Nhung Thi Le

Dr. Nhung Thi Le

Geodesy, Helmholtz Centre Potsdam - GFZ German Research Centre for Geosciences, Potsdam, Brandenburg, Germany
orcid

Research Focus & Metrics

Academic Career Length

15Years (2012-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
21 December 2021 in Remote Sensing
Verified
Journal Article
Open access
PDF file
CNChinh Thai Nguyen
Chinh Thai Nguyen
TOTemitope Seun Oluwadare
Temitope Seun Oluwadare
Nhung Thi LeNhung Thi Le
Nhung Thi Le
MEM.Mahdi Alizadeh Elizei
M.Mahdi Alizadeh Elizei
JWJens Wickert
Jens Wickert
HSHarald Schuh
Harald Schuh
Major advancements in the monitoring of both the occurrence and impacts of space weather can be made by evaluating the occurrence and distribution of ionospheric disturbances. Previous studies have shown that the fluctuations in total electron content (TEC) values estimated from Global Navigation Satellite System (GNSS) observations clearly exhibit the intensity levels of ionospheric irregularities, which vary continuously in both time and space. The duration and intensity of perturbations depend on the geographic location. They are also dependent on the physical activities of the Sun, the Earth’s magnetic activities, as well as the process of transferring energy from the Sun to the Earth. The aim of this study is to establish ionospheric irregularity maps using ROTI (rate of TEC index) values derived from conventional dual-frequency GNSS measurements (30-s interval). The research areas are located in Southeast Asia (15°S–25°N latitude and 95°E–115°E longitude), which is heavily affected by ionospheric scintillations, as well as in other regions around the globe. The regional ROTI map of Southeast Asia clearly indicates that ionospheric disturbances in this region are dominantly concentrated around the two equatorial ionization anomaly (EIA) crests, occurring mainly during the evening hours. Meanwhile, the global ROTI maps reveal the spatial and temporal distributions of ionospheric scintillations. Within the equatorial region, South America is the most vulnerable area (22.6% of total irregularities), followed by West Africa (8.2%), Southeast Asia (4.7%), East Africa (4.1%), the Pacific (3.8%), and South Asia (2.3%). The generated maps show that the scintillation occurrence is low in the mid-latitude areas during the last solar cycle. In the polar regions, ionospheric irregularities occur at any time of the day. To compare ionospheric disturbances between regions, the Earth is divided into ten sectors and their irregularity coefficients are calculated accordingly. The quantification of the degrees of disturbance reveals that about 58 times more ionospheric irregularities are observed in South America than in the southern mid-latitudes (least affected region). The irregularity coefficients in order from largest to smallest are as follows: South America, 3.49; the Arctic, 1.94; West Africa, 1.77; Southeast Asia, 1.27; South Asia, 1.24; the Antarctic, 1.10; East Africa, 0.89; the Pacific, 0.32; northern mid-latitudes, 0.15; southern mid-latitudes, 0.06.

13 March 2024
Verified
Book Chapter
Closed access
PDF file
Nhung Thi LeNhung Thi Le
Nhung Thi Le
BMBenjamin Männel
Benjamin Männel
MJMihaela Jarema
Mihaela Jarema
TLThach Thanh Luong
Thach Thanh Luong
LBLuyen Khac Bui
Luyen Khac Bui
HSHarald Schuh
Harald Schuh

3 March 2021
Verified
Other
Open access
Nhung Thi LeNhung Thi Le
Nhung Thi Le
BMBenjamin Männel
Benjamin Männel
MJMihaela Jarema
Mihaela Jarema
GSGopi Krishna Seemala
Gopi Krishna Seemala
Kosuke HekiKosuke Heki
Kosuke Heki
HSHarald Schuh
Harald Schuh
In data mining, outliers can lead to misleading interpretations of statistical results, particularly in deformation monitoring based on fluctuations and disturbances simulated by numerical models for the analysis of deformations. Therefore, outlier filtering cannot be ignored in data standardization. However, it is not likely that a filtering algorithm is efficient for every data pattern. We investigate five outlier filtering algorithms using MATLAB® (Release 2020a): moving average, moving median, quartiles, Grubbs, and generalized extreme Studentized deviation (GESD) to select the optimal algorithms applied for GNSS time series data. This study is conducted on two types of data used for ionosphere disturbance analysis in the region of the Ring of Fire and crustal deformation monitoring in Germany, one showing seasonal time series patterns and the other presenting the trend models. We apply the simple random sampling method that ensures the principles of unbiased surveying techniques. The optimal algorithm selection is based on the sensitivity of outlier detection and the capability of the central tendency measures. The algorithm robustness is also tested by altering random outliers but maintaining the standard distribution of each dataset. Our results show that the moving median algorithm is most sensitive for outlier detection because it is robust statistics and is not affected by anomalies; followed in turn by quartiles, GESD, and Grubbs. The outlier filtering capability of the moving average algorithm is least efficient, with a percentage of outlier detection below 20% compared to the moving median (corresponding 95% probability). In deformation analysis, disturbances on numerical models are often the basis for motion assessment, while these anomalies are smoothed by moving median filtering. Hence, the quartiles algorithm can be considered in this case. Overall, the moving median is best suited to filter outliers for seasonal and trend time series data; in particular, for deformation analysis, the optimal solution is applying the quartiles or extending the threshold factor and the sliding window of the moving median.Keywords: Outlier filtering, Time series, Deformation analysis, Moving median, Quartiles, MATLAB.

1 August 2023 in IOP Conference Series: Materials Science and Engineering
Verified
Conference Paper
Open access
NVNgoc Quang Vu
Ngoc Quang Vu
Nhung Thi LeNhung Thi Le
Nhung Thi Le
NLNgoc Dung Luong
Ngoc Dung Luong
Analysis and identifying the displacement characteristics play a key role in timely monitoring and detecting the physical responses of the bridge to ensure the safety of the human and structure. Many previous kinds of research used GNSS data to identify displacement and oscillation modelling of the bridge with different algorithms. This study uses GNSS time-series data to determine linear displacement and model oscillation of the bridge using a procedure including filtering outliers, linear regression, and sin function to identify amplitude in three directions, the plane displacement velocity, spatial displacement velocity, and vibration model of the bridge. The data in the research in the GNSS time-series data from three P5 GNSS receivers of the CHC brand on the Dachongyong bridge in Nanning, China with 1646 observations, at one-hour sample intervals in 68 consecutive days. The plane and spatial velocity of the three points DCQ01, DCQ02, and DCQ03 is 0.0181 mm/h, 0.0185 mm/h; 0.0114 mm/h, 0.0173 mm/h; and 0.0071 mm/h, 0.0082 mm/h respectively. The study results are significant in analyzing and identifying the bridge’s displacement characteristics.

24 August 2022
Verified
Conference Paper
Open access
Nhung Thi LeNhung Thi Le
Nhung Thi Le
BMBenjamin Männel
Benjamin Männel
Zhiguo DengZhiguo Deng
Zhiguo Deng
TLThanh Thach Luong
Thanh Thach Luong
HSHarald Schuh
Harald Schuh
 The increasing development of GNSS techniques enables solving geodetic problems on both local and global scales. Parallelly, complex algorithms have been proposed and can also be solved well by Machine Learning (ML). However, ML techniques are sometimes not sensitive enough to gain results with a high probability for some cases, like sparse data or non-stationary GNSS time series. In this study, we use a combination of Human and Machine learning (H&M) to improve the classification performance of continuous GNSS stations. First, 427 permanent GNSS stations are obtained from the EUREF network to train ML models. The models are then applied to classify the quality of 939 continuous observation stations from two projects, EIFEL and IPOC, carried out by the German Research Centre for Geosciences (GFZ), Potsdam, Germany. Next, we independently validate the ML models' reality through a MATLAB program, GNSS metadata, and seismic data. Finally, all data of these 1366 stations are used to re-train the ML models. The main criteria to classify are the number of outliers, jumps in GNSS time series, root mean square errors, observation time-spans, and stability of the crustal motion velocity fields. Applying the approach of the H&M combination improves the performance of the ML models up to 92% while using only ML methods remains ~68%. These ML-based classification models can be applied to estimate the quality of permanent GNSS stations and to manage big databases. The result is the basis for selecting suitable control and monitoring stations in crustal deformation monitoring as well as in civil and industrial applications. Keywords: GNSS station classification, Machine learning, Human & Machine learning combination.

19 December 2024 in GPS Solutions
Verified
Journal Article
Open access
PDF file
Nhung Thi LeNhung Thi Le
Nhung Thi Le
BMBenjamin Männel
Benjamin Männel
LBLuyen Khac Bui
Luyen Khac Bui
HSHarald Schuh
Harald Schuh
The development of Global Navigation Satellite Systems (GNSS) results in large spatial geodetic networks with a distinct range of accuracy. Thus, classification of the GNSS stations is needed to determine which stations are appropriate for geodetic applications. Additionally, advanced Machine Learning (ML) techniques have been proposed. However, ML algorithms may sometimes be less sensitive due to a lack of samples or anomalies in input data. Therefore, this study introduces an approach in which human-based supervision is integrated into ML processes to improve the ML model’s performance in classifying the continuous GNSS stations. The human factor influences the ML processes through two sampling strategies: “suggest-decide” and “correct-retrain”, where the accuracy of ML models will be improved via human-based corrections. The idea is that the unsupervised ML-based clustering techniques are driven by human-based supervision to create samples for training the supervised ML-based classification models. In this study, we develop a MATLAB app to automate the clustering and labeling processes. Our finding demonstrates that applying these sampling strategies can enhance the accuracy of the ML-based classification models from under 50 % up to $$\sim$$ ∼ 99 % after re-training. Also, this study categorizes almost 9000 continuous monitoring stations in the Nevada database, of which 1900 stations in Europe serve as samples for training the ML-based classification models. Furthermore, the methodologies developed in this study can be applied to warning systems, which utilize internal and external human resources to correct errors, address unusual situations, and provide timely feedback for better performance of ML-based forecasts.