Open AccessInterpolation of Irregular Soil Moisture Measurements with Machine Learning Methods
Author(s) -
Milan Čistý,
Frantisek Cyprich,
David Dezericky
Publication year - 2020
Publication title -
iop conference series. materials science and engineering
Language(s) - English
Resource type - Journals
eISSN - 1757-899X
pISSN - 1757-8981
DOI - 10.1088/1757-899x/960/4/042050
Subject(s) - water content , computer science , interpolation (computer graphics) , nonlinear system , gradient boosting , variable (mathematics) , series (stratigraphy) , machine learning , moisture , work (physics) , soil science , artificial intelligence , environmental science , mathematics , random forest , geotechnical engineering , meteorology , engineering , geology , geography , motion (physics) , mechanical engineering , mathematical analysis , paleontology , physics , quantum mechanics
In this work, a method is proposed that uses machine learning techniques intending to make existing soil moisture time series complete. Authors are assuming and solving the usual situation, in which only data irregularly measured are available. Soil moisture is one of the determining variables of the stress on various ecosystems and agriculture systems and a key element of the surface water budget. A time series of this variable is useful for an evaluation of the moisture regime of soil and for decisions regarding building irrigation structures. Interpolation models proposed in this paper were verified using data from the days on which the field measurements were available. Mainly nonlinear machine learning models are proved to be suitable for a solution to this task. The Extreme Gradient Boosting Machines model and simple ensemble model provided the best results.