Leaf optical properties can be used to identify environmental conditions, the effect of light intensities, plant hormone levels, pigment concentrations, and cellular structures. However, the ...reflectance factors can affect the accuracy of predictions for chlorophyll and carotenoid concentrations. In this study, we tested the hypothesis that technology using two hyperspectral sensors for both reflectance and absorbance data would result in more accurate predictions of absorbance spectra. Our findings indicated that the green/yellow regions (500-600 nm) had a greater impact on photosynthetic pigment predictions, while the blue (440-485 nm) and red (626-700 nm) regions had a minor impact. Strong correlations were found between absorbance (R
= 0.87 and 0.91) and reflectance (R
= 0.80 and 0.78) for chlorophyll and carotenoids, respectively. Carotenoids showed particularly high and significant correlation coefficients using the partial least squares regression (PLSR) method (R
= 0.91, R
cv = 0.85, and R
= 0.90) when associated with hyperspectral absorbance data. Our hypothesis was supported, and these results demonstrate the effectiveness of using two hyperspectral sensors for optical leaf profile analysis and predicting the concentration of photosynthetic pigments using multivariate statistical methods. This method for two sensors is more efficient and shows better results compared to traditional single sensor techniques for measuring chloroplast changes and pigment phenotyping in plants.
Spectroscopic measurements of soil samples are reliable because they are highly repeatable and reproducible. They characterise the samples' mineral–organic composition. Estimates of concentrations of ...soil constituents are inevitably less precise than estimates obtained conventionally by chemical analysis. But the cost of each spectroscopic estimate is at most one‐tenth of the cost of a chemical determination. Spectroscopy is cost‐effective when we need many data, despite the costs and errors of calibration. Soil spectroscopists understand the risks of over‐fitting models to highly dimensional multivariate spectra and have command of the mathematical and statistical methods to avoid them. Machine learning has fast become an algorithmic alternative to statistical analysis for estimating concentrations of soil constituents from reflectance spectra. As with any modelling, we need judicious implementation of machine learning as it also carries the risk of over‐fitting predictions to irrelevant elements of the spectra. To use the methods confidently, we need to validate the outcomes with appropriately sampled, independent data sets. Not all machine learning should be considered ‘black boxes’. Their interpretability depends on the algorithm, and some are highly interpretable and explainable. Some are difficult to interpret because of complex transformations or their huge and complicated network of parameters. But there is rapidly advancing research on explainable machine learning, and these methods are finding applications in soil science and spectroscopy. In many parts of the world, soil and environmental scientists recognise the merits of soil spectroscopy. They are building spectral libraries on which they can draw to localise the modelling and derive soil information for new projects within their domains. We hope our article gives readers a more balanced and optimistic perspective of soil spectroscopy and its future.
Highlights
Spectroscopy is reliable because it is a highly repeatable and reproducible analytical technique.
Spectra are calibrated to estimate concentrations of soil properties with known error.
Spectroscopy is cost‐effective for estimating soil properties.
Machine learning is becoming ever more powerful for extracting accurate information from spectra, and methods for interpreting the models exist.
Large libraries of soil spectra provide information that can be used locally to aid estimates from new samples.
Display omitted
•We analyzed the analytical quality of 36 commercial laboratories from Brazil and Paraguay.•Outlying attributes in spectral clusters detected by adjusted method to the data ...distribution.•Spectra-based systematic sampling had attributes equally distributed among the subsets.•Nearly 25% laboratories had the potential to reduce the need for laboratory analysis by at least 50%.
We need innovative solutions to help laboratories to rapidly characterize the soils and adopt internal quality controls with lower costs and ecosystem impacts. Soil spectroscopy emerged as an alternative to wet chemistry. However, soil laboratories still do not widely use this technology in their routines, mainly due to the lack of: i) standards and protocols, ii) spectral libraries, iii) capacity in spectral methods, and iv) professionals with expertise in chemometrics. Therefore, we aimed to provide basic guidelines for the use of soil spectra in laboratory routines regarding internal quality control and samples selection for analysis and prediction. We used 350–2500 nm spectra and unsupervised random forests to compute proximity matrices and cluster spectrally similar soil samples for outlier detection using an adjusted method for skewed distribution, which allowed us to establish an internal quality control. Then, we used the most important wavelengths from unsupervised random forest to order soil samples by laboratory and subset them into different set sizes for training and testing prediction models for clay, sand, organic matter cation exchange capacity and base saturation. Our results indicated that internal quality control based on soil spectra and unsupervised analysis can be implemented for laboratory routines. Spectra-based soil samples selection and attributes predictions can reduce the need for laboratory analysis by half. Soil spectroscopy can become an important alternative to improve traditional analytical results, save time, reduce costs, and mitigate environmental impact with acceptable accuracy. Soil data and full codes from this study are publicly available at https://github.com/raulpoppiel/urssa.
The soil carbon pool which is the sum of soil organic carbon (SOC) and soil inorganic carbon (SIC) is the second largest active store of carbon after the oceans and it is an important component of ...the global carbon cycle. Hence, accurate estimation of SOC and SIC as important carbon reservoirs in terrestrial ecosystems using fast, inexpensive and non-destructive methods is crucial for planning different climate change policies. The aim of the current research was to examine the effectiveness of Vis-NIR (visible and near-infrared spectroscopy: 350–2500 nm) and MIR (mid-infrared spectroscopy: 4000–400 cm
−1
) to characterize and estimate soil organic matter (SOM) and carbonates as main components of soil carbon stocks in Juneqan, Charmahal va Bakhtiari, Iran. To do so, a total of 548 soil samples from this area were collected (October 2015) and analyzed in laboratory (August 2017). In order to develop models capable of predicting SOM and carbonates content, seven spectral preprocessing methods comprising Absorbance (Abs), De-trending (Det), Continuum removal (CR), Savitzky-Golay derivatives (SGD), standard normal variate transformation (SNV), multiplicative scatter correction (MSC) and Normalization by range (NBR) were conducted along with five multivariate methods including Random Forest (RF), Partial Least-Squares Regression (PLSR), Artificial Neural Network (ANN), Support Vector Machine (SVM) and Gaussian Process Regression (GPR). The content of
carbonates
caused spectral reflectance intensity to augment on several ranges of spectrum and strong absorption feature at 2338 nm in the Vis-NIR and 714, 850, 870, 1796, 2150 and 2510 cm
−1
in the MIR spectra range. SOM absorbed energy in several ranges, but also showed specific peaks in MIR. Both facts are associated with the structure of
carbonates
and SOM and its interaction with energy. The best combination of preprocessing and calibration models for carbonates quantification in Vis-NIR spectra was Det/PLSR (
R
2
= 0.74, RPD= 2.19, RMSE= 6.45). For SOM, it was Det/PLSR (
R
2
= 0.82, RPD= 2.41, RMSE= 0.75). The Det/RF (
R
2
= 0.87, RPD= 2.44, RMSE= 0.66) for the quantification of SOM and MSC/RF (
R
2
= 0.84, RPD= 2.84, RMSE= 5.50) for carbonates in MIR spectra range showed the greatest results. The stronger occurrence of spectral bands in MIR as well as the specificity of the absorption features indicated that this range produced better predictions. The obtained results highlighted the significant role of soil spectroscopy technique in predicting SOC and soil carbonates as key components of soil carbon stocks in the study area. Therefore, this technique can be used as a more cost-effective, time saving and nondestructive alternative to traditional methods of soil analysis.
Soil mapping has been identified as key to environmental issues. The determination of soil attributes to achieve the best decision making on land use planning is crucial. The use of remote sensing ...(satellite images) can improve understanding of the surface, since it collects a spectral reflectance fingerprint related to soil properties. However, methodologies still gather spatially fragmented information on bare soil in a single image; thus, there is still room to improve information as a continuous surface. This work has the purpose of developing a procedure using multi-temporal satellite image information, aiming to construct a single synthetic image which would represent soils. The work was carried out in the state of São Paulo, Brazil, on a site covering 14,614 km2. The procedure, designated as Geospatial Soil Sensing System (GEOS3), is based on the following steps: a) creation of a database with Landsat 5 legacy data.; b) filtering of the database to provide images only from the dry season in the region; c) insertion of a set of rules into the system to filter other objects besides soils; d) Each bare soil occurrence for each location along the time-series was used to calculate a Temporal Synthetic Spectral Reflectance (TESS) of the soil surface; e) aggregation of all TESS composes the Synthetic Soil Image (SYSI); f) quantitative and qualitative validation of the SYSI through the correlation between laboratory and TESS, soil line assessment and the principal component analysis (PCA). GEOS3 was able to provide the best representative reflectance of soils for each band during the historical period. Thus, TESS is not the ‘true’ but a synthetic spectral reflectance. The canonical correlation between laboratory and satellite data reached 0.93. A value of up to 0.88 in the Pearson's correlation between laboratory and TESS was also achieved. In a single scene, only 0.5% of area was available as isolated bare soil for spatial analysis. However, SYSI reached 68%. Considering only the sugarcane agricultural areas, a value of 92% was achieved. Our study indicates that a multi-temporal data mining procedure can retrieve soil surface representation. The key to the results was calculating the median spectral reflectance from the bare soil pixels along the period of the time series. GEOS3 products can aid soil evaluation by assisting in digital soil mapping, soil security, precision agriculture, soil attribute quantification, soil conservation, environment monitoring and soil sample allocation, among others.
Overview of the Synthetic Soil Image (SYSI) generated from 27 years of Landsat 5 Thematic Mapper (TM) data. a) Soil surface variations from SYSI, in the RGB 543 composite. Black regions are missing values. b) View for the same extent in figure a by a single Landsat 5 TM image from 22 Feb 2005, in the RGB 321 composite. c) The 2,5D view of SYSI from figure a. d) Elevation gradient for the same extent in figure c (source: digital elevation model from the Shuttle Radar Thematic Mapper (SRTM)). The purple shades are related to soil types. Black regions are missing values for soil surface reflectance. Display omitted
•A soil reflectance composite was created from a time series of multispectral data.•The correlation between laboratory and the composite spectra reached r = 0.66 to 0.88.•Image and laboratory spectra achieved a canonical correlation of up to 0.93.•The system provided imagery for 92% of the cropped area and 68% for the entire region.•Accurate soil spectral reflectance maps are produced at local to regional scales.
•Soil spectral information is sensitive to measurement protocol.•The ISS protocol was tested for spectral harmonization of different SSLs.•Different SSLs may cause prediction bias in external ...samples.•A universal protocol is needed to solve this problem.
Soil spectral libraries (SSLs) are being created worldwide due to their enormous potential to train machine learning algorithms that may benefit environmental and agricultural activities. Accordingly, the extent to which the different SSLs maintain their accuracy for unknown samples is important. Recently, the internal soil standard (ISS) has garnered attention in the soil spectroscopy literature, due to its capacity to rectify systematic effects during spectral measurements. Moreover, a new initiative by the Institute of Electrical and Electronics Engineers Standards Association (IEEE SA) to create an agreed-upon measurement standard and protocol for soil spectral measurements has been launched. Application of the ISS to align one spectrum with another for merging different SSLs is highly needed. This led us to postulate the necessity of evaluating the performance of spectral-based models using different large SSLs that were created following different protocols. Using the random forest (RF) algorithm, we tested four different well-known large SSLs to predict clay content. From these SSLs, two of them were created using the ISS method (GEOCRADLE and Israel) and the other two without (LUCAS and ICRAF–ISRIC). Then, three groups of soils from different regions (Israel, Czech Republic, and Brazil) were used as external test samples to examine the SSLs' performance. The SSLs that were generated following the ISS protocol yielded the best performance in a group of samples that utilized the same method. It was concluded that using an agreed ISS such as the Lucky Bay (LB) soil sample enables a favorable way to merge SSLs.
•Prediction of soil texture via pXRF and Vis-NIR DRS data was evaluated.•Combination of A and B horizons data resulted in predictions with R2 above 0.80.•In general, the best predictions were ...achieved using pXRF data.•RF algorithm outperformed other algorithms for soil texture prediction.•The best prediction models were obtained with pXRF + RF using B horizon data.
Portable X-ray fluorescence (pXRF) spectrometry and visible near-infrared diffuse reflectance spectroscopy (Vis-NIR DRS), used separately or in tandem, have become important techniques for determination and prediction of soil attributes worldwide. However, there is little information available regarding the effectiveness of their combined use in tropical soils. This study aimed to predict soil texture using pXRF and Vis-NIR DRS, evaluating the efficiency of using these proximal sensors separately and in tandem. A total of 315 soil samples were collected from A and B horizons in Brazil. Soil samples were submitted to analyses of texture, pXRF and Vis-NIR DRS. Vis-NIR DRS spectral data pre-processing was evaluated by comparing results delivered by the derivative smoothing methods Savitzky-Golay (WT), Savitzky-Golay with Binning (WB), and data without the pre-processing treatment (WOT). Four algorithms were utilized for predictions: Gaussian Process (Gaussian), Support Vector Machine with linear (SVM-L) and radial (SVM-R) kernels, and Random Forest (RF). In general, models using only pXRF data slightly outperformed models using Vis-NIR DRS (WT, WB, WOT) data alone. Models combining data from both sensors achieved similar results to those obtained by pXRF alone. The best predictions of sand, silt, and clay contents were obtained via pXRF + RF using B horizon data, reaching R2 values of 0.91, 0.81, and 0.83, respectively. Although pXRF alone provided slightly better results, soil texture can be accurately predicted via both pXRF and Vis-NIR DRS data, separately and in tandem. These sensors can contribute to reduce costs and time required for tropical soil texture determination.
We evaluated the use of airborne hyperspectral imaging and non-imaging sensors in the Vis—NIR—SWIR spectral region to assess particle size and soil organic matter in the surface layer of tropical ...soils (Oxisols, Ultisols, Entisols). The study area is near Piracicaba municipality, São Paulo state, Brazil, in a sugarcane cultivation area of 135 hectares. The study area, with bare soil, was imaged in April 2016 by the AisaFENIX aerotransported hyperspectral sensor, with spectral resolution of 3.5 nm between 380 and 970 nm, and 12 nm between 970 and 2500 nm. We collected 66 surface soil samples. The samples were analyzed for particle size and soil organic matter content. Laboratory spectral measurements were performed using a non-imaging spectroradiometer (ASD FieldSpec 3 Jr). Partial Least Square Regression (PLSR) was used to predict clay, silt, sand and soil organic matter (SOM). The PLSR functions developed were applied to the hyperspectral image of the study area, allowing development of a prediction map of clay, sand, and SOM. The developed PLSR models demonstrated the relationship between the predictor variables at the cross-validation step, both for the non-imaging and imaging sensors, when the highest r and R2 values were obtained for clay, sand, and SOM, with R2 over 0.67. We did not obtain a satisfactory model for silt content. For the non-imaging sensor at the prediction step, R2 values for clay and SOM were over 0.7 and sand was lower than 0.54. The imaging sensor yielded models for clay, sand, and SOM with R2 values of 0.62, 0.66, and 0.67, respectively. Pearson correlation between sensors was greater than 0.849 for the prediction of clay, sand, and SOM. Our study successfully generated, from the imaging sensor, a large-scale and detailed predicted soil maps for particle size and SOM, which are important in the management of tropical soils.
Geophysical sensors combined with machine learning
algorithms were used to understand the pedosphere system and landscape
processes and to model soil attributes. In this research, we used parent
...material, terrain attributes, and data from geophysical sensors in different
combinations to test and compare different and novel machine learning
algorithms to model soil attributes. We also analyzed the importance of
pedoenvironmental variables in predictive models. For that, we collected
soil physicochemical and geophysical data (gamma-ray emission from uranium,
thorium, and potassium; magnetic susceptibility and apparent electric
conductivity) by three sensors (gamma-ray spectrometer, RS 230;
susceptibilimeter KT10, Terraplus; and conductivimeter, EM38 Geonics) at
75 points and analyzed the data. The models with the best performance
(R2 0.48, 0.36, 0.44, 0.36, 0.25, and 0.31) varied for clay, sand,
Fe2O3, TiO2, SiO2, and cation exchange capacity
prediction, respectively. Modeling with the selection of covariates at three
phases (variance close to zero, removal by correction, and removal by
importance) was adequate to increase the parsimony. The results were
validated using the method “nested leave-one-out cross-validation”. The
prediction of soil attributes by machine learning algorithms yielded
adequate values for field-collected data, without any sample preparation,
for most of the tested predictors (R2 values ranging from 0.20 to
0.50). Also, the use of four regression algorithms proved to be important
since at least one of the predictors used one of the tested algorithms. The
performance values of the best algorithms for each predictor were higher
than those obtained with the use of a mean value for the entire area
comparing the values of root mean square error (RMSE) and mean absolute
error (MAE). The best combination of sensors that reached the highest model
performance was that of the gamma-ray spectrometer and the
susceptibilimeter. The most important variables for most
predictions were parent material,
digital elevation, standardized height, and magnetic susceptibility. We concluded that soil attributes can be efficiently modeled
by geophysical data using machine learning techniques and geophysical sensor
combinations. This approach can facilitate future soil mapping in a more
time-efficient and environmentally friendly manner.