Nothing Special   »   [go: up one dir, main page]
Next Issue
Volume 12, November-2
Previous Issue
Volume 12, October-2
You seem to have javascript disabled. Please note that many of the page functionalities won't work as expected without javascript enabled.
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
8.3
4.2
Journals
Remote Sensing
Volume 12
Issue 21
remotesensing-logo

Journal Menu

► Journal Menu

Journal Browser

► Journal Browser
share announcement

Remote Sens., Volume 12, Issue 21 (November-1 2020) – 206 articles

Cover Story (view full-size image): The climatological surface solar radiation (SSR) is an important indicator of the solar energy production potential. In the Baltic area, previous studies have indicated lower cloud amounts over seas than over land. Here, we quantify the climatological land–sea contrast of the SSR using two satellite data records of EUMETSAT. Our results show that off-shore locations on average receive higher SSR than inland areas and that the land–sea contrast is strongest during the summer. Furthermore, the land–sea contrast in the summer time SSR exhibits similar behavior in various parts of the Baltic. The annual SSR is 8% higher 20 km off the coastline than 20 km inland. We further find that convective clouds are a key driver of this behavior, as they tend to form over inland areas rather than over the seas during the summer part of the year. View this paper
  • Issues are regarded as officially published after their release is announced to the table of contents alert mailing list.
  • You may sign up for e-mail alerts to receive table of contents of newly released issues.
  • PDF is the official format for papers published in both, html and pdf forms. To view the papers in pdf format, click on the "PDF Full-text" link, and use the free Adobe Reader to open them.
Order results
Result details
Section
Show export options expand_more Show export options expand_less
Select all
Export citation of selected articles as:
19 pages, 8848 KiB  
Article
Reconstruction of Spatiotemporally Continuous MODIS-Band Reflectance in East and South Asia from 2012 to 2015
by Bo Gao, Huili Gong, Jie Zhou, Tianxing Wang, Yuanyuan Liu and Yaokui Cui
Remote Sens. 2020, 12(21), 3674; https://doi.org/10.3390/rs12213674 - 9 Nov 2020
Cited by 1 | Viewed by 2534
Abstract
To reconstruct Moderate Resolution Imaging Spectroradiometer (MODIS) band reflectance with optimal spatiotemporal continuity, three bidirectional reflectance distribution function (BRDF) models—the Ross-Thick-Li-Sparse Reciprocal (RTLSR) model, Gao model, and adjusted BF model—were used to retrieve MODIS-band reflectance for cloudy MODIS pixels according to different inversion [...] Read more.
To reconstruct Moderate Resolution Imaging Spectroradiometer (MODIS) band reflectance with optimal spatiotemporal continuity, three bidirectional reflectance distribution function (BRDF) models—the Ross-Thick-Li-Sparse Reciprocal (RTLSR) model, Gao model, and adjusted BF model—were used to retrieve MODIS-band reflectance for cloudy MODIS pixels according to different inversion conditions with a proposed filling algorithm. Then, a spatiotemporally continuous MODIS-band reflectance dataset for most of Asia with more than 98% spatiotemporal coverage was reconstructed from 2012 to 2015. The validation highlighted an evident improvement in filling cloudy MODIS observations; a reasonable spatial distribution, such as in South Asia and Southeast Asia; and acceptable precision for the filled MODIS pixels, with the root mean square error percentage (RMSE%) at 9.7–9.8% and 12–16% for the Gao and adjusted BF models, respectively. In the course of reconstructing the spatiotemporal continuous MODIS-band reflectance, the differences among the three models were discussed further. For a 16-day period with a stable and unchanged land surface, the RTLSR model, as a basic model, accurately derived land surface reflectance (no more than 10% RMSE% for MCD43C1 V006 band 1) and outperformed the other two models. When the inversion period is sufficiently long (e.g., 108 days, 188 days, 268 days, or a full year), the Gao/adjusted BF model provides better precision than the RTLSR model by considering the normalized difference vegetation index (NDVI) and soil moisture/NDVI as intermediate variables used to adjust the BRDF parameters in real time. The Gao model is optimal when the inversion period is sufficiently long. Based on combining the RTLSR model and Gao/adjusted BF model, we proposed a filling algorithm to derive a dataset of MODIS-band reflectance with optimal spatiotemporal continuity. Full article
►▼ Show Figures

Figure 1

Figure 1
<p>Flowchart of the reconstruction of no-cloud MODIS-band reflectance.</p> Full article ">Figure 2
<p>Experimental area: (<b>a</b>) a true-color image based on the data from MCD43C4CMG V006 on 13 April 2015; (<b>b</b>) land cover in 2012; (<b>c</b>) cumulative number of cloud days from 1 January 2012 to 21 September 2015; and (<b>d</b>) the spatiotemporal distribution of filling values and invalid values from 1 January 2012 to 21 September 2015; and, red vectors cover the several provinces of South China: ‘SC’ for Sichuan province, ‘GZ’ for Guizhou province, ‘YN’ for Yunnan province, ‘GX’ for Guangxi province and ‘CQ’ for Chongqing City, in which there are the most cloudy days in East and South Asia.</p> Full article ">Figure 3
<p>Derived MODIS band-1 reflectance based on three models: (<b>a</b>) Gao model, (<b>b</b>) BF model, and (<b>c</b>) RTLSR model with different inversion periods (108, 188, and 268 days) on the 216th day of the year in 2012.</p> Full article ">Figure 4
<p>Time series of the daily mean MODIS band 1–7 reflectance within no-cloud MODIS pixels including that from MODIS observation (black solid line) and that derived by three models with 268-day inversion period (blue solid line for RTLSR model, orange solid line for Gao model and green solid line for the adjusted BF model); the time series of the difference between daily mean MODIS band 1–7 reflectance within no-cloud MODIS pixels from MODIS observation and that derived by three models with 268-day inversion period (blue dotted line for RTLSR model, orange dotted line for Gao model and green dotted line for the adjusted BF model); the left <span class="html-italic">y</span>-axis means the value of MODIS band reflectance, the right <span class="html-italic">y</span>-axis means the bias which the modelled minus the MODIS observation at no-cloud MODIS pixel.</p> Full article ">Figure 5
<p>Cloud for MODIS observations (white color meaning cloud, black meaning no-cloud) on the 216th day of the year in 2012 (<b>a</b>), QA (white being 255 and white gray being 5) from MCD43C1 V006 (<b>b</b>), the derived MODIS band reflectance on the 216th day of the year in 2012 from MCD43C1 V006 (MODIS band 1) and the proposed algorithm (MODIS band 1–7).</p> Full article ">Figure 6
<p>Daily mean RMSE% of the time series of the derived MODIS band-1 reflectance from the 1st day of the year in 2012 to the 298th day of the year in 2015 by the Gao model, RTLSR model and adjusted BF model with the different inversion periods: 108-day period, 188-day period, and 268-day period; and the statistic of total daily mean RMSE% is show the right part.</p> Full article ">Figure 7
<p>RMSE% spatial distribution of the time series of derived MODIS band-1 reflectance from the 1st day of the year in 2012 to the 298th day of the year in 2015: Gao model, RTLSR model, and adjusted BF model. The corresponding standard deviations of soil moisture and the NDVI are shown.</p> Full article ">Figure 8
<p>Spatial distribution of the AFX derived by MODIS band-1 BRDF parameters on the 80th day and 216th day of the year in 2012: Gao model, RTLSR model and adjusted BF model within a 268-day inversion period.</p> Full article ">Figure 9
<p>Histogram of the spatial distribution of RMSE% for the time series and the precision statistics for land surface reflectance converted from MODIS bands 1–7 reflectance based on MODIS observations and the Gao model with dynamic parameters in 2012.</p> Full article ">Figure 10
<p>Cumulative number of no-cloud MODIS observations (<b>a</b>), the cumulative filled and invalid MODIS-band reflectance values derived from MCD43C1 V006 (<b>b</b>), and the cumulative valid MODIS-band reflectance values derived from MCD43C1 V006 (<b>c</b>), and our algorithm (<b>d</b>) from the 1st day of the year in 2012 to the 298th day of the year in 2015.</p> Full article ">
18 pages, 7156 KiB  
Article
An Improved Spatiotemporal Data Fusion Method Using Surface Heterogeneity Information Based on ESTARFM
by Mengxue Liu, Xiangnan Liu, Xiaobin Dong, Bingyu Zhao, Xinyu Zou, Ling Wu and Hejie Wei
Remote Sens. 2020, 12(21), 3673; https://doi.org/10.3390/rs12213673 - 9 Nov 2020
Cited by 16 | Viewed by 3858
Abstract
The use of the spatiotemporal data fusion method as an effective data interpolation method has received extensive attention in remote sensing (RS) academia. The enhanced spatial and temporal adaptive reflectance fusion model (ESTARFM) is one of the most famous spatiotemporal data fusion methods, [...] Read more.
The use of the spatiotemporal data fusion method as an effective data interpolation method has received extensive attention in remote sensing (RS) academia. The enhanced spatial and temporal adaptive reflectance fusion model (ESTARFM) is one of the most famous spatiotemporal data fusion methods, as it is widely used to generate synthetic data. However, the ESTARFM algorithm uses moving windows with a fixed size to get the information around the central pixel, which hampers the efficiency and precision of spatiotemporal data fusion. In this paper, a modified ESTARFM data fusion algorithm that integrated the surface spatial information via a statistical method was developed. In the modified algorithm, the local variance of pixels around the central one was used as an index to adaptively determine the window size. Satellite images from two regions were acquired by employing the ESTARFM and modified algorithm. Results showed that the images predicted using the modified algorithm obtained more details than ESTARFM, as the frequency of pixels with the absolute difference of mean value of six bands’ reflectance between true observed image and predicted between 0 and 0.04 were 78% by ESTARFM and 85% by modified algorithm, respectively. In addition, the efficiency of the modified algorithm improved and the verification test showed the robustness of the modified algorithm. These promising results demonstrated the superiority of the modified algorithm to provide synthetic images compared with ESTARFM. Our research enriches the spatiotemporal data fusion method, and the automatic selection of moving window strategy lays the foundation of automatic processing of spatiotemporal data fusion on a large scale. Full article
(This article belongs to the Special Issue Multi-Sensor Data Fusion and Analysis of Multi-Temporal Remote Sensed Imagery)
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>Location map of study areas. Study area <b>A</b> is located in Jiujiang, Jiangxi Province, while study area <b>B</b> is located in Langfang, Hebei Province.</p> Full article ">Figure 2
<p>The schematic diagram of the adaptive moving window strategy in the spatiotemporal data fusion method.</p> Full article ">Figure 3
<p>The algorithm flow of an adaptive moving window strategy in the spatiotemporal data fusion method.</p> Full article ">Figure 4
<p>The comparison of actual images and predicted images using different spatiotemporal data fusion algorithms. Note: the upper row are true-color-composites of Landsat and Landsat-like images; the lower row are false-color-composites of Landsat and Landsat-like images; (<b>a</b>,<b>d</b>) are actual observed images; (<b>b</b>,<b>e</b>) are Landsat-like images predicted by ESTARFM; (<b>c</b>,<b>f</b>) are Landsat-like images predicted by the modified algorithm.</p> Full article ">Figure 5
<p>The comparison of actual images and predicted images using different spatiotemporal data fusion algorithms. (<b>a</b>) is the actual image; (<b>b</b>) is the predicted image by ESTARFM; (<b>c</b>) is the predicted images by the modified algorithm.</p> Full article ">Figure 6
<p>Scatter plots of the actual reflectance values and estimated values by the ESTARFM (left column, <b>a</b>,<b>c</b>,<b>e</b>,<b>g</b>,<b>i</b>,<b>k</b>) and modified algorithm (right column, <b>b</b>,<b>d</b>,<b>f</b>,<b>h</b>,<b>j</b>,<b>l</b>).</p> Full article ">Figure 7
<p>The comparison of the absolute difference of <span class="html-italic">R<sub>mean</sub></span> between the true observed image and predicted image by ESTARFM (<b>a</b>) and the modified algorithm (<b>b</b>).</p> Full article ">Figure 8
<p>Frequency statistics histogram of the <span class="html-italic">R<sub>mean</sub></span> absolute difference of results predicted by ESTARFM (<b>a</b>) and the modified algorithm (<b>b</b>) in Jiujiang.</p> Full article ">Figure 9
<p>The comparison of the absolute difference of <span class="html-italic">R<sub>mean</sub></span>between the true observed image and the predicted image by ESTARFM (<b>a</b>) and the modified algorithm (<b>b</b>) in Langfang.</p> Full article ">Figure 10
<p>Frequency statistics histogram of the <span class="html-italic">R<sub>mean</sub></span> absolute difference of results predicted by ESTARFM (<b>a</b>) and the modified algorithm (<b>b</b>) in results of Langfang.</p> Full article ">Figure 11
<p>The use of different window sizes in the algorithm in Jiujiang. (<b>a</b>) is statistical table of number of pixels corresponding to each window size and (<b>b</b>) is the distribution map of the half window size of each pixel in the modified algorithm.</p> Full article ">
19 pages, 3275 KiB  
Article
Development of a Seamless Forecast for Solar Radiation Using ANAKLIM++
by Isabel Urbich, Jörg Bendix and Richard Müller
Remote Sens. 2020, 12(21), 3672; https://doi.org/10.3390/rs12213672 - 9 Nov 2020
Cited by 6 | Viewed by 2817
Abstract
A novel approach for a blending between nowcasting and numerical weather prediction (NWP) for the surface incoming shortwave radiation (SIS) for a forecast horizon of 1–5 h is presented in this study. The blending is performed with a software tool called ANAKLIM++ (Adjustment [...] Read more.
A novel approach for a blending between nowcasting and numerical weather prediction (NWP) for the surface incoming shortwave radiation (SIS) for a forecast horizon of 1–5 h is presented in this study. The blending is performed with a software tool called ANAKLIM++ (Adjustment of Assimilation Software for the Reanalysis of Climate Data) which was originally designed for the efficient assimilation of two-dimensional data sets using a variational approach. A nowcasting for SIS was already presented and validated in earlier publications as seamless solar radiation forecast (SESORA). For our blending, two NWP models, namely the ICON (Icosahedral Non-hydrostatic model) from the German weather Service (DWD) and the IFS (Integrated Forecasting System) from the European Centre for Medium-Range Weather Forecasts (ECMWF), were used. The weights for the input data for ANAKLIM++ vary for every single forecast time and pixel, depending on the error growth of the nowcasting. The results look promising, since the root mean square error (RMSE) and mean absolute error (MAE) of the blending are smaller than the error measures of the nowcasting or NWP models, respectively. Full article
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>Scheme of the temporal availability of the input forecasts for a blending with ANAKLIM++ (adjustment of assimilation software for the reanalysis of climate data). The distance of the depicted forecasts is proportional to the time range in which they are available. The dots on the line mark a break in the time line due to a changing temporal availability and differing runtimes of the utilized NWP (numerical weather prediction) models. For each blending 1 h of NWP, NWC nowcasting and reference needs to be available. The procedure can be repeated every hour.</p> Full article ">Figure 2
<p>Hourly averaged SIS (surface incoming shortwave radiation) of (<b>a</b>) the blending with ANAKLIM++, (<b>b</b>) SARAH-2 (surface radiation data set–Heliosat), (<b>c</b>) NWC, (<b>d</b>) NWC with data gaps, (<b>e</b>) IFS and (<b>f</b>) ICON for 2017-08-07 11 UTC, respectively.</p> Full article ">Figure 3
<p>Hourly averaged SIS of (<b>a</b>) the blending with ANAKLIM++, (<b>b</b>) SARAH-2, (<b>c</b>) NWC, (<b>d</b>) IFS and (<b>e</b>) ICON for 2017-09-30 11 UTC, respectively.</p> Full article ">Figure 4
<p>Mean error measures of all cases against day time for (<b>a</b>) RMSE (root mean square error) with whole NWC, (<b>b</b>) RMSE with gaps in NWC, (<b>c</b>) MAE (mean absolute error) with gaps, (<b>d</b>) bias corrected RMSE with gaps and (<b>e</b>) bias with gaps. The empirical variance is depicted by error bars for (<b>a</b>–<b>d</b>). The validation of SIS was performed with SARAH-2 data and the region of interest is Europe.</p> Full article ">
17 pages, 4833 KiB  
Article
The Ultra-Short-Term Forecasting of Global Horizonal Irradiance Based on Total Sky Images
by Junxia Jiang, Qingquan Lv and Xiaoqing Gao
Remote Sens. 2020, 12(21), 3671; https://doi.org/10.3390/rs12213671 - 9 Nov 2020
Cited by 18 | Viewed by 3866
Abstract
Solar photovoltaics (PV) has advanced at an unprecedented rate and the global cumulative installed PV capacity is growing exponentially. However, the ability to forecast PV power remains a key technical challenge due to the variability and uncertainty of solar irradiance resulting from the [...] Read more.
Solar photovoltaics (PV) has advanced at an unprecedented rate and the global cumulative installed PV capacity is growing exponentially. However, the ability to forecast PV power remains a key technical challenge due to the variability and uncertainty of solar irradiance resulting from the changes of clouds. Ground-based remote sensing with high temporal and spatial resolution may have potential for solar irradiation forecasting, especially under cloudy conditions. To this end, we established two ultra-short-term forecasting models of global horizonal irradiance (GHI) using Ternary Linear Regression (TLR) and Back Propagation Neural Network (BPN), respectively, based on the observation of a ground-based sky imager (TSI-880, Total Sky Imager) and a radiometer at a PV plant in Dunhuang, China. Sky images taken every 1 min (minute) were processed to determine the distribution of clouds with different optical depths (thick, thin) for generating a two-dimensional cloud map. To obtain the forecasted cloud map, the Particle Image Velocity (PIV) method was applied to the two consecutive images and the cloud map was advected to the future. Further, different types of cloud fraction combined with clear sky index derived from the GHI of clear sky conditions were used as the inputs of the two forecasting models. Limited validation on 4 partly cloudy days showed that the average relative root mean square error (rRMSE) of the 4 days ranged from 5% to 36% based on the TLR model and ranged from 12% to 32% based on the BPN model. The forecasting performance of the BPN model was better than the TLR model and the forecasting errors increased with the increase in lead time. Full article
(This article belongs to the Special Issue Forecasting Cloudiness Using Remote Sensing Techniques and Sky Camera Imagery)
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>Appearance and composition of total sky imager (TSI-880).</p> Full article ">Figure 2
<p>TSI image: (<b>a</b>) Raw image, (<b>b</b>) occlusion location, (<b>c</b>) occlusion recovery.</p> Full article ">Figure 3
<p>Relationship of the scattering angle, zenith angle and azimuthal angle in spherical coordinates (from ref. [<a href="#B44-remotesensing-12-03671" class="html-bibr">44</a>]).</p> Full article ">Figure 4
<p>Distribution of RBR (red-blue-ratio) with SPA (Sun-Pixel Angle) and IZA (image zenith angle) in clear sky condition: (<b>a</b>) RBR distribution on 11 August 2015 at SZA = 58° (solar zenith angle); (<b>b</b>) standard deviation of RBR on 4 May, 10 June, 11 August and 24 November 2015 at SZA = 58°.</p> Full article ">Figure 5
<p>Probability distribution of three types of pixel RBR difference on sample images.</p> Full article ">Figure 6
<p>Process of cloud detection (<b>a</b>) image at 15:10 on 7 May 2015 at SZA = 58°, SAA = 239° (solar azimuth). (<b>b</b>) RBR image. (<b>c</b>) The background RBR image at 15:13 on 7 May 7 at SZA = 58°, SAA = 239°. (<b>d</b>) RBR difference of (<b>b</b>,<b>c</b>). (<b>e</b>) Output of cloud detection. (<b>f</b>) Cloud map corrected by SP (sun parameter) (SP = 0.78), white: thick cloud, gray: thin cloud, blue: clear sky.</p> Full article ">Figure 7
<p>Determination of representative velocity vector.</p> Full article ">Figure 8
<p>Cloud map forecast and matching error: the forecast cloud map at <span class="html-italic">T</span> + <span class="html-italic">t</span> (<b>c</b>) is produced by advecting the cloud map at time <span class="html-italic">T</span> (<b>b</b>) in the direction of the motion vector computed from (<b>a</b>) (<span class="html-italic">T</span> − 1) and (<b>b</b>) (<span class="html-italic">T</span>). To determine the forecast error (<b>f</b>), the future binary cloud map at time <span class="html-italic">T</span> + <span class="html-italic">t</span> (<b>e</b>) is compared to the forecast binary cloud map (<b>d</b>). Red and green colors in (<b>f</b>) show forecast errors (red: pixel forecast cloudy but actually clear; green: pixel forecast clear but actually cloudy) and white indicates accurate forecasts.</p> Full article ">Figure 9
<p>Nowcasting irradiance scatter plots on (<b>a</b>) 5 November 2017, (<b>b</b>) 10:00—12:00 on 5 November 2017, (<b>c</b>) 12:00—14:00 on 5 November 2017, (<b>d</b>) 14:00–16:00 on 5 November 2017.</p> Full article ">Figure 10
<p>Nowcasting irradiance scatter plots on (<b>a</b>) 7 November 2017, (<b>b</b>) 10:00—12:00 on 7 November 2017, (<b>c</b>) 12:00—14:00 on 7 November 2017, (<b>d</b>) 14:00—16:00 on 7 November 2017.</p> Full article ">Figure 11
<p>Scatter plots of global horizonal irradiance (GHI) forecasting based on ternary linear regression (TRL) model for 1 min (<b>a</b>), 2 min (<b>b</b>), 3 min (<b>c</b>), 5 min (<b>d</b>) on 5 November 2017.</p> Full article ">Figure 12
<p>Scatter plots of GHI forecasting based on the BPN model for 1 min (<b>a</b>), 2 min (<b>b</b>), 3 min (<b>c</b>), 5 min (<b>d</b>) on 5 November 2017.</p> Full article ">Figure 13
<p>Scatter plots of GHI forecasting based on the BPN model for 1 min (<b>a</b>), 2 min (<b>b</b>), 3 min (<b>c</b>), 5 min (<b>d</b>) on 11 November 2017.</p> Full article ">
22 pages, 11585 KiB  
Article
Spatio-Temporal Evolution, Future Trend and Phenology Regularity of Net Primary Productivity of Forests in Northeast China
by Chunli Wang, Qun’ou Jiang, Xiangzheng Deng, Kexin Lv and Zhonghui Zhang
Remote Sens. 2020, 12(21), 3670; https://doi.org/10.3390/rs12213670 - 9 Nov 2020
Cited by 21 | Viewed by 3726
Abstract
Net Primary Productivity (NPP) is one of the significant indicators to measure environmental changes; thus, the relevant study of NPP in Northeast China, Asia, is essential to climate changes and ecological sustainable development. Based on the Global Production Efficiency (GLO-PEM) model, this study [...] Read more.
Net Primary Productivity (NPP) is one of the significant indicators to measure environmental changes; thus, the relevant study of NPP in Northeast China, Asia, is essential to climate changes and ecological sustainable development. Based on the Global Production Efficiency (GLO-PEM) model, this study firstly estimated the NPP in Northeast China, from 2001 to 2019, and then analyzed its spatio-temporal evolution, future changing trend and phenology regularity. Over the years, the NPP of different forests type in Northeast China showed a gradual increasing trend. Compared with other different time stages, the high-value NPP (700–1300 gC·m−2·a−1) in Changbai Mountain, from 2017 to 2019, is more widely distributed. For instance, the NPP has an increasing rate of 6.92% compared to the stage of 2011–2015. Additionally, there was a significant advance at the start of the vegetation growth season (SOS), and a lag at the end of the vegetation growth season (EOS), from 2001 to 2019. Thus, the whole growth period of forests in Northeast China became prolonged with the change of phenology. Moreover, analysis on the sustainability of NPP in the future indicates that the reverse direction feature of NPP change will be slightly stronger than the co-directional feature, meaning that about 30.68% of the study area will switch from improvement to degradation. To conclude, these above studies could provide an important reference for the sustainable development of forests in Northeast China. Full article
(This article belongs to the Special Issue Advanced Phenology, and Land Cover and Land Use Change Studies)
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>Location of the study area in Northeast China.</p> Full article ">Figure 2
<p>Flowchart of the Global Production Efficiency Model (GLO-PEM) to estimate the Net Primary Productivity (NPP) in Northeast China.</p> Full article ">Figure 3
<p>Variation of daily solar radiation, solar radiation top of atmosphere and solar radiation in clear sky, under different typical six stations, in 2008.</p> Full article ">Figure 4
<p>Comparison of daily estimated photosynthetically active radiation (PAR) and observed PAR in Harbin and Yanji stations, in 2008.</p> Full article ">Figure 5
<p>Comparison between simulated NPP and MOD17 NPP, based on different raster pixels, in 2003 and 2008.</p> Full article ">Figure 6
<p>Inter-annual variation of NPP, from 2001 to 2019, in Northeast China.</p> Full article ">Figure 7
<p>The inter-annual variation of different forest types in Northeast China, during the period of 2001–2019.</p> Full article ">Figure 8
<p>Spatial distribution of NPP in Northeast China, during different periods.</p> Full article ">Figure 9
<p>Changing trend of NPP (<b>a</b>) and its significance (<b>b</b>) in Northeast China, during 2001–2019.</p> Full article ">Figure 10
<p>Hurst index and future trend of NPP in Northeast China, during 2001–2019.</p> Full article ">Figure 11
<p>The changing trend of forests NPP, every eight days, from 2001 to 2019, in Northeast China.</p> Full article ">Figure 12
<p>The variation characteristics of NPP with an eight-day cycle, in SOS, from 2001 to 2019.</p> Full article ">Figure 13
<p>The inter-annual variation characteristics of NPP on daily accumulations, with an eight-day cycle at the end of the vegetation growth season (EOS), from 2001 to 2019.</p> Full article ">
23 pages, 7213 KiB  
Article
Application of Multiple Geomatic Techniques for Coastline Retreat Analysis: The Case of Gerra Beach (Cantabrian Coast, Spain)
by José Juan de Sanjosé Blasco, Enrique Serrano-Cañadas, Manuel Sánchez-Fernández, Manuel Gómez-Lende and Paula Redweik
Remote Sens. 2020, 12(21), 3669; https://doi.org/10.3390/rs12213669 - 9 Nov 2020
Cited by 12 | Viewed by 3429
Abstract
The beaches of the Cantabrian coast (northern Spain) are exposed to strong winter storms that cause the coastline to recede. In this article, the coastal retreat of the Gerra beach (Cantabria) is analyzed through a diachronic study using the following different geomatic techniques: [...] Read more.
The beaches of the Cantabrian coast (northern Spain) are exposed to strong winter storms that cause the coastline to recede. In this article, the coastal retreat of the Gerra beach (Cantabria) is analyzed through a diachronic study using the following different geomatic techniques: orthophotography of the year 1956; photogrammetric flights from 2001, 2005, 2010, 2014, 2017; Light Detection and Ranging (LiDAR) survey from August 2012; Unmanned Aerial Vehicle (UAV) survey from November 2018; and terrestrial laser scanner (TLS) through two dates per year (spring and fall) from April 2012 to April 2020. With the 17 observations of TLS, differences in volume of the beach and the sea cliff are determined during the winter (November–April) and summer (May–October) periods, searching their relationship with the storms in this eight-year period (2012–2020). From the results of this investigation it can be concluded that the retreat of the base of the cliff is insignificant, but this is not the case for the top of the cliff and for the existing beaches in the Cantabrian Sea where the retreat is evident. The retreat of the cliff top line in Gerra beach, between 1956 and 2020 has shown values greater than 40 m. The retreat in other beaches of the Cantabrian Sea, in the same period, has been more than 200 m. With our measurements, investigations carried out on the retreat of the cliffs on the Atlantic coast have been reinforced, where the diversity of the cliff lithology and the aggressive action of the sea (storms) have been responsible for the active erosion on the face cliff. In addition, this research applied geomatic techniques that have appeared commercially during the period (1956–2020), such as aerial photogrammetry, TLS, LiDAR, and UAV and analyzed the results to determine the precision that could be obtained with each method for its application to similar geomorphological structures. Full article
(This article belongs to the Special Issue Remote Sensing Applications in Coastal Environment)
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>Gerra beach and San Vicente-Merón beaches system from the east of abrasion platform “rasas”.</p> Full article ">Figure 2
<p>Geological schematic of the SanVicente-Merón coastal system [<a href="#B41-remotesensing-12-03669" class="html-bibr">41</a>,<a href="#B42-remotesensing-12-03669" class="html-bibr">42</a>].</p> Full article ">Figure 3
<p>Distribution of the control points for the unmanned aerial vehicle (UAV).</p> Full article ">Figure 4
<p>Indication of the most significant landslide zones (red circles) according to terrestrial laser scanner (TLS) surveys of Gerra beach. (<b>a</b>) Digital elevation model (DEM) of October 2018; (<b>b</b>) DEM of April 2019.</p> Full article ">Figure 5
<p>(<b>a</b>) Delimitation of the coastline in the orthophoto of the American flight (1956). DEMs (height contours every 2 m) for the years (<b>b</b>) 2001; (<b>c</b>) 2005; (<b>d</b>) 2010; (<b>e</b>) 2014; (<b>f</b>) 2017.</p> Full article ">Figure 5 Cont.
<p>(<b>a</b>) Delimitation of the coastline in the orthophoto of the American flight (1956). DEMs (height contours every 2 m) for the years (<b>b</b>) 2001; (<b>c</b>) 2005; (<b>d</b>) 2010; (<b>e</b>) 2014; (<b>f</b>) 2017.</p> Full article ">Figure 6
<p>(<b>a</b>) Evolution of the cliff top between A and B. Lines of the years 1956, 2001, 2005, 2010, 2014 and 2017; (<b>b</b>) Comparison of the retreat of the cliff top in the periods 1956–2001, 2001–2010, and 2010–2017.</p> Full article ">Figure 7
<p>(<b>a</b>) Cartography light, detection, and ranging (LiDAR) (National Geographic Institute (IGN)) (August 2012); (<b>b</b>) Height contours from DEM of UAV (November 2018); (<b>c</b>) Comparison of Profile 5 from LiDAR and from UAV; (<b>d</b>) Comparison of Profile 6 from LiDAR and UAV.</p> Full article ">Figure 7 Cont.
<p>(<b>a</b>) Cartography light, detection, and ranging (LiDAR) (National Geographic Institute (IGN)) (August 2012); (<b>b</b>) Height contours from DEM of UAV (November 2018); (<b>c</b>) Comparison of Profile 5 from LiDAR and from UAV; (<b>d</b>) Comparison of Profile 6 from LiDAR and UAV.</p> Full article ">Figure 8
<p>Cliff profile evolution between the more significant campaigns. (<b>a</b>) Profile 5 shows a landslide that occurred in March 2018; (<b>b</b>) Profile 8 contains a rocky substrate.</p> Full article ">Figure 9
<p>Volumetric evolution between the spring of 2012 and the fall of 2019.</p> Full article ">Figure 9 Cont.
<p>Volumetric evolution between the spring of 2012 and the fall of 2019.</p> Full article ">Figure 10
<p>Comparison between the profiles obtained by photogrammetry and by TLS (2014). (<b>a</b>) Profile 1 (north area of the cliff); (<b>b</b>) Profile 9 (south area of the cliff).</p> Full article ">Figure 11
<p>Comparison between the profiles obtained by LiDAR and by TLS (2012). (<b>a</b>) Profile 1 (north area of the cliff); (<b>b</b>) Profile 9 (south area of the cliff).</p> Full article ">Figure 12
<p>Comparison between the profiles obtained by UAV and by TLS (2018). (<b>a</b>) Profile 1 (north area of the cliff); (<b>b</b>) Profile 9 (south area of the cliff).</p> Full article ">
23 pages, 16849 KiB  
Article
Spatiotemporal Changes in 3D Building Density with LiDAR and GEOBIA: A City-Level Analysis
by Karolina Zięba-Kulawik, Konrad Skoczylas, Ahmed Mustafa, Piotr Wężyk, Philippe Gerber, Jacques Teller and Hichem Omrani
Remote Sens. 2020, 12(21), 3668; https://doi.org/10.3390/rs12213668 - 9 Nov 2020
Cited by 6 | Viewed by 4864
Abstract
Understanding how, where, and when a city is expanding can inform better ways to make our cities more resilient, sustainable, and equitable. This paper explores urban volumetry using the Building 3D Density Index (B3DI) in 2001, 2010, 2019, and quantifies changes in the [...] Read more.
Understanding how, where, and when a city is expanding can inform better ways to make our cities more resilient, sustainable, and equitable. This paper explores urban volumetry using the Building 3D Density Index (B3DI) in 2001, 2010, 2019, and quantifies changes in the volume of buildings and urban expansion in Luxembourg City over the last two decades. For this purpose, we use airborne laser scanning (ALS) point cloud (2019) and geographic object-based image analysis (GEOBIA) of aerial orthophotos (2001, 2010) to extract 3D models, footprints of buildings and calculate the volume of individual buildings and B3DI in the frame of a 100 × 100 m grid, at the level of parcels, districts, and city scale. Findings indicate that the B3DI has notably increased in the past 20 years from 0.77 m3/m2 (2001) to 0.9 m3/m2 (2010) to 1.09 m3/m2 (2019). Further, the increase in the volume of buildings between 2001–2019 was +16 million m3. The general trend of changes in the cubic capacity of buildings per resident shows a decrease from 522 m3/resident in 2001, to 460 m3/resident in 2019, which, with the simultaneous appearance of new buildings and fast population growth, represents the dynamic development of the city. Full article
(This article belongs to the Special Issue Understanding Urban Systems Using Remote Sensing)
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>(<b>a</b>) Location of the study area. Left top: location of Luxembourg in Europe, left bottom: Grand Duchy of Luxembourg and Luxembourg City borders; (<b>b</b>) Orthophotomap of Luxembourg City at municipality level (2019).</p> Full article ">Figure 2
<p>Workflow of the performed analyses (ALS—airborne laser scanning; nDSM—normalized digital surface model; LoD2—level of details 2; GEOBIA—geographic object-based image analysis; B3DI—Building 3D Density Index).</p> Full article ">Figure 3
<p>nDSM of building, vegetation and cranes on newly constructed buildings (<b>a</b>); nDSM of buildings only (<b>b</b>).</p> Full article ">Figure 4
<p>Classified ALS point cloud (2019); class (2) ground in pink color; class (6) buildings—in red; and vegetation—in green color.</p> Full article ">Figure 5
<p>Stages of the GEOBIA building segmentation and classification in eCognition software (Trimble Geospatial); <b>1</b>—input data: CIR orthophoto; <b>2</b>—multiresolution segmentation; <b>3</b>—spectral difference segmentation; <b>4</b>—classification: buildings class.</p> Full article ">Figure 6
<p>Flowchart of the GEOBIA classification process.</p> Full article ">Figure 7
<p>GEOBIA classification of buildings based on (<b>a</b>) RGB aerial orthophoto maps in 2001 and (<b>b</b>) CIR orthophoto maps in 2010.</p> Full article ">Figure 8
<p>New and reconstructed buildings between (<b>a</b>) 2001–2010 and (<b>b</b>) 2010–2019 in Luxembourg City.</p> Full article ">Figure 9
<p>A 3D model (LoD2) of the buildings in Luxembourg City based on ALS point cloud (2019); (<b>a</b>) business district at the northeast (Kirchberg) with public buildings; (<b>b</b>) southern part of the city near the central station.</p> Full article ">Figure 10
<p>Residential and non-residential buildings based on ALS point cloud (2019) and cadastral data in Luxembourg City (<b>a</b>); (<b>b</b>) Ville Haute as a residential district (at the right top, b-1), and Gasperich a non-residential part (at the right bottom, b-2).</p> Full article ">Figure 11
<p>Luxembourg City maps (2019) of (<b>a</b>) single buildings volume and (<b>b</b>) rasterized map (100 m GSD).</p> Full article ">Figure 12
<p>Volume of buildings in a grid (100 × 100 m) in 3D view (Luxembourg City, 2019).</p> Full article ">Figure 13
<p>Building 3D Density Index per plot (<b>a</b>) and per district in Luxembourg City in 2019 (<b>b</b>).</p> Full article ">Figure 14
<p>Change of buildings volumetry between 2001 and 2019 in Luxembourg City in a grid (100 × 100 m).</p> Full article ">Figure 15
<p>Changes of the B3DI index in Luxembourg City districts over the last 20 years; differences between 2001–2010 (<b>a</b>); 2010–2019 (<b>b</b>); between 2001–2019 (<b>c</b>).</p> Full article ">Figure 16
<p>The B3DI value of Luxembourg City districts of in years 2001, 2010, and 2019.</p> Full article ">Figure 17
<p>Buildings 3D Density Index in European cities (Source: based on data in publication [<a href="#B44-remotesensing-12-03668" class="html-bibr">44</a>,<a href="#B45-remotesensing-12-03668" class="html-bibr">45</a>,<a href="#B54-remotesensing-12-03668" class="html-bibr">54</a>]).</p> Full article ">Figure 18
<p>Volume of buildings per resident in European cities (Source: based on data in publication [<a href="#B44-remotesensing-12-03668" class="html-bibr">44</a>,<a href="#B45-remotesensing-12-03668" class="html-bibr">45</a>,<a href="#B54-remotesensing-12-03668" class="html-bibr">54</a>]).</p> Full article ">
24 pages, 11348 KiB  
Article
Application of the DIC Technique to Remote Control of the Hydraulic Load System
by Radosław Jasiński, Krzysztof Stebel and Jarosław Domin
Remote Sens. 2020, 12(21), 3667; https://doi.org/10.3390/rs12213667 - 9 Nov 2020
Cited by 4 | Viewed by 2948
Abstract
Displacements or deformations of materials or structures are measured with linear variable differential transducers (LVDT), fibre optic sensors, laser sensors, and confocal sensor systems, while strains are measured with electro-resistant tensometers or wire strain gauges. Measurements significantly limited to a point or a [...] Read more.
Displacements or deformations of materials or structures are measured with linear variable differential transducers (LVDT), fibre optic sensors, laser sensors, and confocal sensor systems, while strains are measured with electro-resistant tensometers or wire strain gauges. Measurements significantly limited to a point or a small area are the obvious disadvantage of these measurements. Such disadvantages are eliminated by performing measurements with optical techniques, such as digital image correlation (DIC) or electronic speckle pattern interferometry (ESPI). Many devices applied to optical measurements only record test results and do not cooperate with the system that exerts and controls load. This paper describes the procedure for preparing a test stand involving the Digital Image Correlation system ARAMIS 6M for remote-controlled loading. The existing hydraulic power pack (ZWICK-ROELL) was adapted by installing the modern NI cRIO-9022 controller operating under its own software developed within the LABVIEW system. The application of the DIC techniques to directly control load on the real structure is the unquestionable innovation of the described solution. This led to the elimination of errors caused by the test stand susceptibility and more precise relations between load and displacements/strains which have not been possible using the previous solutions. This project is a synergistic and successful combination of civil engineering, computer science, automatic control engineering and electrical engineering that provides a new solution class. The prepared stand was tested using two two-span, statically non-determinable reinforced concrete beams loaded under different conditions (force or displacement). The method of load application was demonstrated to affect the redistribution of bending moments. The conducted tests confirmed the suitability of the applied technique for the remote controlling and recording of test results. Regardless of the load control method (with force or displacement), convergent results were obtained for the redistribution of bending moments. Force-controlled rotation of the beam section over the support was over 50% greater than rotation of the second beam controlled with an increase in the displacement. Full article
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>Block diagram of test stand.</p> Full article ">Figure 2
<p>Graphical interpretation of deformations for selected scanning area in a 2D system of coordinates [<a href="#B37-remotesensing-12-03667" class="html-bibr">37</a>]: 1—scanning area, 2—scanning area after deformation, 3—pixel subimages of the structure.</p> Full article ">Figure 3
<p>ARAMIS 6M system and its main components: (<b>a</b>) process diagram [<a href="#B5-remotesensing-12-03667" class="html-bibr">5</a>], (<b>b</b>) main measuring module—two cameras and light source, (<b>c</b>) calibration cross and plate1—test element, 2—calibration cross, 3—set of two cameras; 4, 5—left and right camera image prior to loading, divided into pixels and facets, 6,7—deformed facets recorded by left and right camera after loading.</p> Full article ">Figure 4
<p>Scheme of hydraulic subsystem.</p> Full article ">Figure 5
<p>Block diagram of electrical subsystem.</p> Full article ">Figure 6
<p>Schematic diagram of displacement or load control system.</p> Full article ">Figure 7
<p>Filtering pressure signal in the window of five specimens for 100 ms.</p> Full article ">Figure 8
<p>Filtering of pressure signal within the window of 50 specimens for 1000 ms.</p> Full article ">Figure 9
<p>Static description of control valve: (<b>a</b>) original, before negative feedback occurs, (<b>b</b>) after negative feedback and linearization.</p> Full article ">Figure 10
<p>Geometry and reinforcement of beams.</p> Full article ">Figure 11
<p>Reinforced concrete elements at the test stand: (<b>a</b>) from the side of optical measurements, (<b>b</b>) from the side of making observations of cracks; 1—test element, 2—hydraulic actuator, 3—electro-resistant dynamometer, 4—steel beams for load distribution; 5—supports of the test stand.</p> Full article ">Figure 12
<p>A view of elements of the test stand for exerting load: (<b>a</b>) a beam with cameras from the ARAMIS system, (<b>b</b>) the ARAMIS system and its other elements; 1—test element, 2—hydraulic actuator, 3—the ARAMIS system, 4—hydraulic power pack.</p> Full article ">Figure 13
<p>Changes in bending moments in the span and over the support in statically non-determinable beam.</p> Full article ">Figure 14
<p>Testing programme for: (<b>a</b>) Beam No. 1, (<b>b</b>) Beam No. 2; 1—response of structure.</p> Full article ">Figure 15
<p>Testing programme: (<b>a</b>) general view of the user interface, (<b>b</b>) data on loading/displacement over the function of time.</p> Full article ">Figure 16
<p>Comparison of results from structure measurements using the LVDT transducer and the DIC technique: (<b>a</b>) hydraulic actuators and steel cross beams were measured using the DIC technique, (<b>b</b>) actual results of displacements, (<b>c</b>) displacement results containing the correction factor.</p> Full article ">Figure 17
<p>Changes in imposed load and recorded response of structure in a function of time: (<b>a</b>) Beam No. 1, (<b>b</b>) Beam No. 2.</p> Full article ">Figure 18
<p>Beams at the time of failure: (<b>a</b>) Beam No. 1, (<b>b</b>) Beam No. 2.</p> Full article ">Figure 19
<p>Deformations of beam sections at the time of cracking: (<b>a</b>) at the support section in Beam No. 1, (<b>b</b>) at the span section in Beam No. 1, (<b>c</b>) at the support section in Beam No. 2, (<b>d</b>) at the span section in Beam No. 2.</p> Full article ">Figure 20
<p>Rotations of bottom sections at the time of failure: (<b>a</b>) Beam No. 1, (<b>b</b>) Beam No. 2.</p> Full article ">Figure 21
<p>Changes in support responses: (<b>a</b>) Beam No. 1, (<b>b</b>) Beam No. 2.</p> Full article ">Figure 22
<p>Changes in bending moments of: (<b>a</b>) Beam No. 1, (<b>b</b>) Beam No. 2.</p> Full article ">Figure 23
<p>Comparison of measured values for relative rotations of sections and allowable values specified in the standard EN 1992–1–1:2010 [<a href="#B57-remotesensing-12-03667" class="html-bibr">57</a>]; 1—allowed rotation, 2—unacceptable allowed rotation.</p> Full article ">
23 pages, 8355 KiB  
Article
A New Approach of Ensemble Learning Technique to Resolve the Uncertainties of Paddy Area through Image Classification
by Tsu Chiang Lei, Shiuan Wan, Shih-Chieh Wu and Hsin-Ping Wang
Remote Sens. 2020, 12(21), 3666; https://doi.org/10.3390/rs12213666 - 9 Nov 2020
Cited by 6 | Viewed by 2785
Abstract
Remote sensing technology has rendered lots of information in agriculture. It has usually been used to monitor paddy growing ecosystems in the past few decades. However, there are uncertainties in data fusion techniques which can be resolved in image classification on paddy rice. [...] Read more.
Remote sensing technology has rendered lots of information in agriculture. It has usually been used to monitor paddy growing ecosystems in the past few decades. However, there are uncertainties in data fusion techniques which can be resolved in image classification on paddy rice. In this study, a series of learning concepts integrated by a probability progress Fuzzy Dempster-Shafer (FDS) analysis is presented to upgrade various models and different types of image data which is the goal of this study. More specifically, the study utilized the FDS to generate a series of probability models in the classification of the system. In addition, Logistic Regression (LR), Support Vector Machine (SVM), and Neural Network (NN) approaches are employed into the developed FDS system. Furthermore, two different image types are Satellite Image and Aerial Photo used as the analysis material. The overall classification accuracy has been improved to 97.27%, and the kappa value is 0.93. The overall accuracy of the paddy field image classification for a multi-period of mid-scale satellite images is between 85% and 90%. The overall accuracy of the classification using multi-spectral numerical aerial photos can be between 91% and 95%. The FDS improves the accuracy of the above image classification results. Full article
(This article belongs to the Special Issue Multi-Sensor Data Fusion and Analysis of Multi-Temporal Remote Sensed Imagery)
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>The general model of the data fusion technology [<a href="#B23-remotesensing-12-03666" class="html-bibr">23</a>].</p> Full article ">Figure 2
<p>The research area (Meinong District, Kaohsiung City).</p> Full article ">Figure 3
<p>Research material (<b>a</b>) Satellite image of photo taking at 1 February 2015 (the rice transplanting stage) (<b>b</b>) Satellite image of photo taking at 2 April 2015 (the rice tillering stage). (<b>c</b>) Aerial photo image—DMC of photo taking at 2 April 2015.</p> Full article ">Figure 3 Cont.
<p>Research material (<b>a</b>) Satellite image of photo taking at 1 February 2015 (the rice transplanting stage) (<b>b</b>) Satellite image of photo taking at 2 April 2015 (the rice tillering stage). (<b>c</b>) Aerial photo image—DMC of photo taking at 2 April 2015.</p> Full article ">Figure 4
<p>Data fusion steps of classification analysis through multi-scale images.</p> Full article ">Figure 5
<p>Solutions for the decision procedure of the uncertainty patches. (<b>a</b>) Probability fuzzification diagram. (No. of Patch is 42,643). (<b>b</b>) Probability de-fuzzification diagram. (<b>c</b>) Decision results of the uncertainty patches.</p> Full article ">Figure 6
<p>Inconsistent classification results of (<b>a</b>) satellite images, (<b>b</b>) aerial images, (<b>c</b>) inconsistent classification of six classification results (considering satellite image and aerial image for LR, SVM, ANN).</p> Full article ">Figure 6 Cont.
<p>Inconsistent classification results of (<b>a</b>) satellite images, (<b>b</b>) aerial images, (<b>c</b>) inconsistent classification of six classification results (considering satellite image and aerial image for LR, SVM, ANN).</p> Full article ">Figure 7
<p>Results of FDS. (<b>a</b>) The consistent patches of six classification results (considering satellite image and aerial image with LR, SVM, ANN). (<b>b</b>) The inconsistent patches were fixed by the FDS program. (<b>c</b>) The inconsistent patches were incorrect from FDS program.</p> Full article ">Figure 7 Cont.
<p>Results of FDS. (<b>a</b>) The consistent patches of six classification results (considering satellite image and aerial image with LR, SVM, ANN). (<b>b</b>) The inconsistent patches were fixed by the FDS program. (<b>c</b>) The inconsistent patches were incorrect from FDS program.</p> Full article ">Figure 8
<p>The integration results of FDS outputs for the Data Fusion concept of <a href="#remotesensing-12-03666-f004" class="html-fig">Figure 4</a>.</p> Full article ">Figure 9
<p>Results of image data fusion performed through three classifiers using two data sources. (<b>a</b>) Six different results comparison to FDS. (<b>b</b>) Detail selection of region to observe uncertainty patches</p> Full article ">Figure 9 Cont.
<p>Results of image data fusion performed through three classifiers using two data sources. (<b>a</b>) Six different results comparison to FDS. (<b>b</b>) Detail selection of region to observe uncertainty patches</p> Full article ">
7 pages, 211 KiB  
Editorial
Special Issue “Hyperspectral Remote Sensing of Agriculture and Vegetation”
by Simone Pascucci, Stefano Pignatti, Raffaele Casa, Roshanak Darvishzadeh and Wenjiang Huang
Remote Sens. 2020, 12(21), 3665; https://doi.org/10.3390/rs12213665 - 9 Nov 2020
Cited by 27 | Viewed by 6324
Abstract
The advent of up-to-date hyperspectral technologies, and their increasing performance both spectrally and spatially, allows for new and exciting studies and practical applications in agriculture (soils and crops) and vegetation mapping and monitoring atregional (satellite platforms) andwithin-field (airplanes, drones and ground-based platforms) scales. [...] Read more.
The advent of up-to-date hyperspectral technologies, and their increasing performance both spectrally and spatially, allows for new and exciting studies and practical applications in agriculture (soils and crops) and vegetation mapping and monitoring atregional (satellite platforms) andwithin-field (airplanes, drones and ground-based platforms) scales. Within this context, the special issue has included eleven international research studies using different hyperspectral datasets (from the Visible to the Shortwave Infrared spectral region) for agricultural soil, crop and vegetation modelling, mapping, and monitoring. Different classification methods (Support Vector Machine, Random Forest, Artificial Neural Network, Decision Tree) and crop canopy/leaf biophysical parameters (e.g., chlorophyll content) estimation methods (partial least squares and multiple linear regressions) have been evaluated. Further, drone-based hyperspectral mapping by combining bidirectional reflectance distribution function (BRDF) model for multi-angle remote sensing and object-oriented classification methods are also examined. A review article on the recent advances of hyperspectral imaging technology and applications in agriculture is also included in this issue. The special issue is intended to help researchers and farmers involved in precision agriculture technology and practices to a better comprehension of strengths and limitations of the application of hyperspectral measurements for agriculture and vegetation monitoring. The studies published herein can be used by the agriculture and vegetation research and management communities to improve the characterization and evaluation of biophysical variables and processes, as well as for a more accurate prediction of plant nutrient using existing and forthcoming hyperspectral remote sensing technologies. Full article
(This article belongs to the Special Issue Hyperspectral Remote Sensing of Agriculture and Vegetation)
22 pages, 9671 KiB  
Article
Deformations Prior to the Brumadinho Dam Collapse Revealed by Sentinel-1 InSAR Data Using SBAS and PSI Techniques
by Fábio F. Gama, José C. Mura, Waldir R. Paradella and Cleber G. de Oliveira
Remote Sens. 2020, 12(21), 3664; https://doi.org/10.3390/rs12213664 - 9 Nov 2020
Cited by 29 | Viewed by 6749
Abstract
Differential Interferometric SAR (DInSAR) has been used to monitor surface deformations in open pit mines and tailings dams. In this paper, ground deformations have been detected on the area of tailings Dam-I at the Córrego do Feijão Mine (Brumadinho, Brazil) before its catastrophic [...] Read more.
Differential Interferometric SAR (DInSAR) has been used to monitor surface deformations in open pit mines and tailings dams. In this paper, ground deformations have been detected on the area of tailings Dam-I at the Córrego do Feijão Mine (Brumadinho, Brazil) before its catastrophic failure occurred on 25 January 2019. Two techniques optimized for different scattering models, SBAS (Small BAseline Subset) and PSI (Persistent Scatterer Interferometry), were used to perform the analysis based on 26 Sentinel-1B images in Interferometric Wide Swath (IW) mode, which were acquired on descending orbits from 03 March 2018 to 22 January 2019. A WorldDEM Digital Surface Model (DSM) product was used to remove the topographic phase component. The results provided by both techniques showed a synoptic and informative view of the deformation process affecting the study area, with the detection of persistent trends of deformation on the crest, middle, and bottom sectors of the dam face until its collapse, as well as the settlements on the tailings. It is worth noting the detection of an acceleration in the displacement time-series for a short period near the failure. The maximum accumulated displacements detected along the downstream slope face were −39 mm (SBAS) and −48 mm (PSI). It is reasonable to consider that Sentinel-1 would provide decision makers with complementary motion information to the in situ monitoring system for risk assessment and for a better understanding of the ongoing instability phenomena affecting the tailings dam. Full article
(This article belongs to the Special Issue Recent Advances in Remote Sensing Applied to Geohazards, Vulnerability and Risk Studies)
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>(<b>a</b>) Study area location in Minas Gerais state/Brazil; (<b>b</b>) Location of Paraopeba Iron Complex in the Brumadinho municipality; (<b>c</b>) Paraopeba Iron Complex showing tailings Dam I and hydric Dam VI, ancillary mining structures such as ore treatment installation (ITM), rail network, access roads, and a Sentinel-1B descending orbit ground track.</p> Full article ">Figure 2
<p>Pleiades satellite images taken just before and after the tragedy, showing the water from Dam-VI, the tailings from Dam-I, and the Corrégo do Feijão stream.</p> Full article ">Figure 3
<p>Interferometric pairs (solid lines) chose for SBAS (Small BAseline Subset) analysis; green dots correspond to the S1-B scenes and yellow dot corresponds to the reference scene for the co-registration (31 August 2018).</p> Full article ">Figure 4
<p>Interferometric pairs selected for PSI (Persistent Scatterer Interferometry) analysis; green dots correspond to the S1-B scenes and yellow dot corresponds to the reference scene (31 August 2018).</p> Full article ">Figure 5
<p>Map of the displacement rate (<b>a</b>) and the standard deviation (<b>b</b>) from the SBAS analysis; white circles represent the analyzed points in the dam reservoir (<span class="html-italic">Rs</span>), crest (<span class="html-italic">Ts</span>) and bottom face (<span class="html-italic">Bs</span>).</p> Full article ">Figure 6
<p>3D view of Dam-I showing the accumulated displacement from SBAS processing analysis represented by colored dots.</p> Full article ">Figure 7
<p>Time-series of SBAS line of sight (LoS) displacements for selected measured points (MPs) located at three selected points of the dam, as presented in <a href="#remotesensing-12-03664-f005" class="html-fig">Figure 5</a>a, showing the highest cumulative displacements for the entire period of monitoring and the linear regression of the data (letters A, B, and C are the displacement trends as discussed in the text).</p> Full article ">Figure 8
<p>Maps of the displacement rate (<b>a</b>) and the standard deviation (<b>b</b>) from the PSI analysis; white circles represent the analyzed points in the dam reservoir (<span class="html-italic">Rp</span>), crest (<span class="html-italic">Tp</span>) and bottom face (<span class="html-italic">Bp</span>).</p> Full article ">Figure 9
<p>View in 3D of Dam-I showing the accumulated displacement from PSI analysis represented by colored dots.</p> Full article ">Figure 10
<p>Time-series PSI LoS displacements for three selected points located at three sectors of Dam-I as presented in <a href="#remotesensing-12-03664-f008" class="html-fig">Figure 8</a>a showing the highest cumulative displacements for the entire period of monitoring and the linear regression of the data (letters A, B, and C are the deformation trends as discussed in text).</p> Full article ">Figure 11
<p>SBAS average displacement (<b>a</b>) and inverse of the velocity graph with linear regression (<b>b</b>).</p> Full article ">Figure 12
<p>PSI average displacement (<b>a</b>) and inverse of the velocity graph with linear regression (<b>b</b>).</p> Full article ">Figure 13
<p>SBAS relative frequency distribution of the errors (<b>a</b>) and R<sup>2</sup> coefficient from the inverse velocity predictions (<b>b</b>).</p> Full article ">Figure 14
<p>PSI relative frequency distribution of the errors (<b>a</b>) and R<sup>2</sup> coefficient from the inverse velocity predictions (<b>b</b>).</p> Full article ">Figure 15
<p>SBAS and PSI results combined and the locations of Bs and Bp points (black circles), which are the SBAS and PSI points with the highest accumulated deformation values on the dam bottom; SBAS MPs are represented by square symbols, and PSI MPs are represented by round symbols.</p> Full article ">Figure 16
<p>Time-series of LoS displacement for two selected points located at the bottom of the dam showing the maximum cumulative displacements for the entire monitoring period and the accumulated pluviometric data.</p> Full article ">
15 pages, 7697 KiB  
Article
Automatic High-Resolution Land Cover Production in Madagascar Using Sentinel-2 Time Series, Tile-Based Image Classification and Google Earth Engine
by Meinan Zhang, Huabing Huang, Zhichao Li, Kwame Oppong Hackman, Chong Liu, Roger Lala Andriamiarisoa, Tahiry Ny Aina Nomenjanahary Raherivelo, Yanxia Li and Peng Gong
Remote Sens. 2020, 12(21), 3663; https://doi.org/10.3390/rs12213663 - 8 Nov 2020
Cited by 42 | Viewed by 7616
Abstract
Madagascar, one of Earth’s biodiversity hotpots, is characterized by heterogeneous landscapes and huge land cover change. To date, fine, reliable and timely land cover information is scarce in Madagascar. However, mapping high-resolution land cover map in the tropics has been challenging due to [...] Read more.
Madagascar, one of Earth’s biodiversity hotpots, is characterized by heterogeneous landscapes and huge land cover change. To date, fine, reliable and timely land cover information is scarce in Madagascar. However, mapping high-resolution land cover map in the tropics has been challenging due to limitations associated with heterogeneous landscapes, the volume of satellite data used, and the design of methodology. In this study, we proposed an automatic approach in which the tile-based model was used on each tile (defining an extent of 1° × 1° as a tile) for mapping land cover in Madagascar. We combined spectral-temporal, textural and topographical features derived from all available Sentinel-2 observations (i.e., 11,083 images) on Google Earth Engine (GEE). We generated a 10-m land cover map for Madagascar, with an overall accuracy of 89.2% based on independent validation samples obtained from a field survey and visual interpretation of very high-resolution (0.5–5 m) images. Compared with the conventional approach (i.e., the overall model used in the entire study area), our method enables reduce the misclassifications between several land cover types, including impervious land, grassland and wetland. The proposed approach demonstrates a great potential for mapping land cover in other tropical or subtropical regions. Full article
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>The workflow of the tile-based model used in this study for mapping highly heterogeneous land cover in Madagascar. One tile refers to an extent of 1° × 1°.</p> Full article ">Figure 2
<p>Schematic representation of the hexagon random sampling process in this study. Four steps: (<b>a</b>) generation of hexagons covering the whole study area; (<b>b</b>) visual interpretation of each hexagon overlaied on high-resolution Google Earth imagery; (<b>c</b>) selection of the location of land cover types; and (<b>d</b>) obtention of the sample set of the study area.</p> Full article ">Figure 3
<p>Overview of the 10-m circa 2018 land cover map of Madagascar (i.e., the MDG LC-10 map) derived from the Sentiel-2 dataset. (<b>a</b>) The proportions of the eight major land cover classes over the entire island. The zoomed in windows show the details ranging from the landscape view (<b>b</b>), the urban structure (<b>c</b>), and the fine grain land cover patterns (<b>d</b>).</p> Full article ">Figure 4
<p>Visual comparison of three zoomed areas (<b>a</b>–<b>c</b>) among the MDG LC-10 map, two high resolution land cover maps (i.e., the CCI Africa LC-20 map and the FROM-GLC10 map) and very high resolution imageries from Google Earth.</p> Full article ">Figure 5
<p>Statistical comparison of producer accuracies (PAs) of the available high-resolution land cover maps for Madagascar and the map produced in this study.</p> Full article ">Figure 6
<p>Examples of comparison of the classification performance using our proposed method (i.e., the tile-based model) (<b>c</b>,<b>f</b>) in this study and the conventional method (i.e., the overall model) (<b>b</b>,<b>e</b>). (<b>a</b>,<b>d</b>) are zoomed high-resolution images from Google Earth. Region 1 exhibits an improvement in the misclassification between cropland and wetland. Region 2 shows that the issue of impervious areas misclassification is improved.</p> Full article ">Figure 7
<p>Statistical comparison of producer accuracies (PAs) for the proposed method (i.e., the one tile-based model is applied to one tile) and the conventional method (i.e., one model is applied to the entire study area). The blue and orange bars represent the PAs of eight land cover classes, and the green bars indicate that the differences in PAs.</p> Full article ">Figure 8
<p>Examples of comparisons of the classification performances using Sentinel-2 images with cloud cover below 10% (i.e., n = 4900 images, <b>b</b>,<b>e</b>) and using all available Sentinel-2 images (i.e., n = 11,083 images, <b>c</b>,<b>f</b>) over the entire island. (<b>a</b>,<b>d</b>) were derived from high-resolution Google earth imagery.</p> Full article ">
21 pages, 8039 KiB  
Article
Analysing Urban Development Patterns in a Conflict Zone: A Case Study of Kabul
by Vineet Chaturvedi, Monika Kuffer and Divyani Kohli
Remote Sens. 2020, 12(21), 3662; https://doi.org/10.3390/rs12213662 - 8 Nov 2020
Cited by 15 | Viewed by 5293
Abstract
A large part of the population in low-income countries (LICs) lives in fragile and conflict-affected states. Many cities in these states show high growth dynamics, but little is known about the relation of conflicts and urban growth. In Afghanistan, the Taliban regime, which [...] Read more.
A large part of the population in low-income countries (LICs) lives in fragile and conflict-affected states. Many cities in these states show high growth dynamics, but little is known about the relation of conflicts and urban growth. In Afghanistan, the Taliban regime, which lasted from 1996 to 2001, caused large scale displacement of the population. People from Afghanistan migrated to neighboring countries like Iran and Pakistan, and all developments came to a halt. After the US invasion in October 2001, all the major cities in Afghanistan experienced significant population growth, in particular, driven by the influx of internally displaced persons. Maximum pressure of this influx was felt by the capital city, Kabul. This rapid urbanization, combined with very limited capacity of local authorities to deal with this growth, led to unplanned urbanization and challenges for urban planning and management. This study analyses the patterns of growth between 2001 and 2017, and the factors influencing the growth in the city of Kabul with the help of high-resolution Earth Observation-based data (EO) and spatial logistic regression modelling. We analyze settlement patterns by extracting image features from high-resolution images (aerial photographs of 2017) and terrain features as input to a random forest classifier. The urban growth is analyzed using an available built-up map (extracted from IKONOS images for the year 2001). Results indicate that unplanned settlements have grown 4.5 times during this period, whereas planned settlements have grown only 1.25 times. The unplanned settlements expanded mostly towards the west and north west parts of the city, and the growth of planned settlements happened mainly in the central and eastern parts of the city. Population density and the locations of military bases are the most important factors that influence the growth, of both planned and unplanned settlements. The growth of unplanned settlement occurs predominantly in areas of steeper slopes on the hillside, while planned settlements are on gentle slopes and closer to the institutional areas (central and eastern parts of the city). We conclude that security and availability of infrastructure were the main drivers of growth for planned settlements, whereas unplanned growth, mainly on hillsides, was driven by the availability of land with poor infrastructure. Full article
(This article belongs to the Special Issue Remote Sensing-Based Urban Planning Indicators)
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>Population from the period 1962 to 2010 (Source: Draft Kabul City Master Plan, RECS International Inc. Yachiyo Engineering Co., Ltd.).</p> Full article ">Figure 2
<p>Map showing the geographical location of the city of Kabul.</p> Full article ">Figure 3
<p>Global Human Settlement Layer of 2014 shown on an OpenStreetMap of Kabul City.</p> Full article ">Figure 4
<p>Global Human Settlement Layer (GHSL) of 2014 overlaid on ESRI world imagery showing settlements not captured by Global Human Settlement Layer (Data Source: <a href="https://ghsl.jrc.ec.europa.eu/datasets.php" target="_blank">https://ghsl.jrc.ec.europa.eu/datasets.php</a>).</p> Full article ">Figure 5
<p>Flowchart showing the methodology.</p> Full article ">Figure 6
<p>Aerial photographs for the year 2017 showing the location of (<b>a</b>) training samples and (<b>b</b>) reference samples.</p> Full article ">Figure 7
<p>Population distribution map of Kabul city.</p> Full article ">Figure 8
<p>Permutation Importance.</p> Full article ">Figure 9
<p>Gini Decrease.</p> Full article ">Figure 10
<p>Result of the classification of aerial photograph 2017.</p> Full article ">Figure 11
<p>(<b>a</b>) Built-up extract from IKONOS 2001, (<b>b</b>) change in planned and unplanned from 2001 to 2017.</p> Full article ">Figure 12
<p>Slope map of Kabul.</p> Full article ">
21 pages, 6389 KiB  
Article
Chlorophyll-a Variability during Upwelling Events in the South-Eastern Baltic Sea and in the Curonian Lagoon from Satellite Observations
by Toma Dabuleviciene, Diana Vaiciute and Igor E. Kozlov
Remote Sens. 2020, 12(21), 3661; https://doi.org/10.3390/rs12213661 - 8 Nov 2020
Cited by 13 | Viewed by 3678
Abstract
Based on the analysis of multispectral satellite data, this work demonstrates the influence of coastal upwelling on the variability of chlorophyll-a (Chl-a) concentration in the south-eastern Baltic (SEB) Sea and in the Curonian Lagoon. The analysis of sea surface temperature (SST) data acquired [...] Read more.
Based on the analysis of multispectral satellite data, this work demonstrates the influence of coastal upwelling on the variability of chlorophyll-a (Chl-a) concentration in the south-eastern Baltic (SEB) Sea and in the Curonian Lagoon. The analysis of sea surface temperature (SST) data acquired by the Moderate Resolution Imaging Spectroradiometer (MODIS) onboard Aqua/Terra satellites, together with Chl-a maps from Medium Resolution Imaging Spectrometer (MERIS) onboard Envisat, shows a significant decrease of up to 40–50% in Chl-a concentration in the upwelling zone. This results from the offshore Ekman transport of more productive surface waters, which are replaced by cold and less-productive waters from deeper layers. Due to an active interaction between the Baltic Sea and the Curonian Lagoon which are connected through the Klaipeda Strait, coastal upwelling in the SEB also influences the hydrobiological conditions of the adjacent lagoon. During upwelling inflows, SST drops by approximately 2–8 °C, while Chl-a concentration becomes 2–4 times lower than in pre-upwelling conditions. The joint analysis of remotely sensed Chl-a and SST data reveals that the upwelling-driven reduction in Chl-a concentration leads to the temporary improvement of water quality in terms of Chl-a in the coastal zone and in the hyper-eutrophic Curonian Lagoon. This study demonstrates the benefits of multi-spectral satellite data for upscaling coastal processes and monitoring the environmental status of the Baltic Sea and its largest estuarine lagoon. Full article
(This article belongs to the Special Issue Baltic Sea Remote Sensing)
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>Map of the study site indicating the two subareas: the Curonian Lagoon and the Lithuanian coastal area (denoted as Lithuanian EEZ). The location of the Klaipeda coastal monitoring station is denoted as “Klaipeda HMS”, and the location of the Smalininkai gauging station as “Smalininkai GS”.</p> Full article ">Figure 2
<p>An example of data selection in the (<b>a</b>) MODIS SST and (<b>b</b>) MERIS Chl-a maps acquired, in 25 July 2008.</p> Full article ">Figure 3
<p>Median Chl-a concentration in the upwelling and reference zones (* statistically insignificant differences are indicated with asterisks).</p> Full article ">Figure 4
<p>Wind speed and direction in Klaipeda HMS. ‘F’ symbol in the left corners and full black circles indicate upwelling-favourable northerly winds. The red bar indicates the time interval of the pre-upwelling phase, the green bar the upwelling active phase, and the blue bar the upwelling relaxation phase.</p> Full article ">Figure 5
<p>The evolution of the SST and Chl-a concentrations during the upwelling event of July 2008.</p> Full article ">Figure 6
<p>Variations of Chl-a and SST values along horizontal profiles near Palanga and the Curonian Spit (denoted as “Spit Coast”) during the upwelling event of July–August 2008. The locations of the cross-frontal transects near Palanga and the Curonian Spit are indicated in <a href="#remotesensing-12-03661-f005" class="html-fig">Figure 5</a>. To better illustrate Chl-a variations due to the Curonian Lagoon plume, the Chl-a concentration axes limits in July 31 and August 01 differ from those of the other dates.</p> Full article ">Figure 7
<p>MODIS SST and MERIS Chl-a maps depicting upwelling induced changes of SST and Chl-a concentrations on the coast of the south-eastern Baltic Sea and in the Curonian Lagoon. To better illustrate Chl-a and SST changes, colour bars differ between different dates.</p> Full article ">Figure 8
<p>Satellite-derived Chl-a concentration and SST in the Curonian Lagoon: Chl-a concentration and SST in the upwelling inflow area is denoted as UPW<sub>avg</sub> and in the reference area as REF<sub>avg</sub>.</p> Full article ">Figure 9
<p>Upwelling-associated water quality (WQ) changes in <b>(a)</b> the SE Baltic Sea coast and <b>(b)</b> the Curonian Lagoon based on Chl-a concentration. *Excellent stands for “reference conditions”.</p> Full article ">
22 pages, 1979 KiB  
Article
Fuel Type Classification Using Airborne Laser Scanning and Sentinel 2 Data in Mediterranean Forest Affected by Wildfires
by Darío Domingo, Juan de la Riva, María Teresa Lamelas, Alberto García-Martín, Paloma Ibarra, Maite Echeverría and Raúl Hoffrén
Remote Sens. 2020, 12(21), 3660; https://doi.org/10.3390/rs12213660 - 8 Nov 2020
Cited by 33 | Viewed by 4600
Abstract
Mediterranean forests are recurrently affected by fire. The recurrence of fire in such environments and the number and severity of previous fire events are directly related to fire risk. Fuel type classification is crucial for estimating ignition and fire propagation for sustainable forest [...] Read more.
Mediterranean forests are recurrently affected by fire. The recurrence of fire in such environments and the number and severity of previous fire events are directly related to fire risk. Fuel type classification is crucial for estimating ignition and fire propagation for sustainable forest management of these wildfire prone environments. The aim of this study is to classify fuel types according to Prometheus classification using low-density Airborne Laser Scanner (ALS) data, Sentinel 2 data, and 136 field plots used as ground-truth. The study encompassed three different Mediterranean forests dominated by pines (Pinus halepensis, P. pinaster y P. nigra), oaks (Quercus ilex) and quercus (Q. faginea) in areas affected by wildfires in 1994 and their surroundings. Two metric selection approaches and two non-parametric classification methods with variants were compared to classify fuel types. The best-fitted classification model was obtained using Support Vector Machine method with radial kernel. The model includes three ALS and one Sentinel-2 metrics: the 25th percentile of returns height, the percentage of all returns above mean, rumple structural diversity index and NDVI. The overall accuracy of the model after validation was 59%. The combination of data from active and passive remote sensing sensors as well as the use of adapted structural diversity indices derived from ALS data improved accuracy classification. This approach demonstrates its value for mapping fuel type spatial patterns at a regional scale under different heterogeneous and topographically complex Mediterranean forests. Full article
(This article belongs to the Special Issue Feature Paper Special Issue on Forest Remote Sensing)
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>Study area with the location of areas affected by wildfires, surrounding areas, field plots and pre-fire vegetation.</p> Full article ">Figure 2
<p>Relative presence of fuel types for the different forested areas under study. The value 0 refers to bare soil, and the values 1 to 7 stands for the <span class="html-italic">Prometheus</span> fuel types.</p> Full article ">Figure 3
<p>Fuel type classification mapping using SVM with radial kernel within the forested areas under study.</p> Full article ">Figure 4
<p>Relative presence, expressed between 0 and 1, of fuel types in areas affected by the three wildfires or their surroundings.</p> Full article ">
22 pages, 9976 KiB  
Article
A Deep Learning-Based Method for Quantifying and Mapping the Grain Size on Pebble Beaches
by Antoine Soloy, Imen Turki, Matthieu Fournier, Stéphane Costa, Bastien Peuziat and Nicolas Lecoq
Remote Sens. 2020, 12(21), 3659; https://doi.org/10.3390/rs12213659 - 8 Nov 2020
Cited by 29 | Viewed by 6052
Abstract
This article proposes a new methodological approach to measure and map the size of coarse clasts on a land surface from photographs. This method is based on the use of the Mask Regional Convolutional Neural Network (R-CNN) deep learning algorithm, which allows the [...] Read more.
This article proposes a new methodological approach to measure and map the size of coarse clasts on a land surface from photographs. This method is based on the use of the Mask Regional Convolutional Neural Network (R-CNN) deep learning algorithm, which allows the instance segmentation of objects after an initial training on manually labeled data. The algorithm is capable of identifying and classifying objects present in an image at the pixel scale, without human intervention, in a matter of seconds. This work demonstrates that it is possible to train the model to detect non-overlapping coarse sediments on scaled images, in order to extract their individual size and morphological characteristics with high efficiency (R2 = 0.98; Root Mean Square Error (RMSE) = 3.9 mm). It is then possible to measure element size profiles over a sedimentary body, as it was done on the pebble beach of Etretat (Normandy, France) in order to monitor the granulometric spatial variability before and after a storm. Applied at a larger scale using Unmanned Aerial Vehicle (UAV) derived ortho-images, the method allows the accurate characterization and high-resolution mapping of the surface coarse sediment size, as it was performed on the two pebble beaches of Etretat (D50 = 5.99 cm) and Hautot-sur-Mer (D50 = 7.44 cm) (Normandy, France). Validation results show a very satisfying overall representativity (R2 = 0.45 and 0.75; RMSE = 6.8 mm and 9.3 mm at Etretat and Hautot-sur-Mer, respectively), while the method remains fast, easy to apply and low-cost, although the method remains limited by the image resolution (objects need to be longer than 4 cm), and could still be improved in several ways, for instance by adding more manually labeled data to the training dataset, and by considering more accurate methods than the ellipse fitting for measuring the particle sizes. Full article
(This article belongs to the Special Issue Advances in Remote Sensing in Coastal and Hydraulic Engineering)
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>Example of a close-range top view image of a pebble ridge including a scaling quadra structure (<b>a</b>) before and (<b>b</b>) after orthorectification.</p> Full article ">Figure 2
<p>The Mask Regional Convolutional Neural Network (R-CNN) framework for instance segmentation [<a href="#B32-remotesensing-12-03659" class="html-bibr">32</a>].</p> Full article ">Figure 3
<p>Comparison between sediment particles’ shape as detected with Mask R-CNN (red) and the ellipse fitted around this shape (blue). The ellipses’ major and minor axes are displayed as dashed blue lines.</p> Full article ">Figure 4
<p>Application of Mask R-CNN to manually measured validation data with (<b>a</b>) a non-overlapping disposition of the clasts (colors show the detected pebble instances), (<b>b</b>) an image including overlapping objects, (<b>c</b>) the overlapping scenario tested with the highest number of misidentifications (instances n°4 and n°25) and (<b>d</b>), a comparative distribution of the pebble major axis measured after Mask R-CNN detection vs. the data measured manually using a caliper.</p> Full article ">Figure 5
<p>Location and pictures of the two pebble beaches of Etretat and Hautot-sur-Mer.</p> Full article ">Figure 6
<p>(<b>a</b>) Location of the cross-shore quadra measurement profiles at Etretat on 5 March and 13 March 2020. (<b>b</b>) Hydrodynamic conditions during the measurements: significant wave height (top), wave direction (center) and water level (bottom). (<b>c</b>) Elevation and pebble size profiles of D10, D50 and D90 on 5 March, 2020 (top) and 13 March, 2020 (bottom). Vertical bars show the measurement uncertainty, and horizontal red lines show the intertidal extension of the last tide.</p> Full article ">Figure 7
<p>Unmanned Aerial Vehicle (UAV) derived Ortho-images of the pebble ridges at (<b>a</b>) Etretat and (<b>b</b>) Hautot-sur-Mer.</p> Full article ">Figure 8
<p>Relative scale of the 1 m tiles in comparison to the whole ortho-image at Hautot-sur-Mer.</p> Full article ">Figure 9
<p>Comparison between the image quality of (<b>a</b>) an ortho-image (resolution 5 mm/pixel), and (<b>b</b>) a quadra image (resolution 0.5 mm/pixel). The white square shows the quadra position.</p> Full article ">Figure 10
<p>Comparison between the clast size distribution of a terrestrial sample (yellow) with a UAV sample (blue) at the same location, (<b>a</b>) before and (<b>b</b>) after applying the filtering processing.</p> Full article ">Figure 11
<p>Clast size mapping validation results. (<b>a</b>,<b>b</b>): Comparison of the average grain size as measured on the terrestrial samples (blue “*” symbols), the filtered terrestrial samples (red “o” symbols) and by the UAV samples (black “+”symbols at (<b>a</b>) Etretat and (<b>b</b>) Hautot-sur-Mer. Colored envelops present the standard deviation intervals. Compared mean values of the UAV and filtered terrestrial distributions at (<b>c</b>) Etretat and (<b>d</b>) Hautot-sur-Mer. Quantile–Quantile diagrams comparing the UAV and filtered terrestrial distributions at (<b>e</b>) Etretat and (<b>f</b>) Hautot-sur-Mer. Absolute error on the pebble size measurements as a function of the elevation at (<b>g</b>) Etretat and (<b>h</b>) Hautot-sur-Mer.</p> Full article ">Figure 11 Cont.
<p>Clast size mapping validation results. (<b>a</b>,<b>b</b>): Comparison of the average grain size as measured on the terrestrial samples (blue “*” symbols), the filtered terrestrial samples (red “o” symbols) and by the UAV samples (black “+”symbols at (<b>a</b>) Etretat and (<b>b</b>) Hautot-sur-Mer. Colored envelops present the standard deviation intervals. Compared mean values of the UAV and filtered terrestrial distributions at (<b>c</b>) Etretat and (<b>d</b>) Hautot-sur-Mer. Quantile–Quantile diagrams comparing the UAV and filtered terrestrial distributions at (<b>e</b>) Etretat and (<b>f</b>) Hautot-sur-Mer. Absolute error on the pebble size measurements as a function of the elevation at (<b>g</b>) Etretat and (<b>h</b>) Hautot-sur-Mer.</p> Full article ">Figure 12
<p>Distribution of individual pebbles on the surface of the beaches of (<b>a</b>) Etretat and (<b>b</b>) Hautot-sur-Mer. Histograms of the grain size distribution at (<b>c</b>) Etretat and (<b>d</b>) Hautot-sur-Mer, orange bars show the ellipse major axis dimensions, blue bars refer to the minor axis, and black vertical lines locate the major axis distribution’s D10, D50 and D90 values. Histogram of the grain elongation values (<b>e</b>) and of the grain circularity values (<b>f</b>) Etretat (orange) and Hautot-sur-Mer (blue).</p> Full article ">Figure 12 Cont.
<p>Distribution of individual pebbles on the surface of the beaches of (<b>a</b>) Etretat and (<b>b</b>) Hautot-sur-Mer. Histograms of the grain size distribution at (<b>c</b>) Etretat and (<b>d</b>) Hautot-sur-Mer, orange bars show the ellipse major axis dimensions, blue bars refer to the minor axis, and black vertical lines locate the major axis distribution’s D10, D50 and D90 values. Histogram of the grain elongation values (<b>e</b>) and of the grain circularity values (<b>f</b>) Etretat (orange) and Hautot-sur-Mer (blue).</p> Full article ">Figure 13
<p>Hydrodynamic conditions at (<b>a</b>) Etretat and (<b>b</b>) Hautot-sur-Mer preceding the UAV campaigns. Significant wave height (top), wave direction (center) and water level (bottom). Distribution of individual pebbles on the surface of the beaches of (<b>c</b>) Etretat and (<b>d</b>) Hautot-sur-Mer. Spatial variability of D50 at (<b>e</b>) Etretat and (<b>f</b>) Hautot-sur-Mer.</p> Full article ">
31 pages, 16443 KiB  
Article
Coastal Dam Inundation Assessment for the Yellow River Delta: Measurements, Analysis and Scenario
by Guoyang Wang, Peng Li, Zhenhong Li, Dong Ding, Lulu Qiao, Jishang Xu, Guangxue Li and Houjie Wang
Remote Sens. 2020, 12(21), 3658; https://doi.org/10.3390/rs12213658 - 8 Nov 2020
Cited by 28 | Viewed by 4177
Abstract
Coastal dams along the Yellow River Delta are built to prevent seawater intrusion. However, land subsidence caused by significant oil, gas and brine extraction, as well as sediment compaction, could exacerbate the flooding effects of sea-level rise and storm surge. In order to [...] Read more.
Coastal dams along the Yellow River Delta are built to prevent seawater intrusion. However, land subsidence caused by significant oil, gas and brine extraction, as well as sediment compaction, could exacerbate the flooding effects of sea-level rise and storm surge. In order to evaluate the coastal dam vulnerability, we combined unmanned aerial vehicle (UAV) Light Detection and Ranging (LiDAR) with small baseline subsets (SBAS) interferometric synthetic aperture radar (InSAR) results to generate an accurate coastal dam digital elevation model (DEM) over the next 10, 30 and 80 years. Sea-level simulation was derived from the relative sea-level rise scenarios published by the Intergovernmental Panel on Climate Change (IPCC) and local long-term tide gauge records. Assuming that the current rate of dam vertical deformation and sea-level rise are linear, we then generated different inundation scenarios by the superposition of DEMs and sea-levels at different periods by way of a bathtub model. We found that the overtopping event would likely occur around Year 2050, and the northern part of the dam would lose its protective capability almost entirely by the end of this century. This article provides an alternative cost-effective method for the detection, extraction and monitoring of coastal artificial infrastructure. Full article
(This article belongs to the Special Issue Dam Stability Monitoring with Satellite Geodesy)
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>Location of Gudong Dam taken from Google Earth: (<b>a</b>) general location of the Gudong Oilfield in the Yellow River Delta. The red polyline and yellow pentagram in the lower right subgraph show the Dongying District and location of the Gudong Oilfield, respectively; (<b>b</b>) detailed image of the Gudong Oilfield. Large pentagram indicates the location of tide gauge, whereas small ones show the aerial photo sites in <a href="#remotesensing-12-03658-f003" class="html-fig">Figure 3</a>.</p> Full article ">Figure 2
<p>Pictures of cracks caused by dam deformation taken in January 2020: (<b>a</b>) cracks on asphalt pavement of the dam crest; (<b>b</b>) cracks on retaining wall; (<b>c</b>) longitudinal cracks on the first stage platform of the dam; (<b>d</b>) transverse cracks on the first stage platform of the dam.</p> Full article ">Figure 3
<p>Aerial photos obtained by the unmanned aerial vehicle (UAV) in August 2019: (<b>a</b>) structures of Gudong Dam; (<b>b</b>) Gudong tide gauge station and drainage station built for draining water as a response to the waterlogged condition after Typhoon Lekima (Super Typhoon Lekima, International Code: 1909); (<b>c</b>) damage to the bending section at Gudong Dam caused by Typhoon Lekima; (<b>d</b>) damage to the straight section at Gudong Dam due to Typhoon Lekima.</p> Full article ">Figure 4
<p>Flow chart of inundation assessment in this study.</p> Full article ">Figure 5
<p>(<b>a</b>) DJI Matrice 600 Pro UAV, vehicle of Light Detection and Ranging (LiDAR); (<b>b</b>) control point arrangement; (<b>c</b>) detail points collection; (<b>d</b>) reference station for the Global Navigation Satellite System (GNSS) real time kinematic (RTK).</p> Full article ">Figure 6
<p>Detail point cloud: (<b>a</b>) tide gauge station; (<b>b</b>) southern end of the dam; (<b>c</b>) cross section of the dam in the middle section cut from point cloud.</p> Full article ">Figure 7
<p>Flow chart of LiDAR digital elevation model (DEM) generation.</p> Full article ">Figure 8
<p>Strip DEM of Gudong Dam and its details.</p> Full article ">Figure 9
<p>(<b>a</b>) The distribution of all detail points collected by GNSS RTK along Gudong Dam; (<b>b</b>) typical presentation of detail points.</p> Full article ">Figure 10
<p>Correlation between detailed points collected by GNSS RTK and LiDAR DEM. The number of points utilized in the assessment was 1846.</p> Full article ">Figure 11
<p>Flow chart of small baseline subsets (SBAS) interferometric synthetic aperture radar (InSAR) modified from Berardino et al. [<a href="#B61-remotesensing-12-03658" class="html-bibr">61</a>].</p> Full article ">Figure 12
<p>Vertical deformation rate maps estimated by Sentinel-1 SBAS InSAR: (<b>a</b>) descending track deformation rate map from the Sentinel-1B images; (<b>b</b>) ascending track deformation rate map from the Sentinel-1A images. The base maps are taken from Google Earth.</p> Full article ">Figure 13
<p>(<b>a</b>) Correlation of InSAR deformation rate between the ascending and descending tracks; (<b>b</b>) difference distribution statistics between ascending and descending tracks.</p> Full article ">Figure 14
<p>Schematic diagram of DEM simulation.</p> Full article ">Figure 15
<p>Regular analysis of deformation rate. The position of the four points is the triangle position shown in <a href="#remotesensing-12-03658-f009" class="html-fig">Figure 9</a> (<b>a</b>), (<b>b</b>), (<b>c</b>) and (<b>d</b>).</p> Full article ">Figure 16
<p>Raster calculation process. According to the corresponding position, the values of the two grids are added to obtain the simulated DEM.2.5. Sea-level Simulation.</p> Full article ">Figure 17
<p>Simulated tide change.</p> Full article ">Figure 18
<p>Storm surge records for the Gudong tide gauge station and Dongying port. As the distance is very close, and the law of tidal changes is consistent, the storm surge interval value is taken from the records of the above two stations.</p> Full article ">Figure 19
<p>The yellow grid represents the dam that has not been submerged, and the blue grid represents the dam that has been submerged due to sea-level rise. The numbers in the boxes represent elevation value.</p> Full article ">Figure 20
<p>Profile line of Gudong Dam in the years 2020, 2030, 2050 and 2100, which showed the trend of the dam over time. The vertical axis represents the elevation, and the point number of the horizontal axis represents a segment number from north to south.</p> Full article ">Figure 21
<p>DEM changes superposed with the dam deformation for four fragments of Gudong Dam DEM for the years 2020, 2030, 2050 and 2100.</p> Full article ">Figure 22
<p>Sea-level simulation. The above four figures show the simulated sea-levels. The storm surge is displayed as an interval and is blue filled.</p> Full article ">Figure 23
<p>Schematic diagram of the inundation assessment in this study.</p> Full article ">Figure 24
<p>Static display of dam profile and sea-level in 2020. The profile line is along the center line of Asphalt Road on dam crest, and the horizontal line is the height of two simulated sea-levels; the highest sea-level is 2.48 m and the normal sea-level is 0.98 m. This is an indication of the submerged range from the perspective of the longitudinal section.</p> Full article ">Figure 25
<p>Cross section of the dam and 2D inundation diagram for 2020. The first column shows the DEM of four different part of the dam in 2020, and the second column indicates the inundation range of the dam, due to the high sea-level (2.48 m) shaded in blue. The third column represents a cross section of the dam at Points 10, 49, 88 and 163, in which the right direction is seaward.</p> Full article ">Figure 26
<p>Static display of dam profile and sea-level for 2030. The highest sea-level is 2.58 m and the normal sea-level is 1.08 m.</p> Full article ">Figure 27
<p>Cross section of the dam and 2D inundation diagram for 2030. The first column shows the DEM of four different parts of the dam in 2030, and the second column indicates the inundation range of the dam due to the highest sea-level (2.58 m), shaded in blue. The third column represents a cross section of the dam at points 10, 49, 88 and 163, in which the right direction is seaward.</p> Full article ">Figure 28
<p>Static display of dam profile and sea-level in 2050. The highest sea-level is 2.81 m and the normal sea-level is 1.31 m.</p> Full article ">Figure 29
<p>Cross section of the dam and 2D inundation diagram for 2050. The first column shows the DEM of four different parts of the dam in 2050, and the second column indicates the inundation range of the dam due to the highest sea-level (2.81 m) shaded in blue. The third column represents a cross section of the dam at points 10, 49, 88 and 163, in which the right direction is seaward. The intrusion point was found in the section of point 10.</p> Full article ">Figure 30
<p>Static display of dam profile and sea-level for 2050. The highest sea-level is 3.46 m and the normal sea-level is 1.96 m. We are able to establish the entry point in the high sea-level scenario in 2100 and find that the northern and southern parts of the dam will lose their protective ability completely.</p> Full article ">Figure 31
<p>Cross section of the dam and 2D inundation diagram for 2100. The first column shows the DEM of four different parts of the dam in 2100, and the second column indicates the inundation range of the dam due to the high sea-level (3.46 m), shaded in blue. The third column represents a cross section of the dam at points 10, 49, 88 and 163, in which the right direction is seaward. The dam at points 10 and 163 will lose its protective capacity completely, and seawater will easily overflow the dam crest.</p> Full article ">Figure 32
<p>Simplification of surface deformation process in coastal zone.</p> Full article ">Figure 33
<p>Schematic diagram of danger range of sea level change.</p> Full article ">Figure 34
<p>Inundation assessment in this paper.</p> Full article ">
19 pages, 6525 KiB  
Article
Learning-Based Hyperspectral Imagery Compression through Generative Neural Networks
by Chubo Deng, Yi Cen and Lifu Zhang
Remote Sens. 2020, 12(21), 3657; https://doi.org/10.3390/rs12213657 - 8 Nov 2020
Cited by 16 | Viewed by 3259
Abstract
Hyperspectral images (HSIs), which obtain abundant spectral information for narrow spectral bands (no wider than 10 nm), have greatly improved our ability to qualitatively and quantitatively sense the Earth. Since HSIs are collected by high-resolution instruments over a very large number of wavelengths, [...] Read more.
Hyperspectral images (HSIs), which obtain abundant spectral information for narrow spectral bands (no wider than 10 nm), have greatly improved our ability to qualitatively and quantitatively sense the Earth. Since HSIs are collected by high-resolution instruments over a very large number of wavelengths, the data generated by such sensors is enormous, and the amount of data continues to grow, HSI compression technique will play more crucial role in this trend. The classical method for HSI compression is through compression and reconstruction methods such as three-dimensional wavelet-based techniques or the principle component analysis (PCA) transform. In this paper, we provide an alternative approach for HSI compression via a generative neural network (GNN), which learns the probability distribution of the real data from a random latent code. This is achieved by defining a family of densities and finding the one minimizing the distance between this family and the real data distribution. Then, the well-trained neural network is a representation of the HSI, and the compression ratio is determined by the complexity of the GNN. Moreover, the latent code can be encrypted by embedding a digit with a random distribution, which makes the code confidential. Experimental examples are presented to demonstrate the potential of the GNN to solve image compression problems in the field of HSI. Compared with other algorithms, it has better performance at high compression ratio, and there is still much room left for improvements along with the fast development of deep-learning techniques. Full article
(This article belongs to the Special Issue Learning-Based Hyperspectral Information Extraction: Algorithms and Applications)
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>Variational autoencoder architecture.</p> Full article ">Figure 2
<p>Generative neural network (GNN) architecture.</p> Full article ">Figure 3
<p>Bilinear interpolation.</p> Full article ">Figure 4
<p>Comparison of entropy for a n × m matrix for a fixed n = 1000.</p> Full article ">Figure 5
<p>Weight pruning demonstration.</p> Full article ">Figure 6
<p>Illustration of the embedding technique.</p> Full article ">Figure 7
<p>Batch training for a huge image.</p> Full article ">Figure 8
<p>True color image of Xiong An New Area, Hebei, China (Matiwan Village).</p> Full article ">Figure 9
<p>True color-generated hyperspectral images (HSI) 512 × 512 pixels in size at different epochs.</p> Full article ">Figure 10
<p>Comparison of generated and original pixels in the spectral domain.</p> Full article ">Figure 11
<p>Comparison of the power loss function measured using the mean square error (MSE) criterion.</p> Full article ">Figure 12
<p>Singularity point occurring with a small value of <span class="html-italic">p</span>.</p> Full article ">Figure 13
<p>Batch training demonstration.</p> Full article ">Figure 14
<p>Loss of batch-size training.</p> Full article ">Figure 15
<p>Performance comparison of different GNN structures.</p> Full article ">
17 pages, 412 KiB  
Article
Model Selection in Atmospheric Remote Sensing with Application to Aerosol Retrieval from DSCOVR/EPIC. Part 2: Numerical Analysis
by Sruthy Sasi, Vijay Natraj, Víctor Molina García, Dmitry S. Efremenko, Diego Loyola and Adrian Doicu
Remote Sens. 2020, 12(21), 3656; https://doi.org/10.3390/rs12213656 - 7 Nov 2020
Cited by 5 | Viewed by 2614
Abstract
An algorithm for retrieving aerosol parameters by taking into account the uncertainty in aerosol model selection is applied to the retrieval of aerosol optical thickness and aerosol layer height from synthetic measurements from the EPIC sensor onboard the Deep Space Climate Observatory. The [...] Read more.
An algorithm for retrieving aerosol parameters by taking into account the uncertainty in aerosol model selection is applied to the retrieval of aerosol optical thickness and aerosol layer height from synthetic measurements from the EPIC sensor onboard the Deep Space Climate Observatory. The synthetic measurements are generated using aerosol models derived from AERONET measurements at different sites, while other commonly used aerosol models, such as OPAC, GOCART, OMI, and MODIS databases are used in the retrieval. The numerical analysis is focused on the estimation of retrieval errors when the true aerosol model is unknown. We found that the best aerosol model is the one with a value of the asymmetry parameter and an angular variation of the phase function around the viewing direction that is close to the values corresponding to the reference aerosol model. Full article
(This article belongs to the Special Issue Advances of Remote Sensing Inversion)
►▼ Show Figures

Figure 1

Figure 1
<p>Relative errors <math display="inline"><semantics> <msubsup> <mi>ε</mi> <mrow> <mi>mean</mi> </mrow> <mrow> <mi>τ</mi> <mo>(</mo> <mi>τ</mi> <mo>)</mo> </mrow> </msubsup> </semantics></math> and <math display="inline"><semantics> <msubsup> <mi>ε</mi> <mrow> <mi>mean</mi> </mrow> <mrow> <mi>H</mi> <mo>(</mo> <mi>H</mi> <mo>)</mo> </mrow> </msubsup> </semantics></math> for the first test example. The aerosol databases are OPAC with <math display="inline"><semantics> <mrow> <mi>U</mi> <mo>=</mo> <mn>0.80</mn> </mrow> </semantics></math> (1), <math display="inline"><semantics> <mrow> <mi>U</mi> <mo>=</mo> <mn>0.90</mn> </mrow> </semantics></math> (2), and <math display="inline"><semantics> <mrow> <mi>U</mi> <mo>=</mo> <mn>0.95</mn> </mrow> </semantics></math> (3), GOCART with <math display="inline"><semantics> <mrow> <mi>U</mi> <mo>=</mo> <mn>0.80</mn> </mrow> </semantics></math> (4), <math display="inline"><semantics> <mrow> <mi>U</mi> <mo>=</mo> <mn>0.90</mn> </mrow> </semantics></math> (5), and <math display="inline"><semantics> <mrow> <mi>U</mi> <mo>=</mo> <mn>0.95</mn> </mrow> </semantics></math> (6), OMI (7), and MODIS (8). The plots in the left panels correspond to the extended sets of aerosol models, while those in the right panels correspond to the reduced sets.</p> Full article ">Figure 2
<p>Relative errors <math display="inline"><semantics> <msubsup> <mi>ε</mi> <mrow> <mi>mean</mi> </mrow> <mrow> <mi>τ</mi> <mo>(</mo> <mi>τ</mi> <mo>)</mo> </mrow> </msubsup> </semantics></math>, <math display="inline"><semantics> <msubsup> <mi>ε</mi> <mrow> <mi>max</mi> </mrow> <mrow> <mi>τ</mi> <mo>(</mo> <mi>τ</mi> <mo>)</mo> </mrow> </msubsup> </semantics></math>, <math display="inline"><semantics> <msubsup> <mi>ε</mi> <mrow> <mi>mean</mi> </mrow> <mrow> <mi>H</mi> <mo>(</mo> <mi>H</mi> <mo>)</mo> </mrow> </msubsup> </semantics></math>, and <math display="inline"><semantics> <msubsup> <mi>ε</mi> <mrow> <mi>max</mi> </mrow> <mrow> <mi>H</mi> <mo>(</mo> <mi>H</mi> <mo>)</mo> </mrow> </msubsup> </semantics></math> for the first test example using the extended sets of aerosol models. The aerosol databases are labeled in the same way as in <a href="#remotesensing-12-03656-f001" class="html-fig">Figure 1</a>.</p> Full article ">Figure 3
<p>Same as <a href="#remotesensing-12-03656-f001" class="html-fig">Figure 1</a> but for the second test example.</p> Full article ">Figure 4
<p>Same as in <a href="#remotesensing-12-03656-f001" class="html-fig">Figure 1</a> but for the third test example.</p> Full article ">Figure 5
<p>Phase functions for <math display="inline"><semantics> <mrow> <msub> <mi>τ</mi> <mi mathvariant="normal">t</mi> </msub> <mo>=</mo> <mn>1.0</mn> </mrow> </semantics></math> and <math display="inline"><semantics> <mrow> <msub> <mi>H</mi> <mi mathvariant="normal">t</mi> </msub> <mo>=</mo> <mn>3.0</mn> </mrow> </semantics></math>. The results correspond to the first test example and the extended sets of aerosol models.</p> Full article ">
24 pages, 11578 KiB  
Article
Wavelet Scattering Network-Based Machine Learning for Ground Penetrating Radar Imaging: Application in Pipeline Identification
by Yang Jin and Yunling Duan
Remote Sens. 2020, 12(21), 3655; https://doi.org/10.3390/rs12213655 - 7 Nov 2020
Cited by 26 | Viewed by 5208
Abstract
Automatic and efficient ground penetrating radar (GPR) data analysis remains a bottleneck, especially restricting applications in real-time monitoring systems. Deep learning approaches have good practice in automatic object identification, but their intensive data requirement has reduced their applicability. This paper developed a machine [...] Read more.
Automatic and efficient ground penetrating radar (GPR) data analysis remains a bottleneck, especially restricting applications in real-time monitoring systems. Deep learning approaches have good practice in automatic object identification, but their intensive data requirement has reduced their applicability. This paper developed a machine learning framework based on wavelet scattering networks to analyze GPR data for subsurface pipeline identification. Wavelet scattering network is functionally equivalent to convolutional neural networks, and its null-parameter property is intended for non-intensive datasets. A double-channel framework is designed with wavelet scattering networks followed by support vector machines to determine the existence of pipelines on vertical and horizontal traces separately. Classification accuracy rates arrive around 98% and 95% for datasets without and with noises, respectively, as well as 97% for considering surface roughness. Pipeline locations and diameters are convenient to determine from the reconstructed profiles of both simulated and practical GPR signals. However, the results of 5 cm pipelines are sensitive to noises. Nonetheless, the developed machine learning approach presents promising applicability in subsurface pipeline identification. Full article
(This article belongs to the Special Issue Advanced Techniques for Ground Penetrating Radar Imaging)
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>(<b>a</b>) Example subsurface section with only one pre-buried pipeline, where the brown color represents the soil and the white color represents the pipeline. (<b>b</b>) Example subsurface section with multiple pre-buried pipelines. (<b>c</b>) GPR signal profile corresponding to the subsurface section in panel (a). (<b>d</b>) GPR signal profile corresponding to the subsurface section in panel (b).</p> Full article ">Figure 2
<p>Field GPR profiles of (<b>a</b>) two separate pre-buried pipelines, (<b>b</b>) double-layer concrete cylinders, and (<b>c</b>) three distributed pipelines [<a href="#B28-remotesensing-12-03655" class="html-bibr">28</a>].</p> Full article ">Figure 3
<p>De-“wow” results of a representative GPR trace.</p> Full article ">Figure 4
<p>Example data profiles after preprocessing of (<b>a</b>) <a href="#remotesensing-12-03655-f001" class="html-fig">Figure 1</a>c, (<b>b</b>) <a href="#remotesensing-12-03655-f001" class="html-fig">Figure 1</a>d, (<b>c</b>) <a href="#remotesensing-12-03655-f002" class="html-fig">Figure 2</a>a, and (<b>d</b>) <a href="#remotesensing-12-03655-f002" class="html-fig">Figure 2</a>c.</p> Full article ">Figure 5
<p>A three layer wavelet scattering network. The operator <math display="inline"><semantics> <msub> <mi>U</mi> <msub> <mi>λ</mi> <mn>1</mn> </msub> </msub> </semantics></math> is applied to the original signal <span class="html-italic">x</span> to calculate each <math display="inline"><semantics> <mrow> <mi>U</mi> <mo>[</mo> <msub> <mi>λ</mi> <mn>1</mn> </msub> <mo>]</mo> <mi>x</mi> </mrow> </semantics></math> and output <math display="inline"><semantics> <mrow> <msub> <mi>S</mi> <mi>J</mi> </msub> <mrow> <mo>[</mo> <mo>∅</mo> <mo>]</mo> </mrow> <mi>x</mi> </mrow> </semantics></math>, where ∅ represents an empty set. Then, the operator <math display="inline"><semantics> <msub> <mi>U</mi> <msub> <mi>λ</mi> <mn>2</mn> </msub> </msub> </semantics></math> is applied to each previous layer <math display="inline"><semantics> <mrow> <mi>U</mi> <mo>[</mo> <msub> <mi>λ</mi> <mn>1</mn> </msub> <mo>]</mo> <mi>x</mi> </mrow> </semantics></math> to calculate all <math display="inline"><semantics> <mrow> <mi>U</mi> <mo>[</mo> <msub> <mi>λ</mi> <mn>1</mn> </msub> <mo>,</mo> <msub> <mi>λ</mi> <mn>2</mn> </msub> <mo>]</mo> <mi>x</mi> </mrow> </semantics></math> and output <math display="inline"><semantics> <mrow> <msub> <mi>S</mi> <mi>J</mi> </msub> <mrow> <mo>[</mo> <msub> <mi>λ</mi> <mn>1</mn> </msub> <mo>]</mo> </mrow> <mi>x</mi> </mrow> </semantics></math>. This scattering process is operated iteratively to obtain all convolution results.</p> Full article ">Figure 6
<p>(<b>a</b>) Real component of example Morlet wavelet with parameters <math display="inline"><semantics> <mrow> <msub> <mi>K</mi> <msub> <mi>σ</mi> <mi>t</mi> </msub> </msub> <mo>=</mo> <mn>1</mn> </mrow> </semantics></math>, <math display="inline"><semantics> <mrow> <msub> <mi>σ</mi> <mi>t</mi> </msub> <mo>=</mo> <mn>1</mn> </mrow> </semantics></math>, and <math display="inline"><semantics> <mrow> <mn>2</mn> <mi>π</mi> <mi>f</mi> <mo>=</mo> <mn>5</mn> </mrow> </semantics></math>. (<b>b</b>) Example Gaussian window as scale function <math display="inline"><semantics> <mrow> <mi>ϕ</mi> <mo>(</mo> <mi>t</mi> <mo>)</mo> </mrow> </semantics></math>.</p> Full article ">Figure 7
<p>The machine learning architecture for pipeline identification.</p> Full article ">Figure 8
<p>A brief scheme of profile reconstruction.</p> Full article ">Figure 9
<p>(<b>a</b>) The example input profile with a single pipeline (central coordinates (0.615 m, 0.63 m), diameter 0.25 m). (<b>b</b>) Another example input profile with a single pipeline (central coordinates (0.275 m, 0.667 m), diameter 0.35 m). (<b>c</b>) The reconstructed profile of panel (a) by the proposed learning framework. (<b>d</b>) The reconstructed profile of panel (b) by the proposed learning framework. (<b>e</b>) The reconstructed profile of panel (a) after the independent SVM for comparison. (<b>f</b>) The reconstructed profile of panel (b) after the independent SVM for comparison.</p> Full article ">Figure 10
<p>(<b>a</b>) The example input profile with two pipelines (central coordinates and diameters ((1.355 m, 0.263 m),0.15 m) and ((0.875 m, 0.823 m), 0.4 m)). (<b>b</b>) The example input profile with six pipelines (left five central coordinates and diameters ((0.12 m, 0.56 m), 0.1 m), ((0.505 m, 0.658 m), 0.1 m), ((0.695 m, 0.705 m), 0.2 m), ((1.1 m, 0.66 m), 0.05 m) and ((1.365 m, 0.453 m), 0.05 m), and rightest height and width (0.385 m, 0.37 m)). (<b>c</b>) The reconstructed profile of panel (a) by the proposed learning framework. (<b>d</b>) The reconstructed profile of panel (b) by the proposed learning framework. (<b>e</b>) The reconstructed profile of panel (a) after the independent SVM for comparison. (<b>f</b>) The reconstructed profile of panel (b) after the independent SVM for comparison.</p> Full article ">Figure 11
<p>(<b>a</b>) The example noisy profile with a single pipeline (seen in <a href="#remotesensing-12-03655-f009" class="html-fig">Figure 9</a>a). (<b>b</b>) Another example noisy profile with a single pipeline (seen in <a href="#remotesensing-12-03655-f009" class="html-fig">Figure 9</a>b). (<b>c</b>) The reconstructed profile of panel (a) by the proposed learning framework. (<b>d</b>) The reconstructed profile of panel (b) by the proposed learning framework. (<b>e</b>) The reconstructed profile of panel (a) after the independent SVM for comparison. (<b>f</b>) The reconstructed profile of panel (b) after the independent SVM for comparison.</p> Full article ">Figure 12
<p>(<b>a</b>) The example noisy profile with two pipelines (central coordinates and diameters ((1.355 m, 0.263 m), 0.15 m) &amp; ((0.875 m, 0.823 m), 0.4 m)). (<b>b</b>) The example noisy profile with six pipelines (left five central coordinates and diameters ((0.12 m, 0.56 m), 0.1 m), ((0.505 m, 0.658 m), 0.1 m), ((0.695 m, 0.705 m), 0.2 m), ((1.1 m, 0.66 m), 0.05 m) &amp; ((1.365 m, 0.453 m), 0.05 m), and rightest height and width (0.385 m, 0.37 m)). (<b>c</b>) The reconstructed profile of panel (a) by the proposed learning framework. (<b>d</b>) The reconstructed profile of panel (b) by the proposed learning framework. (<b>e</b>) The reconstructed profile of panel (a) after the independent SVM for comparison. (<b>f</b>) The reconstructed profile of panel (b) after the independent SVM for comparison.</p> Full article ">Figure 13
<p>(<b>a</b>) The example profile with a single pipeline (seen in <a href="#remotesensing-12-03655-f009" class="html-fig">Figure 9</a>a) considering surface roughness. (<b>b</b>) Another example profile with a single pipeline (seen in <a href="#remotesensing-12-03655-f009" class="html-fig">Figure 9</a>b) considering surface roughness. (<b>c</b>) The reconstructed profile of panel (a) by the proposed learning framework. (<b>d</b>) The reconstructed profile of panel (b) by the proposed learning framework. (<b>e</b>) The reconstructed profile of panel (a) after the independent SVM for comparison. (<b>f</b>) The reconstructed profile of panel (b) after the independent SVM for comparison.</p> Full article ">Figure 14
<p>(<b>a</b>) The example profile with two pipelines (central coordinates and diameters ((1.355 m, 0.263 m), 0.15 m) &amp; ((0.875 m, 0.823 m), 0.4 m)) considering surface roughness. (<b>b</b>) The example profile with six pipelines (left five central coordinates and diameters ((0.12 m, 0.56 m), 0.1 m), ((0.505 m, 0.658 m), 0.1 m), ((0.695 m, 0.705 m), 0.2 m), ((1.1 m, 0.66 m), 0.05 m) &amp; ((1.365 m, 0.453 m), 0.05 m), and rightest height and width (0.385 m, 0.37 m)) considering surface roughness. (<b>c</b>) The reconstructed profile of panel (a) by the proposed learning framework. (<b>d</b>) The reconstructed profile of panel (b) by the proposed learning framework. (<b>e</b>) The reconstructed profile of panel (a) after the independent SVM for comparison. (<b>f</b>) The reconstructed profile of panel (b) after the independent SVM for comparison.</p> Full article ">Figure 15
<p>(<b>a</b>) Example input profile with two pipelines (diameter 32 cm). (<b>b</b>) Reconstructed profile of panel (a) by machine learning. (<b>c</b>) Example input profile with 5 × 2 cylinders (diameter 10 cm). (<b>d</b>) Reconstructed profile of panel (c) by machine learning. (<b>e</b>) Example input profile with three pipelines (diameter 50 cm, depth of upper interface 1 m, 1.5 m and 2 m). (<b>f</b>) Reconstructed profile of panel (e) by machine learning.</p> Full article ">
21 pages, 8349 KiB  
Article
Sea Surface Temperature and High Water Temperature Occurrence Prediction Using a Long Short-Term Memory Model
by Minkyu Kim, Hyun Yang and Jonghwa Kim
Remote Sens. 2020, 12(21), 3654; https://doi.org/10.3390/rs12213654 - 7 Nov 2020
Cited by 43 | Viewed by 4800
Abstract
Recent global warming has been accompanied by high water temperatures (HWTs) in coastal areas of Korea, resulting in huge economic losses in the marine fishery industry due to disease outbreaks in aquaculture. To mitigate these losses, it is necessary to predict such outbreaks [...] Read more.
Recent global warming has been accompanied by high water temperatures (HWTs) in coastal areas of Korea, resulting in huge economic losses in the marine fishery industry due to disease outbreaks in aquaculture. To mitigate these losses, it is necessary to predict such outbreaks to prevent or respond to them as early as possible. In the present study, we propose an HWT prediction method that applies sea surface temperatures (SSTs) and deep-learning technology in a long short-term memory (LSTM) model based on a recurrent neural network (RNN). The LSTM model is used to predict time series data for the target areas, including the coastal area from Goheung to Yeosu, Jeollanam-do, Korea, which has experienced frequent HWT occurrences in recent years. To evaluate the performance of the SST prediction model, we compared and analyzed the results of an existing SST prediction model for the SST data, and additional external meteorological data. The proposed model outperformed the existing model in predicting SSTs and HWTs. Although the performance of the proposed model decreased as the prediction interval increased, it consistently showed better performance than the European Center for Medium-Range Weather Forecast (ECMWF) prediction model. Therefore, the method proposed in this study may be applied to prevent future damage to the aquaculture industry. Full article
(This article belongs to the Special Issue Advances in Retrieval, Operationalization, Monitoring and Application of Sea Surface Temperature)
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>Incidence number of sea surface temperature (SST) readings exceeding 28 °C in the seas around the Korean Peninsula from 2014 to 2018 and the target area selected for high water temperature (HWT) prediction.</p> Full article ">Figure 2
<p>Time series of average SSTs in the target area shown in <a href="#remotesensing-12-03654-f001" class="html-fig">Figure 1</a>.</p> Full article ">Figure 3
<p>Ocean depth and current flow patterns around Korea.</p> Full article ">Figure 4
<p>Structure of the long short-term memory (LSTM) model, including the forget, input, and output gates.</p> Full article ">Figure 5
<p>Conceptual model of LSTM training for SST prediction. (<b>a</b>) A typical 1-year SST data series. (<b>b</b>) A schematic diagram of LSTM model training to predict SST.</p> Full article ">Figure 6
<p>Schematic diagrams for SST prediction after <span class="html-italic">m</span> days using the trained LSTM model.</p> Full article ">Figure 7
<p>Actual LSTM network structure used in the experiments to predict SSTs.</p> Full article ">Figure 8
<p>Individual target areas, five in total (<b>a</b>–<b>e</b>; latitude 34.45°, longitude 127.3°–128.3°). The color shown for each area corresponds to the frequency of HWT occurrence.</p> Full article ">Figure 9
<p>Comparison of predicted and real SST data, with scatter diagrams, for 1-day and 7-days prediction intervals using the SST dataset as input.</p> Full article ">Figure 10
<p>Comparison of predicted and real SST data, with scatter diagrams, for 1-day and 7-days prediction intervals using the multi dataset as input.</p> Full article ">Figure 11
<p>Comparison of coefficient of determination (<span class="html-italic">R</span><sup>2</sup>), root mean square error (<span class="html-italic">RMSE</span>), and mean absolute percentage error (<span class="html-italic">MAPE)</span> values between LSTM results produced using the SST and multi datasets as input, for different prediction intervals</p> Full article ">Figure 12
<p>Receiver operating characteristic (ROC) space and plots of HWT occurrence predictions from 1 to 7 days using the true positive rate (<span class="html-italic">TPR</span>) and false positive rate (<span class="html-italic">FPR</span>) in area (a) for 2018.</p> Full article ">Figure 13
<p>Comparison of F1 scores obtained using the two input datasets.</p> Full article ">Figure 14
<p>Comparison of SST and HWT prediction performance between the proposed model (with multi-dataset input) and European Center for Medium-Range Weather Forecast (ECMWF) forecast data.</p> Full article ">Figure A1
<p><span class="html-italic">RMSE</span> and <span class="html-italic">MAPE</span> values between LSTM results produced using the SST and multi datasets as input, for different areas.</p> Full article ">Figure A2
<p>HWT and SST prediction performance for additional test area using multi dataset. (<b>a</b>) Area selected for further experiments. (<b>b</b>) F1 score values for additional test area. (<b>c</b>) <span class="html-italic">R</span><sup>2</sup>, <span class="html-italic">RMSE</span>, and <span class="html-italic">MAPE</span> values for the additional test area.</p> Full article ">Figure A3
<p>Comparison of SST prediction performance for August between the proposed model (with multi-dataset input) and European Center for Medium-Range Weather Forecast (ECMWF) forecast data.</p> Full article ">
21 pages, 9077 KiB  
Article
Early ICESat-2 on-orbit Geolocation Validation Using Ground-Based Corner Cube Retro-Reflectors
by Lori A. Magruder, Kelly M. Brunt and Michael Alonzo
Remote Sens. 2020, 12(21), 3653; https://doi.org/10.3390/rs12213653 - 7 Nov 2020
Cited by 90 | Viewed by 4597
Abstract
The Ice, Cloud and Land Elevation Satellite-2 (ICESat-2), an Earth-observing laser altimetry mission, is currently providing global elevation measurements. Geolocation validation confirms the altimeter’s ability to accurately position the measurement on the surface of the Earth and provides insight into the fidelity of [...] Read more.
The Ice, Cloud and Land Elevation Satellite-2 (ICESat-2), an Earth-observing laser altimetry mission, is currently providing global elevation measurements. Geolocation validation confirms the altimeter’s ability to accurately position the measurement on the surface of the Earth and provides insight into the fidelity of the geolocation determination process. Surfaces well characterized by independent methods are well suited to provide a measure of the ICESat-2 geolocation accuracy through statistical comparison. This study compares airborne lidar data with the ICESat-2 along-track geolocated photon data product to determine the horizontal geolocation accuracy by minimizing the vertical residuals between datasets. At the same location arrays of corner cube retro-reflectors (CCRs) provide unique signal signatures back to the satellite from their known positions to give a deterministic solution of the laser footprint diameter and the geolocation accuracy for those cases where two or more CCRs were illuminated within one ICESat-2 transect. This passive method for diameter recovery and geolocation accuracy assessment is implemented at two locations: White Sands Missile Range (WSMR) in New Mexico and along the 88°S latitude line in Antarctica. This early on-orbit study provides results as a proof of concept for this passive validation technique. For the cases studied the diameter value ranged from 10.6 to 12 m. The variability is attributed to the statistical nature of photon-counting lidar technology and potentially, variations in the atmospheric conditions that impact signal transmission. The geolocation accuracy results from the CCR technique and airborne lidar comparisons are within the mission requirement of 6.5 m. Full article
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>A conceptual view of the along-track relative elevations of ground and corner cube retro-reflector (CCR)-detected photons from a side (<b>a</b>) and top (<b>b</b>) viewpoint. The (<b>a</b>) diagram shows the observed separation of elevation between the terrain and the CCR but (<b>b</b>) reveals that the geolocation process positions all detections at the centerline of the laser footprint despite the relative position of the CCR within the laser spot.</p> Full article ">Figure 2
<p>Time lapse scenario of multiple laser footprints illuminating a single CCR when its position is aligned with the along-track centerline. Each panel (<b>a</b>–f) represents a theoretical time step associated with successive CCR illuminations.</p> Full article ">Figure 3
<p>(<b>a</b>) Scenario when the CCR is aligned with the footprint centerline and there are errors in the ATL03 geolocations. (<b>b</b>) Geolocation accuracy method for using the along (ΔY) and across (ΔX) track distances between the known CCR position and the mid-point of the CCR signature (blue box). (<b>c</b>) Final corrected ATL03 track based on theoretical results.</p> Full article ">Figure 4
<p>Geometric representation of the relative position of a CCR (black dot) to the laser footprint along-track centerline (C<sub>1</sub>) and its corresponding chord length (c), which is less than the full diameter of the spot. The horizontal offset (x) of the CCR position and the center point can be calculated using simple geometry with the diameter/radius of the footprint and measured chord length to determine the horizontal geolocation accuracy.</p> Full article ">Figure 5
<p>(<b>a</b>) Scenario when the CCR is not aligned with the footprint centerline and there are errors in the ATL03 geolocations. (<b>b</b>) Geolocation accuracy method for using the along (ΔY) and across (ΔX) track distances between the known CCR position and the mid-point of the CCR signature (blue box) for the (<b>c</b>) Final corrected ATL03 track based on theoretical results.</p> Full article ">Figure 6
<p>Corner cube signature for a strong beam CCR illumination. The black data are the signal attributed to the ground surface and the green data are those presumed to be from a single CCR based on the location and height of signature.</p> Full article ">Figure 7
<p>Along-track CCR signal distribution statistical extraction of the effective chord length based on the 2σ value for a Gaussian energy profile as a preliminary assessment of the optimal signal distribution.</p> Full article ">Figure 8
<p>Iterative process (<b>a</b>–<b>c</b>) of determining the geolocation offsets associated with the multiple (2) CCR signature characteristics and relative geometry for a given laser footprint diameter.</p> Full article ">Figure 9
<p>Iterative process (<b>a</b>–<b>d</b>) for exploring the relative positioning of the laser ground track and the locations of the illuminated CCRs. The number of combinations is 2<sup>N</sup>, where N is the number of CCRs considered (e.g., 2 CCRs = four possible geometric configurations).</p> Full article ">Figure 10
<p>Ascending and descending ground track intersection pattern that informed the relative positioning of multiple CCR arrays designed to capture multiple CCR signatures for a beam pair (weak and strong) during a single overpass. The GTl and GTr are the Ground Track Left/Right components of the pair.</p> Full article ">Figure 11
<p>CCR array geometry at White Sands Missile Range (WSMR). Each array contains 12 CCRs on varying height poles to ensure the correct CCR location is identified from the CCR signal signature.</p> Full article ">Figure 12
<p>Geographical locations of the CCR arrays along 88S (left), with a representative array also shown (right).</p> Full article ">Figure 13
<p>Thematic output for optimized fit between the airborne lidar reference surface (1 m resolution raster surface) and the ICESat-2 ATL03 elevations for the 2 November, 2018 overpass of WSMR. The solution indicates the geolocation corrections for ATL03.</p> Full article ">Figure 14
<p>Elevation comparison between the reference surface derived from airborne lidar measurements and ATL03 (<b>a</b>) and the error statistics along the transect (<b>b</b>)<b>.</b></p> Full article ">Figure 15
<p>Overview of the 31 Mar 2019 overpass of WSMR CCR arrays. The ground tracks shown here have been adjusted using the airborne lidar survey comparison technique to correct for horizontal offsets. The red points are locations of the CCRs and the black box indicates which arrays contain illuminated CCR for this transect.</p> Full article ">Figure 16
<p>Sequence of three CCR signal signatures at WSMR on 31 March 2019. Each group of detected photons represent multiple laser shots on one CCR on a specific height pole.</p> Full article ">Figure 17
<p>Along-track CCR signal distribution statistical extraction of the effective chord length based on the 2σ value for a Gaussian energy profile as a preliminary assessment of the optimal signal distribution during the March overpass of WSMR for a weak beam.</p> Full article ">Figure 18
<p>Analysis results for CCR geolocation accuracy recovery from 31 March 2018. The black points are the ATL03 reported photon geolocations and the green are the true locations using the CCR positions and the signal signature chord lengths.</p> Full article ">Figure 19
<p>Overview of the 28 September 2019 overpass of WSMR CCR arrays. The ground tracks shown here have been adjusted using the airborne lidar survey comparison technique to correct for horizontal geolocation offsets. The red points are locations of the CCRs and the black box indicates which arrays contain illuminated CCRs.</p> Full article ">Figure 20
<p>Analysis results for CCR geolocation accuracy recovery from 28 September 2019. The black points are the ATL03 reported photon geolocations and the green are the true locations using the CCR positions and the signal signature chord lengths.</p> Full article ">Figure 21
<p>ATL03 geolocation signal photons for track 1111 at 88S. The green points are signal, the blue circles are those signal determined to be from the CCRs and the red x’s are the actual CCR positions.</p> Full article ">
29 pages, 7949 KiB  
Article
Uncertainty and Overfitting in Fluvial Landform Classification Using Laser Scanned Data and Machine Learning: A Comparison of Pixel and Object-Based Approaches
by Zsuzsanna Csatáriné Szabó, Tomáš Mikita, Gábor Négyesi, Orsolya Gyöngyi Varga, Péter Burai, László Takács-Szilágyi and Szilárd Szabó
Remote Sens. 2020, 12(21), 3652; https://doi.org/10.3390/rs12213652 - 7 Nov 2020
Cited by 13 | Viewed by 4624
Abstract
Floodplains are valuable scenes of water management and nature conservation. A better understanding of their geomorphological characteristic helps to understand the main processes involved. We performed a classification of floodplain forms in a naturally developed area in Hungary using a Digital Terrain Model [...] Read more.
Floodplains are valuable scenes of water management and nature conservation. A better understanding of their geomorphological characteristic helps to understand the main processes involved. We performed a classification of floodplain forms in a naturally developed area in Hungary using a Digital Terrain Model (DTM) of aerial laser scanning. We derived 60 geomorphometric variables from the DTM and prepared a geomorphological map of 265 forms (crevasse channels, point bars, swales, levees). Random Forest classification was conducted with Recursive Feature Elimination (RFE) on the objects (mean pixel values by forms) and on the pixels of the variables. We also evaluated the classification probabilities (CP), the spatial uncertainties (SU), and the overfitting in the function of the number of the variables. We found that the object-based method had a better performance (95%) than the pixel-based method (78%). RFE helped to identify the most important 13–20 variables, maintaining the high model performance and reducing the overfitting. However, CP and SU were not efficient measures of classification accuracy as they were not in accordance with the class level accuracy metric. Our results help to understand classification results and the specific limits of laser scanned DTMs. This methodology can be useful in geomorphologic mapping. Full article
(This article belongs to the Special Issue Remote Sensing for Geomorphological Mapping)
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>The location of the study site and the fluvial forms.</p> Full article ">Figure 2
<p>The workflow of the analysis.</p> Full article ">Figure 3
<p>Correlation plot of geomorphometric variables.</p> Full article ">Figure 4
<p>The variables that contributed to reaching the maximum OA in the pixel-based-approach.</p> Full article ">Figure 5
<p>The variables that contributed to reaching the maximum OA in the object-oriented-approach.</p> Full article ">Figure 6
<p>Overall accuracy and number of variables according to the Recursive Feature Elimination variable selection method in pixel-based (<b>a</b>) and object-oriented (<b>b</b>) methods (10-fold cross-validation with 3 repetitions, i.e., 30 models; <span style="color:#5B9BD5">•</span>: highest OA).</p> Full article ">Figure 7
<p>Classification accuracies of different variable sets using a 10-fold cross-validation with 3 repetitions (i.e., 30 models; (<b>a</b>) PB: pixel-based approach; (<b>b</b>) OO: object-oriented approach).</p> Full article ">Figure 8
<p>Landform map of the floodplain using 2 (<b>a</b>) 4 (<b>b</b>) and 20 (<b>c</b>) variables based on the pixel-based approach.</p> Full article ">Figure 9
<p>Landform map of the floodplain using 2 (<b>a</b>) and 13 (<b>b</b>) variables based on the object-oriented approach.</p> Full article ">Figure 10
<p>Maximum probabilities of PB-classifications by fluvial forms in the function of the number of variables (mean ± standard error; v: number of variables).</p> Full article ">Figure 11
<p>F1 class level metric of PB-classifications by fluvial forms in the function of the number of variables (v: number of variables).</p> Full article ">Figure 12
<p>Maximum probability values of landforms calculated by the Random Forest classifier (i.e., these values belonged to the classified pixels; (<b>a</b>): 2-variable, (<b>b</b>): 4-variable, (<b>c</b>): 20-variable solutions). We used a composite to visualize the results (red band: levees; green band: point bar; blue band: swale; and the black color was the crevasse channel).</p> Full article ">Figure 13
<p>Maximum probabilities of OO-classifications by fluvial forms in the function of the number of variables (mean ± standard error; v: number of variables).</p> Full article ">Figure 14
<p>F1 class level metric of OO-classifications by fluvial forms in the function of the number of variables (v: number of variables).</p> Full article ">Figure 15
<p>Proportion of spatially uncertain pixels (classified into different types) related to the total number of pixels by fluvial forms, based on 10 repetitions.</p> Full article ">Figure 16
<p>Change in overfitting and the number of variables in object-based (OO) and pixel-based (PB) approaches.</p> Full article ">
25 pages, 6557 KiB  
Article
Change Points Detected in Decadal and Seasonal Trends of Outlet Glacier Terminus Positions across West Greenland
by Ashley V. York, Karen E. Frey, Sadegh Jamali and Sarah B. Das
Remote Sens. 2020, 12(21), 3651; https://doi.org/10.3390/rs12213651 - 7 Nov 2020
Cited by 3 | Viewed by 3288
Abstract
We investigated the change in terminus position between 1985 and 2015 of 17 marine-terminating glaciers that drain into Disko and Uummannaq Bays, West Greenland, by manually digitizing over 5000 individual frontal positions from over 1200 Landsat images. We find that 15 of 17 [...] Read more.
We investigated the change in terminus position between 1985 and 2015 of 17 marine-terminating glaciers that drain into Disko and Uummannaq Bays, West Greenland, by manually digitizing over 5000 individual frontal positions from over 1200 Landsat images. We find that 15 of 17 glacier termini retreated over the study period, with ~80% of this retreat occurring since 2000. Increased frequency of Landsat observations since 2000 allowed for further investigation of the seasonal variability in terminus position. We identified 10 actively retreating glaciers based on a significant positive relationship between glaciers with cumulative retreat >300 m since 2000 and their average annual amplitude (seasonal range) in terminus position. Finally, using the Detecting Breakpoints and Estimating Segments in Trend (DBEST) program, we investigated whether the 2000–2015 trends in terminus position were explained by the occurrence of change points (significant trend transitions). Based on the change point analysis, we found that nine of 10 glaciers identified as actively retreating also underwent two or three periods of change, during which their terminus positions were characterized by increases in cumulative retreat. Previous literature suggests potential relationships between our identified change dates with anomalous ocean conditions, such as low sea ice concentration and high sea surface temperatures, and our change durations with individual fjord geometry. Full article
(This article belongs to the Special Issue Applications of Remote Sensing in Glaciology)
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>The Disko and Uummannaq Bay study area including reference letters as well as official glacier names and ID numbers [<a href="#B27-remotesensing-12-03651" class="html-bibr">27</a>]. Yellow lines represent terminus positions in August 2014. Background images are from Landsat 8 (August 2014) with mean monthly ice velocity (2014–2015) also shown [<a href="#B30-remotesensing-12-03651" class="html-bibr">30</a>].</p> Full article ">Figure 2
<p>Example of digitized end-of-summer (Aug/Sep) terminus positions for each glacier for each year over the 1985–2015 study period. Background images from Landsat 8 (August 2014).</p> Full article ">Figure 3
<p>(<b>a</b>) Cumulative glacier terminus position (m) for Uummannaq Bay (glaciers a–j) over the 1985–2015 period calculated from 0 at the first observation in 1985. Dotted lines indicate data gaps. (<b>b</b>) Disko Bay (glaciers k–q).</p> Full article ">Figure 4
<p>(<b>a</b>) Cumulative glacier terminus retreat (red) or advance (blue) (m) over the 2000–2015 period. Yellow lines represent terminus positions in August 2014. Background images from Landsat 8 (August 2014). (<b>b</b>) Average annual amplitude in terminus position (m) over the 2000–2015 period. Yellow lines represent terminus positions in August 2014. Background images from Landsat 8 (August 2014).</p> Full article ">Figure 4 Cont.
<p>(<b>a</b>) Cumulative glacier terminus retreat (red) or advance (blue) (m) over the 2000–2015 period. Yellow lines represent terminus positions in August 2014. Background images from Landsat 8 (August 2014). (<b>b</b>) Average annual amplitude in terminus position (m) over the 2000–2015 period. Yellow lines represent terminus positions in August 2014. Background images from Landsat 8 (August 2014).</p> Full article ">Figure 5
<p>The annual amplitude in terminus position (m) for each glacier, in north to south order from top to bottom, for each year over the 2000–2015 period.</p> Full article ">Figure 6
<p>(<b>a</b>) For glaciers (a–j) in Uummannaq Bay, dark black lines indicate cumulative glacier terminus position (m; left axis) over the 2000–2015 period, with * representing actual date of Landsat observations. Gray lines indicate the annual amplitude in terminus position (m; right axis). Red squares and lines indicate the start through end date of a change duration detected by the DBEST program in R software. Please note the different y axes scales. (<b>b</b>) Glaciers (k–q) in Disko Bay.</p> Full article ">Figure 7
<p>Cumulative terminus change (m) compared to the average annual amplitude in terminus position (m) over the 2000–2015 period. Letters refer to individual glaciers (<a href="#remotesensing-12-03651-t0A1" class="html-table">Table A1</a>). Active retreat (hollow circles) are those glaciers that have retreated &gt;300 m (gray dashed line), while stable glaciers (black circles) are those which have advanced or retreated &lt;300 m.</p> Full article ">Figure 8
<p>Letters refer to individual glaciers (<a href="#remotesensing-12-03651-t0A1" class="html-table">Table A1</a>). For the actively retreating (&gt;300 m) glaciers defined in <a href="#remotesensing-12-03651-f007" class="html-fig">Figure 7</a>, the cumulative terminus change (m) and average annual amplitude in terminus position (m) within change durations (hollow circles) and outside change durations (black circles), as defined by the start date and end dates of change durations in DBEST (e.g., red squares in <a href="#remotesensing-12-03651-f006" class="html-fig">Figure 6</a>a;b).</p> Full article ">Figure A1
<p>Outputs of the decomposition process within DBEST showing from top to bottom, the input data, the trend component, the seasonal component, and the remainder component, for (<b>a</b>) glaciers (a, d, e, f, h, and m) and (<b>b</b>) glaciers (n, o, p, and q) with identified change points.</p> Full article ">Figure A1 Cont.
<p>Outputs of the decomposition process within DBEST showing from top to bottom, the input data, the trend component, the seasonal component, and the remainder component, for (<b>a</b>) glaciers (a, d, e, f, h, and m) and (<b>b</b>) glaciers (n, o, p, and q) with identified change points.</p> Full article ">
24 pages, 5780 KiB  
Article
Spatio-Temporal Assessment of Global Precipitation Products over the Largest Agriculture Region in Pakistan
by Zain Nawaz, Xin Li, Yingying Chen, Naima Nawaz, Rabia Gull and Abdelrazek Elnashar
Remote Sens. 2020, 12(21), 3650; https://doi.org/10.3390/rs12213650 - 6 Nov 2020
Cited by 9 | Viewed by 3958
Abstract
Spatial and temporal precipitation data acquisition is highly important for hydro-meteorological applications. Gridded precipitation products (GPPs) offer an opportunity to estimate precipitation at different time and resolution. Though, the products have numerous discrepancies that need to be evaluated against in-situ records. The present [...] Read more.
Spatial and temporal precipitation data acquisition is highly important for hydro-meteorological applications. Gridded precipitation products (GPPs) offer an opportunity to estimate precipitation at different time and resolution. Though, the products have numerous discrepancies that need to be evaluated against in-situ records. The present study is the first of its kind to highlight the performance evaluation of gauge based (GB) and satellite based (SB) GPPs at annual, winter, and summer monsoon scale by using multiple statistical approach during the period of 1979–2017 and 2003–2017, respectively. The result revealed that the temporal magnitude of all the GPPs was different and deviate up to 100–200 mm with overall spatial pattern of underestimation (GB product) and overestimation (SB product) from north to south gradient. The degree of accuracy of GB products with observed precipitation decreases with the increase in the magnitude of precipitation and vice versa for SB precipitation products. Furthermore, the observed precipitation revealed the positive trend with multiple turning points during the period 1979–2005. However, the gentle increase with no obvious break point has been detected during the period of 2005–2017. The large inter-annual variability and trends slope of the reference data series were well captured by Global Precipitation Climatology Centre (GPCC) and Tropical Rainfall Measuring Mission (TRMM) products and outperformed the relative GPPs in terms of higher R2 values of ≥ 0.90 and lower values of estimated RME ≤ 25% at annual and summer monsoon season. However, Climate Research Unit (CRU) performed better during winter estimates as compared with in-situ records. In view of significant error and discrepancies, regional correction factors for each GPPs were introduced that can be useful for future concerned projects over the study region. The study highlights the importance of evaluation by the careful selection of potential GPPs for the future hydro-climate studies over the similar regions like Punjab Province. Full article
(This article belongs to the Special Issue Remote Sensing of Precipitation: Part II)
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>Study area and meteorological stations.</p> Full article ">Figure 2
<p>Mean annual variation in GPPs and reference data.</p> Full article ">Figure 3
<p>Spatial distribution of annual average of (<b>A</b>) GB and (<b>B</b>) SB precipitation products during the whole study periods.</p> Full article ">Figure 4
<p>Statistical indicators for the assessment of (<b>A</b>) (GB) and (<b>B</b>) (SB) precipitation products against reference data (Annual timescale).</p> Full article ">Figure 5
<p>Statistical indicators for the assessment of (<b>A</b>) (GB) and (<b>B</b>) (SB) precipitation products against reference data (winter monsoon).</p> Full article ">Figure 6
<p>Statistical indicators for the assessment of (<b>A</b>) (GB) and (<b>B</b>) (SB) precipitation products against reference data (summer monsoon).</p> Full article ">Figure 7
<p>Statistical evaluation of (<b>A</b>) (GB) and (<b>B</b>) (SB) precipitation products against reference data by using Taylor diagram.</p> Full article ">Figure 8
<p>Spatial statistical indicators for the assessment of (<b>A</b>) (GB) and (<b>B</b>) (SB) precipitation products against reference data.</p> Full article ">Figure 9
<p>Comparative trend assessment of (<b>A</b>) (GB) and (<b>B</b>) (SB) precipitation products and reference data by using MK test (95% confidence Interval).</p> Full article ">Figure 10
<p>Comparison of Abrupt change detection in (<b>A</b>) (GB) and (<b>B</b>) (SB) precipitation products and reference data by using SQMK test (95% confidence Interval).</p> Full article ">
21 pages, 7590 KiB  
Article
Estimates of Forest Canopy Height Using a Combination of ICESat-2/ATLAS Data and Stereo-Photogrammetry
by Xiaojuan Lin, Min Xu, Chunxiang Cao, Yongfeng Dang, Barjeece Bashir, Bo Xie and Zhibin Huang
Remote Sens. 2020, 12(21), 3649; https://doi.org/10.3390/rs12213649 - 6 Nov 2020
Cited by 36 | Viewed by 5606
Abstract
Forest canopy height is an indispensable forest vertical structure parameter for understanding the carbon cycle and forest ecosystem services. A variety of studies based on spaceborne Lidar, such as ICESat, ICESat-2 and airborne Lidar, were conducted to estimate forest canopy height at multiple [...] Read more.
Forest canopy height is an indispensable forest vertical structure parameter for understanding the carbon cycle and forest ecosystem services. A variety of studies based on spaceborne Lidar, such as ICESat, ICESat-2 and airborne Lidar, were conducted to estimate forest canopy height at multiple scales. However, while a few studies have been conducted based on ICESat-2 simulated data from airborne Lidar data, few studies have analyzed ATL08 and ATL03 products derived from the ATLAS sensor onboard ICESat-2 for regional vegetation canopy height mapping. It is necessary and promising to explore how data obtained by ICESat-2 can be applied to estimate forest canopy height. This study proposes a new means to estimate forest canopy height, defined as the mean height of trees within a given forest area, using a combination of ICESat-2 ATL08 and ATL03 data and ZY-3 satellite stereo images. Five procedures were used to estimate the forest canopy height of the city of Nanning in China: (1) Processing ground photons in a 30 m × 30 m grid; (2) Extracting a digital surface model (DSM) using ZY-3 stereo images; (3) Calculating a discontinuous canopy height model (CHM) dataset; (4) Validating the DSM and ground photon height using GEDI data; (5) Estimating the regional wall-to-wall forest canopy height product based on the backpropagation artificial neural network (BP-ANN) model and Landsat 8 vegetation indices and independent accuracy assessments with field measured plots. The validation shows a root mean square error (RMSE) of 3.34 m to 3.47 m and a coefficient of determination R2 = 0.51. The new method shows promise and can be used for large-scale forest canopy height mapping at various resolutions or in combination with other data, such as SAR images. Finally, this study analyzes resolutions and how to filter effective data when ATL08 data are directly used to generate regional or global vegetation height products, which will be the focus of future research. Full article
(This article belongs to the Special Issue Advances in LiDAR Remote Sensing for Forestry and Ecology)
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>Location of the city of Nanning, the study area. The spot mark denotes the position of the field survey sample. Red boxes show the locations of ZY-3 coverage. The Shuttle Radar Topography Mission (SRTM) refers to the value of the elevation product collected by a C-band radar interferometry system.</p> Full article ">Figure 2
<p>ICESat-2 ATLAS ground tracks within the city of Nanning (a shows the distribution of ATL08 data in the study area, b and c show the distribution of ATL03 photons and ZY-3).</p> Full article ">Figure 3
<p>GEDI and ICESat-2 ground tracks. Red dots indicate that the ground photons average value and the GEDI footprint are located within the same 30 m × 30 m pixels of Landsat 8 images, that is, the intersection points (more details are in <a href="#sec2dot3dot1-remotesensing-12-03649" class="html-sec">Section 2.3.1</a>). In pictures (<b>a</b>–<b>d</b>), red dots denote the same. (<b>a</b>) shows the overall distribution while (<b>b</b>–<b>d</b>) show enlarged displays. The green lines in all images are made up of dense GEDI footprints.</p> Full article ">Figure 4
<p>Methodology flowchart. The sequence of the five main steps is numbered and highlighted.</p> Full article ">Figure 5
<p>Examples of locations of ground photons and ground photon average values within a 30 m × 30 m grid. Grid division is based on the row and column IDs of Landsat 8 data. The scale bar at the bottom right of the image matches that shown in (<b>a</b>). (<b>a</b>) shows the spatial distribution of ATL03 ground photons and ZY-3 data, and (<b>b</b>,<b>c</b>) are the positions of ground photons average value and ground photons in each pixel.</p> Full article ">Figure 6
<p>The vertical section profile of the discontinuous canopy height model (CHM) dataset. The interval between the dotted lines indicates a grid resolution of 30 m.</p> Full article ">Figure 7
<p>Photon classification map (on the <b>right</b>) corresponding to atl08 data (on the <b>left</b>).</p> Full article ">Figure 8
<p>ZY-3 DSM masked by two forest area masks and score bands. Discontinuous CHM dataset calculated by DSM subtracting ground photon values. (<b>a</b>,<b>b</b>) shows the distribution of DSM, and (<b>c</b>,<b>d</b>) shows the distribution of discontinuous CHM dataset.</p> Full article ">Figure 9
<p>Comparison of DSM and ground photon average height values with GEDI data. (<b>a</b>) shows the relationship between GEDI terrain elevation and ground photon terrain elevation; (<b>b</b>) shows the relationship between GEDI land surface elevation and ZY-3 DSM; (<b>c</b>) shows the relationship between the SRTM DEM and ground photon terrain elevation.).</p> Full article ">Figure 10
<p>Forest canopy height estimated by the BP-ANN with training samples of the discontinuous CHM dataset.</p> Full article ">Figure 11
<p>Independent accuracy validation for the estimated forest canopy height. Blue dots in (<b>a</b>) were derived from part of the CHM dataset; blue dots in (<b>b</b>) were derived from field measured plots; (<b>c</b>) combines the blue dots from (<b>a</b>,<b>b</b>).</p> Full article ">Figure 11 Cont.
<p>Independent accuracy validation for the estimated forest canopy height. Blue dots in (<b>a</b>) were derived from part of the CHM dataset; blue dots in (<b>b</b>) were derived from field measured plots; (<b>c</b>) combines the blue dots from (<b>a</b>,<b>b</b>).</p> Full article ">Figure 12
<p>Frequency statistics histogram of canopy and ground photons. The <span class="html-italic">x</span>-axis in (<b>a</b>,<b>b</b>) represents the number range of signal photons (canopy and ground photons), and the <span class="html-italic">y</span>-axis in (<b>a</b>,<b>b</b>) represents the frequency of signal photons within 100 m segments of all ATL08 records in the study area.</p> Full article ">
19 pages, 7262 KiB  
Article
Improving Stage–Discharge Relation in The Mekong River Estuary by Remotely Sensed Long-Period Ocean Tides
by Hongrui Peng, Hok Sum Fok, Junyi Gong and Lei Wang
Remote Sens. 2020, 12(21), 3648; https://doi.org/10.3390/rs12213648 - 6 Nov 2020
Cited by 8 | Viewed by 2824
Abstract
Ocean tidal backwater reshapes the stage–discharge relation in the fluvial-to-marine transition zone at estuaries, rendering the cautious use of these data for hydrological studies. While a qualitative explanation is traditionally provided by examining a scatter plot of water discharge against water level, a [...] Read more.
Ocean tidal backwater reshapes the stage–discharge relation in the fluvial-to-marine transition zone at estuaries, rendering the cautious use of these data for hydrological studies. While a qualitative explanation is traditionally provided by examining a scatter plot of water discharge against water level, a quantitative assessment of long-period ocean tidal effect on the stage–discharge relation has been rarely investigated. This study analyzes the relationship among water level, water discharge, and ocean tidal height via their standardized forms in the Mekong Delta. We found that semiannual and annual components of ocean tides contribute significantly to the discrepancy between standardized water level and standardized water discharge time series. This reveals that the long-period ocean tides are the significant factors influencing the stage–discharge relation in the river delta, implying a potential of improving the relation as long as proper long-period ocean tidal components are taken into consideration. By isolating the short-period signals (i.e., less than 15 days) from land surface hydrology and ocean tides, better consistent stage–discharge relations are obtained, in terms of improving the Pearson correlation coefficient (PCC) from ~0.4 to ~0.8 and from ~0.6 to ~0.9 for the stations closest to the estuary and at the Mekong Delta entrance, respectively. By incorporating the long-period ocean tidal height time series generated from a remotely sensed global ocean tide model into the stage–discharge relation, further refined stage–discharge relations are obtained with the PCC higher than 0.9 for all employed stations, suggesting the improvement of daily averaged water level and water discharge while ignoring the short-period intratidal variability. The remotely sensed global ocean tide model, OSU12, which contains annual and semiannual ocean tide components, is capable of generating accurate tidal height time series necessary for the partial recovery of the stage–discharge relation. Full article
(This article belongs to the Special Issue Coastal Waters Monitoring Using Remote Sensing Technology)
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>Map of Mekong Delta (MD), with two pairs of hydrological gauge stations (i.e., Can Tho and Chau Doc, and My Thuan and Tan Chau) situated near the estuaries. (The topography dataset, called earth_relief_30s, is a derived product of SRTM15+ [<a href="#B38-remotesensing-12-03648" class="html-bibr">38</a>], which is obtainable from <a href="http://mirrors.ustc.edu.cn/gmt/data/" target="_blank">http://mirrors.ustc.edu.cn/gmt/data/</a>).</p> Full article ">Figure 2
<p>Low-pass filtered (blue) and original (blue dash) time series of water discharge and water level (red) over (<b>a</b>) Can Tho, (<b>b</b>) My Thuan, (<b>c</b>) Chau Doc, and (<b>d</b>) Tan Chau stations, respectively, and (<b>e</b>) time series of ocean tidal height (sea level) at Vung Tau station spanning from January 2003 to December 2006.</p> Full article ">Figure 3
<p>Spectra of the (<b>a</b>) hourly and (<b>b</b>) daily averaged ocean tidal height time series in Vung Tau tide gauge station.</p> Full article ">Figure 4
<p>(<b>a</b>–<b>d</b>) Relationship between water level (WL) and water discharge (WD) (original daily sampled time series) for the four selected hydrological stations in Mekong Delta.</p> Full article ">Figure 5
<p>Comparison of standardized WD, WL, and tidal height time series in (<b>a</b>) Can Tho, (<b>b</b>) Chau Doc, (<b>c</b>) My Thuan, and (<b>d</b>) Tan Chau station, respectively.</p> Full article ">Figure 6
<p>(<b>a</b>–<b>d</b>) Relationship between WL and WD (low pass filtered time series) for the selected four hydrological stations in the Mekong Delta.</p> Full article ">Figure 7
<p>(<b>a</b>–<b>d</b>) Relationship between <math display="inline"><semantics> <mrow> <msub> <mrow> <mi>WL</mi> </mrow> <mrow> <mi>free</mi> </mrow> </msub> </mrow> </semantics></math> and WD (low pass filtered time series) for the selected four hydrological stations in the Mekong Delta.</p> Full article ">Figure 8
<p>WD and tide-free WL time series from 2003 to 2006 in (<b>a</b>) Can Tho and (<b>b</b>) My Thuan stations.</p> Full article ">Figure 9
<p>The comparison between OSU12 model-derived WL and in-situ WL at (<b>a</b>) Can Tho and (<b>b</b>) My Thuan during 2003–2006.</p> Full article ">Figure 10
<p>Recovered rating curves at (<b>a</b>) Can Tho, (<b>b</b>) My Thuan, (<b>c</b>) Chau Doc, and (<b>d</b>) Tan Chau stations using model-derived ocean tidal height as input.</p> Full article ">Figure 11
<p>(<b>a</b>) Different PCC (presented in color bar) for different <math display="inline"><semantics> <mi mathvariant="normal">b</mi> </semantics></math>, <math display="inline"><semantics> <mi mathvariant="normal">c</mi> </semantics></math> and <math display="inline"><semantics> <mi>α</mi> </semantics></math> using time series from Can Tho station, and (<b>b</b>) slices of (<b>a</b>) for nine chosen α, with maximum PCC for each α shown from the above subplots.</p> Full article ">Figure 12
<p>(<b>a</b>,<b>b</b>) Stage discharge relation from original WL, and (<b>c</b>,<b>d</b>) tide-free WL for Can Tho and My Thuan stations.</p> Full article ">
18 pages, 9460 KiB  
Article
Oil Spill Detection and Mapping: A 50-Year Bibliometric Analysis
by Rodrigo N. Vasconcelos, André T. Cunha Lima, Carlos A. D. Lentini, Garcia V. Miranda, Luís F. Mendonça, Marcus A. Silva, Elaine C. B. Cambuí, José M. Lopes and Milton J. Porsani
Remote Sens. 2020, 12(21), 3647; https://doi.org/10.3390/rs12213647 - 6 Nov 2020
Cited by 41 | Viewed by 5249
Abstract
Oil spill detection and mapping (OSPM) is an extremely relevant issue from a scientific point of view due to the environmental impact on coastal and marine ecosystems. In this study, we present a new approach to assess scientific literature for the past 50 [...] Read more.
Oil spill detection and mapping (OSPM) is an extremely relevant issue from a scientific point of view due to the environmental impact on coastal and marine ecosystems. In this study, we present a new approach to assess scientific literature for the past 50 years. In this sense, our study aims to perform a bibliometric and network analysis using a literature review on the application of OSPM to assess researchers and trends in this field of science. In methodological terms we used the Scopus base to search for articles in the literature, then we used bibliometric tools to access information and reveal quantifying patterns in this field of literature. Our results suggest that the detection of oil in the sea has undergone a great evolution in the last decades and there is a strong relationship between the technological evolution aimed at detection with the improvement of remote sensing data acquisition methods. The most relevant contributions in this field of science involved countries such as China, the United States, and Canada. We revealed aspects of great importance and interest in OSPM literature using a bibliometric and network approach to give a clear overview of this field’s research trends. Full article
(This article belongs to the Special Issue Remote Sensing Applications in Ocean Observation)
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>Logical scheme, methodological approach and data analysis for Phases 1 (blue color) and 2 (yellow color).</p> Full article ">Figure 2
<p>(<b>A</b>) Annual growth rate of OSPM publications (black curve, left y-axis) compared to the cumulative annual growth (red curve, right y-axis) rate of all the manuscripts indexed in the Scopus database per year (1970–2019). (<b>B</b>) Boxplot by decades. Red diamonds inside the polygons represent the mean.</p> Full article ">Figure 3
<p>Word co-occurrence network built for the top 25 papers using words presented in titles, abstracts, keywords, and general feature information of the 50 years of documents published between 1970–2019.</p> Full article ">Figure 4
<p>Total of publications on OSPM and co-authoring collaboration network by countries from documents published between 1970–2019.</p> Full article ">Figure 5
<p>Word co-occurrence network presented in titles, abstracts and keywords for each decade (1970s, 1980s, 1990s, 2000s, and 2010s), as well as for all the scientific articles published in the 50-year time frame (1970–2019).</p> Full article ">Figure 6
<p>Similar to <a href="#remotesensing-12-03647-f004" class="html-fig">Figure 4</a>, except for the topology metrics. Letters (<b>A</b>–<b>H</b>) indicate a topological metric described in <a href="#remotesensing-12-03647-t002" class="html-table">Table 2</a>.</p> Full article ">Figure 7
<p>Number of spill events (<b>A</b>) and amount of oil spill (<b>B</b>, in thousand tonnes) over last 50 years (i.e., 1970–2019 record) based on ITOPF data report and the Max Roser database. Red line indicates the mean of each decade, respectively.</p> Full article ">
15 pages, 6831 KiB  
Article
A More Reliable Orbit Initialization Method for LEO Precise Orbit Determination Using GNSS
by Xuewen Gong, Jizhang Sang, Fuhong Wang and Xingxing Li
Remote Sens. 2020, 12(21), 3646; https://doi.org/10.3390/rs12213646 - 6 Nov 2020
Cited by 2 | Viewed by 2425
Abstract
Precise orbit determination (POD) using GNSS has been rapidly developed and is the mainstream technology for the navigation of low Earth orbit (LEO) satellites. The initialization of orbit parameters is a key prerequisite for LEO POD processing. For a LEO satellite equipped with [...] Read more.
Precise orbit determination (POD) using GNSS has been rapidly developed and is the mainstream technology for the navigation of low Earth orbit (LEO) satellites. The initialization of orbit parameters is a key prerequisite for LEO POD processing. For a LEO satellite equipped with a GNSS receiver, sufficient discrete kinematic positions can be obtained easily by processing space-borne GNSS data, and its orbit parameters can thus be estimated directly in iterative manner. This method of direct iterative estimation is called as the direct approach, which is generally considered highly reliable, but in practical applications it has risk of failure. Stability analyses demonstrate that the direct approach is sensitive to oversized errors in the starting velocity vector at the reference time, which may lead to large errors in design matrix because the reference orbit may be significantly distorted, and eventually cause the divergence of the orbit parameter estimation. In view of this, a more reliable method, termed the progressive approach, is presented in this paper. Instead of estimating the orbit parameters directly, it first fits the discrete kinematic positions to a reference ephemeris in the form of the GNSS broadcast ephemeris, which construct a reference orbit that is smooth and close to the true orbit. Based on the reference orbit, the starting orbit parameters are computed in sufficient accuracy, and then the final orbit parameters are estimated with a high accuracy by using discrete kinematic positions as measurements. The stability analyses show that the design matrix errors are reduced in the progressive approach, which would assure more robust orbit parameter estimation than the direct estimation approach. Various orbit initialization experiments are performed on the KOMPSAT-5 and FY3C satellites. The results have fully verified the high reliability of the proposed progressive approach. Full article
(This article belongs to the Special Issue Advances in GNSS Data Processing and Navigation)
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>The iterative estimation scheme of the direct approach.</p> Full article ">Figure 2
<p>The estimation scheme of the progressive approach.</p> Full article ">Figure 3
<p>The design errors due to starting state errors in orbit parameters estimation.</p> Full article ">Figure 3 Cont.
<p>The design errors due to starting state errors in orbit parameters estimation.</p> Full article ">Figure 4
<p>The design errors due to the starting state error in reference ephemeris estimation.</p> Full article ">Figure 5
<p>The estimated altitude of KOMPSAT-5 after each iteration step of orbit parameter estimation by direct approach.</p> Full article ">Figure 6
<p>The altitude and position error of the reference orbit of FY3C on 2013/287.</p> Full article ">
22 pages, 10625 KiB  
Article
Mapping the Population Density in Mainland China Using NPP/VIIRS and Points-Of-Interest Data Based on a Random Forests Model
by Yunchen Wang, Chunlin Huang, Minyan Zhao, Jinliang Hou, Ying Zhang and Juan Gu
Remote Sens. 2020, 12(21), 3645; https://doi.org/10.3390/rs12213645 - 6 Nov 2020
Cited by 38 | Viewed by 5142
Abstract
Understanding the spatial distribution of populations at a finer spatial scale has important value for many applications, such as disaster risk rescue operations, business decision-making, and regional planning. In this study, a random forest (RF)-based population density mapping method was proposed in order [...] Read more.
Understanding the spatial distribution of populations at a finer spatial scale has important value for many applications, such as disaster risk rescue operations, business decision-making, and regional planning. In this study, a random forest (RF)-based population density mapping method was proposed in order to generate high-precision population density data with a 100 m × 100 m grid in mainland China in 2015 (hereafter referred to as ‘Popi’). Besides the commonly used elevation, slope, Normalized Vegetation Index (NDVI), land use/land cover, roads, and National Polar Orbiting Partnership/Visible Infrared Imaging Radiometer Suite (NPP/VIIRS), 16,101,762 records of points of interest (POIs) and 2867 county-level censuses were used in order to develop the model. Furthermore, 28,505 township-level censuses (74% of the total number of townships) were collected in order to evaluate the accuracy of the Popi product. The results showed that the utilization of multi-source data (especially the combination of POIs and NPP/VIIRS data) can effectively improve the accuracy of population mapping at a finer scale. The feature importances of the POIs and NPP/VIIRS are 0.49 and 0.14, respectively, which are higher values than those obtained for other natural factors. Compared with the Worldpop population dataset, the Popi data exhibited a higher accuracy. The number of accurately-estimated townships was 19,300 (67.7%) in the Popi product and 16,237 (56.9%) in the Worldpop product. The Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE) were 14,839 and 7218, respectively, for Popi, and 18,014 and 8572, respectively, for Worldpop. The research method in this paper could provide a reference for the spatialization of other socioeconomic data (such as GDP). Full article
(This article belongs to the Special Issue Remote Sensing Application to Population Mapping)
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>The classification of 28,505 township-level census data sets in China. The township census data of the Tianjin, Hong Kong, Macau, and Taiwan provinces are missing.</p> Full article ">Figure 2
<p>A flowchart of the mapping of the population density.</p> Full article ">Figure 3
<p>The NPP/VIIRS data in mainland China in 2015. (<b>a</b>) The raw NPP/VIIRS data; (<b>b</b>) the corrected NPP/VIIRS data.</p> Full article ">Figure 4
<p>The out-of-bag values at different bandwidths.</p> Full article ">Figure 5
<p>The out-of-bag values with different parameters. (<b>a</b>) Max features; (<b>b</b>) number of trees; (<b>c</b>) max depth; (<b>d</b>) min samples split; (<b>e</b>) min samples leaf.</p> Full article ">Figure 6
<p>Population density maps with a 100 m × 100 m grid for 2015 in (<b>a</b>) mainland China; and (<b>b</b>) typical cities.</p> Full article ">Figure 7
<p>The spatial distribution of the relative error values at the township level for the (<b>a</b>) Popi product and (<b>b</b>) Worldpop product. The township censuses of Tianjin, Hong Kong, Macau, and Taiwan provinces are missing. Besides this, when RE &lt;− 0.5, the township is considered seriously underestimated. A township with −0.5 &lt; RE &lt; −0.25 is considered to be a slight underestimation. Furthermore, −0.25 &lt; RE &lt; 0.25 is an accurate estimate. When 0.25 &lt; RE &lt; 0.5, it is slightly overestimated. A township with RE &gt; 0.5 is considered to be seriously overestimated.</p> Full article ">Figure 8
<p>The goodness of fit (R<sup>2</sup>) between the township census and the estimated township population of Worldpop and Popi products in 30 provinces (the R<sup>2</sup> values for the Tianjin, Hong Kong, Macau, and Taiwan provinces are missing).</p> Full article ">Figure 9
<p>The result of the relative error at the township level (the township censuses of the Tianjin, Hong Kong, Macau, and Taiwan provinces are missing).</p> Full article ">Figure 10
<p>The feature importance of the independent variables.</p> Full article ">Figure 11
<p>The census data with respect to the National Polar-orbiting Operational Environmental Satellite System Preparatory Project/Visible Infrared Imaging Radiometer (NPP/VIIRS) and point of interest (POI) at the city level (using urban construction land and rural construction land in the 2015 LULC data as masks, the built-up areas and rural areas of 340 cities in mainland China were extracted, respectively). (<b>a</b>) The census data with respect to the NPP/VIIRS data in built-up area; (<b>b</b>) the census data with respect to the POIs data in built-up area; (<b>c</b>) the census data with respect to the NPP/VIIRS data in county; (<b>d</b>) the census data with respect to the POIs data in county.</p> Full article ">Figure 12
<p>The relationship between the census and different type POI data at the city level. (<b>a</b>) POI density layer (the POI density layer is the combination of the 10 individual POI layers using the entropy weight method); (<b>b</b>) the company type POI data; (<b>c</b>) the education type POI data; (<b>d</b>) the food type POI data; (<b>e</b>) the financial type POI data; (<b>f</b>) the hotel type POI data; (<b>g</b>) the government type POI data; (<b>h</b>) the medical type POI data; (<b>i</b>) the village type POI data; (<b>j</b>) the shopping type POI data; (<b>k</b>) the traffic type POI data; (<b>l</b>) the POI density data.</p> Full article ">
Show export options expand_more Show export options expand_less
Select all
Export citation of selected articles as:
 
Displaying articles 1-206
Previous Issue
Next Issue
Back to TopTop