The Use of EPI-Splines to Model Empirical Semivariograms for Optimal Spatial Estimation

Abstract

This research investigates the ability of epi-splines to improve upon current methods of creatingempirical semivariograms for use in optimal spatial estimation (OSE). Models utilizing traditionalmethods of curve fitting for semivariograms (spherical, exponential, and Matrn) used in the spatialestimation process are compared to a proposed model that employs an epi-spline for curve fitting. Theresulting semivariograms are then used for kriging to produce spatial estimation using a sparse number ofmeasurements. The epi-spline model improves upon the mean absolute error, mean standard error, andrange of errors when compared to traditional methods. However, the comparisons indicate that goodnessof fit does not drastically improve the resultant spatial estimation. The benefit of epi-splines, in addition totheir ability to more accurately depict the relationship between data points, is their ability to incorporatesoft information in the form of constraints and the tighter variance of estimates resulting from their use.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Sep 01, 2016
Accession Number
AD1030104

Entities

People

  • Peter Ii M. Tydingco

Organizations

  • Naval Postgraduate School

Tags

Communities of Interest

  • Autonomy
  • Materials and Manufacturing Processes
  • Sensors
  • Weapons Technologies

DTIC Thesaurus Topics

  • Autonomous Underwater Vehicles
  • Curve Fitting
  • Gaussian Distributions
  • Gaussian Processes
  • Global Positioning Systems
  • Inertial Navigation
  • Inertial Navigation Systems
  • Information Science
  • Navigation
  • Numerical Analysis
  • Random Variables
  • Simultaneous Localization And Mapping
  • Standards
  • Surveys
  • Three Dimensional
  • Unmanned Systems
  • Unmanned Underwater Vehicles

Readers

  • Canine Service Warrior Training Program for Wounded Warriors in the Veterinary Industry, Supported by Donors.
  • Computational Modeling and Simulation
  • Computer Vision.