Application of deep learning to large scale riverine flow velocity estimation
نویسندگان
چکیده
Fast and reliable prediction of riverine flow velocities plays an important role in many applications, including flood risk management. The shallow water equations (SWEs) are commonly used for the velocities. However, accurate fast with standard SWE solvers remains challenging cases. Traditional approaches computationally expensive require high-resolution measurement riverbed profile (i.e., bathymetry) predictions. As a result, they poor fit situations where need to be evaluated repetitively due, example, varying boundary condition (BC) scenarios, or when bathymetry is not known certainty. In this work, we propose two-stage process that tackles these issues. First, using principal component geostatistical approach estimate probability density function from velocity measurements, then use multiple machine learning algorithms order obtain solver SWEs, given augmented realizations posterior distribution prescribed range potential BCs. first step proposed allows us predict without any direct bathymetry. Furthermore, augmentation second stage incorporation additional information into improved accuracy generalization, even if changes over time. Here, three different forward solvers, referred as analysis-deep neural network, supervised encoder, variational validate them on reach Savannah river near Augusta, GA. Our results show capable predicting variable BCs good accuracy, at computational cost significantly lower than solving full value problem traditional methods.
منابع مشابه
Deep Optical Flow Estimation Via Multi-Scale Correspondence Structure Learning
As an important and challenging problem in computer vision, learning based optical flow estimation aims to discover the intrinsic correspondence structure between two adjacent video frames through statistical learning. Therefore, a key issue to solve in this area is how to effectively model the multi-scale correspondence structure properties in an adaptive end-to-end learning fashion. Motivated...
متن کاملThe Large - Scale Velocity
1 Bulk Flows as a Cosmological Probe Hubble's Law, now spectacularly connrmed by the work of 27], 35], and 39], tells us that the distances of galaxies are proportional to their observed recession velocities, at least at low redshifts: cz = H 0 r : (1) However, this is not exactly correct. Galaxies have peculiar velocities above and beyond the Hubble ow indicated by Eq. (1). We denote the pecul...
متن کاملapplication of brand personality scale in automobile industry: the study of samand’s brand personality dimensions
این تحقیق شخصیت برند سمند را در ایران با استفاده از مدل پنج بعدی آکر (1997) بعنوان یک چهارچوب بطور توصیفی سنجیده است. بنابر این چهارچوب که دراصل در 42 جزء (42 ویزگی شخصیتی) ودر پنج بعد شخصیتی طراحی شده بود ودر کشورها وصنایع مختلف آزموده شده بود, پرسنامه به زبان فارسی ترجمه شده و با استفاده از روشهای ترجمه معکوس و مصاحبه عمیق با 12 متخصص ایرانی به 38 جزء کاهش یافت. و نظرسنجی ای در پنج نمایندگی ا...
15 صفحه اولThe Large-Scale Velocity Field
where the peculiar velocity of the Local Group itself is v(0), and r̂ is the unit vector to the galaxy in question. In practice, we will measure distances in units of km s, which means that H0 ≡ 1, and the uncertainties in the value of H0 discussed by Freedman and Tammann in this volume are not an issue. Thus measurements of redshifts cz, and of redshift-independent distances via standard candle...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
ژورنال
عنوان ژورنال: Stochastic Environmental Research and Risk Assessment
سال: 2021
ISSN: ['1436-3259', '1436-3240']
DOI: https://doi.org/10.1007/s00477-021-01988-0