Dinh Ho Tong Minh

Mapping Ground Subsidence Phenomena in Ho Chi Minh City through the Radar Interferometry Technique Using ALOS PALSAR Data

The rapidly developing urbanization since the last decade of the 20th century has led to extensive groundwater extraction, resulting in subsidence in Ho Chi Minh City, Vietnam. Recent advances in multi-temporal spaceborne SAR interferometry, especially with a persistent scatters interferometry (PSI) approach, has made this a robust remote sensing technique for measuring large-scale ground subsidence with millimetric accuracy. This work has presented an advanced PSI analysis, to provide an unprecedented spatial extent and continuous temporal coverage of the subsidence in Ho Chi Minh City from 2006 to 2010. The study shows that subsidence is most severe in the Holocene silt loam areas along the Sai Gon River and in the southwest of the city. The groundwater extraction resulting from urbanization and urban growth is mainly responsible for the subsidence. Subsidence in turn leads to more flooding and water nuisance. The correlation between the reference leveling velocity and the estimated PSI result is R2 = 0.88, and the root mean square error is 4.3 (mm/year), confirming their good agreement. From 2006 to 2010, the estimation of the average subsidence rate is -8.0 mm/year, with the maximum value up to -70 mm/year. After four years, in regions along Sai Gon River and in the southwest of the city, the land has sunk up to -12 cm. If not addressed, subsidence leads to the increase of inundation, both in frequency and spatial extent. Finally, regarding climate change, the effects of subsidence should be considered as appreciably greater than those resulting from rising sea level. It is essential to consider these two factors, because the city is inhabited by more than 7.5 million people, where subsidence directly impacts urban structures and infrastructure.

Vertical Structure of P-Band Temporal Decorrelation at the Paracou Forest: Results From TropiScat

In this letter, we present the results from the ground-based European Space Agency campaign TropiScat, which is aimed at evaluating the temporal coherence at P-band in a tropical forest in all polarizations and at different heights within the vegetation layers. The TropiScat equipment has been operated since October 2011 at the Paracou field station, French Guiana, to continuously produce height-range images of the forest with a temporal sampling of 15 min. The forest temporal behavior can be then captured by analyzing the interferometric coherence between the images gathered at different times, considering time scales on the order of hours, days, and months. The results indicate that the vegetation is likely to undergo a significant motion during day hours due to wind and temperature changes, whereas it appears to be definitively more stable during night hours. This result appears to provide a very useful input to the Biomass Monitoring Mission for Carbon Assessment (BIOMASS), as it suggests that the performance over a tropical forest could be optimized by gathering acquisitions in early morning or night hours. The long-term temporal decorrelation has been then evaluated by considering dawn-dawn acquisitions to minimize the impact of wind gusts and by excluding rainy days in order to not confuse forest and system decorrelation. As a result, the temporal coherence at the ground level was found to stay high at about 0.8 at 27 days, whereas the temporal coherence at the canopy height was found to be about 0.8 at 4 days and about 0.65 at 27 days, indicating coherence sensitivity to height.

The Impact of Temporal Decorrelation on BIOMASS Tomography of Tropical Forests

The objective of this letter is to provide a better understanding of the impact of temporal decorrelation on the tomographic phase of the P-band synthetic aperture radar (SAR) mission BIOMASS, selected as the Seventh Earth Explorer by the European Space Agency. In the context of Phase A BIOMASS activities, the tropical forest site of Paracou, French Guiana, was illuminated at P-band during the airborne campaign TropiSAR 2009 and the ground-based campaign TropiScat 2011. P-band data from TropiSAR were used to generate a high-resolution 3-D reconstruction of the Paracou forest, whereas TropiScat data provided information about temporal correlation considering different time lags and different heights within the vegetation layer. The ensemble of the two datasets were used to generate a synthetic SAR data stack that emulates BIOMASS acquisitions over the Paracou forest site, accounting for BIOMASS geometry and resolution, as well as for the forest temporal decorrelation. Different data stacks were produced by varying the revisit time between two consecutive passes from 1 to 17 days. The resulting vertical structure reconstruction and forest height retrieval were observed to yield valuable results as long as the revisit time is 4 days or less.

Temporal Coherence of Tropical Forests at P-Band: Dry and Rainy Seasons

In this letter, the temporal coherence of tropical forest scattering at P-band is addressed by means of a ground-based experiment. The study is based on the TropiScat campaign in French Guiana, designed to support the Biomass mission, which will be the ESA 7th Earth Explorer mission. For Biomass, temporal coherence is a crucial parameter for coherent processing of polarimetric synthetic aperture radar (SAR) interferometry and SAR tomography in repeat-pass acquisitions. During the experiment, data were continuously collected for six months during both the rainy and dry seasons. For the rain-free days in both seasons, the coherence exhibits a daily cycle showing a high decorrelation during daytime, which is likely due to motion in the canopy. Up to a 20-day baseline, the coherence is much higher in the dry season than in the rainy season (> 0.8). From 20 to 40 days, it presents the same order of magnitude in both seasons [0.6, 0.7]. For larger temporal baselines, it becomes lower in the dry season. The results can be used to assess the long-term coherence of repeat-pass observations over a tropical forest. However, an extension of this study to several years and over other forest spots would be necessary to draw more general conclusions.

Deep Recurrent Neural Network for Agricultural Classification using multitemporal SAR Sentinel-1 for Camargue, France

The development and improvement of methods to map agricultural land cover are currently major challenges, especially for radar images. This is due to the speckle noise nature of radar, leading to a less intensive use of radar rather than optical images. The European Space Agency Sentinel-1 constellation, which recently became operational, is a satellite system providing global coverage of Synthetic Aperture Radar (SAR) with a 6-days revisit period at a high spatial resolution of about 20 m. These data are valuable, as they provide spatial information on agricultural crops. The aim of this paper is to provide a better understanding of the capabilities of Sentinel-1 radar images for agricultural land cover mapping through the use of deep learning techniques. The analysis is carried out on multitemporal Sentinel-1 data over an area in Camargue, France. The data set was processed in order to produce an intensity radar data stack from May 2017 to September 2017. We improved this radar time series dataset by exploiting temporal filtering to reduce noise, while retaining as much as possible the fine structures present in the images. We revealed that even with classical machine learning approaches (K nearest neighbors, random forest, and support vector machines), good performance classification could be achieved with F-measure/Accuracy greater than 86% and Kappa coefficient better than 0.82. We found that the results of the two deep recurrent neural network (RNN)-based classifiers clearly outperformed the classical approaches. Finally, our analyses of the Camargue area results show that the same performance was obtained with two different RNN-based classifiers on the Rice class, which is the most dominant crop of this region, with a F-measure metric of 96%. These results thus highlight that in the near future these RNN-based techniques will play an important role in the analysis of remote sensing time series.

Forest Biomass From Radar Remote Sensing

Abstract: Forests play a primordial role for life on Earth. Beyond their contribution as a major source of raw materials and renewable energy, they also hold an inestimable treasure of biodiversity. They ensure the protection of arable land, are a continuous source of water and contribute to improved air quality. Whether for food or pharmacopoeia, forests are the principal source of subsistence for almost 2 billion people.

Tomosar platform supports for Sentinel-1 tops persistent scatterers interferometry

Developing and improving methods to monitor both natural and non-natural environments such as forest and urban in space and time is a timely challenge. To overcome this challenge, we created a software platform — TomoSAR. The kernel of this platform supports the entire processing from SAR, Interferometry, Polarimetry, to Tomography (so called TomoSAR). The objective of this paper is to introduce this platform about its capability in Persistent Scatterers Interferometry (PSI) technique to estimate subsidence using TOPS Sentinel-1 data.

Rice height and biomass estimations using multitemporal SAR Sentinel-1: Camargue case study

The research and improvement of methods to be used for crop monitoring are currently major challenges, especially for radar images due to their speckle noise nature. The European Space Agency’s (ESA) Sentinel1 constellation provides synthetic aperture radar (SAR) images coverage with a 6 days revisit period at a high spatial resolution of pixel spacing 20 m. Sentinel-1 data are considerable useful, as they provide valuable information of the vegetation cover. The objective of this paper is to provide a better understanding of the capabilities of Sentinel-1 radar images for rice height and dry biomass retrievals. To do this, we train Sentinel1 data against ground measurements with classical machine learning techniques (Multiple Linear Regression (MLR), Support Vector Regression (SVR) and Random Forest (RF)) to estimate rice height and dry biomass. The study is carried out on a multi-temporal Sentinel-1 dataset acquired from May 2017 to September 2017 over the Camargue region, southern France. The ground in-situ measurements were made in the same period to collect rice height and dry biomass over 11 rice fields. The images were processed in order to produce an intensity radar data stack in C-band including dual-polarization VV (Vertical receive and Vertical transmit) and VH (Vertical receive and Horizontal transmit) data. We found that non-parametric methods (SVR and RF) had a better performance over the parametric MLR method for rice biophysical parameter retrievals. The accuracy of rice height estimation showed that rice height retrieval was strongly correlated to the in-situ rice height from dual-polarization, in which Random Forest yielded the best performance with correlation coefficient R2 = 0.92 and the root mean square error (RMSE) 16% (7.9 cm). In addition, we demonstrated that the correlation of Sentinel-1 signal to the biomass was also very high in VH polarization with R2 = 0.9 and RMSE = 18% (162 g.m−2 ) (with Random Forest method). Such results indicate that the highly qualified Sentinel-1 radar data could be well exploited for rice biomass and height retrieval and they could be used for operational tasks

Estimation of Rice Height and Biomass Using Multitemporal SAR Sentinel-1 for Camargue, Southern France

The research and improvement of methods to be used for crop monitoring are currently major challenges, especially for radar images due to their speckle noise nature. The European Space Agency’s (ESA) Sentinel-1 constellation provides synthetic aperture radar (SAR) images coverage with a 6-day revisit period at a high spatial resolution of pixel spacing of 20 m. Sentinel-1 data are considerably useful, as they provide valuable information of the vegetation cover. The objective of this work is to study the capabilities of multitemporal radar images for rice height and dry biomass retrievals using Sentinel-1 data. To do this, we train Sentinel-1 data against ground measurements with classical machine learning techniques (Multiple Linear Regression (MLR), Support Vector Regression (SVR) and Random Forest (RF)) to estimate rice height and dry biomass. The study is carried out on a multitemporal Sentinel-1 dataset acquired from May 2017 to September 2017 over the Camargue region, southern France. The ground in-situ measurements were made in the same period to collect rice height and dry biomass over 11 rice fields. The images were processed in order to produce a radar stack in C-band including dual-polarization VV (Vertical receive and Vertical transmit) and VH (Vertical receive and Horizontal transmit) data. We found that non-parametric methods (SVR and RF) had a better performance over the parametric MLR method for rice biophysical parameter retrievals. The accuracy of rice height estimation showed that rice height retrieval was strongly correlated to the in-situ rice height from dual-polarization, in which Random Forest yielded the best performance with correlation coefficient R2 = 0.92 and the root mean square error (RMSE) 16% (7.9 cm). In addition, we demonstrated that the correlation of Sentinel-1 signal to the biomass was also very high in VH polarization with R2 = 0.9 and RMSE = 18% (162 g·m−2) (with Random Forest method). Such results indicate that the highly qualified Sentinel-1 radar data could be well exploited for rice biomass and height retrieval and they could be used for operational tasks.

Applying deep learning for agricultural classification using multitemporal SAR Sentinel-1 for Camargue, France

The aim of this paper is to provide a better understanding of potentialities of the new Sentinel-1 radar images for mapping the different crops in the Camargue region in the South France. The originality relies on deep learning techniques. The analysis is carried out on multitemporal Sentinel-1 data over an area in Camargue,France.50 Sentinel-1 images processed in order to produce an intensity radar data stack from May 2017 to September 2017. We revealed that even with classical machine learning approaches (K nearest neighbors, random forest, and support vector machine), good performance classification could be achieved with F-measure/Accuracy greater than 86 % and Kappa coefficient better than 0.82. We found that the results of the two deep recurrent neural network (RNN)-based classifiers clearly outperformed the classical approaches. Finally, our analyses of Camargue area results show that the same performance was obtained with two different RNN-based classifiers on the Rice class, which is the most dominant crop of this region, with a F-measure metric of 96 %. These results thus highlight that in the near future, these RNN-based techniques will play an important role in the analysis of remote sensing time series.