The successful determination of soil properties by visible and near-infrared (Vis–NIR) reflectance spectroscopy (350–2500nm) depends on the selection of an appropriate multivariate calibration ...technique. In this study, four multivariate techniques (principal components regression, PCR; partial least squares regression, PLSR; back-propagation neural network, BPNN; and support vector machine regression, SVMR) were compared with the aim of rapidly and accurately predicting soil properties, including soil organic matter (SOM), total nitrogen (TN), total phosphorus (TP), and total potassium (TK). A total of 148 intact soil cores (8.4cm internal diameter and 40cm long) collected from paddy fields in Yujiang, China were used as the dataset for the calibration-validation procedure. The Vis–NIR spectra were measured on flat, horizontal surfaces of soil core sections at four depths (i.e., 5, 10, 15, and 20cm) in the laboratory. The coefficient of determination (R2), root mean square error (RMSE), and residual prediction deviation (RPD) were used to evaluate the accuracy of the calibration models. Both the cross-validation and independent validation data sets showed that the SVMR models outperformed the BPNN, PCR, and PLSR models for SOM, TN, and TP predictions, whereas BPNN outperformed the other models for TK. Furthermore, BPNN and SVMR provided better performance than PCR and PLSR. The best predictions were obtained by the SVMR model for SOM (R2P=0.88; RMSEP=4.87; RPDP=2.84) and TN (R2P=0.86; RMSEP=0.31; RPDP=2.69), which were classified as good model predictions. The predictions of TP (R2P=0.76; RMSEP=0.080; RPDP=2.03) by SVMR were approximately quantitative predictions, whereas the TK (R2P=0.65; RMSEP=3.54; RPDP=1.65) prediction with BPNN was unsuccessful. Vis–NIR spectroscopy combined with SVMR has great potential to accurately determine the selected soil properties of intact soil cores of paddy fields.
•We tested Vis–NIR to predict soil properties of intact soil cores of paddy fields.•PCR, PLSR, BPNN, and SVMR were compared to predict SOM, TN, TP, and TK.•SOM, TN, and TP are best predicted by SVMR and TK is best predicted by BPNN.•BPNN and SVMR models perform better than PCR and PLSR models.•Vis–NIR spectroscopy combined with SVMR technique is recommended.
Land-use/land-cover changes (LUCCs) have links to both human and nature inter- actions. China's Land-Use/cover Datasets (CLUDs) were updated regularly at 5-year inter- vals from the late 1980s to ...2010, with standard procedures based on Landsat TM/ETM+ im- ages. A land-use dynamic regionalization method was proposed to analyze major land-use conversions. The spatiotemporal characteristics, differences, and causes of land-use changes at a national scale were then examined. The main findings are summarized as fol- lows. Land-use changes (LUCs) across China indicated a significant variation in spatial and temporal characteristics in the last 20 years (1990-2010). The area of cropland change de- creased in the south and increased in the north, but the total area remained almost un- changed. The reclaimed cropland was shifted from the northeast to the northwest. The built-up lands expanded rapidly, were mainly distributed in the east, and gradually spread out to central and western China. Woodland decreased first, and then increased, but desert area was the opposite. Grassland continued decreasing. Different spatial patterns of LUC in China were found between the late 20th century and the early 21st century. The original 13 LUC zones were replaced by 15 units with changes of boundaries in some zones. The main spatial characteristics of these changes included (1) an accelerated expansion of built-up land in the Huang-Huai-Hai region, the southeastern coastal areas, the midstream area of the Yangtze River, and the Sichuan Basin; (2) shifted land reclamation in the north from northeast China and eastern Inner Mongolia to the oasis agricultural areas in northwest China; (3) continuous transformation from rain-fed farmlands in northeast China to paddy fields; and (4) effective- ness of the "Grain for Green" project in the southern agricultural-pastoral ecotones of Inner Mongolia, the Loess Plateau, and southwestern mountainous areas. In the last two decades, although climate change in the north affected the change in cropland, policy regulation and economic driving forces were still the primary causes of LUC across China. During the first decade of the 21st century, the anthropogenic factors that drove variations in land-use pat- terns have shifted the emphasis from one-way land development to both development and conservation. The "dynamic regionalization method" was used to analyze changes in the spatial patterns of zoning boundaries, the internal characteristics of zones, and the growth and decrease of units. The results revealed "the pattern of the change process," namely the process of LUC and regional differences in characteristics at different stages. The growth and decrease of zones during this dynamic LUC zoning, variations in unit boundaries, and the characteristics of change intensities between the former and latter decades were examined. The patterns of alternative transformation between the "pattern" and "process" of land use and the causes for changes in different types and different regions of land use were explored.
Soil iron (Fe) performs vital functions in the biogeochemical cycles of soil environments. The amount and profile allocation of various Fe parameters can be used as sensitive indicators of soil ...development and pedogenic processes. This study aimed to evaluate the potential of ground‐based hyperspectral imaging (HSI: 400–1010 nm) spectroscopy to predict and map six Fe parameters indicative of pedogenic processes: total Fe (Fet), dithionite‐citrate‐bicarbonate (DCB)‐extracted Fe (Fed), oxalate‐extracted Fe (Feo), weathering index (FeW), active ratio (FeA) and crystallinity ratio (FeC). In total, 17 intact soil profiles at a depth of 100 ± 5 cm were collected to acquire HSI images. Four non‐linear machine learning techniques, namely, random forest (RF), XGBoost, CatBoost and support vector machine regression (SVMR), were implemented and compared with linear partial least squares (PLS) to identify the models with the best performance for different soil Fe parameters. Our results indicate that the four non‐linear machine learning models outperformed PLS for most soil Fe parameters, with low root mean square error (RMSE) and high Lin's concordance correlation coefficient (LCCC) values. Based on the testing set, SVMR showed better performance over the other tested models for Fet (RMSEP = 2.645 g kg−1, LCCCP = 0.89), Fed (RMSEP = 0.972 g kg−1, LCCCP = 0.97), Feo (RMSEP = 0.273 g kg−1, LCCCP = 0.97), FeW (RMSEP = 0.035, LCCCP = 0.97), FeA (RMSEP = 0.033, LCCCP = 0.97) and FeC (RMSEP = 0.031, LCCCP = 0.97). According to the LCCCP values, soil Fet was predicted to be in substantial agreement by SVMR, and the other soil Fe predictions were considered to be in near perfect agreement. Moreover, SVMR required lower computational costs. Given these results, the combination of HSI spectroscopy and SVMR is recommended due to its more reliable estimation and profile mapping of the selected soil Fe parameters than that of PLS, RF, XGBoost and CatBoost.
Highlights
Hyperspectral imaging was used to map six soil Fe parameters of intact profiles.
Nonlinear machine learning models were compared to select the most suitable model.
Nonlinear techniques outperformed the linear PLS model in most cases.
SVMR showed higher comprehensive performance than other models.
Hyperspectral imaging combined with SVMR is recommended to map soil Fe in profiles.
Assessment and monitoring of soil organic matter (SOM) quality are important for understanding SOM dynamics and developing management practices that will enhance and maintain the productivity of ...agricultural soils. Visible and near-infrared (Vis-NIR) diffuse reflectance spectroscopy (350-2500 nm) has received increasing attention over the recent decades as a promising technique for SOM analysis. While heterogeneity of sample sets is one critical factor that complicates the prediction of soil properties from Vis-NIR spectra, a spectral library representing the local soil diversity needs to be constructed. The study area, covering a surface of 927 km2 and located in Yujiang County of Jiangsu Province, is characterized by a hilly area with different soil parent materials (e.g., red sandstone, shale, Quaternary red clay, and river alluvium). In total, 232 topsoil (0-20 cm) samples were collected for SOM analysis and scanned with a Vis-NIR spectrometer in the laboratory. Reflectance data were related to surface SOM content by means of a partial least square regression (PLSR) method and several data pre-processing techniques, such as first and second derivatives with a smoothing filter. The performance of the PLSR model was tested under different combinations of calibration/validation sets (global and local calibrations stratified according to parent materials). The results showed that the models based on the global calibrations can only make approximate predictions for SOM content (RMSE (root mean squared error) = 4.23-4.69 g kg-1; R2 (coefficient of determination) = 0.80-0.84; RPD (ratio of standard deviation to RMSE) = 2.19-2.44; RPIQ (ratio of performance to inter-quartile distance) = 2.88-3.08). Under the local calibrations, the individual PLSR models for each parent material improved SOM predictions (RMSE = 2.55-3.49 g kg-1; R2 = 0.87-0.93; RPD = 2.67-3.12; RPIQ = 3.15-4.02). Among the four different parent materials, the largest R2 and the smallest RMSE were observed for the shale soils, which had the lowest coefficient of variation (CV) values for clay (18.95%), free iron oxides (15.93%), and pH (1.04%). This demonstrates the importance of a practical subsetting strategy for the continued improvement of SOM prediction with Vis-NIR spectroscopy.
•Annual soil erosion rates decreased with tree species richness.•Tree diversity reduced soil erosion by affecting tree canopy and biological soil crust development.•Restoring species-rich plantations ...may be beneficial for soil erosion control.
Biodiversity plays a crucial role in forest ecosystem sustainability. However, it is unclear how tree diversity and especially the relationship between diversity and ecosystem functioning affect soil erosion. Based on a forest biodiversity and ecosystem functioning experiment established in subtropical China (BEF China), we measured soil erosion at four tree species richness levels (monocultures, 8 tree species, 16 tree species and 24 species stands) during the rainy seasons from 2013 to 2015. The result showed that mean annual soil erosion rates were detected to decrease with tree species richness significantly over the observed three years. Leaf area index (LAI) and biological soil crusts (BSCs) were the two main factors driving soil erosion within tree stands of different species richness. Positive effects of tree species richness on tree canopy structure and BSCs might drive the reduction of soil erosion in the earlier successional stage after afforestation of tree plantations. Therefore, we highlight the important influence of tree species richness on soil erosion control, hydrologic processes and thus sustainable ecology services.
Information and knowledge about the spatial variability and multi-scale sources of soil heavy metal variability are important for risk assessment, soil remediation, as well as effective management ...recommendations. Topsoil samples (0–20
cm) (
n
=
330) were collected from Luhe County, China. Geostatistical multivariate factorial kriging with robust estimates of variograms, spatial overly analysis and indicator kriging were applied to explore the correlations among soil heavy metals (Cu, Zn, Pb, Cr, Ni, Cd, and Hg) across different spatial scales, identify the sources of spatial variability, and evaluate the potential risk of soil contamination. High variability of soil heavy metals was observed and the correlations among the selected soil heavy metals in the study area depend on spatial scale. Copper, Zn, Cr, and Ni at short-range scale (2
km) are mainly controlled by geology (bed rock), while land use greatly affected the long-range variations of Cd and Hg. The strong correlations between Cr and Ni at short- and long-range (11
km) scale indicate that the sources of Cr and Ni are predominately geochemical. However, anthropogenic activities had profound impacts on the concentrations of Cu, Zn, Pb, Cd, and Hg. The potential contamination risk, defined here as measured soil heavy metal concentrations which exceeded the background values of Chinese Environmental Quality Standard for Soils, was observed in Cu, Zn and Pb, and the areas covered 3%, 3.4%, and 1.4%, respectively. The possible sources of Cu and Zn contamination can be considered as the current and historic industrial emissions from Dachang industrial zone, and the potential areas of Pb contamination could be attributed to the industrial emissions from Dachang industrial zone, small factories and heavy traffic flows in/nearby downtown area of the county. While the Hg contamination risk was closely related to the widely use of Hg-related pesticide during 1940–1970s.
•We assess the degradation of farming terraces in the Three Gorges ecosystem.•We combine field mapping and random forests successfully.•Indicators for anthropogenic effects and natural drivers were ...combined.•Terrace degradation distinctly ranges from well maintained to completely collapsed.•Distances to roads, settlements, and rivers explain the varying terrace conditions.
Due to resettlements, construction of new infrastructure, and new land reclamation the rapid agricultural changes in the Three Georges Area (TGA) in Central China are expected to force the degradation of the cultivated terraced landscape. Consequently, increased soil erosion can hamper a sustainable land management in the mountainous TGA. This paper presents the model framework TerraCE (Terrace Condition Erosion) for determining the causes for different terrace conditions and terrace degradation based on field surveys and spatial data mining. For a total of 987 bench terrace plots in the Xiangxi catchment we collected data on their state of maintenance and terrace design to account for terrace stability and thus capability of soil conservation. Assessing the driving factors of terrace degradation was done by embedding terrain-based predictors and distance-transforms of remote-sensing data as indicators of environmental and anthropogenic influences. Random forests classification and regression models were applied for data mining. Terrace degradation in the Xiangxi catchment is obvious. The sequence of degradation ranges from ‘well maintained’ (21%), ‘fairly maintained’ (44%), and ‘partially collapsed’ (23%) to ‘completely collapsed’ (11%) terraces. The cross-validation error of the supervised TerraCE model is below 8%, allowing for reasonable and valid interpretations of the causes of terrace degradation. Data mining reveals indicators for anthropogenic effects such as the distance to settlements or to roads as major drivers for the spatial distribution of terrace conditions. The effect of relief, which can be regarded as the major natural driver for terrace degradation by erosive action is tributary but altered and overlaid by land use dynamics associated with the Three Gorges Dam. An important indicator representing a combined effect of terrain and human activity is the distance to stream channels of different orders. Applying TerraCE we reveal mechanisms of terrace degradation in disturbed environments and present a framework for standardized mapping and analysis of terrace degradation under cultivation. The approach might also be used to develop guidelines for terrace planning in mountainous terraced landscapes of limited carrying capacity, with respect to socio-economic as well as environmental conditions.
Understanding the impacts of climate change and agricultural management practices on soil organic carbon (SOC) dynamics is critical for implementing optimal farming practices and maintaining ...agricultural productivity. This study examines the influence of climatic variables and agricultural management on carbon sequestration potentials in Tai-Lake Paddy soils of China using the DeNitrification-DeComposition (DNDC, version 9.1) model, with a high-resolution soil database (1:50,000). Model simulations considered the effects of no-tillage, the application rates of manure, N fertilization, and crop residue, water management, and changes in temperature and precipitation. We found that the carbon sequestration potential in the top soils (0–30cm) for the 2.32Mha paddy soils of the Tai-Lake region varied from 4.71 to 44.31Tg C under the feasible management practices during the period of 2001–2019. The sequestration potential significantly increased with increasing application of N-fertilizer, manure, conservation tillage, and crop residues, with an annual average SOC changes ranged from 107 to 121kgCha−1yr−1, 159 to 326kgCha−1yr−1, 78 to 128kgCha−1yr−1, and 489 to 1005kgCha−1yr−1, respectively. Toward mitigating greenhouse emissions and N losses, no-tillage and increase of crop residue return to soils as well as manure application are recommended for agricultural practice in this region. Our analysis of climate impacts on SOC sequestration suggests that the rice paddies in this region will continue to be a carbon sink under future warming conditions. Specifically, with rising air temperature of 2.0°C and 4°C, the average annual SOC changes were 52 and 21kgCha−1yr−1, respectively.
•High-resolution soil database was used to drive DeNitrification–DeComposition model.•The C sink varied from 4.71 to 44.31Tg C under the feasible management practices.•No-tillage, increasing crop residue and manure application significantly increase SOC.•The region remains C sink even if the temperature increases.
As limited resources, soils are the largest terrestrial sinks of organic carbon. In this respect, 3D modelling of soil organic carbon (SOC) offers substantial improvements in the understanding and ...assessment of the spatial distribution of SOC stocks. Previous three-dimensional SOC modelling approaches usually averaged each depth increment for multi-layer two-dimensional predictions. Therefore, these models are limited in their vertical resolution and thus in the interpretability of the soil as a volume as well as in the accuracy of the SOC stock predictions. So far, only few approaches used spatially modelled depth functions for SOC predictions. This study implemented and evaluated an approach that compared polynomial, logarithmic and exponential depth functions using non-linear machine learning techniques, i.e. multivariate adaptive regression splines, random forests and support vector machines to quantify SOC stocks spatially and depth-related in the context of biodiversity and ecosystem functioning research. The legacy datasets used for modelling include profile data for SOC and bulk density (BD), sampled at five depth increments (0-5, 5-10, 10-20, 20-30, 30-50 cm). The samples were taken in an experimental forest in the Chinese subtropics as part of the biodiversity and ecosystem functioning (BEF) China experiment. Here we compared the depth functions by means of the results of the different machine learning approaches obtained based on multi-layer 2D models as well as 3D models. The main findings were (i) that 3rd degree polynomials provided the best results for SOC and BD (R2 = 0.99 and R2 = 0.98; RMSE = 0.36% and 0.07 g cm-3). However, they did not adequately describe the general asymptotic trend of SOC and BD. In this respect the exponential (SOC: R2 = 0.94; RMSE = 0.56%) and logarithmic (BD: R2 = 84; RMSE = 0.21 g cm-3) functions provided more reliable estimates. (ii) random forests with the exponential function for SOC correlated better with the corresponding 2.5D predictions (R2: 0.96 to 0.75), compared to the 3rd degree polynomials (R2: 0.89 to 0.15) which support vector machines fitted best. We recommend not to use polynomial functions with sparsely sampled profiles, as they have many turning points and tend to overfit the data on a given profile. This may limit the spatial prediction capacities. Instead, less adaptive functions with a higher degree of generalisation such as exponential and logarithmic functions should be used to spatially map sparse vertical soil profile datasets. We conclude that spatial prediction of SOC using exponential depth functions, in conjunction with random forests is well suited for 3D SOC stock modelling, and provides much finer vertical resolutions compared to 2.5D approaches.
Display omitted
•Soil C and aggregate properties were investigated after land-use change.•The PVF had higher SOC, MBC and aggregate-associated C than the OVF.•A greater decrease in O-alkyl C occurred ...in the OVF than in the PVF.•SOC fractions had obvious effects on soil aggregate formation.•Soil-aggregate C was positively affected by the SOC and MBC contents.
Conversion of rice-wheat rotation (RWR) to conventional vegetable cultivation, especially vegetable cultivation in plastic sheds, readily causes soil structure degradation, and organic manure is commonly applied to mitigate such degradation. Although many studies have focused on changes in SOC properties or structural soil parameters, additional details about the changes, such as the relationships between SOC fractions and soil aggregates during unique land-use changes, are unknown. Here, we studied the changes in soil aggregation, and SOC parameters in 14-year-old plastic-shed vegetable fields (PVFs) and open-air vegetable fields (OVFs) covered with organic manure, with adjacent RWR fields serving as a control. The vegetable fields were converted from RWR fields. SOC fractions were analyzed by using 13C solid-state NMR spectroscopy, and the aggregate classes were divided as follows: large macro-aggregates (>2 mm), small macro-aggregates (2–0.25 mm), micro-aggregates (0.25–0.053 mm), and silt and clay particles (<0.053 mm). PVFs had higher SOC and MBC than OVFs and RWR fields. In terms of the SOC fractions, the proportion of O-alkyl C decreased as RWR (53.90%) > PVF (43.21%) > OVF (35.37%), in contrast to the trends in the carbonyl C and aromatic C fractions. The aggregate-associated C, especially that associated with large macro-aggregates and micro-aggregates, in the PVFs was highest among the treatments. The O-alkyl C fraction in large macro-aggregates decreased in the order RWR (44.67%) > OVF (35.76%) > PVF (32.40%), consistent with the results for micro-aggregates. Furthermore, there was a significant (P ≤ 0.05) positive relationship between macro-aggregates (>0.25 mm) and the active C fractions, in contrast to the relationships with stable fractions. It suggested that plastic-shed cultivation might have a more positive effect than open-air cultivation on soil structure and the carbon stock when large amounts of organic manure are applied; moreover, the SOC quantity and quality affect soil aggregation differently.