Premium
A Real–time Adaptive Sampling Method for Field Mapping in Patchy, Heterogeneous Environments
Author(s) -
Graniero P A,
Robinson V B
Publication year - 2003
Publication title -
transactions in gis
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.721
H-Index - 63
eISSN - 1467-9671
pISSN - 1361-1682
DOI - 10.1111/1467-9671.00128
Subject(s) - transect , sampling (signal processing) , traverse , adaptive sampling , sample (material) , computer science , parameterized complexity , sample size determination , field (mathematics) , point (geometry) , algorithm , statistics , data mining , mathematics , geography , cartography , monte carlo method , geology , computer vision , physics , geometry , oceanography , filter (signal processing) , pure mathematics , thermodynamics
Many environmental studies require detailed maps describing the spatial distribution of various environmental characteristics. These distributions tend to be ‘patchy’; that is, their structure and their relationships vary from place to place according to the influences of the local setting. We present a simple sampling method that adapts the sample spacing on a point–by–point basis as the data are collected. The resulting sample is denser in areas of higher variability and sparser in more ‘well–behaved’ areas, and is collected in a single traverse of the transect. It uses a combination of simple fuzzy functions representing the concepts ‘too close’ and ‘too far’ that are adaptively parameterized based on current measurements. The adaptive sampler produced better representations for 47% of simulated reference transects than uniform or random samples of the same size under perfect measurement conditions, increasing to best performance for 71% of the transects when measurement error was severe with only minimal increase in sampling density. The rapid calculations can be easily incorporated into real–time data acquisition software, and the method may be extended to achieve some type of compromise when faced with the need to sample multiple simultaneous variables.