1.
Mîrț, Andrei; Reiche, Johannes; Verbesselt, Jan; Herold, Martin
A Downsampling Method Addressing the Modifiable Areal Unit Problem in Remote Sensing Journal Article
In: Remote Sensing, vol. 14, iss. 22, no. 5538, 2022.
Abstract | Links | BibTeX | Tags: big data, downsampling, modifiable areal unit problem, multscale, remote sensing
@article{nokey,
title = {A Downsampling Method Addressing the Modifiable Areal Unit Problem in Remote Sensing},
author = {Andrei Mîrț and Johannes Reiche and Jan Verbesselt and Martin Herold},
url = {https://www.mdpi.com/2072-4292/14/21/5538},
doi = {https://doi.org/10.3390/rs14215538},
year = {2022},
date = {2022-11-03},
journal = {Remote Sensing},
volume = {14},
number = {5538},
issue = {22},
abstract = {Handling multiple scales efficiently is one avenue for processing big remote sensing imagery data. Unfortunately, imagery is also affected by the infamous modifiable areal unit problem, which creates unpredictable errors at different scales. We developed a downsampling method that attempts to keep the data distribution in a downsampled image constant, reducing the modifiable areal unit problem. We tested our method against classic downsampling methods (mean, central pixel selection, random) under a range of typical remote sensing scenarios. Under our experimental conditions, our downsampling method consistently outperformed the classical downsampling methods within a 95% confidence level. The downsampling method can be used in most typical situations where downsampling is needed, but it is likely to shine when used as a pyramid building policy in geocomputing platforms, such as Google Earth Engine.},
keywords = {big data, downsampling, modifiable areal unit problem, multscale, remote sensing},
pubstate = {published},
tppubtype = {article}
}
Handling multiple scales efficiently is one avenue for processing big remote sensing imagery data. Unfortunately, imagery is also affected by the infamous modifiable areal unit problem, which creates unpredictable errors at different scales. We developed a downsampling method that attempts to keep the data distribution in a downsampled image constant, reducing the modifiable areal unit problem. We tested our method against classic downsampling methods (mean, central pixel selection, random) under a range of typical remote sensing scenarios. Under our experimental conditions, our downsampling method consistently outperformed the classical downsampling methods within a 95% confidence level. The downsampling method can be used in most typical situations where downsampling is needed, but it is likely to shine when used as a pyramid building policy in geocomputing platforms, such as Google Earth Engine.