Coefficient Of Determination In Validation Research Articles

Exploration of Global and Specialized Near-Infrared Calibrations for the Quantification of Nutritional Content in Grapevine Organs, Berry Phenological Stages, and Shoot Lignification.

Current infrared spectroscopy applications in the field of viticulture are moving toward direct in-field measuring techniques. However, limited research is available on quantitative applications using direct measurement of fresh tissue. The few studies conducted have combined the spectral data from various cultivars, growing regions, grapevine organs, and phenological stages during model development. The spectral data from these heterogeneous samples are combined into a single data set and analyzed jointly during quantitative analysis. Combining the spectral information of these diverse samples into a global data set could be an unsuitable approach and could yield less accurate prediction results. Spectral differences among samples could be overlooked during model development and quantitative analysis. The development of specialized calibrations should be considered and could lead to more accurate quantitative analyses. This study explored a model optimization strategy attempting global and specialized calibrations. Global calibrations, containing data from multiple organs, berry phenological, and shoot lignification stages, were compared to specialized calibrations per organ or stage. The global calibration for organs contained data from shoots, leaves, and berries and produced moderately accurate prediction results for nitrogen, carbon, and hydrogen. The specialized calibrations per organ yielded more accurate calibrations with a coefficient of determination in validation (R2val) at 90.65% and a root mean square error of prediction (RMSEP) at 0.32% dry matter (DM) for the berries' carbon calibrations. The leaves and shoots carbon calibrations had R2val and RMSEP at 84.99%, 0.34% DM, and 90.06%, 0.37% DM, respectively. The specialized calibrations for nitrogen and hydrogen showed similar improvements in prediction accuracy per organ. Specialized calibrations per phenological and lignification stage were also explored. Not all stages showed improvement, however, most stages had comparable or improved results for the specialized calibrations compared to the global calibrations containing all phenological or lignification stages. The results indicated that both global and specialized calibrations should be considered during model development to optimize prediction accuracy.

Prediction of physical attributes in fresh grapevine (Vitis vinifera L.) organs using infrared spectroscopy and chemometrics

Spectra obtained from fresh grapevine organs provide information on chemical composition but could also contain valuable information on the morphological and physical attributes. The prediction of grapevine organs physical attributes using infrared spectroscopy is explored for the first time in this study. Near infrared spectroscopy (NIR) using a solid probe (NIR-SP) and a rotating integrating sphere (NIR-RS) and mid infrared (MIR) were used to obtain spectra from fresh and intact grapevine shoots, leaves, and berries. Linear partial least squares (PLS) and non-linear least absolute shrinkage and selection operator (LASSO), and extreme gradient boost (XGBoost) were implemented to predict relevant physical attributes in grapevine organs. NIR-RS using XGBoost showed coefficients of determination in validation (R2val) of 91.01% and root mean square error of prediction (RMSEP) of 0.71 mm (6.80%) for berry diameter. Shoot diameter was predicted at R2val of 62.08% and RMSEP at 0.82 mm (12.75%) using NIR-RS with LASSO regression. Monitoring these attributes throughout the growing season can lead to important viticultural information on grapevine yield, growth, and health.

Predicting leaf nitrogen content in wolfberry trees by hyperspectral transformation and machine learning for precision agriculture.

Leaf nitrogen content (LNC) is an important indicator for scientific diagnosis of the nutrition status of crops. It plays an important role in the growth, yield and quality of wolfberry. This study aimed to develop new spectral indices (NSIs) and constructed machine learning regression (MLR) models for predicting wolfberry tree LNC. By utilizing four smoothing methods and five mathematic transformation methods, we obtained the original spectral dataset and five spectral transformation datasets for quantitative analysis and model establishment. Subsequently, published vegetation indices (PVIs) were acquired, sensitive wavelengths (SWs) were screened and NSIs were calculated based on SWs. Then MLR models were constructed by combining NSIs from six spectral datasets with three machine learning algorithms. Finally, a comparison was made among the MLR models. The study indicated that the application of mathematical transformation highlighted the differences in spectra, the square root, first-derivative and second-derivative transformation improved the prediction accuracy of MLR models constructed based on NSIs (MLR-NSIs models). However, these transformations had little impact on improving the prediction ability of MLR models constructed based on PVIs (MLR-PVIs models). Additionally, The optimal model for predicting the LNC of wolfberry tree was obtained by using the Random Forest (RF) algorithm combined with NSIs developed by first-derivative transformation spectra. The determination coefficient of validation (Rv2) and ratio of percentage deviation (RPD) was 0.71 and 1.90, respectively. In conclusion, this study has demonstrated that the combination of hyperspectral transformation and machine learning is useful for improving the accuracy of LNC estimation in wolfberry trees.

Inversion of soil water and salt information based on UAV hyperspectral remote sensing and machine lear-ning.

Accurate diagnosis of water and salt information in saline agricultural lands is crucial for long-term soil quality improvement and arable land conservation. In this study, we extracted field-scale vegetation canopy spectral information by UAV hyperspectral information, transforming the reflectance (R) to standard normal variate transformation (SNV), multiplicative scatter correction (MSC), first derivative of reflectance (FDR) and second derivative of reflectance (SDR). We determined the optimal spectral transformation forms of soil water content (SWC), soil pH, and soil salt content (SSC) by the maximum absolute correlation coefficient (MACC), and extracted the feature bands by competitive adaptive reweighted sampling (CARS). We constructed an inversion model of soil water and salt information by partial least squares regression (PLSR), random forest (RF), and extreme gradient boosting (XGBoost). The results showed that R, FDR and MSC were the best spectral transformation types for soil water content, soil pH, and soil salt content, and the corresponding MACC were 0.730, 0.472 and 0.654, respectively. The CARS algorithm effectively eliminated the irrelevant variables, optimally selecting 16-17 feature bands from 150 spectral bands. Both soil water content and soil pH performed best with XGBoost model, achieving determination coefficient of validation (Rp2) 0.927 and 0.743, and the relative percentage difference (RPD) amounted to 3.93 and 2.45. For soil salt content, the RF model emerged as the best inversion method with Rp2 and RPD of 0.427 and 1.64, respectively. The study could provide a reference solution for the integrated remote sensing monitoring of soil water and salt information in space and sky, serving as a scientific guide for the amelioration and sustainable management of saline lands.

Mapping mangrove leaf area index (LAI) by combining remote sensing images with PROSAIL‐D and XGBoost methods

AbstractLeaf area index (LAI) is a vital parameter reflecting vegetation structure, physio‐ecological process and growth development. Accurate estimation of mangrove LAI is fundamental for assessing the ecological restoration and sustainable development of mangrove ecosystems. To date, very few studies have explored the hybrid method of radiative transfer model (RTM) and machine‐learning model in retrieving mangrove LAI with different satellite sensors. This study investigated the capabilities of combining the PROSAIL‐D model, XGBoost (extreme gradient boosting) and remote sensing images in estimating mangrove LAI, considering the spatial resolutions and spectral vegetation indices (VIs) of Landsat‐8, Sentinel‐2, Worldview‐2 and Zhuhai‐1 images, and further explored the role of eco‐environmental factors in the spatial distribution of LAI in Gaoqiao Mangrove Reserve, China. The results showed that the Zhuhai‐1 acquires the best estimation accuracy ( (the determination coefficient of validation) = 0.86, RPD (residual prediction deviation) = 3.36 and RMSE (root mean square error) = 0.31), followed by Worldview‐2 ( = 0.84, RPD = 2.64 and RMSE = 0.33), Sentinel‐2 ( = 0.34, RPD = 1.33 and RMSE = 0.62) and Landsat‐8 ( = 0.29, RPD = 1.03 and RMSE = 0.71). The newly developed three‐band VIs ( with Landsat‐8, with Sentinel‐2, with Worldview‐2 and with Zhuhai‐1) were efficient estimators of mangrove LAI. Moreover, elevation and species composition could greatly affect the spatial distribution of mangrove LAI. We concluded that the hybrid method of PROSAIL‐D and XGBoost model using VIs derived from Zhuhai‐1 hyperspectral image could be deemed as basic method and input variables of mapping mangrove LAI, and could be effectively and widely applied in generating mangrove LAI products at the regional and national scales.

Assessment of the ED‐XRF technique to quantify mineral elements in nonlyophilised milk and cheese

The interest in analytical methods that accurately measure or predict the elemental profile of dairy foods has been steadily rising. The purpose of this study was to assess the robustness of the energy‐dispersive X‐ray fluorescence (ED‐XRF) technique for the prediction of mineral elements in milk and cheese, avoiding any sample preparation steps. Results highlighted relatively low accuracy of the ED‐XRF technique for the quantification of milk mineral elements, with coefficients of determination in validation () from 0.03 (magnesium) to 0.39 (phosphorus). Greater accuracies were obtained for the quantification of cheese minerals, with from 0.26 (sodium) to 0.69 (calcium).

NEAR INFRARED SPECTROSCOPY MEASUREMENT AND KINETIC MODELING FOR PHYSIOCHEMICAL PROPERTIES OF TABTIM FISH (HYBRID TILAPIA OREOCHROMIS SP.) FILLETS DURING CHILLING STORAGE

This research is aimed to investigate the physiochemical properties of Tabtim fish fillets [pH, moisture content, water holding capacity (WHC) and texture profile analysis] during storage at 0, 5, and 10ºC for up to 18 days. Near infrared spectroscopy (NIR) was used as a nondestructive measurement and kinetics was also applied for determining the quality changes. Results showed that the pH value increased with an increase of storage time and storage temperature (P≤0.05) while the hardness and WHC were decreased as a result of increasing time and storage temperature (P≤0.05). Change in pH, hardness, gumminess, and chewiness during storage can be described by zero- or first-order reactions. NIR spectra of fish fillets (180 samples) were collected in the reflectance mode with both the short-wavelength region (700-1000 nm) and long-wavelength region (1000-2500 nm). A partial regression model with cross validation was created. The models for pH, hardness, adhesiveness, and gumminess demonstrated that the value for the determination coefficient of validation ranged from 0.61-0.73 while the standard error of prediction was 0.14 for pH, 2.58 N for hardness, 0.21 N.mm for adhesiveness, and 1.68 N for gumminess. The model developed using NIR could predict pH, hardness, adhesiveness, and gumminess as basic for a screening test.

Concurrent starch accumulation in stump and high fruit production in coffee (Coffea arabica).

In coffee, fruit production on a given shoot drops after some years of high yield, triggering pruning to induce resprouting. The timing of pruning is a crucial farmer's decision affecting yield and labour. One reason for fruit production drop could be the exhaustion of resources, particularly the non-structural carbohydrates (NSC). To test this hypothesis in a Coffea L. arabica agroforestry system, we measured the concentrations of NSC, carbon (C) and nitrogen (N) in leaves, stems and stumps of the coffee plants, 2 and 5years after pruning. We also compared shaded vs full sun plants. For that purpose, both analytical reference and visible and near infrared reflectance spectroscopy (VNIRS) methods were used. As expected, concentrations of biochemical variables linked to photosynthesis activity (N, glucose, fructose, sucrose) decreased from leaves to stems, and then to stumps. In contrast, variables linked more closely to plant structure and reserves (total C, C:N ratio, starch concentration) were higher in long lifespan organs like stumps. Shading had little effect on most measured parameters, contrary to expectations. Concentrations of N, glucose and fructose were higher in 2-year-old organs. Conversely, starch concentration in perennial stumps was three times higher 5years after pruning than 2years after pruning, despite high fruit production. Therefore, the drop in fruit production occurring after 5-6years was not due to a lack of NSC on plant scale. Starch accumulation in perennial organs concurrently to other sinks, such as fruit growth, could be considered as a 'survival' strategy, which may be a relic of the behaviour of wild coffee (a tropical shade-tolerant plant). This study confirmed that VNIRS is a promisingly rapid and cost-effective option for starch monitoring (coefficient of determination for validation, R2val=0.91), whereas predictions were less accurate for soluble sugars, probably due to their too similar spectral signature.

Potential of Multivariate Statistical Technique Based on the Effective Spectra Bands to Estimate the Plant Water Content of Wheat Under Different Irrigation Regimes

Real-time, nondestructive, and accurate estimation of plant water status is important to the precision irrigation of winter wheat. The objective of this study was to develop a method to estimate plant water content (PWC) by using canopy spectral proximal sensing data. Two experiments under different water stresses were conducted in 2014–2015 and 2015–2016. The PWC and canopy reflectance of winter wheat were collected at different growth stages (the jointing, booting, heading, flowering, and filling stages in 2015 and the jointing, booting, flowering, and filling stages in 2016). The performance of different spectral transformation approaches was further compared. Based on the optimal pretreatment, partial least squares regression (PLSR) and four combination methods [i.e., PLSR-stepwise regression (SR), PLSR-successive projections algorithm (SPA), PLSR-random frog (RF), and PLSR-uninformative variables elimination (UVE)] were used to extract the sensitive bands of PWC. The results showed that all transformed spectra were closely correlated to PWC. The PLSR models based on the first derivative transformation method exhibited the best performance (coefficient of determination in calibration, R2C = 0.96; root mean square error in calibration, RMSEC = 20.49%; ratio of performance to interquartile distance in calibration, RPIQC = 9.19; and coefficient of determination in validation, R2V = 0.86; root mean square error in validation, RMSEV = 46.27%; ratio of performance to interquartile distance in validation, RPIQV = 4.34). Among the combination models, the PLSR model established with the sensitive bands from PLSR-RF demonstrated a good performance for calibration and validation (R2C = 0.99, RMSEC = 11.53%, and RPIQC = 16.34; and R2V = 0.84, RMSEV = 44.40%, and RPIQV = 4.52, respectively). This study provides a theoretical basis and a reference for estimating PWC of winter wheat by using canopy spectral proximal sensing data.

Application of near infrared hyperspectral imaging for identifying and quantifying red clover contained in experimental poultry refusals

For laying hens, free-range systems are promoted to improve both animal behavior and egg quality. In this context, estimation of feed intake, and particularly the proportion of herbage in the diet, has become a hot topic in poultry feeding management. During experiments, animals are housed in individual cages to measure daily ingestion, calculated by the difference between offered feed and refusals (non-ingested feed). Due to their natural behavior, laying hens tend to put offered feed on the ground, mixing grains and herbage with impurities (wood-chip litter and droppings) which leads to actual ingestion being underestimated. Manual sorting of mixed refusals is tedious. The aim of this study was thus to propose a procedure to identify herbage, red clover in this case, among impurities and to estimate the weight of red clover in entire mixed refusals by combining near infrared reflectance (NIR) imaging systems. A discriminant analysis (based on 8074 spectra) allowed each pixel corresponding to red clover to be identified. The results showed that more than 90 % of pixels were correctly classified. On the basis of the number of pixels predicted as red clover, a linear regression was built to convert pixels into red clover weight in order to recalculate the actual feed intake. A determination coefficient (R²) of 0.95 was achieved. Model validation was based on 12 composite refusals, and resulted in a determination coefficient of validation (R²v) equal to 0.83 and a root mean square error of prediction (RMSEP) equivalent to 3.4 g of dry matter (DM). The equation was considered satisfactory considering the possible bias in the manual sorting method. The main advantage of this procedure is the reduction in the time taken by the procedure by a factor of 4, while maintaining a high level of accuracy.

Simultaneous Determination of Caffeine and Taurine in Energy Drinks by FT-MIR Spectroscopy Coupled with Multivariate Analysis

Energy drinks have been studied due to their damaging side effects on the health of their consumers when consumed in excess or when combined with alcohol. Our objective was to develop chemometric models, based on Fourier-transform mid-infrared (FT-MIR) spectroscopy, to quantify the taurine and caffeine content in energy drinks rapidly and simultaneously. The taurine and caffeine content in the 50 samples ranged between 0 and 69.51 mg/100 mL and 14.92 and 1126.17 mg/100 mL, respectively. The best prediction model was obtained with the partial least squares (PLS1) algorithm; for taurine, the following values were obtained: determination coefficient of calibration (Rc2) = 0.9999, standard error of calibration (SEC) = 0.15, determination coefficient of validation (Rv2) = 0.9997, and standard error of prediction (SEP) = 0.16; for caffeine, Rc2 = 0.9999, SEC = 0.26, Rv2 = 0.9999, and SEP = 0.32. The model developed with PLS1 showed certainty in predictions during the validation stage and during application to external samples. FT-MIR coupled to chemometrics is a reliable and fast technique (compared to conventional techniques) to quantify taurine and caffeine in energy drinks simultaneously.

Comparison of the Potential Abilities of Three Spectroscopy Methods: Near-Infrared, Mid-Infrared, and Molecular Fluorescence, to Predict Carotenoid, Vitamin and Fatty Acid Contents in Cow Milk.

The objective of this work is to compare the ability of three spectroscopy techniques: molecular fluorescence, near-infrared (NIR), and mid-infrared with attenuated total reflectance (MIR-ATR) spectroscopy to predict the concentrations of 8 carotenoids, 6 vitamins and 22 fatty acids (FA) in cow’s milk. A dataset was built through the analysis of 242 frozen milk samples from different experiments. The milk compounds were analysed using reference methods and by NIR, MIR-ATR, and fluorescence to establish different predictive models. NIR spectroscopy allowed for better prediction of cis9-β-carotene, β-cryptoxanthin and the sum of carotenoids than the other techniques, with a coefficient of cross-validation in calibration (R2CV) > 0.60 and a coefficient of determination in validation (R2V) > 0.50. Their standard errors of prediction (SEP) were equal to 0.01, except for the sum of carotenoids (SEP = 0.15). However, MIR-ATR and fluorescence seem usable for the prediction of lutein and all-trans-β-carotene, respectively. These three spectroscopy methods did not allow us to predict (R2CV < 0.30) vitamin contents except, for vitamin A (the best R²CV = 0.65 with NIR and SEP = 0.15) and α-tocopherol (the best R²CV = 0.56 with MIR-ATR and SEP = 0.41), but all R²V were <0.30. NIR spectroscopy yielded the best prediction of the selected milk FA.

Influences of Detection Position and Double Detection Regions on Determining Soluble Solids Content (SSC) for Apples Using On-line Visible/Near-Infrared (Vis/NIR) Spectroscopy

With increasing living standards, more attention has been drawn to the eating quality of fruits. To meet the standards of market and consumers, on-line fruits grading in terms of the internal qualities, such as soluble solids content (SSC), is necessary in the industrial scale. Near-infrared (NIR) spectroscopy technology has advantages of being rapid, safe, non-destructive, and environmentally friendly; thus, it was widely applied in this area. Therefore, in the present work, an on-line near-infrared SSC detection system for apples based on bicone roller transportation was used to compare the influences of detection positions and double detection regions with the help of mono-branch optical fiber and binary-branch optical fiber on the performance of the prediction models. Better prediction models were obtained at the specific position than those at the random one. The system with the binary-branch optical fiber proved superior robustness, while that with the mono-branch optical fiber proved superior accuracy. The optimal model, when considering robustness and accuracy, had a root mean square error of calibration (RMSEC), determination coefficient of calibration ( $$ {R}_C^2 $$ ), root mean square error of validation (RMSEP), and determination coefficient of validation ( $$ {R}_P^2 $$ ) of 0.568%, 0.7319, 0.610%, and 0.6295, respectively. It was established by data simultaneously detected in two different regions utilizing the binary-branch optical fiber at the specific position. This work provided a possible solution on improving the model robustness in the practical application of on-line NIR detection for the SSC measurement of apples.

FT-MIR spectroscopy and multivariate analysis for determination of bioactive compounds and antioxidant capacity in Cabernet Sauvignon wines

ABSTRACTThirty samples of Cabernet Sauvignon Mexican wines from 5 wine regions, 19 wineries and vintage between 2004 and 2012 were characterised regarding total phenolic compounds (TPC), anthocyanins (AC), tannins (TC), flavonoids (FC) and antioxidant capacity (ABTS and DPPH assays). Samples were processed using Fourier transform mid-infrared spectroscopy combined with partial least squares regression (PLS1 and PLS2) and principal components regression algorithms to obtain one optimised chemometric model for each algorithm. The spectral region selected for each model was between 1550 and 824 cm−1. The PLS1 model showed the best predictive capacity for simultaneous prediction of the six parameters. Excellent values of the coefficient of determination in validation (Rv2; 0.93–0.95) and residual predictive deviation (RPD; 4.28–6.95) were obtained for TPC, FC, ABTS and DPPH. Suitable values of Rv2 (0.91–0.92) and RPD (3.75–3.84) were obtained for TC and AC.

Glycogen Concentration in Freeze-Dried Tissues of Eastern Oyster (Crassostrea virginica) Using Near Infrared Reflectance Spectroscopy to Determine the Relationship Between Concentrations of the Tissues Excised for Histological Sampling and the Remaining Tissues

To improve the accuracy and reproducibility of the previous near infrared reflectance spectroscopy (NIRS) model for glycogen in the oyster species Crassostrea virginica, a new model using freeze-dried samples was developed. The NIRS glycogen calibration model was developed using 380 individual oyster samples collected between 2014 and 2016 from several locations in the Chesapeake Bay. Homogenized freeze-dried samples were scanned in the near infrared region between 1,000 and 2,500 nm. In parallel, glycogen concentration (GC), measured as percent dry weight, was determined using laboratory-based methods. The two sets of data allowed us to build a NIRS model based on freeze-dried oyster meats, and the model gave a strong prediction of GC [coefficient of determination for validation (R2val) = 0.96 and residual predictive deviation (RPD) = 5.2]. The second part of the study applied the model to determine GC among 39 diploid and 40 triploid C. virginica and determined the strength of the relationship between the GC of tissues excised for histological sampling to the remaining tissue (corpus) to verify assumptions made throughout the literature. There was an estimated R2 = 0.99 between the GC in the corpus and the tissues of whole oyster meat. Among the samples, two factors, ploidy and size (shell height), had a significant effect on GC.

Prediction of pH of fresh chicken breast fillets by VNIR hyperspectral imaging

Visible and near-infrared (VNIR) hyperspectral imaging (400–900 nm) was used to evaluate pH of fresh chicken breast fillets (pectoralis major muscle) from the bone (dorsal) side of individual fillets. After the principal component analysis (PCA), a band threshold method was applied to the first principal component (PC1) score image in order to get the region of interest (ROI). Then, the average reflective spectrum of ROI of each image was acquired by inverse PCA transform. Eight pretreatment algorithms were evaluated for partial least squares regression (PLSR) models. The PLSR model with the pretreatment of multiplicative scatter correction followed by second derivative showed the best performance with coefficients of determination for validation (Rv2) of 0.87, root mean square error for validation (RMSEv) of 0.16 and the ratio of percentage deviation (RPD) of 2.02. Optimal 20 wavelengths were selected using competitive adaptive reweighed sampling (CARS) method to develop a new multispectral PLSR model, leading to an enhanced result with Rv2 of 0.94, RMSEv of 0.06 and RPD of 3.55. To assess the performance of the prediction models, new ROIs where pH values were measured using a pH probe, were defined and corresponding mean spectra were used as an independent test set of the new multispectral PLSR model. Coefficients of determination for independent test set (Rp2) and root mean square error for independent test set (RMSEp) were 0.73 and 0.29, respectively. The prediction image showing the spatial distribution of the predicted pH values was generated to analyze the spatial context of pH values as well as the overall pH level of each fillet. The results demonstrated that VNIR hyperspectral imaging could be used to predict spatial and global pH values of fresh chicken breast meat.

Near infrared spectroscopy combined with multivariate analysis for monitoring the ethanol precipitation process of fraction I + II + III supernatant in human albumin separation

Nowadays, as a powerful process analytical tool, near infrared spectroscopy (NIRS) has been widely applied in process monitoring. In present work, NIRS combined with multivariate analysis was used to monitor the ethanol precipitation process of fraction I+II+III (FI+II+III) supernatant in human albumin (HA) separation to achieve qualitative and quantitative monitoring at the same time and assure the product's quality. First, a qualitative model was established by using principal component analysis (PCA) with 6 of 8 normal batches samples, and evaluated by the remaining 2 normal batches and 3 abnormal batches. The results showed that the first principal component (PC1) score chart could be successfully used for fault detection and diagnosis. Then, two quantitative models were built with 6 of 8 normal batches to determine the content of the total protein (TP) and HA separately by using partial least squares regression (PLS-R) strategy, and the models were validated by 2 remaining normal batches. The determination coefficient of validation (Rp2), root mean square error of cross validation (RMSECV), root mean square error of prediction (RMSEP) and ratio of performance deviation (RPD) were 0.975, 0.501g/L, 0.465g/L and 5.57 for TP, and 0.969, 0.530g/L, 0.341g/L and 5.47 for HA, respectively. The results showed that the established models could give a rapid and accurate measurement of the content of TP and HA. The results of this study indicated that NIRS is an effective tool and could be successfully used for qualitative and quantitative monitoring the ethanol precipitation process of FI+II+III supernatant simultaneously. This research has significant reference value for assuring the quality and improving the recovery ratio of HA in industrialization scale by using NIRS.

Visible near infrared reflectance spectroscopy prediction of soil heavy metal concentrations in paper mill biosolid- and liming by-product-amended agricultural soils

Continuous assessment of soil heavy metal concentrations in agricultural soils is critical for ensuring ecosystem health and minimizing potential adverse effects. However, conventional laboratory determinations are time-consuming, expensive and involve hazardous chemicals. We evaluated the potential of visible near infrared reflectance spectroscopy (VNIRS) to rapidly and simultaneously predict total, Mehlich-3 (M3), DTPA, and water-soluble (H2O) soil heavy metal concentrations. Soil samples were collected from a loam soil (n=64) and a loamy sand (n=32) after repeated annual applications (3–9years) of paper mill biosolids (PB) and liming by-products. Models were developed using modified partial least squares regression on 75% of the total sample set and also for each soil, after removal of the spectral outliers (n=3), with the remainder used for validation. The coefficient of determination in validation (R2V) and the ratio of performance to deviation (RPD) were used to assess model accuracy. Predictions for the total set were excellent for total Zn and Cd, M3-Cu and -Cd, and DTPA-Cd (R2V>0.90, RPD>3.0), good for total Cu, Ni and Mo, M3-Ni, DTPA-Cu and -Ni, and H2O-Cu and -Ni (R2V=0.70 to 0.90, RPD=1.70 to 3.0), but poor for M3-Zn, DTPA-Zn, and H2O-Zn, -Cd and -Mo (R2V<0.70, RPD<1.70). Except for one case, site-specific models were either more accurate or provided similar results with the model of the total set in predicting site-specific concentrations. The relationship with soil organic matter (SOM) was the major predictive mechanism for Cd in our dataset, while the prediction of other soil heavy metals was due to the combined influence of SOM and Fe oxides. We conclude that VNIRS has the potential to predict total, M3, DTPA and to a lesser extent water-soluble soil heavy metals, especially when developed using site-specific samples.

Improving the prediction of arsenic contents in agricultural soils by combining the reflectance spectroscopy of soils and rice plants

Visible and near-infrared reflectance spectroscopy provides a beneficial tool for investigating soil heavy metal contamination. This study aimed to investigate mechanisms of soil arsenic prediction using laboratory based soil and leaf spectra, compare the prediction of arsenic content using soil spectra with that using rice plant spectra, and determine whether the combination of both could improve the prediction of soil arsenic content. A total of 100 samples were collected and the reflectance spectra of soils and rice plants were measured using a FieldSpec3 portable spectroradiometer (350–2500nm). After eliminating spectral outliers, the reflectance spectra were divided into calibration (n=62) and validation (n=32) data sets using the Kennard-Stone algorithm. Genetic algorithm (GA) was used to select useful spectral variables for soil arsenic prediction. Thereafter, the GA-selected spectral variables of the soil and leaf spectra were individually and jointly employed to calibrate the partial least squares regression (PLSR) models using the calibration data set. The regression models were validated and compared using independent validation data set. Furthermore, the correlation coefficients of soil arsenic against soil organic matter, leaf arsenic and leaf chlorophyll were calculated, and the important wavelengths for PLSR modeling were extracted. Results showed that arsenic prediction using the leaf spectra (coefficient of determination in validation, Rv2=0.54; root mean square error in validation, RMSEv=12.99mgkg−1; and residual prediction deviation in validation, RPDv=1.35) was slightly better than using the soil spectra (Rv2=0.42, RMSEv=13.35mgkg−1, and RPDv=1.31). However, results also showed that the combinational use of soil and leaf spectra resulted in higher arsenic prediction (Rv2=0.63, RMSEv=11.94mgkg−1, RPDv=1.47) compared with either soil or leaf spectra alone. Soil spectral bands near 480, 600, 670, 810, 1980, 2050 and 2290nm, leaf spectral bands near 700, 890 and 900nm in PLSR models were important wavelengths for soil arsenic prediction. Moreover, soil arsenic showed significantly positive correlations with soil organic matter (r=0.62, p<0.01) and leaf arsenic (r=0.77, p<0.01), and a significantly negative correlation with leaf chlorophyll (r=−0.67, p<0.01). The results showed that the prediction of arsenic contents using soil and leaf spectra may be based on their relationships with soil organic matter and leaf chlorophyll contents, respectively. Although RPD of 1.47 was below the recommended RPD of >2 for soil analysis, arsenic prediction in agricultural soils can be improved by combining the leaf and soil spectra.

Successive projections algorithm-based three-band vegetation index for foliar phosphorus estimation

Phosphorus (P) is essential for plant growth and development. Very few studies have reported the use of hyperspectral three-band vegetation indices (TBVIs) in foliar P estimation. Further, the optimal TBVI is generally chosen from millions of all possible band combinations. This study aimed to investigate resampling and two wavelength selection methods (genetic algorithm (GA) and successive projections algorithm (SPA)) in deriving TBVIs for foliar P estimation and further to compare the performances of the newly developed TBVIs and published VIs. A total of 137 field-based canopy hyperspectral reflectance (350–2500nm) of Carex (C. cinerascens) were obtained and reduced to 1603 wavelengths due to spectral noises. Considering both the original and first derivative reflectance spectra, their resampled wavelengths and selected wavelengths by GA and SPA were employed to derive TBVIs. A total of 24 selected TBVI models were calibrated for foliar P estimation with the training dataset, and they were independently validated with the test dataset. The root mean square error of validation (RMSEVal), determination coefficient of validation (RVal2) and residual prediction deviation (RPD) values were calculated to evaluate the performance of each model. The results demonstrated that 5474, 1972 and 1.2s in average was taken in calculating all possible TBVIs using resampling, GA and SPA, respectively. Two SPA-based TBVIs, i.e. (ρ760−ρ2387)/(ρ723−ρ2387) (ρλ, original reflectance) and (ρ′729−ρ′1319+2ρ′714)/(ρ′729+ρ′1319−2ρ′714) (ρ′λ, first derivative reflectance), had the best model performances (RVal2=0.680, RMSEVal=0.040%, RPD=1.75; RVal2=0.692, RMSEVal=0.039%, RPD=1.80) in foliar P estimation among the 24 TBVIs. Compared with 15 published VIs (RVal2<0.64, RPD<1.64), the two SPA-based TBVIs exhibited better validation performances. We concluded that SPA has the great potential for TBVI derivation due to the reduction of computation time, and the use of SPA in TBVI derivation is recommended for NDVI derivation or other biochemical parameter estimation.