@InProceedings{MoserMcNiLiOlVi:2015:EsCaSa,
author = "Moser, Paolo and McRoberts, Ronald and Nicoletti, Adilson Luiz and
Lingner, D{\'e}bora Vanessa and Oliveira, Laio Zimermann and
Vibrans, Alexander Christian",
title = "M{\'e}todos para quantifica{\c{c}}{\~a}o da acur{\'a}cia de
estimativas de cobertura florestal a partir mapas baseados em
sensoriamento remoto e dados de invent{\'a}rio terrestre: um
estudo de caso em Santa Catarina, Brasil",
booktitle = "Anais...",
year = "2015",
editor = "Gherardi, Douglas Francisco Marcolino and Arag{\~a}o, Luiz
Eduardo Oliveira e Cruz de",
pages = "5951--5958",
organization = "Simp{\'o}sio Brasileiro de Sensoriamento Remoto, 17. (SBSR)",
publisher = "Instituto Nacional de Pesquisas Espaciais (INPE)",
address = "S{\~a}o Jos{\'e} dos Campos",
abstract = "Estimation of area of forest cover from remote sensing-based maps
is a very useful technique when data and qualified are available.
However, there is no a definitive conclusion about techniques that
compensate misclassification errors in the process. In addition,
estimating the bias of the estimates is an incipient technique. A
case study was conducted in the state of Santa Catarina in
southern Brazil, where estimates of four remote sensing-based land
cover maps were compared, using the estimate of Santa Catarina
Forest and Floristic Inventory as the ground truth. The inventory
data to assess the overall accuracy and the bias for the four
maps. The routine of analysis is based in X steps: (1)
classification errors; (2) construction of error matrices, (3)
calculating accuracy measures (4) estimating bias and (5)
constructing 95% confidence intervals for the estimates. To
correct the existing bias, a model-assisted estimator was used.
After the adjustments, the estimates for three of the four maps
were closer to the plot-based, simple random sampling estimates
than before adjustment. However, an independent t-test showed that
three of the four maps were statistically significantly different
from the simple random sampling estimate. The model-assisted
estimator is an easily implemented technique for adjusting for
estimated classification bias and for constructing real confidence
intervals.",
conference-location = "Jo{\~a}o Pessoa",
conference-year = "25-29 abr. 2015",
isbn = "978-85-17-0076-8",
label = "1228",
language = "pt",
organisation = "Instituto Nacional de Pesquisas Espaciais (INPE)",
ibi = "8JMKD3MGP6W34M/3JM4ERD",
url = "http://urlib.net/ibi/8JMKD3MGP6W34M/3JM4ERD",
targetfile = "p1228.pdf",
type = "Floresta e vegeta{\c{c}}{\~a}o",
urlaccessdate = "20 maio 2024"
}