Search

Visible and near-infrared (vis-NIR) spectroscopy has proven to be a straightforward method for sample preparation and scaling soil testing, while the increasing availability of high-resolution remote sensing (RS) data has further facilitated the understanding of spatial variability in soil organic carbon (SOC) and total nitrogen (TN) across landscapes. However, the impact of combining vis-NIR spectroscopy with high-resolution RS data for SOC and TN prediction remains an open question. This study evaluated the effects of incorporating a high-resolution LiDAR-derived digital elevation model (DEM) and a medium-resolution SRTM-derived DEM with vis-NIR spectroscopy for predicting SOC and TN in peatlands. A total of 57 soil cores, comprising 262 samples from various horizons (<2 m), were collected and analysed for SOC and TN content using traditional methods and ASD Fieldspec® 4. The 262 observations, along with elevation data from LiDAR and SRTM, were divided into 80% training and 20% testing datasets. By employing the Cubist modelling approach, the results demonstrated that incorporating high-resolution LiDAR data with vis-NIR spectra improved predictions of SOC (RMSE: 4.60%, RPIQ: 9.00) and TN (RMSE: 3.06 g kg−1, RPIQ: 7.05). In conclusion, the integration of LiDAR and soil spectroscopy holds significant potential for enhancing soil mapping and promoting sustainable soil management. [ABSTRACT FROM AUTHOR]

The number of samples used in the calibration data set affects the quality of the generated predictive models using visible, near and shortwave infrared (VIS–NIR–SWIR) spectroscopy for soil attributes. Recently, the convolutional neural network (CNN) has been regarded as a highly accurate model for predicting soil properties on a large database. However, it has not yet been ascertained how large the sample size should be for CNN model to be effective. This paper investigates the effect of the training sample size on the accuracy of deep learning and machine learning models. It aims at providing an estimate of how many calibration samples are needed to improve the model performance of soil properties predictions with CNN as compared to conventional machine learning models. In addition, this paper also looks at a way to interpret the CNN models, which are commonly labelled as a black box. It is hypothesised that the performance of machine learning models will increase with an increasing number of training samples, but it will plateau when it reaches a certain number, while the performance of CNN will keep improving. The performances of two machine learning models (partial least squares regression – PLSR; Cubist) are compared against the CNN model. A VIS–NIR–SWIR spectra library from Brazil, containing 4251 unique sites with averages of two to three samples per depth (a total of 12 044 samples), was divided into calibration (3188 sites) and validation (1063 sites) sets. A subset of the calibration data set was then created to represent a smaller calibration data set ranging from 125, 300, 500, 1000, 1500, 2000, 2500 and 2700 unique sites, which is equivalent to a sample size of approximately 350, 840, 1400, 2800, 4200, 5600, 7000 and 7650. All three models (PLSR, Cubist and CNN) were generated for each sample size of the unique sites for the prediction of five different soil properties, i.e. cation exchange capacity, organic carbon, sand, silt and clay content. These calibration subset sampling processes and modelling were repeated 10 times to provide a better representation of the model performances. Learning curves showed that the accuracy increased with an increasing number of training samples. At a lower number of samples (< 1000), PLSR and Cubist performed better than CNN. The performance of CNN outweighed the PLSR and Cubist model at a sample size of 1500 and 1800, respectively. It can be recommended that deep learning is most efficient for spectra modelling for sample sizes above 2000. The accuracy of the PLSR and Cubist model seems to reach a plateau above sample sizes of 4200 and 5000, respectively, while the accuracy of CNN has not plateaued. A sensitivity analysis of the CNN model demonstrated its ability to determine important wavelengths region that affected the predictions of various soil attributes. [ABSTRACT FROM AUTHOR]

Soil maps at regional and farm levels are vital for the best management agricultural practices (BMAP). The soil is the substrate for plant growth and essential to ensure food security. In this context, soil maps require a detailed cartographic scale for BMAP. This study (i) investigated the use of digital soil mapping (DSM) products, such as soil chemical and physical attributes, indices, mineralogy, and properties to extrapolate late soil survey maps at 1:20,000 scale; (ii) created the digital yield environment for sugarcane based on the DSM products; and (iii) evaluated qualitatively the predict soil maps and relationship with previous studies and the predicted yield environment. The region of interest covers eight municipalities and almost 2598 km2 in São Paulo State, Brazil. The soil survey at farm level conducted covered almost 86.52 km2, ∼3.33% of the total area (96.67% of the unmapped area). We created a point grid (centroid) with the same spatial resolution (30 m) of the rasters used as covariates for soil mapping unit (SMU) predictions. This grid intended to retrieve the representative soil mapping unit of each geometric polygon. It was retrieved 117,413 points representing 27 SMU of seven soil orders at a first categorical level, according to the Brazilian Classification System, and seven yield environment for sugarcane production. SMU predictions and their respective soil orders were performed using the random forest machine learning regression method. The level of association between SMU and yield environments was 0.34 (α = 0.01) by the Cramer's V coefficient with a very strong relationship. Our approach could provide the first digital yield environment for sugarcane based on the DSM products. Furthermore, a qualitative evaluation of our framework was substantiated with previous research in the same study site. This framework could be replicated and fulfil the need for DSM at regional and farm levels for policy-makers and farmers. [Display omitted] • Soil mapping units (SMU) and yield environments had very strong association. • Digital soil mapping (DSM) products extrapolated late soil survey maps. • DSM products were used to create digital yield environment for sugarcane. • DSM products worked effectively as predictors of the SMU and yield environment. [ABSTRACT FROM AUTHOR]

Soil contamination by potentially toxic elements (PTEs) is one of the greatest threats to environmental degradation. Knowing where PTEs accumulated in soil can mitigate their adverse effects on plants, animals, and human health. We evaluated the potential of using long-term remote sensing images that reveal the bare soils, to detect and map PTEs in agricultural fields. In this study, 360 soil samples were collected at the superficial layer (0–20 cm) in a 2574 km2 agricultural area located in São Paulo State, Brazil. We tested the Soil Synthetic Image (SYSI) using Landsat TM/ETM/ETM+, Landsat OLI, and Sentinel 2 images. The three products have different spectral, temporal, and spatial resolutions. The time series multispectral images were used to reveal areas with bare soil and their spectra were used as predictors of soil chromium, iron, nickel, and zinc contents. We observed a strong linear relationship (−0.26 > r > −0.62) between the selected PTEs and the near infrared (NIR) and shortwave infrared (SWIR) bands of Sentinel (ensemble of 4 years of data), Landsat TM (35 years data), and Landsat OLI (4 years data). The clearest discrimination of soil PTEs was obtained from SYSI using a long term Landsat 5 collection over 35 years. Satellite data could efficiently detect the contents of PTEs in soils due to their relation with soil attributes and parent materials. Therefore, distinct satellite sensors could map the PTEs on tropics and assist in understanding their spatial dynamics and environmental effects. [Display omitted] • Bare soil pixels were able to retrieve potential toxic elements from space. • Landsat and Sentinel presented important data to detect topsoil toxic elements on soils. • Long term images generated better spectra to detect toxic element. • Spatial dynamics of toxic elements in agriculture areas are presented. [ABSTRACT FROM AUTHOR]

Soil organic carbon (SOC) stocks are a remarkable property for soil and environmental monitoring. The understanding of their dynamics in crop soils must go forward. The objective of this study was to determine the impact of temporal environmental controlling factors obtained by satellite images over the SOC stocks along soil depth, using machine learning algorithms. The work was carried out in São Paulo state (Brazil) in an area of 2577 km2. We obtained a dataset of boreholes with soil analyses from topsoil to subsoil (0–100 cm). Additionally, remote sensing covariates (30 years of land use history, vegetation indexes), soil properties (i.e., clay, sand, mineralogy), soil types (classification), geology, climate and relief information were used. All covariates were confronted with SOC stocks contents, to identify their impact. Afterwards, the abilities of the predictive models were tested by splitting soil samples into two random groups (70 for training and 30% for model testing). We observed that the mean values of SOC stocks decreased by increasing the depth in all land use and land cover (LULC) historical classes. The results indicated that the random forest with recursive features elimination (RFE) was an accurate technique for predicting SOC stocks and finding controlling factors. We also found that the soil properties (especially clay and CEC), terrain attributes, geology, bioclimatic parameters and land use history were the most critical factors in controlling the SOC stocks in all LULC history and soil depths. We concluded that random forest coupled with RFE could be a functional approach to detect, map and monitor SOC stocks using environmental and remote sensing data. [ABSTRACT FROM AUTHOR]

Soil minerals are usually quantified by the conventional laboratory soil analyses. However, developments in interpretations and analyses of the visible, near-infrared, and short-infrared (Vis-NIR-SWIR) diffuse reflectance have allowed the quantification of some soil minerals. In this study, we aimed to implement a novel framework using Vis-NIR-SWIR spectroscopy to quantify the main soil minerals. We also assessed the application of this framework to create new environmental variables for digital soil mapping (DSM). The soil spectra database comprised 2701 samples from 1008 sites in the spectral range of 350–2500 nm at 0–20, 40–60, and 80–100 cm depths. The specific bands in the Vis-NIR-SWIR spectra that identify the presence of soil mineral were selected based on the literature with the United States Geological Survey Spectral Library Version 7 and in the strong maxima and minima of the second-derivative curves of the soil mineral standards using the Savitzky-Golay method. We proposed an estimation and conversion of the measurement unit of soil minerals in amplitude to g kg−1 using a small dataset of mineral content quantified via X-Ray Powder Diffraction. We selected randomly 85 samples out of 2701 available at 0–20 cm depth and sent to conventional laboratory analyses to calibrate the final estimation, using the kaolinite soil mineral as an example. Therefore, a constant factor was determined to estimate mineral content in soils displaying RMSE, R2 adj , the Lin's concordance coefficient (CCC), Bias, and RPIQ values of 7612 g kg−1, 0.28, 0.50, 13.09 g kg−1, and 0.56, respectively. This evaluation was assessed by splitting 85 samples into 80% to determine and 20% to validate the constant factor. For the DSM procedure, we used 2701 samples split into 80% and 20% for calibration and validation, respectively, of the models for each of the nine minerals. This study showed that the proposed framework using Vis-NIR-SWIR spectroscopy to estimate soil minerals is promising due to higher CCC and lower RMSE values obtained. Furthermore, the spectral amplitude for each mineral provides important information to be used as environmental variables for the prediction of soil attributes, soil types, and soil properties. • Soil mineral content estimate was suggested based on XRD and spectroscopy. • Diffuse reflectance spectroscopy quantified the soil minerals. • Integrating remote and proximal sensing helped to characterise soil mineralogy. [ABSTRACT FROM AUTHOR]

Soil mapping demands large-scale surveys that are costly and time consuming. It is necessary to identify strategies with reduced costs to obtain detailed information for soil mapping. We aimed to compare multispectral satellite image and relief parameters for the quantification and mapping of clay and sand contents. The Temporal Synthetic Spectral (TESS) reflectance and Synthetic Soil Image (SYSI) approaches were used to identify and characterize texture spectral signatures at the image level. Soil samples were collected (0–20 cm depth, 919 points) from an area of 14,614 km2 in Brazil for reference and model calibration. We compared different prediction approaches: (a) TESS and SYSI; (b) Relief-Derived Covariates (RDC); and (c) SYSI plus RDC. The TESS method produced highly similar behavior to the laboratory convolved data. The sandy textural class showed a greater increase in average spectral reflectance from Band 1 to 7 compared with the clayey class. The prediction using SYSI produced a better result for clay (R2 = 0.83; RMSE = 65.0 g kg−1) and sand (R2 = 0.86; RMSE = 79.9 g kg−1). Multispectral satellite images were more stable for the identification of soil properties than relief parameters. [ABSTRACT FROM AUTHOR]