中国遥感卫星专辑
- The Paper
- Abstract：On the basis of multisource ocean dynamic satellite observations of sea surface wind fields in China, the National Satellite Ocean Application Service has developed a high-quality Level-4 global sea surface wind field fusion product (Multisatellite Ocean Wind Vector, MULOWV). As a novel wind field dataset, MULOWV integrates the strengths of multiple data sources, offering potential utilization in wind speed and wind direction measurements. However, its reliability and performance in practical applications remain to be validated. To assess the application potential of MULOWV, this study conducts detailed comparisons and evaluations of METOP and moored buoy observations. Furthermore, triple collocation analysis is performed to calibrate and assess the biases, scaling factors, and random errors among the three datasets, revealing their consistency and providing a reliable basis for multisource data fusion and optimization.To ensure spatial and temporal consistency between satellite observations and in situ measurements, this study defines appropriate spatiotemporal collocation windows on the basis of satellite spatial resolution and the variability of sea surface winds. Through triple collocation analysis, unbiased error estimates are derived for each dataset, enabling the assessment of errors and systematic biases without requiring a reference truth. Standard deviation, mean bias error, Root-Mean-Square Error (RMSE), and correlation coefficient are selected as evaluation metrics. These indicators are applied to wind speed and wind direction and analyzed from an overall perspective and across different wind speed regimes.Results show that in comparison with BUOY and METOP data, MULOWV data exhibit small biases, with wind speed RMSE below 1.6 m/s and wind direction RMSE below 15°. MULOWV demonstrates high consistency and stability in collocations and strong adaptability for large-scale and multisource validation needs. In the triple collocation analysis, buoy data serve as the baseline reference, providing critical guidance for calibrating MULOWV and METOP; however, buoy data exhibit large random errors in low-wind-speed conditions, suggesting the need for joint calibration with other sources. Across the wind speed regimes, data quality varies. MULOWV and METOP show strong overall correlations, but slight deviations occur under extreme-wind conditions. Under moderate-to-high wind speeds, MULOWV achieves good random error characteristics and bias corrections, demonstrating high consistency and accuracy, making it suitable for use as a primary data source for high-precision wind field monitoring and model validation.In conclusion, the MULOWV fusion product provides high accuracy and strong multisource consistency, supporting its use as a major data source for wind field monitoring and model evaluation. MULOWV is applicable for large-scale wind field estimation, but for high-precision applications, multisource integration is still needed to enhance reliability. Buoy data remain suitable for use as calibration references in localized regions. By applying triple collocation analysis to wind speed and direction, this study provides valuable insights for wind field monitoring and model validation under diverse conditions. Future efforts should focus on improving stability under extreme-wind scenarios to enhance MULOWV’s applicability in complex atmospheric environments.Keywords：Ocean Satellite;Blended Sea Surface Wind Field;Triple Collocation Analysis;MULOWV;microwave scatterometer;Wind Field Validation567|1042|0
<HTML><L-PDF><Enhanced-PDF><Meta-XML>Updated：2026-02-09 - Abstract：Chlorophyll-a (Chl-a) concentration is an essential climate variable and fundamental to global carbon cycle studies and ocean environmental monitoring. HY-1C/1D satellites, equipped with the Chinese Coastal Ocean Color and Temperature Scanner (COCTS), enable global ocean color monitoring at kilometer-scale resolution. High-accuracy remote sensing algorithms for Chl-a concentration on the basis of COCTS data need to be developed to fully leverage these Chinese autonomous satellites for ocean monitoring and climate research.This study developed a Chl-a retrieval algorithm on the basis of a multilayer perceptron neural network (MLP-NN) for the COCTS sensor. The model inputs included remote sensing reflectance (Rrs) at COCTS center bands and environmental variables, such as geolocation, Sea Surface Temperature (SST), and Photosynthetically Active Radiation (PAR). The model was trained using 2,165 in-situ measurements collected from the global ocean. After a comparative analysis of mainstream machine learning models, MLP was selected as the core architecture for the NN framework. A multidimensional feature fusion strategy was implemented to construct the MLP-NN model. Given that multidimensional inputs could introduce redundancy, sensitivity analysis was conducted to quantify the contribution of each input, identify the optimal input set, and improve the model’s efficiency and generalization.The sensitivity analysis identified the following optimal combination for MLP-NN: Rrs at 412, 443, 490, 520, 565, and 670 nm; latitude; month; average SST from the previous month; and climatological PAR from the previous month. Validation indicated that Chl-a estimated by MLP-NN achieved a Root Mean Square Difference (RMSD) of 0.22 and a Median Absolute Percentage Difference (MAPD) of 29.1% for log-transformed Chl-a, which are 0.1 and 16.9% lower than those estimated by the NASA operational Ocean Color Index (OCI) algorithm, respectively. Further validation using satellite and in-situ matchups confirmed that MLP-NN outperformed OCI, reducing RMSD and MAPD by 0.09 and 9.8%, respectively, highlighting its improved robustness. In China’s Bohai Sea, both algorithms effectively captured the spatial distribution patterns of Chl-a. However, OCI exhibited systematic bias, underestimating Chl-a concentrations at high and low extremes. By contrast, the MLP-NN model demonstrated high accuracy in retrieving extreme Chl-a values.Overall, the MLP-NN model developed in this study substantially improves the estimation of Chl-a concentrations from HY-1C/1D satellite observations. It offers valuable algorithmic support for leveraging domestic satellites in ocean ecological monitoring.Keywords：Chlorophyll-a (Chl-a);remote sensing reflectance;retrieval algorithm;HY-1C/1D satellites;neural network;COCTS;ocean color849|1546|0<HTML><L-PDF><Enhanced-PDF><Meta-XML>Updated：2026-02-09
- Abstract：Forest Soil Organic Carbon (SOC) is a critical indicator of forest soil quality, significantly affecting the growth of forest trees and playing an essential role in the sustainable development of forestry. Investigating the potential of utilizing hyperspectral images to accurately determine the SOC in natural secondary forests is imperative. This investigation would aid in providing technical assistance for estimating forest SOC on a long-term and large-scale basis. With a focus on the SOC of natural secondary forests, this study randomly selected a total of 67 samples in the Maoershan Experimental Forest Farm of Northeast Forestry University. Soil samples were collected from three different depths: 0—5 cm, 5—15 cm, and 15—30 cm. The SOC content of each sample was measured, and the mean of the three layers was calculated as the SOC content for 0—30 cm depth. The hyperspectral image of ZY-1F was analyzed to calculate the first-order differential, second-order differential, reciprocal logarithm of the spectral curve, and vegetation indices. The Recursive Feature Elimination (RFE) method was then employed to screen the features, taking into account the Digital Elevation Model (DEM), soil moisture, and forest Aboveground Biomass (AGB) datasets. Three machine learning models, namely, random forest, Extreme Gradient Boosting (XGBoost), and support vector regression, and ordinary least squares regression were employed to estimate the SOC content, and the best model was chosen to estimate the SOC at various depths. Results showed that XGBoost had the highest accuracy in various soil depths: the R2 for the soil depths of 0—30 cm, 0—5 cm, 5—15 cm, and 15—30 cm were 0.54, 0.54, 0.46, and 0.30, respectively, and the RMSEs were 21.28, 44.25, 15.72, and 12.56 g/kg, respectively. The average SOC values in the natural secondary forest of Maoershan Forest Farm were estimated to be 67.20, 88.87, 46.92, and 40.12 g/kg for the 0—30 cm, 0—5 cm, 5—15 cm, and 15—30 cm soil layers, respectively. The SOC concentration in the forest declined as the soil depth increased. Variations in SOC content exist across various forest types, and the SOC is ordered in descending order as follows: broad-leaved forest, mixed coniferous and broad-leaved forest, and coniferous forest. The band information from hyperspectral images enables the estimation of the SOC in forests. However, the large number of bands leads to data redundancy, which in turn reduces the accuracy of the model’s estimates. The RFE method can be employed to identify the optimal combination of features, thereby reducing the amount of features and enhancing the accuracy of model estimate. The differential characteristics of the 710—850 nm bands in hyperspectral images are extremely beneficial for accurately estimating the SOC in natural secondary forests. Topographic factors exert a more significant influence on the SOC at depths above 15 cm, while soil moisture and AGB have a more pronounced effect on SOC in the 5—15 cm layer compared with other factors. The integration of hyperspectral images with DEM, soil moisture, and AGB data facilitates the accurate estimation of the SOC content of natural secondary forests. This approach offers valuable support for estimating long-term and large-scale SOC of natural secondary forests on the basis of multiperiod hyperspectral images.Keywords：natural secondary forest;SOC;hyperspectral;machine learning model;feature selection1376|3067|0<HTML><L-PDF><Enhanced-PDF><Meta-XML>Updated：2026-08-03
- Abstract：Land Surface Temperature (LST) is a critical parameter for understanding surface energy balance and water cycle processes. This study aimed to develop LST retrieval algorithms using thermal infrared data from the Visual and Infrared Multispectral Imager (VIMI) onboard Chinese Gaofen-5B (GF-5B) satellite, providing a new data source for global high-spatial-resolution LST product generation.A simulated dataset was first constructed on basis of the conventional thermal infrared radiative transfer model, TIGR-3 atmospheric profiles, and 74 typical surface emissivity samples selected from the ASTER spectral library. Two split-window algorithms (called as SW-1 and SW-2) for GF-5B thermal infrared data were developed based on the simulated dataset and sensor characteristics, with both algorithms achieving an RMSE of 1.13 K. Considering the influence of Column Water Vapor (CWV) on LST retrieval, algorithm coefficients were derived under different CWV conditions. Comparison of RMSE under identical conditions demonstrated that SW-2 generally outperformed SW-1 in accuracy. The algorithm’s performance varies with CWV, with RMSE values of SW-1 (SW-2) decreasing from 1.41 K (1.34 K) in high CWV conditions (4.5—6.3 g/cm²) to 0.40 K (0.39 K) in low CWV conditions (0.0—2.5 g/cm²). The SW-2 algorithm was thus recommended as the preferred choice for GF5B-based LST retrieval applications. Key parameters acquisition methods were also proposed, including land surface emissivity (via a modified NDVI-NDWI threshold method) and atmospheric CWV (via a water vapor split-window covariance-variance ratio method).The proposed algorithm was applied to four experimental regions with different land cover types. Combined with false-color composite imagery (green, red, and near-infrared bands) derived from VIMI’s visible and near-infrared data, the spatial distribution of LST retrieval results was evaluated in relation to surface characteristics. Artificial surfaces and bare soil exhibited relatively higher LST values, while water bodies and vegetation showed lower LST values, consistent with thermal radiative characteristics. Validation was conducted using in-situ LST measurements from HiWATER sites and MODIS LST products. Results demonstrated that the LST errors for daytime and nighttime retrievals were 1.88 K and 0.99 K, respectively. Among all validated data, the maximum daytime temperature difference reached 2.86 K, while the nighttime maximum difference was 1.27 K. Additionally, variations in validation accuracy were observed across imagery acquired at different periods for the same site. For cross-comparison with MODIS MOD11A1 LST products, observational discrepancies were minimized by aligning acquisition times and viewing angles, and spatial aggregation was applied to reduce the impact of resolution mismatches. High-resolution Sentinel-2 land cover products and temperature gradients between adjacent pixels were utilized as screening criteria. Except for the Tangshan (Hebei) region, which exhibited significant deviations due to cloud contamination, temperature differences in the remaining three regions remained within 1.5 K.The LST retrieval results across the four experimental regions exhibited reasonable spatial distributions and strong correlations with land cover types. Validation with limited in-situ HiWATER data confirmed higher accuracy for nighttime retrievals compared to daytime. Cross-comparison with MOD11A1 LST products demonstrated acceptable agreement (differences ≤1.5 K) in three regions, indicating that the proposed algorithm achieves satisfactory accuracy.Keywords：GF-5B;thermal infrared data;Land Surface Temperature (LST);split-windows algorithm;land surface emissivity estimation;atmospheric water vapor content (CWV) estimation;accuracy validation;in-situ LST measurements validation;cross-comparison of LST967|1545|0<HTML><L-PDF><Enhanced-PDF><Meta-XML>Updated：2026-08-03
- Abstract：The accuracy of Digital Surface Models (DSMs) reconstructed from satellite stereo imagery is generally lower in rugged gully-dominated landscapes than in flat areas, especially in the absence of Ground Control Points (GCPs). Moreover, collecting GCPs in large and rugged areas is often operationally challenging or costly. To improve the accuracy of DSMs derived from satellite stereo imagery, this study proposes a method that integrates laser altimetry data with DSMs generated from Gaofen-7 (GF-7) satellite stereo imagery using a backpropagation (BP) neural network. The proposed method seeks to improve terrain DSM accuracy in areas where GCPs are either scarce or costly to obtain, providing an efficient solution for terrain modeling in such regions.Considering the high-precision elevation accuracy of laser altimetry data, the proposed method improves the DSM accuracy without GCPs through establishing relationships between multiple factors and elevation data from the Global Ecosystem Dynamics Investigation mission. These factors include the DSM generated from GF-7 satellite stereo imagery without GCPs, geographic coordinates (longitude and latitude), terrain slope, and terrain errors. This fusion is achieved through the use of a BP neural network, which is trained to model and enhance the accuracy of the DSM under uncontrolled conditions. The study area is located in the Loess Plateau region spanning Shaanxi Province and Inner Mongolia Autonomous Region, China. The terrain has the characteristics of severely eroded loess gullies, intensive mining activities, and dramatic undulations. The proposed method is tested across regions covered with different numbers of images, and its performance is validated through comparisons with ground-truth data measured by RTK.Experimental results show that the elevation error of the GF-7 stereo-derived DSM in gully-developed mining areas without GCPs can reach up to 20.49 m. By contrast, the average elevation accuracy of the fused DSM is significantly improved to 1.63 m after applying the fusion technique using the BP neural network, which is comparable to the accuracy of the DSM generated with GCPs (1.44 m). When quality-filtered laser altimetry points are utilized as substitute GCPs directly, the elevation accuracy of DSM can also be refined to 2.40 m. However, the proposed comprehensive multifactor fusion method yields better performance (1.63 m vs. 2.40 m), confirming the advantage of fusion approaches. The improvement in elevation accuracy demonstrates that, whether used directly as GCP substitutes or through data fusion approaches, laser altimetry data can effectively enhance the vertical accuracy of DSMs derived from GF-7 satellite stereo imagery without GCPs.The findings of this study show that the proposed BP neural network-based fusion method significantly enhances the accuracy of DSMs generated from GF-7 satellite stereo imagery in areas without GCPs. This approach effectively solves the problem of low elevation accuracy in regions with complex topography, such as gully-developed mining areas. This study not only provides an innovative solution for terrain modeling in areas lacking GCPs but also offers a new way to utilize domestic high-resolution satellite imagery for high-precision terrain reconstruction. This method contributes to the advancement of terrain modeling techniques, especially in regions where obtaining GCPs is difficult or impossible, and provides a viable solution for reducing labor costs, improving operational efficiency, and reconstructing high-quality terrain DSMs.Keywords：digital surface model;ground control points;neural network;gully-developed areas;GF-7 satellite stereo imagery;GEDI(Global Ecosystem Dynamics Investigation)774|2619|0<HTML><L-PDF><Enhanced-PDF><Meta-XML>Updated：2026-08-03
- Abstract：The advanced geostationary orbit radiometer (AGRI) of FY-4A satellite has been on orbit for 6 years, and the radiation performance of some reflective channels have significantly degraded, affecting the accuracy of quantitative remote sensing product applications. On-orbit vicarious calibration methods based on Deep Convective Cloud (DCC) targets can track and correct the radiometric response of spaceborne optical sensors for attenuation. This method relies on large-sample statistical analysis, and conducting sensitivity studies on factors influencing the calibration accuracy and stability in this method and developing optimal solutions holds significant importance.The procedure of the fundamental DCC calibration and tracking method is outlined as follows. Initially, DCC target pixels were extracted from FY-4A/AGRI L1 level data, the reflectance of the target pixels was calculated, and anisotropic correction was executed using the DCC Angle Distribution Model (ADM). Subsequently, daily or monthly Probability Density Functions (PDFs) of DCC reflectance were constructed, and the trend in peak reflectance (also known as mode) or reflectance mean was tracked to monitor and evaluate the radiometric performance of the FY-4A/AGRI instrument. To improve the calibration accuracy and stability, the sensitivity research scheme for infrared brightness temperature threshold, pixel uniformity conditions, and DCC ADM was proposed. Lastly, the DCC model was corrected, and an optimal solution was established according to the sensitivity analysis results.Results indicate that for the infrared brightness temperature threshold, the sensitivity of DCC mean reflectance is lower than that of PDF peak reflectance in the visible light channel, and in the short-wave infrared channel, the sensitivity of DCC PDF peak reflectance is slightly lower than that of reflectance mean. In the visible-near-infrared band, the CERES ADM can better correct the effect of DCC reflectance anisotropy and is significantly better than the Hu model. However, neither of the two ADMs has obvious correction effect in the short-wave infrared band. Based on the above sensitivity studies, the threshold selection and ADM correction strategy in the DCC method are determined. The radiation response of FY-4A/AGRI reflected bands from March 2017 to April 2023 is tracked and evaluated. Results show that the radiation response of 0.47, 0.65, and 2.25 μm channels degrades significantly, with total attenuation rates of 45.55%, 26.22%, and 6.362%, respectively. This result provides a reference for updating the AGRI operation calibration coefficient.A sensitivity analysis on the key factors in the radiometric calibration tracking method was conducted based on DCC for satellite optical sensors, enhancing calibration accuracy and stability through the establishment of an optimal solution. By utilizing optimization methods, the variations in the radiometric response performance in the reflectance band of the FY-4A/AGRI were quantitatively evaluated, providing valuable reference for updating the operational calibration coefficients of this instrument.Keywords：remote sensing and sensors;radiometric calibration;Deep convective cloud;advanced geostationary radiation imager;angular distribution model;top of atmosphere reflectance;reflective solar bands1739|2205|0<HTML><L-PDF><Enhanced-PDF><Meta-XML>Updated：2025-11-03
- Abstract：The clear-sky image synthesis in a single day is of great significance for daily water body recognition and other business applications. This paper proposes a clear-sky image synthesis algorithm based on a binary Gaussian mixture model for a 1-minute continuous imaging sequence data of the Geostationary High-speed Imager (GHI) of the FY-4B satellite. The GHI is the world’s first quantitative remote sensing instrument capable of high-frequency imaging of geostationary orbit during day and night. It is designed to meet short-term forecasting needs, with a total of seven visible and infrared channels, providing continuous observation of multiple spectral bands at 1-minute intervals in a 2000 × 2000 km area.Generally, except for ice and snow, the reflectivity of clouds is higher than that of the underlying surface. For the same location, the change in reflectivity is relatively small, whereas the change in reflectivity of passing clouds is significant. That is, the distribution of surface reflectance shows low mean and small variance (with relatively concentrated samples), whereas the distribution of cloud reflectance passing through shows high mean and large variance (with scattered samples). Therefore, the algorithm in this article assumes that the reflectance sequence samples of a single pixel within a single day are composed of clear-sky reflectance samples and cloud reflectance samples, which satisfy Gaussian distributions. The problem of synthesizing clear-sky images is transformed into estimating the distribution parameters of clear-sky reflectance samples. The algorithm is divided into three main steps: initial guess value (Step I), pixel classification (Step C), and parameter update (Step U). In the initial parameter estimation, a simple threshold method is used to initialize the clear-sky binary Gaussian distribution parameters. For the sequential processing of time-series images, Gaussian distribution function is used to identify the clear-sky type to which new image pixels belong. The average, standard deviation, and other parameters of the two types of clear sky at the location are updated based on the new identification results. Finally, when the sequential processing of all intraday image data are completed, the average reflectance of the clear-sky type is used as the estimated reflectance of the clear-sky composite result for that location.The method has linear time and memory space complexity, and the effective clear-sky pixel ratio and image information entropy of the clear-sky composite image gradually increase. Compared with typical clear sky algorithms, it has higher robustness in distinguishing clear sky and filtering ability for cloud edge shadows.High-frequency single-day clear-sky composite images can be applied in ecological remote sensing applications such as vegetation, water environment, and water monitoring.Keywords：Clear Sky Synthesis Image;FY-4B;GHI;Gaussian model;Water Body Identification;multi-temporal remote sensing data471|2744|0<HTML><L-PDF><Enhanced-PDF><Meta-XML>Updated：2025-11-03
- Abstract：In the current research on sea surface height inversion from satellite-borne GNSS reflected signals, classical algorithms are usually used to invert sea surface height. However, due to the existence of multiple complex errors, such as inaccurate receiver orbit, system error, ionosphere error, and troposphere error, the results inverted using classical algorithms are mostly of low accuracy. Therefore, an error model is needed to correct the inversion results. Classic error models generally improve the accuracy of sea surface height inversion by correcting common errors, such as tropospheric error, ionosphere error, and antenna baseline attitude error, but there remain large errors that cannot be corrected. To address this problem, this paper proposes an error compensation model based on the combined training of neural networks and Attention Mechanisms (AMs) to correct the sea surface height inversion results.This paper proposes a training method that combines a Convolutional Neural Network (CNN) model with an AM to accurately train the error of sea surface height inversion from satellite-borne GNSS reflection signals. An error compensation model is generated to replace the classical error model, thereby improving the accuracy of sea surface height inversion.The proposed model was compared with the classic error model, CNN model, and random forest model and tested on about 2 million delay Doppler mapping data of the FY-3E dataset. The evaluation indicators used Mean Absolute Error (MAE) and Root Mean Square Error (RMSE). For the Global Positioning System reflected signal data corrected using the error compensation model, the MAE was 1.74 m, and the RMSE was 2.25 m. For the Beidou Navigation Satellite System reflected signal data, the MAE was 0.97 m, and the RMSE was 2.16 m. Compared with the classic error model, the correction accuracy was improved by about 80%. Compared with the random forest model and CNN model, the accuracy was also slightly improved.This paper proposes an error compensation model based on the training of CNN and AM to correct the sea surface height inversion results. Experiments show that the proposed error compensation model effectively corrects the sea surface height inversion error of space-borne GNSS-R.Keywords：GNSS-R;neural network;satellite-based;FY-3E;sea surface height inversion;Error;DDM;Beidou943|3689|0<HTML><L-PDF><Enhanced-PDF><Meta-XML>Updated：2025-11-03
- Abstract：The Sea Surface Temperature (SST) is an important indicator for studying ocean dynamics, ocean-atmosphere interaction, and climate change, which is closely related to multiple marine environmental factors, such as ocean currents, salinity, and nutrient distribution, collectively affecting the balance and evolution of marine ecosystems. Although the traditional SST acquisition methods are precise, they are limited by the number and coverage of sampling points, making it difficult to meet the requirements of large-scale and high-resolution ocean research. Satellite remote sensing data can cover global waters with high update frequency and is widely used in ocean research. However, during the collection process of satellite remote sensing data, SST data are often missing due to factors such as weather conditions, satellite scanning orbit range, and satellite sensor operation failures, which limits the use of data to some extent. Therefore, precise reconstruction of missing data in satellite remote sensing SST data to obtain high-quality and fully covered SST datasets is of great significance for ocean research. This study incorporates an Inception module into a deep interpolation convolutional autoencoder (DINCAE) and proposes the improved DINCAE (I-DINCAE) model used for data reconstruction of SST products with the FY-3C satellite in the South China Sea. The I-DINCAE is used to reconstruct the missing SST data in the South China Sea from 2014 to 2020, and the reconstruction accuracy of the DINCAE and I-DINCAE models is compared and analyzed. To further improve the accuracy of SST data, deep neural networks (DNNs) are used to calibrate satellite data in combination with the measured data, thereby optimizing the quality of the SST dataset. Finally, based on the corrected SST data, spatiotemporal variation analysis is conducted to reveal the characteristics of SST changes. At the same time, combined with many years of measured data, DNN model is used to calibrate the reconstructed temperature data of the new model. A dataset of 11,993 independent measured data points is used for testing. Results show that the RMSE, MAE, and R² of the reconstructed SST and measured SST are 1.27 ℃, 0.96 ℃, and 0.84, which decreased to 0.57 ℃, 0.43 ℃, and 0.92 after the DNN model correction, respectively. Based on the corrected SST data, the spatiotemporal distribution and variation characteristics of SST in the South China Sea at monthly and quarterly scales are analyzed from two dimensions of time and space. Results show that on the seasonal scale, the SST of the South China Sea has obvious variation characteristics. This shows that the SST reaches the highest value in the summer, and the SST decreases to the lowest value in the winter. On the monthly scale, the SST variation in the South China Sea presents a sine (cosine) wave form, with SST usually reaching a maximum value in June and a minimum value in January. This study not only reveals the uniqueness of the marine environment in the South China Sea but also provides an important basis for understanding the marine ecosystem and climate change in the South China Sea.Keywords：sea surface temperature;data reconstruction;deep learning;FY-3C;spatio-temporal variation806|3826|0<HTML><L-PDF><Enhanced-PDF><Meta-XML>Updated：2025-11-03
- Abstract：Snow Depth (SD) and Snow Water Equivalent (SWE) are crucial parameters for describing snow cover information. High-precision SD and SWE data are critical in investigating weather forecast, hydrology, surface processes, and other applications. Passive microwave remote sensing is an effective means of observing SD and SWE. Since April 2019, the National Satellite Meteorological Center has released passive microwave global SD and SWE products of microwave radiation imager aboard the Fengyun-3D Satellite (FY-3D). Compared with the FY-3B SD retrieval algorithm, the operational algorithm of the FY-3D introduces fractional forest cover for performing empirical correction on forest influence in Northeast China. This study investigates the performance of the improved FY-3D SD and SWE operational algorithms and verify the accuracy of the corresponding products in the forest area in Northeast China.This article obtains the situation of SD in the study area over the years through observation data from meteorological stations in Yichun, Heilongjiang Province. The FY-3D SD and SWE operational products are validated through measured snow course and SD data observed by meteorological stations in the forest areas. Moreover, the uncertainty of FY-3D SD products and the representativeness of meteorological stations are analyzed.Results indicate a strong temporal heterogeneity in SD distribution in the Yichun Region. The verification results indicate that the FY-3D SD product exhibits an overall underestimation, and the RMSE is 5 cm and 13.2 cm when compared with the measurements of snow course and the observations of meteorological station, respectively. Conversely, the RMSE between the FY-3D SWE product and the snow course data is 2.1 mm. The FY-3D SD operational algorithm, as a semi-empirical algorithm, cannot eliminate the influence of forests on microwave radiation brightness temperature. Although forest radiometric correction can enhance the correlation between brightness temperature gradient and SD, the empirical nature of forest radiometric correction also increases the uncertainty of SD inversion results.Analysis shows that the FY-3D algorithm has a lag in response to sudden snow drops due to its lack of response to new snow with an exponential correlation length of 0.11 mm. At the beginning of the snow season, when the SD remains below 5 cm, the change in brightness temperature gradient caused by soil freezing can be misjudged by the inversion algorithm, leading to overestimation of SD during this period. In the preliminary exploration of site representativeness, the analysis of the differences between point and surface combined with field observations show that snow in forest areas is deeply influenced by various factors, leading to strong local spatial heterogeneity. This work can provide reference for improving the SD inversion algorithm in forest regions based on domestic FY-3D brightness temperature data in the future.Keywords：FY-3D/MWRI;snow depth;snow water equivalent;product validation;forest region717|1948|0<HTML><L-PDF><Enhanced-PDF><Meta-XML>Updated：2025-11-03
