@PhDThesis{Klippel:2017:DiAtPr,
author = "Klippel, Sandro",
title = "Modelos de distribui{\c{c}}{\~a}o de esp{\'e}cies de peixes
demersais marinhos da plataforma sul do Brasil:
distribui{\c{c}}{\~a}o atual e proje{\c{c}}{\~o}es futuras em
cen{\'a}rios de mudan{\c{c}}as clim{\'a}ticas",
school = "Instituto Nacional de Pesquisas Espaciais (INPE)",
year = "2017",
address = "S{\~a}o Jos{\'e} dos Campos",
month = "2016-08-04",
keywords = "rela{\c{c}}{\~o}es esp{\'e}cie-ambiente, habitat essencial de
peixes, boosted regression trees, modelos aditivos generalizados,
MAXENT, species-environment relationships, essential fish habitat,
generalized additive model.",
abstract = "Modelos de distribui{\c{c}}{\~a}o de esp{\'e}cies (MDE)
s{\~a}o ferramentas para a obten{\c{c}}{\~a}o de mapas de
aptid{\~a}o de habitat com base em ocorr{\^e}ncias
hist{\'o}ricas das esp{\'e}cies e vari{\'a}veis ambientais.
Neste trabalho, quatro t{\'e}cnicas diferentes foram comparadas:
O procedimento AquaMaps, o algoritmo de m{\'a}xima entropia
(MAXENT), modelos aditivos generalizados (GAM), e as
\${''}\$Boosted Regression Trees\${''}\$ (BRT). AquaMaps e
MAXENT foram aplicados utilizando apenas dados de presen{\c{c}}a
enquanto GAM e BRT utilizaram dados de
presen{\c{c}}a-aus{\^e}ncia. Usando essas t{\'e}cnicas, foram
desenvolvidos MDEs para 65 peixes marinhos. Dados
extra{\'{\i}}dos de 3.167 lances de arrasto de fundo de
pesquisas cient{\'{\i}}ficas realizadas na plataforma sul do
Brasil foram utilizados nos modelos. A capacidade preditiva dos
modelos foi avaliada usando a {\'a}rea sob a curva
caracter{\'{\i}}stica de opera{\c{c}}{\~a}o do receptor (AUC)
e valida{\c{c}}{\~a}o cruzada. Embora as abordagens de
presen{\c{c}}a-aus{\^e}ncia tenham produzido um n{\'u}mero
maior de modelos com boa capacidade preditiva do que as
t{\'e}cnicas que utilizam apenas dados de presen{\c{c}}a; alguns
modelos que utilizam apenas dados de presen{\c{c}}a tiveram um
desempenho equivalente. MDEs foram usadas para projetar a
{\'a}rea de distribui{\c{c}}{\~a}o de peixes demersais marinhos
sob diferentes cen{\'a}rios de aquecimento das {\'a}guas do
Atl{\^a}ntico Sudoeste. As mudan{\c{c}}as projetadas para
2\$^{o}\$C de aquecimento incluem uma redu{\c{c}}{\~a}o de
15-34\% no habitat adequado para tr{\^e}s esp{\'e}cies
comercialmente importantes de peixes marinhos na plataforma sul do
Brasil, considerando prefer{\^e}ncias t{\'e}rmicas e de
profundidade, e um deslocamento para o sul do centro da {\'a}rea
de distribui{\c{c}}{\~a}o que chega a 200 km. A
redu{\c{c}}{\~a}o e o deslocamento das {\'a}reas de
distribui{\c{c}}{\~a}o dessas esp{\'e}cies, que s{\~a}o os
principais recursos demersais explorados pela pesca comercial na
regi{\~a}o, pode reduzir a disponibilidade das esp{\'e}cies para
as frotas nacionais com graves consequ{\^e}ncias econ{\^o}micas
e sociais. Al{\'e}m disso, foram desenvolvidos MDEs para cinco
esp{\'e}cies de elasmobr{\^a}nquios amea{\c{c}}adas, e
comparados com o conhecimento atual sobre essas esp{\'e}cies.
Esses modelos fizeram previs{\~o}es razo{\'a}veis usando a
grande cobertura espacial e temporal de sensoriamento remoto. Tal
desempenho {\'e} particularmente {\'u}til para restringir
{\'a}reas de pesca, ou mesmo para deslocar os pescadores para
{\'a}reas com menor probabilidade de capturas acidentais.
ABSTRACT: Species distribution models (SDM) are tools to obtain
habitat suitability maps based on historical species occurrences
and environmental variables. In this work, four different
techniques were compared: The AquaMaps procedure, maximum entropy
algorithm (MAXENT), general additive models (GAM), and Boosted
Regression Trees (BRT). AquaMaps and MAXENT were applied using
presence-only data whereas GAM and BRT used presence-absence data.
Using these techniques, SDMs for 65 marine fishes were developed.
Data drawn from 3,167 bottom trawl hauls of scientific surveys
carried out on South Brazil Shelf were used in the models. The
predictive ability of the models was assessed using the area under
receiver operating characteristic curve (AUC) and 10-fold
cross-validation. While presence-absence approaches have been
found to produce more models with greater predictive ability than
presence-only approaches; some presence-only models can perform
almost as well. SDMs were used to project the distribution area of
marine demersal fishes under different warming scenarios of the
Southwestern Atlantic waters. The projected changes under
2\$^{o}\$C warming include a decrease of 1534\% in suitable
habitat for three commercially important marine fish species in
Southern Brazil, considering thermal and depth preferences, and a
shift to the south of the center of the distribution area that
reaches 200 km. The reduction and displacement of distribution
areas of those species, which are the main demersal resources
exploited by commercial fisheries in the region, may reduce the
availability of those species to national fleets with serious
economic and social consequences. In addition, SDMs for five
endangered elasmobranchs species were developed and compared with
the current knowledge about those species. Those models make
reasonable predictions using the great spatial and temporal
coverage of satellite data. Such performance is particularly
useful to restrict fishing grounds, or even to prompt fishers to
move to areas with lesser probability of incidental catches.",
committee = "Valeriano, Dalton de Morisson (presidente) and Vinhas, L{\'u}bia
(orientadora) and Kampel, Silvana Amaral (orientadora) and Kampel,
Milton and Santos, Roberta Aguiar dos and Silva, Ant{\^o}nio
Olinto {\'A}vila da",
copyholder = "SID/SCD",
englishtitle = "Species distribution models of marine demersal fishes of South
Brazil Shelf: current distribution and future projections under
climate change scenarios",
language = "pt",
pages = "295",
ibi = "8JMKD3MGP3W34P/3M7GPK5",
url = "http://urlib.net/ibi/8JMKD3MGP3W34P/3M7GPK5",
targetfile = "publicacao.pdf",
urlaccessdate = "27 abr. 2024"
}