Incoherent scatter modeling of Jicamarca radar spectra

Jicamarca;Radar spectra;k-Nearest Neighbors (KNN) Machine Learning;Ionosphere;Collision Frequency;Machine Learning;Electron ion collision

Aggarwal, Deepanshu ; Kudeki ; Erhan

关键词: Jicamarca;    Radar spectra;    k-Nearest Neighbors (KNN) Machine Learning;    Ionosphere;    Collision Frequency;    Machine Learning;    Electron ion collision;   

Others  :  https://www.ideals.illinois.edu/bitstream/handle/2142/46790/Deepanshu_Aggarwal.pdf?sequence=1&isAllowed=y

美国|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

This thesis describes a project on the modeling of spectral characteristics of electrondensity irregularities of the topside equatorial ionosphere probed by the Jicamarca IncoherentScatter Radar (ISR) located near Lima, Peru. The topside equatorial ionosphereis a multi-ion plasma and the spectrum of its electron density irregularities canbe modeled by extending the single-ion spectral model developed by Kudeki and Milla(2011) for a collisional equatorial F-region ionosphere. This single-ion model of Kudekiand Milla captures the essential physics of the equatorial F-region ionosphere where randomdisplacements of the dominant oxygen ions are characterized as a Brownian-motionprocess, while electron displacements are non-Brownian and described in terms of a numericallibrary (constructed using a Monte-Carlo approach) of single-electron ACFs (autocorrelationfunctions) parametrized by five state parameters of the F-region consisting ofionospheric electron density, geomagnetic flux density, electron and ion temperatures, andthe deviation angle of the radar boresight direction from the plane perpendicular to thegeomagnetic field, the so-called magnetic aspect angle. While the extension of the modelto the multi-ion case is straightforward, the discrete nature of the numerical electron ACFlibrary defined over a grid of input parameters precludes the evaluation of the extendedmodel with arbitrary and continuously varying input parameters. To overcome this difficultya machine learning (ML) based interpolation procedure is developed. The thesisdescribes the ML algorithm, the associated training and testing steps, and finally presentsa suite of examples of multi-ion IS spectra obtained with the extended model.