Field-scale soil moisture (SM) estimation from Sentinel-1 C-band SAR alone is challenged by vegetation, roughness, and spatial heterogeneity. This study proposes a Stacked Additive Boosting-based Model (SABM) that combines Sentinel-1 SAR, Sentinel-2 optical, and ancillary geophysical features with temporal lag SM features (SMlag1, SMlag2) derived from a station’s own antecedent in situ record, exploiting SM persistence at 12-day Sentinel-1 repeat intervals; the framework is accordingly intended for instrumented sites with historical SM observations rather than as a satellite-only retrieval method for ungauged locations. Validated at Little Washita Watershed, Oklahoma, USA (21 stations, 2016–2023), lag features improved temporal hold-out validation from R2 = 0.6255, RMSE = 0.0535 m3 m−3 to R2 = 0.7771, RMSE = 0.0413 m3 m−3 (23% RMSE reduction) and improved accuracy at held-out, instrumented stations from mean R2 = −0.6024 to R2 = 0.5134 across five station-level hold-out folds. Extended Triple Collocation indicated lower point-scale error for SABM (ETC RMSE = 0.0347 m3 m−3) than SMAP L3 Enhanced (0.0419 m3 m−3) at 13 of 20 stations at a nominal ~900× finer native pixel resolution (10 m Sentinel-1 pixel spacing vs. ~9 km SMAP footprint), though the effective support scale of the extracted features is coarser due to spatial averaging of the input features, and the ground reference’s own error term could not be reliably resolved at most stations, a limitation attributable to SABM’s use of antecedent ground observations as predictors. With cross-site transfer to REMEDHUS, Spain yielded R2 = 0.6729 without retraining, though this accuracy was attributable primarily to SM persistence rather than transferred SAR/optical relationships. Boruta–SHAP identified soil pH as the top-ranked predictor by SHAP importance, though follow-up ablation indicates that this ranking substantially reflects a replaceable, station-identity-correlated proxy rather than an indispensable physical driver; SHAP dependence patterns for seasonal forcing, vegetation density, and SAR polarisation ratio were physically coherent. These findings demonstrate that temporal lag features substantially enhance retrieval accuracy at instrumented, model-unseen stations, providing a robust and interpretable framework for operational field-scale SM monitoring at sites with antecedent soil moisture records.

Shallow water depth is a key dataset for marine science and coastal management. Active–passive fusion machine learning approaches that integrate ICESat-2 LiDAR with passive optical imagery provide an efficient solution for bathymetric retrieval in remote areas. However, current bathymetry research suffers from the problem of insufficient utilization of remote sensing features in the spatial dimension. To overcome this drawback, we propose a multi-feature convolutional neural network (CNN) for active–passive bathymetric retrieval. Reference depths derived from ICESat-2 ATL03 data are used as training labels, while 5×5 pixel-sized image patches (consisting of spectral reflectance and spatial position information) extracted from Sentinel-2 imagery serve as model inputs. Experimental results from the Buck Island provide a clear demonstration of the proposed method's superiority over traditional models like support vector regression, random forest, and XGBoost. The results indicate that fully mining remote sensing features in the spatial dimension through CNN can significantly improve the accuracy of bathymetry retrieval.

Vegetation is a critical factor influencing the identification of rock outcrops using hyperspectral remote sensing data. When mixed pixels containing both vegetation and rock are formed, the spectral signatures of vegetation can partially or fully obscure the diagnostic absorption features of rocks. Based on GaoFen-5 (GF-5) Advanced Hyperspectral Imager (AHSI) data, this study employs a linear spectral mixture model to simulate sparse vegetation–rock mixed pixels. The potential of high-frequency components derived from discrete wavelet transform (DWT) to enhance lithological discrimination within sparse vegetation–rock mixed spectra was analyzed, and the findings were validated using image spectra. The results show that andesite spectra are the most susceptible to vegetation interference. Absorption features in the 2.0–2.4 μm wavelength range were identified as critical indicators for distinguishing lithologies from mixed spectra. High-frequency components extracted through the DWT of the simulated mixed spectra using the Daubechies 8 wavelet function were found to significantly improve classification performance. As vegetation content (including green grass, golden grass, bushes, and lichens) increased from 5% to 60%, the average overall accuracy improved by 15% (from 0.51 to 0.66) after using high-frequency features. The average F1-scores for granite and sandstone increased by 0.12 (from 0.68 to 0.80) and 0.20 (from 0.48 to 0.68), respectively. For AHSI image spectra, the use of high-frequency features resulted in F1-score improvements of 0.48, 0.11, and 0.09 for tuff, granite, and limestone, respectively. Although the identification of andesite remains challenging, this study provides a promising approach for improving lithological mapping accuracy using GF-5 hyperspectral data, particularly in humid and semi-humid regions.

Water depth in shallow marine environments is a fundamental parameter for oceanographic research and coastal engineering applications. High-resolution satellite imagery and long-term medium-resolution imagery offer significant potential for detailed bathymetric mapping and monitoring spatiotemporal variations in bathymetry. However, most of these images contain only three visible bands (blue, green, and red), making bathymetric mapping from such images challenging in practical applications. For the empirical approach, high-quality in situ depth calibration data, which are essential for establishing a reliable empirical bathymetric model, are either unavailable or excessively expensive. For the physics-based approach, images containing only three visible bands can be problematic in accurately deriving depths. To address this limitation, this study proposes a novel semi-analytical-empirical hybrid model for water depth retrieval. The core of the proposed method is the integration of a semi-analytical model with a physics-based dual-band model. This integration quantifies the relative depth relationships among pixels and uses them as a physical constraint. Through this constraint, the method identifies physically reliable depth estimates from the multiple numerical solutions of the semi-analytical model for a subset of shallow-water pixels, which then serve as an in situ–free calibration dataset. This dataset is subsequently used to determine the empirically based optimal retrieval model, which is finally applied to generate the complete bathymetric map. The results from four typical coral reef regions—Buck Island, Yongxing Island, Kaneohe Bay, and Yongle Atoll—demonstrated that the proposed model achieved root-mean-square errors (RMSE) of 0.98–1.62 m, mean absolute errors (MAE) of 0.73–1.13 m, and coefficients of determination (R2) of 0.91–0.95 in comparison to in situ measurements. Compared to both the physics-based dual-band model and the L-S model (i.e., the bathymetry mapping approach combining Log-ratio and Semi-analytical models), the proposed model reduced the RMSE by 9–55%, reduced the MAE by 4–56%, and improved the R2 by 0.01–0.29. Additionally, the accuracy of the proposed model surpasses that of both the physics-based dual-band model and the L-S model across all depth intervals, particularly in deeper depth waters (>15 m). This study offers a robust solution for bathymetric mapping in areas lacking in situ depth data and contributes significantly to advancing optical remote sensing techniques for underwater topography detection.

Hyperspectral data contain multiple similar bands, making texture feature extraction while avoiding data redundancy a challenge for rock classification. Three-dimensional Discrete Wavelet Transform (3D-DWT) can both extract spatial and spectral features from hyperspectral data, enhancing rock image textures and providing more usable information for rock classification. This paper utilizes Gaofen-5 Advanced HyperSpectral Imager (AHSI) data from the Eastern Tianshan region of China to extract multi-scale, 3D wavelet features for different rock units. Jeffries-Matusita distance is applied for feature ranking, with a Bayesian-optimized random forest classifier selected for classification. The traditional texture extraction method—the grayscale co-occurrence matrix (GLCM) —is used for comparison. For lithological mapping, the classification results using 3D-DWT statistical textures achieved an overall accuracy of 79.71%, outperforming both original statistical texture features (42.72%) and original pixel values (52.08%). For rock unit classification, neither reflectance images nor wavelet-transformed images were suitable for texture extraction by GLCM. When texture features can assist hyperspectral data in rock classification, remote sensing rock classification accuracy in complex terrain will be further enhanced.

A procedure designed to develop probabilistic thresholds for rainfall-induced landslides adopting novel rainfall variables is proposed and tested in an area of Campania region, southern Italy. The dataset used comprises 180 rainfall-induced landslides in the period 2010–2020, derived from the FraneItalia catalogue ( https://franeitalia.wordpress.com/ ), and rainfall records retrieved from a network of 22 regional rain gauges active in the area. The procedure starts by reconstructing rainfall events responsible and not responsible for landslides in the period of analysis. Subsequently, a large number of rainfall variables are derived automatically, adopting a time series feature extraction tool, and their significance in identifying landslide-triggering events is evaluated. Then, adopting a Bayesian approach, the most significant rainfall variables are used to define a set of probabilistic thresholds in one, two, and three dimensions. Finally, the most effective thresholds are identified by means of standard performance indicators. The probabilistic thresholds developed using novel rainfall variables outperform those employing conventional variables. Specifically, looking at the minimum distance from the perfect classification point (δ), thresholds employing novel variables yield a minimum δ of 0.230, while those adopting conventional variables lead to a minimum δ of 0.292. The results achieved herein demonstrate that the use of novel rainfall variables within territorial landslide warning models can represent a promising option for improving the performance of these models.

Shallow water bathymetry is critical for environmental monitoring and maritime security. Current widely used statistical models based on passive optical satellite remote sensing often rely on prior bathymetric data, limiting their application to regions lacking such information. In contrast, the physics-based dual-band log-linear analytical model (P-DLA) can estimate shallow water bathymetry without in situ measurements, offering significant potential. However, the quasi-analytical algorithm (QAA) used in the P-DLA is sensitive to non-ideal pixels, resulting in unstable bathymetry estimation. To address this issue and evaluate the potential of SuperDove imagery for bathymetry estimation in regions without prior bathymetric data, this study proposes an improved physics-based dual-band model (IPDB). The IPDB replaces the QAA with a spectral optimization algorithm that integrates deep and shallow water sample pixels to estimate diffuse attenuation coefficients for the blue and green bands. This allows for more accurate estimation of shallow water bathymetry. The IPDB was tested on SuperDove images of Dongdao Island, Yongxing Island, and Yongle Atoll. The results showed that SuperDove images are capable of estimating shallow water bathymetry in regions without prior bathymetric data. The IPDB achieved Root Mean Square Error (RMSE) values below 1.7 m and R2 values above 0.89 in all three study areas, indicating strong performance in bathymetric estimation. Notably, the IPDB outperformed the standard P-DLA model in accuracy. Furthermore, this study outlines four sampling principles that, when followed, ensure that variations in the spatial distribution of sampling pixels do not significantly impact model performance. This study also showed that the blue–green band combination is optimal for the analytical expression of the physics-based dual-band model.

The Lueyang and Xunyang counties in the Shaanxi province (China) are highly susceptible to rainfall-induced landslides. Rainfall thresholds are the most used tool to predict the occurrence of rainfall-induced landslides over large areas. However, the definition of robust thresholds may be difficult for unbalanced datasets, for which the number of non-landslide observations is much higher than the number of landslide observations. This study aims at defining adequate rainfall thresholds for the two study areas using landslide datasets that are strongly unbalanced in terms of occurrences vs. non-occurrences. Two types of rainfall thresholds are determined using a frequentist method at several non-exceedance and exceedance probabilities, separately considering rainfall events responsible for landslides (positive thresholds) and rainfall events not responsible for landslides (negative thresholds). The comparison between the two sets of thresholds shows that the method based on non-triggering events allows defining rainfall thresholds characterized by lower uncertainties and a better performance than the ones defined considering the triggering events, in both the study areas. In particular, the best-performing thresholds are the negative threshold defined at 15% exceedance probability for the Lueyang county and the negative threshold defined at 20% exceedance probability for the Xunyang county.

The high intra-class variability of rock spectra is an important factor affecting classification accuracy. The discrete wavelet transform (DWT) can capture abrupt changes in the signal and obtain subtle differences between the spectra of different rocks. Taking laboratory spectra and hyperspectral data as examples, high-frequency features after DWT were used to improve the discrimination accuracy of rocks. Various decomposition levels, mother wavelet functions, and reconstruction methods were used to compare the accuracy. The intra-class variability was measured using the intra-class Spectral Angle Mapper (SAM). Our results show that the high-frequency features could improve the discrimination accuracy of laboratory spectra by 13.4% (from 46.5% to 59.9%), compared to the original spectral features. The accuracy of image spectra in two study areas increased by 8.6% (from 68.3% to 76.9%) and 7.2% (from 81.3% to 88.5%), respectively. Haar wavelets highlighted the spectral differences between different rocks. After DWT, intra-class SAM reduced and intra-class variability of rocks decreased. The Pearson correlation coefficient indicated a negative correlation between intra-class variability and overall accuracy. It suggested that improving classification accuracy by reducing intra-class variability was feasible. Though the result of lithological mapping still leaves room for improvement, this study provides a new approach to reduce intra-class variability, whether using laboratory spectra or hyperspectral data.

Nearshore water depth plays a crucial role in scientific research, navigation management, coastal zone protection, and coastal disaster mitigation. This study aims to address the challenge of insufficient feature extraction from remote sensing data in nearshore water depth inversion. To achieve this, a convolutional neural network with spatial location integration (CNN-SLI) is proposed. The CNN-SLI is designed to extract deep features from remote sensing data by considering the spatial dimension. In this approach, the spatial location information of pixels is utilized as two additional channels, which are concatenated with the input feature image. The resulting concatenated image data are then used as the input for the convolutional neural network. Using GF-6 remote sensing images and measured water depth data from electronic nautical charts, a nearshore water depth inversion experiment was conducted in the waters near Nanshan Port. The results of the proposed method were compared with those of the Lyzenga, MLP, and CNN models. The CNN-SLI model demonstrated outstanding performance in water depth inversion, with impressive metrics: an RMSE of 1.34 m, MAE of 0.94 m, and R2 of 0.97. It outperformed all other models in terms of overall inversion accuracy and regression fit. Regardless of the water depth intervals, CNN-SLI consistently achieved the lowest RMSE and MAE values, indicating excellent performance in both shallow and deep waters. Comparative analysis with Kriging confirmed that the CNN-SLI model best matched the interpolated water depth, further establishing its superiority over the Lyzenga, MLP, and CNN models. Notably, in this study area, the CNN-SLI model exhibited significant performance advantages when trained with at least 250 samples, resulting in optimal inversion results. Accuracy evaluation on an independent dataset shows that the CNN-SLI model has better generalization ability than the Lyzenga, MLP, and CNN models under different conditions. These results demonstrate the superiority of CNN-SLI for nearshore water depth inversion and highlight the importance of integrating spatial location information into convolutional neural networks for improved performance.

Multispectral satellite data allow experts to discriminate rock units based on their spectral signature characteristics. Here, Sentinel-2, ASTER and the Landsat-8 Operational Land Imager (OLI) were assessed for lithological mapping by using a random forest (RF) classifier for a study area located in Xitieshan, Northwest China. The classification accuracy of Sentinel-2 was 60.71%, which was 5.24% and 4.77% higher than the accuracies for ASTER and the Landsat-8 OLI, respectively. Three image enhancement techniques, namely, principal component analysis (PCA), independent component analysis (ICA) and minimum noise fraction (MNF), were used with grey-level cooccurrence matrices (GLCMs) to increase the quality of the input datasets. The ICA could discriminate between rock unit datasets better than the other approaches. In contrast, GLCM performed poorly when used independently. The overall classification accuracies were 60.71%, 62.63%, 64.34%, 65.21% and 58.87% for the 10 bands of Sentinel-2, PCA, MNF, ICA and GLCM, respectively. Then, five datasets were combined as a single group and applied in RF classification. Sentinel-2 obtained an overall accuracy of 73.96% and performed better than the other single-dataset approaches used in this study. Furthermore, the classification result of RF was achieved better performance than that of the support vector machine algorithm (SVM). During feature selection processing, ReliefF, the most successful pre-processing algorithm, was employed to preliminarily perform feature screening. Then, the optimal dataset was selected on the basis of the importance ranking of RF. A total of 20 more important predictors were selected from 114 original features using the ReliefF-RF model. These predictors were used in the lithological mapping, and an overall accuracy of 77.63% was reached.

Most of the land surface water in the plateau region is frozen. Due to the huge difference in the dielectric constant of water and ice, this paper generated a ten-year surface permafrost change map from 1992 to 2001 based on SSM/I passive microwave remote sensing data, and the results showed that from 1992 to 2001, the annual average frozen area in Tibet showed a downward trend, which also verified the general trend of global warming. According to the comparison between the images generated by the two algorithms and the actual situation, it can be seen that the freeze-thaw discriminant algorithm has better accuracy than the double-index algorithm under the condition of ascending orbit. The results of this study can have important reference value for analyzing the changes of permafrost in Tibet for a long time series and selecting the appropriate soil freeze-thaw discrimination algorithm, and also provide an SSM/I sensor image processing method.

With continuous improvements in oil production, the environmental problems caused by oil exploitation are becoming increasingly serious. Rapid and accurate estimation of soil petroleum hydrocarbon content is of great significance to the investigation and restoration of environments in oil-producing areas. In this study, the content of petroleum hydrocarbon and the hyperspectral data of soil samples collected from an oil-producing area were measured. For the hyperspectral data, spectral transforms, including continuum removal (CR), first- and second-order differential (CR-FD, CR-SD), and Napierian logarithm (CR-LN), were applied to eliminate background noise. At present, there are some shortcomings in the method of feature band selection, such as large quantity, time of calculation, and unclear importance of each feature band obtained. Meanwhile, redundant bands easily exist in the feature set, which seriously affects the accuracy of the inversion algorithm. In order to solve the above problems, a new method (GARF) for hyperspectral characteristic band selection was proposed. It combined the advantage that the grouping search algorithm can effectively reduce the calculation time with the advantage that the point-by-point search algorithm can determine the importance of each band, which provided a clearer direction for further spectroscopic research. The 17 selected bands were used as the input data of partial least squares regression (PLSR) and K-nearest neighbor (KNN) algorithms to estimate soil petroleum hydrocarbon content, and the leave-one-out method was used for cross-validation. The root mean squared error (RMSE) and coefficient of determination (R2) of the estimation result were 3.52 and 0.90, which implemented a high accuracy with only 8.37% of the entire bands. The results showed that compared with the traditional characteristic band selection methods, GARF can effectively reduce the redundant bands and screen out the optimal characteristic bands in the hyperspectral data of soil petroleum hydrocarbon with the method of importance assessment, which retained the physical meaning. It provided a new idea for the research of other substances in soil.

As a supplement to empirical-statistical methods, physically based methods can be employed to define rainfall thresholds for triggering landslides in areas lacking records of landslides. The transient rainfall infiltration and grid-based regional slope-stability model (TRIGRS), as a physically based model, has been applied to define rainfall thresholds at the basin, slope unit, and grid cell scales. However, as far as we know to date, the influence of soil mechanical and hydraulic parameters on defining rainfall thresholds at various scales has not been comprehensively evaluated. In this study, TRIGRS was used to define rainfall intensity (I) and duration (D) thresholds at various scales for Buzhe village, Pu’an county, Guizhou province, China, under the conditions of different soil physical parameters. The results show that the number of rainfall thresholds decreased with cohesion (c) and internal friction angle (φ) and increased with soil unit weight (γs), excluding the basin scale. Threshold position varied positively with c and φ and negatively with γs. Soil mechanical parameters have a greater influence on the definition of rainfall thresholds based on TRIGRS than hydraulic parameters.

The development of advanced and efficient methods for mapping and monitoring wetland regions is essential for wetland resources conservation, management, and sustainable development. Although remote sensing technology has been widely used for detecting wetlands information, it remains a challenge for wetlands classification due to the extremely complex spatial patterns and fuzzy boundaries. This study aims to implement a comprehensive and effective classification scheme for wetland land covers. To achieve this goal, a novel object-based multigrained cascade forest (OGCF) method with multisensor data (including Sentinel-2 and Radarsat-2 remote sensing imagery) was proposed to classify the wetlands and their adjacent land cover classes in the wetland National Natural Reserve. Moreover, a hybrid selection method (ReliefF-RF) was proposed to optimize the feature set in which the spectral and polarimetric decomposition features are contained. We obtained six spectral features from visible and shortwave infrared bands and 10 polarimetric decomposition features from the H/A/Alpha, Pauli, and Krogager decomposition methods. The experimental results showed that the OGCF method with multisource features for land cover classification in wetland regions achieved the overall accuracy and kappa coefficient of 88.20% and 0.86, respectively, which outperformed the support vector machine (SVM), extreme gradient boosting (XGBoost), random forest (RF), and deep neural network (DNN). The accuracy of the wetland classes ranged from 75.00% to 97.53%. The proposed OGCF method exhibits a good application potential for wetland land cover classification. The classification scheme in this study will make a positive contribution to wetland inventory and monitoring and be able to provide technical support for protecting and developing natural resources.

Earthquake disasters are frequent, and the seismic intensity is large in Northeast China. Earthquake activity research is an important aspect of earthquake disaster management. We chose some unconventional means to study fault activity, to find updated activity evidence. The Ms 5.3 earthquake occurred near the Fuyu North Fault (FNF) of China on May 27, 2018. Using the Sentinel-1B descending orbit data from 2016 to 2019, the line-of-sight (LOS) surface deformation in the study area was calculated by using the small baseline subset (SBAS) method. After transforming to the horizontal EW deformation, the variance component estimation method was used for fusion reconstruction with the EW data of the surrounding GNSS stations. The polynomial least square method is used to fit the fault slip rate of three EW data on the surface trace of the FNF. The fitting results of the three regions show that the horizontal eastward distribution rate of the upper plate is significantly greater than that of the lower plate, which is left-lateral clockwise torsion. The vertical structural deformation caused by the growth strata of the upper and lower plates of the upper SYT2 seismic profile of the FNF is quantitatively calculated, and the thrust rate of the upper plate is 0.2 mm/y relative to that of the lower plate. Based on the Li Siguang chessboard structure model, we found that the compression stress in the north-south direction is gradually weakened, and the compression stress in the east-west direction is gradually enhanced. Through the Coulomb stress analysis, the three events of CMT only induced the historical focal location of the surrounding part. The events of 2017 did not induce the events of 2018, but the events of 2019 were related to the induced effects of 2017 and 2018.

The Transient Rainfall Infiltration and Grid-Based Regional Slope-Stability (TRIGRS) model has been widely used to define rainfall thresholds for triggering shallow landslides. In this study, the rainfall intensity(I)-duration(D) thresholds for multiple slope units of an area in Pu’an County, Guizhou Province, China were defined based on TRIGRS. Given that TRIGRS is used to simulate the slope stability under the conditions of a given increasing sequence of I-D data, if the slope reaches instability at I = a, D = b, it will also become unstable in the case of I = a, D > b or I > a, D = b. To explore the effect of these I-D data with the same I or D values on the definition of I-D thresholds and the best method to exclude these data, two screening methods were used to exclude the I-D data that caused instability in the TRIGTS simulation. First, I-D data with the same I values when D values are greater than a certain limit value were excluded. Second, several D values were selected to exclude I-D data with the same I values for a slope unit. Then, an I value was selected to exclude I-D data with the same D values. After screening, two different I-D thresholds were defined. The comparison with the thresholds defined without screening shows that the I-D data with the same I or D values will reduce the accuracy of thresholds. Moreover, the second screening method can entirely exclude these data.

Northeast China is an important ecological barrier and commodity grain base in China. The coupling coordination relationship between geology-geomorphology and ecology has become a critical background condition for ecosystem protection and sustainable development. Taking Northeast China as a case (accounting for about 13% of China's land area), 9 divisions are divided according to the characteristics of regional ecology and geology-geomorphology, and 17 indicators are selected to build an evaluation index system. Methods of analytic hierarchy process, entropy weight and game theory are used to determine the index weights. Based on the coupling coordination degree (CCD) model, the spatial coupling coordination characteristics of geology-geomorphology and ecology are studied. The variation characteristics of the Normalized Difference Vegetation Index (NDVI) are evaluated by Sen+Mann-Kendall (Sen+MK) method. Our results are as follows. (1) The coupling between geology-geomorphology and ecology is strong, but the spatial differentiation of CCD is obvious. Nine divisions are evaluated as two high-level, three medium-level and three low-level coordination types and one mild imbalance type. (2) The plain divisions Ⅰ and Ⅳ where the typical black soil belt is located are high coordination types. Restricted by geology-geomorphological conditions or ecological conditions, mountain divisions Ⅲ and Ⅶ and plain division Ⅴ are moderate coordination types, mountain divisions Ⅱ and Ⅷ and plateau division Ⅸ are low coordination types, and mountain division Ⅵ is mild imbalance type. (3) The variation trend of NDVI shows a significant increase in divisions Ⅲ, Ⅴ, Ⅰ, Ⅱ and Ⅶ. it shows a significant decrease in part of divisions Ⅳ, Ⅵ, Ⅷ and Ⅸ, and ecological management and construction should be strengthened in these divisions. The research shows that the CCD model method is feasible for evaluating the relationship between geology-geomorphology and ecology and can provide eco-geological background information for Northeast China.

为实现黑土有机质含量更准确的估测,提出了基于小波变换和连续投影算法的高光谱估测方法.以典型黑土区采集的土壤样品为研究对象、分析光谱设备(analytical spectral devices,ASD)光谱仪获取的可见光—近红外区间光谱数据和经化学分析得到的土壤有机质含量为数据源,首先采用小波变换提取1～7层小波低频系数,然后利用连续投影算法分别对土壤全谱和1～7层小波低频系数进行变量筛选,最后分别基于土壤全谱、1～7层小波低频系数、连续投影算法选择的变量,利用偏最小二乘和支持向量机两种方法建立估测模型.结果表明:经小波变换和连续投影算法处理后,不仅变量数目得到了大幅降低,而且模型精度也进一步得到了提高,采用偏最小二乘法时,R2由土壤全谱的0.79提高至第6层小波低频系数的0.93,RMSE由6.06 g·kg-1降低至3.48 g·kg-1;采用支持向量机方法时,R2由土壤全谱的0.75提高至第3层小波低频系数的0.91,RMSE由7.46 g·kg-1降低至4.12 g·kg-1,说明提出的方法能有效用于黑土有机质含量高光谱估测.