An automated machine learning based ensemble approach for improving estimates of soil water retention parameters

<p>Accurately mapping soil water retention parameters is vital for modeling atmosphere-land interactions but is challenged by limited measurements and simulations globally. Ensemble pedotransfer functions (PTFs) have been highly recommended for use due to the higher reliability of ensemble models and the error compensation among ensemble members. However, conventional ensemble approaches assign a fixed weight to each PTF and may not fully utilize the strengths of individual PTFs. In this work, we developed a new ensemble approach based on an automated machine learning workflow to assign varying weights to assemble 13 widely used PTFs. The AutoML-assisted ensemble approach (AutoML-Ens), as well as the simple average (MEAN), Bayesian Model Average (BMA), and the hierarchical multi-model ensemble approach (HMME), were evaluated using the global coverage National Cooperative Soil Surbey (NCSS) Soil Characterization Database. Results indicate that AutoML-Ens approach performs better than the conventional approaches in terms of the coefficient of determination (R<sup>2</sup>) and root mean square error (RMSE). Three soil hydraulic parameters, i.e., saturated water content, field capacity, and wilting points, and their corresponding uncertainties, were further derived through the AutoML-Ens approach at a 30’’×30’’ geographical spatial resolution based on a global soil composition database (SoilGrids), which can be applied in the Earth System Modeling. This study demonstrated the necessity of dynamic weights assigning in ensemble approaches and the great potential of coupling data-driven (here, the AutoML) and modeling (empirically or physically-based PTFs) approaches in mapping global soil water retention-like parameters.</p>

Assessment of Lipophilicity Descriptors of Selected NSAIDs Obtained at Different TLC Stationary Phases

Lipophilicity study of selected NSAIDs, the group of the bioactive compounds usually used in humans and animals medicine, with the use of experimental and calculation methods was evaluated. LogP values are proposed and compared as descriptors of the lipophilicity of eleven compounds (from oxicams and coxibs). Obtained data were designated by thin-layer chromatography (TLC) in various chromatographic conditions, with stationary phases with different properties. The mobile phase systems were prepared by mixing the respective amounts of water and organic modifier, methanol and acetone, in the range of 30 to 80% (v/v) in 5% increments. Retention parameters (RF, RM and RM0) were calculated and statistically evaluated to establish correlations. All experimentally determined RM0 values were compared with partition coefficients obtained by computational methods using linear regression analysis. Moreover, in order to extract information about the lipophilicity of compounds from large retention datasets, two chemometric approaches, namely principal component analysis (PCA) and cluster analysis (CA) were carried out. Established models of lipophilicity may have the potential to predict the biological activity of a number of drugs. The presented knowledge may also be of use during drug discovery processes, broadening the knowledge of potential ways to modify the physicochemical properties of chemical compounds.

Controlling the correctness of retention parameters variations in reversed phase HPLC using recurrent relations

Recurrent approximation of analytes’ net retention times (tR) in reversed phase high performance liquid chromatography (RP HPLC) at different contents of organic constitu­ent in the eluent (C) is recommended as a method of revealing the reversible hydrate for­ma­tion. The criterion of that are the deviations of the dependences tR(С + DС) = atR(С) + b (*) from linearity, where DС is the constant increment of concentration variations; in our case DС = 5%. However, such deviations are rather small and, hence their measuring requires high robustness of the equipment involved. Besides hydrate formation, there are additional reasons for deviations, namely discrepancies between the real and the selected flows of the eluent. Compa­ring tR values obtained for the same analytes using the same chromatogra­phic column at the same conditions, but with different HPLC instruments using the systems methanol – water as the eluent confirms that tR values of one data set are equal only to approx. 76-98% tR values of another data set. Therefore, the eluent flow in the second case exceeds that in the first regime at the same pro­por­tion. The simple method for revealing such flow deviations is proposed. It is based on the recurrent approximation of tR = f(C) data sets for any compounds forming no hydrates in RP HPLC conditions (chlorobenzene was selected). The absence of the influ­en­ce of any distorted factors is confirmed with values of correlation coefficients for re­cur­rent depen­dencies (*) exceeding 0.999.

Modification of the MONERIS Nutrient Emission Model for a Lowland Country (Hungary) to Support River Basin Management Planning in the Danube River Basin

The contamination of waters with nutrients, especially nitrogen and phosphorus originating from various diffuse and point sources, has become a worldwide issue in recent decades. Due to the complexity of the processes involved, watershed models are gaining an increasing role in their analysis. The goal set by the EU Water Framework Directive (to reach “good status” of all water bodies) requires spatially detailed information on the fate of contaminants. In this study, the watershed nutrient model MONERIS was applied to the Hungarian part of the Danube River Basin. The spatial resolution was 1078 water bodies (mean area of 86 km2); two subsequent 4 year periods (2009–2012 and 2013–2016) were modeled. Various elements/parameters of the model were adjusted and tested against surface and subsurface water quality measurements conducted all over the country, namely (i) the water balance equations (surface and subsurface runoff), (ii) the nitrogen retention parameters of the subsurface pathways (excluding tile drainage), (iii) the shallow groundwater phosphorus concentrations, and (iv) the surface water retention parameters. The study revealed that (i) digital-filter-based separation of surface and subsurface runoff yielded different values of these components, but this change did not influence nutrient loads significantly; (ii) shallow groundwater phosphorus concentrations in the sandy soils of Hungary differ from those of the MONERIS default values; (iii) a significant change of the phosphorus in-stream retention parameters was needed to approach measured in-stream phosphorus load values. Local emissions and pathways were analyzed and compared with previous model results.