Cannot see the random forest for the decision trees: selecting predictive models for restoration ecology

Amount, distribution and controls of the soil organic carbon storage loss in the degraded China's grasslands

More than 80 % of China's grasslands are classified as degraded, and the loss of soil carbon storage due to degradation has a significant impact on China's terrestrial carbon sinks as well as carbon neutrality targets. The loss of soil carbon storage in degraded grasslands can serve as a benchmark for quantifying the carbon sequestration capacity of restored grasslands in the future. Here, above- and below-ground biomass, soil organic carbon (SOC) content at various depths (0-100 cm) and soil bulk density were collected from 226 degradation sequences around China. The above information was integrated and statistically analyzed to quantify the difference of SOC storage between the degraded and natural grassland at national scale. The result showed that grassland degradation led to a significant reduction in SOC storage across different depths. SOC (0-100 cm) of degraded grassland decreased by 39 % compared to that of natural grassland, ranging from 21 % in the lightly degraded sites to 59 % of the extremely degraded sites. 15 potential predictors were used to estimate the national amount of these differences of 0-20 cm depth SOC storage as 5.29 ± 1.59 Pg C. This considerable carbon storage gap implies the necessity of China's grassland restoration project in achieving carbon neutrality goals in the future.

Predicting landscape-scale biodiversity recovery by natural tropical forest regrowth

Natural forest regrowth is a cost-effective, nature-based solution for biodiversity recovery, yet different socioenvironmental factors can lead to variable outcomes. A critical knowledge gap in forest restoration planning is how to predict where natural forest regrowth is likely to lead to high levels of biodiversity recovery, which is an indicator of conservation value and the potential provisioning of diverse ecosystem services. We sought to predict and map landscape-scale recovery of species richness and total abundance of vertebrates, invertebrates, and plants in tropical and subtropical second-growth forests to inform spatial restoration planning. First, we conducted a global meta-analysis to quantify the extent to which recovery of species richness and total abundance in second-growth forests deviated from biodiversity values in reference old-growth forests in the same landscape. Second, we employed a machine-learning algorithm and a comprehensive set of socioenvironmental factors to spatially predict landscape-scale deviation and map it. Models explained on average 34% of observed variance in recovery (range 9-51%). Landscape-scale biodiversity recovery in second-growth forests was spatially predicted based on socioenvironmental landscape factors (human demography, land use and cover, anthropogenic and natural disturbance, ecosystem productivity, and topography and soil chemistry); was significantly higher for species richness than for total abundance for vertebrates (median range-adjusted predicted deviation 0.09 vs. 0.34) and invertebrates (0.2 vs. 0.35) but not for plants (which showed a similar recovery for both metrics [0.24 vs. 0.25]); and was positively correlated for total abundance of plant and vertebrate species (Pearson r = 0.45, p = 0.001). Our approach can help identify tropical and subtropical forest landscapes with high potential for biodiversity recovery through natural forest regrowth.

The interplay of flow processes shapes aquatic invertebrate successions in floodplain channels - A modelling applied to restoration scenarios

The high biotic diversity supported by floodplains is ruled by the interplay of geomorphic and hydrological processes at various time scales, from daily fluctuations to decennial successions. Because understanding such processes is a key question in river restoration, we attempted to model changes in taxonomic richness in an assemblage of 58 macroinvertebrate taxa (21 gastropoda and 37 ephemeroptera, plecoptera and trichoptera, EPT) along two successional sequences typical for former braided channels. Individual models relating the occurrence of taxa to overflow and backflow durations were developed from field measurements in 19 floodplain channels of the Rhône floodplain (France) monitored over 10 years. The models were combined to simulate diversity changes along a progressive alluviation and disconnection sequence after the reconnection with the main river of a previously isolated channel. Two scenarios were considered: (i) an upstream + downstream reconnection creating a lotic channel, (ii) a downstream reconnection creating a semi-lotic channel. Reconnection led to a direct increase in invertebrate richness (on average x2.5). However, taxonomical richness showed a constant decrease as isolation progressed and reached an average of 2 for EPT and 7 for gastropods at the end of the scenarios. With more than 80% of the taxonomic models with an AUC equal or higher than 0.7 and slopes of linear relations between observed and predicted richness of 0.75 (gastropods) and 1 (EPT), the Boosted Regression Trees (BRT) provided a good basis for prediction of species assemblages. These models can be used to quantify a priori the sustainability and ecological efficiency of restoration actions and help floodplain restoration planning and management.

A machine learning-based framework for Predicting Treatment Failure in tuberculosis: A case study of six countries

Tuberculosis is ranked as the 2nd deadliest disease in the world and is responsible for ten million deaths in 2017. Treatment failure is one of a main reason behind these deaths. Reasons of treatment failure are still unknown and the death rate due to TB is increasing. Machine learning and data analytics approaches are proved to be useful in healthcare domain in finding the associations among different attributes that can affect the outcome of any disease. Timely identification of reasons can save a patient's life. This study aims to find features that are strongly correlated with treatment failure using feature selection techniques. The validation of features is demonstrated using different classification algorithms. Moreover, this study provides a demographic based feature association of six highly burdened treatment failure countries. A verified real-life patient's dataset gathered from different countries including Azerbaijan, Belarus, Georgia, India, Moldova, and Romania is utilized to address the problem. Two types of experimentation are performed on combined dataset by achieving an average accuracy of 78% and an accuracy of 92% on Romania's data. Results shows the importance of features obtained through this study are highly influential in leading a patient towards treatment failure.

Soft Sensor Application in Identification of the Activated Sludge Bulking Considering the Technological and Economical Aspects of Smart Systems Functioning

The paper presented the methodology for the construction of a soft sensor used for activated sludge bulking identification. Devising such solutions fits within the current trends and development of a smart system and infrastructure within smart cities. In order to optimize the selection of the data-mining method depending on the data collected within a wastewater treatment plant (WWTP), a number of methods were considered, including: artificial neural networks, support vector machines, random forests, boosted trees, and logistic regression. The analysis conducted sought the combinations of independent variables for which the devised soft sensor is characterized with high accuracy and at a relatively low cost of determination. With the measurement results pertaining to the quantity and quality of wastewater as well as the temperature in the activated sludge chambers, a good fit can be achieved with the boosted trees method. In order to simplify the selection of an optimal method for the identification of activated sludge bulking depending on the model requirements and the data collected within the WWTP, an original system of weight estimation was proposed, enabling a reduction in the number of independent variables in a model—quantity and quality of wastewater, operational parameters, and the cost of conducting measurements.

Combination of an Automated 3D Field Phenotyping Workflow and Predictive Modelling for High-Throughput and Non-Invasive Phenotyping of Grape Bunches

In grapevine breeding, loose grape bunch architecture is one of the most important selection traits, contributing to an increased resilience towards Botrytis bunch rot. Grape bunch architecture is mainly influenced by the berry number, berry size, the total berry volume, and bunch width and length. For an objective, precise, and high-throughput assessment of these architectural traits, the 3D imaging sensor Artec® Spider was applied to gather dense point clouds of the visible side of grape bunches directly in the field. Data acquisition in the field is much faster and non-destructive in comparison to lab applications but results in incomplete point clouds and, thus, mostly incomplete phenotypic values. Therefore, lab scans of whole bunches (360°) were used as ground truth. We observed strong correlations between field and lab data but also shifts in mean and max values, especially for the berry number and total berry volume. For this reason, the present study is focused on the training and validation of different predictive regression models using 3D data from approximately 2000 different grape bunches in order to predict incomplete bunch traits from field data. Modeling concepts included simple linear regression and machine learning-based approaches. The support vector machine was the best and most robust regression model, predicting the phenotypic traits with an R2 of 0.70–0.91. As a breeding orientated proof-of-concept, we additionally performed a Quantitative Trait Loci (QTL)-analysis with both the field modeled and lab data. All types of data resulted in joint QTL regions, indicating that this innovative, fast, and non-destructive phenotyping method is also applicable for molecular marker development and grapevine breeding research.