Validation R2 Research Articles

Strategies of method selection for fine-scale PM<sub>2.5</sub> mapping in an intra-urban area using crowdsourced monitoring

Abstract. Fine particulate matter (PM2.5) is of great concern to the public due to its significant risk to human health. Numerous methods have been developed to estimate spatial PM2.5 concentrations in unobserved locations due to the sparse number of fixed monitoring stations. Due to an increase in low-cost sensing for air pollution monitoring, crowdsourced monitoring of exposure control has been gradually introduced into cities. However, the optimal mapping method for conventional sparse fixed measurements may not be suitable for this new high-density monitoring approach. This study presents a crowdsourced sampling campaign and strategies of method selection for 100 m scale PM2.5 mapping in an intra-urban area of China. During this process, PM2.5 concentrations were measured by laser air quality monitors through a group of volunteers during two 5 h periods. Three extensively employed modelling methods (ordinary kriging, OK; land use regression, LUR; and regression kriging, RK) were adopted to evaluate the performance. An interesting finding is that PM2.5 concentrations in micro-environments varied in the intra-urban area. These local PM2.5 variations can be easily identified by crowdsourced sampling rather than national air quality monitoring stations. The selection of models for fine-scale PM2.5 concentration mapping should be adjusted according to the changing sampling and pollution circumstances. During this project, OK interpolation performs best in conditions with non-peak traffic situations during a lightly polluted period (holdout validation R2: 0.47–0.82), while the RK modelling can perform better during the heavily polluted period (0.32–0.68) and in conditions with peak traffic and relatively few sampling sites (fewer than ∼100) during the lightly polluted period (0.40–0.69). Additionally, the LUR model demonstrates limited ability in estimating PM2.5 concentrations on very fine spatial and temporal scales in this study (0.04–0.55), which challenges the traditional point about the good performance of the LUR model for air pollution mapping. This method selection strategy provides empirical evidence for the best method selection for PM2.5 mapping using crowdsourced monitoring, and this provides a promising way to reduce the exposure risks for individuals in their daily life.

Cluster-based bagging of constrained mixed-effects models for high spatiotemporal resolution nitrogen oxides prediction over large regions

BackgroundAccurate estimation of nitrogen dioxide (NO2) and nitrogen oxide (NOx) concentrations at high spatiotemporal resolutions is crucial for improving evaluation of their health effects, particularly with respect to short-term exposures and acute health outcomes. For estimation over large regions like California, high spatial density field campaign measurements can be combined with more sparse routine monitoring network measurements to capture spatiotemporal variability of NO2 and NOx concentrations. However, monitors in spatially dense field sampling are often highly clustered and their uneven distribution creates a challenge for such combined use. Furthermore, heterogeneities due to seasonal patterns of meteorology and source mixtures between sub-regions (e.g. southern vs. northern California) need to be addressed. ObjectivesIn this study, we aim to develop highly accurate and adaptive machine learning models to predict high-resolution NO2 and NOx concentrations over large geographic regions using measurements from different sources that contain samples with heterogeneous spatiotemporal distributions and clustering patterns. MethodsWe used a comprehensive Kruskal-K-means method to cluster the measurement samples from multiple heterogeneous sources. Spatiotemporal cluster-based bootstrap aggregating (bagging) of the base mixed-effects models was then applied, leveraging the clusters to obtain balanced and less correlated training samples for less bias and improvement in generalization. Further, we used the machine learning technique of grid search to find the optimal interaction of temporal basis functions and the scale of spatial effects, which, together with spatiotemporal covariates, adequately captured spatiotemporal variability in NO2 and NOx at the state and local levels. ResultsWe found an optimal combination of four temporal basis functions and 200 m scale spatial effects for the base mixed-effects models. With the cluster-based bagging of the base models, we obtained robust predictions with an ensemble cross validation R2 of 0.88 for both NO2 and NOx [RMSE (RMSEIQR): 3.62 ppb (0.28) and 9.63 ppb (0.37) respectively]. In independent tests of random sampling, our models achieved similarly strong performance (R2 of 0.87–0.90; RMSE of 3.97–9.69 ppb; RMSEIQR of 0.21–0.27), illustrating minimal over-fitting. ConclusionsOur approach has important implications for fusing data from highly clustered and heterogeneous measurement samples from multiple data sources to produce highly accurate concentration estimates of air pollutants such as NO2 and NOx at high resolution over a large region.

A spatially structured adaptive two-stage model for retrieving ground-level PM2.5 concentrations from VIIRS AOD in China

While the aerosol optical depth (AOD) product from the Visible Infrared Imaging Suite (VIIRS) instrument has proven effective for estimating regional ground-level particle concentrations with aerodynamic diameters less than 2.5 μm (PM2.5), its performance at larger spatial scales remains unclear. Despite the wide application of statistical models in building ground-level PM2.5 satellite remote sensing retrieval models, a limited number of studies have considered the spatiotemporal heterogeneities for model structures. Taking China as the study area, we used the VIIRS AOD, together with multi-source auxiliary variables, to develop a spatially structured adaptive two-stage model to estimate ground-level PM2.5 concentrations at a 6-km spatial resolution. To this end, we first defined and calculated a dual distance from the ground-level PM2.5 monitoring data. We then applied the unweighted pair-group method with arithmetic means on dual distances and obtained 13 spatial clusters. Subsequently, we combined the time fixed effects regression (TEFR) model and geographically weighted regression (GWR) model to develop the spatially structured adaptive two-stage model. For each spatial cluster, we examined all possible combinations of auxiliary variables and determined the best model structure according to multiple statistical test results. Finally, we obtained the PM2.5 estimates through regression mapping. At least seven model-fitting data records per day made a good threshold that could best overcome the model overfitting induced by the second-stage GWR model at the minimum price of losing samples. The overall model fitting and ten-fold cross validation (CV) R2 were 0.82 and 0.60, respectively, under that threshold. Model performances among different spatial clusters differed to a certain extent. High-CV R2 values always exceeded 0.6 while low-CV R2 values less than 0.5 also existed. Both the size of the model-fitting data records and the extent of urban-industrial characteristics of spatial clusters accounted for these differences. The PM2.5 estimates agreed well with the PM2.5 observations with correlation coefficients all exceeding 0.5 at the monthly, seasonal, and annual scales. East of Hu’s line and north of the Yangtze River were characterized by high PM2.5 concentrations. This study contributes to the understanding of how well VIIRS AOD can retrieve ground-level PM2.5 concentrations at the national scale and strategies for building ground-level PM2.5 satellite remote sensing retrieval models.

Photonic crystal-enhanced fluorescence imaging immunoassay for cardiovascular disease biomarker screening with machine learning analysis

When myocardial walls experience stress due to cardiovascular diseases, like heart failure, hormone N-terminal pro-B-type natriuretic peptide (NT-proBNP) is secreted into the blood. Early detection of NT-proBNP can assist diagnosis of heart failure and enable early medical intervention. A simple, cost-effective detection technique such as the widely used fluorescence imaging immunoassay is yet to be developed to detect clinically relevant levels of NT-proBNP. In this work, we demonstrate photonic crystal-enhanced fluorescence imaging immunoassay using diatom biosilica, which is capable of detecting low levels of NT-proBNP in solution with the concentration range of 0˜100 pg/mL. By analyzing the fluorescence images in the spatial and spatial frequency domain with principle component analysis (PCA) and partial least squares regression (PLSR) algorithms, we create a predictive model that achieves great linearity with a validation R2 value of 0.86 and a predictive root mean square error of 14.47, allowing for good analyte quantification. To demonstrate the potential of the fluorescence immunoassay biosensor for clinical usage, we conducted qualitative screening of high and low concentrations of NT-proBNP in human plasma. A more advanced machine learning algorithm, the support vector machine classification, was paired with the PCA and trained by 160 fluorescence images. In the 40 testing images, we achieved excellent specificity of 93%, as well as decent accuracy and sensitivity of 78% and 65% respectively. Therefore, the photonic crystal-enhanced fluorescence imaging immunoassay reported in this article is feasible to screen clinically relevant levels of NT-proBNP in body fluid and evaluate the risk of heart failure.

Comparative analysis of index and chemometric techniques-based assessment of leaf area index (LAI) in wheat through field spectroradiometer, Landsat-8, Sentinel-2 and Hyperion bands

Successful retrieval of leaf area index (LAI) from hyperspectral remote sensing relies on the proper selection of indices or multivariate models. The objectives of the research work were to identify best vegetation index and multivariate model based on canopy reflectance and LAI measured at different growth stages of wheat. Comparison of existing indices revealed optimized soil-adjusted vegetation index (OSAVI) as the best index based on R2 of calibration, validation and root mean square error of validation. Proposed ratio index (RI; R670, R845) and normalized difference index (NDI; R670, R845) provided comparable performance with the existing vegetation indices (R2 = 0.65 and 0.62 for RI and NDI, respectively, during validation). Among the multivariate models, partial least squares regression (PLSR) model with Hyperion band configuration performed the best during validation (R2 = 0.80 and RMSE = 0.58 m2 m−2). Our results manifested the opportunities for developing biophysical products based on satellite sensors.

Estimate of winter-wheat above-ground biomass based on UAV ultrahigh-ground-resolution image textures and vegetation indices

When dealing with multiple growth stages, estimates of above-ground biomass (AGB) based on optical vegetation indices (VIs) are difficult for two reasons: (i) optical VIs saturate at medium-to-high canopy cover, and (ii) organs that grow vertically (e.g., biomass of reproductive organs and stems) are difficult to detect by canopy spectral VIs. Although several significant improvements have been made for estimating AGB by using narrow-band hyperspectral VIs, synthetic aperture radar, laser intensity direction and ranging, the crop surface model technique, and combinations thereof, applications of these new techniques have been limited by cost, availability, data-processing difficulties, and high dimensionality. The present study thus evaluates the use of ultrahigh-ground-resolution image textures, VIs, and combinations thereof to make multiple temporal estimates and maps of AGB covering three winter-wheat growth stages. The selected gray-tone spatial-dependence matrix-based image textures (e.g., variance, entropy, data range, homogeneity, second moment, dissimilarity, contrast, correlation) are calculated from 1-, 2-, 5-, 10-, 15-, 20-, 25-, and 30-cm-ground-resolution images acquired by using an inexpensive RGB sensor mounted on an unmanned aerial vehicle (UAV). Optical-VI data were obtained by using a ground spectrometer to analyze UAV-acquired RGB images. The accuracy of AGB estimates based on optical VIs varies, with validation R2: 0.59–0.78, root mean square error (RMSE): 1.22–1.59 t/ha, and mean absolute error (MAE): 1.03–1.27 t/ha. The most accurate AGB estimate was obtained by combining image textures and VIs, which gave R2: 0.89, MAE: 0.67 t/ha, and RMSE: 0.82 t/ha. The results show that (i) the eight selected textures from ultrahigh-ground-resolution images were significantly related to AGB, (ii) the combined use of image textures from 1- to 30-cm-ground-resolution images and VIs can improve the accuracy of AGB estimates as compared with using only optical VIs or image textures alone; and (iii) high AGB values from winter-wheat reproductive growth stages can be accurately estimated by using this method; (iv) high estimates of winter-wheat AGB (8–14 t/ha) using the proposed combined method (DIS1, SE30, B460, B560, B670, EVI2 using MSR) show a 22.63% (nRMSE) improvement compared with using only spectral VIs (LCI, NDVI using MSR), and a 21.24% (nRMSE) improvement compared with using only image textures (COR1, DIS1, SE30, EN30 using MSR). Thus, the combined use of image textures and VIs can help improve estimates of AGB under conditions of high canopy coverage.

Accurate Prediction of the Adsorption Capabilities of Synthetic Organic Contaminants by Single‐Walled Carbon Nanotubes

AbstractIn this work, quantitative structure‐property relationship (QSPR) was employed to predict the adsorption capabilities of 40 synthetic organic contaminants (SOCs) by single‐walled carbon nanotubes (SWCNTs). Total of 1081 molecular descriptors that consist of 102 constitutional and physicochemical properties, 342 topological properties, and 637 three‐dimensional properties were used to characterize the chemical information of the 40 SOCs. After removing redundant variables by stepwise multiple regression (SMR), an optimal partial least squares (PLS) model was established by using only 3 decriptors, i. e., BEHv, Dm, and Mor30e. The determinant R2 of the training samples, Leave‐one‐out cross validation R2 (Q2), and R2pred of test samples were 0.775, 0.675, and 0.713, respectively. Y‐random permutation test showed that the optimal PLS model is not caused by chance. Based on the optimal PLS model, it can be inferred that topological, steric, and electronic properties are dominant factors affecting the adsorption capabilities of the SOCs by SWCNTs.

Multi-Crop Green LAI Estimation with a New Simple Sentinel-2 LAI Index (SeLI).

The spatial quantification of green leaf area index (LAIgreen), the total green photosynthetically active leaf area per ground area, is a crucial biophysical variable for agroecosystem monitoring. The Sentinel-2 mission is with (1) a temporal resolution lower than a week, (2) a spatial resolution of up to 10 m, and (3) narrow bands in the red and red-edge region, a highly promising mission for agricultural monitoring. The aim of this work is to define an easy implementable LAIgreen index for the Sentinel-2 mission. Two large and independent multi-crop datasets of in situ collected LAIgreen measurements were used. Commonly used LAIgreen indices applied on the Sentinel-2 10 m × 10 m pixel resulted in a validation R2 lower than 0.6. By calculating all Sentinel-2 band combinations to identify high correlation and physical basis with LAIgreen, the new Sentinel-2 LAIgreen Index (SeLI) was defined. SeLI is a normalized index that uses the 705 nm and 865 nm centered bands, exploiting the red-edge region for low-saturating absorption sensitivity to photosynthetic vegetation. A R2 of 0.708 (root mean squared error (RMSE) = 0.67) and a R2 of 0.732 (RMSE = 0.69) were obtained with a linear fitting for the calibration and validation datasets, respectively, outperforming established indices. Sentinel-2 LAIgreen maps are presented.

Sediment Yield Modeling Using SWAT Model: Case of Changjiang River Basin

Soil erosion is posing a severe challenge to the productivity of land throughout the word, this alarming situation requires urgent interventions in order to preserve water and soil resources. This study presents an application of the model Soil and Water Assessment Tool(SWAT) to simulate daily and monthly water flow and sediment load from 1980 to 2012 in the Changjiang River basin, which is a sub-basin of the Poyang Lake basin, China. The sensitivity analysis is carried out by using the LH-OAT method. The model was calibrated and validated based on a comparison of simulated and measured runoff and sediment data of the Tankou and Dufengken hydrological station in Changjiang River. Indeed, for the monthly time step application, the relative error of the model to the runoff simulation was less than 10.21% in the calibration period and the validation period. The correlation coefficient(R2) and Nash-Suttcliffe(Ens) were higher than or equal to 0.91, and The relative error of the model to the sediment simulation was less than 26.0% in the calibration period and the validation period, R2 and Ens were higher than 0.76. The research indicates that SWAT model can be used to simulate the runoff and sediment, water loss and soil erosion are mainly caused by rainfall in flood season, and the sediment is up to 99.26% of the total year, and it has a good correlation among rainfall, runoff and sediment in flood period.

Incorporating satellite-derived data with annual and monthly land use regression models for estimating spatial distribution of air pollution

The purpose of this study was to assess the performance of annual and monthly land use regression (LUR) models for estimating the spatial distribution of NO2 and PM2.5 in Taiwan. Samples were collected at 73 air quality monitoring sites in 2015. Data transformation coupled with extracting principle components and satellite-derived data were integrated with LUR modeling and applied to increase PM2.5 model performance. Results indicated that NO2 exhibited more robust model performance compared with PM2.5. Leave-one-out cross validation (LOOCV) R2 of NO2 annual model was 0.76 and ranged from 0.56 to 0.81 for monthly models. The LOOCV R2 of PM2.5 annual model was improved from 0.13 to 0.56 by applying principle component analysis and adding satellite data (i.e., percentage of sunshine coverage and aerosol optical depth). These approaches also improved the performance of PM2.5 monthly models. The median LOOCV R2 increased from 0.12 to 0.49.

Data fusion strategy in quantitative analysis of spectroscopy relevant to olive oil adulteration

Olive oil adulteration with various less expensive edible oils represents a great danger for consumers. Spectrometry has been used to detect olive oil adulteration with other oil, but we need more robust and accurate model. Therefore, this work investigated the combination of infrared (NIR) and mid infrared (MIR) spectroscopy for the quantification of rapeseed oil in olive oil blends. Furthermore, a partial least squares (PLS) model was established to predict the concentration of the adulterant. Models constructed using baseline correction by combination of standard normal variate (SNV), SG smoothing and vector normalization pretreatments, respectively. Three data fusion strategies (low, mid and high-level) have been applied to take advantage of the synergistic effect of the information obtained from NIR and MIR. We chose algorithm (SPA) to extract spectral features for mid-level data fusion. Binary linear regression used in high-level data fusion. We selected the best pretreatment for final evaluation according to the evaluation parameters (R2 of calibration and validation, RMSECV and RMSEP). NIR, MIR and data fusion models were evaluated by comparing the R2 of validation and RMSEP (root mean square error of prediction). The RMSEP of low-level (3.44), high-level (2.86) data fusion were better than NIR(7.09), MIR(4.04), mid-level(6.09)and the validation coefficient of determination R2 of low-level data fusion (0.975) and high-level data fusion (0.988) are better than the NIR (0.896) and MIR (0.966). Results showed that：(1) NIR and MIR are fast and non-destructive testing tools to detect the extra-virgin olive oil adulteration with rapeseed oil. (2) Low-level data fusion can effectively improve model prediction accuracy. (3) SPA reduced the number of variables, but it did not improved the results. (4) High-level data fusion strategy can be used as a reliable tool for quantitative analysis.

Estimating Total Leaf Chlorophyll Content of Gannan Navel Orange Leaves Using Hyperspectral Data Based on Partial Least Squares Regression

The goal of this study was to model the total leaf chlorophyll content (LCCtot) of Gannan navel orange leaves using a field imaging spectroscopy system in the visible and near-infrared domain. The spectral range from 400 to 1000 nm with 176 wavebands (a wavelength interval of 3.41 nm) or 360 wavebands (a wavelength interval of 1.67 nm), labeled as “Datasets_1.67” and “Datasets_3.41”, respectively, were used. Although different spectral data types were used, better prediction results for LCCtot were based on Datasets_1.67 for LCCtot prediction. Several prediction models of LCCtot were built based on partial least squares regression (PLSR), artificial neural networks (ANN), ordinary least squares regression (OLSR), and stepwise linear regression (SLR) using full spectral and effective wavelength (EW) data (raw spectral (RS), first derivative spectral (FDS) and second derivative spectral (SDS) data). The determination coefficient (R 2 ), the root mean square error (RMSE) and the residual predictive deviation (RPD) were used to evaluate the reliability and accuracy of the predicted LCCtot values. As a result, 14 (7 obtained from Datasets_1.67, 7 obtained from Datasets_3.41), 39 (21 obtained from Datasets_1.67, 18 obtained from Datasets_3.41) and 50 (27 obtained from Datasets_1.67, 23 obtained from Datasets_3.41) wavebands were selected from the RS data, FDS data and SDS data, respectively, as the EWs for LCCtot prediction of navel orange leaves. After that, PLSR and ANN predictive models were established using full spectra, and OLSR and SLR predictive models were built using the selected EWs. The experimental results demonstrated that these various regression methods were useful for estimating LCCtot in the order of PLSR models established using full spectra from RS data (F-RS-PLSR) > PLSR models established using full spectra from SDS data (F-SDS-PLSR) > PLSR models established using full spectra from FDS data (F-FDS-PLSR) > SLR models established using EWs by RS data (EWs-RS-SLR). However, models built with ANN and OLSR, where the RPD values were less than 3, cause the models to be inaccurate. Finally, in comparison, the F-RS-PLSR model exhibited the best performance of LCCtot estimation; with the number of principal components (Pcs) = 5, this model provided high values of the R 2 of calibration (C-R 2 ) = 0.92 and the R 2 of validation (V-R 2 ) = 0.96, small values of the RMSE of calibration (C-RMSE)=0.05 mg/g and the RMSE of validation (V-RMSE) = 0.19 mg/g, and sufficient the RPD of calibration (C-RPD)=17.00 and the RPD of validation (V-RPD)=3.63 values. Overall, the best modeling method was PLSR. Hence, the PLSR applicability for assessing chlorophyll content in navel orange leaves was demonstrated.

Feedstock and inoculum characteristics and process parameters as predictors for methane yield in mesophilic solid-state anaerobic digestion

Abstract: In this study, several linear regression models were tested to predict the cumulative 30-day methane yield produced in mesophilic solid-state anaerobic digestion, employing diverse lignocellulosic biomass sources. Data collected from 13 studies were utilized, totalizing 86 experimental points, divided into regression and validation. Models containing higher order terms, the inverse of variables and interactions among all eleven input variables were tested. Simple linear models utilizing a single variable were unable to describe the methane production, giving an R² lower than 0.37. However, combinations of multiple variables and its inverses as only independent variable permitted an increase in simple linear models predictive capacity up to 63% of experimental variability. Higher order models presented an improvement in predictive quality: for a fourth-order multiple linear model, a validation R² of 0.8329 was achieved. In view of the obtained results, the proposed linear regression models consist in an attractive tool to propose experimental routines and to investigate new biomass sources for methane production using solid-state anaerobic digestion, significantly reducing time and cost requirements to experiments’ execution.

Local variation of PM2.5 and NO2 concentrations within metropolitan Beijing

Studies on long-term effects of air pollutants in China are limited, in part due to a lack of appropriate spatial exposure models. Elsewhere, Land Use Regression (LUR) models have been used extensively to estimate individual-level air pollutant exposure variation within urban areas by combining air pollution measurements with geospatial predictors. Here, LUR models were developed to estimate spatial variation in fine particulate matter (PM2.5) and nitrogen dioxide (NO2) concentrations during the heating and non-heating seasons in Beijing, China. Daily routine PM2.5 and NO2 measurements at 35 monitoring stations from October 2012 to December 2013 were collected as dependent variables, with 191 geospatial measures generated as potential independent variables. The PM2.5 LUR model included latitude and forest land (within a 5000m buffer) in the heating season; and latitude, residential land use (5000m) and main road length (100m) in the non-heating season. The NO2 LUR model included latitude, population density (5000m), water bodies (2000m) and road network factors in the heating season; and residential land use (5000m), water bodies (1000m) and road network factors in the non-heating season. Adjusted R2 and cross validation R2 values of all seasonal models were above 0.80, except for the heating season NO2 model (adjusted R2 = 0.87, cross validation R2 = 0.77). Different spatial patterns between NO2 and PM2.5 were evident in Beijing, with a more distinct local pattern for NO2 than PM2.5 and during the non-heating season, when regional pollution was less dominant. These 250m resolution LUR models can be used for health impact assessment or to estimate PM2.5 and NO2 concentrations at the residential addresses of participants in epidemiological cohorts of Beijing.

Spatiotemporally mapping of the relationship between NO2 pollution and urbanization for a megacity in Southwest China during 2005–2016

Nitrogen dioxide (NO2) significantly contributes to air pollution. Long-term NO2 exposure is harmful to human health. The NO2 pollution in China has surpassed developed countries and attracts international attention. To understand the spatial and temporal distributions of NO2 across Chengdu in Southwest China, a random forest (RF) model was developed based on NO2 environmental monitoring data, the Ozone Monitoring Instrument (OMI) satellite retrievals, and geographic covariates. The RF model showed good performance with a cross validation R2 of 0.77, and a root mean square error (RMSE) of 11.0 μg/m3. The ground-level NO2 concentrations of Chengdu for 2005–2016 were predicted using the developed model with the multiyear population weighted NO2 concentration being 41.7 ± 11.7 μg/m3. The predicted NO2 concentrations exhibited a clear seasonal variation trend with winter being the highest and summer being the lowest. Furthermore, higher NO2 concentrations in the downtown areas were observed than that in the rural areas indicating the former being attributed to more anthropogenic sources. The population weighted NO2 concentrations with deseasonlization were relatively high during 2011–2013. The NO2 concentration increased at a rate of 0.81 μg/m3/year before 2011 (43.4 ± 11.2 μg/m3) and decreased at a rate of −1.03 μg/m3/year after 2013 (44.8 ± 12.8 μg/m3).

Evaluation of random forest regression and multiple linear regression for predicting indoor fine particulate matter concentrations in a highly polluted city

BackgroundIndoor and outdoor fine particulate matter (PM2.5) are both leading risk factors for death and disease, but making indoor measurements is often infeasible for large study populations. MethodsWe developed models to predict indoor PM2.5 concentrations for pregnant women who were part of a randomized controlled trial of portable air cleaners in Ulaanbaatar, Mongolia. We used multiple linear regression (MLR) and random forest regression (RFR) to model indoor PM2.5 concentrations with 447 independent 7-day PM2.5 measurements and 87 potential predictor variables obtained from outdoor monitoring data, questionnaires, home assessments, and geographic data sets. We also developed blended models that combined the MLR and RFR approaches. All models were evaluated in a 10-fold cross-validation. ResultsThe predictors in the MLR model were season, outdoor PM2.5 concentration, the number of air cleaners deployed, and the density of gers (traditional felt-lined yurts) surrounding the apartments. MLR and RFR had similar performance in cross-validation (R2 = 50.2%, R2 = 48.9% respectively). The blended MLR model that included RFR predictions had the best performance (cross validation R2 = 81.5%). Intervention status alone explained only 6.0% of the variation in indoor PM2.5 concentrations. ConclusionsWe predicted a moderate amount of variation in indoor PM2.5 concentrations using easily obtained predictor variables and the models explained substantially more variation than intervention status alone. While RFR shows promise for modelling indoor concentrations, our results highlight the importance of out-of-sample validation when evaluating model performance. We also demonstrate the improved performance of blended MLR/RFR models in predicting indoor air pollution.

Application of fractional-order derivative in the quantitative estimation of soil organic matter content through visible and near-infrared spectroscopy

The spectral preprocessing method has become an integral component of soil analysis through visible and near-infrared (Vis-NIR) spectroscopy. Various spectral pretreatment techniques are applied to improve model accuracy by removing undesired side effects, such as spectral noise, baseline shift, and light scattering, and by accentuating spectral features. Conventional integer-order derivatives (i.e., first and second derivatives), which represent specific cases of fractional-order derivatives (FODs), may neglect some detailed spectral information related to target variables. The objective of this study was to compare the performance of FOD in the estimation of soil organic matter (SOM) with that of conventional first and second derivatives. A total of 258 soil samples (180 for calibration and 78 for validation) were collected from Jianghan Plain, Central China. The reflectance spectra and SOM concentrations of the samples were obtained in the laboratory. Two regression techniques, namely, partial least squares (PLS) and PLS–support vector machine (PLS–SVM), and eight FOD transformation processes for spectral data were combined and compared. Results indicated that as the derivative order increased, the details of the FOD spectra changed and the spectral resolution of reflectance curves improved; the intensity of the spectral signals, however, weakened. The correlation between SOM and FOD spectra was enhanced at some specific wavelengths (e.g., the absolute value of the best one-dimensional correlation coefficient of the 1.25-order derivative spectra could reach 0.65 but that of the original reflectance spectra was only 0.47). In most cases, the PLS–SVM models performed better in SOM estimation than the PLS models. The PLS–SVM model with the 1.25-order derivative spectra exhibited the best model performance and provided the validation R2 and ratio of performance to an interquartile range of 0.79 and 3.03, respectively. FOD offers greater advantages in balancing the spectral resolution and the magnitude of spectral strength than traditional integer-order derivatives. Furthermore, the FOD algorithm has strong application potential in soil Vis-NIR spectroscopy and other types of spectroscopy.

Development of a hydrology and water quality model for a large transboundary river watershed to investigate the impacts of climate change – A SWAT application

The aim of this study is to assess a different approach of SWAT model setup with enhanced Hydrologic Response Unit (HRU) definition procedure. The method uses a set of customizable MATLAB scripts (hence, SWAT-LAB) to produce HRUs from a combination of topographic, soil, landuse and administrative unit rasterized datasets. We used this approach to setup, run and calibrate a large-scale transboundary Nemunas River model, consisting of eleven sub-models. Nemunas River is the major contributory that discharges into the Curonian Lagoon, which is the largest European coastal lagoon. Belarus, Lithuania, Poland and the Russian Federation Kaliningrad Oblast share Nemunas River basin area. The basin is experiencing nutrient load from different sources in the riparian countries, nevertheless, the burden of improving the water quality of the river falls mainly on Lithuania. This article focuses on assessing the practicability of SWAT-LAB for creating a high-resolution large-scale transboundary hydrological and water quality model, where data availability is limited and fragmented. We demonstrate the model performance on a case study of one of the Nemunas River tributaries: Neris or Vilija River basin, which is situated largely in the Republic of Belarus. Model performance was evaluated graphically, using hydrographs and percent exceedance curves, and quantified using coefficient of determination (R2) and Nash–Sutcliffe efficiency (NS). We achieved good model performance for monthly (calibration R2 = 0.80, NS = 0.83; validation R2 = 0.80, NS = 0.76) and daily (calibration R2 = 0.66, NS = 0.66; validation R2 = 0.67, NS = 0.66) flow. Satisfactory performance results were achieved in modeling monthly loads of suspended sediments (calibration R2 = 0.4, NS = 0.44; validation R2 = 0.58, NS = 0.34), total phosphorus (calibration R2 = 0.71, NS = 0.61; validation R2 = 0.53, NS = 0.56) and total nitrogen (calibration R2 = 0.55, NS = 0.42; validation R2 = 0.48, NS = 0.4). Furthermore, we use the model to assess possible future nutrient loads that could be transported from Belarus to Lithuania under two climate change scenarios (RCP 4.5 and 8.5) and present the results.

Estimating lead and zinc concentrations in peri-urban agricultural soils through reflectance spectroscopy: Effects of fractional-order derivative and random forest

Heavy metal contamination of peri-urban agricultural soil is detrimental to soil environmental quality and human health. A rapid assessment of soil pollution status is fundamental for soil remediation. Heavy metals can be monitored by visible and near-infrared spectroscopy coupled with chemometric models. First and second derivatives are two commonly used spectral preprocessing methods for resolving overlapping peaks. However, these methods may lose the detailed spectral information of heavy metals. Here, we proposed a fractional-order derivative (FOD) algorithm for preprocessing reflectance spectra. A total of 170 soil samples were collected from a typical peri-urban agricultural area in Wuhan City, Hubei Province. The reflectance spectra and lead (Pb) and zinc (Zn) concentrations of the samples were obtained in the laboratory. Two calibration methods, namely, partial least square regression and random forest (RF), were used to establish the relation between the spectral data and the two heavy metals. In addition, we aimed to explore the use of spectral estimation mechanism to predict the Pb and Zn concentrations. Three model evaluation parameters, namely, coefficient of determination (R2), root mean squared error, and ratio of performance to inter-quartile range (RPIQ), were used. Overall, the spectral reflectance decreased with the increase in Pb and Zn contents. The FOD algorithm gradually removed spectral baseline drifts and overlapping peaks. However, the spectral strength slowly decreased with the increase in fractional order. High fractional-order spectra underwent more spectral noises than low fractional-order spectra. The optimal prediction accuracies were achieved by the 0.25- and 0.5-order reflectance RF models for Pb (validation R2 = 0.82, RPIQ = 2.49) and Zn (validation R2 = 0.83, RPIQ = 2.93), respectively. A spectral detection of Pb and Zn mainly relied on their covariation with soil organic matter, followed by Fe. In summary, our results provided theoretical bases for the rapid investigation of Pb and Zn pollution areas in peri-urban agricultural soils.

AMSR2 Soil Moisture Downscaling Using Temperature and Vegetation Data

Soil moisture (SM) applications in terrestrial hydrology require higher spatial resolution soil moisture products than those provided by passive microwave remote sensing instruments (grid resolution of 9 km or larger). In this investigation, an innovative algorithm that uses visible/infrared remote sensing observations to downscale Advanced Microwave Scanning Radiometer 2 (AMSR2) coarse spatial resolution SM products was developed and implemented for use with data provided by the Advanced Microwave Scanning Radiometer 2 (AMSR2). The method is based on using the Normalized Difference Vegetation Index (NDVI) modulated relationships between day/night SM and temperature change at corresponding times. Land surface model output variables from the North America Land Data Assimilation System (NLDAS), remote sensing data from the Moderate-Resolution Imaging Spectroradiometer (MODIS), and Advanced Very High Resolution Radiometer (AVHRR) were used in this methodology. The functional relationships developed using NLDAS data at a grid resolution of 12.5 km were applied to downscale AMSR2 JAXA (Japan Aerospace Exploration Agency) SM product (25 km) using MODIS land surface temperature (LST) and NDVI observations (1 km) to produce the 1 km SM estimates. The downscaled SM estimates were validated by comparing them with ISMN (International Soil Moisture Network) in situ SM in the Black Bear–Red Rock watershed, central Oklahoma between 2015–2017. The overall statistical variables of the downscaled AMSR2 SM validation R2, slope, RMSE and bias, demonstrate good accuracy. The downscaled SM better characterized the spatial and temporal variability of SM at watershed scales than the original SM product.