Water is the main limiting factor in agricultural production as well as a scarce resource that needs to be optimized. The measurement of soil water with sensors is an efficient way for optimal irrigation management. However, commercial sensors are still too expensive for most farmers. This paper presents the design, development and calibration of a new capacitive low-cost soil moisture sensor that incorporates SDI-12 communication, allowing one to select the calibration equation for different soils. The sensor was calibrated in three different soils and its variability and accuracy were evaluated. Lower but cost-compensated accuracy was observed in comparing it with commercial sensors. Field tests have demonstrated the temperature influence on the sensor and its capability to efficiently detect irrigation and rainfall events.

The number of sensors, ground-based and remote, exploiting the relationship between soil dielectric response and soil water content continues to grow. Empirical expressions for this relationship generally work well in coarse-textured soils but can break down for high-surface area and intricate materials such as clayey soils. Dielectric mixing models are helpful for exploring mechanisms and developing new understanding of the dielectric response in porous media that do not conform to a simple empirical approach, such as clayey soils. Here, we explore the dielectric response of clay minerals and clayey soils using the mixing model approach in the frequency domain. Our modeling focuses on the use of mixing models to explore geometrical effects. New spectroscopic data are presented for clay minerals (talc, kaolinite, illite and montmorillonite) and soils dominated by these clay minerals in the 1 MHz–6 GHz bandwidth. We also present a new typology for the way water is held in soils that we hope will act as a framework for furthering discussion on sensor design. We found that the frequency-domain response can be mostly accounted for by adjusting model structural parameters, which needs to be conducted to describe the Maxwell–Wagner (MW) relaxation effects. The work supports the importance of accounting for soil structural properties to understand and predict soil dielectric response and ultimately to find models that can describe the dielectric–water content relationship in fine-textured soils measured with sensors.

A low-cost handheld e-nose unit for assessing composition and quality of food in the field or for automatic discrimination of different foodstuffs would have significant commercial advantages and provide a positive impact on consumer safety and confidence. Whilst many studies have focused on the freshness of specific foodstuffs, the preliminary results presented in this report indicate that a low-cost (circa 30 euro) e-nose, based on MQ series gas sensors, using sinusoidally varying heater voltages, can successfully differentiate between samples of unspoiled lamb, hake, salmon, beef, pork and chicken, even using default, non-optimized control parameters, which could be applied in food adulteration and fraudulent labelling detection or in smart appliances to automatically select cooking parameters. Samples were exposed to the e-nose for five cycles of two minutes each and, by subjecting the response waveforms to Discrete Fourier Transform (DFT) analysis, harmonic component amplitudes were extracted. These data items were then fed into two different Machine Learning supervised classification methods (discriminant analysis and random forest) to determine which foodstuff was present. The relevance of the different DFT harmonic components and MQ sensors in the foodstuffs classification was quantified. Classification accuracies in excess of 95% were demonstrated.

The crop water stress index (CWSI) is one of the parameters measured in deficit irrigation and it is obtained from crop canopy temperature. However, image segmentation is required for non-leaf region exclusion in temperature measurement, as it is critical to obtain the temperature values for the calculation of the CWSI. To this end, two image-segmentation models based on support vector machine (SVM) and deep learning have been studied in this article. The models have been trained with different parameters (encoder depth, optimizer, learning rate, weight decay, validation frequency and validation patience), and several indicators (accuracy, precision, recall and F₁ score/dice coefficient), as well as prediction, training and data preparation times are discussed. The results of the F₁ score indicator are 83.11% for SVM and 86.27% for deep-learning models. More accurate results are expected for the deep-learning model by increasing the dataset, whereas the SVM model is worthwhile in terms of reduced data preparation times.

Water is a limited resource in arid and semi-arid regions, as is the case in the Mediterranean Basin, where demographic and climatic conditions make it ideal for growing fruits and vegetables, but a greater volume of water is required. Deficit irrigation strategies have proven to be successful in optimizing available water without pernicious impact on yield and harvest quality, but it is essential to control the water stress of the crop. The direct measurement of crop water status is currently performed using midday stem water potential, which is costly in terms of time and labor; therefore, indirect methods are needed for automatic monitoring of crop water stress. In this study, we present a novel approach to indirectly estimate the water stress of 15-year-old mature sweet cherry trees from a time series of soil water status and meteorological variables by using Machine Learning methods (Random Forest and Support Vector Machine). Time information was accounted for by integrating soil and meteorological measurements within arbitrary periods of 3, 6 and 10 days. Supervised binary classification and regression approaches were applied. The binary classification approach allowed for the definition of a model that alerts the farmer when a dangerous crop water stress episode is about to happen a day in advance. Performance metrics F2 and recall of up to 0.735 and 0.769, respectively, were obtained. With the regression approach a R² of up to 0.817 was achieved.