Validation R2 Research Articles

Estimating the Daily NO2 Concentration with High Spatial Resolution in the Beijing–Tianjin–Hebei Region Using an Ensemble Learning Model

Nitrogen dioxide (NO2) is an important pollutant related to human activities, which has short-term and long-term effects on human health. An ensemble learning model was constructed and applied to estimate daily NO2 concentrations in the Beijing–Tianjin–Hebei region between 2010 and 2016. A variety of predictive variables included satellite-based troposphere NO2 vertical column concentration, meteorology, elevation, gross domestic product (GDP), population, land-use variables, and road network. The ensemble learning model achieved two things: a 0.01° × 0.01° grid resolution and the estimation of historical data for the years 2010–2013. The ensemble model showed good performance, whereby the R2 of tenfold cross-validation was 0.72 and the R2 of test validation was 0.71. Meteorological hysteretic effects were incorporated into the model, where the one-day lagged boundary layer height contributed the most. The annual NO2 estimation showed little change from 2010 to 2016. The seasonal NO2 estimation from highest to lowest occurred in winter, autumn, spring, and summer. In the annual maps and seasonal maps, the NO2 estimations in the northwest region were lower than those in the southeast region, and there was a heavily polluted band in the south of the Taihang Mountains. In coastal areas, the annual NO2 estimations were higher than the NO2 monitored values. The drawback of the model is underestimation at high values and overestimation at low values. This study indicates that the ensemble learning model has excellent performance in the simulation of NO2 with high spatial and temporal resolution. Furthermore, the research framework in this study can be a generally applied for drawing implications for other regions, especially for other cities in China.

Characterization of field-scale soil variation using a stepwise multi-sensor fusion approach and a cost-benefit analysis

The potential of a stepwise fusion of proximally sensed portable X-ray fluorescence (pXRF) spectra and electromagnetic induction (EMI) with remote Sentinel-2 bands and a digital elevation model (DEM) was investigated for predicting soil physicochemical properties in pedons and across a heterogeneous 80-ha crop field in Wisconsin, USA. We found that pXRF spectra with partial least squares regression (PLSR) models can predict sand, total nitrogen (TN), organic carbon (OC), silt contents, and clay with validation R2 of 0.81, 0.74, 0.73, 0.68, and 0.64 at the pedon scale but performed less well for soil pH (R2 = 0.51). A combination of EMI, Sentinel-2, and DEM data showed promise in mapping sand, silt contents, and TN at two depths and Ap horizon thickness and soil depth across the field. A clustering analysis using combinations of mapped soil properties or proximal and remote sensing data suggested that data fusion improved the characterization of field-scale variability of soil properties. The cost-benefit analysis showed that the most accurate management zones (MZs) for topsoil can be generated only using estimated soil property maps while it was the most costly as compared to other data sources. For an intermediate-high (for topsoil) and high (subsoil) accuracy and a moderate economic budget, the combination of sensors (proximal + remote sensing + DEM) might be a better approach for effective MZs generation than collecting soil samples for laboratory analysis while the latter produced the most accurate maps for topsoil. It can be concluded that pXRF spectra can be useful for predicting key soil properties (e.g., sand, TN, OC, silt, clay) at different soil depths, and a combination of proximal and remote sensing provides an effective way to delineate soil MZs that are useful for decision-making.

Machine learning for design of thin-film nanocomposite membranes

In this study, a novel machine learning approach is proposed for estimation of the permeate flux and foulant rejection in nanocomposite filtration membranes. Nine independent variables are fed to artificial neural networks (ANNs) including support, nanoparticles concentration, concentration of organic phase trimesoyl chloride (TMC) in-n-hexane (TMC in n-hexane), operation pressure, contact angle, thin layer thickness, location of the nanoparticles (NPs), post-treatment temperature and duration, and permeate flux and foulant rejection were derived as the outputs of the ANNs. The proposed method was evaluated on two datasets across training, validation and test datasets, and an unseen dataset. 2250 different initial weights and number of the neurons in the hidden layer for the proposed ANN models were considered and compared to find the optimized ANN models. The mean squared error (MSE) and coefficient of determination (R2) were employed to select the best 20 ANN models for further analysis. The proposed ANN models resulted in accurate estimates for both permeate flux and foulant rejection with R2 of 0.9958 and 0.9412 in all data included in the training, validation and test datasets and R2 of 0.9938 and 0.9811 in unseen dataset, respectively. In addition, results of sensitivity analysis revealed that post treatment temperature and contact angle were found the most important input variables for estimation of permeate flux and foulant rejection. The proposed method can provide valuable insights for formulating permeate flux and foulant rejection and considering the effects of each experimental condition on nanocomposite filtration membranes without doing real experiments, which is time-consuming and expensive.

Planting time optimization on tidal agricultural land

Tidal agricultural land (TAL) is a marginal farming field that is highly available but yet to be optimized in Indonesia. Productivity enhancement of TAL is achievable through agriculture technologies, but water management is the key solution to TAL main problems, i.e. flood and drought cycle. The main objective of this research was to predict tidal cycle for available planting time optimization in TAL. Research was conducted on 100 hectares TAL in Jejangkit, Barito Kuala, South Kalimantan, with the Alalak river become the main subject of the hydrodynamics monitoring. Watershed water balance simulation model run on 10-day interval was utilized to predict the tidal cycle with 4 input parameters including water-level, precipitation, percolation, and evapotranspiration. The prediction result showed that the available planting months in Jejangkit TAL was about 8 months, between the first or second 10-day of May and third 10-day of January or first 10-day of February (calibration and validation R2 were 0.78 and 0.92, respectively). On-field validation in Jejangkit recorded flood event started during the third 10-day of January 2020, which indicated that the prediction was reliable. This result shows that tidal cycle could be simulated using watershed water balance model to assist farmers on TAL’s productivity enhancement.

Determination of ethyl octanoate in Chinese liquor using FT-NIR spectroscopy

To quantitatively detect ethyl octanoate in Chinese liquor, Fourier-transform near-infrared (FT-NIR) spectroscopy was performed in the present work, with 162 Chinese liquor samples selected from Luoyang Dukang Distillery. The chemical values of ethyl octanoate were determined by gas chromatography (GC), and spectral data from 12,000 to 4000 cm-1 were collected. The calibration model was established with partial least squares (PLS) regression, and then validated using internal cross-validation. The predictability of the model was further confirmed by the validation set as external validation. After comparing the effects of the models set up with sample data under different pre-processing methods, the model was built within the spectral region of 6101.7 - 5449.8 cm-1; based on the SNV pre-processing method which was selected as the optimal model. The coefficient of determination (R2) for cross-validation of the model was 0.9507, and the corresponding root mean square errors of cross-validation (RMSECV) was 3.91 mg L-1. The R2 for external validation was 0.9537, and the root mean square errors of prediction (RMSEP) was 3.62 mg L-1. The results demonstrated that using NIR spectroscopy to determine ethyl octanoate in Chinese liquor is feasible and can achieve satisfactory results.

Hyperspectral imagery to monitor crop nutrient status within and across growing seasons

Imaging spectroscopy provides the opportunity to monitor nutrient status of vegetation. In crops, prior studies have generally been limited in scope, either to a small wavelength range (e.g., 400–1300 nm), a small number of crop cultivars, a single growth stage or single growing season. Methods that are not time- or site-specific are needed to use imaging spectroscopy for routine monitoring of crop status. Using data from four cultivars of potatoes (Solanum tuberosum L.), three growth stages and two growing seasons, we demonstrate the capacity of full-range (400–2350 nm) imaging spectroscopy to quantify nutrient status (petiole nitrate, whole leaf and vine total nitrogen) and predict tuber yield in potatoes across cultivars, growth stages and growing seasons. We specifically tested the capabilities of: (1) ordinary least-squares regression (OLSR) using traditional hyperspectral vegetation indices (VIs); (2) partial least-squares regression (PLSR) using full spectrum (400–2350 nm), VNIR- (visible-to-near infrared: 400–1300 nm) or SWIR-only (shortwave infrared: 1400–2350 nm) wavelengths; (3) predictive models developed for one potato type or planting season on withheld data from a different type or season. Our results show that OLSR models produced poor predictions with data from all dates pooled together (validation R2 < 0.01). Single-date OLSR models performed better (R2 = 0.20–0.60, relative RMSE = 15–30%). PLSR models performed well and were comparable using different spectral regions (full-spectrum, VNIR-only and SWIR-only), with validation R2 = 0.68–0.82 and RRMSE = 12–25%. Testing across potato types, models produced reliable predictions (R2 = 0.45–0.75, RRMSE = 13–30%), but with some bias. Cross-season models had validation R2 = 0.46–0.75 and RRMSE = 17–100%, with a more significant bias than the cross-potato type models. To achieve models that are generalizable and robust, we recommend: (1) obtaining ground measurements that capture the full range of plant growth conditions and developmental stages, and (2) ensuring that image processing approaches minimize spectral discrepancies among dates.

A comparison of regionalization methods in monsoon dominated tropical river basins

AbstractThe present study evaluated five regionalization methods: global averaging; regression; spatial proximity; behavioral similarity and artificial neural network (ANN) for Soil and Water Assessment Tool (SWAT), using data from 24 river basins in monsoon dominated tropical river basins of peninsular India. Regionalization was performed for each basin using the remaining 23 basins. The performance of the calibration and thus the regionalization method is limited by the unreliable or erroneous data at the basins. Overall, we found that the regression method outperforms other regionalization methods in terms of predicting the daily as well as peak discharges. It was found that despite showing a better R2 in training, testing and validation, the ANN method performed poorly probably due to a lower number of training data. Therefore, it is suggested that the ANN should be avoided for regionalization in the absence of sufficient training data. Moreover, the regression equations developed in the present study can be utilized to predict SWAT parameters of basins located in the vicinity of the study area. However, the basins located far away from the group of catchments or having diverse characteristics should be avoided for regionalization.

Hydrological Response of Land Use Land Cover Change on Stream Flow of Anger Watershed, Blue Nile Basin, Ethiopia

Stream flow of Anger watershed (7,982 km2), located in Western Oromia Regional State, Ethiopia was simulated by using soil and water assessment tool (SWAT) model. Assessment of layer staking, mosaicking, image classification and accuracy was carried by ArcGIS. The main target of this study was to determine the SWAT Model applicability and impact of past land-use land-cover change on stream flow of the Anger watershed. The time series data from 1990 to 1999 was used for model parameter calibration, and data from 2000 to 2004 were used to validate the model using the input parameter set. The performance of the SWAT model was evaluated by both R2 time series plots and ENS the statistical measures. The results of the model during calibration for monthly stream flow was R2= 0.90, ENS= 0.76 and during validation R2= 0.84, ENS= 0.63 respectively. Validation of image classification is achieved through Mosaicking image and Google Earth image. The performance of the Arc GIS model was evaluated by using the overall accuracy and kapa coefficient. The agreement between classified land use image and ground truth of the year 1986, 2000 and 2010 were showed a good agreement with result of overall accuracy=88%, 85% and 87% and kapa coefficient=81%, 79% and 82% respectively. Simulation of steam flow of the watershed under the impacts of past land-use land-cover changes can be determined by using SWAT model as a tool for water resources planning and management in this watershed.

Age estimation of barramundi (

The age of whole otoliths from barramundi (Lates calcarifer) obtained from the southern Gulf of Carpentaria were estimated using Fourier transform near-infrared (FT-NIR) spectroscopy. Otoliths from 1716 barramundi collected in 2006, 2009 and 2012–2015 were used in this study. Partial least-squares regression models (PLS-R) and multiple linear regression models (MLR) were developed from the diffuse reflectance spectra and the age was obtained from traditional sectioned otoliths. Calibration models were built up over consecutive years (2012–2015) by using a subset of the samples and used to predict the age of the remaining samples and samples from the following year. Results suggest that when seasonal (temporal) variability is incorporated into the calibration model, FT-NIR has the ability to predict barramundi age (validation R2 ranged from 0.73 to 0.78; RMSEP ranged from 6.92 to 7.64 months). The predicted age class was within 1 year of the reference age in over 96% of the samples. These models were also able to predict the age of otoliths from 2006 and 2009, which were retrieved from long-term storage (validation R2 ranged from 0.77 to 0.84; RMSEP ranged from 8.66 to 10.88 months). The results from this study have shown the potential for barramundi from the southern Gulf of Carpentaria to be aged quickly and accurately by using FT-NIR.

Thermal vision of oil palm fruits under difference ripeness quality

Indonesia is the primary producer of palm oil. The edible oil in the palm fruits mesocarp obtained through mechanical extraction, where the quantity and quality of oil regulated by the fruit ripeness upon harvest. When oil forms and accumulate in the mesocarp, it replaces the moisture until the optimum ripeness reached before oil started to deteriorate. However, due to its nature, harvesting the fruits at optimum ripeness is challenging. Since water and oil have different thermal properties, in this study, we developed a Thermal-Vision system to observe the thermal properties of the fruits before harvest. Five harvest windows selected, namely 110-130, 131-150, 151-170, 171-190, and 191-200 days after anthesis (DAA). The recorded images then processed to determine the surface temperature of each fruit. The oil obtained from fruits as quality parameter evaluated. Additionally, Moisture Content (MC) of fruits mesocarp measured. Models were developed using the Multilayer Perceptron Artificial Neural Network algorithm to correlate fruits' thermal properties with measured parameters (Oil Content). The coefficient determination (R2) of FFB ripeness with the OC of 0.9058 and OC with temperature of 0.8039.The models successfully predict with R2 upon calibration was 0.7818 with SEC of 0.0831. While upon validation R2 value was 0.9535 and SEP of 0.0003.

Route Duration Prediction in a Stochastic and Dynamic Vehicle Routing Problem with Short Delivery Deadlines

Abstract We are facing a real-world vehicle routing problem where orders arrive dynamically over the day at an online store and have to be delivered within short time. Stochastic information in form of the expected number and weight of orders and the traffic congestion level is available upfront. The goal is to predict the average time needed to deliver an order for a given time and day. This information is desirable for both routing decisions in the short horizon and planning vehicle drivers’ shifts with just the right capacity prior to the actual day. We compare a white box linear regression model and a neural network based black box model on historic route data collected over three months. We employ a hourly data aggregation approach with sampling statistics to estimate the ground truth and features. The weighted mean square error is used as loss function to favor samples with less uncertainty. A mean validation R2 score over 10 x 5-fold cross-validations of 0.53 indicates a substantial amount of unexplained variance. Both predictors are slightly optimistic and produce median standardized absolute residuals of about one.

Non destructive evaluation quality of oil palm fresh fruit bunch (FFB) (Elaeis guineensis Jack) based on optical properties using artificial neural network (ANN)

Oil palm (Elaeis guineensis Jacq) is one of the oil-producing plants with high productivity. The quality of palm fresh fruit bunches (FFB) is affected by harvesting activities. Generally, the method of harvesting oil palm by visual observation (maturity fraction) and harvest rotation systems. This study evaluated the quality of oil palm FFB non-destructively based on optical properties. Oil palm Fresh Fruit Bunches harvested in 5 ripeness ranges (110-130 DAA, 131-150 DAA, 151-170 DAA, 171-190 DAA, and 191-200 DAA). The oil palm FFB image recorded using a mobile camera with a minimum resolution of 25 megapixels, and then the image was processed using a digital image processing program. Furthermore, the quality parameters of oil palm FFB tested, and data were analyzed using SPSS Statistics 20.0 with the Artificial Neural Network (ANN) method. The model’s R2 upon calibration was 0.6934. While upon validation R2 value was 0.7211. The model was considered appropriate since R2 value both in calibration and validation were high.

On-site visible–near IR prediction of iodine number and fatty acid composition of subcutaneous fat of raw hams as phenotypes for a heavy pig breeding program

The quality of subcutaneous fat of raw hams is a trait of interest in selective breeding programs for pig lines used in dry-cured ham production, and rapid, non-invasive methods for its assessment are available. However, the efficacy of such methods to provide indicator traits for breeding programs needs to be proven. The study investigated the accuracy of on-site visible–near IR spectroscopy predictions of iodine number and fatty acid (FA) composition of raw ham subcutaneous fat, and it evaluated their effectiveness as indicator traits of ham fat quality in a pig breeding program. Prediction equations were developed using visible–near IR spectra acquired at the slaughterhouse from five sites in subcutaneous fat of raw hams of 1025 crossbred pigs. Pigs were raised, under standardized rearing and feeding conditions, in the sib-testing program of the Goland C21 boar line and slaughtered at nine months of age and average body weight of 166 ± 15 kg. Accuracy was generally relatively poor, but R2 in external validation was >0.7 for iodine number and concentration of C18:2n-6, polyunsaturated FAs and omega-6 FAs. To assess the effectiveness of the on-site predictions as indicator traits in a breeding program, (co)variance components of the measured traits (OBS) and of their predictions using in-lab (in-lab-PR) or on-site (on-site-PR) spectrometers were estimated. Available records for OBS were 6814 and 2048, for iodine number and FA composition, respectively. Predictions using in-lab were available for pigs slaughtered between 2006 and 2014, for a total of 10 153 records. Predictions using on-site were obtained from spectra collected since 2011, for a total of 10 296 records. The estimated heritabilities for the investigated traits ranged from 0.34 to 0.50 and were greater for on-site-PR than for OBS. Genetic correlations between OBS and in-lab-PR were very close to 1.00 for all the investigated traits, whereas those between OBS and on-site-PRED ranged from 0.86 to 0.94. On-site visible-IR predictions are accurate enough to support the use of this technique for large-scale phenotyping of raw ham fat quality, even when dealing with animals of a single genetic line raised in standardized conditions, and may be implemented as indicator traits in breeding programs.

Estimating hourly PM2.5 concentrations in Beijing with satellite aerosol optical depth and a random forest approach

Assessing short-term exposure to PM2.5 requires the concentration distribution at a high spatiotemporal resolution. Abundant researches have derived the daily predictions of fine particles, but estimating hourly PM2.5 is still a challenge restrained by the input data. The recent aerosol optical depth (AOD) product from Himawari-8 provides hourly satellite observations informative to modelling. In this study, we developed separate random forest models with and without AOD and combined the estimates to obtain a full-coverage hourly PM2.5 distribution. 10-fold cross validation R2 ranged from 0.92 to 0.95 and root mean square errors from 14.1 to 16.9 μg/m3, indicating the good model performance. Spatial convolutional layers of PM2.5 measurements and temporal accumulation effects of meteorological features were added into the model. They turned out to be of the most important predictors and improved the performance significantly. Finally, we mapped hourly PM2.5 at a 1-km resolution in Beijing during a pollution episode in 2019 and studied the pollution pattern. The study proposed a method to obtain 24-h full-coverage hourly PM2.5 estimates which are useful for acute exposure assessment in epidemiological researches.

Texture evaluation of cooked parboiled rice using nondestructive milled whole grain near infrared spectroscopy

One consumer acceptability criterion of cooked parboiled rice is its texture, particularly hardness and toughness. The samples were obtained from parboiled rice factory for export. The hardness and toughness calibration models based on milled whole grain near infrared spectroscopy was developed. The ISO 11747 Rice-Determination of Rice Kernel Resistance to Extrusion after Cooking method was used as reference test. The models were established using partial least squares regression (PLSR), principal component regression (PCR) and support vector machine regression (SVM). The PLSR optimal calibration model of hardness with moving average smoothing pre-processing gave coefficient of determination of validation (r2), root mean square error of prediction (RMSEP) and ratio of prediction to deviation (RPD) of 0.70, 7.24 N and 1.93, respectively. The PCR optimal model of toughness using mean normalization preprocessing provided r2, RMSEP and RPD of 0.66, 38.00 Nmm and 1.75, respectively. According to RPD threshold, the models were fair for prediction application. This feasibility study suggested that the NIR protocol developed was applicable for real use due to the error of the NIR scanning and other unexplained errors was only 5% and 1% for hardness and toughness models, respectively. However, the sample preparation before texture analysis has to be improved.

The use of FTIR spectroscopy in combination with chemometrics for the authentication of milk fat from palm oil

This research was aimed to develop Fourier transform infrared spectroscopy (FTIR) combined with chemometrics of linear discriminant analysis (LDA), partial least square (PLS), and principal component regression (PCR) for authentication of milk fat from palm oil adulterant. FTIR spectroscopy and LDA have been successfully used to detect the presence of palm oil in MF. All the adulterated samples were clearly separated with authentic MF shown by the Cooman’s plot. Chemometrics of PLS at the wavenumber of 3033-692 cm-1 using first derivative spectra was successfully applied for the quantification of palm oil in MF. The suitability of the model was presented by its high R2 value both for calibration and validation models, accounting for 1 and 0.9998 respectively and its lower RMSEC (root mean square error of calibration) and RMSEP (root mean square error of prediction) value, accounting for 0.154 and 0.743 respectively. Quantification of palm oil was also successfully performed using chemometrics of PCR. The model showed high R2 in both calibration (0.9998) and validation (0.9997) values with lower RMSEC (0.671) and RMSEP (0.905) values. It can be concluded that a combination of FTIR spectroscopy with chemometrics could be used for the authentication of milk fat adulteration.

PROSPECT-PRO for estimating content of nitrogen-containing leaf proteins and other carbon-based constituents

Models of radiative transfer (RT) are important tools for remote sensing of vegetation, allowing for forward simulations of remotely sensed data as well as inverse estimation of biophysical and biochemical traits from vegetation optical properties. Estimation of foliar protein content is the key to monitor the nitrogen cycle in terrestrial ecosystems, in particular to assess the photosynthetic capacity of plants and to improve nitrogen management in agriculture. However, until now physically based leaf RT models have not allowed for proper spectral decomposition and estimation of leaf dry matter as nitrogen-based proteins and other carbon-based constituents (CBC) from optical properties of fresh and dry foliage. Such an achievement is the key for subsequent upscaling to canopy level and for development of new Earth observation applications.Therefore, we developed a new version of the PROSPECT model, named PROSPECT-PRO, which separates the nitrogen-based constituents (proteins) from CBC (including cellulose, lignin, hemicellulose, starch and sugars). PROSPECT-PRO was calibrated and validated on subsets of the LOPEX dataset, accounting for both fresh and dry broadleaf and grass samples. We applied an iterative model inversion optimization algorithm and identified the optimal spectral ranges for retrieval of proteins and CBC. When combining leaf reflectance and transmittance within the selected optimal spectral domains, PROSPECT-PRO inversions revealed similarly accurate CBC estimates of fresh and dry leaf samples (respective validation R2 = 0.96 and 0.95, NRMSE = 9.6% and 13.4%), whereas a better performance was obtained for fresh than for dry leaves when estimating proteins (respective validation R2 = 0.79 and 0.57, NRMSE = 15.1% and 26.1%). The accurate estimation of leaf constituents for fresh samples is attributed to the optimal spectral feature selection procedure.We further tested the ability of PROSPECT-PRO to estimate leaf mass per area (LMA) as the sum of proteins and CBC using independent datasets acquired for numerous plant species. Results showed that both PROSPECT-PRO and PROSPECT-D inversions were able to produce comparable LMA estimates across an independent dataset gathering 1685 leaf samples (validation R2 = 0.90 and NRMSE = 16.5% for PROSPECT-PRO, and R2 = 0.90 and NRMSE = 18.3% for PROSPECT-D). Findings also revealed that PROSPECT-PRO is capable of assessing the carbon-to‑nitrogen ratio based on the retrieved CBC-to-proteins ratio (R2 = 0.87 and NRMSE = 15.7% for fresh leaves, and R2 = 0.65 and NRMSE = 28.1% for dry leaves). The performance assessment of newly designed PROSPECT-PRO demonstrates a promising potential for its involvement in precision agriculture and ecological applications aiming at estimation of leaf carbon and nitrogen contents from observations of current and forthcoming airborne and satellite imaging spectroscopy sensors.

Evaluation of physiological properties and texture traits of durian pulp using near-infrared spectra of the pulp and intact fruit

This study aimed to investigate the feasibility of non-destructively predicting physiological properties (color, dry matter, and soluble solids) and texture properties (initial firmness, average firmness, rupture force, rupture distance, toughness, average penetrating force, and penetrating energy) of ‘Monthong’ durian using Fourier transform near-infrared spectroscopy of the pulp of the largest locule, the intact fruit at the largest locule, and the stylar end of the intact fruit. Based on partial least squares regression modeling, the internal quality evaluation of durian obtained by scanning the pulp could provide rough screening capability, with coefficient of determination of validation (r2), root mean square error of prediction (RMSEP), and the ratio of standard error of validation to standard deviation (RPD) values for the dry matter content, average penetrating force, and rupture force of 0.89, 3.60%, and 3.27; 0.73, 5.53 N, and 1.95; and 0.74, 6.15 N, and 1.96, respectively. Only the dry matter content of the pulp could be reasonably predicted based on scanning the intact durian fruit at the largest locule, with r2, RMSEP, and RPD values of 0.79, 5.23%, and 2.18, respectively. This finding could be applied at the first stage of trade between durian agriculturalists and exporters to prevent the exportation of immature durian and would be helpful for the industries producing frozen durian pulp and intact durian fruit for export.

Rapid soil fertility prediction using X-ray fluorescence data and machine learning algorithms

Conventional soil fertility analyses are laborious, costly, time-consuming, and produce hazardous waste. The high demand of these laboratory-based analyses prompted us to investigate an environment-friendly, rapid, and inexpensive methodology for soil fertility assessment. Portable X-ray fluorescence (pXRF) spectrometry allows the determination of total elemental concentration in soils quickly, simply and without hazardous waste production. However, incipient usage of this technology for the prediction of soil fertility properties has been reported for tropical conditions. Soil samples were collected from seven Brazilian states (n = 1975) aiming to use pXRF data to predict contents of available or exchangeable Ca2+, Mg2+, Al3+, K+ and P by testing different algorithms using 70% of the samples for model training, and the remaining 30% for model validation. In addition to point data predictions, the best performing models were applied to data obtained from a farm within the studied regions with a known cropping history to create soil fertility maps and illustrate another applicability of this approach. The attested use of pXRF data and machine learning algorithms stepwise Generalized Linear Model (GLM) and Random Forest (RF) to predict the contents of relevant soil fertility properties exhibited great potential. Validation of the models corroborated that RF resulted in more accurate predictions than GLM. Validation R2 values ranged from 0.59 to 0.82. Maps created were coherent with expected distributions of soil fertility attributes. This environment-friendly methodology may be used for the assessment of soil fertility properties in a wide range of tropical and subtropical soils with minimum waste generation and reduced costs.

English

Accurate quantification of heavy metals is essential for optimizing remedial efforts at polluted waterfronts. Although surface water quality models can be helpful tools for rapid quantification of heavy metals in streams, they cannot be extrapolated to other water bodies. This study estimated Mg, Cd, and Ni levels in one of the polluted urban waterfronts in the Niger Delta Province, Nigeria (5.317&deg;N, 6.467&deg;E) using one-dimensional transport model. To do this, water samples were collected along the waterfront in both dry and rainy seasons and analyzed for the selected heavy metals using a SHIMADZU&reg; AA6800 Atomic Absorption Spectrophotometer according to international standards. Model parameters were determined by simple linear regression using the experimental dataset. Results showed that the average concentrations of Mg (5.964 mg/L), Cd (0.088 mg/L), and Ni (0.071 mg/L) in the waterfront in both dry and rainy seasons exceeded both local and international regulatory values. The seasonal variation of the heavy metals in the waterfront showed that the concentrations of Mg and Cd were higher during the dry season. Model validation of experimental data showed reasonable prediction precision with root mean square error (RMSE) of 0.0016-0.2254 mg/L in the dry season, RMSE of 0.0026-0.3259 mg/L in the rainy season, and corresponding average validation r2 of 0.985 in both seasons. These results suggest that the waterfront is laden with heavy metals and the one-dimensional transport model is a useful tool to rapidly estimate the levels of Mg, Cd, and Ni in the waterfront for pollution control. Key words: Environmental monitoring, heavy metals, water pollution, water quality modelling. &nbsp