The use of sensory perception indicators for improving the characterization and modelling of total petroleum hydrocarbon (TPH) grade in soils.

This paper proposes a multistep approach for creating a 3D stochastic model of total petroleum hydrocarbon (TPH) grade in potentially polluted soils of a deactivated oil storage site by using chemical analysis results as primary or hard data and classes of sensory perception variables as secondary or soft data. First, the statistical relationship between the sensory perception variables (e.g. colour, odour and oil-water reaction) and TPH grade is analysed, after which the sensory perception variable exhibiting the highest correlation is selected (oil-water reaction in this case study). The probabilities of cells belonging to classes of oil-water reaction are then estimated for the entire soil volume using indicator kriging. Next, local histograms of TPH grade for each grid cell are computed, combining the probabilities of belonging to a specific sensory perception indicator class and conditional to the simulated values of TPH grade. Finally, simulated images of TPH grade are generated by using the P-field simulation algorithm, utilising the local histograms of TPH grade for each grid cell. The set of simulated TPH values allows several calculations to be performed, such as average values, local uncertainties and the probability of the TPH grade of the soil exceeding a specific threshold value.

Determination of accurate baseline representation for three Central Iowa watersheds within a HAWQS-based SWAT analyses.

Improving food systems to address food insecurity and minimize environmental impacts is still a challenge in the 21st century. Ecohydrological models are a key tool for accurate system representation and impact measurement. We used a multi-phase testing approach to represent baseline hydrologic conditions across three agricultural basins that drain parts of north central and central Iowa, U.S.: the Des Moines River Basin (DMRB), the South Skunk River Basin (SSRB), and the North Skunk River Basin (NSRB). The Soil and Water Assessment Tool (SWAT) ecohydrological model was applied using a framework consisting of the Hydrologic and Water Quality System (HAWQS) online platform, 40 streamflow gauges, the alternative runoff curve number method, additional tile drainage and fertilizer application. In addition, ten SWAT baselines were created to analyze both the HAWQS parameters (baseline 1) and nine alternative baseline configurations (considering the framework). Most of the models achieved acceptable statistical replication of measured (close to the outlet) streamflows, with Nash-Sutcliffe (NS) values ranging up to 0.80 for baseline 9 in the DMRB and SSRB, and 0.78 for baseline 7 in the NSRB. However, water balance and other hydrologic indicators revealed that careful selection of management data and other inputs are essential for obtaining the most accurate representation of baseline conditions for the simulated stream systems. Using cumulative distribution curves as a criterion, baselines 7 to 10 showed the best fit for the SSRB and NSRB, but none of the baselines accurately represented 20% of low flows for the DMRB. Analysis of snowmelt and growing season periods showed that baselines 3 and 4 resulted in poor simulations across all three basins using four common statistical measures (NS, KGE, Pbias, and R2), and that baseline 9 was characterized by the most satisfactory statistical results, followed by baselines 5, 7 and 1.

COVID-19 mortality analysis from soft-data multivariate curve regression and machine learning.

A multiple objective space-time forecasting approach is presented involving cyclical curve log-regression, and multivariate time series spatial residual correlation analysis. Specifically, the mean quadratic loss function is minimized in the framework of trigonometric regression. While, in our subsequent spatial residual correlation analysis, maximization of the likelihood allows us to compute the posterior mode in a Bayesian multivariate time series soft-data framework. The presented approach is applied to the analysis of COVID-19 mortality in the first wave affecting the Spanish Communities, since March 8, 2020 until May 13, 2020. An empirical comparative study with Machine Learning (ML) regression, based on random k-fold cross-validation, and bootstrapping confidence interval and probability density estimation, is carried out. This empirical analysis also investigates the performance of ML regression models in a hard- and soft-data frameworks. The results could be extrapolated to other counts, countries, and posterior COVID-19 waves. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1007/s00477-021-02021-0.

Western Lake Erie Basin: Soft-data-constrained, NHDPlus resolution watershed modeling and exploration of applicable conservation scenarios.

Complex watershed simulation models are powerful tools that can help scientists and policy-makers address challenging topics, such as land use management and water security. In the Western Lake Erie Basin (WLEB), complex hydrological models have been applied at various scales to help describe relationships between land use and water, nutrient, and sediment dynamics. This manuscript evaluated the capacity of the current Soil and Water Assessment Tool (SWAT) to predict hydrological and water quality processes within WLEB at the finest resolution watershed boundary unit (NHDPlus) along with the current conditions and conservation scenarios. The process based SWAT model was capable of the fine-scale computation and complex routing used in this project, as indicated by measured data at five gaging stations. The level of detail required for fine-scale spatial simulation made the use of both hard and soft data necessary in model calibration, alongside other model adaptations. Limitations to the model's predictive capacity were due to a paucity of data in the region at the NHDPlus scale rather than due to SWAT functionality. Results of treatment scenarios demonstrate variable effects of structural practices and nutrient management on sediment and nutrient loss dynamics. Targeting treatment to acres with critical outstanding conservation needs provides the largest return on investment in terms of nutrient loss reduction per dollar spent, relative to treating acres with lower inherent nutrient loss vulnerabilities. Importantly, this research raises considerations about use of models to guide land management decisions at very fine spatial scales. Decision makers using these results should be aware of data limitations that hinder fine-scale model interpretation.