Next Article in Journal
Identification of a Protein Content-Associated Gene in Peanut (Arachis hypogaea) and Functional Characterization of the Gene and Its Homolog in Arabidopsis thaliana
Previous Article in Journal
Novel Insights into Botanical Seed Production of Xanthosoma spp. in Cuba
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Agronomy
Volume 15
Issue 6
10.3390/agronomy15061367
agronomy-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Identification of Mattic Epipedon Degradation on the Northeastern Qinghai–Tibetan Plateau Using Hyperspectral Data

by
Junjun Zhi
1,
Hong Zhu
1,
Jingwen Yang
1,
Qiuchen Yan
1,
Dandan Zhi
2,
Zhongbao Sun
1,*,
Liangwei Ge
1 and
Chengwen Lv
1
1
School of Geography and Tourism, Anhui Normal University, Wuhu 241002, China
2
Shanghai Institute of Satellite Engineering, Shanghai 200240, China
*
Author to whom correspondence should be addressed.
Agronomy 2025, 15(6), 1367; https://doi.org/10.3390/agronomy15061367
Submission received: 27 April 2025 / Revised: 30 May 2025 / Accepted: 30 May 2025 / Published: 2 June 2025
(This article belongs to the Section Agroecology Innovation: Achieving System Resilience)
Browse Figures
Versions Notes

Abstract

:
Accurate identification of mattic epipedon degradation is critically important for addressing ecological issues such as the weakening of alpine grassland carbon sink capacity and reduced soil and water conservation. However, efficient and rapid methods for its detection remain limited. This study aimed to clarify the hyperspectral response mechanisms of mattic epipedon degradation and, based on hyperspectral technology, to construct models for identifying degraded mattic epipedon and screen preprocessing methods suitable for different moisture conditions. The results showed the following: (1) The XGBoost model with preprocessing using multiplicative scatter correction combined with second derivative transformation (MSC+SD) performed best, achieving an identification accuracy and Kappa coefficient of 0.85 and 0.82, respectively. The characteristic bands were concentrated in the visible light range (446–450 nm) and short-wave infrared range (2134 nm, 2267–2269 nm), which are closely related to the spectral responses of organic carbon and mineral components. (2) Spectral reflectance was significantly negatively correlated with moisture content, and model accuracy decreased as moisture content increased. (3) After correction using the EPO algorithm, the model accuracy for the high-moisture group improved by 13.2–16.7%, whereas that for the low-moisture group (<15%) decreased by 7.5%, verifying 15% moisture content as the critical threshold for water interference. This study elucidated the impact mechanism of moisture on the hyperspectral characteristics of the mattic epipedon. The established MSC+SD-XGBoost model adapts to varying moisture conditions, providing technical support for the rapid monitoring of mattic epipedon degradation and holding significant practical value for carbon management in alpine ecosystems.

1. Introduction

The Qinghai–Tibet Plateau, often referred to as the “Roof of the World” and the “Water Tower of Asia,” plays a pivotal role in the global ecological landscape. Its alpine meadow ecosystem serves as a crucial platform for climate regulation, carbon sequestration, and soil and water conservation [1,2]. However, in recent decades, this ecosystem has experienced increasing degradation driven by global environmental changes—such as atmospheric nitrogen deposition and altered precipitation patterns—compounded by anthropogenic pressures including overgrazing and intensified land use. As a result, approximately one-third of the alpine meadows now exhibit signs of compound degradation, characterized by vegetation dwarfism and soil carbon loss [3,4,5,6]. Among the most significant indicators of this degradation is the mattic epipedon, a diagnostic soil horizon unique to alpine meadows. Its degradation is accompanied by the depletion of soil organic carbon pools and the breakdown of root system networks, ultimately disrupting the regional carbon balance. Consequently, the mattic epipedon has been widely recognized as a key diagnostic indicator of alpine meadow ecosystem degradation [7,8]. The degradation of the mattic epipedon is commonly attributed to inappropriate human activities, such as overgrazing [9,10,11], climate change [12,13,14], or their combined effects [15,16,17]. The classification of non-degraded and degraded mattic epipedon is based on its developmental types and grading criteria [18]. Typically, non-degraded mattic epipedon is characterized by high species richness, dominated by Kobresia pygmaea meadows, with vegetation cover ranging from 85% to 90%. Surface fissures are occasionally observed, but signs of erosion and rodent activity are absent. In contrast, degraded mattic epipedons are dominated by Salix alpine and Kobresia pygmaea, with frequent surface fissures and a high density of rodent burrows (2–3 per m2), leading to reduced pastoral value. In cases of severe degradation, rodent activity becomes intense, vegetation is nearly completely replaced with miscellaneous forbs, and the surface exhibits dark brown, patchy bare areas known as “black soil patches”. To date, research on mattic epipedon degradation has primarily focused on degradation and erosion processes [19,20], changes in physical and chemical properties [21], and the underlying causes [22]. However, rapid and non-destructive methods for identifying degraded mattic epipedons remain underdeveloped. Existing identification approaches largely rely on expert assessment and laboratory analysis, which are time-consuming and costly, thereby limiting their applicability for large-scale monitoring efforts.
Soil hyperspectral analysis technology has been widely recognized as an effective method for the rapid and non-destructive acquisition of soil properties, owing to its simplicity, environmental friendliness, efficiency, and cost-effectiveness. The research team led by Zhou Shi [23,24] conducted nationwide measurements of soil samples using ASD spectrometers, capturing reflectance spectra across the visible to near-infrared (VNIR) range. By analyzing soil spectral radiative transfer and applying geostatistical and machine learning techniques, they successfully predicted key soil properties such as soil organic matter, total nitrogen, moisture content, and salinity. Malmir [25] and Javadi [26] have demonstrated that statistical models developed from hyperspectral inversion data can rapidly and cost-effectively estimate concentrations of soil heavy metals—including copper, arsenic, nickel, lead, and zinc—and accurately predict their spatial distribution. Research conducted by Wang [27] and Gholizadeh [28] further indicated that VNIR spectral data can provide robust quantitative indicators for soil classification, achieving high classification accuracy. Although this technology has been successfully applied to the estimation of soil organic matter [29,30,31], moisture [32] texture [33,34], and heavy metal content [35,36,37], studies specifically targeting its application in the rapid and non-destructive identification of soil degradation types remain limited. In field conditions, spectral data acquisition is often influenced by environmental factors, instrument variability, and operational inconsistencies. As such, spectral preprocessing techniques have been widely employed in soil spectral analysis to enhance data quality. These techniques primarily serve to transform signals, eliminate noise, enhance features, and extract relevant information for quantitative prediction models. For mattic epipedon spectral data, commonly used preprocessing methods include Savitzky–Golay (SG) smoothing, first derivatives (FDs), standard normal variates (SNVs), and multiplicative scatter correction (MSC) [38]. These techniques effectively suppress high-frequency noise, correct baseline drift, and reduce scattering effects, thereby significantly improving the quality of spectral data and the accuracy of model predictions. Notably, although some researchers have optimized model performance by comparing the feature selection results under different preprocessing methods, a systematic investigation has yet to be conducted on feature band identification specific to mattic epipedon degradation. The existing literature has not yet elucidated the specific impacts of different preprocessing techniques on the extraction of degradation-sensitive spectral bands for mattic epipedons. This represents a key methodological gap in the application of hyperspectral technology for monitoring ecosystem degradation in alpine environments [39,40]. Alla [41], working within the framework of the Richards equation, optimized irrigation strategies through dynamic boundary control and found that soil moisture content directly influences soil water potential and root water uptake efficiency, thereby altering water absorption characteristics in soil spectra. Similarly, Lopes [42] developed an optimal control-based irrigation model for agricultural fields in Lisbon, Portugal, demonstrating that soil spectral signals are sensitive to both plant transpiration and soil water loss under water stress conditions. Their state-constrained optimization algorithm enhanced the accuracy of water demand predictions during drought years. While previous studies have uncovered the modulation mechanisms of moisture on spectral signals, the influence of moisture on the hyperspectral characteristics of mattic epipedons—and the development of models capable of adapting to varying moisture conditions—remains underexplored.
To address the aforementioned challenges, this study targeted the critical bottlenecks in the identification of mattic epipedon degradation and aimed to achieve the following two research objectives through systematic scientific approaches and technical methodologies: (1) Accurately identify mattic epipedon degradation based on hyperspectral data, thereby overcoming the current reliance on traditional laboratory analyses for degradation assessment. (2) Investigate the impact of soil moisture on the hyperspectral detection of mattic epipedon degradation and develop a moisture interference correction strategy that enhances the environmental adaptability of the proposed method across diverse alpine conditions.

2. Materials and Methods

2.1. Study Area

The study area is located in the northeast of the Qinghai–Tibet Plateau, 37°30′~38°36′ N, 99°22′~101°08′ E. Figure 1 shows an overview of the study area. It has a plateau continental climate, humid and cold, with a mean annual temperature (MAT) of about −4 °C. Precipitation is concentrated in July and August., with a mean annual precipitation (MAP) of about 393 nm. The altitude is higher in the north and lower in the middle and south, with an average altitude of about 3670 m [5]. The vegetation in the study area is mainly alpine grassland and coniferous forest, among which the grassland types include alpine meadow, alpine steppe, and mountain meadow, with Qinghai spruce and Qilian juniper as the main species [5]. Due to the suitable altitude and precipitation, mattic epipedon is widely distributed in the area.

2.2. Data Collection

2.2.1. Soil Sample Collection and Preparation

Soil sample collection was implemented in July 2023, and a total of 227 mattic epipedon sampling points were set up in the study area. Figure 2 shows the on-site collected photos. The standard knife ring method (inner diameter 5 cm, volume 100 cm3) was used to collect undisturbed soil samples, and two knife ring soil samples were collected at each sampling point. The collected soil samples were immediately packed into polyethylene self-sealing bags, and waterproof labels containing the sampling point number, date, and other information were affixed. The samples were immediately preserved using portable refrigerated containers maintained at a low temperature of −4 ± 2 °C with ice packs, transported to the laboratory within 24 h, and subsequently stored in a freezer to ensure the stability of their physicochemical properties. The physical and chemical properties of the mattic epipedon were analyzed. Table 1 lists descriptive statistics of the soil, including physical and chemical properties. The bulk density, moisture content, root volume ratio, pH, organic carbon, singular fractal dimension, nitrogen (N), and carbon (C) were measured in the laboratory, while the thickness of the mattic epipedon was assessed on-site by experts. This study also utilized the Long-term High-resolution Grazing Intensity (LHGI) dataset for China, released by the National Ecological Science Data Center [43]. This dataset provides a quantitative representation of the spatial and temporal distribution of grazing intensity on grasslands. Moreover, through principal component analysis, five indicators—total carbon, root volume ratio, single fractal dimension, grazing intensity, and pH—were selected as the minimum dataset for evaluating mattic epipedon degradation. The subjective [44,45,46,47] and objective weights were calculated using hierarchical analysis and the CRITIC method, respectively, and the comprehensive weight was obtained based on the principle of minimum information entropy. The soil quality index was calculated using the membership function combined with the comprehensive weight, and the degradation state of the mattic epipedon was distinguished with a threshold of 0.5 [44,45,46,47]. Finally, 94 samples were defined as non-degraded mattic epipedons and 133 as degraded mattic epipedons.

2.2.2. Soil Hyperspectral Data

Soil hyperspectral data were measured using an American ASD Field Spec 4 portable ground feature spectrometer [48,49]. In this study, the American ASD Field Spec 4 portable ground feature spectrometer manufactured by ASD Inc. (Analytical Spectral Devices, Inc.) from the United States was used for spectral data collection. The headquarters of the manufacturer is located in Boulder, CO, USA. This instrument has a wavelength range of 350 to 2500 nm and a spectral resolution of 1 nm. The hyperspectral data were measured in a dark laboratory. The indoor light source was a 50 W halogen lamp with a light source height of 130 cm. The ASD probe was stabilized on a tripod and aimed at the center of the soil sample dish. The soil surface was first scraped flat with a ruler before measurement. Before each measurement, the spectrometer was optimized and calibrated with a white board [50]. The spectrum of each soil sample was measured in four directions (90° rotation each time, a total of 3 rotations), and two spectral curves were saved in each direction. Eight spectral curves were repeatedly collected and averaged as the original spectral data of the soil sample. During the spectral acquisition process, the edge bands of 350 to 399 nm and 2401 to 2500 nm were removed because of the large signal-to-noise ratio. The remaining band of 400 to 2400 nm data was retained as the original soil spectrum. To further reduce noise interference, the original soil spectrum was smoothed via the Savitzky–Golay (SG) filtering method. All soil samples collected from the 227 sampling sites in this study were subjected to hyperspectral analysis following the aforementioned procedures and standards.

2.2.3. Hyperspectral Data Preprocessing

In this study, six spectral preprocessing methods were employed: Savitzky–Golay (SG) filtering, first-order differential transformation (First Derivative, D1), second-order differential transformation (Second Derivative, D2), continuum removal transformation (CR), standard normal variate transformation (SNV), and multiplicative scatter correction (MSC). The data processing procedure was as follows: SG convolution smoothing was first applied to all spectral data for baseline correction, followed sequentially by D1, D2, CR, SNV, and MSC processing. Additionally, D1 and D2 differential transformations were further superimposed onto the CR-, SNV-, and MSC-processed data. This progressive preprocessing strategy was designed to suppress both environmental and instrumental noise interference while enhancing the spectral response characteristics of the target features.

2.2.4. Hyperspectral Data Characteristic Band Screening

Hyperspectral data contain a vast amount of spectral band information, some of which are predictive, while the remainder are redundant or noisy. Such redundancy and noise not only increase model complexity but also contribute to overfitting, thereby reducing the model’s generalization capability. Therefore, the screening of characteristic bands is essential for effective hyperspectral data analysis. In this study, the Boruta algorithm [51] was applied to identify the characteristic bands from hyperspectral data. The Boruta algorithm is a wrapper-based feature selection method that utilizes random forest classifiers. It identifies informative spectral bands by comparing the importance of original features with that of randomly permuted shadow features. Compared with other feature selection algorithms, Boruta offers the advantage of retaining all potentially relevant feature bands while effectively reducing the risk of overfitting on training data.

2.2.5. Moisture Correction of Hyperspectral Data

In this study, the external parameter orthogonalization (EPO) algorithm [52] was applied to correct the interference caused by moisture during the identification of degraded mattic epipedons based on spectral data. Typically, the spectral matrix of a water-containing mattic epipedon, based on the data collected with a spectrometer, can be expressed as Equation (1):
X = X P + X Q + r
where X represents the spectral matrix of the water-containing mattic epipedon, XP denotes the projection matrix containing useful information for degradation identification, XQ represents the projection matrix of moisture content, and r refers to a redundant matrix that can be disregarded.
The steps for calculating the EPO algorithm are as follows: First, the difference spectrum matrix of the mattic epipedon under varying moisture conditions is calculated:
D = X m − X 0
where D represents the difference spectrum matrix of the mattic epipedon, X m is the spectral matrix under varying moisture contents, and X 0 is the spectral matrix of the dried soil sample.
Next, the covariance matrix of the difference spectrum is computed, followed by singular value decomposition (SVD):
S V D D T D = U S V T
The projection matrix representing the moisture content of the mattic epipedon is then expressed as the following:
Q = V g V g T
where V g denotes the matrix composed of the leading singular vectors, and the cumulative sum of the selected singular values is required to reach 99% of the total variance in the dataset.
Finally, the spectral matrix of the water-bearing mattic epipedon is projected onto the orthogonal subspace to eliminate moisture interference, yielding the corrected spectrum matrix:
P = I − Q
X * = X P
where I is the unit matrix and X * is the target matrix representing the corrected spectra.

2.3. Model Building and Validation

Machine learning algorithms have provided a crucial technical pathway for the efficient and accurate identification of mattic epipedon degradation, with numerous studies conducted in this field [53,54,55] that have compared the performances of various machine learning algorithms in spectral data classification. Ultimately, the XGBoost model was selected as the core algorithm. Additionally, the effects of different spectral preprocessing methods on the accuracy of mattic epipedon degradation identification were analyzed to determine the optimal preprocessing combination, thereby enhancing the predictive performance of the model.

2.3.1. The XGBoost Model

The accuracy of hyperspectral models for mattic epipedon degradation is influenced by the composition of the mattic epipedon, the representativeness of the sampling data, and the choice of identification method. Direct physical modeling of mattic epipedon degradation remains challenging. XGBoost [56] is a gradient-boosting framework whose objective function consists of both a loss function, which quantifies the difference between the predicted and actual labels, and a regularization term, which constrains model complexity. This study employed R language (Version: 4.2.1) for data modeling, with the XGBoost model implemented using the xgboost package. Built upon the gradient boosting algorithm framework, this package facilitates efficient handling of large-scale datasets and optimizes model parameters (such as learning rate, tree depth, regularization parameters, etc.) through cross-validation. Model accuracy is improved iteratively by correcting the prediction errors of the preceding decision trees. The scores assigned to leaf nodes after each tree split are aggregated and converted into class probabilities through a logistic function. Decisions are then made based on probability thresholds. XGBoost is capable of addressing both regression and classification tasks and effectively handles complex nonlinear datasets. Accordingly, the XGBoost algorithm was selected in this study to establish a hyperspectral identification model for mattic epipedon degradation.

2.3.2. Model Validation

In this study, we compared the mattic epipedon degradation samples classified based on the soil degradation index with the model identification results, evaluating the model’s inversion accuracy using metrics such as accuracy, precision, and recall. Accuracy is defined as the proportion of correctly predicted samples relative to the total number of samples, serving as an indicator of overall model performance. The corresponding formulas are defined as follows:
A c c u r a c y = T P + T N T P + F P + F N + T N
P r e c i s i o n = T P T P + F P
R e c a l l = T P T P + F N
In these formulas, true positive (TP) refers to the number of samples that are actually degraded and correctly predicted as degraded; true negative (TN) denotes the number of non-degraded samples that are correctly predicted as non-degraded; false positive (FP) represents the number of non-degraded samples incorrectly predicted as degraded; and false negative (FN) denotes the number of degraded samples that are incorrectly predicted as non-degraded.

3. Results

3.1. Identification of Mattic Epipedon Degradation Under Water-Free Conditions

The original spectral reflectance curves of the mattic epipedon are presented in Figure 3, where Figure 3a,b illustrate the curves for the non-degraded and degraded mattic epipedon layers, respectively. These layers exhibit similar spectral morphological characteristics within the 400–2400 nm band, although slight differences are present. The original spectral curves indicate that reflectance decreases with increasing wavelength in the 400–500 nm range, increases sharply between 500 and 1000 nm, rises gradually from 1000 to 2200 nm, and reaches a peak near 1800 nm. Additionally, three distinct absorption valleys are observed at 1400 nm, 1900 nm, and 2200 nm. The absorption valleys located at 1400 nm and 1900 nm are associated with hydroxyl groups and water, whereas the absorption feature at 2200 nm is primarily attributed to organic matter. It is noteworthy that although the spectral trends of the two mattic epipedon types are largely consistent, the overall reflectance of the degraded mattic epipedon is generally higher than that of the non-degraded samples, providing a critical spectral basis for degradation identification.
The hyperspectral reflectance curves of the mattic epipedon layer, following various preprocessing methods, are presented in Figure 4: Figure 4a displays the spectral reflectance curve following Savitzky–Golay (SG) filtering, which significantly reduces random noise interference and results in a smoother spectral profile. Figure 4b illustrates the curve after first-order differential processing, which eliminates baseline offsets and enhances the edge information of the absorption features. Figure 4c shows the curve after second-order differential processing, which further accentuates the curvature characteristics of the absorption valleys. Figure 4d presents the curve after continuum removal processing, which accurately characterizes the relative depth and morphological properties of the absorption features. Finally, Figure 4e,f depict the spectral reflectance curves following multivariate scatter correction (MSC) and standard normal variate (SNV) transformation, respectively. Both MSC and SNV transformations effectively reduce the influence of soil surface scattering effects and particle size variations. Each preprocessing method offers distinct advantages and provides a foundation for subsequent feature selection.
The Boruta algorithm was employed to screen the characteristic bands, and the results revealed significant differences among the bands selected using different preprocessing methods. Among these methods, the MSC+FD preprocessing combination exhibited the strongest feature selection capability, with its selected bands spanning the entire range of visible, near-infrared, and short-wave infrared spectra. This study noted that the bands at 446–450 nm, 637–639 nm, 662–663 nm, 2134 nm, and 2267–2269 nm appeared repeatedly across multiple preprocessing results, suggesting that these bands play a critical role in indicating mattic epipedon degradation. Specifically, the 2134 nm band is closely associated with variations in organic matter content, while the 2267–2269 nm band reflects changes in mineral composition. These findings are highly consistent with previous studies. The 2134 nm wavelength, located in the near-infrared region, corresponds to the combination and overtone absorption bands of hydrogen-containing functional groups such as O–H and C–H. Organic matter molecules in soil or vegetation—such as humic substances and cellulose—are rich in these groups, and variations in their content directly affect spectral reflectance at this wavelength [57,58]. Previous research has demonstrated that the near-infrared region, particularly the 2000–2300 nm range, is a sensitive spectral window for monitoring soil organic matter content [59]. When mattic epipedon degradation leads to organic matter decomposition or loss, reflectance in this band tends to increase due to the reduction in organic matter.
These key characteristic bands provide a critical foundation for the establishment of a high-precision degradation identification model for the mattic epipedon layer. The research findings demonstrated that the combination of appropriate spectral preprocessing methods and feature selection algorithms can effectively identify mattic epipedon degradation characteristics, offering a novel approach for the rapid monitoring of alpine grassland ecosystems. The final selected characteristic bands are summarized in Table 2. These feature bands were used as input variables for constructing the XGBoost model, and the corresponding results are presented in Table 3. The comparison indicates that model recognition accuracy significantly improved when second-order differential transformation preprocessing was applied. Subsequent research will focus on samples pretreated with the MSC+SD combination, using the XGBoost model as a unified benchmark framework.

3.2. Identification of Mattic Epipedon Degradation Under Different Water Content Conditions

The hyperspectral reflectance curves of mattic epipedon under varying moisture conditions are presented in Figure 5. Within the moisture content range of 0–30%, the spectral reflectance generally exhibits a decreasing trend as the moisture content increases; however, the change is notably less pronounced in the 5–10% and 20–25% intervals, suggesting the presence of a staged saturation effect induced by moisture. Under varying water conditions, the spectral curve maintains low reflectance in the visible light region (400–700 nm), exhibits a rapid increase followed by a deceleration in the near-infrared region (700–1400 nm), and displays continuous fluctuations in the short-wave infrared region (1400–2400 nm). Notably, significant absorption peaks are observed at 1400 nm, 1900 nm, and 2200 nm, which correspond to the vibrational characteristics of hydroxyl groups, interlayer water, and aluminum/magnesium hydroxyl groups, respectively. These characteristic absorption peaks provide a critical spectral basis for the inversion of water content. The spectral reflectance generally increases with rising soil moisture content. However, as the soil moisture increases, the spectral reflectance initially decreases before increasing again. When the soil moisture reaches approximately 30%, the reflectance begins to rise. This phenomenon is primarily attributed to the formation of a dual-structure in the soil moisture regime, where water content exceeds the surface water layer, leading to changes in the soil’s optical properties [60].
Under varying moisture conditions, the selection of characteristic hyperspectral bands for the mattic epipedon was uniformly conducted using a combined spectral preprocessing approach of Multiplicative Scatter Correction (MSC) and second derivative transformation. As shown in Table 4, the analysis revealed that under low moisture content (5–10%), the characteristic bands are primarily concentrated in the shortwave infrared region (2140–2359 nm), which closely corresponds to the molecular vibration features of cellulose and lignin. At moderate moisture levels (15–20%), the characteristic bands extend into the visible and near-infrared regions. Specifically, the 571–574 nm range reflects spectral features associated with chlorophyll degradation products, while the 2203–2210 nm range indicates coupling effects between water molecules and organic matter. Under high moisture conditions (25–30%), the characteristic bands are predominantly centered around 2200 nm, where strong water absorption dominates. Furthermore, the number of characteristic bands increases from 10 to 13 as the moisture content rises from 5% to 20%, but then declines and becomes more concentrated when the moisture exceeds 25%. This trend is closely related to the masking effect of water on the spectral peaks of the organic components.
The hyperspectral identification results for mattic epipedon degradation under varying water content conditions are presented in Table 5. Water content exerts a significant influence on degradation identification accuracy. The model achieved its highest accuracy (0.81) at a moisture content of 5%, while performance progressively declined with increasing moisture, reaching a minimum value of 0.65 at 30%. Among these, the highest precision (0.76) was observed at a moisture content of 10%, while the highest recall rate (0.83) occurred at 15%. This phenomenon can be attributed to the dual influence of moisture on spectral features: Moisture absorption bands may mask critical degradation-related signals, while variations in moisture responses across degradation states contribute to feature confusion. These results suggest that monitoring mattic epipedon degradation under low-moisture conditions yields more reliable identification outcomes, thereby providing an important reference for the design of field sampling strategies aimed at monitoring alpine grassland ecosystems.

3.3. Identification of Mattic Epipedon Degradation Using EPO Moisture Correction Algorithm

The difference in the hyperspectral reflectance curves of mattic epipedon under both drying and humidification conditions is presented in Figure 6. The analysis indicates that the hyperspectral reflectance decreases overall as the moisture content increases, with particularly pronounced changes occurring in the near-infrared and short-wave infrared regions. The hyperspectral reflectance curve of mattic epipedon after EPO correction is presented in Figure 7. The analysis revealed that, compared with the original spectrum, the range of reflectance values is significantly expanded following EPO correction, which effectively reduces moisture-induced noise and substantially improves the signal-to-noise ratio.
Using a preprocessing method that combines multivariate scatter correction (MSC) and standard normal variate (SNV) transformation, along with external parameter orthogonal (EPO) correction, this study systematically analyzed the characteristic band screening patterns of mattic epipedon hyperspectral data under varying moisture conditions. The screening results for the characteristic bands are presented in Table 6. The results indicate that the characteristic bands screened under different moisture conditions exhibit varying distribution patterns, and their quantity shows a nonlinear relationship with moisture content. At 5% moisture content, the characteristic bands are primarily concentrated in the short-wave infrared region, whereas at 10–30% moisture levels, the bands are distributed across multiple spectral regions. High moisture content may suppress spectral differentiation, leading to an overall decreasing trend in the number of characteristic bands.
The identification results for characteristic bands screened from the hyperspectral data of mattic epipedon under varying moisture conditions are presented in Table 7. The results indicate that when the water content is below 10%, recognition accuracy decreases following EPO correction, which may be attributed to the algorithm’s limited ability to accurately extract the interference subspace and the complex interactions between water and the soil matrix. When the water content exceeds 10%, the recognition accuracy improves, with the most pronounced optimization observed under 30% moisture conditions, suggesting that EPO effectively separates water interference and enhances the expression of degradation characteristics. Overall, under low-moisture conditions, EPO correction may lead to over-adjustment, resulting in reduced accuracy, whereas under high-moisture conditions, more degradation-related information is retained after correction, thereby improving recognition accuracy.

4. Discussion

This study identified key spectral indicators and response mechanisms associated with the degradation of mattic epipedon. By integrating spectral preprocessing, machine learning modeling, and moisture noise correction techniques, a high-accuracy classification of degraded mattic epipedon was achieved. The main conclusions are as follows:
It is worth noting that spectral preprocessing is essential for enhancing informative features within the spectra and improving their correlation with mattic epipedon degradation. In this study, eight different preprocessing techniques were applied to the raw hyperspectral data. Each method improved the correlation between spectral features and degradation to varying degrees, among which the combination of Multiplicative Scatter Correction (MSC) and second derivative transformation yielded the most significant enhancement. It was also observed that the choice of preprocessing method had a substantial influence on the selection of characteristic bands, a finding consistent with the results reported by Bouslihim [61]. This is primarily because spectral data acquired in the field are influenced by instrument noise, environmental interference, and the physical properties of the samples—such as particle size and surface roughness—leading to raw spectra containing substantial redundant information and noise. Different preprocessing techniques target specific sources of interference and modify the mathematical representation of the spectral data, thereby affecting the sensitivity and representativeness of the subsequently selected feature bands [62,63].
In addition to spectral preprocessing, variations in moisture content significantly affect the spectral reflectance of the mattic epipedon and the accuracy of degradation identification models. The analysis indicates that under low-moisture conditions (5–10%), the characteristic bands are primarily concentrated in the short-wave infrared region (2140–2359 nm), which closely corresponds to the molecular vibration features of cellulose and lignin. Krzyszczak [64] found that in soils with low moisture content, the absorption depth within the 2200–2300 nm range is positively correlated with the cellulose content, corroborating our findings. Under low moisture conditions, mattic epipedon is likely in the early stages of mild degradation, where vegetation residues decompose slowly and cellulose and lignin are not fully mineralized. The structural integrity of these molecules leads to pronounced short-wave infrared spectral features, making this band combination a potential spectral marker for early degradation warning.
Under moderate moisture conditions (15–20%), the characteristic bands extend to the visible and near-infrared regions. According to the research findings of Dalal and Henry [65], the absorption peak near 2200 nm can be decomposed into “water-independent absorption” and “organic matter–water complex absorption” components. The expanded response of this band in our study confirms the modulating effect of moisture on organic matter spectral characteristics.
At high moisture levels (25–30%), within the 5–20% moisture range, the number of characteristic bands increases from 10 to 13, whereas above 25%, the number of bands decreases, although their distribution becomes more concentrated. This phenomenon is closely related to the masking effect of water on organic spectral peaks. Under high moisture conditions, water acts as a strong absorbing medium dominating the spectral signal, suppressing the spectral features of other components such as organic matter and minerals. This observation aligns with the findings of Wei [66], who reported that for sandy soils with moisture content exceeding 30%, spectral curves tend to approach the smooth, low-reflectance pattern characteristic of water bodies.
After moisture correction using the EPO algorithm, the model performance exhibited heterogeneous responses. After moisture correction using the External Parameter Orthogonalization (EPO) algorithm, the model performance exhibited heterogeneous responses. Under low moisture conditions, where water-related spectral signals are inherently weak, the application of EPO resulted in reduced classification accuracy. In such environments, the spectral influence of moisture is relatively limited, and signals related to target identification tend to dominate. The EPO algorithm operates by constructing a transformation matrix based on the difference between wet and dry spectra to isolate water-related components from the moist spectral data. However, due to the weak water signals in low moisture conditions, EPO may inadvertently remove subtle yet relevant features that contribute to degradation identification, mistaking them for moisture interference, thus leading to decreased model performance [67,68].
Conversely, under high moisture conditions, where water signals are strong and have a pronounced impact on the spectra, EPO correction improved model accuracy. In these cases, water absorption bands can obscure key spectral features associated with degradation, and variations in moisture responses across different degradation levels may introduce feature confusion. The EPO algorithm effectively separates water-related information from wet spectra while preserving features relevant to degradation detection [67,68,69]. This reduces moisture interference and enables the model to more accurately capture spectral signatures linked to the degradation of the mattic epipedon, thereby enhancing classification performance. These results confirm the efficacy of the EPO algorithm under conditions of high moisture interference.

5. Conclusions

This study focused on the degradation of mattic epipedon in the Qilian Mountains region of the Qinghai–Tibet Plateau, employing hyperspectral technology to construct quantitative models for the identification of degraded mattic epipedon in this area. Through the integration of multiple hyperspectral preprocessing methods with the XGBoost algorithm, a robust quantitative model was successfully developed, achieving high-accuracy discrimination of degraded mattic epipedons. Furthermore, this study revealed the significant influence of soil moisture on the hyperspectral features of mattic epipedons and established an MSC+SD-XGBoost model tailored to different moisture conditions, demonstrating strong adaptability across varying soil moisture levels.
The results provide critical technical support for the precise identification and early warning of mattic epipedon degradation, laying a foundation for ecological restoration efforts. However, certain limitations remain, such as the lack of moisture interference correction under low-moisture conditions and the need to enhance the generalizability of the proposed method. Future research could explore multimodal data fusion strategies through integrating hyperspectral, thermal infrared, and radar remote sensing data to develop physically interpretable models for degradation feature interpretation. This approach could overcome the limitations posed by single-spectral-dimension sensitivity to confounding factors like moisture and organic matter.

Author Contributions

Conceptualization, J.Z., H.Z. and Z.S.; methodology, J.Z. and H.Z.; software, H.Z., Q.Y., L.G. and D.Z.; validation, J.Y. and Q.Y.; formal analysis, H.Z. and J.Z.; investigation, J.Z., J.Y., H.Z., Q.Y., D.Z., Z.S. and L.G.; resources, J.Z., C.L. and Z.S.; data curation, L.G., Q.Y., D.Z. and J.Y.; writing—original draft preparation, H.Z.; writing—review and editing, J.Z., Z.S. and C.L.; visualization, H.Z., J.Y. and L.G.; supervision, J.Z.; project administration, J.Z. and Z.S.; funding acquisition, J.Z. and Z.S. All authors have read and agreed to the published version of the manuscript.

Funding

This work was financed by the Natural Science Foundation of China (Nos. 42271060 and 42401445) and the Natural Science Foundation of Anhui Province (No. 2208085MD91).

Data Availability Statement

Data are contained within the article.

Conflicts of Interest

The authors declare no conflicts of interest.

References

  1. Ma, Y.; Wang, B.; Chen, X.; Zhong, L.; Hu, Z.; Ma, W.; Han, C.; Li, M. Strengthening the three-dimensional comprehensive observation system of multi-layer interaction on the Tibetan Plateau to cope with the warming and wetting trend. Atmos. Ocean. Sci. Lett. 2022, 15, 100224. [Google Scholar] [CrossRef]
  2. Qiu, J. The third pole. Nature 2008, 454, 393–396. [Google Scholar] [CrossRef] [PubMed]
  3. He, L.; Zhong, L.; Ma, Y.; Qi, Y.; Liu, J.; Li, P. A study on the simulation of carbon and water fluxes of Dangxiong alpine meadow and its response to climate change. Atmos. Ocean. Sci. Lett. 2024, 17, 100507. [Google Scholar] [CrossRef]
  4. Wang, X.; Zhu, J.; Peng, S.; Zheng, T.; Qi, Z.; Hu, J.; Ji, C. Patterns of grassland community composition and structure along an elevational gradient on the Qinghai-Tibet Plateau. J. Plant Ecol. 2022, 15, 808–817. [Google Scholar] [CrossRef]
  5. Wang, Y.; Xue, K.; Hu, R.; Ding, B.; Zeng, H.; Li, R.; Xu, B.; Pang, Z.; Song, X.; Li, C.; et al. Vegetation structural shift tells environmental changes on the Tibetan Plateau over 40 years. Sci. Bull. 2023, 68, 1928–1937. [Google Scholar] [CrossRef]
  6. Zhang, M.Y.; Ma, Y.J.; Chen, P.; Shi, F.Z.; Wei, J.Q. Growing-season carbon budget of alpine meadow ecosystem in the Qinghai Lake Basin: A continued carbon sink through this century according to the Biome-BGC model. Carbon Balance Manag. 2023, 18, 25. [Google Scholar] [CrossRef]
  7. Xie, Y.; Huang, J.; Wu, G.; Liu, Y.; Dong, W.; Lu, M.; He, B.; Su, Z.; Bao, Q.; Zhao, Q.; et al. Oceanic repeaters boost the global climatic impact of the Tibetan Plateau. Sci. Bull. 2023, 68, 2225–2235. [Google Scholar] [CrossRef]
  8. Zhi, J.; Zhou, Z.; Cao, X. Exploring the determinants and distribution patterns of soil mattic horizon thickness in a typical alpine environment using boosted regression trees. Ecol. Indic. 2021, 133, 108373. [Google Scholar] [CrossRef]
  9. Shen, H.; Dong, S.; DiTommaso, A.; Xiao, J.; Zhi, Y. N deposition may accelerate grassland degradation succession from grasses-and sedges-dominated into forbs-dominated in overgrazed alpine grassland systems on Qinghai-Tibetan Plateau. Ecol. Indic. 2021, 129, 107898. [Google Scholar] [CrossRef]
  10. Niu, Y.; Zhu, H.; Yang, S.; Ma, S.; Zhou, J.; Chu, B.; Hua, R.; Hua, L. Overgrazing leads to soil cracking that later triggers the severe degradation of alpine meadows on the Tibetan Plateau. Land Degrad. Dev. 2019, 30, 1243–1257. [Google Scholar] [CrossRef]
  11. Dong, L.; Li, J.; Sun, J.; Yang, C. Soil degradation influences soil bacterial and fungal community diversity in overgrazed alpine meadows of the Qinghai-Tibet Plateau. Sci. Rep. 2021, 11, 11538. [Google Scholar] [CrossRef] [PubMed]
  12. Abegaz, A.; Winowiecki, L.A.; Vågen, T.-G.; Langan, S.; Smith, J.U. Spatial and temporal dynamics of soil organic carbon in landscapes of the upper Blue Nile Basin of the Ethiopian Highlands. Agric. Ecosyst. Environ. 2016, 218, 190–208. [Google Scholar] [CrossRef]
  13. Gao, S.; Jiang, G.; Zhang, Z.; Xu, X.; Wang, L.; Wu, Q. Permafrost temperature dynamics and its climate relations in various Tibetan alpine grasslands. CATENA 2024, 243, 108148. [Google Scholar] [CrossRef]
  14. Duan, H.; Xue, X.; Wang, T.; Kang, W.; Liao, J.; Liu, S. Spatial and Temporal Differences in Alpine Meadow, Alpine Steppe and All Vegetation of the Qinghai-Tibetan Plateau and Their Responses to Climate Change. Remote Sens. 2021, 13, 669. [Google Scholar] [CrossRef]
  15. Zhang, Y.; Dong, S.; Gao, Q.; Liu, S.; Zhou, H.; Ganjurjav, H.; Wang, X. Climate change and human activities altered the diversity and composition of soil microbial community in alpine grasslands of the Qinghai-Tibetan Plateau. Sci. Total Environ. 2016, 562, 353–363. [Google Scholar] [CrossRef]
  16. Ji, W.; Li, S.; Chen, S.; Shi, Z.; Rossel, R.A.V.; Mouazen, A.M. Prediction of soil attributes using the Chinese soil spectral library and standardized spectra recorded at field conditions. Soil Tillage Res. 2016, 155, 492–500. [Google Scholar] [CrossRef]
  17. Teng, H.; Viscarra Rossel, R.A.; Shi, Z.; Behrens, T.; Chappell, A.; Bui, E. Assimilating satellite imagery and visible–near infrared spectroscopy to model and map soil loss by water erosion in Australia. Environ. Model. Softw. 2016, 77, 156–167. [Google Scholar] [CrossRef]
  18. Wang, J.-b.; Zhang, D.-g.; Cao, G.-m.; Tian, Q. Regional characteristics of the alpine meadow degradation succession on the Qinghai-Tibetan Plateau. Acta Prataculturae Sin. 2013, 22, 1–10. [Google Scholar]
  19. Chen, K.; Zhang, J.; Muneer, M.A.; Xue, K.; Niu, H.; Ji, B. Plant community and soil available nutrients drive arbuscular mycorrhizal fungal community shifts during alpine meadow degradation. Fungal Ecol. 2023, 62, 101211. [Google Scholar] [CrossRef]
  20. Dai, L.; Guo, X.; Ke, X.; Du, Y.; Mang, F.; Cao, G. The variation in soil water retention of alpine shrub meadow under different degrees of degradation on northeastern Qinghai-Tibetan plateau. Plant Soil 2021, 458, 231–244. [Google Scholar] [CrossRef]
  21. Wu, Y.; Yang, W.; Li, Q.; Qiao, Q.; Zhao, S.; Zhang, Y.; Yu, Y.; Zhang, S.; Li, X.; Kou, J. Microbial Community Response to Alpine Meadow Degradation and Its Impact on Soil Nutrient Cycling. Agronomy 2024, 14, 195. [Google Scholar] [CrossRef]
  22. Du, C.; Gao, Y. Grazing exclusion alters ecological stoichiometry of plant and soil in degraded alpine grassland. Agric. Ecosyst. Environ. 2021, 308, 107256. [Google Scholar] [CrossRef]
  23. Xu, D.; Ma, W.; Chen, S.; Jiang, Q.; He, K.; Shi, Z. Assessment of important soil properties related to Chinese Soil Taxonomy based on vis-NIR reflectance spectroscopy. Comput. Electron. Agric. 2018, 144, 1–8. [Google Scholar] [CrossRef]
  24. Jia, X.; O’Connor, D.; Shi, Z.; Hou, D. VIRS based detection in combination with machine learning for mapping soil pollution. Environ. Pollut. 2021, 268, 115845. [Google Scholar] [CrossRef] [PubMed]
  25. Malmir, M.; Tahmasbian, I.; Xu, Z.; Farrar, M.B.; Bai, S.H. Prediction of soil macro- and micro-elements in sieved and ground air-dried soils using laboratory-based hyperspectral imaging technique. Geoderma 2019, 340, 70–80. [Google Scholar] [CrossRef]
  26. Javadi, S.H.; Munnaf, M.A.; Mouazen, A.M. Fusion of Vis-NIR and XRF spectra for estimation of key soil attributes. Geoderma 2021, 385, 114851. [Google Scholar] [CrossRef]
  27. Wang, C.; Wang, J.; Zhang, F.; Yang, Y.; Luo, F.; Li, Y.; Li, J. Stability response of alpine meadow communities to temperature and precipitation changes on the Northern Tibetan Plateau. Ecol. Evol. 2022, 12, e8592. [Google Scholar] [CrossRef]
  28. Gholizadeh, A.; Rossel, R.A.V.; Saberioon, M.; Boruvka, L.; Kratina, J.; Pavlu, L. National-scale spectroscopic assessment of soil organic carbon in forests of the Czech Republic. Geoderma 2021, 385, 114832. [Google Scholar] [CrossRef]
  29. Ye, M.; Zhu, L.; Li, X.; Ke, Y.; Huang, Y.; Chen, B.; Yu, H.; Li, H.; Feng, H. Estimation of the soil arsenic concentration using a geographically weighted XGBoost model based on hyperspectral data. Sci. Total Environ. 2023, 858, 159798. [Google Scholar] [CrossRef]
  30. Munnaf, M.A.; Mouazen, A.M. Removal of external influences from on-line vis-NIR spectra for predicting soil organic carbon using machine learning. CATENA 2022, 211, 106015. [Google Scholar] [CrossRef]
  31. Nanni, M.R.; Demattê, J.A.M.; Rodrigues, M.; Santos, G.L.A.A.d.; Reis, A.S.; Oliveira, K.M.d.; Cezar, E.; Furlanetto, R.H.; Crusiol, L.G.T.; Sun, L. Mapping Particle Size and Soil Organic Matter in Tropical Soil Based on Hyperspectral Imaging and Non-Imaging Sensors. Remote Sens. 2021, 13, 1782. [Google Scholar] [CrossRef]
  32. Jiang, C.-L.; Zhao, J.-Y.; Ding, Y.-Y.; Zhao, Q.-H.; Ma, H.-Y. Study on Soil Water Retrieval Technology of Yellow River Source Based on SPA Algorithm and Machine Learning. Spectrosc. Spectr. Anal. 2023, 43, 1961–1967. [Google Scholar]
  33. Forkuor, G.; Hounkpatin, O.K.; Welp, G.; Thiel, M. High Resolution Mapping of Soil Properties Using Remote Sensing Variables in South-Western Burkina Faso: A Comparison of Machine Learning and Multiple Linear Regression Models. PLoS ONE 2017, 12, e0170478. [Google Scholar] [CrossRef] [PubMed]
  34. Shen, Q.; Shang, K.; Xiao, C.; Tang, H.; Wu, T.; Wang, C. A novel hyperspectral remote sensing estimation model for surface soil texture using AHSI/ZY1-02D satellite image. Int. J. Appl. Earth Obs. Geoinf. 2025, 138, 104453. [Google Scholar] [CrossRef]
  35. Feng, Y.; Wang, J.; Tang, Y. Estimation and inversion of soil heavy metal arsenic (As) based on UAV hyperspectral platform. Microchem. J. 2024, 207, 112027. [Google Scholar] [CrossRef]
  36. Zou, Z.; Wang, Q.; Wu, Q.; Li, M.; Zhen, J.; Yuan, D.; Zhou, M.; Xu, C.; Wang, Y.; Zhao, Y.; et al. Inversion of heavy metal content in soil using hyperspectral characteristic bands-based machine learning method. J. Environ. Manag. 2024, 355, 120503. [Google Scholar] [CrossRef]
  37. Shen, Q.; Zhang, S.-w.; Ge, C.; Liu, H.-l.; Zhou, Y.; Chen, Y.-p.; Hu, Q.-q.; Ye, H.-c.; Huang, Y.-f. Hyperspectral Inversion of Heavy Metal Content in Soils Reconstituted by Mining Wasteland. Spectrosc. Spectr. Anal. 2019, 39, 1214–1221. [Google Scholar]
  38. Xie, S.; Ding, F.; Chen, S.; Wang, X.; Li, Y.; Ma, K. Prediction of soil organic matter content based on characteristic band selection method. Spectrochim. Acta A Mol. Biomol. Spectrosc. 2022, 273, 120949. [Google Scholar] [CrossRef]
  39. Shen, L.; Gao, M.; Yan, J.; Li, Z.-L.; Leng, P.; Yang, Q.; Duan, S.-B. Hyperspectral Estimation of Soil Organic Matter Content using Different Spectral Preprocessing Techniques and PLSR Method. Remote Sens. 2020, 12, 1206. [Google Scholar] [CrossRef]
  40. Gholizadeh, A.; Borůvka, L.; Saberioon, M.M.; Kozák, J.; Vašát, R.; Němeček, K. Comparing different data preprocessing methods for monitoring soil heavy metals based on soil spectral features. Soil Water Res. 2015, 10, 218–227. [Google Scholar] [CrossRef]
  41. Alla, A.; Berardi, M.; Saluzzi, L. State Dependent Riccati for dynamic boundary control to optimize irrigation in Richards’ equation framework. Math. Comput. Simul. 2025, 232, 261–275. [Google Scholar] [CrossRef]
  42. Lopes, S.; Fontes, F.; Pereira, R.; De Pinho, M.d.R.; Gonçalves, A.M. Optimal Control Applied to an Irrigation Planning Problem. Math. Probl. Eng. 2016, 2016, 5076879. [Google Scholar] [CrossRef]
  43. Wang, D.; Peng, Q.; Li, X.; Zhang, W.; Xia, X.; Qin, Z.; Ren, P.; Liang, S.; Yuan, W. A long-term high-resolution dataset of grasslands grazing intensity in China. Sci. Data 2024, 11, 1194. [Google Scholar] [CrossRef] [PubMed]
  44. Nascimento, C.M.; de Sousa Mendes, W.; Quiñonez Silvero, N.E.; Poppiel, R.R.; Sayão, V.M.; Dotto, A.C.; Valadares dos Santos, N.; Accorsi Amorim, M.T.; Demattê, J.A.M. Soil degradation index developed by multitemporal remote sensing images, climate variables, terrain and soil atributes. J. Environ. Manag. 2021, 277, 111316. [Google Scholar] [CrossRef]
  45. Jin, Q.; Wang, C.; Sardans, J.; Vancov, T.; Fang, Y.; Wu, L.; Huang, X.; Gargallo-Garriga, A.; Peñuelas, J.; Wang, W. Effect of soil degradation on the carbon concentration and retention of nitrogen and phosphorus across Chinese rice paddy fields. CATENA 2022, 209, 105810. [Google Scholar] [CrossRef]
  46. Gadal, S.; Gbetkom, P.; Ngandam Mfondoum, A.H. A New Soil Degradation Method Analysis by Sentinel 2 Images Combining Spectral Indices and Statistics Analysis: Application to the Cameroonians Shores of Lake Chad and Its Hinterland. In Proceedings of the 7th International Conference on Geographical Information Systems Theory, Applications and Management (GISTAM 2021), Virtual Event, 23–25 April 2021. [Google Scholar]
  47. Wang, J.; Zhen, J.; Hu, W.; Chen, S.; Lizaga, I.; Zeraatpisheh, M.; Yang, X. Remote sensing of soil degradation: Progress and perspective. Int. Soil Water Conserv. Res. 2023, 11, 429–454. [Google Scholar] [CrossRef]
  48. Danner, M.; Locherer, M.; Hank, T.; Berger, K. Spectral Sampling with the ASD FIELDSPEC 4; EnMAP Field Guides Technical Report; EnMAP Consortium, GFZ Data Services: Potsdam, Germany, 2019. [Google Scholar]
  49. Dhangar, V.; Barode, D.; Dhole, P.; Kshirsagar, A.; Gawali, B. Exploring the Spectral Database of Food Samples Using ASD Field Spec 4 Spectroradiometer. Int. J. Innov. Technol. Explor. Eng. (IJITEE) 2024, 13, 15–20. [Google Scholar] [CrossRef]
  50. Guindo, M.L.; Kabir, M.H.; Chen, R.; Liu, F. Potential of Vis-NIR to measure heavy metals in different varieties of organic-fertilizers using Boruta and deep belief network. Ecotoxicol. Environ. Saf. 2021, 228, 112996. [Google Scholar] [CrossRef]
  51. Jin, H.; Peng, J.; Bi, R.; Tian, H.; Zhu, H.; Ding, H. Comparing Laboratory and Satellite Hyperspectral Predictions of Soil Organic Carbon in Farmland. Agronomy 2024, 14, 175. [Google Scholar] [CrossRef]
  52. Yang, M.; Chen, S.; Xu, D.; Zhao, X.; Shi, Z.; Qian, H.; Zhang, Z. Removing the moisture effect on predicting soil organic matter using vis-NIR spectroscopy with external parameter orthogonalization. Geoderma Reg. 2024, 37, e00797. [Google Scholar] [CrossRef]
  53. Demattê, J.A.M.; Sousa, A.A.; Alves, M.C.; Nanni, M.R.; Fiorio, P.R.; Campos, R.C. Determining soil water status and other soil characteristics by spectral proximal sensing. Geoderma 2006, 135, 179–195. [Google Scholar] [CrossRef]
  54. Wang, C.; Feng, M.; Yang, W.-d.; Ding, G.-w.; Wang, H.-q.; Li, Z.-h.; Hui, S.; Shi, C.-c. Use of Spectral Character to Evaluate Soil Organic Matter. Soil Sci. Soc. Am. J. 2016, 80, 1078–1088. [Google Scholar] [CrossRef]
  55. Viscarra Rossel, R.A.; Behrens, T.; Ben-Dor, E.; Brown, D.J.; Demattê, J.A.M.; Shepherd, K.D.; Shi, Z.; Stenberg, B.; Stevens, A.; Adamchuk, V.; et al. A global spectral library to characterize the world’s soil. Earth-Sci. Rev. 2016, 155, 198–230. [Google Scholar] [CrossRef]
  56. Zhu, J.; Yin, Y.; Ma, T.; Wang, D. A novel maintenance decision model for asphalt pavement considering crack causes based on random forest and XGBoost. Constr. Build. Mater. 2025, 477, 140610. [Google Scholar] [CrossRef]
  57. Sun, W.; Liu, S.; Zhang, X.; Li, Y. Estimation of soil organic matter content using selected spectral subset of hyperspectral data. Geoderma 2022, 409, 115653. [Google Scholar] [CrossRef]
  58. Chen, Y.; Wang, J.; Liu, G.; Yang, Y.; Liu, Z.; Deng, H. Hyperspectral Estimation Model of Forest Soil Organic Matter in Northwest Yunnan Province, China. Forests 2019, 10, 217. [Google Scholar] [CrossRef]
  59. Zhao, M.; Gao, Y.; Lu, Y.; Wang, S. Hyperspectral Modeling of Soil Organic Matter Based on Characteristic Wavelength in East China. Sustainability 2022, 14, 8455. [Google Scholar] [CrossRef]
  60. Wu, L.-g.; Wang, S.-l.; He, J.-g.; Kazuhiro, N. Soil Moisture Mechanism and Establishment of Model Based on Hyperspectral Imaging Technique. Chin. J. Lumin. 2017, 38, 1366. [Google Scholar]
  61. Bouslihim, Y.; Bouasria, A. Potential of EnMAP Hyperspectral Imagery for Regional-Scale Soil Organic Matter Mapping. Remote Sens. 2025, 17, 1600. [Google Scholar] [CrossRef]
  62. Saleem, M.; Wang, C.; Feng, M.; Yang, W. Hyperspectral prediction of soil organic matter in winter wheat field based on continuum-removed method. Fresenius Environ. Bull. 2021, 30, 7979–7986. [Google Scholar]
  63. Wang, X.; Zhang, M.-W.; Guo, Q.; Yang, H.-L.; Wang, H.-L.; Sun, X.-L. Estimation of soil organic matter by in situ Vis-NIR spectroscopy using an automatically optimized hybrid model of convolutional neural network and long short-term memory network. Comput. Electron. Agric. 2023, 214, 108350. [Google Scholar] [CrossRef]
  64. Krzyszczak, J.; Baranowski, P.; Pastuszka, J.; Wesołowska, M.; Cymerman, J.; Sławiński, C.; Siedliska, A. Assessment of soil water retention characteristics based on VNIR/SWIR hyperspectral imaging of soil surface. Soil Tillage Res. 2023, 233, 105789. [Google Scholar] [CrossRef]
  65. Dalal, R.; Henry, R. Simultaneous Determination of Moisture, Organic Carbon, and Total Nitrogen by Near Infrared Reflectance Spectrophotometry. Soil Sci. Soc. Am. J.-SSSAJ 1986, 50, 120–123. [Google Scholar] [CrossRef]
  66. Wei, Z.; Meng, Y.; Zhang, W.; Peng, J.; Meng, L. Downscaling SMAP soil moisture estimation with gradient boosting decision tree regression over the Tibetan Plateau. Remote Sens. Environ. 2019, 225, 30–44. [Google Scholar] [CrossRef]
  67. Zhang, J.; Liu, T.; Wang, J.; Li, H.; Wang, Z.; Zhang, F.; Yuan, H. A simple but effective evaluation criterion for parameters optimization of EPO and its application to moisture insensitive prediction of soil organic matter. Chemom. Intell. Lab. Syst. 2023, 236, 104794. [Google Scholar] [CrossRef]
  68. Mirzaei, S.; Darvishi Boloorani, A.; Bahrami, H.A.; Alavipanah, S.K.; Mousivand, A.; Mouazen, A.M. Minimising the effect of moisture on soil property prediction accuracy using external parameter orthogonalization. Soil Tillage Res. 2022, 215, 105225. [Google Scholar] [CrossRef]
  69. Wijewardane, N.K.; Ge, Y.; Morgan, C.L.S. Moisture insensitive prediction of soil properties from VNIR reflectance spectra based on external parameter orthogonalization. Geoderma 2016, 267, 92–101. [Google Scholar] [CrossRef]
Agronomy 15 01367 g001
Figure 1. Overview of the study area. Geographic location of the Qinghai–Tibet Plateau (a). Location of the study area and distribution of field sampling points (b). Digital Elevation Model (DEM) of the study area (c). Spatial distribution of annual mean temperature (d). Spatial distribution of annual mean precipitation (e).
Figure 1. Overview of the study area. Geographic location of the Qinghai–Tibet Plateau (a). Location of the study area and distribution of field sampling points (b). Digital Elevation Model (DEM) of the study area (c). Spatial distribution of annual mean temperature (d). Spatial distribution of annual mean precipitation (e).
Agronomy 15 01367 g001
Agronomy 15 01367 g002
Figure 2. Field sampling images.
Figure 2. Field sampling images.
Agronomy 15 01367 g002
Agronomy 15 01367 g003
Figure 3. Original spectral reflectance curves of non-degraded (a) and degraded (b) mattic epipedon layers.
Figure 3. Original spectral reflectance curves of non-degraded (a) and degraded (b) mattic epipedon layers.
Agronomy 15 01367 g003
Agronomy 15 01367 g004
Figure 4. Hyperspectral reflectance profiles of mattic epipedon after different pretreatments: SG filtering (a), first-order differential processing (b), second-order differential processing (c), continuum removal processing (d), MSC transformation (e), and SNV transformation (f). Each line indicates a mattic epipedon layer.
Figure 4. Hyperspectral reflectance profiles of mattic epipedon after different pretreatments: SG filtering (a), first-order differential processing (b), second-order differential processing (c), continuum removal processing (d), MSC transformation (e), and SNV transformation (f). Each line indicates a mattic epipedon layer.
Agronomy 15 01367 g004
Agronomy 15 01367 g005
Figure 5. Hyperspectral reflectance curves of mattic epipedon under different water content conditions.
Figure 5. Hyperspectral reflectance curves of mattic epipedon under different water content conditions.
Agronomy 15 01367 g005
Agronomy 15 01367 g006
Figure 6. Differential hyperspectral reflectance curves of mattic epipedon under different water content conditions.
Figure 6. Differential hyperspectral reflectance curves of mattic epipedon under different water content conditions.
Agronomy 15 01367 g006
Agronomy 15 01367 g007
Figure 7. Hyperspectral reflectance curves of mattic epipedon processed via EPO moisture correction.
Figure 7. Hyperspectral reflectance curves of mattic epipedon processed via EPO moisture correction.
Agronomy 15 01367 g007
Table 1. Descriptive statistics of soil physical and chemical properties.
Table 1. Descriptive statistics of soil physical and chemical properties.
Determination IndexAverage ValueMinimum ValueMaximum ValueStandard DeviationCoefficient of Variation
Bulk density (g/cm3)0.72 0.24 1.30 0.18 0.25
Moisture content (%)0.23 0.02 1.60 0.24 1.04
Root volume ratio0.25 0.10 0.70 0.11 0.55
pH7.06 5.95 8.24 0.48 0.07
Organic carbon (g/kg)96.6616.72 319.21 48.74 0.50
Single fractal2.47 2.15 3.00 0.08 0.03
N (%)0.63 0.22 2.29 0.22 0.35
C (%)7.13 1.83 18.47 2.61 0.37
Thickness of felt layer (cm)10.45 4.40 31.90 3.71 0.36
Grazing intensity1.010.003.820.690.68
Table 2. Results of feature band screening using different preprocessing methods.
Table 2. Results of feature band screening using different preprocessing methods.
Preprocessing MethodsNumber of Characteristic BandsCharacteristic Bands (nm)
FD6446–450, 1914
SD10638–639, 662, 1921–1927
CR+FD22446–448, 506–511, 763, 768, 780–781, 792–793, 795–796, 1430–1433, 1436
CR+SD7637–639, 662–663, 2000, 2267
MSC+FD11446–449, 784, 792, 801, 2193–2196
MSC+SD13637–638, 661–663, 2134–2135, 2137, 2213, 2267–2269, 2295
SNV+FD7446–447, 449, 2194–2197
SNV+SD12638, 661–663, 2134, 2136–2137, 2213, 2267–2269, 2295
Table 3. Identification of mattic epipedon degradation using different preprocessing methods with the XGBoost model.
Table 3. Identification of mattic epipedon degradation using different preprocessing methods with the XGBoost model.
Preprocessing MethodsAccuracyPrecisionRecall
FD0.790.680.74
SD0.710.550.74
CR+FD0.750.650.57
CR+SD0.770.610.83
MSC+FD0.750.610.74
MSC+SD0.850.740.87
SNV+FD0.710.570.57
SNV+SD0.780.650.74
Table 4. Results of spectral characterization band screening under different water content conditions.
Table 4. Results of spectral characterization band screening under different water content conditions.
Moisture ContentNumber of Characteristic BandsCharacteristic Bands (nm)
5%102142, 2146–2147, 2213, 2265–2268, 2358–2359
10%102012, 2137–2139, 2141, 2178, 2204–2205, 2212–2213
15%12571–574, 771, 2154, 2274–2275, 2301–2302, 2307–2308
20%131424–1425, 1637, 2143–2144, 2203–2210
25%10615, 773, 1368, 2137, 2203–2205, 2207, 2308–2309
30%11728–729, 1427, 2138, 2209–2214, 2266
Table 5. Results of hyperspectral identification of mattic epipedon under different water content conditions.
Table 5. Results of hyperspectral identification of mattic epipedon under different water content conditions.
Moisture ContentAccuracyPrecisionRecall
5%0.810.70.8
10%0.810.760.73
15%0.750.590.83
20%0.760.740.69
25%0.680.640.57
30%0.650.60.43
Table 6. Results of spectral characterization band screening under different water content conditions, preceded by the EPO moisture correction algorithm.
Table 6. Results of spectral characterization band screening under different water content conditions, preceded by the EPO moisture correction algorithm.
Moisture ContentNumber of Characteristic BandsCharacteristic Bands (nm)
EPO—5%141429, 1435–1438, 1919–1920, 1992, 2011–2012, 2026, 2203–2204, 2312
EPO—10%16597, 639–643, 1114, 1421–1422, 1434, 1436–1438, 2259–2261
EPO—15%8686, 698–699, 1342–1344, 1961, 2302
EPO—20%4615, 1435–1436, 1901
EPO—25%8631, 1434–1436, 2210–2213
EPO—30%7617, 1066, 1432, 1699, 1855, 2358, 2360
Table 7. Results of hyperspectral identification of mattic epipedon under different water content conditions after EPO correction.
Table 7. Results of hyperspectral identification of mattic epipedon under different water content conditions after EPO correction.
Moisture ContentAccuracyPrecisionRecall
EPO—5%0.690.620.64
EPO—10%0.750.680.76
EPO—15%0.770.790.6
EPO—20%0.770.740.68
EPO—25%0.710.670.56
EPO—30%0.740.770.52
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Zhi, J.; Zhu, H.; Yang, J.; Yan, Q.; Zhi, D.; Sun, Z.; Ge, L.; Lv, C. Identification of Mattic Epipedon Degradation on the Northeastern Qinghai–Tibetan Plateau Using Hyperspectral Data. Agronomy 2025, 15, 1367. https://doi.org/10.3390/agronomy15061367

AMA Style

Zhi J, Zhu H, Yang J, Yan Q, Zhi D, Sun Z, Ge L, Lv C. Identification of Mattic Epipedon Degradation on the Northeastern Qinghai–Tibetan Plateau Using Hyperspectral Data. Agronomy. 2025; 15(6):1367. https://doi.org/10.3390/agronomy15061367

Chicago/Turabian Style

Zhi, Junjun, Hong Zhu, Jingwen Yang, Qiuchen Yan, Dandan Zhi, Zhongbao Sun, Liangwei Ge, and Chengwen Lv. 2025. "Identification of Mattic Epipedon Degradation on the Northeastern Qinghai–Tibetan Plateau Using Hyperspectral Data" Agronomy 15, no. 6: 1367. https://doi.org/10.3390/agronomy15061367

APA Style

Zhi, J., Zhu, H., Yang, J., Yan, Q., Zhi, D., Sun, Z., Ge, L., & Lv, C. (2025). Identification of Mattic Epipedon Degradation on the Northeastern Qinghai–Tibetan Plateau Using Hyperspectral Data. Agronomy, 15(6), 1367. https://doi.org/10.3390/agronomy15061367

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop