scholarly journals Satellite retrieval of aerosol microphysical and optical parameters using neural networks: a new methodology applied to the Sahara desert dust peak

2014 ◽  
Vol 7 (9) ◽  
pp. 3151-3175 ◽  
Author(s):  
M. Taylor ◽  
S. Kazadzis ◽  
A. Tsekeri ◽  
A. Gkikas ◽  
V. Amiridis
Keyword(s):  
Neural Network ◽  
Remote Sensing ◽  
Neural Networks ◽  
Refractive Index ◽  
Size Distribution ◽  
Aerosol Optical Depth ◽  
Principal Components ◽  
Complex Refractive Index ◽  
Remote Sensing Data ◽  
Optical Parameters

Abstract. In order to exploit the full-earth viewing potential of satellite instruments to globally characterise aerosols, new algorithms are required to deduce key microphysical parameters like the particle size distribution and optical parameters associated with scattering and absorption from space remote sensing data. Here, a methodology based on neural networks is developed to retrieve such parameters from satellite inputs and to validate them with ground-based remote sensing data. For key combinations of input variables available from the MODerate resolution Imaging Spectro-radiometer (MODIS) and the Ozone Measuring Instrument (OMI) Level 3 data sets, a grid of 100 feed-forward neural network architectures is produced, each having a different number of neurons and training proportion. The networks are trained with principal components accounting for 98% of the variance of the inputs together with principal components formed from 38 AErosol RObotic NETwork (AERONET) Level 2.0 (Version 2) retrieved parameters as outputs. Daily averaged, co-located and synchronous data drawn from a cluster of AERONET sites centred on the peak of dust extinction in Northern Africa is used for network training and validation, and the optimal network architecture for each input parameter combination is identified with reference to the lowest mean squared error. The trained networks are then fed with unseen data at the coastal dust site Dakar to test their simulation performance. A neural network (NN), trained with co-located and synchronous satellite inputs comprising three aerosol optical depth measurements at 470, 550 and 660 nm, plus the columnar water vapour (from MODIS) and the modelled absorption aerosol optical depth at 500 nm (from OMI), was able to simultaneously retrieve the daily averaged size distribution, the coarse mode volume, the imaginary part of the complex refractive index, and the spectral single scattering albedo – with moderate precision: correlation coefficients in the range 0.368 ≤ R ≤ 0.514. The network failed to recover the spectral behaviour of the real part of the complex refractive index. This new methodological approach appears to offer some potential for moderately accurate daily retrieval of the total volume concentration of the coarse mode of aerosol at the Saharan dust peak in the area of Northern Africa.

scholarly journals Satellite retrieval of aerosol microphysical and optical parameters using neural networks: a new methodology applied to the Sahara desert dust peak

2013 ◽  
Vol 6 (6) ◽  
pp. 10955-11010
Author(s):  
M. Taylor ◽  
S. Kazadzis ◽  
A. Tsekeri ◽  
A. Gkikas ◽  
V. Amiridis
Keyword(s):  
Remote Sensing ◽  
Neural Networks ◽  
Refractive Index ◽  
Size Distribution ◽  
Aerosol Optical Depth ◽  
Principal Components ◽  
Complex Refractive Index ◽  
Remote Sensing Data ◽  
Optical Parameters ◽  
Desert Dust

Abstract. In order to exploit the full-Earth viewing potential of satellite instruments to globally characterise aerosols, new algorithms are required to deduce key microphysical parameters like the particle size distribution and optical parameters associated with scattering and absorption from space remote sensing data. Here, a methodology based on neural networks is developed to retrieve such parameters from satellite inputs and to validate them with ground-based remote sensing data. For key combinations of input variables available from MODIS and OMI Level 3 datasets, a grid of 100 feed-forward neural network architectures is produced, each having a different number of neurons and training proportion. The networks are trained with principal components accounting for 98% of the variance of the inputs together with principal components formed from 38 AERONET Level 2.0 (Version 2) retrieved parameters as outputs. Daily-averaged, co-located and synchronous data drawn from a cluster of AERONET sites centred on the peak of dust extinction in Northern Africa is used for network training and validation, and the optimal network architecture for each input parameter combination is identified with reference to the lowest mean squared error. The trained networks are then fed with unseen data at the coastal dust site Dakar to test their simulation performance. A NN, trained with co-located and synchronous satellite inputs comprising three aerosol optical depth measurements at 470, 500 and 660 nm, plus the columnar water vapour (from MODIS) and the modelled absorption aerosol optical depth at 500 nm (from OMI), was able to simultaneously retrieve the daily-averaged size distribution, the coarse mode volume, the imaginary part of the complex refractive index, and the spectral single scattering albedo – with moderate precision: correlation coefficients in the range 0.368 ≤ R ≤ 0.514. The network failed to recover the spectral behaviour of the real part of the complex refractive index with only 39–45% of the data falling within the acceptable level of uncertainty relative to ground-truth data at the daily timescale. In the context of Saharan desert dust, this new methodological approach appears to offer some potential for moderately accurate daily retrieval of previously inaccessible aerosol parameters from space.

scholarly journals Global scale variability of the mineral dust long-wave refractive index: a new dataset of in situ measurements for climate modeling and remote sensing

2017 ◽  
Vol 17 (3) ◽  
pp. 1901-1929 ◽  
Author(s):  
Claudia Di Biagio ◽  
Paola Formenti ◽  
Yves Balkanski ◽  
Lorenzo Caponi ◽  
Mathieu Cazaunau ◽  
...  
Keyword(s):  
Remote Sensing ◽  
Refractive Index ◽  
Size Distribution ◽  
Climate Models ◽  
Mineral Dust ◽  
Complex Refractive Index ◽  
Mineralogical Composition ◽  
Global Scale ◽  
Long Wave

Abstract. Modeling the interaction of dust with long-wave (LW) radiation is still a challenge because of the scarcity of information on the complex refractive index of dust from different source regions. In particular, little is known about the variability of the refractive index as a function of the dust mineralogical composition, which depends on the specific emission source, and its size distribution, which is modified during transport. As a consequence, to date, climate models and remote sensing retrievals generally use a spatially invariant and time-constant value for the dust LW refractive index. In this paper, the variability of the mineral dust LW refractive index as a function of its mineralogical composition and size distribution is explored by in situ measurements in a large smog chamber. Mineral dust aerosols were generated from 19 natural soils from 8 regions: northern Africa, the Sahel, eastern Africa and the Middle East, eastern Asia, North and South America, southern Africa, and Australia. Soil samples were selected from a total of 137 available samples in order to represent the diversity of sources from arid and semi-arid areas worldwide and to account for the heterogeneity of the soil composition at the global scale. Aerosol samples generated from soils were re-suspended in the chamber, where their LW extinction spectra (3–15 µm), size distribution, and mineralogical composition were measured. The generated aerosol exhibits a realistic size distribution and mineralogy, including both the sub- and super-micron fractions, and represents in typical atmospheric proportions the main LW-active minerals, such as clays, quartz, and calcite. The complex refractive index of the aerosol is obtained by an optical inversion based upon the measured extinction spectrum and size distribution. Results from the present study show that the imaginary LW refractive index (k) of dust varies greatly both in magnitude and spectral shape from sample to sample, reflecting the differences in particle composition. In the 3–15 µm spectral range, k is between ∼ 0.001 and 0.92. The strength of the dust absorption at ∼ 7 and 11.4 µm depends on the amount of calcite within the samples, while the absorption between 8 and 14 µm is determined by the relative abundance of quartz and clays. The imaginary part (k) is observed to vary both from region to region and for varying sources within the same region. Conversely, for the real part (n), which is in the range 0.84–1.94, values are observed to agree for all dust samples across most of the spectrum within the error bars. This implies that while a constant n can be probably assumed for dust from different sources, a varying k should be used both at the global and the regional scale. A linear relationship between the magnitude of the imaginary refractive index at 7.0, 9.2, and 11.4 µm and the mass concentration of calcite and quartz absorbing at these wavelengths was found. We suggest that this may lead to predictive rules to estimate the LW refractive index of dust in specific bands based on an assumed or predicted mineralogical composition, or conversely, to estimate the dust composition from measurements of the LW extinction at specific wavebands. Based on the results of the present study, we recommend that climate models and remote sensing instruments operating at infrared wavelengths, such as IASI (infrared atmospheric sounder interferometer), use regionally dependent refractive indices rather than generic values. Our observations also suggest that the refractive index of dust in the LW does not change as a result of the loss of coarse particles by gravitational settling, so that constant values of n and k could be assumed close to sources and following transport. The whole dataset of the dust complex refractive indices presented in this paper is made available to the scientific community in the Supplement.

scholarly journals Remote sensing data retouching based on image inpainting algorithms in the forgery generation problem

2020 ◽  
Vol 44 (5) ◽  
pp. 763-771
Author(s):  
A.V. Kuznetsov ◽  
M.V. Gashnikov
Keyword(s):  
Neural Network ◽  
Remote Sensing ◽  
Neural Networks ◽  
Image Inpainting ◽  
Remote Sensing Data ◽  
Earth Remote Sensing ◽  
Sensing Data ◽  
Natural Data ◽  
Generation Problem

We investigate image retouching algorithms for generating forgery Earth remote sensing data. We provide an overview of existing neural network solutions in the field of generation and inpainting of remote sensing images. To retouch Earth remote sensing data, we use imageinpainting algorithms based on convolutional neural networks and generative-adversarial neural networks. We pay special attention to a generative neural network with a separate contour prediction block that includes two series-connected generative-adversarial subnets. The first subnet inpaints contours of the image within the retouched area. The second subnet uses the inpainted contours to generate the resulting retouch area. As a basis for comparison, we use exemplar-based algorithms of image inpainting. We carry out computational experiments to study the effectiveness of these algorithms when retouching natural data of remote sensing of various types. We perform a comparative analysis of the quality of the algorithms considered, depending on the type, shape and size of the retouched objects and areas. We give qualitative and quantitative characteristics of the efficiency of the studied image inpainting algorithms when retouching Earth remote sensing data. We experimentally prove the advantage of generative-competitive neural networks in the construction of forgery remote sensing data.

scholarly journals Global scale variability of the mineral dust longwave refractive index: a new dataset of in situ measurements for climate modelling and remote sensing

2016 ◽  
Author(s):  
Claudia Di Biagio ◽  
Paola Formenti ◽  
Yves Balkanski ◽  
Lorenzo Caponi ◽  
Mathieu Cazaunau ◽  
...  
Keyword(s):  
Remote Sensing ◽  
Refractive Index ◽  
Size Distribution ◽  
Climate Models ◽  
Mineral Dust ◽  
Complex Refractive Index ◽  
Mineralogical Composition ◽  
Global Scale ◽  
Source Regions

Abstract. Modelling the interaction of dust with longwave (LW) radiation is still a challenge due to the scarcity of information on the complex refractive index of dust from different source regions. In particular, little is known on the variability of the refractive index as a function of the dust mineralogical composition, depending on the source region of emission, and the dust size distribution, which is modified during transport. As a consequence, to date, climate models and remote sensing retrievals generally use a spatially-invariant and time-constant value for the dust LW refractive index. In this paper the variability of the mineral dust LW refractive index as a function of its mineralogical composition and size distribution is explored by in situ measurements in a large smog chamber. Mineral dust aerosols were generated from nineteen natural soils from Northern Africa, Sahel, Middle East, Eastern Asia, North and South America, Southern Africa, and Australia. Soil samples were selected from a total of 137 samples available in order to represent the diversity of sources from arid and semi-arid areas worldwide and to account for the heterogeneity of the soil composition at the global scale. Aerosol samples generated from soils were re-suspended in the chamber, where their LW extinction spectra (2–16 µm), size distribution, and mineralogical composition were measured. The generated aerosol exhibits a realistic size distribution and mineralogy, including both the sub- and super-micron fractions, and represents in typical atmospheric proportions the main LW-active minerals, such as clays, quartz, and calcite. The complex refractive index of the aerosol is obtained by an optical inversion based upon the measured extinction spectrum and size distribution. Results from the present study show that the LW refractive index of dust varies greatly both in magnitude and spectral shape from sample to sample, following the changes in the measured particle composition. The real part (n) of the refractive index is between 0.84 and 1.94, while the imaginary part (k) is ~ 0.001 and 0.92. For instance, the strength of the absorption at ~ 7 and 11.4 µm depends on the amount of calcite within the samples, while the absorption between 8 and 14 µm is determined by the relative abundance of quartz and clays. A linear relationship between the magnitude of the refractive index at 7.0, 9.2, and 11.4 µm and the mass concentration of calcite and quartz absorbing at these wavelengths was found. We suggest that this may lead to predictive rules to estimate the LW refractive index of dust in specific bands based on an assumed or predicted mineralogical composition, or conversely, to estimate the dust composition from measurements of the LW extinction at specific wavebands. Based on the results of the present study, we recommend using refractive indices specific for the different source regions, rather than generic values, in climate models and remote sensing applications. Our observations also suggest that the refractive index of dust in the LW does not change due to the loss of coarse particles by gravitational settling, so that a constant value could be assumed close to sources and during transport. The results of the present study also clearly suggest that the LW refractive index of dust varies at the regional scale. This regional variability has to be characterized further in order to better assess the influence of dust on regional climate, as well as to increase the accuracy of satellite retrievals over regions affected by dust. We make the whole dataset of the dust complex refractive indices obtained here available to the scientific community by publishing it in the supplementary material to this paper.

scholarly journals Improved inversion of aerosol components in the atmospheric column from remote sensing data

2020 ◽  
Author(s):  
Ying Zhang ◽  
Zhengqiang Li ◽  
Yu Chen ◽  
Gerrit de Leeuw ◽  
Chi Zhang ◽  
...  
Keyword(s):  
Remote Sensing ◽  
Refractive Index ◽  
Environmental Control ◽  
Complex Refractive Index ◽  
Regional Distribution ◽  
Remote Sensing Data ◽  
Northern Region ◽  
Water Soluble ◽  
Optical Remote Sensing ◽  
Sensing Data

Abstract. A method to calculate the complex refractive index of a multicomponent liquid system is introduced into the retrieval of atmospheric columnar aerosol components from optical remote sensing data. The aerosol components comprise five species, combining eight sub-components including black carbon (BC), water-soluble (WSOM) and water-insoluble organic matter (WIOM), ammonium nitrate (AN), sodium chloride (SC), dust-like (DU), and aerosol water in the fine and coarse modes (AWf and AWc). The uncertainty in the retrieval results is analyzed based on the simulation of typical models, showing that the complex refractive index obtained from instantaneous optical-physical inversion compares very well with that obtained from chemical estimation. The algorithm was used to retrieve the columnar aerosol components over China using the ground-based remote sensing measurements from the Sun-sky radiometer Observation NETwork (SONET) in the period from 2010 to 2016. The results were used to analyze the regional distribution and interannual variation. The analysis shows that the atmospheric columnar DU component is dominant in the northern region of China, whereas the AW is higher in the southern coastal region. The AN significantly decreased from 2011 to 2016, by 26.8 mg m−2, which is inseparable from China's environmental control policies.

scholarly journals Sub-Mode Aerosol Volume Size Distribution and Complex Refractive Index from the Three-Year Ground-Based Measurements in Chengdu China

Atmosphere ◽  
2019 ◽  
Vol 10 (2) ◽  
pp. 46 ◽  
Author(s):  
Chi Zhang ◽  
Ying Zhang ◽  
Zhengqiang Li ◽  
Yongqian Wang ◽  
Hua Xu ◽  
...  
Keyword(s):  
Refractive Index ◽  
Size Distribution ◽  
Aerosol Optical Depth ◽  
Optical Depth ◽  
Complex Refractive Index ◽  
Observation Data ◽  
Aerosol Model ◽  
Coarse Mode ◽  
Fine Mode ◽  
Volume Size

Chengdu is a typical basin city of Southwest China with rare observations of remote sensing measurements. To assess the climate change and establish a region aerosol model, a deeper understanding of the separated volume size distribution (VSD) and complex refractive index (CRI) is required. In this study, we employed the sub-mode VSD and CRI in Chengdu based on the three years observation data to investigate the sub-mode characteristics and climate effects. The annual average fraction of the fine-mode aerosol optical depth (AODf) is 92%, which has the same monthly tendency as the total AOD. But the coarse-mode aerosol optical depth (AODc) has little variation in different months. There are four distinguishing modes of VSD in Chengdu; the median radii are 0.17 μm ± 0.05, 0.31 μm ± 0.12, 1.62 μm ± 0.45, 3.25 μm ± 0.99, respectively. The multi-year average and seasonal variations of fine- and coarse-mode VSD and CRI are also analyzed to characterize aerosols over this region. The fine-mode single scattering albedos (SSAs) are higher than the coarse-mode ones, which suggests that the coarse-mode aerosols have a stronger absorbing effect on solar light than the small-size aerosol particles in Chengdu.

scholarly journals THE REASON OF THE FIREWORK TYPE OF DEPTH EDUCATION IN PROCESSING THE DATA OF REMOTE SURFACES OF THE EARTH

2019 ◽  
pp. 70-79
Author(s):  
M. P. Romanchuk
Keyword(s):  
Neural Network ◽  
Remote Sensing ◽  
Neural Networks ◽  
Deep Learning ◽  
Remote Sensing Data ◽  
Earth Remote Sensing ◽  
Technical Problems ◽  
Sensing Data ◽  
Advantages And Disadvantages ◽  
Software Frameworks

An important task in the processing of Earth remote sensing data is the automation of the decoding process of aerospace images, in particular the detection and recognition of objects in military decoding. In the article the directions of automation of decryption of photos are considered and promising from them is selected, which is based on the use of neural networks of deep learning, and also analyzed the technical problems that arise during the creation of algorithms and the deployment of trained models on a variety of mobile devices. The important role of deep-instruction software frameworks in the process of training of neural network models is aimed at facilitating development and deployment. The changes in the popularity of software frameworks in recent years have been analyzed and the need to analyze their dynamically changing capabilities has been analyzed. The most widely used software frameworks for the implementation of deep learning approaches, their advantages and disadvantages for solving tasks of thematic decryption on accessible computational resources are explored. The types of computational graphs, which use the software of deep learning, and programming languages, with the help of which it is allowed to create and deploy models of neural networks are considered. The analysis of the frameworks according to selected criteria was performed: distributed execution, architecture optimization, reflection of the learning process, joint support and portability. As a result, the software framework to be used in conducting research is highlighted, and the conclusion is drawn about the predominant framework for industrial use in the course of in-depth training of the neural network for the processing of Earth remote sensing data.

scholarly journals Improved inversion of aerosol components in the atmospheric column from remote sensing data

2020 ◽  
Vol 20 (21) ◽  
pp. 12795-12811 ◽  
Author(s):  
Ying Zhang ◽  
Zhengqiang Li ◽  
Yu Chen ◽  
Gerrit de Leeuw ◽  
Chi Zhang ◽  
...  
Keyword(s):  
Remote Sensing ◽  
Refractive Index ◽  
Organic Matter ◽  
Environmental Control ◽  
Complex Refractive Index ◽  
Regional Distribution ◽  
Remote Sensing Data ◽  
Liquid System ◽  
Water Soluble ◽  
Sensing Data

Abstract. Knowledge of the composition of atmospheric aerosols is important for reducing uncertainty in climate assessment. In this study, an improved algorithm is developed for the retrieval of atmospheric columnar aerosol components from optical remote sensing data. This is achieved by using the complex refractive index (CRI) of a multicomponent liquid system in the forward model and minimizing the differences with the observations. The aerosol components in this algorithm comprise five species, combining eight subcomponents including black carbon (BC), water-soluble organic matter (WSOM) and water-insoluble organic matter (WIOM), ammonium nitrate (AN), sodium chloride (SC), dust-like content (DU), and aerosol water content in the fine and coarse modes (AWf and AWc). The calculation of the CRI in the multicomponent liquid system allows for the separation of the water-soluble components (AN, WSOM and AWf) in the fine mode and SC and AWc in the coarse mode. The uncertainty in the retrieval results is analyzed based on the simulation of typical models, showing that the complex refractive index obtained from instantaneous optical–physical inversion compares well with that obtained from chemical estimation. The algorithm was used to retrieve the columnar aerosol components over China using the ground-based remote sensing measurements from the Sun–sky radiometer Observation NETwork (SONET) in the period from 2010 to 2016. The results were used to analyze the regional distribution and interannual variation. The analysis shows that the atmospheric columnar DU component is dominant in the northern region of China, whereas the AW is higher in the southern coastal region. The SC component retrieved over the desert in northwest China originates from a paleomarine source. The AN significantly decreased from 2011 to 2016, by 21.9 mg m−2, which is inseparable from China's environmental control policies.

scholarly journals Convolutional Neural Network for the Semantic Segmentation of Remote Sensing Images

2021 ◽  
Vol 26 (1) ◽  
pp. 200-215
Author(s):  
Muhammad Alam ◽  
Jian-Feng Wang ◽  
Cong Guangpei ◽  
LV Yunrong ◽  
Yuanfang Chen
Keyword(s):  
Neural Network ◽  
Remote Sensing ◽  
Neural Networks ◽  
Image Processing ◽  
Deep Learning ◽  
Semantic Segmentation ◽  
Natural Scene ◽  
Remote Sensing Images ◽  
Advantages And Disadvantages ◽  
Target Segmentation

AbstractIn recent years, the success of deep learning in natural scene image processing boosted its application in the analysis of remote sensing images. In this paper, we applied Convolutional Neural Networks (CNN) on the semantic segmentation of remote sensing images. We improve the Encoder- Decoder CNN structure SegNet with index pooling and U-net to make them suitable for multi-targets semantic segmentation of remote sensing images. The results show that these two models have their own advantages and disadvantages on the segmentation of different objects. In addition, we propose an integrated algorithm that integrates these two models. Experimental results show that the presented integrated algorithm can exploite the advantages of both the models for multi-target segmentation and achieve a better segmentation compared to these two models.

scholarly journals Road Extraction from Unmanned Aerial Vehicle Remote Sensing Images Based on Improved Neural Networks

Sensors ◽  
2019 ◽  
Vol 19 (19) ◽  
pp. 4115 ◽  
Author(s):  
Yuxia Li ◽  
Bo Peng ◽  
Lei He ◽  
Kunlong Fan ◽  
Zhenxu Li ◽  
...  
Keyword(s):  
Neural Network ◽  
Remote Sensing ◽  
Neural Networks ◽  
Unmanned Aerial Vehicle ◽  
Computational Efficiency ◽  
Neural Nets ◽  
Road Extraction ◽  
Remote Sensing Images ◽  
Feature Maps ◽  
Aerial Vehicle

Roads are vital components of infrastructure, the extraction of which has become a topic of significant interest in the field of remote sensing. Because deep learning has been a popular method in image processing and information extraction, researchers have paid more attention to extracting road using neural networks. This article proposes the improvement of neural networks to extract roads from Unmanned Aerial Vehicle (UAV) remote sensing images. D-Linknet was first considered for its high performance; however, the huge scale of the net reduced computational efficiency. With a focus on the low computational efficiency problem of the popular D-LinkNet, this article made some improvements: (1) Replace the initial block with a stem block. (2) Rebuild the entire network based on ResNet units with a new structure, allowing for the construction of an improved neural network D-Linknetplus. (3) Add a 1 × 1 convolution layer before DBlock to reduce the input feature maps, reducing parameters and improving computational efficiency. Add another 1 × 1 convolution layer after DBlock to recover the required number of output channels. Accordingly, another improved neural network B-D-LinknetPlus was built. Comparisons were performed between the neural nets, and the verification were made with the Massachusetts Roads Dataset. The results show improved neural networks are helpful in reducing the network size and developing the precision needed for road extraction.

Sign in / Sign up

Export Citation Format

Share Document