Three-dimensional （3D） remote sensing mechanism model is an important tool for teaching quantitative remote sensing inversion and conducting virtual experiments on new methods. Based on Radiosity theory and computer graphics algorithms， the RAPID model proposed the concept of porous individual object for vegetation， which greatly reduced the calculation of 3D radiation transfer. The simulation ability of RAPID has been extended from optical， thermal infrared to microwave bands to achieve the unified simulation of reflectivity， brightness temperature， point cloud， waveform and backscattering coefficient. RAPID is very suitable for quantitative remote sensing teaching， simple model validation， complex scene simulation and multi-source data fusion exploration. This paper generally introduced the principle， input and output as well as common application methods of RAPID， the full-band and multi-sensor 3D remote sensing mechanism model.
Three-dimensional （3D） radiative transfer model can accurately describe the interactions between solar radiation and heterogeneous land surfaces. Recently， it has become an important tool for quantitative remote sensing studies. LESS is a ray-tracing based 3D radiative transfer model， which take full advantage of the forward ray-tracing techniques for simulating radiative budget and backward ray-tracing for simulating large-scale images， which makes it possible to simulate various remote sensing data in a single model. Currently， LESS can simulate multi-angle Bidirectional Reflectance Factor （BRF）， multi-spectral/high-spectral images， fish-eye cameras， upwelling/downwelling shortwave radiation in rugged terrains and layered FPAR， etc. This simulated dataset can be used for validating physical modes， developing parameterized models， as well as training neural networks. This paper presents the fundamentals of LESS and its applications. LESS can be downloaded from www.lessrt.org.
Aiming at the problem of low efficiency for estimating large-area forest Above-Ground Biomass（AGB） using multi-mode remote sensing， this study fully integrated multi-dimensional observation characteristics of forest AGB from active and passive remotely sensed features， in order to improve the regional estimation result. Based on an analysis on two temporal estimation results， this study disclosed the spatial patterns of the regional forest AGB changes. It could provide data supports for the scientific assessments on the regional eco-environmental protection projects （i.e.， the Natural Forest Protection Project） and for improving the ability of continuous dynamic monitoring and early warning the national eco-environment by use of remote sensing. The study area is located at the Great Khingan， the Inner Monolia. Based on the active and passive multi-mode remotely sensed features extracted from the Landsat-TM5（TM） and ALOS-1 PALSAR mainly acquired in 2009，and the Gaofen-1（GF-1）and ALOS-2 PALSAR data mainly acquired in 2014， respectively， the k- Nearest Neighbor with Fast Iterative Features Selection （KNN-FIFS） method was applied to fast select the features composition to establish the optimal estimating model. The 7th and 8th National Forest resource Inventory （NFI） data were applied to training and validating （by Leave One Out method， LOO） the optimal KNN-FIFS for estimating two-temporal forest（arbor forest） AGB over study area. Based on the comparison between the two-temporal AGB results， the local forest changes from 2009 to 2014 at pixel and regional scales were quantitatively analyzed. At pixel scale， the validation based on NFI and LOO method showed that， estimates obtained a R2=0.56 and Root-Mean-Square Error （RMSE） = 25.95 t/ha， and a R2=0.64； RMSE=24.55 t/ha for 2009 and 2014， respectively. Meanwhile， as compared with NFI measurements， the average of 2009 results was over-estimated （predictions： 81.59 t/ha VS NFI measurements：78.64 t/ha）， but the average of 2014 was under-estimated （predictions： 79.63 t/ha VS NFI measurements：82.48 t/ha）. At regional scale， the overall averages of 2009 and 2014 were 88.33 t/ha， 94.61 t/ha respectively， with a increment of 6.28 t/ha，which were closed to those from previous studies using the Biomass Expansion Factor method， 87.14 t/ha for 2008， and 92.20 t/ha for 2013， respectively. The KNN-FIFS method used in this study， could largely improve the efficiency for selecting the optimal composition from high-dimensional multi-mode remotely sensed features. Full integration of the multi-dimensional observation characteristics from active and passive remotely sensed information， could improve the estimating accuracy and saturation level of forest AGB. Validation based the LOO method at pixel scale made the KNN-FIFS more robust with avoiding the random errors brought form the selection of training and validation data set. From 2009 to 2014， the local vegetation fractional coverage got to increase obviously， as well as the local forest AGB. Thanks to the implement of National Forest Protection Project， the situation of the local forest resource was effectively improved， although some forest fire were occasionally witnessed by the study years.
The Populus euphratica forest in the Tarim River Basin is a typical forest resource in the desert area. The canopy size and plant number information of Populus euphratica is of great significance for forest resource monitoring， ecological protection and restoration in the Tarim River Basin. Due to the complexity of the distribution of arbor， shrub and grass communities in the area， it is difficult to achieve accurate segmentation of canopy in dense Populus euphratica and large-scale plant number extraction. Taking the Populus euphratica forest in the middle of Tarim River as the research area， several typical Populus euphratica forest areas were selected， and the integrated processing methods of fusion deep learning and watershed segmentation were proposed. The precise segmentation of dense Populus euphratica and the extraction of Populus euphratica were carefully discussed in depth. First， the drone images （spatial resolution 0.16 m） are seamlessly stitched together to generate an orthophoto. Then U-Net convolutional neural network was used to accurately segment the canopy cover area of ??Populus euphratica. Furthermore， the marker segmentation method was used to automatically re-segment and count the intensive Populus canopy， and the number of Populus euphratica in the selected study area was calculated and accurately positioned. The results show that the average accuracy of the extraction of all canopy regions of Populus euphratica by integrated U-Net convolutional neural network is up to 94.1%. The overall accuracy of the calculation of Populus euphratica by the marker watershed segmentation method is 93.3%. The study suggests that the combination of deep learning and marker watershed methods can provide new ideas and lessons for the automation of large-scale forest resource monitoring.
A strategy of forest carbon and water fluxes simulation was proposed aiming at taking into account soil vertical movements and improving the simulation of forest carbon and water fluxes. Forest carbon and water fluxes at Changbai Mountain forest site was simulated using Biome-BGC MuSo model， which was composed of multi-layer soil module， phenology module and management module. Then multi-layer soil observed parameters were assimilated into Biome-BGC MuSo model， and modeled carbon and water fluxes were evaluated against eddy covariance data. The results demonstrated that Biome-BGC MuSo improved simulations of Net Ecosystem Exchange （NEE）， Ecosystem Respiration （ER）， and evapotranspiration （ET）. After data assimilation， carbon and water fluxed were improved at the Changbai Mountain forest site （NEE： R2 = 0.70， RMSE = 1.16 gC·m–2·d–1； ER： R2 = 0.85， RMSE = 1.97 gC·m–2·d–1； ET： R2 = 0.81， RMSE = 0.70 mm·d–1）. Data-model assimilation provides scientific technology in simulation of forest carbon and water fluxes.
Aiming at exploring the potentials of Gaofen-1 （GF-1） WFV data and Gaogen-6 （GF-6） satellite data in quantitative inversion of Forest Canopy Closure （FCC）， based on GF-1data， the simulated GF-6 data by adding two Sentinel-2A red-edge bands （RE） into GF-1 WFV multispectral data and the extracted relevant Texture Information （TI） ，Vegetation Index （VI） and Red edge Index （RI） ， a k- Nearest Neighbor with Fast Iterative Features Selection（ KNN-FIFS） method， was used to estimate the Forest Canopy Closure （FCC） in the Genhe of the Great Khingan， Inner Mongolia. Besides that， the impact of terrain was further explored by adding Topographic Factors （TF） into the feature compositions. The verification using 44 field samples and the Leave-One-Out （LOO） method showed that： FCC estimation based on GF-1 WFV is in good agreement with measured data， with R2 = 0.52， RMSE = 0.08； the GF-1 WFV+VI+TI’s has R2 = 0.56， RMSE = 0.08； the GF-1 WFV+RE+RI+TI’s has been significantly improved with R2=0.63 and RMSE=0.07； and highest accuracy from the GF-1 WFV+RE+RI+TI+TF composition with R2=0.68 and RMSE=0.07 was superior to the results from both stepwise multiple linear regressions （SMLR） （R2=0.39， RMSE=0.10） and support vector machine （SVM） （R2=0.49， RMSE=0.10） methods. It indicated that the KNN-FIFS method is more reliable for FCC estimation than both SMLR and SVM methods， and the simulated GF-6 data with red-edge information can effectively improve the estimation accuracy of FCC， especially after topographic correction.
In order to promote the application of Chinese Gaofen data in the classification of forest tree species， The six GF-2 images of the main area of Badaling National Forest Park in Yanqing District， Beijing were used as the data source， we used support vector machine-recursive feature elimination， C5.0 decision tree and feature space optimization three feature optimization methods to accomplish the object-oriented Support Vector Machines （SVM） and Random Forest （RF） forest tree classification from four feature dimensions on the basis of the hierarchical classification. we can achieve good classification results that the average Overall Accuracy of the study was 83.65%， the Producer's Accuracy of specific tree species was between 93.75% （Apricot） and 38.10% （Locust）， and the Use's Accuracy of specific tree species was between 100% （North China Larch） and 44.74% （Elm）. The results showed the C5.0 feature selection took the shortest time（0.01 h） and features selected by it could be applied to the highest classification accuracy （86.90%）. Under different feature dimensions， the Overall Accuracy of SVM classification was 3.28% higher than the RF.SVM and RF were both insensitive to feature dimensions， but good feature optimization results will still have a large impact on the classification efficiency of SVM（Highest improvement was 86.98%） and the classification accuracy of RF（Highest improvement was 9.22%）.
remote sensing images are affected by the atmospheric absorption， scattering and topographic changes， so that the radiation signals received by the sensors contain both the information of the ground features and the information of the atmosphere and terrain.In order to improve the retrieval accuracy of land surface reflectance， remote sensing images need to be pretreated.In this paper， a method of atmospheric correction for Landsat8 remote sensing images based on Look-Up Table（LUT） is proposed.The method generates LUT by the 6S radiative transfer model， in which the input parameters include water vapor content， ozone concentration， and aerosol optical depth （AOT） retrieved from the MODIS atmospheric two stage product.The atmospheric parameter table established by traditional method usually considers only a few factors， which is not applicable to atmospheric correction using MODIS product as input parameter.Therefore， this paper established a five dimensional LUT most of the input parameters， with high generality for Landsat-8 OLI sensor， and spectral analysis， to verify the the accuracy of the model of USGS surface reflectance products. The correlation （R2） between model-based NDVI and USGS-based NDVI was as high as 0.802 6.The verification results show that the method can effectively accurate inversion of surface reflectance products. It is also found that the calculated NDVI based on 6S radiation transmission model is more in line with the spectral characteristics of typical vegetation than the NDVI based on apparent reflectance. Finally， using visual interpretation， statistical analysis and Shadow-eliminated Vegetation Index（SEVI） correction results will do comparative analysis， compare the terrain subtractive effect. The results show that the 6S radiative transfer model and SEVI have little difference in the effect of terrain attenuation.
Lithium is an important strategic emerging resource. Jiajika ore deposit which is located in Sichuan province becomes one of the richest areas of Liresources in China and even in the world．The spectral measurement of its typical drill hole ZK1101 in the biggest X03 vein was conducted using ASD FieldSpec-4 portable spectroradiometer， and the spectral characteristics of pegmatite containing spodumene， pegmatite without spodumene， surrounding rocks were analyzed. It indicated that the spectrum of spodumene had three absorptions at 1 413 nm， 1 910 nm， 2 207 nm， and the discrimination of these three kinds of rocks was the spectral absorption features on 1 900 nm. Then the quantitative estimation model of lithium was built based on the correlation analysis between absorption depth of 1 900 nm and content of lithium. The study will start a new perspective of spodumene which plays an important role in ore prediction. What’s more， it will provide hyperspectral basis for quantitative estimation of content of lithium in the feature.
In order to explore the feasibility of estimating the heavy metal Cu content in soil by hyperspectral data， based on the study of the cinnamon soil of the water source protected area in Shijiazhuang， the correlation analysis between the different spectral data and the heavy metal copper content was made. The univariate partial least squares model of soil heavy metal Cu and a partial least squares model of multivariate were established. The results showed that the correlation between the spectral reflectance and the Cu content was improved by the Reciprocal Transformation First Derivative （RTFD）. The spectral sensitivity bands were 418， 427， 435， 446， 490， 673， 1 909， 1 920， 2 221 nm， which was located in the characteristic absorption region of soil iron oxide and clay minerals. The univariate partial least squares model with the best estimation effect on soil heavy metal Cu content was RTFD model， and its model determination coefficient R2 was 0.649， Root Mean Square Error （RMSE） was 1.477. The multivariate partial least squares model R2 and RMSE were 0.751 and 1.162， and the modeling effect was better than the univariate model. The research results can provide a reference for the hyperspectral estimation of heavy metal Cu in cinnamon soil in northern China.
To accurately analyze the spatial distribution and temporal variation of the aerosol in China， firstly， the Himawari-8 Level 3 Aerosol Optical Depth （AOD） products were validated by the Level 1.5 AERONET （Aerosol Robotic Network） sunphotometer measurements at 9 observation sites all over the China. Then， the hourly Himawari-8 AOD products from July 2015 to April 2018 were selected for further analyzing the spatial and temporal variation of AOD in China. The result shows that： （1） The Himawari-8 AOD agreed well with those from the AERONET， with a slope of 0.57~0.68 and a high correlation coefficient R ranging from 0.64~0.91. （2） The correlation between Himawari AOD and AERONET AOD is relatively low at noon compared to other time periods； In winter， the AOD estimates from Himawari 8 in the northern regions of China is relative worse than that in summer， but in the southern regions. （3） The values of annual-averaged AOD in China are highest in the eastern regions and lowest in the western regions； the AOD in spring and summer is obviously higher than that in autumn and winter， with the highest in summer and the second in spring. Moreover， the difference in the variation of monthly-averaged AOD between different regions of China was also significant； in most regions， the diurnal variation of daily-averaged AOD showed a trend of decreasing first， then rising and then decreasing. Besides， the highest value of daily AOD appeared at 14~16 in the afternoon， and the lowest value occurred at 18 o’clock. The results from this study can not only provide a new reference for understanding the spatial and temporal variation of atmospheric aerosols in China， but for the air pollution monitoring methods throughout the day.
Lithological mapping of mafic-ultramafic rocks is a hotspot in remote sensing petrology， and this lithological information is significant for the prediction of magmatic Ni-Cu sulfide ore deposit. This paper proposed a full-spectrum remote sensing identification model based on Landsat-8， ASTER， and PALSAR-2 data to extract mafic-ultramafic rocks. Firstly， mafic-ultramafic rock indices were proposed by feature space model， Bayesian linear discriminant analysis， and stepwise partial least squares regression analysis， based on analysis of the laboratory reflectance and emissivity spectra of field rock samples measured by portable spectrometer and related rocks and minerals spectra from spectral library in the visible to thermal infrared region， combined with the microwave scattering properties of the rock surface. And then these mafic-ultramafic rock indices were used to preliminary obtain the mafic-ultramafic rock information. Finally， the final mafic-ultramafic rock information was obtained after multi-information fusion processing based on Bayesian decision theory. The results show that the full-spectrum remote sensing identification model can accurately locate mafic-ultramafic rocks， and the extraction accuracy of the mafic-ultramafic rock information is above 94%. Furthermore，Based on mafic-ultramafic rock extraction of this identification model and structure interpretation， mineral exploration prospect also infer around Chishishan-Xiaochangshan-Zhongposhan area and Xiaoqingshan area， comparing with metallogenic rock masses of known Ni-Cu sulfide ore deposits in study area.
Convex geometry theory is the foundation of geometric spectral unmixing approaches in highly dimensional feature space for hyperspectral imagery. It is an important research field of spectral umixing for the geometric methods， and they have the characteristics of intuition， simplicity and high performance. A review of convex geometry-based spectral unmixing methods is summarized， and some primary problems the researchers usually confronted are concluded in the paper. Principally， three main problems are briefly analyzed herein： （1） the influence of Dimensionality Reduction （DR） on the geometric endmember extraction methods for hyperspectral imagery； （2） the difference of two classic simplex volume criterions for spectral unmixing； （3） Three formulas and their relationships for simplex volume calculating of spectral unmixing for the hyperspectral imagery. Finally， some elementary analysis results are obtained in the paper.
Vein-type alterations are extensively distributed and normally very thin， the extraction of these alteration zones is affected by the spatial resolution of remote sensing data seriously. To evaluate the ability of different spatial resolutions for alteration mineral mapping， taking Jintanzi area as the study area， the pixel of airborne CASI/SASI hyperspectral data were resampled to 5 m， 10 m， 15 m， 20 m and 30 m. The spectrum of muscovite from JPL spectral library and matched filtering method were utilized to extract the distribution of muscovite minerals. Spectra of pixels show that for alteration of large areas the spectral features are influenced weakly by the changing of spatial resolution and all the absorption features could be retained. Comparably， for the thin vein-type alteration， with the degradation of spatial resolution， the effect of the mixed pixel is more serious. Thus， the absorption feature of image spectra is very shallow. When the spatial resolution is 30 m， the absorption is weakest and difficult to be identified. The mapping results of the muscovite show that the thin vein-type alteration （about 1 to 5 meters wide） could be identified in the images with resolution of 5 m， 10 m and 15 m， while it is difficult for the images with resolution of 20 m and 30 m to detect.
With the increasingly maturing of the consumer multi-rotors drone techniques， its application for photogrammetry has been a reality. The paper employs consumer four-rotors drone equipped with ordinary camera for the 1∶500 large scale mapping task. Based on the image collecting characteristic of consumer drone and non-measure camera， the implementation plan is set out and the data is produced. The practical result indicates that： it is feasible to make large scale mapping with consumer drone， and its mapping precision can be qualified with the standard requirement； three-dimensional model of real-scenery generated through collecting image can be an auxiliary data for mapping inside， which can largely reduce the work of investigation and survey outside. Research shows that the program improves the efficiency of large-scale mapping， reduces the overall cost of the project， and has good economic benefits and application prospects.
Influenced by the rainy and cloudy monsoonal climate， optical remote sensing has limitations in the mapping of land use and land cover change of Nanjing City under rapid urbanization. Alternatively， Synthetic Aperture Radar （SAR） provides a feasible solution owing to the operation capacity in all-time and all-weather conditions. In this study， taking Hexi New Town and Jiangbei developing District （Nanjing） as example， we jointly applied coherent and intensity RC composition for the SAR change detection using thirteen COSMO-SkyMed images acquired in the period from 2016 to 2018. To reduce impacts from the speckle， we respectively applied Statistically Homogeneous Pixel Selection （SHPS） and Non-Local （NL） filters to coherence and intensity SAR images. The performance comparison of two aforementioned change detection approaches indicate that both of them achieved a reliable correct detection probability （up to 94%）， in particular when the adaptive filters were employed. However， the difference of the omission probability from them were evident， resulting in 66.8% of the coherent compared to 29.6 % of the intensity RC composition. Generally， the intensity RC composition is more sensitive to the change of small patches and linear features. In addition， this method is not constrained by data processing and acquisitions， such as the requirements of spatiotemporal baselines. In summary， the intensity RC composition has a better performance and potential in the urban change detection， in particular for regions where rainy and cloudy climate is prevailing.
Studying the changes of croplands in the western Jilin Province and the transformation process of paddy field in dry farmland are of great significance to the rational utilization of regional natural resources. Using remote sensing imagery as a data source， this paper uses a combination of object-oriented and decision trees to obtain information on land use. This study analysis the characteristic changes of cropland and paddy land and dry farmland and driving factors. The results show that： during the period from 1990 to 2015， the area of croplands in the western Jilin Province has increased by 2 157.33 km2， and the growth rate has gradually slowed down. The area of dry farmland has increased slightly during 1990~2000 and 2000~2010. However， there was a decreasing trend during 2010~2015. The area of paddy field has continued to expand， which has increased by 1 139.39 km2（51.7%）in 25 years. The area of the net conversion of dry farmland into paddy field is continuously increasing. It was 69.13 km2 during 1990~2000， 156.19 km2 during 2000~2010， and 288.27 km2 during 2010~2015. Population and economic growth are the main reasons for the rapid growth of croplands. The driving factors that affect the expansion of paddy field and the conversion of dry farmland to paddy field are： scientific and technological progress， construction of water conservancy facilities， policy orientation and interest-driven. This paper puts forward suggestions for the protection of croplands in the western part of Jilin Province， and provides scientific reference for regional agricultural production and ecological construction.
It is of great significance to extracting new construction land information rapidly and accurately for urbanization monitor research. Because of the low accuracy of original samples in the updating land cover method based on Change Vector Analysis in Posterior Probability Space （CVAPS）， and the poor accuracy of change detection of CVAPS， the paper proposed a new automatic approach applied to extract new construction land information effectively. This method was improved from the updating land cover method based on CVAPS by combining with Multivariate Alteration Detection （MAD）. The method firstly introduced MAD results of bi-temporal images to improve the accuracy of the initial samples， then added MAD results into the process of iterating samples selection in order to improve the accuracy of change detection and reclassification， thereby extracting new construction land more precisely. A case study of bi-temporal GF-1 images and land use/cover map in Jiaxing area was conducted to test performance of the improved method， and compared this method with CVAPS method. The experimental results show that the new construction land extracted by improved method in 2017 has higher accuracy， its overall accuracy of the updated construction land in 2017 reached 85% and its kappa coefficient is above 0.7. The accuracy of change detection is significantly higher than CVAPS method. Meanwhile， the proposed method reduced number of iteration and raised extraction efficiency significantly.
Research shows that lightning activity is generally ahead of the strong convection center， and strong lightning activity has a good correspondence with heavy precipitation. It is of great significance to apply lightning data in severe weather monitoring on a large scale. A dense network of lightning monitoring stations has been established in China at present， it provides accurate location and frequency of lightning that occurs nearby. But it’s difficult to give a comprehensive lightning distribution image， due to the limit of ground environment. In the field of view from the static orbit platform， the Lightning Mapping Imager （LMI） on FY4 makes up for the lack of ground monitoring and provides important information for severe convective weather monitoring while making continuous and uninterrupted observation of lightning. Taking the rainstorm in Xiamen on May 7， 2018 as a case study， FY-4 bright temperature data， automatic weather station precipitation data and the fusion data of FY-4 lightning data and ground lightning location network data are used to analyze the application of lightning data in monitoring and warning heavy precipitation. The study shows that data fusion of FY-4 lightning data and ground lightning can effectively reduce the incompleteness， uncertainty and error of the respective data of the both. The moving track of lightning is consistent with that of convective cloud， and the former always lies ahead of the latter. The intensity of lightning is stronger in deep convection and areas with large temperature gradient， and it is positively proportional to the intensity of rain. The peak frequency of lightning mostly occurs about 45 minutes before the peak of precipitation.
In order to estimate the instantaneous precipitation rates brought by the typhoon， the Level 1 brightness temperatures from the Microwave Humidity and Temperature Sounder （MWHTS） onboard the FY-3C satellite and the Tropical Rainfall Measuring Mission （TRMM） Multi-Satellite Precipitation Analysis （TMPA） 3B42 precipitation product data are used to retrieve the precipitation rates in the typhoon area using the multiple linear regression and BP neural network retrieval algorithms. The results show that the precipitation distribution maps retrieved by these two algorithms can be clearly observed the location， distribution and structural information of the typhoons such as typhoon center， cloud wall and spiral rain belt， which are consistent with the TMPA 3B42 precipitation product data. In addition， from a quantitative point of view， the TMPA 3B42 precipitation data and surface precipitation rate （mm/hr） retrieved by these two precipitation retrieval algorithms reach higher correlation and smaller deviations and root mean square errors， and the retrieval accuracy is higher. Therefore， these two retrieval algorithms can be used to retrieve the precipitation in the typhoon area. It also shows that microwave on-orbit observation data from the FY-3C MWHTS can play a high application value in typhoon monitoring and precipitation research.
The performance of non-precipitation echo identification algorithm directly affects the results of downstream applications （such as quantitative precipitation estimation）. Therefore， it is important to evaluate the algorithm performance objectively and quantitatively. The algorithm used in this paper is based on the quality control algorithm used in SWAN system. The method of clear air echoes identification was added in the algorithm， the echo extension was used and the height limitation of the echoes was added to identify and remove the clear air echoes. The spatial and temporal matching between radar data from the Precipitation Radar （PR） on board the Tropical Rainfall Measuring Mission （TRMM） satellite and the ground-based radar （GR） before and after the non-precipitation echo removal was carried out to objectively assess the performance the algorithm visually and quantitatively. According to observations of PR， the algorithm has a good identifying performance for the AP echoes which is not aliased to the precipitation echoes， and the result is better than that AP echoes aliased to the precipitation echoes. Part of the convective echo information lost after processing by the algorithm because of the factor TDBZ. For the situation of precipitation aliased to AP echoes， the evaluation was carried out by comparing the PR reflectivity factor values with those of GR before and after the identifying. The result indicated that the observations after identifying was closer to those of PR. The evaluation results can provide basis for the applicability analysis and improvement of the identifying algorithm.
Based on the correlation analysis and relative error methods， the accuracy of TRMM 3B43 v7 precipitation data at the watershed and grid scale in the Yellow River Basin was validated using 90 meteorological stations data. The spatial distribution characteristics of the accuracy evaluation indexes were analyzed. The influence of elevation and precipitation intensity on the accuracy of TRMM precipitation was discussed. The results show that： （1） At the basin scale， the TRMM monthly precipitation data is highly linearly related to the measured precipitation data at the site. TRMM precipitation data is slightly higher than the site-measured precipitation data. （2） At the grid scale， the TRMM monthly precipitation data of most grids have a high correlation coefficient with the measured precipitation data at the site. The deviation between TRMM precipitation and in site measured precipitation is small. （3） The accuracy of TRMM precipitation is related to precipitation intensity and elevation. The average absolute error between TRMM precipitation and measured precipitation is decreasing from southeast to northwest， which is consistent with precipitation distribution in the Yellow River Basin. The relative error， average error and average absolute error tend to decrease with the increase of elevation. Overall， over the Yellow River Basin， TRMM data tend to underestimate precipitation as precipitation increases. TRMM data underestimates precipitation in high altitude areas， overestimates precipitation in low altitude areas. By assessing the accuracy of TRMM satellite precipitation products in the Yellow River Basin， it provides an effective supplement for ground precipitation products in the region.
Heavy rainfall attacked Hunan province during late June to early July in 2017， causing various secondary disasters and severe financial losses. Here， three GSMaP （Global Satellite Mapping of Precipitation） datasets （i.e.， GSMaP_NRT， GSMaP_MVK and GSMaP_Gauge） were investigated based on several statistical metrics and the error decomposition model， in an effort to analyze their error structure and variation characteristics， assess the capability of GSMaP products for monitoring heavy rain events. Results show that： （1） All datasets can well capture the spatial distribution and temporal characteristics of the heavy rainfall. （2） Due to the interference of orographic convection， all three datasets show uncertainties over mountainous regions. In the error structure， hit bias contributes most to the total error. （3） GSMaP_Gauge performs best for monitoring extremely heavy rainfall， missed error has a significant decrease after applying the gauge adjustments. Compared to IMERG products， GSMaP products shows much higher accuracy in monitoring heavy rainfall. We expected the results documented here can provide feedback for further improving the GSMaP retrieving algorithm and strengthening data quality during heavy rainfall periods.
