Pseudo-invariant Features Research Articles

Detection of Coral Reef Bleaching by Multitemporal Sentinel-2 Data Using the PU-Bagging Algorithm: A Feasibility Study at Lizard Island

Coral reef bleaching events have become more frequent all over the world and pose a serious threat to coral reef ecosystems. Therefore, there is an urgent need for better detection of coral reef bleaching in a time- and cost-saving manner. In recent years, remote sensing technology has often been utilized and gained recognition for coral reef bleaching detection. However, bleaching corals in the water always have weak spectral change signals, causing difficulties in using remote sensing data. Additionally, uneven change samples make it challenging to adequately capture the details of coral reef bleaching detection and produce thematic maps. To resolve these problems, a novel method named coral reef bleaching detection by positive-unlabeled bagging (CBD-PUB) is proposed in this paper. To test the capacity of the method, a series of multi-temporal Sentinel-2 remote sensing images are utilized, and Lizard Island in Australia is taken as a case study area. The pseudo-invariant feature atmospheric correction (PIF) algorithm is adopted to improve coral reef bleaching spectral signals. After that, CBD-PUB is employed to effectively explore coral reef bleaching variation and its corresponding influence relations. The experimental results show that the overall accuracy of bleaching detection by the proposed algorithm reaches 92.1% and outperforms the traditional method. It fully demonstrates the feasibility of the model for the field of coral reef bleaching detection and provides assistance in the monitoring and protection of coral environments.

Federated learning via reweighting information bottleneck with domain generalization

Federated learning (FL) plays an important role in collaborative distributed modeling. However, most studies cannot address poor generalization of out-of-distribution (OoD) data. Efforts have been exerted to address data heterogeneity among participants, but yielding limited success. Here, we propose an information bottleneck based FL method (FedIB), which aims to build a model with better OoD generalization. We extract the domain-invariance of different source domains to mitigate the domain heterogeneity under the cross-silo scenarios. Next, given the scale imbalance, we balance the representation importance of different domains with reweighting a better invariance across multiple domains. In addition, the convergence of FedIB is analyzed. As opposed to aligning distributions or eliminating redundancy by previous methods, FedIB achieves better domain generalization explicitly by eliminating the pseudo-invariant features. Finally, we conduct extensive experiments on various datasets revealing that FedIB has superior performances facing OoD and scale imbalance scenarios in distributed modeling.

LIRRN: Location-Independent Relative Radiometric Normalization of Bitemporal Remote-Sensing Images.

Relative radiometric normalization (RRN) is a critical pre-processing step that enables accurate comparisons of multitemporal remote-sensing (RS) images through unsupervised change detection. Although existing RRN methods generally have promising results in most cases, their effectiveness depends on specific conditions, especially in scenarios with land cover/land use (LULC) in image pairs in different locations. These methods often overlook these complexities, potentially introducing biases to RRN results, mainly because of the use of spatially aligned pseudo-invariant features (PIFs) for modeling. To address this, we introduce a location-independent RRN (LIRRN) method in this study that can automatically identify non-spatially matched PIFs based on brightness characteristics. Additionally, as a fast and coregistration-free model, LIRRN complements keypoint-based RRN for more accurate results in applications where coregistration is crucial. The LIRRN process starts with segmenting reference and subject images into dark, gray, and bright zones using the multi-Otsu threshold technique. PIFs are then efficiently extracted from each zone using nearest-distance-based image content matching without any spatial constraints. These PIFs construct a linear model during subject-image calibration on a band-by-band basis. The performance evaluation involved tests on five registered/unregistered bitemporal satellite images, comparing results from three conventional methods: histogram matching (HM), blockwise KAZE, and keypoint-based RRN algorithms. Experimental results consistently demonstrated LIRRN's superior performance, particularly in handling unregistered datasets. LIRRN also exhibited faster execution times than blockwise KAZE and keypoint-based approaches while yielding results comparable to those of HM in estimating normalization coefficients. Combining LIRRN and keypoint-based RRN models resulted in even more accurate and reliable results, albeit with a slight lengthening of the computational time. To investigate and further develop LIRRN, its code, and some sample datasets are available at link in Data Availability Statement.

A Novel Atmospheric Correction for Turbid Water Remote Sensing

For the remote sensing of turbid waters, the atmospheric correction (AC) is a key issue. The “black pixel” assumption helps to solve the AC for turbid waters. It has proved to be inaccurate to regard all water pixels in the SWIR (Short Wave Infrared) band as black pixels. It is necessary to perform atmospheric correction in the visible bands after removing the radiation contributions of water in the SWIR band. Here, the modified ACZI (m-ACZI) algorithm was developed. The m-ACZI assumes the spatial homogeneity of aerosol types and employs the BPI (Black Pixel Index) and PIFs (Pseudo-Invariant Features) to identify the “black pixel”. Then, the radiation contributions of waters in the SWIR band are removed to complete the atmospheric correction for turbid waters. The results showed that the m-ACZI had better performance than the SeaDAS (SeaWiFS Data Analysis System) -SWIR and the EXP (exponential extrapolation) algorithm in the visible band (sMAPE < 30.71%, RMSE < 0.0111 sr−1) and is similar to the DSF (Dark Spectrum Fitting) algorithm in floating algae waters. The m-ACZI algorithm is suitable for turbid inland waters.

Radiometric Normalization Using a Pseudo−Invariant Polygon Features−Based Algorithm with Contemporaneous Sentinel−2A and Landsat−8 OLI Imagery

As sensor parameters and atmospheric conditions create uncertainties for at−sensor radiation detection, radiometric consistency among satellite images is difficult to achieve. Relative radiometric normalization is a method that can improve multi−image consistency with accurate pseudo−invariant features (PIFs), especially over large areas or long time series satellite images. Although there are algorithms that manually or automatically select PIFs, the spatial mismatch of satellite images can affect PIF extraction, particularly with artificial pixels. To alleviate this problem, we proposed to use Landsat−8 OLI as the reference image and Sentinel−2A as the subject image, to apply pseudo−invariant features−based algorithms with polygon features through the single−band and multiple−band regression. Compared to pseudo−invariant point features, hyperspectral library, and histogram matching approaches, the results demonstrate the superiority of the proposed algorithms with correlation coefficients of 0.9948 and 0.9945, and an RMSE of 0.0097 and 0.0095 with multiple− and single−band regression, respectively. We also found more accurate linear fitting and better shape matching through band scattering and reflectance frequency analysis. The proposed algorithms are a significant improvement in radiometric normalization, within artificial pixels, achieving spectral signature consistency.

Unsupervised Domain Adaptation Semantic Segmentation of High-Resolution Remote Sensing Imagery With Invariant Domain-Level Prototype Memory

Semantic segmentation is a key technique involved in automatic interpretation of high-resolution remote sensing (HRS) imagery and has drawn much attention in the remote sensing community. Deep convolutional neural networks (DCNNs) have been successfully applied to the HRS imagery semantic segmentation task due to their hierarchical representation ability. However, the heavy dependency on a large number of training data with dense annotation and the sensitiveness to the variation of data distribution severely restrict the potential application of DCNNs for the semantic segmentation of HRS imagery. This study proposes a novel unsupervised domain adaptation semantic segmentation network (MemoryAdaptNet) for the semantic segmentation of HRS imagery. MemoryAdaptNet constructs an output space adversarial learning scheme to bridge the domain distribution discrepancy between source domain and target domain and to narrow the influence of domain shift. Specifically, we embed an invariant feature memory module to store invariant domain-level context information because the features obtained from adversarial learning only tend to represent the variant feature of current limited inputs. This module is integrated by a category attention-driven invariant domain-level context aggregation module to current pseudo invariant feature for further augmenting the pixel representations. An entropy-based pseudo label filtering strategy is used to update the memory module with high-confident pseudo invariant feature of current target images. Extensive experiments under three cross-domain tasks indicate that our proposed MemoryAdaptNet is remarkably superior to the state-of-the-art methods.

Building consistent time series night-time light data from average DMSP/OLS images for indicating human activities in a large-scale oceanic area

• Random Forest algorithm and a stepwise intercalibration approach were combined. • Multi-filter and bathymetric images were composited to improve intercalibration. • Noise impact on the intercalibration was greatly eliminated. • Consistent time series night-time light data specialized for large-scale oceanic area was build. • Total light index was implied as an effective indicator of ocean fishery activities. Human activities in the ocean have never been chronically and continuously investigated on a large scale. Night-time light (NTL) images collected by the Defense Meteorological Satellite Program/Operational Linescan System (DMSP/OLS) have been used as a proxy for monitoring the distribution and intensity of some human activities in the ocean from 1992 to 2013. However, systematic radiometric biases exist among the average visible-light DMSP/OLS NTL images (DMSP avg ) derived from different satellites. Moreover, the high randomness of fishing vessel locations and the large amount of noise impede the intercalibration of DMSP avg . To address these issues, this study has developed a method for generating a series of consistent NTL images from 1992 to 2013 for a large-scale oceanic area. A composite image was first constructed by combining the original DMSP avg , median, and standard deviation filter images derived from the DMSP avg , and a bathymetry image. Thereafter, Random Forest (RF) algorithm was employed to classify the composite image into effective and noisy pixels. Finally, a stepwise intercalibration method was adopted to reduce the systematic radiometric biases in the denoised images. The experimental results showed that RF had an overall accuracy of 96% and a Kappa coefficient of 0.775. Furthermore, the intercalibration was shown to significantly reduce the systematic radiometric biases owing to the noises being effectively discarded by the RF. Specifically, the Sum Normalized Different Index (SNDI) of the images intercalibrated by the proposed method can reach 0.61, which is 68.2% less than that of the original DMSP avg . In addition, the correlation coefficients between the intercalibrated DMSP avg and fishery catches in the exclusive economic zones (EEZs) of Japan and Malaysia can reach 0.949 and 0.901, respectively, which are higher than other values, such as the one intercalibrated using the Pseudo-Invariant Features (PIFs) method. In summary, the proposed method has been proven to be effective and feasible for generating consistent time-series NTL data for a large-scale oceanic area, and the derived Total Light Index (TLI) is an effective indicator of ocean fishery activities for ocean ecosystem research and related applications.

The radiometric accuracy of the 8-band multi-spectral surface reflectance from the planet SuperDove constellation

• The absolute mean error of the SuperDove surface reflectance were 6%. • The SuperDove surface reflectance was on average almost identical to Sentinel-2. • The probability was 90% of acquiring stable multi-satellite surface reflectance. • 68% of single-satellite data were more likely to acquire stable surface reflectance. • The SuperDove time series was strongly correlated to Sentinel-2 on vegetation. Advances in the capabilities of commercial CubeSat constellations have enabled the retrieval of multi-spectral surface reflectance data over the Earth’s terrestrial surfaces on an almost daily basis. For example, while the earliest versions of Planet’s CubeSats provided tri-band and then quad-band optical image data, the most recent iterations deliver imaging capabilities with eight unique spectral bands for Earth system monitoring. To determine the utility of these rich geospatial data collections for a range of applications, it is necessary to characterise their radiometric accuracy. Leveraging on-ground spectroradiometer measurements of radiometrically pseudo-invariant features, we assess the absolute accuracy of an annual sequence of Planet SuperDove surface reflectance data. Date-coincident and spectrally overlapping Sentinel-2 image data were also used to assess the relative radiometric accuracy of the CubeSat reflectance data. Additionally, confidence levels for acquiring consistent SuperDove surface reflectance data were calculated, and the multi-temporal patterns of surface reflectance between the SuperDove and Sentinel-2 surface reflectance products were evaluated by examining their rank correlation. Our findings demonstrate that (1) the average accuracy of the Coastal-Blue, Blue, Green I, and Green II SuperDove surface reflectance bands was 5% higher than for the Yellow, Red, Red-Edge, and Near-Infrared bands; (2) the SuperDove surface reflectance data was on average almost identical to the surface reflectance derived from the coincident Sentinel-2 data; and (3) the radiometric accuracies (i.e., mean errors) can be improved by using band combinations (e.g. normalised difference vegetation index). However, due to the different radiometric performance between the spectral bands, some vegetation indices (e.g. the Yellowness Index) did not provide a linear relationship between the SuperDove and reference data. The probability of acquiring surface reflectance data with less than 5% reflectance variation was approximately 90% for the annual multi-satellite dataset. On average, 68% of the data derived from a single satellite had at least 95% probability of acquiring surface reflectance with less than 5% variation. The time series patterns of the SuperDove reflectance data had an average rank correlation coefficient of 0.67 for all types of surface units and 0.77 for vegetated surfaces when compared with the spectrally overlapping bands of the Sentinel-2 surface reflectance data. In addition to providing confidence levels for users to directly assess the radiometric accuracies of SuperDove surface reflectance products, we also provide suggestions to improve the data quality from an end-user’s perspective.

A consistent and corrected nighttime light dataset (CCNL 1992–2013) from DMSP-OLS data

Remote sensing of nighttime light can observe the artificial lights at night on the planet’s surface. The Defense Meteorological Satellite Program’s Operational Line Scan (DMSP-OLS) data (1992–2013) provide planet-scale nighttime light data over a long-time span and have been widely used in areas such as urbanization monitoring, socio-economic parameters estimation, and disaster assessment. However, due to the lack of an on-board calibration system, sensor design defects, limited light detection range, and inadequate quantization levels, the applications of DMSP-OLS data are greatly limited by interannual inconsistency, saturation, and blooming problems. To address these issues, we used the power function model based on pseudo-invariant feature, the saturation correction method based on regression model and radiance-calibrated data (SARMRC), and the self-adjusting model (SEAM) to improve the quality of DMSP data, and generated a Consistent and Corrected Nighttime Light dataset (CCNL 1992–2013). CCNL dataset shows good performance in interannual consistency, spatial details of urban centers, and light blooming, which is helpful to fully explore the application potentials of long time series nighttime light data.

Invariant Information Bottleneck for Domain Generalization

Invariant risk minimization (IRM) has recently emerged as a promising alternative for domain generalization. Nevertheless, the loss function is difficult to optimize for nonlinear classifiers and the original optimization objective could fail when pseudo-invariant features and geometric skews exist. Inspired by IRM, in this paper we propose a novel formulation for domain generalization, dubbed invariant information bottleneck (IIB). IIB aims at minimizing invariant risks for nonlinear classifiers and simultaneously mitigating the impact of pseudo-invariant features and geometric skews. Specifically, we first present a novel formulation for invariant causal prediction via mutual information. Then we adopt the variational formulation of the mutual information to develop a tractable loss function for nonlinear classifiers. To overcome the failure modes of IRM, we propose to minimize the mutual information between the inputs and the corresponding representations. IIB significantly outperforms IRM on synthetic datasets, where the pseudo-invariant features and geometric skews occur, showing the effectiveness of proposed formulation in overcoming failure modes of IRM. Furthermore, experiments on DomainBed show that IIB outperforms 13 baselines by 0.9% on average across 7 real datasets.

Automatic Relative Radiometric Normalization of Bi-Temporal Satellite Images Using a Coarse-to-Fine Pseudo-Invariant Features Selection and Fuzzy Integral Fusion Strategies

Relative radiometric normalization (RRN) is important for pre-processing and analyzing multitemporal remote sensing (RS) images. Multitemporal RS images usually include different land use/land cover (LULC) types; therefore, considering an identical linear relationship during RRN modeling may result in potential errors in the RRN results. To resolve this issue, we proposed a new automatic RRN technique that efficiently selects the clustered pseudo-invariant features (PIFs) through a coarse-to-fine strategy and uses them in a fusion-based RRN modeling approach. In the coarse stage, an efficient difference index was first generated from the down-sampled reference and target images by combining the spectral correlation, spectral angle mapper (SAM), and Chebyshev distance. This index was then categorized into three groups of changed, unchanged, and uncertain classes using a fast multiple thresholding technique. In the fine stage, the subject image was first segmented into different clusters by the histogram-based fuzzy c-means (HFCM) algorithm. The optimal PIFs were then selected from unchanged and uncertain regions using each cluster’s bivariate joint distribution analysis. In the RRN modeling step, two normalized subject images were first produced using the robust linear regression (RLR) and cluster-wise-RLR (CRLR) methods based on the clustered PIFs. Finally, the normalized images were fused using the Choquet fuzzy integral fusion strategy for overwhelming the discontinuity between clusters in the final results and keeping the radiometric rectification optimal. Several experiments were implemented on four different bi-temporal satellite images and a simulated dataset to demonstrate the efficiency of the proposed method. The results showed that the proposed method yielded superior RRN results and outperformed other considered well-known RRN algorithms in terms of both accuracy level and execution time.

A Multiscale Normalization Method of a Mixed-Effects Model for Monitoring Forest Fires Using Multi-Sensor Data

This paper points out the shortcomings of existing normalization methods, and proposes a brightness temperature inversion normalization method for multi-source remote sensing monitoring of forest fires. This method can satisfy both radiation normalization and observation angle normalization, and reduce the discrepancies in forest fire monitoring between multi-source sensors. The study was based on Himawari-8 data; the longitude, latitude, solar zenith angle, solar azimuth angle, emissivity, slope, aspect, elevation, and brightness temperature values were collected as modeling parameters. The mixed-effects brightness temperature inversion normalization (MEMN) model based on FY-4A and Himawari-8 satellite sensors is fitted by multiple stepwise regression and mixed-effects modeling methods. The results show that, when the model is tested by Himawari-8 data, the coefficient of determination (R2) reaches 0.8418, and when it is tested by FY-4A data, R2 reaches 0.8045. At the same time, through comparison and analysis, the accuracy of the MEMN method is higher than that of the random forest normalization method (RF) (R2=0.7318), the pseudo-invariant feature method (PIF) (R2=0.7264), and the automatic control scatter regression method (ASCR) (R2=0.6841). The MEMN model can not only reduce the discrepancies in forest fire monitoring owing to different satellite sensors between FY-4A and Himawari-8, but also improve the accuracy and timeliness of forest fire monitoring.

MACA: A Relative Radiometric Correction Method for Multiflight Unmanned Aerial Vehicle Images Based on Concurrent Satellite Imagery

Unmanned aerial vehicle (UAV) offers an unprecedented observing potential with ultrahigh spatial and temporal resolutions and high flexibility. However, it remains difficult to solve the radiometric inconsistency of multiflight UAV imagery. This study proposed a relative radiometric correction method for multiflight UAV images based on concurrent satellite imagery (MACA) consisting of two steps, i.e., cross-sensor spectral fitting (CSF) and fine-resolution spectral calibration (FSC). In CSF, relationships of multiflight UAV reflectance and the concurrent satellite reflectance were established for generating the fine-resolution reference imagery. Subsequently, a relative radiometric correction model was constructed in the FSC step to correct multiflight UAV imagery. The performance of MACA was evaluated using multiflight UAV datasets acquired on six cropland sites and a concurrent Sentinel-2 image. Compared with four typical or state-of-the-art relative correction methods, the correction using MACA yielded better consistency between UAV and Sentinel-2 data, regardless of individual spectral bands (<inline-formula> <tex-math notation="LaTeX">$\text{R}^{2} =0.79$ </tex-math></inline-formula>&#x2013;0.86, root mean square error (RMSE) <inline-formula> <tex-math notation="LaTeX">$=0.004$ </tex-math></inline-formula>&#x2013;0.019) or vegetation indices (VIs) (<inline-formula> <tex-math notation="LaTeX">$\text{R}^{2} =0.80$ </tex-math></inline-formula>&#x2013;0.86, RMSE <inline-formula> <tex-math notation="LaTeX">$=0.024$ </tex-math></inline-formula>&#x2013;0.054). Moreover, the prediction of plant nitrogen accumulation (PNA) based on the MACA-corrected UAV data had the highest accuracy and showed the spatial variation most significantly within and between fields for all sites. The results demonstrated that MACA was more robust in reducing spectral mismatch across sensors and eliminating the subjective error of pseudo-invariant features (PIFs) selection. MACA has the potential to be used to cross-calibrate multisensor data into a consistent standard, which will benefit multisensor synergies.

Integrated Preprocessing of Multitemporal Very-High-Resolution Satellite Images via Conjugate Points-Based Pseudo-Invariant Feature Extraction

Multitemporal very-high-resolution (VHR) satellite images are used as core data in the field of remote sensing because they express the topography and features of the region of interest in detail. However, geometric misalignment and radiometric dissimilarity occur when acquiring multitemporal VHR satellite images owing to external environmental factors, and these errors cause various inaccuracies, thereby hindering the effective use of multitemporal VHR satellite images. Such errors can be minimized by applying preprocessing methods such as image registration and relative radiometric normalization (RRN). However, as the data used in image registration and RRN differ, data consistency and computational efficiency are impaired, particularly when processing large amounts of data, such as a large volume of multitemporal VHR satellite images. To resolve these issues, we proposed an integrated preprocessing method by extracting pseudo-invariant features (PIFs), used for RRN, based on the conjugate points (CPs) extracted for image registration. To this end, the image registration was performed using CPs extracted using the speeded-up robust feature algorithm. Then, PIFs were extracted based on the CPs by removing vegetation areas followed by application of the region growing algorithm. Experiments were conducted on two sites constructed under different acquisition conditions to confirm the robustness of the proposed method. Various analyses based on visual and quantitative evaluation of the experimental results were performed from geometric and radiometric perspectives. The results evidence the successful integration of the image registration and RRN preprocessing steps by achieving a reasonable and stable performance.

A novel automatic method on pseudo-invariant features extraction for enhancing the relative radiometric normalization of high-resolution images

ABSTRACT Relative radiometric normalization (RRN) is a critical preprocessing step that is widely applied to remote sensing data. Essential to RRN are the pseudo-invariant features (PIFs) – ground objects with invariant reflectance across the acquisition dates of multiple images. To meet the requirements of RRN for change detection in multi-temporal images, this paper proposed a novel automatic method for PIF extraction: first, we calculated spectral, spatial, and visual features for the images. Next, a set of change vectors were generated to quantify the features’ differences. Then, the initial PIF was obtained by synthesizing similar pixels from each change vector. Finally, an iterative linear regression was performed to generate the final PIFs. Furthermore, we employed the random sample consensus (RANSAC) algorithm to establish the RRN model for normalizing the target image. The methods and workflow of PIF extraction and the RRN model establishment were validated with two TripleSat-2 multi-spectral images with a spatial resolution of 3.2 metres. The applicability of the proposed PIF extraction method was further validated by Landsat-8 OLI, GF-1 WFV, Sentinel-2 MSI, and WorldView-2 images. The results showed that: (1) The proposed PIF extraction method can automatically obtain the PIFs with high precision. Based on the extracted PIFs, the RANSAC algorithm is capable of building the RRN model with a significant linear relationship. (2) The PIF extraction method is applicable to multi-spectral images with different spatial resolutions. (3) Compared with the multivariate alteration detection (MAD) and the iteratively reweighted MAD (IR-MAD), the proposed PIF extraction method achieved a better performance in selecting the pixels with smaller radiometric differences, which is helpful to reach a higher accuracy of the RRN model. (4) Differences in image brightness, spectral domain, and spatial domain are considered in the PIF extraction method with fewer parameters and high portability. (5) Spectral features are dominant to obtain ideal PIFs, and feature combinations can get the advantages to enhance the effectiveness of the PIFs.

Detection of Coral Reef Bleaching Based on Sentinel-2 Multi-Temporal Imagery: Simulation and Case Study

Sentinel-2 mission has been shown to have promising applications in coral reef remote sensing because of its superior properties. It has a 5-day revisit time, spatial resolution of 10 m, free data, etc. In this study, Sentinel-2 imagery was investigated for bleaching detection through simulations and a case study over the Lizard Island, Australia. The spectral and image simulations based on the semianalytical (SA) model and the sensor spectral response function, respectively, confirmed that coral bleaching cannot be detected only using one image, and the change analysis was proposed for detection because there will be a featured change signal for bleached corals. Band 2 of Sentinel-2 is superior to its other bands for the overall consideration of signal attenuation and spatial resolution. However, the detection capability of Sentinel-2 is still limited by the water depth. With rapid signal attenuation due to the water absorption effect, the applicable water depth for bleaching detection was recommended to be less than 10 m. The change analysis was conducted using two methods: one radiometric normalization with pseudo invariant features (PIFs) and the other with multi-temporal depth invariant indices (DII). The former performed better than the latter in terms of classification. The bleached corals maps obtained using the PIFs and DII approaches had an overall accuracy of 88.9 and 57.1%, respectively. Compared with the change analysis based on two dated images, the use of a third image that recorded the spectral signals of recovered corals or corals overgrown by algae after bleaching significantly improved the detection accuracy. All the preliminary results of this article will aid in the future studies on coral bleaching detection based on remote sensing.

南方红壤典型水土流失区马尾松林地上林木碳储量的遥感监测——以长汀县河田镇为例

森林碳储量动态变化对揭示区域水土流失治理成效具有重要指示意义。以长汀县河田镇为例，2017年随机设置34个马尾松林样本作为建模集，分别与同期Landsat影像的原始波段、植被指数及主成分因子进行回归分析，构建马尾松（（Pinus massoniana））林地上林木碳储量的最佳反演模型，基于伪不变特征原理的线性归一化法实现该模型在2003、2010年影像上的适用性校正转换，实现研究区2003、2010、2017年马尾松林地上林木碳储量的反演及时空分异特征的研究。结果表明：研究区2017年马尾松林地上林木碳储量最佳遥感反演模型是以绿色植被指数（GNDVI）为自变量构建的指数模型：C₂₀₁₇=0.006e^{14.357GNDVI₂₀₁₇}，该模型拟合的决定系数为0.57，平均相对精度为82.19%；2003年、2010年马尾松林地上林木碳储量遥感估测模型为：C₂₀₀₃=0.006e^{（16.4086GNDVI₂₀₀₃+1.1428）}、C₂₀₁₀=0.006e^{（15.1677GNDVI₂₀₁₀+1.5821）}，两期校正模型的决定系数均在0.85以上；2003、2010及2017年碳储量分别为8.24 t/hm²、11.34t/hm²、16.14 t/hm²，整体呈上升趋势；地上林木碳储量随海拔、坡度的升高而增加，向阳坡地上林木碳储量高于背阴坡；碳储量增长率随海拔、坡度的升高而降低，背阴坡碳储量增长率高于向阳坡。;The dynamic changes of forest carbon storage have important indications for revealing the effectiveness of regional soil erosion control. Take Hetian Town, Changting County as an example. In 2017, 34 Masson pine (Pinus massoniana) forests samples were randomly set as the modeling set, and the original waveband, vegetation index and principal component factor regression analysis of the Landsat images of the same period were used to construct the best inversion model for the aboveground forest carbon storage of Masson pine forests. The best inversion model, the linear normalization method based on the principle of pseudo invariant feature (PIF), realizes the applicability correction conversion of the model on the images of 2003 and 2010, and realizes the inversion of the carbon storage of the Masson pine forests in the study area in 2003, 2010 and 2017. Research on spatio-temporal differentiation characteristics. The results show that the best remote sensing inversion model for the aboveground forest carbon storage of Masson pine forests in the study area in 2017 is an index model constructed with the Green Normalized Vegetation Index (GNDVI) as the independent variable:C₂₀₁₇=0.006e^{14.357GNDVI₂₀₁₇},the fitting coefficient of determination of this model is 0.57, and the average relative accuracy is 82.19%; In 2003 and 2010, the remote sensing estimation models of the aboveground forest carbon storage of Masson pine forests are: C₂₀₀₃=0.006e^{(16.4086GNDVI₂₀₀₃+1.1428)},C₂₀₁₀=0.006e^{(15.1677GNDVI₂₀₁₀+1.5821)}, The coefficient of determination of the two-period calibration model is above 0.85; Carbon storage in 2003, 2010 and 2017 were 8.24 t/hm², 11.34 t/hm² and 16.14 t/hm², respectively, showing an overall upward trend; the aboveground forest carbon storage increases with the elevation and slope, and the aboveground forest carbon storage on the sunny slope is higher than that on the shady slope; The growth rate of carbon storage decreases with the increase of altitude and slope, and the growth rate of carbon storage on the shady slope is higher than that on the sunny slope.

Radiometric normalization of Landsat thermal imagery for detection of tundra land cover changes: experience from West Siberia

ABSTRACT The potentiality and specificity of using multitemporal stacks of thermal infrared (TIR) satellite scenes for modelling changes in tundra land cover and predicting geohazards in areas under development are studied using the example of two areas in northern West Siberia. The study aims at developing and testing a method for relative radiometric normalization of Land Remote-Sensing Satellite (Landsat) TIR images applied to estimate changes in land surface temperature (LST) in the tundra with regard to the high-latitude climate conditions. The suggested normalization method (RNUC) differs from other similar approaches in the use of unsupervised classification to determine pseudo-invariant features where master and slave images are spectrally identical. The error in LST contrasts is minimized iteratively by choosing the optimal number of land-cover classes and linear regression coefficients by cross-validation. The normalization procedure ensures error reduction by a factor of two to four, which is evidence of its high performance. The results of processing and analysis of 19 pairs of Landsat thermal images at the sites of the Marre-Sale and Urengoy oil and gas-condensate fields using the RNUC method showed that the relative error of LST contrast normalization could be in the range of approximately 3% to 10%. Based on the time series of normalized values of LST and the normalized difference vegetation index (NDVI) of sites with pyrogenic damage in the area of the Urengoy oil and gas-condensate field, it is shown that the process of stabilization of the tundra cover is characterized by a negative correlation of these parameters and lasts 25 to 28 years. A positive correlation between LST and NDVI characterizes a different type of landscape transformation; in particular, it may be an increase in the thickness of secondary vegetation cover or a change in the hydrological conditions.

A High Spatial Resolution Satellite Remote Sensing Time Series Analysis of Cape Bounty, Melville Island, Nunavut (2004–2018)

Changes in vegetation have been observed in areas of the Arctic due to changing climate. This study examines a normalized difference vegetation index (NDVI) time series (2004–2018) of high spatial resolution satellite data (i.e., IKONOS, WorldView-2, WorldView-3) to determine if vegetation abundance has changed over the Cape Bounty Arctic Watershed Observatory, Melville Island, Nunavut. Image data were corrected to top-of-atmosphere reflectance and normalized for time series analysis using the pseudo-invariant feature (PIF) method. Percent vegetation cover measurements and indices derived from local climate data (growing degree days base 5 °C; GDD5) were used to contextualize NDVI trends in different vegetation types and within active layer detachments (ALDs). NDVI showed similar patterns within the different vegetation types and across the ALDs. There was no significant change in NDVI nor in GDD5 over time. However, there were statistically significant (p < 0.05) relationships between the GDD5 and NDVI for all vegetation types. Using field measurements with high spatial resolution remote sensing data helps link changes in NDVI with changes to vegetation and earth surface processes. The challenges of integrating high spatial resolution satellite data from different sensors in a time series analysis are also discussed.

A Relative Radiation Normalization Method of ISS Nighttime Light Images Based on Pseudo Invariant Features

The International Space Station (ISS) offers a unique view from space that provides nighttime light (NTL) images of many parts of the globe. Compared with other NTL remote sensing data, ISS NTL multispectral images taken by astronauts with commercial digital single-lens reflex (DSLR) cameras have the characteristics of free access, high spatial resolution, abundant data and no light saturation, so it plays a unique advantage in the research of small-scale urban planning, optimization of lighting resource allocation and blue light pollution. In order to improve the radiation consistency of ISS NTL images, a relative radiation normalization method of ISS NTL images is proposed in this paper. Pseudo invariant features (PIF) were identified in the cloud-free Defense Meteorological Satellite Program/Operational Linescan System (DMSP/OLS) time series NTL remote sensing annual composite product, and then they were used to derive the relative radiation normalization model of ISS NTL images. The results show that the radiation brightness of ISS NTL images in different regions is normalized to the same gray level with that of DMSP/OLS NTL remote sensing images in the same year, which improves the radiation brightness comparability between different regions of ISS NTL images. This method is universally applicable to all ISS NTL images, which is beneficial to the NTL comparability of ISS NTL image in the regional horizontal and temporal vertical.