Real-time processing capability is a key enabler for the next generation of intelligent satellites. Full-waveform spaceborne laser altimetry has become an indispensable tool in a wide range of scientific and engineering applications, including terrain mapping, biomass estimation, and Earth system monitoring. Within this context, accurate and efficient laser footprint localization is essential for ensuring the geometric reliability of altimetric measurements. However, conventional waveform-matching methods suffer from severe computational burdens, rendering them unsuitable for high-frequency on-orbit calibration and difficult to deploy in real-time onboard scenarios. To address these challenges, this paper proposes SLA-FLNet - a physics-guided deep learning framework that integrates key physical mechanisms of laser pulse propagation, terrain modulation, and echo formation through a multi-branch spatiotemporal architecture. Each module of SLA-FLNet explicitly encodes a physically interpretable process, enabling accurate, interpretable, and scalable footprint localization. To support supervised training in the absence of ground-truth labels, pseudo-labels were generated using a classical waveform-matching algorithm. The model was evaluated on 2379 laser footprints from 18 beams of the GaoFen-7 (GF-7) satellite, spanning 12 U.S. states with diverse terrain. SLA-FLNet achieved high prediction accuracy and delivered footprint localization results consistent with waveform matching, even in unseen geographic regions. An ablation study further highlighted the critical role of terrain-encoding in enhancing structural fidelity and cross-regional generalization. Compared to traditional methods, SLA-FLNet achieved over 100,000 & times; inference speedup on modern GPUs, demonstrating strong potential for real-time onboard processing. In summary, SLA-FLNet provides a physically consistent, computationally efficient, and deployment-ready solution for full-waveform footprint localization and on-orbit calibration, supporting future autonomous Earth observation missions.

Highlights What are the main findings? A dual-branch spatial-spectral fusion framework (SSF-TransUNet) was proposed for cross-source crop classification using high-resolution and multispectral imagery. The spatial-spectral fusion strategy improves crop discrimination by integrating spatial structures from GF-2 imagery and spectral information from Sentinel-2 observations. What is the implication of the main finding? Cross-source spatial-spectral feature learning provides an effective solution for fine crop classification when spatial and spectral information are acquired by different sensors. The proposed framework offers a practical approach for high-resolution crop mapping in heterogeneous agricultural environments.Highlights What are the main findings? A dual-branch spatial-spectral fusion framework (SSF-TransUNet) was proposed for cross-source crop classification using high-resolution and multispectral imagery. The spatial-spectral fusion strategy improves crop discrimination by integrating spatial structures from GF-2 imagery and spectral information from Sentinel-2 observations. What is the implication of the main finding? Cross-source spatial-spectral feature learning provides an effective solution for fine crop classification when spatial and spectral information are acquired by different sensors. The proposed framework offers a practical approach for high-resolution crop mapping in heterogeneous agricultural environments.Abstract Accurate exploitation of spatial structures and spectral characteristics is essential for fine-grained crop classification using remote sensing imagery. Although multi-source remote sensing data provide complementary information, most existing methods implicitly assume homogeneous data sources with consistent spatial resolution. In practice, high spatial resolution and rich spectral information are usually provided by different sensors, making cross-source spatial-spectral fusion a non-trivial challenge. To address this issue, we propose SSF-TransUnet, a dual-branch spatial-spectral joint modeling framework for fine crop classification. The proposed network explicitly decouples spatial structure extraction and spectral discriminability learning by jointly utilizing high spatial resolution imagery and multi-spectral observations acquired from different satellite sensors within a unified architecture. To support model training and evaluation, we construct SSCR-Agri, a spatial-spectral complementary resolution agricultural dataset integrating meter-level GF-2 imagery and multi-spectral Sentinel-2 data from five representative agricultural regions in northern China, covering five crop categories including corn, rice, wheat, potato, and others. Extensive experiments demonstrate that SSF-TransUnet consistently outperforms representative CNN-based and hybrid CNN-Transformer models. The proposed method achieves an overall accuracy (OA) of 81.84% and a mean Intersection over Union (mIoU) of 0.6954 in fine-grained crop classification, effectively distinguishing crops. These results highlight the effectiveness of spatial-spectral joint modeling for high-resolution crop mapping and demonstrate its potential for precision agriculture and large-scale agricultural monitoring applications, and shows a promising mechanism when combined with multi-temporal observations.

Accurate shallow-water tidal fields are essential for coastal vertical-datum realization and hydrographic applications, but tide gauges provide only pointwise constraints and global tidal models often degrade near complex coastlines. ICESat-2 photon-counting LiDAR offers dense along-track sea-surface observations, yet its sparse and irregular temporal sampling prevents direct conventional harmonic analysis. This study proposes a spatially smoothed regularized least-squares harmonic analysis framework, termed ReLSHA, to correct and reconstruct shallow-water tidal harmonics using ICESat-2 ATL12 observations. Detrended and sea-state-bias (SSB)-corrected ATL12 sea-surface heights (SSHs) are used as relative corrective observations, while tide-gauge (TG) harmonic constants and FES2014b are introduced as strong and weak priors, respectively. A 2-D Laplacian regularization is imposed to stabilize the inversion and maintain spatially continuous amplitude-phase fields. Applied to the northern Gulf of Mexico, ReLSHA improves the reconstruction of dominant tidal constituents, particularly M2, and yields vector root mean square errors (RMSEs) generally within 0.07-0.12 m for the principal constituents. The synthesized tidal time series achieves an RMSE of approximately 0.13 m at independent tide gauges, and the derived relative lowest astronomical tide (LAT) datum reaches an accuracy of about 0.10 m. These results indicate that ICESat-2, combined with TG constraints, tidal-model priors, and spatial regularization, can support shallow-water tidal correction and continuous relative datum reconstruction in gauge-sparse coastal regions.

The Ice, Cloud, and Land Elevation Satellite-2 (ICESat-2) photon-counting lidar provides abundant canopy height estimates for assessing forest structure and carbon stocks. However, the nonlinear response of photon-counting detectors introduces vertical sampling errors, leading to underestimation of derived canopy heights, especially for the needleleaf forests. To address this issue, a canopy height correction algorithm based on the geometricoptical and radiative-transfer (GORT) model is proposed. By extracting the surface roughness and the gap probability from the ICESat-2 photon data, the entire signal waveform including the ground and canopy return is recovered according to the assumed Gaussian ground echo and the GORT model. The first photon height (FPH) is reconstructed from the recovered signal waveform and compared with the measured FPH from the ICESat-2 product. Based on the criterion that the difference between the measured and reconstructed FPH is minimal, the canopy height correction can be derived. The proposed algorithm is implemented using ICESat-2 data and reference airborne lidar data over the Eldorado National Forest, California. The results show that ATL08-derived canopy heights are underestimated with the biases of -2.77 m, whereas our proposed algorithm can substantially reduce such biases to -0.24 m. Specifically, the effects of the pulse strength and canopy cover on the corrected canopy heights are quantitatively investigated. Our proposed algorithm can correct the underestimation of the canopy heights for different pulse strengths and canopy covers, which is physically applicable to improve the precision of the canopy height in needleleaf forests.

Synthetic aperture radar (SAR) imaging is susceptible to various types of jamming, which can severely degrade image quality and hinder downstream tasks. To address this issue, this article proposes a jamming suppression method through a stochastic differential equation (SDE) based diffusion model trained on pseudopaired SAR images. First, candidate jamming regions are identified in suppression jamming SAR images through energy concentration and low-rank characteristics. Then, pseudopaired SAR images representing low and high jamming states are constructed by combining these candidate regions with the original SAR images (referred to as clean images in the following text). Last, a diffusion model, with images evolving from the low jamming state to the high jamming state during the forward process and allowing the reverse process to effectively reconstruct clean images from heavily corrupted inputs, is trained to learn the transition between states. This yields a network capable of progressively suppressing jamming and recovering the clean images. Experiments on simulated SAR images with multiple active suppression jamming types and practical Sentinel-1 datasets demonstrate that the proposed method adapts well to diverse jamming types and intensity levels, exhibiting notable effectiveness, robustness, and practical applicability. The training strategy eliminates the need for prior knowledge of suppression jamming patterns and the availability of real paired SAR images, making it especially suitable for complex real-world scenarios, where jamming characteristics are difficult to characterize.

Synthetic Aperture Radar (SAR) imaging relies on using focusing algorithms to transform raw measurement data into radar images. These algorithms require knowledge of SAR system parameters, such as wavelength, center slant range, fast time sampling rate, pulse repetition interval, waveform, and platform speed. However, in non-cooperative scenarios or when metadata is corrupted, these parameters are unavailable, rendering traditional algorithms ineffective. To address this challenge, this article presents a novel parameter-free method for recovering SAR images from raw data without the requirement of any SAR system parameters. Firstly, we introduce an approximated matched filtering model that leverages the shift-invariance properties of SAR echoes, enabling image formation via convolving the raw data with an unknown reference echo. Secondly, we develop a Principal Component Maximization (PCM) method that exploits the low-dimensional structure of SAR signals to estimate the reference echo. The PCM method employs a three-stage procedure: 1) segment raw data into blocks; 2) normalize the energy of each block; and 3) maximize the principal component's energy across all blocks, enabling robust estimation of the reference echo under non-stationary clutter. Experimental results on various SAR datasets demonstrate that our method can effectively recover SAR images from raw data without any system parameters. To facilitate reproducibility, the Matlab program is available at https://github.com/huizhangyang/pcm.

Deep learning methods have been widely applied to polarimetric synthetic aperture radar land cover classification. However, their performance is often constrained by information loss and the presence of visually similar patterns caused by similar scattering mechanisms. In contrast, directly leveraging rich polarimetric decomposition features introduces challenges due to their high redundancy and the difficulty of identifying the most discriminative feature subsets effectively. To address these issues, we propose a novel Physics-guided dual-branch network. This network comprises a spatial branch that processes Pauli images and a polarimetric branch dedicated to processing 23-dimensional polarimetric features. Within the polarimetric branch, we design a local group-adaptive feature Selection mechanism that dynamically selects discriminative features during end-to-end training. Crucially, we introduce a multifamily diversity constraint that groups polarimetric features according to their underlying physical scattering mechanisms. We enforce group regularization, sparsity regularization, and physical correlation regularization to ensure that the selected features are physically consistent and complementary. Experiments conducted on two agricultural datasets (AIRSAR and RADARSAT-2) demonstrate that the proposed method significantly outperforms state-of-the-art approaches in terms of overall accuracy and Mean Intersection over Union. Furthermore, visualization analysis confirms that the selected features closely align with the scattering mechanisms of different crop types, substantially enhancing both classification performance and physical interpretability.

The ICESat-2 (Ice, Cloud, and land Elevation Satellite-2) carries a revolutionary photon-counting lidar and diverse studies have demonstrated its great observational capabilities on Earth observations. However, it is still difficult to obtain high quality land surface reflectance in that a single measurement (e.g., the laser signal derived apparent surface reflectance) is characterized by two main unknowns, i.e., the SR (surface reflectance) and atmospheric transmission. As the radiative properties of ICESat-2 laser signal and solar noise are simultaneously obtained, we combine signal and noise information to achieve the co-observations, which show how two measurements help to decouple the reflectivity of atmosphere from surfaces and thus resolve this inherent ill-posed problem. Specifically, we propose the theoretical laser signal and solar noise models for spaceborne lidars that link the two measurements (the signal count and background rate) to the two unknowns. Then, a method and workflow applied for ICESat-2 is designed to retrieve the AOTs (aerosol optical thickness) and SRs. The performance is validated against the MODIS (Moderate-resolution Imaging Spectroradiometer) and AERONET (Aerosol Robotic NETwork) product with the average MAPE (mean absolute percentage errors) of less than 30% for AOTs and the average MAPE of less than 15% for SRs in different land cover types. In addition, the ability of this method to identify snow-covered or cloud-covered areas is explored and validated. This study provides a reference for the active and passive co-observations. In the future, satellites carrying both lidar and multispectral cameras could enable higher quality Earth observations, with the lidar enabling more accurate isolation of atmospheric and surface contributions.

Photomultiplier tube (PMT)-based single-photon lidars (SPL) hold significant potential for airborne and spaceborne remote sensing. However, under high-flux conditions, the pulse pile-up or dead-time effect causing conventional leading-edge discrimination methods to suffer from severe intensity-dependent range walk errors. This mechanism fundamentally constrains the ranging accuracy in strong-return scenarios. By extracting and processing both the leading- and trailing-edge timing information of the PMT anode pulse, the underestimation and overestimation intensity-dependent range errors are theoretically neutralized. Monte Carlo simulations demonstrate that this dual-edge approach effectively mitigates range walk errors across diverse signal intensities and pulse widths, maintaining high stability regardless of photon flux variations. Furthermore, a custom high-speed dual-edge discrimination module was developed and integrated with a commercial PMT. Experimental results indicate that for the average signal photon number ranging from 0.39 to 19.7 counts, the dual-edge method maintains ranging biases within ±1 cm, effectively decoupling the range walk error from signal intensity. This method provides a robust and hardware-efficient solution for pulse-pile-up compensation, significantly enhancing the ranging accuracy of SPL systems with minimal architectural complexity.

Satellite altimetry has been widely used for coastal sea level monitoring, which is significantly affected by the sea state bias (SSB). SSB represents systematic height errors caused by the interaction between the altimeter footprint and the dynamically rough sea surface. To address this issue, the ICESat-2 Ocean Surface Height (L3A) product, ATL12, includes an SSB correction based on photon return rates observed across different sea surface locations. However, as the influence of the complex geometry of the sea surface is not fully considered yet, there is still up to +/- 0.1 m systematic bias remaining in ICESat-2 data compared with the tide gauge stations along the North American coast, which may affect further application in shallow waters that are always associated with complex terrain and need more accurate measurements. To address this problem, a generalized additive model (GAM) is trained to characterize the nonlinear relationships between remaining SSB errors in ICESat-2 data and the relevant sea state variables. According to the verification results, the trained GAM improves the elevation accuracy of the sea surface survey: the residual errors between ATL12 sea surface height products and tide gauge observations are further reduced from -0.060 m to -0.001 m in mean error and from 0.196 m to 0.083 m in RMSE. Sensitivity analyses further indicate that higher-order statistical features (e.g. skewness, kurtosis) significantly enhance the model's generalization capability. At least 4 years of training data are required to ensure stable performance. Overall, the proposed method effectively mitigates the residual SSB errors in ATL12, enhancing its suitability for coastal and shallow-water applications.

Active-passive satellite-derived bathymetry (SDB), effectively combining the high vertical accuracy of the spaceborne lidar with the wide spatial coverage of multispectral imagery, has emerged as an important solution for shallow-water bathymetry on a global scale. However, the unreliable depth estimates beyond a certain depth severely restrict their utility for subsequent scientific and operational applications. When the remotely sensed signal originating from the seafloor attenuates below a critical signal-to-noise threshold, the sensor physically loses its ability to correctly resolve the depth. Accurately identifying this maximum reliable depth and deriving a boundary for the inversion results is essential. This study proposed an SDB method with its self-consistent reliable depth boundary. Recognizing the tradeoff between local reliability versus global applicability, a framework is developed through two complementary strategies. First, in the reliability-oriented image-specific strategy, Ice, Cloud, and Land Elevation Satellite-2 (ICESat-2) seafloor photons are exploited in a dual-purpose utilization. The radiometric intensity is used to retrieve the diffuse attenuation coefficient K-d (denotes the water quality), while its geometric information was served as the seed points in the training SDB model. Combining the uncertainty (sigma(total)) of the remote sensing reflectance derived from deep-water statistics of Sentinel-2 imagery, we construct a physical model to calculate the maximum reliable depth (H-max) without requiring external ancillary data. This model is validated across six diverse shallow-water regions, and the results demonstrate that the calculated depth boundary effectively identifies and masks unreliable deep-water pixels, whose root-mean-square errors (RMSEs) increase exceeding three times the RMSEs from the reliable pixels within the depth boundary. Second, extending to a scalability-oriented global strategy, we utilize global ocean color products and theoretical instrument noise models (instead of local environmental and instrument parameters) to estimate the theoretical bathymetric capability of Sentinel-2 on a global scale. This methodology provides a robust, self-consistent framework for generating highly reliable nearshore bathymetric products and assessing the global detection limits of satellite sensors.

Since heavy clutter seriously restricts the ability of radar to detect targets, it is significant to build the target detector under heavy clutter. For practical situations without the secondary data or prior knowledge of target and clutter, this paper proposes a polarization-space-time detector. First, a general radar model is constructed for multiple pulses, multiple arrays, and multiple polarizations. Based on the theory of ternary hypothesis, the secondary data free (SDF) GLRT detector is proposed, which can maintain the constant false alarm probability (CFAR) in inhomogeneous clutter. Then, this paper proposes a matrix transform operator and an adaptive detection method using sliding window. These two approaches do not need to know the steering vector of radar and the noncentral parameter of clutter in advance, so the SDF-GLRT detector can adapt to different application scenarios. In addition, this paper optimizes the polarization waveform of the radar system by constructing a projection matrix. This method yields closed-form solutions of the optimal polarization and worst polarization, rather than relying on numerical solution. Finally, the performances of the SDF-GLRT detector and three other detectors are compared by simulated and real data. The proposed SDF-GLRT maintains PFA of 5.4×10−3 and 2.6×10−3 on two IPIX datasets (#54 and #310) at a design PFA=10−3, whereas the other detectors deviate to 0.0249–0.7405. The optimal polarization yields a detection-probability gain of more than 0.22 over the worst polarization at SCR=0 dB.

Polarization offers rich information for enhancing target recognition in synthetic aperture radar (SAR) imagery. However, most existing SAR target recognition methods rely on single-channel data, and the potential of multichannel polarimetric images remains underexplored. In this article, we propose an end-to-end target recognition framework for polarimetric SAR (PolSAR) images based on a quaternion convolutional neural network (QCNN) operating in the Poincar & eacute; sphere parameter domain. The QCNN is constructed with a sequence of quaternion operation layers and incorporates a specialized loss function designed for quaternion-valued representations. To address the mismatch problem in the quaternion field, we introduce quaternion maximum pooling and quaternion average pooling operations. To the best of our knowledge, this is the first QCNN developed for PolSAR target recognition. Experiments on both simulated and real datasets demonstrate that the proposed QCNN achieves recognition performance comparable to state-of-the-art real- and complex-valued models while requiring significantly fewer parameters and offering enhanced physical interpretability, thereby validating the effectiveness and superiority of the proposed approach.

Clouds, covering two-thirds of the Earth, are the primary obstacle to the development of all-weather spaceborne photon-counting lidars (SPLs) by severely attenuating signals and increasing noise levels, thereby complicating the signal detection and extraction. Consequently, most data applications of SPLs, such as those from the Ice, Cloud, and land Elevation Satellite-2 (ICESat-2), typically avoid cloudy scenarios, limiting the full potential of SPLs. To quantitatively model and estimate the cloud-penetrating capability of SPLs, we integrate the radiative transfer model, the detection model, and the signal extraction performance into a standardized and rigorous framework and transform the issue into signal extraction capability under different signal-to-noise ratios (SNRs). First, we derive a radiative transfer model for SPLs in complex atmospheric environments using Monte Carlo simulations to calculate cloud optical parameters and considering three surface types: ocean, snow-free land, and snow-covered land. The accuracy of the model is verified by 12 ICESat-2 tracks worldwide, with an average mean absolute percentage error (MAPE) of 27.56% for signal photon counts and 19.64% for background noise rates. Second, we develop an assessment model that integrates hardware-based detection and algorithm-based extraction for SPL performance, deriving theoretical relationships between signal and noise levels, photon-counting detector characteristics, algorithm parameters, and the F-score. The cloud-penetration capability of ICESat-2 is estimated across three typical surfaces, with discrepancies between theoretical and measured values below 15% and a mean error of 9.0%. A quantitative analysis evaluates the impact of system and environmental parameters on cloud-penetration capability, complemented by seasonal analysis of ICESat-2 globally. The strong beam of ICESat-2 can penetrate clouds with optical depths of approximately 0.7-0.9 over oceans, 0.3-0.8 over snow-free land, and 1.35-1.55 over snow-covered land. The proposed models provide critical scientific insights and practical value for SPL development in breaking through climate limitations, enhancing data quality control, and optimizing system design.

This article proposes a novel high-frequency scattering model for targets on complex rough surfaces. The core innovation is a local ray incidence-based technique that computes the reflection and multireflection components from targets and the coupling scattering components between targets and the background. This approach generates ray tubes exclusively from the target facet model. The background facet model is subsequently processed to perform ray-triangle intersection tests, while coupling scattering components are modified using the four-path principle. Background scattering is calculated via coherent-incoherent models, and diffraction components are computed using the equivalent electromagnetic current (EEC) method. The accuracy of the proposed method is validated against full-wave numerical solvers, showing good agreement in radar cross section (RCS) fluctuations. Compared with the traditional shooting and bouncing rays (SBRs) technique with EEC, the proposed method achieves a significant reduction in simulation time. This capability enables the rapid simulation of SAR images for targets on complex rough surfaces.

Received waveforms of spaceborne Light Detection and Ranging (lidar) over mountainous vegetated areas comprise multiple overlapped components. It encounters a challenge to extract the ground elevation from the complex waveform due to the difficulty in the determination of the ground return component location and the ground return extent. This study proposes an algorithm to resolve such issue, which includes: 1) based on abundant simulated lidar waveforms, we establish a regression model of the ground return extent with regard to the signal-to-noise ratio (SNR) of the received waveform and the surface slope. 2) we decompose the waveform components by the pattern matching and nonlinear least-square fitting methods, and then, 3) we extract the centroid of the decomposed waveform components within the ground return extent as the ground Elevation. Using the Global Ecosystem Dynamics Investigation (GEDI) lidar waveforms in Henry W. Coe State Park of USA, the ground elevations are resolved and compared to those by the existing methods and the GEDI products. The results demonstrate that our proposed method outperforms other methods, especially for the lower-SNR waveforms (<16 dB) with the over steep terrains (>20°). The bias, the root mean square error and the mean absolute error of the derived ground elevations by our proposed algorithm are -0.08 m, 3.33 m and 2.31 m (laser shot number is 22291), which have the reductions of 96.7 %, 30.9%, and 34.6% than the GEDI product, respectively. It indicates that our proposed algorithm can effectively improve the accuracy of the ground elevation over rugged mountainous vegetated areas.

Objective Benefiting from the extremely high sensitivity of single-photon detectors, it has become possible to detect the weak profile received signals from the water body in satellite platforms. The new generation of spaceborne single-photon lidar promises to provide an independent observational data source for the diffuse attenuation coefficient (K-d) of the water body, while addressing the limitation of passive ocean color satellites in achieving continuous day-night observations. However, the extreme sensitivity of single-photon detectors also introduces significant solar background noise during the daytime, which can overwhelm the signal photons of water bodies and render the retrieval of K-d nearly impossible. Current research on the retrieval of water optical parameters by spaceborne single-photon lidar primarily focuses on nighttime scenarios, with only a few studies qualitatively analyzing the feasibility of such retrievals during daytime. We hope that the conclusions drawn from our research can help evaluate the capability to retrieve the water optical parameters of existing single-photon lidar systems and provide critical reference for the parameter design of future Chinese spaceborne single-photon lidar missions. Methods Challenges in acquiring ground-truth data and comprehensive noise testing lead us to simulate photon clouds via Monte Carlo, enabling quantitative noise influence analysis by adopting a standardized K-d retrieval framework. First, based on the theory of JONSWAP (Joint North Sea Wave Project) ocean wave spectrum and Cox-Munk reflection model, the signal photons on the sea surface under different wind speeds are simulated. Then, based on the bio-optical model of typical waters, different inherent optical properties (IOPs) of the water body are simulated by taking chlorophyll-a mass concentration as a variable parameter. The received signal intensity distribution at different water depths is simulated by adopting a Monte Carlo method, and the photon clouds are finally generated based on the received signal intensity. It should be noted that the photons in the dead time are deleted according to the response model of the PMT single-photon detector. Finally, the pseudo-waveforms are generated statistically from the photon clouds, and the Kd values are estimated under different noise levels, yielding a maximum tolerable background for single-photon lidar. Results and Discussions The K-d retrieval of typical scenes shows that under the same water environment parameters, the effective signal of water bodies is gradually submerged with the increasing background noise, making it impossible to retrieve K-d under high background noise. The exponential decline trend of received signal intensity of water is gradually weakened, which results in the gradual underestimation of Kd until it loses its physical representation of the true attenuation properties. By taking the Kd value of pure water as a limitation (0.046 m-1@532 nm), the maximum resistance of the background noise rate for ICESat-2 system is analyzed. As the chlorophyll-a mass concentration in the water body increases, the backscatter signal intensifies, thereby enhancing the corresponding noise resistance. Combined with the background noise model of spaceborne single-photon lidar over the ocean, the background noise rate under different solar elevation angles in typical waters can be estimated. Thus, the maximum resistance of the background noise rates along with their corresponding solar elevation angles is shown for typical chlorophyll-a mass concentrations. For the spaceborne single-photon lidar system ICESat-2, its maximum noise resistance is below 100 kHz in oligotrophic waters, and below 500 kHz in chlorophyll-rich regions, which indicates that Kd can be effectively retrieved only at night or under small solar elevation angle scenarios. Additionally, the maximum resistance of the background noise rate is gradually enhanced with the rising received energy. Taking the ICESat-2 system as an example, if the laser emission energy is increased by a factor of five, the maximum noise immunity can exceed 1 MHz. This means it can adapt to most clear-sky scenarios, enabling the satellite to essentially measure water optical parameters during daylight hours. Conclusions The diffuse attenuation coefficient K-d is a fundamental parameter in ocean color remote sensing applications and holds significance in the field of marine optics. Spaceborne single-photon lidar provides an independent observational capability for the diffuse attenuation coefficient K-d. However, its performance in varying background noise conditions has not been thoroughly explored. Based on the Monte Carlo method, we simulate water bodies' photon clouds under different environmental parameters and quantitatively analyze the influence of background noise rates on the retrieved photon-counting values by adopting standard water diffuse attenuation coefficient processing algorithms. Our study finds that as background noise increases, the phenomenon of decreasing photon point cloud density with water depth becomes progressively obscured, and the retrieved photon-counting values gradually decrease until they fall below the pure water attenuation coefficient, rendering them practically meaningless. In oligotrophic waters, the maximum tolerable noise level for ICESat-2 is less than 100 kHz, which implies that obtaining valid results during daylight hours is nearly impossible. In chlorophyll-rich regions, the maximum tolerable noise is also less than 500 kHz, requiring a lower solar elevation angle for the effective retrieval of the diffuse attenuation coefficient. Our study provides a quantitative assessment basis for evaluating the environmental adaptability and engineering application requirements of water optical parameter retrieval. Finally, we analyze the influence of modifying system parameters on noise resistance capability. Taking the ICESat-2 system as an example, if the emission energy is increased fivefold, the maximum noise resistance capability exceeds 1 MHz, making it adaptable to clear-sky daytime scenarios. This can serve as a reference for the parameter design of future spaceborne ocean single-photon lidar systems and provide a quantitative evaluation of water optical parameter retrieval capabilities in different working conditions.

Constructing functional connectivity networks from electroencephalogram (EEG) channels and using graph neural networks for emotion recognition have emerged as a significant technical route in EEG emotion recognition. However, most existing approaches are limited to estimating brain graph networks based on EEG full-channel signals, failing to adequately explore the representations between and within brain regions. To address this limitation and further investigate the interactions between channels and regions, an explainable cross-level topological network (ECTN) is proposed for EEG emotion recognition, which is designed to capture EEG functional interactions from channel-level to region-level. Within the ECTN framework, three modules are designed, namely cross-region topological feature fusion module, specific-region position-guided attention module, and bidirectional gated fusion module. Specifically, EEG functional interactions are explicitly decoupled into two complementary views: global region interactions and local region dynamics. Additionally, the bidirectional gated fusion module leverages the inclusion relationships between channels and brain regions to further integrate region-level and channel-level features. The ECTN model is evaluated on the publicly available SEED series of datasets, SEED, SEED-IV and SEED-V. Experimental results indicate that our method achieves superior performance, effectively validating the benefits of exploring channel-wise and region-wise interactions.

Synthetic aperture radar (SAR) object detection plays a crucial role in remote sensing applications. However, conventional methods often require high computational and memory costs, limiting their deployment in resource constrained environments. The challenges of SAR imagery such as sparse object distribution, speckle noise, and multiscale variations make it difficult for existing lightweight detectors to achieve both high accuracy and efficiency. To address this issue, we propose LightKD-SAR, a lightweight SAR object detection framework that combines an efficient network architecture with enhanced instance selection based knowledge distillation. Specifically, we design a lightweight detection network using customized inverted residual modules, and further reduce computational complexity through optimized feature extraction and fusion strategies while maintaining robust detection performance. In addition, we introduce an improved instance selection mechanism combined with multidimensional knowledge transfer, focusing on samples with large prediction discrepancies to enhance learning of ambiguous objects and complex backgrounds in SAR images. Extensive experiments on the large-scale SARDet-100 k dataset demonstrate that LightKD-SAR achieves a mAP of 50.92% with only 15.7 GFLOPs and 11.43 M parameters. Compared with state-of-the-art methods, the proposed framework demonstrates superior tradeoff between detection accuracy and computational efficiency, making it well-suited for practical deployment in real-world SAR-based remote sensing systems.

The spaceborne photon-counting lidar (SPL) is capable of acquiring Earth's surface elevation with decimeter-level accuracy, while the swath capability is limited under the current configuration. Anticipatedly, future SPLs are increasingly advancing toward a multibeam configuration with many tens to a few hundreds of laser beams, which will greatly enhance the Earth's observation capabilities. However, there is a significant gap in the data denoising algorithms for the dense along-track (meters-level) and sparse cross-track (tens of meters-level) point clouds. The challenge of leveraging cross-track information to improve the efficiency and capability of denoising algorithms, especially under low signal-to-noise ratio (SNR) conditions over steep mountainous terrains, remains to be addressed. In this study, we propose a binary-coding-based signal extraction algorithm that incorporates the cross-track topographic information, which rapidly partitions the space through quadtree isolation (QI) while applying Morton coding to photons to encode their spatial distribution information. The utilization of the binary-coding method not only enhances the performance of signal extraction but also offers significant advantages for hardware implementation, making it suitable for onboard field-programmable gate array (FPGA) based processing on satellite platforms. Its performance is verified using multiple repeated track data from ICESat-2's identical reference ground track (RGT) in mountainous regions of Alaska, Idaho, and Utah in the USA, which demonstrate stable F -scores with an average of 0.95 across varying along-track slopes and SNR conditions. Compared to currently existing methods, the proposed algorithm shows superior robustness, especially in low SNR regions. Furthermore, we analyze the relationship between F -scores and cross-track distance under some extreme conditions. In summary, this study lays a methodological foundation for future onboard data processing of multibeam SPLs and offers significant insights for the design of future multibeam SPL systems.