Machine Learning and Feature Selection for soil spectroscopy. An evaluation of Random Forest wrappers to predict soil organic matter, clay, and carbonates.

Soil spectroscopy estimates soil properties using the absorption features in soil spectra. However, modelling soil properties with soil spectroscopy is challenging due to the high dimensionality of spectral data. Feature Selection wrapper methods are promising approaches to reduce the dimensionality but are barely used in soil spectroscopy. The aim of this study is to evaluate the performance of two feature selection wrapper methods, Sequential Forward Selection (SFS) and Sequential Flotant Forward Selection (SFFS) built using the Random Forest (RF) algorithm, for dimensionality reduction of spectral data and predictive modelling of modelling soil organic matter (SOM), clay and carbonates. The reflectance of 100 soil samples, acquired from Sierra de las Nieves (Spain), was measured under laboratory conditions using ASD FieldSpec Pro JR. Four different datasets were obtained after applying two spectral preprocessing methods to raw spectra: raw spectra, Continuum Removal (CR), Multiplicative Scatter Correction (MSC), and a so-called "Global" dataset composed of raw, CR and MSC features. The performance of RF models built with feature selection methods was compared to that of Partial Least Squares Regression (PLSR) and RF (alone). RF models built with SFS and SFFS outperformed PLSR and RF alone models: The best RF models with feature selection had a respective ratio of performance to interquartile distance of 1.93, 0.38 and 2.56. PLSR models had an accuracy of 1.41, 0.29 and 1.81 for SOM, carbonates, and clay, respectively. RF alone had a respective performance of 1.29, 0.29 and 1.81. The application of feature selection wrapper methods reduced the number of features to less than 1 % of the starting features. Features were selected across all spectra for SOM and clay, and around 900 nm, 1900 nm, and 2350 nm for carbonates. However, feature selection highlighted features around 1100 nm in SOM modelling, as well as other features around 2200 nm, which is considered a main absorption feature of clay. The application of feature selection with Random Forest was very important in improving modelling accuracy, reducing the redundant features and avoiding the curse of dimensionality or Hughes effect. Thus, this research showed an alternative to dimensionality reduction approaches that have been applied to date to model soil properties with spectroscopy and paves the way for further scientific investigation based on feature selection methods and machine learning.

Integrated modeling of labile and glycated hemoglobin with glucose for enhanced diabetes detection and short-term monitoring.

Metabolic biomarkers, particularly glycated hemoglobin and fasting plasma glucose, are pivotal in the diagnosis and control of diabetes mellitus. Despite their importance, they exhibit limitations in assessing short-term glucose variations. In this study, we propose labile hemoglobin as an additional biomarker, providing insightful perspectives into these fluctuations. By utilizing datasets from 40,652 retrospective general participants and conducting glucose tolerance tests on 60 prospective pediatric subjects, we explored the relationship between plasma glucose and labile hemoglobin. A mathematical model was developed to encapsulate short-term glucose kinetics in the pediatric group. Applying dimensionality reduction techniques, we successfully identified participant subclusters, facilitating the differentiation between diabetic and non-diabetic individuals. Intriguingly, by integrating labile hemoglobin measurements with plasma glucose values, we were able to predict the likelihood of diabetes in pediatric subjects, underscoring the potential of labile hemoglobin as a significant glycemic biomarker for diabetes research.

Variable selection for nonlinear dimensionality reduction of biological datasets through bootstrapping of correlation networks.

Identifying the most relevant variables or features in massive datasets for dimensionality reduction can lead to improved and more informative display, faster computation times, and more explainable models of complex systems. Despite significant advances and available algorithms, this task generally remains challenging, especially in unsupervised settings. In this work, we propose a method that constructs correlation networks using all intervening variables and then selects the most informative ones based on network bootstrapping. The method can be applied in both supervised and unsupervised scenarios. We demonstrate its functionality by applying Uniform Manifold Approximation and Projection for dimensionality reduction to several high-dimensional biological datasets, derived from 4D live imaging recordings of hundreds of morpho-kinetic variables, describing the dynamics of thousands of individual leukocytes at sites of prominent inflammation. We compare our method with other standard ones in the field, such as Principal Component Analysis and Elastic Net, showing that it outperforms them. The proposed method can be employed in a wide range of applications, encompassing data analysis and machine learning.

Can Citizen Science Contribute to Avian Influenza Surveillance?

Global change is an important driver of the increase in emerging infectious diseases in recent decades. In parallel, interest in nature has increased, and different citizen science platforms have been developed to record wildlife observations from the general public. Some of these platforms also allow registering the observations of dead or sick birds. Here, we test the utility of live, sick and dead observations of birds recorded on the platform Observation.org for the early detection of highly pathogenic avian influenza virus (HPAIV) outbreaks in the wild in Belgium and The Netherlands. There were no significant differences in the morbidity/mortality rate through Observation.org one to four weeks in advance. However, the results show that the HPAIV outbreaks officially reported by the World Organisation for Animal Health (WOAH) overlapped in time with sudden increases in the records of sick and dead birds in the wild. In addition, in two of the five main HPAIV outbreaks recorded between 2016 and 2021, wild Anseriformes mortality increased one to two months before outbreak declaration. Although we cannot exclude that this increase was related to other causes such as other infectious diseases, we propose that Observation.org is a useful nature platform to complement animal health surveillance in wild birds. We propose possible approaches to improve the utility of the platform for pathogen surveillance in wildlife and discuss the potential for HPAIV outbreak detection systems based on citizen science to complement current surveillance programs of health authorities.

Thirty-four years of Landsat monitoring reveal long-term effects of groundwater abstractions on a World Heritage Site wetland.

Wetlands are among the most biodiverse yet endangered ecosystems on Earth. Despite being the most important wetland in Europe, the Doñana National Park (southwestern Spain) is no exception, and the increase of nearby groundwater abstractions for intensive agriculture and human supply has raised international concerns about the conservation of this iconic wetland. It is thus needed to assess wetlands' long-term trends and responses to global and local factors to make informed management decisions. In this paper, we used 442 Landsat satellite images to analyze the historical trends and drivers of the date of desiccation and maximum flooded area in 316 ponds located in Doñana National Park during a 34-year period (1985-2018) and found that 59 % of the ponds studied are currently desiccated. The use of Generalized Additive Mixed Models (GAMMs) showed inter-annual variation in rainfall and temperature as the most important factors determining pond flooding. However, GAMMS also showed that intensive agriculture and the nearby tourist resort were related to the desiccation or shrinking of ponds all over Doñana, finding that the strongest negative flooding anomalies (i.e. ponds flooding less than explained by climate alone) were located in proximity to pumping areas. These results suggest that current levels of groundwater exploitation may be unsustainable and require urgent measures to control abstractions to ensure the integrity of the Doñana pond network, and the persistence of >600 wetland-dependent species.

Evaluating the BFAST method to detect and characterise changing trends in water time series: A case study on the impact of droughts on the Mediterranean climate.

Mediterranean climate regions are facing increased aridity conditions and water scarcity, thus needing integrated management of water resources. Detecting and characterising changes in water resources over time is the natural first step towards identifying the drivers of these changes and understanding the mechanism of change. The aim of this study is to evaluate the potential of Breaks For Additive Seasonal and Trend (BFAST) method to identify gradual (trend) and abrupt (step- change) changes in the freshwater resources time series over a long-term period. This research shows an alternative to the Pettitt's test, LOESS (locally estimated scatterplot smoothing) filter, Mann-Kendall trend test among other common methods for change detection in hydrological data, and paves the way for further scientific investigation related to climate variability and its influence on water resources. We used the monthly accumulated stored water in three reservoirs, the monthly groundwater levels of three hydrological settings and a standardized precipitation index to show BFAST performance. BFAST was successfully applied, enabling: (1) assessment of the suitability of past management decisions when tackling drought events; (2) detection of recovery and drawdown periods (duration and magnitude values) of accumulated stored water in reservoirs and groundwater bodies after wet and dry periods; 3) measurement of resilience to drought conditions; (4) establishment of similarities/differences in trends between different reservoirs and groundwater bodies with regard to drought events.

Behavioural immune landscapes of inflammation.

Transcriptional and proteomic profiling of individual cells have revolutionized interpretation of biological phenomena by providing cellular landscapes of healthy and diseased tissues1,2. These approaches, however, do not describe dynamic scenarios in which cells continuously change their biochemical properties and downstream 'behavioural' outputs3-5. Here we used 4D live imaging to record tens to hundreds of morpho-kinetic parameters describing the dynamics of individual leukocytes at sites of active inflammation. By analysing more than 100,000 reconstructions of cell shapes and tracks over time, we obtained behavioural descriptors of individual cells and used these high-dimensional datasets to build behavioural landscapes. These landscapes recognized leukocyte identities in the inflamed skin and trachea, and uncovered a continuum of neutrophil states inside blood vessels, including a large, sessile state that was embraced by the underlying endothelium and associated with pathogenic inflammation. Behavioural screening in 24 mouse mutants identified the kinase Fgr as a driver of this pathogenic state, and interference with Fgr protected mice from inflammatory injury. Thus, behavioural landscapes report distinct properties of dynamic environments at high cellular resolution.

SURFWET: A biokinetic model for surface flow constructed wetlands.

Constructed wetlands are an alternative biotechnology for wastewater treatment that have several advantages over conventional systems. In this work, a biokinetic model for surface flow constructed wetlands is presented (SURFWET). SURFWET belongs to a class of models that are not only interesting from a theoretical viewpoint, as they allow to improve the understanding of the underlying processes; but also from a practical viewpoint, because they can be useful for optimal designs of constructed wetlands, complementing current empirical methods. The proposed model is centered on the intervening physical and biochemical processes involved in pollutant removal in wastewater (organic matter, nitrogen, phosphorus, suspended solids), capturing the interplay of the main agents on contaminant removal (bacteria, macrophytes and phytoplankton). Furthermore, the hydraulic model considers water volume as a variable depending on the outlet hydraulic capacity, and dissolved oxygen has also been introduced as a key driver of reaction kinetics of wetlands. Beyond putting forward a theoretical framework, SURFWET has been applied to simulate a specific case to demonstrate its robustness, in a 12-year-interval simulation. The results show the typical seasonality of this biotechnology, highlighting the importance of dissolved oxygen, which is a key limiting factor on a large number of biochemical processes.

Effects of flooding regime and meteorological variability on the removal efficiency of treatment wetlands under a Mediterranean climate.

The modeling of free-water surface constructed wetlands (FWS-CWs) provides an improved understanding of their processes and constitutes a useful tool for the design and management of these systems. In this work, a dynamic simulation model for FWS-CWs was developed and used to simulate the operation of a FWS-CW proposed for improving the treatment of sewage effluents entering the Tablas de Daimiel National Park in central Spain. The process-based model simulates carbon, nitrogen and phosphorus dynamics, including key hydrological processes for wetlands under a fluctuating Mediterranean semiarid climate. The model allows for the simulation of the operation of FWS-CWs with variable flooding regimes, relating the surface water level to the flooded area and the water outflow. Simulations of the proposed FWS-CW under different water management schemes and scenarios were run, and the consequences of those management strategies on the treatment efficiency were analyzed. Under the Mediterranean climate and geology of the study area, namely, high water losses through evapotranspiration and infiltration, the decrease in nutrient concentrations was higher when the flooded area was reduced in summer than when a constant flooded area was maintained. Moreover, the meteorological variability introduced in different scenarios produced different results in terms of water outflow, but differences in terms of nutrient concentrations were not significant. The ability of the model to simulate different hydrological scenarios and their consequences on water quality makes it a useful decision-support tool.

Predictive models of turbidity and water depth in the Doñana marshes using Landsat TM and ETM+ images.

We have used Landsat-5 TM and Landsat-7 ETM+ images together with simultaneous ground-truth data at sample points in the Doñana marshes to predict water turbidity and depth from band reflectance using Generalized Additive Models. We have point samples for 12 different dates simultaneous with 7 Landsat-5 and 5 Landsat-7 overpasses. The best model for water turbidity in the marsh explained 38% of variance in ground-truth data and included as predictors band 3 (630-690 nm), band 5 (1550-1750 nm) and the ratio between bands 1 (450-520 nm) and 4 (760-900 nm). Water turbidity is easier to predict for water bodies like the Guadalquivir River and artificial ponds that are deep and not affected by bottom soil reflectance and aquatic vegetation. For the latter, a simple model using band 3 reflectance explains 78.6% of the variance. Water depth is easier to predict than turbidity. The best model for water depth in the marsh explains 78% of the variance and includes as predictors band 1, band 5, the ratio between band 2 (520-600 nm) and band 4, and bottom soil reflectance in band 4 in September, when the marsh is dry. The water turbidity and water depth models have been developed in order to reconstruct historical changes in Doñana wetlands during the last 30 years using the Landsat satellite images time series.