Modeling the distribution of four vegetation alliances using generalized linear models and classification trees with spatial dependence

Jennifer Miller, Janet Franklin

Research output: Contribution to journalArticle

167 Citations (Scopus)

Abstract

Generalized linear models (GLMs) and classification trees were developed to predict the presence of four vegetation alliances in a section of the Mojave Desert in California. Generalized additive models were used to provide response shapes for parameterizing GLMs. Environmental variables used to model the distribution of the alliances included temperature, precipitation, elevation, elevation-derived terrain variables (slope, transformed aspect, topographic moisture index, solar radiation, and landscape position), and categorical landform/surface composition variables. Vegetation distributions exhibit spatial dependence and therefore we used indicator kriging to derive neighborhood values of "presence" also used as predictors in the models. The models were developed using 2859 observations coded present or absent for each of the four alliances, and assessed using 960 observations. In general, all of the models were improved with the addition of the kriged dependence term. However, models that relied heavily on the kriged dependence term were less generalizable for predictive purposes. Classification tree models had higher classification accuracy with the training data, but were less robust when used for predictions with the test data. Each of the models was used to generate a map of predictions for each alliance and the results were often quite different. The predicted maps with the kriged dependence terms looked unrealistically smooth, particularly in the classification tree models where they were often selected as the most important variables, and therefore heavily influenced the spatial pattern of the resulting map predictions.

Original languageEnglish (US)
Pages (from-to)227-247
Number of pages21
JournalEcological Modelling
Volume157
Issue number2-3
DOIs
StatePublished - Nov 30 2002
Externally publishedYes

Fingerprint

linear models
taxonomy
vegetation
modeling
prediction
topographic aspect
Mojave Desert
distribution
landscape position
landforms
kriging
solar radiation
landform
desert
environmental factors
moisture
spatial distribution
temperature
testing

Keywords

  • Classification tree
  • Generalized linear model
  • Indicator kriging
  • Mojave Desert
  • Predictive vegetation modeling

ASJC Scopus subject areas

  • Ecology, Evolution, Behavior and Systematics
  • Ecological Modeling
  • Ecology

Cite this

Modeling the distribution of four vegetation alliances using generalized linear models and classification trees with spatial dependence. / Miller, Jennifer; Franklin, Janet.

In: Ecological Modelling, Vol. 157, No. 2-3, 30.11.2002, p. 227-247.

Research output: Contribution to journalArticle

@article{daa956077bc94dc5a1d9fa8ebf3a92e4,
title = "Modeling the distribution of four vegetation alliances using generalized linear models and classification trees with spatial dependence",
abstract = "Generalized linear models (GLMs) and classification trees were developed to predict the presence of four vegetation alliances in a section of the Mojave Desert in California. Generalized additive models were used to provide response shapes for parameterizing GLMs. Environmental variables used to model the distribution of the alliances included temperature, precipitation, elevation, elevation-derived terrain variables (slope, transformed aspect, topographic moisture index, solar radiation, and landscape position), and categorical landform/surface composition variables. Vegetation distributions exhibit spatial dependence and therefore we used indicator kriging to derive neighborhood values of {"}presence{"} also used as predictors in the models. The models were developed using 2859 observations coded present or absent for each of the four alliances, and assessed using 960 observations. In general, all of the models were improved with the addition of the kriged dependence term. However, models that relied heavily on the kriged dependence term were less generalizable for predictive purposes. Classification tree models had higher classification accuracy with the training data, but were less robust when used for predictions with the test data. Each of the models was used to generate a map of predictions for each alliance and the results were often quite different. The predicted maps with the kriged dependence terms looked unrealistically smooth, particularly in the classification tree models where they were often selected as the most important variables, and therefore heavily influenced the spatial pattern of the resulting map predictions.",
keywords = "Classification tree, Generalized linear model, Indicator kriging, Mojave Desert, Predictive vegetation modeling",
author = "Jennifer Miller and Janet Franklin",
year = "2002",
month = "11",
day = "30",
doi = "10.1016/S0304-3800(02)00196-5",
language = "English (US)",
volume = "157",
pages = "227--247",
journal = "Ecological Modelling",
issn = "0304-3800",
publisher = "Elsevier",
number = "2-3",

}

TY - JOUR

T1 - Modeling the distribution of four vegetation alliances using generalized linear models and classification trees with spatial dependence

AU - Miller, Jennifer

AU - Franklin, Janet

PY - 2002/11/30

Y1 - 2002/11/30

N2 - Generalized linear models (GLMs) and classification trees were developed to predict the presence of four vegetation alliances in a section of the Mojave Desert in California. Generalized additive models were used to provide response shapes for parameterizing GLMs. Environmental variables used to model the distribution of the alliances included temperature, precipitation, elevation, elevation-derived terrain variables (slope, transformed aspect, topographic moisture index, solar radiation, and landscape position), and categorical landform/surface composition variables. Vegetation distributions exhibit spatial dependence and therefore we used indicator kriging to derive neighborhood values of "presence" also used as predictors in the models. The models were developed using 2859 observations coded present or absent for each of the four alliances, and assessed using 960 observations. In general, all of the models were improved with the addition of the kriged dependence term. However, models that relied heavily on the kriged dependence term were less generalizable for predictive purposes. Classification tree models had higher classification accuracy with the training data, but were less robust when used for predictions with the test data. Each of the models was used to generate a map of predictions for each alliance and the results were often quite different. The predicted maps with the kriged dependence terms looked unrealistically smooth, particularly in the classification tree models where they were often selected as the most important variables, and therefore heavily influenced the spatial pattern of the resulting map predictions.

AB - Generalized linear models (GLMs) and classification trees were developed to predict the presence of four vegetation alliances in a section of the Mojave Desert in California. Generalized additive models were used to provide response shapes for parameterizing GLMs. Environmental variables used to model the distribution of the alliances included temperature, precipitation, elevation, elevation-derived terrain variables (slope, transformed aspect, topographic moisture index, solar radiation, and landscape position), and categorical landform/surface composition variables. Vegetation distributions exhibit spatial dependence and therefore we used indicator kriging to derive neighborhood values of "presence" also used as predictors in the models. The models were developed using 2859 observations coded present or absent for each of the four alliances, and assessed using 960 observations. In general, all of the models were improved with the addition of the kriged dependence term. However, models that relied heavily on the kriged dependence term were less generalizable for predictive purposes. Classification tree models had higher classification accuracy with the training data, but were less robust when used for predictions with the test data. Each of the models was used to generate a map of predictions for each alliance and the results were often quite different. The predicted maps with the kriged dependence terms looked unrealistically smooth, particularly in the classification tree models where they were often selected as the most important variables, and therefore heavily influenced the spatial pattern of the resulting map predictions.

KW - Classification tree

KW - Generalized linear model

KW - Indicator kriging

KW - Mojave Desert

KW - Predictive vegetation modeling

UR - http://www.scopus.com/inward/record.url?scp=0037202441&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0037202441&partnerID=8YFLogxK

U2 - 10.1016/S0304-3800(02)00196-5

DO - 10.1016/S0304-3800(02)00196-5

M3 - Article

VL - 157

SP - 227

EP - 247

JO - Ecological Modelling

JF - Ecological Modelling

SN - 0304-3800

IS - 2-3

ER -