Gonzalo Astray

Geographical origin determination for seafood products is a fundamental aspect due to its implications for fraud prevention, ensuring food safety, and promoting resource sustainable management. In this research, different machine learning (ML) models based on random forests, support vector machines, and artificial neural networks were fed with trace element fingerprinting (TEF) and stable isotope ratio analysis (SIRA) to determine the origin of mussels that have been farmed in eight regions and ten locations around the world (areas of the European Atlantic coast, the Mediterranean Sea, and the Pacific coast of Chile). Fourteen trace elements in shells and carbon and nitrogen isotope ratios of mussel tissue were used singly, in combination, or reduced to develop the different approach models. All the selected models present high prediction accuracies for the independent variables (except for SIRA models), for their combination, or for their optimisation, highlighting the artificial neural network and random forest models that presented a 100% accuracy for all cases using a combination of variables selected based on a random forest model TEF to predict region and location, respectively. This fact confirms that ML models are suitable approximation techniques to determine the region and location of Mediterranean mussel origin, with key applications in food safety and global sustainability.

Pollen forecasting models are helpful tools to predict environmental processes and allergenic risk events. Parietaria belongs to the Urticaceae family, and due to its high-level pollen production, is responsible for many cases of severe pollinosis reactions. This research aims to develop different machine learning models such as the random forest—RF, support vector machine—SVM, and artificial neural network—ANN models, to predict Parietaria pollen concentrations in the atmosphere of northwest Spain using 24 years of data from 1999 to 2022. The results obtained show an increase in the duration and intensity of the Parietaria main pollen season in the Mediterranean region (Ourense). Machine learning models exhibited their capacity to forecast Parietaria pollen concentrations at one, two, and three days ahead. The best selected models presented high correlation coefficients between 0.713 and 0.859, with root mean squared errors between 5.55 and 7.66 pollen grains·m⁻³ for the testing phase. The models developed could be improved by increasing the number of years, studying other hyperparameter ranges, or analyzing different data distributions.

The shiitake mushroom has gained popularity in the last decade, ranking second in the world for mushrooms consumed, providing consumers with a wide variety of nutritional and healthy benefits. It is often not clear the origin of these mushrooms, so it becomes of great importance to the consumers. In this research, different machine learning algorithms were developed to determine the geographical origin of shiitake mushrooms (Lentinula edodes) consumed in Korea, based on experimental data reported in the literature (δ¹³C, δ¹⁵N, δ¹⁸O, δ³⁴S, and origin). Regarding the origin of shiitake in three categories (Korean, Chinese, and mushrooms from Chinese inoculated sawdust blocks), the random forest model presents the highest accuracy value (0.940) and the highest kappa value (0.908) for the validation phase. To determine the origin of shiitake mushrooms in two categories (Korean and Chinese, including mushrooms from Chinese inoculated sawdust blocks in the latter ones), the support vector machine model is chosen as the best model due to the high accuracy (0.988) and kappa (0.975) values for the validation phase. Finally, to determine the origin in two categories (Korean and Chinese, but this time including the mushrooms from Chinese inoculated sawdust blocks in the Korean ones), the best model is the random forest due to its higher accuracy value (0.952) in the validation phase (kappa value of 0.869). The accuracy values in the testing phase for the best selected models are acceptable (between 0.839 and 0.964); therefore, the predictive capacity of the models could be acceptable for their use in real applications. This allows us to affirm that machine learning algorithms would be suitable modeling instruments to determine the geographical origin of shiitake.

The study of the phenolic compounds present in olive leaves (Olea europaea) is of great interest due to their health benefits. In this research, different machine learning algorithms such as RF, SVM, and ANN, with temperature, time, and volume as input variables, were developed to model the extract yield and the total phenolic content (TPC) from experimental data reported in the literature. In terms of extract yield, the neural network-based ANN_Z-L model presents the lowest root mean square error (RMSE) value in the validation phase (9.44 mg/g DL), which corresponds with a mean absolute percentage error (MAPE) of 3.7%. On the other hand, the best model to determine the TPC value was the neural network-based model ANN_R, with an RMSE of 0.89 mg GAE/g DL in the validation phase (MAPE of 2.9%). Both models obtain, for the test phase, MAPE values of 4.9 and 3.5%, respectively. This affirms that ANN models would be good modelling tools to determine the extract yield and TPC value of the ultrasound-assisted extraction (UAE) process of olive leaves under different temperatures, times, and solvents.

Reservoirs play an important function in human society due to their ability to hold and regulate the flow. This will play a key role in the future decades due to climate change. Therefore, having reliable predictions of the outflow from a reservoir is necessary for early warning systems and adequate water management. In this sense, this study uses three approaches machine learning (ML)-based techniques—Random Forest (RF), Support Vector Machine (SVM) and artificial neural network (ANN)—to predict outflow one day ahead of eight different dams belonging to the Miño-Sil Hydrographic Confederation (Galicia, Spain), using three input variables of the current day. Mostly, the results obtained showed that the suggested models work correctly in predicting reservoir outflow in normal conditions. Among the different ML approaches analyzed, ANN was the most appropriate technique since it was the one that provided the best model in five reservoirs.

Nowadays, there is an extensive production and use of plastic materials for different industrial activities. These plastics, either from their primary production sources or through their own degradation processes, can contaminate ecosystems with micro- and nanoplastics. Once in the aquatic environment, these microplastics can be the basis for the adsorption of chemical pollutants, favoring that these chemical pollutants disperse more quickly in the environment and can affect living beings. Due to the lack of information on adsorption, three machine learning models (random forest, support vector machine, and artificial neural network) were developed to predict different microplastic/water partition coefficients (log K_d) using two different approximations (based on the number of input variables). The best-selected machine learning models present, in general, correlation coefficients above 0.92 in the query phase, which indicates that these types of models could be used for the rapid estimation of the absorption of organic contaminants on microplastics.

Nowadays, there is extensive production and use of plastic materials for different industrial activities. These plastics, either from their primary production sources or through degradation processes of the plastics themselves, can contaminate the ecosystem with micro and nanoplastics. Once in the aquatic environment, these microplastics can be the basis for the adsorption of chemical pollutants, favoring that these chemical pollutants disperse more quickly in the environment and can affect living beings. Due to the lack of information on adsorption, three machine learning models (random forest, support vector machine, and artificial neural network) to predict different microplastic/water partition coefficients (log Kd) were developed using two different approximations (based on the number of input variables). The best-selected machine learning models present, in general, correlation coefficients upper than 0.92 in the query phase, which indicate that these type of models could be used for a rapid estimation of the absorption of organic contaminants on microplastics.

Global solar irradiation is an important variable that can be used to determine the suitability of an area to install solar systems; nevertheless, due to the limitations of requiring measurement stations around the entire world, it can be correlated with different meteorological parameters. To confront this issue, different locations in Rias Baixas (Autonomous Community of Galicia, Spain) and combinations of parameters (month and average temperature, among others) were used to develop various machine learning models (random forest -RF-, support vector machine -SVM- and artificial neural network -ANN-). These three approaches were used to model and predict (one month ahead) monthly global solar irradiation using the data from six measurement stations. Afterwards, these models were applied to seven different measurement stations to check if the knowledge acquired could be extrapolated to other locations. In general, the ANN models offered the best results for the development and testing phases of the model, as well as for the phase of knowledge extrapolation to other locations. In this sense, the selected ANNs obtained a mean absolute percentage error (MAPE) value between 3.9 and 13.8% for the model development and an overall MAPE between 4.1 and 12.5% for the other seven locations. ANNs can be a capable tool for modelling and predicting monthly global solar irradiation in areas where data are available and for extrapolating this knowledge to nearby areas.

The commercialization of products with cannabidiol (CBD) has undergone a significant increase. These products can be presented in different forms such as baked goods, gummies or beverages (such as kombucha, beer or teas, among others) using wide concentrations ranges. The use of CBD in edibles favors its consumption, for medicinal users, during the work week, avoid its possible social stigma and facilitates its transport. These products can be purchased on store shelves and online. There is a large number of specialized studies, in which the possible advantages of CBD consumption are described in the preclinical and clinical trials. It is also necessary to recognize the existence of other works revealing that the excessive consumption of CBD could have some repercussions on health. In this review, it is analyzed the composition and properties of Cannabis sativa L., the health benefits of cannabinoids (focusing on CBD), its consumption, its possible toxicological effects, a brief exposition of the extraction process, and a collection of different products that contain CBD in its composition.