Earth & Space Science

Permanent URI for this collection

https://hdl.handle.net/10315/27626

Browse

Now showing 1 - 20 of 173

Open Access
3D Classification of Power Line Scene Using Airborne Lidar Data
(2015-08-28) Kim, Heungsik; Sohn, Gunho
Failure to adequately maintain vegetation within a power line corridor has been identified as a main cause of the August 14, 2003 electric power blackout. Such that, timely and accurate corridor mapping and monitoring are indispensible to mitigate such disaster. Moreover, airborne LiDAR (Light Detection And Ranging) has been recently introduced and widely utilized in industries and academies thanks to its potential to automate the data processing for scene analysis including power line corridor mapping. However, today’s corridor mapping practice using LiDAR in industries still remains an expensive manual process that is not suitable for the large-scale, rapid commercial compilation of corridor maps. Additionally, in academies only few studies have developed algorithms capable of recognizing corridor objects in the power line scene, which are mostly based on 2-dimensional classification. Thus, the objective of this dissertation is to develop a 3-dimensional classification system which is able to automatically identify key objects in the power line corridor from large-scale LiDAR data. This dissertation introduces new features for power structures, especially for the electric pylon, and existing features which are derived through diverse piecewise (i.e., point, line and plane) feature extraction, and then constructs a classification model pool by building individual models according to the piecewise feature sets and diverse voltage training samples using Random Forests. Finally, this dissertation proposes a Multiple Classifier System (MCS) which provides an optimal committee of models from the model pool for classification of new incoming power line scene. The proposed MCS has been tested on a power line corridor where medium voltage transmission lines (115 kV and 230 kV) pass. The classification results based on the MCS applied by optimally selecting the pre-built classification models according to the voltage type of the test corridor demonstrate a good accuracy (89.07%) and computationally effective time cost (approximately 4 hours/km) without additional training fees.
Open Access
3D Reconstruction of Building Rooftop and Power Line Models in Right-of-Ways Using Airborne LiDAR Data
(2016-11-25) Jwa, Yoonseok; Sohn, Gunho
The research objectives aimed to achieve thorough the thesis are to develop methods for reconstructing models of building and PL objects of interest in the power line (PL) corridor area from airborne LiDAR data. For this, it is mainly concerned with the model selection problem for which model is more optimal in representing the given data set. This means that the parametric relations and geometry of object shapes are unknowns and optimally determined by the verification of hypothetical models. Therefore, the proposed method achieves high adaptability to the complex geometric forms of building and PL objects. For the building modeling, the method of implicit geometric regularization is proposed to rectify noisy building outline vectors which are due to noisy data. A cost function for the regularization process is designed based on Minimum Description Length (MDL) theory, which favours smaller deviation between a model and observation as well as orthogonal and parallel properties between polylines. Next, a new approach, called Piecewise Model Growing (PMG), is proposed for 3D PL model reconstruction using a catenary curve model. It piece-wisely grows to capture all PL points of interest and thus produces a full PL 3D model. However, the proposed method is limited to the PL scene complexity, which causes PL modeling errors such as partial, under- and over-modeling errors. To correct the incompletion of PL models, the inner and across span analysis are carried out, which leads to replace erroneous PL segments by precise PL models. The inner span analysis is performed based on the MDL theory to correct under- and over-modeling errors. The across span analysis is subsequently carried out to correct partial-modeling errors by finding start and end positions of PLs which denotes Point Of Attachment (POA). As a result, this thesis addresses not only geometrically describing building and PL objects but also dealing with noisy data which causes the incompletion of models. In the practical aspects, the results of building and PL modeling should be essential to effectively analyze a PL scene and quickly alleviate the potentially hazardous scenarios jeopardizing the PL system.
Open Access
A BIM - GIS Integrated Information Model Using Semantic Web and RDF Graph Databases
(2023-03-28) Hor, Abdelhadi; Sohn, Gunho
In recent years, 3D virtual indoor and outdoor urban modelling has become an essential geospatial information framework for civil and engineering applications such as emergency response, evacuation planning, and facility management. Building multi-sourced and multi-scale 3D urban models are in high demand among architects, engineers, and construction professionals to achieve these tasks and provide relevant information to decision support systems. Spatial modelling technologies such as Building Information Modelling (BIM) and Geographical Information Systems (GIS) are frequently used to meet such high demands. However, sharing data and information between these two domains is still challenging. At the same time, the semantic or syntactic strategies for inter-communication between BIM and GIS do not fully provide rich semantic and geometric information exchange of BIM into GIS or vice-versa. This research study proposes a novel approach for integrating BIM and GIS using semantic web technologies and Resources Description Framework (RDF) graph databases. The suggested solution's originality and novelty come from combining the advantages of integrating BIM and GIS models into a semantically unified data model using a semantic framework and ontology engineering approaches. The new model will be named Integrated Geospatial Information Model (IGIM). It is constructed through three stages. The first stage requires BIMRDF and GISRDF graphs generation from BIM and GIS datasets. Then graph integration from BIM and GIS semantic models creates IGIMRDF. Lastly, the information from IGIMRDF unified graph is filtered using a graph query language and graph data analytics tools. The linkage between BIMRDF and GISRDF is completed through SPARQL endpoints defined by queries using elements and entity classes with similar or complementary information from properties, relationships, and geometries from an ontology-matching process during model construction. The resulting model (or sub-model) can be managed in a graph database system and used in the backend as a data-tier serving web services feeding a front-tier domain-oriented application. A case study was designed, developed, and tested using the semantic integrated information model for validating the newly proposed solution, architecture, and performance.
Open Access
A Comparison Study on Control Moment Gyroscope Arrays and Steering Laws
(2020-05-11) Moorthy, Chitiiran Krishna; Lee, Regina S. K.
Current reaction wheels and magnetorquers for microsatellite are limited by low slew rate and heavily depends on orbital parameters for coverage area. Control moment gyroscope (CMG) clusters offer an alternative solution for high slew rates and rapid retargeting. Though CMGs are often used in large space missions, their use in microsatellites is limited due to the stringent mass budget. Most literature reports only on pyramid configuration, and there are no definite cross-comparison studies between various CMG clusters and steering laws. In this research, a generic tool in Matlab and Simulink is developed to further understand CMG configurations and steering laws for a microsat mission. Various steering laws necessary for mitigating singularities in CMG clusters are compared in two distinct missions. The simulation results were evaluated based on the pointing accuracy, platform jitter, and pointing stability achieved by the spacecraft for each combination of CMG clusters, and steering laws and trajectories. The simulation results demonstrate that the pyramid cluster is marginally better than the rooftop cluster in pointing accuracy. The comparison of steering laws shows that, counterintuitively, Singularity Robust steering law, which passes through singularities, outperforms both Moore-Penrose and Local Gradient methods for almost all evaluation criteria for the two missions it was tested on. The simulation results would aid systems engineers in designing low-cost actuation systems and corresponding control software, which can increase the data acquisition rate of remote sensing missions.
Open Access
A Deep Learning Approach to the Detection and Tracking of Moving Objects in 2D Point Clouds
(2022-12-14) Schofield, Hunter Liam; Shan, Jinjun
The detection and tracking of moving objects (DATMO) are crucial tasks that any autonomous vehicle must perform. Autonomous vehicles must detect and track all obstacles to ensure safety within the environment while also completing their tasks efficiently. In autonomous driving research, LiDAR is becoming increasingly popular due to its high resolution and accuracy. There are many state-of-the-art DATMO methods using LiDAR, however, most methods are designed for 3D LiDAR sensors. Methods that work for 2D LiDAR sensors are not as robust as their 3D counterparts or require too many computational resources to run efficiently on less powerful robots. This research presents two robust solutions to the DATMO problem based on deep learning techniques that can scale to meet a variety of hardware constraints. The first solution, detect while track (DWT), combines a convolutional neural network (CNN) with a multiple hypothesis tracking (MHT) approach and Kalman filter. The second solution, pixel predictions for future-oriented bounding boxes (PIXFOR), combines a CNN with a recurrent network architecture to solve both detection and tracking problems in a single forward pass. Both methods are experimentally validated on an unmanned ground vehicle (UGV) operating on an intersection scenario and a highway scenario using 2D point clouds collected from simulation and hardware environments. The run-time performance of both methods is also validated different hardware platforms to show that the methods can scale to meet different hardware constraints. When compared to state-of-the-art DATMO methods, the newly proposed methods outperform in the object detection and tracking tasks, while operating at a faster run time on equivalent hardware.
Open Access
A Large Scale Simulation of Satellites Tracking Vessels and Other Targets
(2015-01-26) Wisnarama, Sriyan Indrajith; Lee, Regina
This research outlines the design of a large scale simulation of satellites tracking large amounts of dynamic targets. The use of such a simulation is presented and current solutions available are presented. The research sets out a list of objectives to meet by creating an application programming interface (API) that have the requirements of being efficient, scalable, flexible, and easy to use for the implementer. Methods of creating sections of the simulation such as the attitude motion of a satellite based on the physical characteristics of nanosatellites is explored and developed. The creation of targets that are contained only on certain land features are also developed and tested. The objectives set out are tested by creating a simulation using the API developed and the results are presented.
Open Access
A Novel Controlled Environment Study of Prebiotic Organic Material in the Tagish Lake Meteorite Using Raman and Fluorescence Spectroscopy: Implications for Asteroid-Return Sample Analysis
(2023-12-08) Lymer, Elizabeth Anne; Daly, Michael
Several ambitious missions such as OSIRIS-REx, Hayabusa2 and Mars Sample Return are currently collecting or planning to collect the most pristine planetary samples to analyze on Earth since the Apollo missions. Therefore, it is essential to create instrumentation that will allow preliminary compositional analysis of these materials without the need for sample preparation and in a minimally destructive way, while preventing oxidation, alteration, or contamination from Earth’s environment. At York University, a small, transportable environmental chamber with temperature, pressure, and atmosphere control has been integrated on an optical table with a combined UV (266 nm) and Green (532 nm) Raman, laser-induced fluorescence, and time-resolved laser-induced fluorescence instrument. This system can collect high-resolution 2D spectroscopic maps, long accumulation point analysis and time-resolved fluorescence ‘fingerprinting’ of minerals and organics. The sample chamber is capable of pressures < 10E-4 mbar, temperatures < -20C, and different atmosphere compositions such nitrogen, argon or carbon dioxide via gas hookup. The intended use of the system is to detect and identify minerals and organics within sensitive or fragile planetary samples using minimally destructive laser-based techniques, while maintaining specific environmental conditions to preserve and maintain the pristineness of the material being analyzed, such as samples returned from planetary missions. To validate the functionality of the system as a laboratory tool for samples returned from space or for planetary science research, several experiments were conducted using different materials to showcase the different modalities of the instrumentation. One such experiment is conducted on a non-pristine piece of the Tagish Lake meteorite, containing some of the most primitive materials in our solar system. The meteorite fragment was kept frozen and in an inert atmosphere during Raman (UV and green) and Fluorescence (UV) analyses. These spectroscopic techniques resulted in high-resolution 2D maps of the surface of the meteorite fragment, showing mineral localities and groupings, as well as organic constituents present. UV-Raman in particular results in spectra containing aromatic and aliphatic hydrocarbon, ketones, and cyano radical peaks, all of which are the basic building blocks of important organic constituents like amino acids which play an integral role in biotic life on Earth. The detection of such compounds in the Tagish Lake meteorite using UV-Raman spectroscopy is novel and provides a unique tool to analyze organic constituents within sensitive materials in a very minimally destructive way with no sample preparation.
Open Access
A Study of Buried Organics on Mars - A Computational And Experimental Approach
(2023-12-08) Das, Ankita; Sapers, Haley; Moore, John
This thesis consists of two related projects investigating exogenous organic burial on the surface of Mars and potential detection methods. Mars receives a significant quantity of organic material through exogenous sources such as micrometeorites. This work investigates the preservation of such material within sedimentary structures formed through the gradual airfall of dust particles. The computer model devised indicates rapid burial of organics and presents amount of organic carbon preserved at selected locations on Mars after 10 Martian years (99% preserved). The model is further extended globally to point at sites on Mars which are most favorable for organic preservation and hence possible sites for robotic exploration. The experimental section of this thesis investigates organic detection (Tryptophan) in simulated Martian dust matrix (JSC-1 Mars) through Ultraviolet Light Induced Fluorescence (LIF) spectroscopy. Instrument used was unable to detect Tryptophan in 100 ppm concentrations owing to the UV absorptive nature of JSC-1 Mars.
Open Access
Abstracting Cubesat Operations: A Path to Real Cubesat Interoperability
(2019-11-22) Jain, Vidushi; Newland, Franz
Introduction of the CubeSat form factor brought a paradigm shift in the industry. With the size becoming a standard, cost and development time were able to be reduced significantly. However, the industry has not yet fully realized the potential of this new paradigm. Many other simplifications or standardizations can be made, whilst still meeting CubeSat mission requirements. One area that can be addressed without significant change is CubeSat mission operations. Many operations activities for CubeSat busses are common, or the differences between missions are close enough to benefit from common streamlining. This thesis proposes abstracted operations sequence for CubeSats. The sequence is demonstrated by applying it to an upcoming CubeSat mission - DESCENT. Simplifications made as a result of this abstraction are demonstrated. The thesis also points to some of the other improvements that could be made longer-term for CubeSat mission designers and operators through further industry standardization.
Open Access
Accuracy Improvement of Terrestrial Mobile Lidar System in Engineering Surveys
(2015-12-16) Liu, Guannan; Wang, Jian-Guo
In this thesis, a number of effective algorithms and strategies were developed to improve the accuracy of terrestrial mobile LiDAR solutions in the field of engineering surveys. A detailed analysis for error budget of the terrestrial mobile LiDAR system has been presented in order to well interpret the effects of individual error sources. Firstly, the 3D conformal coordinate transformation (3DCCT) through Least Squares Method (LSM) was applied by employing the ground control points incorporating with feature constraints. Secondly, the multistrip adjustment (MA) algorithm was developed by taking advantage of the overlapped data strips and the repeated data acquisition over the same survey area using both of tie points and tie features. Lastly, the boresight angles of a terrestrial LiDAR system was preliminarily calibrated by using the planar and/or line features of two scans acquired during consecutive runs in opposite driving directions at the post-processing stage proposed by Keller et al. (2013).
Open Access
Active Reinforcement Learning for the Semantic Segmentation of Images Captured by Mobile Sensors
(2023-03-28) Jodeiri Rad, Mahya; Armenakis, Costas
Neural Networks have been employed to attain acceptable performance on semantic segmentation. To perform well, many supervised learning algorithms require a large amount of annotated data. Furthermore, real-world datasets are frequently severely unbalanced, resulting in poor detection of underrepresented classes. The annotation task requires time-consuming human labor. This thesis investigates the use of a reinforced active learning as region selection method to reduce human labor while achieving competitive results. A Deep Query Network (DQN) is utilized to identify the best strategy to label the most informative regions of the image. A Mean Intersection over Union (MIoU) training performance equivalent to 98% of the fully supervised segmentation network was achieved with labeling only 8% of dataset. Another 8% of labelled dataset was used for training the DQN. The performance of all three segmentation networks trained with regions selected by Frequency Weighted Average (FWA) IoU is better in comparison with baseline methods.
Open Access
Advanced Data Fusion Methods to Improve Wetland Classification Using Multi-Source Remotely Sensed Data
(2023-12-08) Judah, Aaron Jonathan; Hu, Baoxin
The goal of this research was to improve wetland classification accuracy and the reduction of classification errors and uncertainty by fully exploiting multi-source remotely sensed, and ancillary data and image metrics using advanced data analysis techniques. This PhD research executed in three phases: 1. Explorations of data type selections and significance in support of wetland classification. 2. The development of a hierarchically-based classification approach to best exploit the data identified and characterized through the first study. 3. The development of an ensemble classifier incorporating the aforementioned developments with Dempster-Shafer (D-S) theory in order to reduce errors and streamline computations. The first phase explored the most effective data features, and metrics or families of data features in support of wetland classification. It was found that wetlands were best classified using the NDVI calculated from optical imagery obtained in the spring months, radar backscatter coefficients, surface temperature, and ancillary data such as surface slope, computed through either a Random Forest (RF) or Support Vector Machine (SVM) classifier. This work was also able to produce a wetland land cover map with an accuracy of 87.51% - an improvement from the ~82% typical of similar datasets and landcover types. In the second phase a more effective approach to classify the aforementioned features in order to fully utilize the discriminant power of those features was explored. This was done through two hierarchically based classification strategies. The second hierarchically based RF classification methodology produced the most accurate classification result (91.94%). The third phase focused on how to better exploit broad class separations and to reduce the propagation of errors and uncertainty which cascades through the classification hierarchy. These classifiers were integrated using D–S theory. Classification resulted in an overall accuracy of ~93% an improvement of 5% when compared to a traditional classification method. High level of confidence (>85%) misclassified pixels were reduced by ~10%. The major contribution of this research was the improvement of classification accuracy and the reduction of classification errors and uncertainty through use of multiple classifiers, designed to best exploit broad class separations, through selected data features computed within a D-S framework.
Open Access
Agent-Based Modelling and Simulation of Sidewalk Delivery Robots' Interaction with Pedestrians
(2023-12-08) Hassan, Ali; Sohn, Gunho
In the evolving urban landscape, the surge of Sidewalk Autonomous Delivery Robots (SADRs) calls for insights into their pedestrian interactions. This thesis explores these dynamics and presents a novel web-based simulator, "TwinWalk", to emulate such interactions. Rooted in a GIS-enhanced digital twin of a campus-like urban setting, TwinWalk employs agent-based modeling, steering behaviours, and the Predictive Avoidance Model (PAM) to illustrate collision scenarios and human-robot interactions. Experiments reveal that while SADRs maintain safety buffers, they pose collision risks. Notably, pedestrians often breach safety distances, implicating them in proximity issues. This emphasizes the need for further research and specialized safety measures. The simulator aids urban planners and researchers in assessing design interventions and policies. The study suggests that enlarged safety zones around robots can reduce collisions and enhance pedestrian flow, promoting harmonious robot-human urban coexistence.
Open Access
An Examination of the ThermopiezoelectricEffect in Piezoelectric Actuators Via Finite Elements
(2023-12-08) Toledo, Rafael Chaves; Orszulik, Ryan
The application of piezoelectric actuators in smart structures is a rapidly developing field, especially in aerospace environments. Since thermal effects play an important role in aerospace environments, thermopiezoelectricity has been studied as it takes into account the thermal field in addition to the mechanical and electrical fields. As a result, the coupling effects among these three fields have to be considered, including the pyroelectric (change in the electric potential due to the presence of a thermal field) and electrocaloric (change in the temperature when an electric field is applied) effects. This thesis presents an examination of how these coupled effects can affect the performance of piezoelectric bender and stack actuators in varying external environments. More specifically, this thesis investigates the influence of the pyroelectric and electrocaloric effects on the positioning and dynamic performance of these actuators in static and dynamic cases by using a custom written finite element code that considers the three fully coupled field equations of thermopiezoelectricity.
Open Access
An Extensible Self-Arbitrating Architecture for Redundant Onboard Computers In Nanosatellites
(2021-11-15) Bindra, Udai; Zhu, George Z.H.
This thesis explores the use of a novel N-modular cold redundancy architecture using Commercial-Off-the-Shelf onboard computers in Nanosatellites. The proposed architecture explores a software-based decentralization of the arbitration logic. This eliminates the need for an external supervising device to perform the arbitration amongst the onboard computers (OBC). Instead, the architecture is reliant only on a memory-less watchdog timer to reset the spacecraft power bus in case an OBC becomes unresponsive. Moreover, the architecture assesses the health of each redundant OBC and provides precedence to the critical capabilities of the OBC. Finally, the proposed redundancy architecture enables extensibility and is near platform agnostic, allowing reusability across missions that utilize different OBC platforms, without the need for major modifications in the arbitration mechanism. The arbitration mechanism of proposed architecture was successfully validated in the lab by emulating faults using Raspberry Pi Zero W modules in triple modular redundancy. An additional OBC was added to demonstrate extensibility of the proposed redundancy architecture.
Open Access
An Imaging Fourier Transform Spectrometer (IFTS) for Climate Observations
(2019-03-05) Singh, Gurpreet; McElroy, C. Thomas
Climate change is an ongoing global phenomenon having a greater impact at higher latitudes. The instrument development reported herein is aimed at demonstrating the feasibility of using an Imaging Fourier Transform Spectrometer (IFTS) to measure carbon dioxide (CO2), and methane (CH4) mixing ratios at high latitudes using oxygen A-band measurements as a surface pressure reference. This thesis details the optical design, instrumental setup, and development criteria for the IFTS. The development of a software package to control and acquire data is also discussed. The instrument is developed to achieve the Technology Readiness Level 4 standard which covers the breadboard validation of a space system in a laboratory environment. Hardware specifications and software algorithms of the instrument are presented. Results from an external Helium-Neon (HeNe) laser and a broadband light source limited by spectral bandpass filters are presented. Finally, recommendations and future improvements to this research and development program are listed.
Open Access
An Investigation of Turbulence and Diffusion within Vehicle Wakes and On-Road Measurements using an Instrumented Mobile Car and a Stationary Roadside Monitoring System
(2022-12-14) Miller, Stefan John; Gordon, Mark
Moving motor vehicles emit pollutants that negatively impact human health. Stationary roadside measurements alone are not sufficient to quantify the pollutant–flow interactions that occur behind moving vehicles. The instrumented mobile car however is well–suited for on–road measurements, but has been underutilized for this purpose since limited studies have investigated its accuracy at high vehicle speeds. Thus, this work details two on–road measurement campaigns using an instrumented car, with three main objectives: (1) study the vehicle momentum wake and vehicle–induced turbulence (VIT), (2) investigate the accuracy of the mobile system for measuring atmospheric means, variances and covariances, and (3) quantify the emission of aerosols and CO2 by on–road vehicles and their subsequent diffusion. Measurements behind on–road vehicles demonstrate that VIT decays with increasing distance following a power law relationship. Comparison of measurements with prior on–road studies suggests a height dependence of VIT in vehicle wakes, and an extended parameterization is outlined that describes the total on–road turbulent kinetic energy (TKE) enhancement due to a composition of vehicles, including a vertical dependence on the magnitude of TKE. Next, a wavelet–based approach to remove the effects of sporadic passing traffic is developed and applied to a measurement period during which a heavy–duty truck passes in the opposite highway lane; removing the times with traffic in this measurement period gives a 10% reduction in the TKE. When sampling uncertainties are considered, the vertical momentum flux measured on the car is found to be not different from roadside measurements in the 95% confidence interval. The first on–road and in–traffic measurements of the vertical turbulent particle number flux and the vertical turbulent CO2 flux are presented and the results suggest this technique could be further developed to measure individual vehicle emission rates while driving. The lateral width of the wake generated by each passing vehicle is estimated using the stationary roadside measurements, and is determined to be a factor of 5 times greater for heavy–duty trucks relative to sport utility vehicles and passenger cars at a distance of 150 m behind the vehicle.
Open Access
Analysis and Conditioning of GNSS Measurements from Smartphones for Precise Point Positioning in Realistic Environments
(2020-11-13) Shinghal, Ganga; Bisnath, Sunil B.
With the availability of raw GNSS code and carrier-phase measurements from dual-frequency chips in smartphones, Precise Point Positioning (PPP) can be used to improve the accuracy of the positioning solution offered by them, without any additional reference infrastructure. In realistic applications in suburban and urban environments, with the smartphone in hand or on the dashboard of a car, there are numerous challenges with smartphone data such as missing measurements, poor multipath suppression and low and irregular carrier-to-noise-density ratio. The measurements are analysed and conditioned or pre-processed by implementing a prediction technique for filling in data gaps and a C/N0-based stochastic model for assigning realistic a priori weights to the measurements in the PPP processing engine. Finally, the post-processed PPP solution accuracy and availability have been compared with the positioning solution obtained from other GNSS positioning techniques. After conditioning, the smartphone PPP positioning solution is seen to have nearly 100% solution availability and 50% more accuracy.
Open Access
Aphelion Cloud Formation and Swiss Cheese Sublimation: Martian Atmospheric Water Vapour Processes
(2022-03-03) Innanen, Alex Cummings; Moores, John
This thesis studies the behaviour of atmospheric water vapour at two points in the Martian water cycle: the formation of the Aphelion Cloud Belt (ACB) in Northern spring and summer and the sublimation of the south polar cap in Northern autumn and winter. Firstly, comparing the phase functions of ACB clouds before and after the global dust storm (GDS) of Mars year 34 finds little difference between the two, indicating the dominant ice crystal habit was not impacted by the GDS. The exact nature of this geometry is still not known. Secondly, we mapped the current extent of Swiss cheese features and determined a range of temperatures of exposed water ice within them. From this we estimated a current sublimation rate and found the water vapour contribution from them is negligible. We also determined the area of exposed water ice needed to sublimate 30 precipitable microns of water vapour.
Open Access
Application of physics-informed neural network approach in soil moisture retrieval using GNSS reflectometry
(2024-03-16) Gavili Kilane, Narin; Bisnath, Sunil B.
This research aims to obtain soil moisture from reflected GNSS signals using physics-informed neural networks (PINN). GNSS reflectometry (GNSS-R) signals can be considered as a new remote sensing source to study soil moisture. Despite the high sensitivity between GNSS reflected signal power and soil moisture, the model between measurements and parameters is difficult to solve mathematically due to the complexity of the electromagnetic relationships. Although Neural Network (NN) algorithms have been applied successfully in GNSS-R soil moisture retrieval, neural networks are trained without respecting any laws of physics. In this work, a new framework referred to as “physics-informed neural networks (PINN)” was used which adds governing physical relationships between data parameters to neural networks to generate more robust models, with less data. The proposed research advances GNSS-R soil moisture estimations, exploiting Cyclone Global Navigation Satellite Systems (CYGNSS) satellite signals using PINN methodology. In PINN’s structure, reflected GPS signals from CYGNSS and land surface geophysical parameters are used as input features. Since reflected signal power variations are not only sensitive to changes in soil moisture, but also to changes in vegetation, surface roughness, soil texture, and elevation angle, the effects of land surface geophysical parameters involved in physical relationships are considered in the model. For reference data, soil moisture measurements of the International Soil Moisture Network (ISMN) were used in both training and validation. The proposed PINN model generates daily soil moisture values with a root mean squared error (RMSE) of 0.05 〖cm〗^3/ 〖cm〗^3, which is an improvement from 0.0774 〖cm〗^3/ 〖cm〗^3 for the underlying NN model due to adding physical models. Four different soil dielectric constant models (Dobson, Hallikainen, Mironov, and Wong models) have been used to investigate the impact of soil dielectric constant models as part of physical relations. The RMSE distinction and correlation coefficient difference of the best model (Hallikainen) and worst model (Mironov) is 0.02 and 0.13, respectively demonstrating PINN sensitivity to different soil dielectric constant models. Consequently, the soil dielectric constant model selection influences overall PINN results. Thus, calibration of soil dielectric models is necessary for GNSS-R soil moisture retrieval in the future.