摘要:The Amundsen impact crater, with a diameter of about 103 km, is one of the most significant geomorphological features in the south polar region of the Moon. The crater floor hosts extensive plain terrains and permanent shadowed regions, with the enrichment of hydrogen elements, making it one of the most prioritized candidate landing areas for future volatile prospecting missions in the lunar south polar region. However, the detailed geological characteristics and evolution history of this region are poorly constrained. To serve future exploration needs of the lunar polar regions, we systematically investigated the geological context and characteristics of the Amundsen region using multi-source high-resolution remote sensing data including topography, spectroscopy, and solar illumination conditions. The topographical signature of the Amundsen crater is well preserved and is one of the representative complex impact craters on the Moon. Over 2000 km2 of plain terrains occur on the crater floor, with an average slope less than 5°. On the basis of composition and albedo analyses, we found that these floor plains are characterized with high albedo and low iron content, obviously different from mare plains, indicating that these plains probably originated from ejecta materials from distant impact craters/basins, especially the Schrodinger basin, rather than volcanic eruption. In addition, the Amundsen crater area is also characterized with elevated abundance of hydrogen, showing its great potential for water ice prospecting. We also cratered a new, large-scale geological map of the Amundsen area, and sorts out the regional multi-stage geological evolutionary history. Through comparison with the Von Kármán crater mare plain where China’s Chang'e-4 mission landed, we suggest that the Amundsen crater floor plain has the topographical conditions for soft landing efforts, while its special characteristics of solar illumination condition, surface temperature, and the provenance of the plains materials bring both scientific opportunities and technological challenges for landing exploration missions.
Wu Hucheng,Wang Renfang,Qiu Hong,Wang Feng,GAO Guang,Wu Dun
Corrected Proof
DOI:10.11834/jrs.20243419
摘要:Semantic change detection of remote sensing images plays an important role in ecological environment, land use, land cover monitoring and so on. In recent years, deep learning-based change detection methods are the hotspot of remote sensing intelligent interpretation concern, however, the existing three-branch semantic change detection methods lack of modeling the consistency of change branch and semantic branches, which leads to the self-contradiction of the bi-temporal semantic change detection.Method To address this problem, this paper proposes a remote sensing image semantic change detection algorithm based on Siamese CNN and Transformer. In the encoding stage, the Siamese ResNet34 network is firstly designed to extract the multi-scale features of the image, and the Difference Enhancement Module is embedded to increase the attention of the change information; then the semantic tokenizer is used to map the feature map into compact semantic tokens, and the Transformer encoder is used to combine the bi-temporal semantics and the change information to model the "semantics-change" consistency. In the decoding stage, different fine-grained semantic information is fused by Transformer decoder using hopping connection to generate a refined semantic feature map. Finally, the result of bi-temporal semantic change is obtained after up-sampling recovery and mask multiplication.Result The experimental results on the remote sensing semantic change detection public dataset SECOND and LandSat-SCD show that the algorithm proposed in this paper can effectively focus on the change region, maintain the consistency between the change results and the semantic results, and achieve excellent evaluation indexes and visual effects.Conclusion We can draw the following conclusions: (1) The proposed Difference Enhancement Module can enhance the difference characteristics of bi-temporal remote sensing images and improve the network's focus on change information. (2) The proposed bi-temporal Transformer module maps the difference information and bi-temporal semantic information into semantic tokens and fuses them to jointly model the "semantic-change" information of the whole spatio-temporal domain in the token space, effectively modeling the long-range dependencies in the images and modeling the bi-temporal contextual correlations. The long-range dependency in the image is effectively modeled, and the bi-temporal contextual relevance is modeled. The ReTNet network designed accordingly pays more attention to the change area, and can accurately detect the change location and recognize the change element type of the bi-temporal remote sensing image.
LU Lingxiao,QIN Kai,COHEN Jason Blake,LI Xiaolu,ZHOU Chunyan
Corrected Proof
DOI:10.11834/jrs.20244018
摘要:Anthropogenic emissions, primarily resulting from the combustion of fossil fuel, have led to a rapid and accelerating rise in atmospheric carbon dioxide (CO2) concentration in recent decades. Utilizing satellite remote sensing technology is crucial in estimating CO2 emissions from fossil fuel consumption, which is essential for achieving the "dual carbon" targets. However, the long lifetime of CO2 makes it challenging to directly estimate CO2 emissions from satellite measurements, and the existing CO2 satellite sensors have insufficient spatial resolution. Considering that fossil fuel combustion emits both CO2 and nitrogen oxides (NOx), the lifespan of NOx is much shorter and the feasibility of estimating its emissions through satellite remote sensing is more favorable. Therefore, this study aims to indirectly estimate daily CO2 emissions based on TROPOMI NO2 observations.In this work, we focus on 28 eastern cities and one western energy-intensive region, known as the "Energy Golden Triangle", to indirectly estimate daily CO2 emissions from fossil fuel consumption based on TROPOMI nitrogen dioxide (NO2) column concentrations. There are three steps in our methodology: First, we utilize TROPOMI observations of NO2 tropospheric column concentrations, and inverts the daily NOx emissions in 2019 based on a MCMFE (Mass-Conserving Model Free approximation of Emissions) approach. This approach is proposed by the authors' team based on the principle of atmospheric component mass conservation. Secondly, the MEIC inventory (Multi-resolution Emission Inventory for China) is employed to compute and analyze the emission relationships between CO2 and NOx. Finally, this work estimates the daily CO2 emissions from fossil fuel consumption in these regions by using the NOx emission (MCMFE-NOx) and CO2-to-NOx ratio computed.This work analyzes the estimation results of 28 eastern cities and three type of sources (power plants, iron and steel factories and coal mines) in the Energy Golden Triangle separately. The findings indicate that the estimations align with the spatial distribution of CO2 emissions in the MEIC inventory, yet offer higher spatial resolution and temporal frequency, revealing emerging and smaller emission sources missed in the inventories. In 28 eastern cities, significant emerging emission sources have surfaced in suburban areas due to recent urbanization expansion and economic development, exhibiting substantial emission volumes. For instance, in eastern cities like Beijing, remote sensing estimations in suburban areas exceeded the MEIC inventory by approximately 104%, indicating the emergence of numerous new emission sources accompanying the rapid expansion of these urban centers. In the Energy Golden Triangle, it was discovered that there are small-scale power plants and industrial sources overlooked in the MEIC inventory in places like Baotou, Yulin, Yinchuan, and Wuzhong. Exemplified by Yulin, estimations from grids containing power plants, steel factories, and coal mines surpassed the MEIC inventory by around 41%, indicating that some smaller power plants and industrial sources not included in the emission inventory were captured through satellite remote sensing.This paper combines the NOx emission and uncertainty results with the "bottom-up" CO2-to-NOx emission ratio, and derives the daily CO2 emission estimation results of the study areas in 2019 (whole year). It performs a statistical analysis of emissions in 28 large cities, and sources in the Energy Golden Triangle. The findings indicate that the estimations align with the spatial distribution of CO2 emissions in the MEIC inventory, yet offer higher spatial resolution and temporal frequency, revealing emerging and smaller emission sources missed in the inventories. This study provides technical support for carbon emission accounting related to fossil fuel consumption in China.Objective:Method:Result:Conclusion
摘要:ObjectiveSnow seasonal evolution is one of the key factors influencing hydrological dynamics in mountainous areas and controlling terrestrial ecology. Accurate information on snowmelt is essential for meteorological, hydrological and global climate change studies as well as for disaster prediction and early warning. The traditional snowmelt detection approach based on time series SAR suffers from the influence of vegetation cover, rugged terrain and long revisit time in some regions. In this study, we propose a new snowmelt detection method based on high resolution Sentinel-2 optical remote sensing data.MethodThe time series snow surface grain size variation is used to detect snowmelt events. As snow starts to melt, the liquid water in snow tend to significantly increase the optical equivalent grain size retrieved from optical remote sensing. When the wet snow refreezes, the optical equivalent grain size remains significantly larger than dry snow. This provides the theoretical basis for snowmelt onset detection from optical remote sensing. The snow surface optical equivalent grain size is retrieved by applying snow reflectance models, with bidirectional reflectance, sun zenith angle, sensor viewing angle and relative azimuth ange as inputs. Pure snow pixels are selected for snow optical equivalent grain size retrieval and snowmelt detection. In this paper, snowmelt detection results of Altay Mountain is presented and analyzed. Snowmelt onset detection based on optical remote sensing is also compared with SAR method, the advantages and shortcomings of the two methods are analyzed.ResultConclusionThe snowmelt onset date retrieved using Sentinel-2 data is similar to those retrieved from the SAR method using Sentinel-1 data, and they show similar dependencies on elevation and aspect. The new method based on Sentinel-2 data also shows advantages compared with SAR method, such as the optical method is less affected by speckle noise, mixed pixels, and vegetation cover, and it provides more spatial details about the snowmelt onset data. The proposed snowmelt detection method based on optical data suffers from cloud cover, but it offers an alternative way to detect wet snow with high spatial resolution other than SAR. The snowmelt detection based on SAR and Sentinel-2 data can be complementary to each other, and the snowmelt detection in mountainous area can be improved by combining both methods.The main advantages of the proposed new method in this study are as follows The new method based on optical remote sensing is more sensitive in detecting the occurrence of snowmelt and provides significantly richer information about the snowmelt process. The choice of snow reflectance model can introduce differences in the retrieved snow grain size due to variations in modeling snow particle shape and light scattering and absorption. Additionally, the selection of threshold values for distinguishing between wet and dry snow grain sizes can also affect the results. Therefore, the use of different snow reflectance models can lead to certain differences in the results of snowmelt detection. The snowmelt onset dates retrieved using the optical method show overall similarity to those retrieved from the SAR method, exhibiting similar dependencies on elevation and aspect. However, there are also some differences between the two methods, which can be attributed to variations in detection principles and data sources. Compared to the SAR method, the optical method is less affected by speckle noise, mixed pixels, and vegetation cover. Particularly in low-elevation and vegetated areas, the proposed method demonstrates superior capability in detecting snowmelt events compared to the SAR method.
摘要:Carbon Dioxide (CO2) is an important greenhouse gas. Satellite remote sensing of atmospheric CO2 has the advantages of long-term and wide spatial range observation, which is of great significance for verifying emission reduction strategies to cope with global warming. Aerosol scattering in the atmosphere is considered to be a major obstacle for remote sensing retrieval of CO2 with high accuracy. Previous studies have shown that over areas with high surface albedo, such as desert regions, satellite retrievals of atmospheric column-average dry-air mole fraction of CO2 (XCO2) are systematically overestimated, and the bias can reach 50% of the allowable error to meet the practical application requirements. However, sufficient understandings and quantitative analysis of the systematic bias are still lacking. Focusing on this difficult problem, this thesis analyzes and quantifies the bias of XCO2 retrievals caused by the scattering effect of dust aerosol over desert regions using an accurate atmospheric radiative transfer model and a retrieval algorithm based on optimal estimation. This study starts from three important representative variables of aerosols, including Aerosol Optical Depth (AOD), Aerosol Layer Height (ALH) and Single Scattering Albedo (SSA), to illustrate the physical mechanism of dust aerosol scattering effects on XCO2 remote sensing retrievals. From the perspective of spectra radiance generated from forward radiative transfer model, increasing AOD will lead to decrease in the spectrum continuum level (defined as radiance of channels where gas absorption can be neglected) in the case of high surface albedo through its extinction effect. Increasing in ALH will cause smaller relative absorption depth (defined as the ratio of radiance difference between continuum level and absorption channels to continuum level) which is closely related to the XCO2 retrievals. From the perspective of retrieval model, this thesis carries out separate retrieval experiments using the O2 A Band and the WCO2 Band, respectively, and joint retrieval experiment using both bands. The results show that the underestimation of AOD or ALH of dust aerosols or the overestimation of SSA in satellite retrieval algorithms can be possible causes of the overestimation of XCO2 over deserts. Specifically, we show that: (1) in the case of not considering aerosol in the retrieval algorithm, XCO2 retrievals will be overestimated by more than 1% when the actual AOD is larger than 1.0; (2) when AOD is underestimated by a value between 0.3 and 0.5, XCO2 retrievals will be overestimated by 0.15% - 1.28%; (3) when ALH is underestimated by more than 0.6 km, XCO2 retrievals will be overestimated by more than 1%; (4) when SSA is overestimated, XCO2 retrievals will also be overestimated, but by no more than 0.15%. It can be shown from these simulation experiments that accurate aerosol information is of significant importance to achieving accurate atmospheric XCO2 retrievals. Additionally, this thesis also discusses the impact of potential “critical albedo” on retrievals and demonstrated that its effect is probably the cause of the bias in extracting useful aerosol information from CO2 monitoring satellites. This thesis proposes that to address this difficult problem, observations from aerosol observing instruments should be included in actual retrievals to further constrain the aerosol information to improve the accuracy of XCO2 retrievals.
关键词:dust aerosol;Remote Sensing Retrieval Algorithm;Scattering Effect;radiative transfer model
摘要:The thermal physical properties of lunar regolith are crucial for unraveling the Moon’s thermal history, geological evolution process, and viability of in-situ resource utilization, with the thermal anomaly being a key area of interest. Microwave observations can reveal subsurface characteristics about lunar regolith from centimeters to meters in depth, significantly enhancing our understanding of these properties. Microwave Radiometer (MRM) onboard Chang’E-2 (CE-2) conducted passive microwave remote sensing of the Moon at frequencies of 3.0, 7.8, 19.35 and 37 GHz from 2010 to 2011. The employment of multi-channel and multi-temporal MRM data provides a new perspective on thermal anomalies.Considering that brightness temperature (TB) is strongly affected by latitude, the difference between the TB values of the same frequency generated at two different local times is introduced, named dTB, which indicates a good description of the thermophysical characteristics of the mare deposits in the wavelength-related penetration depth. Utilizing dTB obtained at 37 GHz, along with numerical simulations of brightness temperature, integrating with Clementine UV-VIS、LRO Diviner and DEM data, the significantly low dTB at 37GHz exists in the eastern and southern highlands of the Moscoviense basin. Through statistical analysis of the anomalous region and the western highlands, the corresponding FeO, TiO2, rock abundance and DEM data do not show the same trend of change. This indicates that the known compositional and topographical data cannot explain the genesis of the low dTB anomaly.The existence of low dTB anomalies was confirmed in Mare Smythii. Taking the highlands at the same latitude and the Chang'e-5 landing region with normal brightness temperature differences as references, EM unit in Mare Smythii exhibits lower dTB at 19.35 and 37GHz. Furthermore, comparative analysis with the compositional data from Chang’e-5 landing region reveals that currently known compositions cannot account for the low dTB anomalies. It is speculated that the anomaly may likely be due to a component that has not yet been detected or identified.Additionally, the discovery of low dTB anomalies in the Apollo basin and Balmer region provided valuable clues for understanding these phenomena. Based on the impact cratering process model, the low dTB anomaly in the Apollo basin is likely due to the rebound of deep lunar crustal materials during the impact process. The anomaly in the Balmer region shows a clear correspondence with the location of the impact crater ejecta coverage, suggesting that the material causing the low dTB anomaly likely originated from the excavation of deep lunar crustal materials by impact events, indicating vertical heterogeneity in the lunar crust's composition.The discovery of low dTB anomaly provides a new insight for further research on thermal radiative properties of lunar regolith, the localized and previously undetected material offers innovative scientific views for investigating the Moon’s impact evolution history and the properties of shallow crust.
关键词:Low brightness temperature difference anomaly;Microwave radiometer;brightness temperature;Regolith composition;Geological significance
摘要:Outgoing longwave radiation (OLR) of the atmosphere top (TOA) is an important component of the radiation energy balance. The Fengyun-3D (FY-3D) and Fengyun-3E (FY-3E) polar-orbiting meteorological satellites, launched in November 2017 and July 2021 respectively, carry the Medium Resolution Spectral Imager (MERSI) II and Low Light (LL) instruments. Both instruments have the capability to retrieve outgoing longwave radiation (OLR) using two water vapor channels and two window channels. In this paper, based on the introduction of the MERSI OLR inversion algorithm of Fengyun satellite, the instantaneous OLR retrieval accuracy of FY-3D and FY-3E MERSI is compared by using the instantaneous observation data of Aqua CERES OLR. The comparison results show that the instantaneous OLR retrieval accuracy of FY-3D and FY-3E is basically the same as the instantaneous OLR data of Aqua CERES. The RMSE of FY-3D and FY-3E MERSI OLR is between 6-7 W•m-2 compared with that of Aqua CERES OLR. It reflects that although there are differences between the performance of the MERSI instruments of the FY-3D and FY-3E satellites, the OLR retrieval capabilities of the two satellites are comparable. The comparison results of the daily average OLR data based on CERES between the single and joint calculations of FY-3D and FY-3E show that the daily average OLR calculated based on the four times of the two satellites per day is 3-4 W•m-2 higher than that calculated twice a day for a single satellite. The global daily average OLR data for the 7 months from June to December 2022 were selected as an example. Compared with the daily average OLR of CERES, the daily average OLR obtained by FY-3 two satellites is significantly improved compared with the root mean square error of the daily average OLR calculated by a single satellite. It shows that the daily observation data of multiple polar-orbiting meteorological satellites can better reflect the diurnal variation characteristics of OLR. The comparison process involves the spatio-temporal matching of data, reprojection and resampling schemes will affect the validation results. Compared with CERES, the difference between the retrieval OLR of the low-temperature target is higher than that of the high-temperature target regardless of the daily average OLR calculated based on the single or double satellites of Fengyun-3. The reasons for this are multifaceted and need to be further analyzed. The research results show that the joint application of multiple polar-orbiting satellites with a certain instantaneous observation interval can effectively improve the calculation accuracy of the daily average OLR in the cloud area. But how to construct the daily variation model of OLR and deepen the calculation method of the daily average OLR needs to be further studied.
Sun jiaxin,Feng Li,Zhang Xiao,Zhou Yanan,Feng Hairong
Corrected Proof
DOI:10.11834/jrs.20243487
摘要:Objective Most of previous studies on the cooling efficiency of urban vegetation are based on the large-scale of satellite remote sensing data, leaving a gap in fine-scale investigations at the micro-level. Therefore, this paper took the vegetation of typical residential areas in Jiangning District of Nanjing City as the research object, UAV visible light data was used to obtain the fine classification of residential green space, and the Regional Green Plot Ratio (RGPR) index was constructed.Method And then We proposed an improved calculating method of vegetation cooling efficiency—Regional Cooling Efficiency (RCE) based on hour-by-hour UAV thermal infrared data. We applied this method to calculate the cooling efficiency of vegetation in residential areas, investigating the response relationship between the surface temperature of different local climate zones (LCZs) residential areas and the RGPR indicator and further determining the optimal thresholds of cooling efficiency for different RGPRs in residential areas.Result The results showed that: (1) the daily variation curve of the cooling efficiency of vegetation in the residential area shows a "peak" pattern, and the RCE increases with the increase of solar radiation. During the observation period, the minimum value of RCE appeared at 8:00 a.m. (all lower than 1.0℃), and the maximum value appeared at 14:00 p.m. (all higher than 1.4℃). At the level of local climate zone, the RCE of compact residential areas was higher than that of open residential areas, and in the case of open residential areas, the RCE decreased with the increase of the average height of the buildings in the area. (2) Under compact local climate zone, the higher the RGPR, the stronger the cooling effect brought by vegetation, while in open residential areas, regardless of the height of the building, there is a certain RGPR threshold for the cooling efficiency of vegetation, and when the RGPR reaches a certain threshold, the cooling efficiency of the RGPR on surface temperatures reaches its maximum intensity.Conclusion Overall, this study proposes a method for calculating the cooling efficiency of vegetation in residential areas at the micro-scale for high-resolution unmanned aerial vehicle (UAV) data, and investigates the cooling efficiency of vegetation and the optimal cooling efficiency thresholds for different types of residential areas at the level of localized climate zones. This study can help to improve the urban thermal environment and provide more specific and scientific theoretical guidance and support for enhancing urban sustainability.
关键词:Urban residential areas;Microthermal environment;Regional Cooling Efficiency (RCE);Local Climate Zone;Regional Green Plot Ratio(RGPR);UAV thermal infrared remote sensing;Cooling efficiency threshold;Green space planning in urban residential areas
摘要:Reducing the reliance on in-situ crop type samples is critical for remotely sensed crop type classification over large areas. This study used Suihua, a major grain-producing city in Heilongjiang Province, as an example to investigate the effect of sample size on crop type classification and to test the possibility of extrapolating supervised classification models trained on a small region to a larger area. Specifically, this study trained the crop type classification model in Beijing District and then extrapolated it to the whole of Suihua. First, a parameter-optimized random forest model was trained and used to identify the spatial distribution of crops in Beilin District in 2022 using Sentinel-2 remote sensing imagery from sowing to mid-tasseling of maize. We found that the overall accuracy (OA) gradually increased as the proportion of samples participating in the random forest training increased from 10% to 50% of the GVG samples in Beilin District. The model achieved the best performance with a maximum OA of 94.6% when 50% of the GVG samples in Beilin District were used for crop classification, where maize, rice and soybean had approximately 130 training samples. Thereafter, the performance of the model remained stable even as the number of in-situ crop samples increased. We also found that the most important features in the classification of maize, soybean and rice were REP at the tassel stage of maize, SWIR1 at the pod stage of soybean and LSWI at the transplanting stage of rice. Secondly, we extrapolated the best trained model in Beilin District to classify crop types in the whole of Suihua. We found that the model extrapolation achieved an OA of 93.7% for crop type classification in Suihua, which was only 1.3% lower than the model trained directly in Suihua. The similarity of the spatial and probability distribution maps of the crops between the Beilin model and the Suihua model indicated that the extrapolation of the crop classification model in a small area can achieve a comparable classification result to the crop classification model trained directly in a large area. Finally, we carefully examined the effects of distance, spatial representativeness and number of samples, and similarity of crop structure between small area and target expansion area on model extrapolation. We found that different crops have different sensitivities to distance, and the classification effect of rice was insensitive to changes in distance due to significant differences between the LSWI and SWIR1 of rice and other crops, while the classification effects of maize and soybean show an overall decreasing trend of change with increasing extrapolation distance. In short, when building crop classification models in small regions with similar crop structures in both source and target areas, not only the number of samples should be considered, but also the representativeness of their spatial distribution. This would ensure that the model is adequately trained and can achieve a better spatial extrapolation effect. The results of the study provide a cost-effective and efficient method to accurately classify crops over large areas using remote sensing. In addition, the study provides a scientific basis for developing crop sampling strategies, selecting sensitive bands and determining the classification time window. It also provides a valuable reference for the development of model extrapolation methods with greater robustness and generalizability.
Teng Chenkai,Xiao Yueyao,Zhang Jialong,He Yunrun,Chen Chaoqing
Corrected Proof
DOI:10.11834/jrs.20244006
摘要:Time series remote sensing data has good applications in accurately estimating forest carbon storage, providing data support for a deeper understanding of the carbon cycle process of forest ecosystems, scientific management, and protection of forest resources, but there is a lot of noise in remote sensing time-series data, in order to enhance the accuracy of estimating carbon storage, a filtering algorithm was developed to reduce the interference of noise in Landsat time-series data from high-altitude areas. Based on the continuous inventory fixed plots data of National Forest Inventory in Shangri-La area for the years 1987, 1992, 1997, 2002, 2007, 2012, and 2017, as well as Landsat time series images from 1987 to 2017. In this study, the Adaptive Topography Convolution(ATC)algorithm was developed using Python, which takes into account the impact of terrain factors on image quality, removes the noise from the image while retaining as much detail as possible, and uses Savitzky Golay filtering and median filtering to filter Landsat time series data, using the Random Forest Regression (RFR) algorithm, a carbon storage estimation model for Pinus Densata in Shangri La City was constructed. The optimal estimation model was selected to invert and map the carbon storage of Pinus Densata in 1987, 1992, 1997, 2002, 2007, 2012, and 2017. The results showed that 1) According to the average absolute error (MAE) of the image quality evaluation index, the image quality is the best after filtering with the ATC algorithm. In addition, the PSNR value of the time series data filtered by the ATC algorithm is relatively high, indicating an improvement in data quality; 2) When using the random forest regression algorithm, the filtered data showed higher fitting and prediction accuracy than the original data; 3) When using the random forest regression algorithm, the time series data filtered based on the ATC algorithm has the best estimation accuracy when selecting the top 10 feature factors with contribution and the feature factors with cumulative contribution reaching 70% for modeling; 4) The estimation model constructed based on ATC filtered time series and remote sensing features (Number of features 10) and random forest algorithm has the best performance in research, with a determination coefficient R2 of 0.867, root mean square error RMSE of 15.527 t/ha, prediction accuracy P of 73.54%, and relative root mean square error rRMSE of 41.14%; 5) The carbon storage inversion results of Shangri La Pinus Densata based on the optimal estimation model are as follows: 6.77 million tons (1987), 7.16 million tons (1992), 7.22 million tons (1997), 4.36 million tons (2002), 7.20 million tons (2007), 7.11 million tons (2012), 7.53 million tons (2017). From the inversion results, it can be observed that during the period from 1987 to 1997, the carbon storage of Shangri-La Pinus Densata showed a gradually increasing trend. However, from 2002 to 2017, the carbon storage exhibited a significant fluctuating trend. The use of ATC filtering method can effectively remove noise in time series images of high-altitude areas, thereby reducing the uncertainty of time series images and improving the accuracy of remote sensing estimation of Pinus Densata carbon storage.
关键词:Landsat time series;Filter;Pinus Densata;carbon storage;ATC
摘要:Mining can cause severe ground subsidence, which is often accompanied by widespread and uneven characteristics, posing a great threat to the production and life of the mining community. Timely and accurate monitoring and prediction of ground subsidence in mining areas are crucial to mitigating its adverse effects. However, traditional spatio-temporal prediction models for ground subsidence often struggle with capturing comprehensive spatio-temporal information and learning the intricate features associated with this phenomenon.To address these challenges, this study incorporates a temporal decomposition strategy into a deep learning network model, resulting in the development of the Spatio-Temporal Forecasting Framework (SFF-PredRNN) model. This innovative approach takes into account seasonal displacement features, enhancing the model's ability to capture complex spatio-temporal patterns accurately. By integrating this advanced methodology, the SFF-PredRNN model offers improved predictive capabilities, allowing for more effective mitigation measures against ground subsidence and its associated risks.In this study, the focus is on the Micun coal mine located in Xinmi City, a region characterized by extensive mineral resource extraction and distinct seasonal variations in rainfall. The summer season contributes significantly to the annual rainfall, accounting for 60.9%. Certain mining areas within this region have experienced notable ground subsidence issues. Using the small baseline set interference technique algorithm, ground subsidence data from 2018 to 2021 were collected for the study area. The analysis revealed distinct spatial differences in subsidence patterns, particularly in the Mengzhuang and Zhangpocun coal mines at the center and the Wangzhuang coal mine in the southwest. These areas exhibited severe ground subsidence problems, with the maximum subsidence reaching 256 mm, while the surrounding regions did not show significant ground subsidence. A spatio-temporal dataset of ground subsidence was constructed based on the collected information, and the developed SFF-PredRNN model was employed for prediction. The model's accuracy was assessed using metrics such as MAE, RMSE, PSNR, and SSIM. Meanwhile, in order to assist in verifying the advantages of the model in the spatio-temporal prediction of the mine area, we selected a profile line crossing the mine area in the horizontal and vertical directions respectively, and selected equal spacing to take out a certain number of subsidence points, and extracted the subsidence values predicted by the model through these points and verified the results.The results demonstrated that the SFF-PredRNN model, as proposed in this study, exhibited superior accuracy in predicting subsidence for the years 2019, 2020, and 2021. This highlights the model's strengths in both temporal and spatial predictions of ground subsidence. The predictions for the upcoming year indicated a continued trend of subsidence in the mining areas of Mengzhuang, Wangzhuang, and Zhangpocun, with an expected maximum cumulative subsidence of 274.3 mm. The spatial distribution of settlement in the study area remained consistent with previous patterns.In conclusion, the SFF-PredRNN model proposed in this paper shows good performance in spatial-temporal prediction of ground subsidence, which can be used as an effective method for spatial-temporal prediction of ground subsidence, and provides effective methodological guidance for the prevention and early warning of ground subsidence disasters in mining areas. In the future, we can construct the prediction model by integrating more data of ground subsidence influencing factors to realize more accurate spatio-temporal prediction on a large scale.
摘要:Objective The successful launch of the Ice, Cloud, and land Elevation Satellite-2 (ICESat-2), carrying the Advanced Topographic Laser Altimeter System (ATLAS), has made it possible to accurately quantify global vegetation structure. However, due to limitations of the sensitive photon detection system, its data contains a large amount of background noise photons. Aiming at the problem that ICESat-2/ATLAS has low signal extraction accuracy in mountain forest area, which leads to the difficulty of ground and canopy surface detection, the method of ground and canopy surface detection based on direction adaptive ordering points to identify the clustering structure (OPTICS) is proposed.Method Firstly, the initial ground surface is obtained through segmented curve fitting based on Random Sample Consensus (RANSAC), which is used to construct a direction adaptive elliptical searching area to replace the traditional circle in OPTICS, forming the direction adaptive OPTICS algorithm. Based on this algorithm, the reachability distance (RD) of all photon point clouds is obtained first, and then potential ground signals and potential ground surface are obtained successively by a "two-step method", that is, Nobuyuki Otsu method (OTSU) is first introduced to obtain potential ground signals, and then potential ground surface is obtained by fitting potential ground signals based on RANSAC. Iterate the "two-step method" several times until the similarity between the potential ground surface obtained before and after is greater than 90%, indicating that the potential ground surface is considered a fine ground surface. Secondly, the effect of terrain on photons is eliminated by reference to the fine ground surface, and then the vegetation signal is extracted by vertical elliptical OPTICS. Finally, based on the vegetation signal, the surface of canopy is detected by combining the elevation percentile and Piecewise Cubic Hermite Interpolation (PCHIP) curve fitting.Result The ATLAS data of Mengjiagang Forest Farm in Heilongjiang Province and Fushun Forest Farm in Liaoning Province are used as the research objects to carry out experiments, and the accuracy is verified by manually labeled samples and Unmanned Aerial Vehicle (UAV) products. The results show that the extraction accuracy (F) of ground surface and vegetation signals in the mountainous forest area is 0.97, which is about 0.07 higher than that of the OPTICS based on elliptical searching area. In addition, the RMSE of ground and canopy surface detected by the proposed method are 1.08m and 2.33m, respectively, compared with 1.92m and 3.29m of ATL08, which significantly improves the accuracy.Conclusion Therefore, compared with the OPTICS based on elliptical searching area and ATL08 results, the proposed method has higher precision in extracting vegetation signal photons and detecting the surface of ground and canopy. It is more suitable for areas with large gradient changes such as mountain forest area, and can provide a reliable data foundation for the subsequent inversion of forest spatial structure.
关键词:ICESat-2;ATLAS;direction adaptive;OPTICS;iterative refinement;terrain slope;photon-counting;mountain forest area
WANG Yingqi,HUANG Huiping,ZHU Wenlu,YANG Guang,YU Kun
Corrected Proof
DOI:10.11834/jrs.20243529
摘要:Remote Sensing Ecological Index (RSEI) is the most widely used ecological environment quality assessment model. Generally, the four indicators (greenness, wetness, heat and dryness) of RSEI are calculated from Landsat images to construct an index that comprehensively reflects the ecological environment condition in pixel units. Since high-resolution remote sensing images generally lack the short-wave infrared and thermal infrared bands involved in the calculation of RSEI, the application of RSEI in high-resolution ecological environment quality assessment is limited. It is undoubtedly a great waste that the advantages of high-resolution remote sensing data cannot be fully utilized due to the limitation of spectral resolution.In order to solve the problem of mismatch between high-resolution remote sensing image bands and the bands required for RSEI calculation, this paper established a multi-resolution band fusion model with scale-invariant features. Based on Landsat8 and GF-2 remote sensing images, the short-wave infrared bands and surface temperature with high resolution (4m) were generated utilizing the statistical relationship between the bands. And the High-resolution Remote Sensing Ecological Index (HRSEI) was constructed based on the principle of RSEI, which fills the gap of RSEI research at fine scale. This method was applied in Fan County, Henan Province. The results showed that:(1) Utilizing the multi-resolution band fusion technique, high-resolution short-wave infrared band and surface temperature can be generated. The correlation coefficients between the fitted image and the original image were higher than 0.7, indicating that the machine learning model based on the random forest algorithm was effective. The obtained high-resolution band/product can be used in the subsequent ecological environment quality evaluation work. This method can effectively make up for the disadvantage of band absence of high-resolution images, breaking through the limitation of RSEI application in fine scale, and expanding the application scenario of high-resolution remote sensing data.(2) The calculation results of the first principal component of HRSEI showed that, the loadings of greenness and wetness were positive, and the loadings of heat and dryness were negative, which indicated that greenness and wetness promoted ecological environment quality, and heat and dryness impeded ecological environment quality. The above results are consistent with the objective actual pattern, and coincided with the trend of the results of RSEI. The Pearson correlation coefficient showed that HRSEI and RSEI were highly correlated (R=0.74). The contrast and information entropy of HRSEI for the three typical areas (built-up area, village and beach area) were greater than that of RSEI. It is sufficiently demonstrated that, with maintaining high relevance and consistency, the information abundance presented by HRSEI generated by 4m GF-2 data is significantly higher than that of 30m Landsat data.(3) The results of HRSEI in 2016 and 2023 showed that the ecological environment quality of Fan County had been generally improved. However, there were still some areas where the ecological environment quality deteriorated. There are two main factors that contribute to the deterioration. Firstly, urbanization has led to the expansion of built-up land, with previously cultivated or forested land being changed to impervious surfaces. Secondly, due to the policy of relocation and reclamation in the Yellow River beach area, the villages near the Yellow River have carried out the demolition of old villages and the construction of resettlement areas. Especially, the lack of timely reclamation after the demolition of old villages has seriously expanded the scope of the deterioration of the ecological environment quality.
摘要:Lakes, known as "sentinels" of global climate change, are significant contributors to the worldwide water cycle. Accurately estimating lake water storage and its fluctuations is crucial for forecasting global climate change. The Surface Water and Ocean Topography (SWOT) mission, launched in December 2022, offers comprehensive observations of the world's lakes, providing a major advancement in our understanding of lakes on a global scale. In order to better utilize SWOT data in the future, this article will conduct simulations and evaluate the application potential in lake storage estimation of SWOT mission. We will generate SWOT lake data, estimate the simulated water storage of four lakes located on the Qinghai-Tibet Plateau, analyze the errors of lake water storage derived from SWOT simulation data, and suggest necessary considerations for future use of SWOT mission in estimating lake water storage. By addressing these issues, we hope to offer valuable insights for more accurate utilization of SWOT data in estimating lake water storage in the future.In this experiment, we employed the CNES SWOT hydrology toolbox to generate simulated data of lakes. The toolbox contains three primary components: the Large Scale Simulator, RiverObs, and LOCNES. The tool computes the lake extent data intersected by the SWOT wide swath. Subsequently, the tool will generate point cloud data within the overlapping area. Each point cloud dataset contains detailed information, including the water surface height. Next, the RiverObs tool is employed to create SWOT river data in shapefile format. Finally, the LOCNES tool is used to generate lake data from the data that is not categorized as part of the river system.The National Tibetan Plateau Scientific Data Center (TPDC) provides bathymetry point data for four lakes. This study uses Topo to Raster in ArcGIS to generate the lake bathymetry. Besides, this article also uses the global lake bathymetry GloBathy. we use the true maximum depth data of four lakes provided by the TPDC to regenerate modified GloBathy data. In summary, this experiment gets three sets of lake bathymetry data of each four lakes, including true bathymetry(tru), the initial GloBathy(ori), and the modified GloBathy(mod).At last, this paper uses the SWOT simulated data to estimate the lake water storage. we analyzed the impact of errors in SWOT mission on lake storage estimation. In the SWOT PIXC data, the errors of majority of water surface elevation measurements are less than 1 m. After averaging at the scale of the lakes, the errors were mostly within 0.02 m. The error of the water surface height measured by most SWOT point clouds is less than 1 m. The correlation coefficient between the simulated water surface height sequence of the SWOT mission and the real sequence exceeds 0.9. It shows that SWOT mission can well reflect the seasonal changes in the lake water surface height. The relative error in estimating lake area from SWOT observations are less than 10 % due to the dark water effect. In this study, SWOT simulated data were used to estimate lake water storage using three types of lake bathymetry. It showed that the errors in water surface elevation had a relatively small impact on the accuracy of lake water volume estimation, while the errors in estimating the lake topography had a more significant impact on the accuracy of lake water volume estimation.The research indicates that SWOT mission have significant prospects for lake water volume estimation. Obtaining higher-precision prior data on lake depths is critical for improving accuracy in lake water storage estimation in the future. In the future, we can combine SWOT mission data with other hydrological satellite data and surface measurement data to get more accurate water storage changes in surface water.
关键词:lakes;Surface Water and Ocean Topography (SWOT);CNES SWOT Hydrology Toolbox
摘要:The Arabian plate continues to squeeze the Eurasian plate northward, which promotes stress field changes, local stress locking and rupture instability. This resulted in the magnitude 6.0 shallow earthquake on June 21, 2022 in Paktita province on the Afghanistan-Pakistan border, which is the largest earthquake in the region in the past 10 years. It is very important to study the phenomenon and mechanism of earthquake. Through analyzing the earthquake case with multi-source data, the eternal earth science topic of earthquake perception and cognition is studied deeply.In this study, the microwave brightness temperature (MBT) data collected by the AMSR-2 radiometer of the Aqua satellite was used to extract pre-earthquake and post-earthquake MBT residuals within more than a million square kilometers around the epicenter by using spatio-temporally weighted two-step method, and reveal the spatio-temporal evolution characteristics of MBT and the polymorphism of positive MBT anomalies. Based on the data of precipitation, soil moisture, regional geological map, land cover and greenhouse gas such as CH4 and CO, the attribution analysis of positive polymorphic MBT anomalies was discriminated one by one.In this study, the temporal and spatial evolution of the MBT of the Paktika earthquake was analyzed. The results showed that the MBT positive anomalies was affected by many factors such as regional plate structure activity, geological lithology and surface land cover. The MBT anomaly in seismogenic stage was polymorphic, and needed to be carefully screened using multi-source information and multi-parameters. The study is of great significance for observing and identifying the seismic anomaly in West Asia, and has reference value for the seismic remote sensing monitoring and anomaly recognition in other parts of the world.We reached that 1) The positive MBT anomaly in the Indus Plain in the southeast of the epicenter and the positive MBT anomaly in the Karakum desert to the northwest of the epicenter could be attributed to the cavity particles (P-hole) activated by the seismogenic stress, transferring from the seismogenic area to the Quaternary overburden along the stress gradient, which reduce the dielectric constant of the dielectric constant in shallow surface layer; 2) The positive MBT anomaly in the alpine area during the earthquake period could be attributed to the transfer and accumulation of stress-activated P-hole to the alpine low-temperature area, which resulted in the decrease of microwave dielectric constant of the sandy layer;3) The positive MBT anomaly along the Herat Fault in the northwest of the epicenter was related to the fault stretching during the imminent earthquake, and might had been affected by the greenhouse effect caused by the degassing of coal-bearing formations along fault and coal mines.
关键词:microwave brightness temperature;Seismic anomaly;P-hole;microwave dielectric;greenhouse effect;crustal stress field alteration
DENG Songwen,YANG Fei,WANG Yinghui,ZHANG Wei,WANG Wenhuan
Corrected Proof
DOI:10.11834/jrs.20243293
摘要:Mangroves are important blue carbon ecosystems that play a significant role in maintaining global marine carbon cycles and mitigating the rate of climate change. Remote sensing, due to its advantages of good repeatability, high resolution, and low cost, can better facilitate the monitoring and management of mangrove carbon resources. This paper reviews the research progress of remote sensing-based mangrove carbon reservoirs and categorizes the development into three stages based on the research content and depth: the early exploration stage (2007-2012), which primarily focused on global mangrove mapping and the extraction of spatial structural information; the mid-term application research stage (2013-2015), which estimated mangrove carbon stocks based on previous research achievements; and the comprehensive development stage (after 2016), characterized by improving accuracy in carbon stock estimation and a research focus on the impact mechanisms of environmental factors on mangrove carbon reservoirs. The current status of optical remote sensing and radar remote sensing methods is reviewed, and the degree of improvement in results through the fusion analysis of these two remote sensing techniques is explored. Furthermore, the performance of various mangrove carbon models in estimating carbon stocks and simulating carbon cycling in mangroves is discussed. Starting from the two important carbon reservoirs of biomass and soil in mangroves, relevant research on their carbon stocks is reviewed. The biomass carbon reservoir is primarily composed of carbon stored in vegetation roots, stems, and leaves, and it is a major influencing factor in mangrove primary productivity. However, the biomass carbon stock is highly affected by human activities and natural influences, resulting in significant fluctuations. The soil carbon reservoir, which accounts for approximately 49-98% of the total carbon stock in mangroves, is the largest carbon reservoir in mangrove ecosystems. Nevertheless, research on soil carbon reservoirs is relatively limited compared to biomass carbon reservoirs, primarily due to challenges in acquiring remote sensing data and dealing with complex optical characteristics. Considering the significant role of mangrove ecosystems in carbon sequestration and the achievement of carbon-related goals, the need for improvements in applying mangrove carbon sinks to carbon accounting and statistics is analyzed, and the potential applications of unmanned aerial vehicle remote sensing technology and artificial intelligence in mangrove carbon stock estimation are explored.
HE Jinchen,ZHANG Shuhang,FENG Wei,YAN Xingyuan,JIN Zehui,LIN Jiayuan
Corrected Proof
DOI:10.11834/jrs.20243170
摘要:High-resolution and non-contact water depth monitoring is crucial for the management and protection of tufa lake landscapes. Satellite-derived bathymetry cannot capture the subtle underwater sedimentary characteristics of tufa lakes. In recent years, the remote sensing technology of light and small unmanned aerial vehicles (UAV) has gradually been applied to ultra-high-resolution bathymetric mapping in shallow water areas. However, the classic logarithmic model in water depth inversion is difficult to adapt to the widespread Rayleigh scattering phenomenon in tufa lakes. Therefore, in this article machine learning methods are used to construct bathymetric inversion models of tufa lakes based on UAV imagery.Taking Spark Lake in Jiuzhaigou National Nature Reserve, Sichuan Province, China as the experimental area, aerial image data for bathymetric model construction were extracted from UAV platforms. Based on the pre- and post-earthquake UAV images, the pre-earthquake orthophoto with water and the post-earthquake surface model without water are generated by the Structure-from-Motion algorithm, respectively. After excluding anomalous areas, sample points for the bathymetric inversion were randomly selected. Each sample data has both the red, green, and blue band (RGB) digital number (DN) values of the pre-earthquake orthophoto and the relative depth values of the post-earthquake exposed terrain relative to the pre-earthquake water surface. Based on this dataset, machine learning regression models based on random forest (RF), support vector machine (SVM), and multi-layer perceptron (MLP) are constructed respectively. The above machine learning models are trained repeatedly to determine their respective optimal parameters. Finally, the accuracy of the estimated bathymetry was verified using the exposed lake terrain after the earthquake.The results indicate that the water depth distribution of the three models has small differences in shallow water areas, and the areas with significant differences are mainly deep-water regions. The bathymetric map simulated by the RF model is susceptible to high-frequency signals, while the bathymetric maps simulated by SVM and MLP models suffer from localized overestimation of water depth distribution. In terms of accuracy assessment, the RF, SVM, and MLP models have root mean square errors (RMSE) of 0.816 m, 0.945 m, and 0.832 m, and coefficients of determination (R2) of 0.948, 0.930, and 0.946. The RF and MLP models have relatively good consistency across the entire depth range, while the SVM model has an overestimation of depth in general within the interval of 6-9 m.To sum up, machine learning models have higher accuracy in water depth retrieval compared to traditional logarithmic models. Among them, the RF and MLP models are more suitable than the SVM model for water depth retrieval of tufa lakes based on UAV imagery. Unlike the models only utilizing blue and green bands, the introduction of red bands into machine learning models both improves the accuracy of the shallow-water bathymetry and at the same time increases the local bathymetry uncertainty. In the future, further research using UAV multispectral imagery with the coastal blue band is necessary. Given an adequate dataset, it is proposed to construct a deep convolutional neural network-based bathymetric inversion model for tufa lakes.
LI Kewen,ZHU Guanglei,WANG Hui,ZHU Rui,DI Xiyao,ZHANG Tianjian,XUE Zhaohui
Corrected Proof
DOI:10.11834/jrs.20243043
摘要:Objective Due to the imaging characteristics and the limitations of spatial resolution, feature extraction for small target is very hard, which increases the difficulty of small target detection. The existing deep learning target detection network architectures are mostly based on natural images and it is insufficient in research and exploration of small target in remote sensing images. To overcome the above issues, this paper proposes a remote sensing image small target detection algorithm that combines dual attention mechanism and bidirectional feature pyramid.Method The innovative contributions are: 1) Aiming to sovle the problem of small target occupytion in the remote sensing image and the usually huge parameter size, we introduce LKG bottleneck and GIoU loss function into YOLOv3 and propose LKGNet-YOLO network. 2) Aiming to sovle the disturbance of noise and the drawback of feature fusion, we introduce DA-LKGNet bottleneck and BiFPN into LKGNet-YOLO and propose DA-LKGNet-YOLO network.Result In order to validate the effectiveness of the proposed method in the field of remote sensing image small target detection,, the remote sensing image dataset (UCAS-AOD) released by the University of Chinese Academy of Sciences in 2014 and the AI-TOD dataset released by Wuhan University in 2021 is used for experiments. The experimental results demonstrate that the proposed method achieves a mean average precision (mAP) of 96.21% at a threshold of 0.5 on the UA dataset. On the AI-TOD dataset, the mAP0.5~0.95 at thresholds ranging from 0.5 to 0.95 is 9.51%. Compared to YOLOv3, RFBNet, SSD, FSSD, RetinaNet and RefineDet, the mAP accuracy of the proposed method is 3.45%-7.52% and 1.36%-4.84% higher, achieving significant performance improvement. Meanwhile, the detection level on small-scale targets is better than Faster-RCNN and YOLOv7 algorithms. Compared with the original YOLOv3, this method reduces the number of floating-point operations per second by 48% and the number of parameters is reduced by 42%.Conclusion A small target detection method for remote sensing image is proposed in this paper. Its contribution lies in the utilization of the YOLOv3 model with LKGNet as the backbone network, along with the design of dual attention mechanism and bi-directional feature pyramid network, resulting in a lightweight DA-LKGNet-YOLO model. This model can effectively extract small-sized objects in complex remote sensing images. The experimental results confirm the effectiveness of the method proposed in this paper.
摘要:Knowledge and data are the two main elements that have characterized the development of remote sensing image interpretation for decades. With the continuous enrichment of sensor platforms and rapid breakthroughs in deep learning, big data, multi-modal, and long time-series methodologies, data-driven intelligent remote sensing image interpretation has become a hot research direction in recent years. However, in the deepening and expanding research and applications, the limitations of data-driven methods such as difficult reuse between different scenarios, strong training sample dependence, and weak interpretability are beginning to emerge. Various types of knowledge accumulated in the long-term remote sensing image interpretation practice have the characteristics of objective reality, certainty, scene adaptability, interpretability, etc., which can be complemented with data-driven approaches, and the dual-driven of knowledge and data is becoming a new direction of remote sensing image interpretation. This paper first reviews the major stages in the development of remote sensing image interpretation and the respective roles of knowledge and data in each of these stages. Then the main types of knowledge involved in remote sensing image interpretation are summarized and categorized into fourteen types. The fusion of knowledge and deep learning is an important path to achieve the dual-drive of knowledge and data, and this paper summarizes five categories and fifteen subcategories of knowledge and deep neural network fusion methods with relevant cases. From the perspective of knowledge types, this paper further provides an overview of existing applications of remote sensing interpretation with joint knowledge and data. The effectiveness and capability increment of fusing knowledge and data is demonstrated by the analyses of typical examples. Lastly, this paper gives a systematic prospect on the framework and key techniques for knowledge and data compound driven remote sensing image interpretation.
摘要:The abstract of this study contains four sections objective, method,result and conclusion High Mountain Asia (HMA) is the richest high altitude region in the world except for the poles in terms of glacier and snow resources, The accurate monitoring of HMA snowpack distribution is important for HMA snowmelt runoff simulation, climate change prediction and ecosystem evolution. Fractional Snow Cover (FSC) can quantitatively describe the extent of snow cover at the sub-image scale, and is more suitable for reflecting the distribution of snow in complex mountainous areas than binary snow. The objective of this study is to develop a new HMA snow area ratio inversion algorithm and integrate the algorithm into Google Earth Engine to prepare a set of long time series HMA snow area ratio products.Method Considering the influence of HMA topography and sub-bedding type on the accuracy of snow accumulation information extraction, this paper proposes a Multivariate Adaptive Regression Splines (MARS) model LC-MARS to invert the proportion of snow accumulation area in Asia by integrating topography correction and subland class feature extraction. The FSC extracted by Landsat-8 is used as the true value, and the LC-MARS model is tested for inversion FSC accuracy using binary and error validation methods, and the performance of linear regression models trained with the same training samples and the LC-MARS model for inversion HMAFSC accuracy is compared, and the accuracy of the FSC inversion of the LC-MARS model with SnowCCI and MOD10A1 is also compared.Result (1) The overall accuracy of FSC binary validation of LC-MARS model inversion showed that Accuracy and Recall were 93.4% and 97.1%, respectively, and the overall accuracy of error validation showed that RMSE was 0.148 and MAE was 0.093, both binary validation and error validation indicated that the FSC accuracy of LC-MARS model inversion was higher. (2) The LC-MARS model trained based on the same training samples has higher FSC accuracy than the linear regression model in forest area, vegetation and bare land inversions, indicating that the LC-MARS model is more suitable for FSC inversions in mountain and forest areas. (3) The overall RMSE of MOD10A1 is 0.178 and MAE is 0.096; the overall RMSE of SnowCCI is 0.247 and MAE is 0.131. The accuracy of FSC prepared by LC-MARS is higher than that of MOD10A1 and SnowCCI, indicating that FSC inversion by LC-MARS has some application value.Conclusion The LC-MARS model can fit high-dimensional nonlinear relationships and significantly improve the inversion accuracy of FSC in mountain and forest areas. The computational efficiency of the LC-MARS model based on Google Earth Engine is high, and it is suitable for preparing FSC products of large scale long time series. In this study, the day-by-day MODIS FSC products of HMA from 2000 to 2021 were prepared based on the LC-MARS model, which provides important data support for the study of climate change, hydrological and water resources in HMA.