Realizaram-se dois estudos, e primeiro deles objetivou verificar a influência do comportamento próximo ao comedouro e bebedouro sobre o consumo de água (CA) e avaliar a adequação das equações de predição para estimar o CA em bovinos Brangus no subtrópico. Os dados foram coletados em dois experimentos realizados no Rio Grande do Sul, na EEA-UFRGS, o primeiro em 2017, com 60 novilhas da raça Brangus e o segundo em 2018 com 30 bezerros da mesma raça. Os valores preditos foram calculados utilizando 6 equações previamente publicadas na literatura científica e comparados com os valores de CA medidos nos bebedouros automáticos. As análises de regressão linear entre CA medido e os valores preditos mostraram que todas as equações superestimaram o CA medido, devido a diferenças ambientais e genotípicas e fatores não considerados nas equações. Com as informações geradas nos experimentos, foram propostas equações de predição: CA= - 2,44 + (0,009 x PC) + (0,84 x CMS) – (0,10 x UR) + (0,64 x TMAX) e CA= - 2,52 + (0,96 x CMS) – (0,09 x UR) + (0,45 x TMAX) + (0,76 x NVCB) + (0,18 x TCB) - (0,02 x NVCC) + (1,81 x TI) para novilhas; e CA1= - 4,23 + (0,98 x CMS) + (0,50 x TMAX) - (0,98 x PP) e CA2 = 13,07 + (0,61 x CMS) - (0,14 x UR) + (0,34 x TMAX) – (0,91 x VV) - (0,09 x RS) + (0,99 x NVCB) para bezerros. As equações propostas para os bezerros foram validadas com dados coletados em outro período, com novilhos Brangus. Entre os dois modelos propostos no estudo, o modelo comportamental (CA2) apresentou maior coeficiente de determinação, com média estimada de CA de 9,49 kg para um CA medido de 19,55 kg/d. O segundo estudo objetivou avaliar a influência do temperamento sobre o consumo de água e alimentos de bezerros de corte confinados e analisar o efeito do temperamento sobre as características produtivas dos animais. Os dados foram coletados de 30 bezerros da raça Brangus no ano de 2018 em três avaliações. O temperamento foi avaliado como o escore composto de balança (escores de 1 a 5, de calmos a muito reativos) e, posteriormente, os animais foram categorizados em calmos, intermediários e reativos. O temperamento não influenciou o consumo de água e alimentar dos animais. Animais calmos ganharam menos peso no primeiro período de avaliação que os demais. Animais calmos permaneceram menos tempo no cocho que os demais. O presente estudo confirmou o baixo coeficiente de determinação dos modelos de predição de consumo de água e a superestimação do consumo. O consumo de água e alimentos e a maior parte dos atributos comportamentais ligados à ingestão não foram influenciado pelo temperamento dos bovinos. | Two studies were carried out, the first one aimed to verify the influence of the behavior close to the feeder and drinker on water intake (WI) and to evaluate the adequacy of the prediction equations to estimate the WI in Brangus cattle in the subtropical region. Data were collected in two experiments carried out in Rio Grande do Sul, at EEA- UFRGS. The first was in 2017 studying 60 Brangus heifers and the second was in 2018 studying 30 calves of the same breed. The predicted values were calculated using six previously published equations and compared with the WI values measured in the automatic drinkers. Linear regression analyzes between WI and predicted values demonstrated that all equations overestimated WI, due to both environmental and genotypic differences as well as to factors not considered in the equations. From the information generated in the experiments, new prediction equations were proposed: WI = - 2,44 + (0,009 x BW) + (0,84 x DMI) – (0,10 x HU) + (0,64 x MT) e WI = - 2,52 + (0,96 x DMI) – (0,09 x HU) + (0,45 x MT) + (0,76 x NVCB) + (0,18 x TCB) - (0,02 x NVCC) + (1,81 x TI) for heifers; and WI1 = - 4,23 + (0,98 x DMI) + (0,50 x MT) - (0,98 x PP) e WI2 = 13,07 + (0,61 x DMI) - (0,14 x HU) + (0,34 x MT) – (0,91 x WS) - (0,09 x SR) + (0,99 x NVCB) for calves. The proposed equations for the calves were validated with data collected in other period with Brangus steers. Among the two models proposed in the study, the behavioral model (WI2) presented the highest coefficient of determination, with an estimated mean of 9.49 kg of WI for an observed overall WI of 19.55 kg/d. The second article aimed to evaluate the influence of temperament on the consumption of water and feed of beef calves and to analyze the effect of temperament on the productive characteristics of animals. Data were collected from 30 Brangus calves in 2018 in three evaluations. The temperament was evaluated as the balance composed score (scores from 1 to 5 from calm to very reactive) and, later, the animals were categorized into calm, intermediate and reactive. The temperament did not influence the water and food consumption of the animals. Calm animals gained less weight in the first evaluation period than the rest of them. Calm animals spent less time in the feeder than the others. The present study confirmed the low coefficient of determination of the water intake prediction models and the overestimation of water consumption. The intake of water and food and most of the behavioral attributes linked to ingestion were not influenced by temperament in beef cattle raised in the subtropics.

The use of artificial intelligence has stood out as a powerful tool in predicting outcomes through machine learning, especially when dealing with large volumes of data. The integration of artificial intelligence techniques, such as neural networks, with traditional statistical methods, like principal component analysis (PCA) and clustering algorithms, such as k-means, in developing predictive models to understand hydrological processes in plant substrates, proves to be a promising approach to comprehend the relationship between volumetric moisture and different matrix potentials. By training these models with comprehensive and representative datasets, capturing complex patterns in the data and making more accurate predictions about water behavior is expected. Additionally, the combination of neural networks with clustering algorithms, such as k-means, allows for identifying patterns in the data that may not be easily perceptible to the naked eye, which is useful for grouping moisture data, enabling a detailed analysis of variations in water distribution. Principal component analysis (PCA) complements this process by aiding in reducing data dimensionality and identifying the main variables that influence water retention, facilitating result interpretation and identifying important relationships between variables. In the context of agriculture, these techniques can have broad applications, from efficient irrigation and drainage management to crop planning and yield prediction. Thus, the main objective of this work is to present methodological approaches with advanced artificial intelligence techniques to accurately predict the water retention curve of plant substrates, aiming to contribute to the advancement of precision agriculture and the development of more sustainable and efficient agricultural practices. The analysis of the prediction results conducted on the five clusters (K1 to K5) revealed valuable information about the relationship between matrix potentials and the moisture of formulations. Neural networks demonstrated an impressive ability to model and predict moisture under different conditions, as represented by the various matrix potentials. The coefficients of determination (R²) obtained for the training, testing, and validation data reflect the model's effectiveness in explaining variability in the data and providing accurate predictions. Identifying consistent patterns between observed and predicted values, even with small databases (LI et al., 2016), highlights the robustness and generalization capability of neural networks in generating the water retention curve in plant substrates. These results suggest that neural networks are a powerful and versatile tool for understanding and modeling moisture in different waste materials, providing important information for producers, aiding in water management, and environmental conservation. This study convincingly emphasizes the vital role of neural networks in advancing sciences related to irrigation, drainage, and agricultural practices by offering a deeper and more accurate understanding of the processes involved in substrate formulation, as well as the neural networks that promote a promising approach. Therefore, with this innovative tool, we have the ability to significantly improve our agricultural practices, driving efficiency, productivity, and sustainability. | O uso da inteligência artificial tem se destacado como uma ferramenta poderosa na previsão de resultados através da aprendizagem de máquina, especialmente quando lidamos com grandes volumes de dados. A integração de técnicas de inteligência artificial, como redes neurais, com métodos estatísticos tradicionais, como análise de componentes principais (ACP) e algoritmos de agrupamento, como k-means, no desenvolvimento de modelos preditivos para compreender os processos hidrológicos em substratos para plantas, mostra-se uma abordagem promissora para entender a relação da umidade volumétrica em diferentes potenciais matriciais. Ao treinar esses modelos com conjuntos de dados abrangentes e representativos, espera-se capturar padrões complexos nos dados e fazer previsões mais precisas sobre o comportamento da água. Além disso, a combinação de redes neurais com algoritmos de agrupamento, como k-means, permite identificar padrões nos dados que podem não ser facilmente perceptíveis a olho nu, o que é útil para agrupar dados de umidade, permitindo uma análise detalhada das variações na distribuição da água. A análise de componentes principais (ACP) complementa esse processo ao ajudar na redução da dimensionalidade dos dados e na identificação das principais variáveis que influenciam na retenção de água, facilitando a interpretação dos resultados e a identificação de relações importantes entre variáveis. No contexto da agricultura, essas técnicas podem ter amplas aplicações, desde o manejo eficiente da irrigação e drenagem até o planejamento de cultivos e a previsão de safras. Assim, o objetivo principal deste trabalho é apresentar abordagens metodológicas com técnicas avançadas de inteligência artificial para prever com precisão a curva de retenção de água de substratos para plantas, visando contribuir para o avanço da agricultura de precisão e o desenvolvimento de práticas agrícolas mais sustentáveis e eficientes. A análise dos resultados da predição realizada nos cinco clusters (K1 a K5) revelou informações valiosas sobre a relação entre os potenciais matriciais e a umidade das formulações. As redes neurais mostraram uma notável capacidade de modelar e prever a umidade em diferentes condições, representadas pelos diversos potenciais matriciais. Os coeficientes de determinação (R²) obtidos para os dados de treinamento, teste e validação refletem a eficácia do modelo em explicar a variabilidade nos dados e fornecer previsões precisas. A identificação de padrões consistentes entre os valores observados e previstos, mesmo com bases de dados pequenas (LI et al., 2016), destaca a robustez e a capacidade de generalização das redes neurais na geração da curva de retenção de água em substratos para plantas. Esses resultados sugerem que as redes neurais são uma ferramenta poderosa e versátil para entender e modelar a umidade em diferentes materiais oriundos de descarte, fornecendo informações importantes para os produtores, auxiliando na gestão hídrica e na conservação do meio ambiente. Este estudo ressalta o papel vital das redes neurais no avanço das ciências relacionadas à irrigação, drenagem e práticas agrícolas, oferecendo uma compreensão mais profunda e precisa dos processos que envolvem a formulação de substratos, bem como das redes neurais que promovem uma abordagem promissora. Com essa ferramenta inovadora, temos a capacidade de melhorar significativamente nossas práticas agrícolas, impulsionando a eficiência, produtividade e sustentabilidade.

The application of statistical classification methods is investigated—in comparison also to spatial interpolation methods—for predicting the acceptability of well-water quality in a situation where an effective quantitative model of the hydrogeological system under consideration cannot be developed. In the example area in northern Italy, in particular, the aquifer is locally affected by saline water and the concentration of chloride is the main indicator of both saltwater occurrence and groundwater quality. The goal is to predict if the chloride concentration in a water well will exceed the allowable concentration so that the water is unfit for the intended use. A statistical classification algorithm achieved the best predictive performances and the results of the study show that statistical classification methods provide further tools for dealing with groundwater quality problems concerning hydrogeological systems that are too difficult to describe analytically or to simulate effectively.

Nonlinear complex behavior of pore-water pressure responses to rainfall was modelled using support vector regression (SVR). Pore-water pressure can rise to disturbing levels that may result in slope failure during or after rainfall. Traditionally, monitoring slope pore-water pressure responses to rainfall is tedious and expensive, in that the slope must be instrumented with necessary monitors. Data on rainfall and corresponding responses of pore-water pressure were collected from such a monitoring program at a slope site in Malaysia and used to develop SVR models to predict pore-water pressure fluctuations. Three models, based on their different input configurations, were developed. SVR optimum meta-parameters were obtained using k-fold cross validation and a grid search. Model type 3 was adjudged the best among the models and was used to predict three other points on the slope. For each point, lag intervals of 30 min, 1 h and 2 h were used to make the predictions. The SVR model predictions were compared with predictions made by an artificial neural network model; overall, the SVR model showed slightly better results. Uncertainty quantification analysis was also performed for further model assessment. The uncertainty components were found to be low and tolerable, with d-factor of 0.14 and 74 % of observed data falling within the 95 % confidence bound. The study demonstrated that the SVR model is effective in providing an accurate and quick means of obtaining pore-water pressure response, which may be vital in systems where response information is urgently needed.

Groundwater resources in semi-arid areas and especially in the Mediterranean face a growing demand for irrigated agriculture and, to a lesser extent, for domestic uses. Consequently, groundwater reserves are affected and water-table drops are widely observed. This leads to strong constraints on groundwater access for farmers, while managers worry about the future evolution of the water resources. A common problem for building proper groundwater management plans is the difficulty in assessing individual groundwater withdrawals at regional scale. Predicting future trends of these groundwater withdrawals is even more challenging. The basic question is how to assess the water budget variables and their evolution when they are deeply linked to human activities, themselves driven by countless factors (access to natural resources, public policies, market, etc.). This study provides some possible answers by focusing on the assessment of groundwater withdrawals for irrigated agriculture at three sites in North Africa (Morocco, Tunisia and Algeria). Efforts were made to understand the different features that influence irrigation practices, and an adaptive user-oriented methodology was used to monitor groundwater withdrawals. For each site, different key factors affecting the regional groundwater abstraction and its past evolution were identified by involving farmers’ knowledge. Factors such as farmer access to land and groundwater or development of public infrastructures (electrical distribution network) are crucial to decode the results of well inventories and assess the regional groundwater abstraction and its future trend. This leads one to look with caution at the number of wells cited in the literature, which could be oversimplified.

To date, there has been no agreement on the best way to simulate saltwater intrusion (SWI) in karst aquifers. An equivalent porous medium (EPM) is usually assumed without justification of its applicability. In this paper, SWI in a poorly karstified aquifer in Lebanon is simulated in various ways and compared to measurements. Time series analysis of rainfall and aquifer response is recommended to decide whether quickflow through conduits can be safely ignored. This aids in justifying the selection of the exemplified EPM model. To examine the improvement of SWI representation when discrete features (DFs) are embedded in the model domain, the results of a coupled discrete-continuum (CDC) approach (a hybrid EPM-DF approach) are compared to the EPM model. The two approaches yielded reasonable patterns of hydraulic head and groundwater salinity, which seem trustworthy enough for management purposes. The CDC model also reproduced some local anomalous chloride patterns, being more adaptable with respect to the measurements. It improved the overall accuracy of salinity predictions at wells and better represented the fresh–brackish water interface. Therefore, the CDC approach can be beneficial in modeling SWI in poorly karstified aquifers, and should be compared with the results of the EPM method to decide whether the differences in the outcome at local scale warrant its (more complicated) application. The simulation utilized the SEAWAT code since it is density dependent and public domain, and it enjoys widespread application. Including DFs necessitated manual handling because the selected code has no built-in option for such features.

The modeling of changes in surface water and groundwater in the areas of inter-basin water diversion projects is quite difficult because surface water and groundwater models are run separately most of the time and the lack of sufficient data limits the application of complex surface-water/groundwater coupling models based on physical laws, especially for developing countries. In this study, a distributed surface-water and groundwater coupling model, named the distributed time variant gain model–groundwater model (DTVGM-GWM), was used to assess the influence of climate change and inter-basin water diversion on a watershed hydrological cycle. The DTVGM-GWM model can reflect the interaction processes of surface water and groundwater at basin scale. The model was applied to the Haihe River Basin (HRB) in eastern China. The possible influences of climate change and the South-to-North Water Diversion Project (SNWDP) on surface water and groundwater in the HRB were analyzed under various scenarios. The results showed that the newly constructed model DTVGM-GWM can reasonably simulate the surface and river runoff, and describe the spatiotemporal distribution characteristics of groundwater level, groundwater storage and phreatic recharge. The prediction results under different scenarios showed a decline in annual groundwater exploitation and also runoff in the HRB, while an increase of groundwater storage and groundwater level after the SNWDP’s operation. Additionally, as the project also addresses future scenarios, a slight increase is predicted in the actual evapotranspiration, soil water content and phreatic recharge. This study provides valuable insights for developing sustainable groundwater management options for the HRB.

Geothermal resources are practical and competitive clean-energy alternatives to fossil fuels, and study on the recharge sources of geothermal water supports its sustainable exploitation. In order to provide evidence on the recharge source of water and circulation dynamics of the Tangshan Geothermal System (TGS) near Nanjing (China), a comprehensive investigation was carried out using multiple chemical and isotopic tracers (δ²H, δ¹⁸O, δ³⁴S, ⁸⁷Sr/⁸⁶Sr, δ¹³C, ¹⁴C and ³H). The results confirm that a local (rather than regional) recharge source feeds the system from the exposed Cambrian and Ordovician carbonate rocks area on the upper part of Tangshan Mountain. The reservoir temperature up to 87 °C, obtained using empirical as well as theoretical chemical geothermometers, requires a groundwater circulation depth of around 2.5 km. The temperature of the geothermal water is lowered during upwelling as a consequence of mixing with shallow cold water up to a 63% dilution. The corrected ¹⁴C age shows that the geothermal water travels at a very slow pace (millennial scale) and has a low circulation rate, allowing sufficient time for the water to become heated in the system. This study has provided key information on the genesis of TGS and the results are instructive to the effective management of the geothermal resources. Further confirmation and even prediction associated with the sustainability of the system could be achieved through continuous monitoring and modeling of the responses of the karstic geothermal reservoir to hot-water mining.

Sustainable management of groundwater resources is vital for development of areas at risk from water-resource over-exploitation. In northeast Thailand, the Phu Thok aquifer is an important water source, particularly in the Thaphra area, where increased groundwater withdrawals may result in water-level decline and saline-water upconing. Three-dimensional finite-difference flow models were developed with MODFLOW to predict the impacts of future pumping on hydraulic heads. Four scenarios of pumping and recharge were defined to evaluate the system response to future usage and climate conditions. Primary model simulations show that groundwater heads will continue to decrease by 4–12 m by the year 2040 at the center of the highly exploited area, under conditions of both increasing pumping and drought. To quantify predictive uncertainty in these estimates, in addition to the primary conceptual model, three alternative conceptual models were used in the simulation of sustainable yields. These alternative models show that, for this case study, a reasonable degree of uncertainty in hydrostratigraphic interpretation is more impactful than uncertainty in recharge distribution or boundary conditions. The uncertainty-analysis results strongly support addressing conceptual-model uncertainty in the practice of groundwater-management modeling. Doing so will better assist decision makers in selecting and implementing robust sustainable strategies.

An accurate estimate of the groundwater inflow to a tunnel is one of the most challenging but essential tasks in tunnel design and construction. Most of the numerical or analytical solutions that have been developed ignore tunnel seepage conditions, material properties and hydraulic-head changes along the tunnel route during the excavation process, leading to inaccurate prediction of inflow rates. A method is introduced that uses MODFLOW code of GMS software to predict inflow rate as the tunnel boring machine (TBM) gradually advances. In this method, the tunnel boundary condition is conceptualized and defined using Drain package, which is simulated by dividing the drilling process into a series of successive intervals based on the tunnel excavation rates. In addition, the drain elevations are specified as the respective tunnel elevations, and the conductance parameters are assigned to intervals, depending on the TBM type and the tunnel seepage condition. The Qomroud water conveyance tunnel, located in Lorestan province of Iran, is 36 km in length. Since the Qomroud tunnel involved groundwater inrush during excavating, it is considered as a good case study to evaluate the presented method. The groundwater inflow to this tunnel during the TBM advance is simulated using the proposed method and the predicted rates are compared with observed rates. The results show that the presented method can satisfactorily predict the inflow rates as the TBM advances.