Visible-near infrared reflectance spectroscopy for rapid, nondestructive assessment of wetland soil quality

Determination of soil pH from Vis-NIR spectroscopy by extreme learning machine and variable selection: A case study in lime concretion black soil

Variable selection is widely accepted as an important step in the quantitative analysis of visible and near-infrared (Vis-NIR) spectroscopy, as it tends to improve the model's robustness and predictive ability. In this study, a total of 140 lime concretion black soil samples were collected from two towns in Guoyang County, China. The Vis-NIR spectra measured in the laboratory were used to estimate soil pH by an extreme learning machine (ELM). First, the soil spectra were treated by the optimized continuous wavelet transform (CWT), and then four spectral feature selection methods (competitive adaptive reweighted sampling, CARS; successive projections algorithm, SPA; Monte Carlo uninformative variable elimination, MCUVE; genetic algorithm, GA) were applied with ELM in the CWT domain to determine the techniques with most predictions. For comparison, The PLS and SVM models were also developed. The coefficient of determination (R²), root mean square error (RMSE), and residual prediction deviation (RPD) were used to evaluate the model performance. Based on the validation dataset, the performance of the ELM models was superior to that of the PLS and SVM models expect SPA and MCUVE. In the ELM models, the order of the prediction accuracy was GA-ELM (R²_p = 0.86; RMSE_p = 0.1484; RPD = 2.64), CARS-ELM (R²_p = 0.84; RMSE_p = 0.1565; RPD = 2.50), ELM (R²_p = 0.84; RMSE_p = 0.1572; RPD = 2.49), SPA-ELM (R²_p = 0.84; RMSE_p = 0.1589; RPD = 2.47) and MCUVE-ELM (R²_p = 0.83; RMSE_p = 0.1599; RPD = 2.45). The proposed method of CARS-ELM had a relatively strong ability for spectral variable selection while retaining excellent prediction accuracy and short computing time (0.39 s). In addition, the variables selected by the four methods (CARS, SPA, MCUVE and GA) indicated the prediction mechanism for pH in lime concretion black soil may be the relation between pH and iron oxides and organic matter. In conclusion, CARS-ELM has great potential to accurately determine the pH in lime concretion black soil using Vis-NIR spectroscopy.

Estimating Forest Soil Properties for Humus Assessment—Is Vis-NIR the Way to Go?

Recently, forest management faces new challenges resulting from increasing temperatures and drought occurrences. For sustainable, site-specific management strategies, the availability of up to date soil information is crucial. Proximal soil sensing techniques are a promising approach for rapid and inexpensive collection of data, and could facilitate the provision of the necessary information. This study evaluates the potential of visual and near-infrared spectroscopy (vis-NIRS) for estimating soil parameters relevant for humus mapping in Saxon forests. Therefore, soil samples from the organic layer are included. So far there is little knowledge about the applicability of vis-NIRS in the humus layer of forests. We investigate the spectral behaviour of samples from organic (Oh) and mineral (0–5 cm, Ah) horizons, pointing out differences in the occurring absorption features. Further, we identify and assess the accuracy of selected soil properties based on vis-NIRS for forest sites, compare the outcome of different regression methods, investigate the implications for forest soils due to the presence and different composition of the humus layer and organic horizons and interpret the results regarding their usefulness for soil mapping and monitoring purposes. For this, we used retained humus soil samples of forests from Saxony. Regression models were built with Partial Least Squares Regression, Support Vector Machine and Cubist. Investigated properties were carbon (C) and nitrogen (N) content, C/N ratio, pH value, cation exchange capacity (CEC) and base saturation (BS) due to their importance for assessing humus conditions in forests. In organic Oh horizons, prediction results for C and N content achieved R² values between 0.44 and 0.58, with corresponding RPIQ ranging from 1.58 to 2.06 depending on the used algorithm. Estimations of C/N ratio were more precise with R² = 0.65 and RMSE = 2.16. Best results were reported for pH value, with R² = 0.90 and RMSE = 0.20. Regarding BS, the best model accuracy was R² = 0.71, with RMSE = 13.97. In mineral topsoil, C and N content models achieved higher values of R² = 0.59 to 0.72, with RPIQ values between 2.22 and 2.54. However, prediction accuracy was lower for C/N ratio (R² = 0.50, RMSE = 3.52) and pH values (R² = 0.62, RMSE = 0.29). Models for CEC achieved R² = 0.65, with RPIQ = 2.81. In general, prediction precision varied dependent on the used algorithm, without showing clear tendencies. Classification into pH classes was exemplified since this offers a new perspective for humus mapping on forest soils. Balanced accuracy for the defined classes ranged from 0.50 to 0.87. We show that vis-NIR spectroscopy is suitable for assessing humus conditions in Saxon forests (Germany), in particular not only for mineral horizons but also for organic Oh horizons.

Fine-scale spatial homogenization of microbial habitats: a multivariate index of headwater wetland complex condition

With growing public awareness that wetlands are important to society, there are intensifying efforts to understand the ecological condition of those wetlands that remain, and to develop indicators of wetland condition. Indicators based on soils are not well developed and are absent in some current assessment protocols; these could be advantageous, particularly for soils, which are complex habitats for plants, invertebrates, and microbial communities. In this study, we examine whether multivariate soil indicators, correlated with microbial biomass and community composition, can be used to distinguish reference standard (i.e., high condition) headwater wetland complexes from impacted headwater wetland complexes in central Pennsylvania, USA. Our reference standard sites existed in forested landscapes, while our impacted sites were situated in multi-use landscapes and were affected by a range of land-use legacies in the 1900s. We found that current assessment protocols are likely underrepresenting sampling needs to accurately represent site mean soil properties. On average, more samples were required to represent soil property means in reference standard sites compared to impacted sites. Reference standard and impacted sites also had noticeably different types of microbial habitats for the two multivariate soil indices assessed, and impacted sites were more homogenized in terms of the fine-scale (i.e., 1 and 5 m) spatial variability of these indices. Our study shows promise for the use of multivariate soil indices as indicators of wetland condition and provides insights into the sample sizes and scales at which soil sampling should occur during assessments. Future work is needed to test the generalizability of these findings across wetland types and ecoregions and establish definitive links between structural changes in microbial habitats and changes in wetland soil functioning.

Machine-learning-based quantitative estimation of soil organic carbon content by VIS/NIR spectroscopy

Soil organic carbon (SOC) is an important soil property that has profound impact on soil quality and plant growth. With 140 soil samples collected from Ebinur Lake Wetland National Nature Reserve, Xinjiang Uyghur Autonomous Region of China, this research evaluated the feasibility of visible/near infrared (VIS/NIR) spectroscopy data (350-2,500 nm) and simulated EO-1 Hyperion data to estimate SOC in arid wetland regions. Three machine learning algorithms including Ant Colony Optimization-interval Partial Least Squares (ACO-iPLS), Recursive Feature Elimination-Support Vector Machine (RF-SVM), and Random Forest (RF) were employed to select spectral features and further estimate SOC. Results indicated that the feature wavelengths pertaining to SOC were mainly within the ranges of 745-910 nm and 1,911-2,254 nm. The combination of RF-SVM and first derivative pre-processing produced the highest estimation accuracy with the optimal values of Rt (correlation coefficient of testing set), RMSE t and RPD of 0.91, 0.27% and 2.41, respectively. The simulated EO-1 Hyperion data combined with Support Vector Machine (SVM) based recursive feature elimination algorithm produced the most accurate estimate of SOC content. For the testing set, Rt was 0.79, RMSE t was 0.19%, and RPD was 1.61. This practice provides an efficient, low-cost approach with potentially high accuracy to estimate SOC contents and hence supports better management and protection strategies for desert wetland ecosystems.

Spectral prediction of sediment chemistry in Lake Okeechobee, Florida

High-resolution diffuse reflectance spectra in the visible and near-infrared wavelengths were used to predict chemical properties of sediment samples obtained from Lake Okeechobee (FL, USA). Chemometric models yielded highly effective prediction (relative percent difference (RPD) = SD/RMSE >2) for some sediment properties including total magnesium (Mg), total calcium (Ca), total nitrogen (TN), total carbon (TC), and organic matter content (loss on ignition (LOI)). Predictions for iron (Fe), aluminum (Al), and various forms of phosphorus (total P (TP), HCl-extractable P (HCl-P), and KCl-extractable P (KCl-P)) were also sufficiently accurate (RPD > 1.5) to be considered useful; predictions for other P fractions as well as all pore water properties were poor. Notably, scanning wet sediments resulted in only a 7 % decline in RPD scores. Moreover, interpolation maps based on values predicted from wet sediment spectra captured the same spatial patterns for Ca, Mg, TC, TN, and TP as maps derived directly from wet chemistry, suggesting that field scanning of perpetually saturated sediments may be a viable option for expediting sample analysis and greatly reducing mapping costs. Indeed, the accuracy of spectral model predictions compared favorably with the accuracy of kriging model predictions derived from wet chemistry observations suggesting that, for some analytes, higher density spatial sampling enabled by use of field spectroscopy could increase the geographic accuracy of monitoring for changes in lake sediment chemical properties.

In situ measurement of some soil properties in paddy soil using visible and near-infrared spectroscopy

In situ measurements with visible and near-infrared spectroscopy (vis-NIR) provide an efficient way for acquiring soil information of paddy soils in the short time gap between the harvest and following rotation. The aim of this study was to evaluate its feasibility to predict a series of soil properties including organic matter (OM), organic carbon (OC), total nitrogen (TN), available nitrogen (AN), available phosphorus (AP), available potassium (AK) and pH of paddy soils in Zhejiang province, China. Firstly, the linear partial least squares regression (PLSR) was performed on the in situ spectra and the predictions were compared to those with laboratory-based recorded spectra. Then, the non-linear least-square support vector machine (LS-SVM) algorithm was carried out aiming to extract more useful information from the in situ spectra and improve predictions. Results show that in terms of OC, OM, TN, AN and pH, (i) the predictions were worse using in situ spectra compared to laboratory-based spectra with PLSR algorithm (ii) the prediction accuracy using LS-SVM (R²>0.75, RPD>1.90) was obviously improved with in situ vis-NIR spectra compared to PLSR algorithm, and comparable or even better than results generated using laboratory-based spectra with PLSR; (iii) in terms of AP and AK, poor predictions were obtained with in situ spectra (R²<0.5, RPD<1.50) either using PLSR or LS-SVM. The results highlight the use of LS-SVM for in situ vis-NIR spectroscopic estimation of soil properties of paddy soils.

Visible, near-infrared, and mid-infrared spectroscopy applications for soil assessment with emphasis on soil organic matter content and quality: state-of-the-art and key issues

Visible near-infrared (Vis-NIR) reflection spectroscopy and mid-infrared (mid-IR) reflection spectroscopy are cost- and time-effective and environmentally friendly techniques that could be alternatives to conventional soil analysis methods. Successful determination of spectrally active soil components, including soil organic matter (SOM), depends on the selection of suitable pretreatment and multivariate calibration techniques. The objective of the present review is to critically examine the suitability of Vis-NIR (350-2500 nm) and mid-IR (4000-400 cm(-1)) spectroscopy as a tool for SOM quantity and quality determination. Particular attention is paid to different pretreatment and calibration procedures and methods, and their ability to predict SOM content from Vis-NIR and mid-IR data is discussed. We then review the most recent research using spectroscopy in different calibration scales (local, regional, or global). Finally, accuracy and robustness, as well as uncertainty in Vis-NIR and mid-IR spectroscopy, are considered. We conclude that spectroscopy, especially the mid-IR technique in association with Savitzky-Golay smoothing and derivatization and the least squares support vector machine (LS-SVM) algorithm, can be useful in determining SOM quantity and quality. Future research conducted for the standardization of protocols and soil conditions will allow more accurate and reliable results on a global and international scale.

Changes in soil microbial community structure following the abandonment of agricultural terraces in mountainous areas of Eastern Spain

In Eastern Spain, almond trees have been cultivated in terraced orchards for centuries, forming an integral part of the Mediterranean forest scene. In the last decades, orchards have been abandoned due to changes in society. This study investigates effects of changes in land use from forest to agricultural land and the posterior land abandonment on soil microbial community, and the influence of soil physico-chemical properties on the microbial community composition (assessed as abundances of phospholipids fatty acids, PLFA). For this purpose, three land uses (forest, agricultural and abandoned agricultural) at four locations in SE Spain were selected. Multivariate analysis showed a substantial level of differentiation in microbial community structure according to land use. The microbial communities of forest soils were highly associated with soil organic matter content. However, we have not found any physical or chemical soil property capable of explaining the differences between agricultural and abandoned agricultural soils. Thus, it was suggested that the cessation of the perturbation caused by agriculture and shifts in vegetation may have led to changes in the microbial community structure. PLFAs indicative of fungi and ratio of fungal to bacterial PLFAs were higher in abandoned agricultural soils, whereas the relative abundance of bacteria was higher in agricultural soils. Actinomycetes were generally lower in abandoned agricultural soils, while the proportions of vesicular-arbuscular mycorrhyzal fungi were, as a general trend, higher in agricultural and abandoned agricultural soils than in forests. Total microbial biomass and richness increased as agricultural < abandoned agricultural < forest soils.

Direct comparison between visible near- and mid-infrared spectroscopy for describing diuron sorption in soils

Both visible near-infrared (VNIR) and mid-infrared (MIR) spectroscopy have been claimed to better predict pesticide sorption in soils than other methods. We compared the performances of VNIR and MIR spectroscopy for predicting both organic carbon content (foc) and the sorption affinity (Kd) of diuron in 112 surface soils from South Australia. Separate calibration models were developed between VNIR and MIR spectra, and foc and Kd using partial least-squares (PLS) regression. MIR clearly outperformed VNIR for predictions of both foc and Kd in soils. Correlation (R2) and accuracy (RPD) indices were 0.4 and 1.3 for the VNIR-PLS model versus 0.8 and 2.3 for the MIR-PLS model, respectively, for Kd prediction. PLS loadings for sorption prediction were compared in terms of the soil information they contained. While VNIR loading did not include any direct spectral information regarding soil minerals, MIR loading included peaks associated with sand, clays, and carbonates. Perhaps by better predicting foc and integrating the effects of OC as well as minerals, the MIR-PLS model provided a better prediction for diuron Kd values in our calibration set.

Building spectral libraries for wetlands land cover classification and hyperspectral remote sensing

Recent advances in remote sensing provide opportunities to map plant species and vegetation within wetlands at management relevant scales and resolutions. Hyperspectral imagers, currently available on airborne platforms, provide increased spectral resolution over existing space-based sensors that can document detailed information on the distribution of vegetation community types, and sometimes species. Development of spectral libraries of wetland species is a key component needed to facilitate advanced analytical techniques to monitor wetlands. Canopy and leaf spectra at five sites in California, Texas, and Mississippi were sampled to create a common spectral library for mapping wetlands from remotely sensed data. An extensive library of spectra (n=1336) for coastal wetland communities, across a range of bioclimatic, edaphic, and disturbance conditions were measured. The wetland spectral libraries were used to classify and delineate vegetation at a separate location, the Pacheco Creek wetland in the Sacramento Delta, California, using a PROBE-1 airborne hyperspectral data set (5m pixel resolution, 128 bands). This study discusses sampling and collection methodologies for building libraries, and illustrates the potential of advanced sensors to map wetland composition. The importance of developing comprehensive wetland spectral libraries, across diverse ecosystems is highlighted. In tandem with improved analytical tools these libraries provide a physical basis for interpretation that is less subject to conditions of specific data sets. To facilitate a global approach to the application of hyperspectral imagers to mapping wetlands, we suggest that criteria for and compilation of wetland spectral libraries should proceed today in anticipation of the wider availability and eventual space-based deployment of advanced hyperspectral high spatial resolution sensors.

Evaluation of the nutrient metal content in Chinese animal manure compost using near infrared spectroscopy (NIRS)

This study explored the feasibility of determining the content of several nutrient metals (K, Ca, Mg, Fe and Zn) in animal manure compost products in China using near infrared spectroscopy (NIRS). Samples of 120 compost products were collected from 22 provinces in China. The spectra were scanned obtained with a FT-NIRS system. For fresh samples, the validation coefficient of determination (r2), standard error of prediction (SEP) and the ratio of the standard deviation in the validation set to the standard error of prediction (RPD) were 0.69, 5.08g/kg and 1.83 for K, 0.46, 17.99g/kg and 1.33 for Ca, 0.54, 1.73g/kg and 1.42 for Mg, 0.86, 1.14g/kg and 2.55 for Fe and 0.65, 62.95mg/kg and 1.71 for Zn, respectively. For dried samples, the r2, SEP and RPD were 0.68, 5.68g/kg and 1.79 for K, 0.78, 15.34g/kg and 2.12 for Ca, 0.72, 2.07g/kg and 1.81 for Mg, 0.84, 1.49g/kg and 2.27 for Fe and 0.57, 86.32mg/kg and 1.52 for Zn, respectively. The results showed that the NIRS technique is a potential method for predicting nutrient metal content of animal manure compost products. However, further research is needed to improve the prediction precision of calibration models by enlarging the number of samples and using other chemometric methods.

Near infrared spectroscopy for determination of various physical, chemical and biochemical properties in Mediterranean soils

The potential of near infrared (NIR) reflectance spectroscopy to predict various physical, chemical and biochemical properties in Mediterranean soils from SE Spain was evaluated. Soil samples (n=393) were obtained by sampling thirteen locations during three years (2003-2005 period). These samples had a wide range of soil characteristics due to variations in land use, vegetation cover and specific climatic conditions. Biochemical properties also included microbial biomarkers based on phospholipid fatty acids (PLFA). Partial least squares (PLS) regression with cross validation was used to establish relationships between the NIR spectra and the reference data from physical, chemical and biochemical analyses. Based on the values of coefficient of determination (r(2)) and the ratio of standard deviation of validation set to root mean square error of cross validation (RPD), predicted results were evaluated as excellent (r(2)>0.90 and RPD>3) for soil organic carbon, Kjeldahl nitrogen, soil moisture, cation exchange capacity, microbial biomass carbon, basal soil respiration, acid phosphatase activity, β-glucosidase activity and PLFA biomarkers for total bacteria, Gram positive bacteria, actinomycetes, vesicular-arbuscular mycorrhizal fungi and total PLFA biomass. Good predictions (0.81<r(2)<0.90 and 2.5<RPD<3) were obtained for exchangeable calcium and magnesium, water soluble carbon, water holding capacity and urease activity. Resultant models for protozoa and fungi were not accurate enough to satisfactorily estimate these variables, only permitting approximate predictions (0.66<r(2)<0.80 and 2.0<RPD<2.5). Electrical conductivity, pH, exchangeable phosphorus and sodium, metabolic quotient and Gram negative bacteria were poorly predicted (r(2)<0.66 and RPD<2). Thus, the results obtained in this study reflect that NIR reflectance spectroscopy could be used as a rapid, inexpensive and non-destructive technique to predict some physical, chemical and biochemical soil properties for Mediterranean soils, including variables related to the composition of the soil microbial community composition.