@InProceedings{FelgueirasOrtiCama:2016:SpPrCa,
author = "Felgueiras, Carlos Alberto and Ortiz, Jussara de Oliveira and
Camargo, Eduardo Celso Gerbi",
affiliation = "{Instituto Nacional de Pesquisas Espaciais (INPE)} and {Instituto
Nacional de Pesquisas Espaciais (INPE)} and {Instituto Nacional de
Pesquisas Espaciais (INPE)}",
title = "Spatial predictions of categorical attributes constrained to
uncertainty assessments",
booktitle = "Proceedings...",
year = "2016",
organization = "Simp{\'o}sio Internacional SELPER, 17.",
note = "{Informa{\c{c}}{\~o}es Adicionais: Abstract} and This article
explores the use of nonlinear geostatistical procedures, known as
kriging and simulation indicator approaches, for spatial modeling
of categorical attributes. The categorical information is
initially represented by a set of sample points observed within a
spatial region of interest. The original sample set is used to
generate indicator fields take into account the classes of the
categorical data. The indicator fields, or indicator samples,
contain 0 and 1 attribute values according to the class they are
representing. Empirical and theoretical semivariograms are built
from the indicator samples to represent the spatial variation of
each class in relation to the others. The geostatistical
procedures, making use of the samples and the theoretical
semivariograms, allow obtaining an approximation of the stochastic
model, the conditioned probability distribution function (cpdf) of
the categorical attribute at any desired spatial location. From
any cpdf it is possible to assess optimal prediction, or estimate,
and uncertainty values associated to the stochastic model. Optimal
prediction as mean, median or any quantile values can be assessed.
Uncertainty values are obtained by means of the maximum cpdf
probability, Shannon entropy, or another criterion. The
uncertainty values can be used to qualify the predictions and can
also be considered to generate constrained spatial predictions, or
constrained classifications, that are important in decision
makings related to environmental planning activities, for example.
The concepts here presented are applied and tested in a case study
developed for a sample set of soil texture observed in an
experimental farm in the region of S{\~a}o Carlos city in
S{\~a}o Paulo State, Brazil. Four classes of soil texture are
considered, sandy, medium clay, clay and too clay, in order to get
the cpdf values. Some maps derived by constraints are presented
and analyzed considering different probability values from the
attribute stocha.",
keywords = "Spatial Analyzes, indicator geostatistics, Spatial Modeling of
Categorical Attributes, Uncertainty Assesments, Constrained
Classifications, Decision Making in Environmental Planning.",
conference-location = "Puerto Iguaz{\'u}, Misiones",
conference-year = "7-11 nov.",
label = "lattes: 2916855460918534 1 FelgueirasOrtiCama:2016:SPPRCA",
language = "en",
targetfile = "ConstrainedPredictionsv3.pdf",
url = "https://selperargentina2016.org/trabajos-aceptados/",
urlaccessdate = "15 jun. 2024"
}