The development of suitability species models look for the availability to growth in a study area. These models can be used for different targets. In this research, a suitability model of Eucalyptus has been developed to soils contaminated by trace elements management. Guadiamar Green Corridor has been selected due to the huge data available regarding trace elements, forestry species and so on. Logistic regression (LR) and Random Forest (RF), as popular machine learning model, were applied in a geodatabase from Guadiamar Green Corridor with more of 20 years of data. This database is composed by soil physical and chemical variables, climate (temperature min and max, annual precipitation), forestry species. The results show the poor performance of LR and RF applied directly over the unbalanced training set. However, when Up-sampling or SMOTE are applied, both procedures improve its sensitivity, however, RF show more improve that LR. The methodology applied can help to determine the potential distribution of Eucalyptus in similar Mediterranean areas and extended to different areas according to Soil, Climate and Trace Elements data. Finally, the models developed under this research work can be used to reduce human and environmental health by trace elements taking into account local conditions but also climate change scenarios.

During the past years, air quality has become an important global issue, due to its impact on people's lives and the environment, and has caused severe problems for humans. As a prevention to effectively control air pollution, forecasting models have been developed as a base for decision-makers and urban managers during the past decades. In general, these methods can be divided into three classes: statistical methods, machine learning methods and hybrid methods. This study's primary intent is to supply an overview of air pollution prediction techniques in urban areas and their advantages and disadvantages. A comparison has also been made between the methods in terms of error assessment and the use of geospatial information systems (GIS). In addition, several approaches were applied to actual data, and the findings were compared to those acquired from previous published literatures. The results showed that forecasting using machine learning and hybrid methods has provided better results. It has also been demonstrated that GIS can improve the results of the forecasting methods.

The sediment transport, involving the movement of the bedload and suspended sediment in the basins, is a critical environmental concern that worsens water scarcity and leads to degradation of land and its ecosystems. Machine learning (ML) algorithms have emerged as powerful tools for predicting sediment yield. However, their use by decision-makers can be attributed to concerns regarding their consistency with the involved physical processes. In light of this issue, this study aims to develop a physics-informed ML approach for predicting sediment yield. To achieve this objective, Gaussian, Center, Regular, and Direct Copulas were employed to generate virtual combinations of physical of the sub-basins and hydrological datasets. These datasets were then utilized to train deep neural network (DNN), conventional neural network (CNN), Extra Tree, and XGBoost (XGB) models. The performance of these models was compared with the modified universal soil loss equation (MUSLE), which serves as a process-based model. The results demonstrated that the ML models outperformed the MUSLE model, exhibiting improvements in Nash–Sutcliffe efficiency (NSE) of approximately 10%, 18%, 32%, and 41% for the DNN, CNN, Extra Tree, and XGB models, respectively. Furthermore, through Sobol sensitivity and Shapley additive explanation–based interpretability analyses, it was revealed that the Extra Tree model displayed greater consistency with the physical processes underlying sediment transport as modeled by MUSLE. The proposed framework provides new insights into enhancing the accuracy and applicability of ML models in forecasting sediment yield while maintaining consistency with natural processes. Consequently, it can prove valuable in simulating process-related strategies aimed at mitigating sediment transport at watershed scales, such as the implementation of best management practices.

The prediction and identification of the factors controlling heavy metal transfer in soil-crop ecosystems are of critical importance. In this study, random forest (RF), gradient boosted machine (GBM), and generalised linear (GLM) models were compared after being used to model and identify prior factors that affect the transfer of heavy metals (HMs) in soil-crop systems in the Yangtze River Delta, China, based on 13 covariates with 1822 pairs of soil-crop samples. The mean bioaccumulation factors (BAFs) for all crops followed the order Cd > Zn > As > Cu > Ni > Hg > Cr > Pb. The RF model showed the best prediction ability for the BAFs of HMs in soil-crop ecosystems, followed by GBM and GLM. The R2 values of the RF models for the BAFs of Zn, Cu, Cr, Ni, Hg, Cd, As, and Pb were 0.84, 0.66, 0.59, 0.58, 0.58, 0.51, 0.30, and 0.17, respectively. The primary controlling factor in soil-to-crop transfer of all HMs under study was plant type, followed by soil heavy metal content and soil organic materials. The model used herein could be used to assist the prediction of heavy metal contents in crops based on heavy metal contents in soil and other covariates, and can significantly reduce the cost, labour, and time requirements involved with laboratory analysis. It can also be used to quantify the importance of variables and identify potential control factors in heavy metal bioaccumulation in soil-crop ecosystems.

International audience | Soil contamination posed by potentially toxic elements is becoming more serious under continuously development of industrialization and the abuse of fertilizers and pesticides. The investigation of soil potentially toxic elements is therefore urgently needed to ensure human and other organisms’ health. In this study, we investigated the feasibility of the separate and combined use of portable X-ray fluorescence (pXRF) and visible near-infrared reflectance (vis-NIR) sensors for measuring eight potentially toxic elements in soil. Low-level fusion was achieved by the direct combination of the pXRF and vis-NIR spectra; middle-level fusion was achieved by the combination of selected bands of the pXRF and vis-NIR spectra using the Boruta feature selection algorithm; and high-level fusion was conducted by outer-product analysis (OPA) and Granger–Ramanathan averaging (GRA). The estimation accuracy for the eight considered elements were in the following order: Zn > Cu > Ni > Cr > As > Cd > Pb > Hg. The measurement for Cu and Zn could be achieved by pXRF spectra alone with Lin’s concordance correlation coefficient (LCCC) values of 0.96 and 0.98, and ratio of performance to interquartile distance (RPIQ) values of 2.36 and 2.69, respectively. The measurement of Ni had the highest model performance for high-level fusion GRA with LCCC of 0.89 and RPIQ of 3.42. The measurements of Cr using middle- and high-level fusion were similar, with LCCC of 0.86 and RPIQ of 2.97. The best estimation accuracy for As, Cd, and Pb were obtained by high-level fusion using OPA, with LCCC >0.72 and RPIQ >1.2. However, Hg measurement by these techniques failed, having an unacceptable performance of LCCC <0.20 and RPIQ <0.75. These results confirm the effectiveness of using portable spectrometers to determine the contents of several potentially toxic elements in soils.

International audience

Nitrous oxide is a potent greenhouse gas, with a global warming potential 298 times greater than that of CO2. In agricultural soils, N2O emissions are influenced by a large number of environmental characteristics and crop management techniques that are not systematically reported in experiments. Random Forest (RF) is a machine learning method that can handle missing data and ranks input variables on the basis of their importance. We aimed to predict N2O emission on the basis of local information, to rank environmental and crop management variables according to their influence on N2O emission, and to compare the performances of RI: with several regression models. RF outperformed the regression models for predictive purposes, and this approach led to the identification of three important input variables: N fertilization, type of crop, and experiment duration. This method could be used in the future for prediction of N2O emissions from local information. (c) 2013 Elsevier Ltd. All rights reserved.

peer reviewed | A dust storm is a major environmental problem affecting many arid regions worldwide. The novel contribution of this study is combining indicators extracted from RS- and statistic-based predictive models to spatial mapping of land susceptibility to dust emissions in a very important dust source area in the borders of Iran and Iraq (Khuzestan province in Iran and Al-Basrah and Maysan provinces in Iraq). In this research, remote sensing (RS) techniques and machine learning techniques, including multivariate adaptive regression spline (MARS), random forest (RF), and logistic regression (LR), were used for dust source identification and susceptibility map preparation. To this end, 152 DSA for the period of 2005-2020 were identified in the study area. Of these DSA data, 70% was assigned to the Dust Source Susceptibility Mapping (DSSM) (training dataset) and 30% to model validation. Consequently, six factors (i.e., soil, lithology, slope, normalized vegetation differential index (NDVI), geomorphology, and land use units) were prepared as DSA's independent and effective variables. The results of all three models indicated that land use had the most impact on DSA. The validation results of these models using the test data showed sub-curves of 0.92, 0.86, and 0.76 for the RF, MARS, and LR models, respectively. Also, results showed that the RF model outperformed MARS (AUC = 0.89) and LR (AUC = 0.78) methods. In all three models, high and very high susceptibility classes generally covered a large percentage of the case study. The highest percentage of dust source points was also in this susceptibility category. Overall, the results of this study can be useful for planners and managers to control and reduce the risk of negative dust consequences.

Ecological risk assessment of contaminated sediment has become a fundamental component of water quality management programs, supporting decision-making for management actions or prompting additional investigations. In this study, we proposed a machine learning (ML)-based approach to assess the ecological risk of contaminated sediment as an alternative to existing index-based methods and costly toxicity testing. The performance of three widely used index-based methods (the pollution load index, potential ecological risk index, and mean probable effect concentration) and three ML algorithms (random forest, support vector machine, and extreme gradient boosting [XGB]) were compared in their prediction of sediment toxicity using 327 nationwide data sets from Korea consisting of 14 sediment quality parameters and sediment toxicity testing data. We also compared the performances of classifiers and regressors in predicting the toxicity for each of RF, SVM, and XGB algorithms. For all algorithms, the classifiers poorly classified toxic and non-toxic samples due to limited information on the sediment composition and the small training dataset. The regressors with a given classification threshold provided better classification, with the XGB regressor outperforming the other models in the classification. A permutation feature importance analysis revealed that Cr, Cu, Pb, and Zn were major contributors to toxicity prediction. The ML-based approach has the potential to be even more useful in the future with the expected increase in available sediment data.

Dissolved oxygen (DO) is an effective indicator for water pollution. However, since DO is a non-optically active parameter and has little impact on the spectrum captured by satellite sensors, research on estimating DO by remote sensing at multiple spatiotemporal scales is limited. In this study, the support vector regression (SVR) models were developed and validated using the remote sensing reflectance derived from both Landsat and Moderate Resolution Imaging Spectroradiometer (MODIS) data and synchronous DO measurements (N = 188) and water temperature of Lake Huron and three other inland waterbodies (N = 282) covering latitude between 22–45 °N. Using the developed models, spatial distributions of the annual and monthly DO variability since 1984 and the annual monthly DO variability since 2000 in Lake Huron were reconstructed for the first time. The impacts of five climate factors on long-term DO trends were analyzed. Results showed that the developed SVR-based models had good robustness and generalization (average R² = 0.91, root mean square percentage error = 2.65%, mean absolute percentage error = 4.21%), and performed better than random forest and multiple linear regression. The monthly DO estimates by Landsat and MODIS data were highly consistent (average R² = 0.88). From 1984 to 2019, the oxygen loss in Lake Huron was 6.56%. Air temperature, incident shortwave radiation flux density, and precipitation were the main climate factors affecting annual DO of Lake Huron. This study demonstrated that using SVR-based models, Landsat and MODIS data could be used for long-term DO retrieval at multiple spatial and temporal scales. As data-driven models, combining spectrum and water temperature as well as extending the training set to cover more DO conditions could effectively improve model robustness and generalization.