounts of energy and water. The selection of more efficient backwashing operational conditions requires accurate information of the pressure drop and the bed expansion, the latter being not measured in commercial filters. An experimental study with a scaled filter that used a window to observe the bed expansion was conducted with three porous media types (glass microspheres and two silica sands), two packed media bed heights (200 mm and 300 mm) and four nozzles (one commercial and three prototypes). The 24 combinations of the filter experimental configuration were investigated for different superficial velocities. Both data and video recordings for all the 705 tests conducted were carefully analysed to obtain mean values and standard deviations of the height of the expanded bed. The behaviour of the fluidised bed dynamics was characterised. Results indicated that the nozzle design had a strong influence on the pressure drop, and, in consequence, on the power required for backwashing. It also had an observable impact on the fluidised bed dynamics although its effect on determining the overall height of the expanded bed was limited, this being more dependent on the type of the porous media. The most effective combination in terms of energy efficiency and porosity of the expanded bed was obtained with microspheres, though its retention efficiency might be questionable from the literature review, and the frustoconical nozzle geometry.

In micro-irrigation systems, distinct media filters and filtering materials are employed to remove suspended solids from irrigation water and thereby avoid emitter obstruction. Turbidity is related to suspended solids and dissolved oxygen depends on organic matter load. At this time, no models exist that are trustworthy enough to forecast the dissolved oxygen and turbidity at the outlet when utilising various media configurations and filter types. The objective of this investigation was to construct a model that can identify turbidity and dissolved oxygen at the filter outlet in advance. This study presents an algorithm for meta-heuristic optimisation inspired by populations termed Differential Evolution (DE) in conjunction with Support Vector Regression (SVR) (DE/SVR-relied model). This is an effective machine learning method, with seven kernel types for calculating the output turbidity (Turbo) and the output dissolved oxygen (DOo) from a dataset comprising 1,016 samples of various reclaimed water-using filter types. The type of media and filter, the height of the filter bed, the cycle duration, and the filtration velocity, as well as the electrical conductivity at the filter inlet, pH, inlet dissolved oxygen, water temperature, and the input turbidity are all tracked and analysed in order to achieve this. The best-fitted DE/SVR-relied model was constructed to predict the Turbo and DOo as well as the input variables' relative importance. Determination coefficients for the best-fitted DE/SVR-relied model for the testing dataset were 0.89 and 0.92 for outlet turbidity (Turbo) and outlet dissolved oxygen (DOo), respectively, showing a good predictive performance which are of great importance for the management of drip irrigation systems.

Rapid deep bed filtration is a common process of drip irrigation systems to prevent emitter clogging. The particle retention in the porous media increases the pressure difference between the filter's inlet and outlet. Commercial operational instructions preset a threshold value of this pressure difference to define the end of a filtration cycle. Accurate particle retention models may contribute in the determination of this setpoint to improve the energy efficiency of a filtration cycle. A two-step method was developed to calibrate a phenomenological particle retention model with transient data of the filter pressure drop, and retained mass in different media slices at the end of the filtration cycle. The first step used the vertical profile of accumulated mass in the media to fit the input parameters of the specific deposit rate equation. The second step applied the pressure data to find the input parameters of the pressure drop equation. The use of constant values of the input parameters provided reasonable results for all tests. When the end of the filtration cycle was set by applying a threshold pressure drop, both particle mass accumulation and removal efficiency decreased as the flow rate increased. By using the threshold pressure drop criterion, the energy consumption in a filtration cycle also decreased as the flow rate increased, but the energy consumption per unit of filtered water volume increased. For high setpoint pressure drop values, a minimum of the energy consumption per unit of filtered water volume as a function of the flow rate was found.

Different media filters and filtration media are used to eliminate suspended particles in microirrigation and therefore prevent clogging in the emitter. The water volume by filtration cycle is a parameter related to the filter and media capacity to retain particles, while turbidity is a variable related to particles in suspension in the water. Since turbidity can be measured easily and quickly, it is commonly mentioned in recommendations for the reuse of effluents in microirrigation. There are currently no models that are reliable enough to predict the filtered volume in each irrigation filter and outlet turbidity when using different filter types and media configurations. The object of this work was to propose a model that can detect early the filtered volume and the turbidity at the outlet values. This investigation presents an effective machine learning method, the Random Forest regression (RFR) in combination with the population-inspired metaheuristic optimization algorithm, called Differential Evolution (DE), for estimating the output turbidity and the filtered volume from a dataset with 1,016 samples of distinct media filters that use reclaimed effluent. The same experimental dataset was also fitted with Elastic-net, Lasso and Ridge regression machine learning methods also in combination with DE optimizer for comparison. This optimization performs the parameter tuning in the RFR using the training dataset, which considerably improves the accuracy of the regression. To achieve this, the most relevant operation input variables are tracked and analyzed: the kind of medium and filter, filtration velocity (v), height of the filter bed (H), cycle duration and the electrical conductivity for the filter inlet (ECi), pHi, dissolved oxygen (DOi), water temperature (Ti) and the input turbidity (Turbi). There are two kinds of results. Firstly, the importance ranking of the input variables on the outlet turbidity and filtered volume is presented using the DE/RFR model. Secondly, an innovative model for the prediction of the outlet turbidity and filtered volume was built and a regression with optimized parameters was done and coefficients of determination of 0.9331 and 0.8712 for filtered volume and outlet turbidity were obtained with this DE/RFR–based model, respectively. Additionally, the outcomes from the Elastic-net, Lasso and Ridge models are worse than DE/RFR–relied model estimations. The DE/RFR-based model's strong performance was confirmed by the agreement between experimental data and the latter results.

Accurate estimation of soil water content (SWC) is essential for effective agriculture and water resources management. While various methods have been developed for in-situ SWC measurement, practical limitations and the need for comprehensive water sensor networks make their use complicated. To overcome these challenges, heuristic data-driven models may provide a suitable alternative to practical methods for SWC simulation under different cultivation conditions. In this paper, the application of gene expression programming (GEP) methodology was proposed to simulate SWC at three different depths in rice fields using information related to weather and groundwater. A modeling study was conducted that applied the robust k-fold testing data assignment method, considering two different chronologic strategies of "k" defining to evaluate both strategies. The first one was based on the definition of the "k" values based on yearly data partitioning, while the second one considered growing stages as the "k" definition criterion. Besides evaluating the models using error statistics, a further uncertainty analysis was also conducted to check stability and confidence. The obtained results revealed that selection of "k" based on growing stages produced more accurate and stable results. Among the target parameters, water content at the third layer was predicted with higher accuracy.

Agricultural irrigation systems help provide food to meet the growing demands of the global population [...]

Rapid water filtration with pressurised porous media filters is extensively applied in drip irrigation systems. In double-chamber filters, the underdrains are fixed to the base of the inner plate to sustain the media above while draining water. Here, a new underdrain design intended to reduce the filter energy consumption is presented. The main difference with commercial underdrain units corresponds to the distribution of the slots, being in a horizontal plate to uniformise the flow trajectories inside the porous media. Both commercial and new underdrain designs have been tested in laboratory in both filtration and backwashing modes with three media types, two media heights, and superficial velocities ranging from 20 to 120 m h−1. In filtration mode, results indicate that the new design reduces the filter pressure drop by 31

Pressurized sand media filters are commonly used in drip irrigation systems to prevent emitter clogging. However, the performance of these filters may be improved with more information about the retention of solids at different bed depths under different filter operation conditions and irrigation water sources. In this study, experiments in a scaled sand media filter were conducted to clog the filter with two different filtration velocities (30 and 60 m h−1) and two-particle types (inorganic from A4 coarse sand dust and organic from a reclaimed effluent). The suspended solids retained in slices of 5 mm (in the first 20 mm of the bed) and 20 mm (from 20 to 200 mm depth) thick were determined following the van Staden and Haarhoff (2011) procedure. The solids retained in each slice per mass of media were significantly (p < 0.05) affected by the interaction between the filtration velocity, the bed depth, and the particle type. The solids retained in the first 5 mm of the bed were significantly higher than at other depths. Moreover, inorganic solids were retained more in upper slices than organic ones. Therefore, media depths may be adjusted depending on the irrigation water source to optimize media use.

Accurate model predictions are fundamental when designing porous media filters in drip irrigation systems that reduce both energy and water consumption. Many studies have focused on improving filter hydraulics under clean water conditions but further advances may require consideration of particle retention by the granular media. Rapid deep bed filtration models employ conservative equations and empirical correlations to determine the behaviour of particle depositions on the media. These models involve many input parameters, some of which have an inherent uncertainty range. Therefore, thorough model sensitivity analyses must be carried out prior to their use as predictors for the assessment of new filter designs. This paper applies both local and global (variance-based Sobol indices) sensitivity methods to a comprehensive particle retention model that is able to describe the main three stages of the filtration process. Uncertainty ranges of 15 input variables were defined. Three model outputs were analysed (flow particle concentration at the filter's outlet, mass of retained particles per unit area, and total pressure drop through the porous media) at different flow times. The results of the global sensitivity analysis indicated that the relevant model parameters vary depending on the filter stage. The rank of influential input variables also varied depending on the chosen output variable. The least absolute shrinkage and selection operator (LASSO) regression analysis method was also applied but the high non-linearity of the model reduced its predictive capacity in most of the situations analysed. Conclusions from the global sensitivity analysis were employed for model calibration with experimental data.

Abstract In the Mediterranean basin, rice is cultivated in approximately 1,000,000 hectares. The most important rice-producing countries in the region are Egypt, Italy, Türkiye and Spain. In all these areas, rice is traditionally cultivated under continuous flooding, requiring larger irrigation water volumes compared to non-ponded crops. In the framework of the MEDWATERICE project (https://www.medwaterice.org/), innovative irrigation methods to reduce irrigation water use and other rice environmental impacts were experimented and benchmarked to the traditional wet-seeding and continuous flooding (WFL) in seven case studies (CSs) representative of different rice agroecosystems in five Mediterranean countries. The most promising irrigation options for each CS were selected with the support of local Stake-Holder Panels, and tailored to site-specific conditions. Alternate wetting and drying (AWD), dry-seeding and delayed flooding (DFL), reduction of irrigation discharge input/output (WIR), hybrid irrigation (HYBRID), sprinkler irrigation (SPRINKLER), surface drip irrigation (DRIP) and subsurface drip irrigation (SDI) were investigated. Results suggest that AWD, DFL and WIR, which are flooding techniques rather simple to implement, might be sound alternatives to WFL leading to an increase in Water Productivity (WP) and safeguarding the yield production. Both SPRINKLER and HYBRID resulted in an increase in WP of about 50% while maintaining or increasing the yield. DRIP and SDI showed a great potential in reducing water use, increasing WP up to 260%; however, yield may sometimes be notably reduced. Nevertheless, for each technique and in particular for localized irrigation methods, site-specific conditions must be carefully evaluated to properly select, design and manage irrigation strategies.