Model Selection for Predictive Species Range Mapping - PowerPoint PPT Presentation

Loading...

PPT – Model Selection for Predictive Species Range Mapping PowerPoint presentation | free to download - id: 164939-ZDc1Z



Loading


The Adobe Flash plugin is needed to view this content

Get the plugin now

View by Category
About This Presentation
Title:

Model Selection for Predictive Species Range Mapping

Description:

Model Selection for Predictive Species Range Mapping – PowerPoint PPT presentation

Number of Views:29
Avg rating:3.0/5.0
Slides: 26
Provided by: dharm9
Learn more at: http://www.iaia.org
Category:

less

Write a Comment
User Comments (0)
Transcript and Presenter's Notes

Title: Model Selection for Predictive Species Range Mapping


1
Model Selection for Predictive Species Range
Mapping
  • Changwan Seo, UOS
  • James Thorne, UC Davis
  • David Stoms, UCSB
  • Wilfried Thuiller, CNRS France
  • Frank Davis, UCSB
  • Lee Hannah, Conservation International/UCSB

2007. 06. 07 IAIA07
2
Whole Earth
Apollo 17, December 1972
3
What species at risk, where and when?
4
(No Transcript)
5
Aloe dichotoma (quiver tree)
?
6
Purposes
  • 1. To find the reasonable model for species
    modeling
  • Building and validating individual models
  • Reducing the uncertainty of each model
  • 2. To detect the change of species distribution
    under the climate change
  • Projecting a model to the future environment

7
The process of study
Selecting species environmental data
GLM/GAM/GBM/CART/ANN
Building validating models
Bayesian Approach
Projecting models to the future environment
Optimizing model outputs
Consensus Model
Estimating the change of species range
8
Study Area
9
Tree species
Symbol GENUS SPECIES suffix
ABMA Abies magnifica  
ACNEC2 Acer negundo var. californicum
AECA Aesculus californica
COSE3 Cornus sessilis  
FRDI2 Fraxinus dipetala
JUCA(southern) Juglans californica
LIDE3 Lithocarpus densiflorus  
PIAT Pinus attenuata
PIBA Pinus balfouriana  
PIBR Pivea breweriana
PICO3 Pinus coulteri
PILA Pinus lambertiana
PIMO Pinus monophylla  
PIMU Pinus muricata
PISA2 Pinus sabiniana
PLRA Platanus racemosa
PSMA Pseudotsuga macrocarpa
QUAG Quercus agrifolia
QUCH2 Quercus chrysolepus  
QUDO Quercus douglasii
QUEN Quercus engelmannii  
QUGA4 Quercus garryana  
QULO Quercus lobata
QUWI2 Quercus wislizenii
SESE3 Sequoia sempervirens  
TOCA Torreya californica
UMCA Umbellularia california  
27 endemic species
10
Tree Plots
18,000 1km scale-up points
11
Climate data
Environmental Data
BIO5 Max Temp of Warmest Month
BIO6 Min Temp of Coldest Month
BIO7 Temp Annual Range 5-6
BIO8 Mean Temp of Wettest Quarter
BIO9 Mean Temp of Driest Quarter
BIO16 Precipitation of Wettest Quarter
BIO18 Precipitation of Warmest Quarter
Soil data
AWCL Available water capacitylow level (cm)
DEPL Soil depthlow level (m)
PHL pHlow level
SALL Salinitylow level (mmhos/cm )
WTDL Depth to water tablelow level (m)
12
Bio6 Min Temp of Coldest Month
13
Bio16 Precipitation of Wettest Quarter
14
AWCL PHL
15
BioMOD (Thuiller et al 2003)
  1. GLM (Generalized Linear Model)
  2. GAM (Generalized Addictive Model)
  3. GBM (Generalized Boosted Model)
  4. CART (Classification And Regression Tree)
  5. ANN (Artificial Neural Network)

16
Individual Model Suitability
The probability map under the present climate
1) GLM 2) GAM
3) CART 4) ANN
The probability map under the future climate
17
Bayesian Approach (Pereira and Itami 1991)
Pnew The revised (combined) probability
estimate Pvar1 The probability derived from a
first model Pvar2 The probability from a second
model
18
Bayesian Suitability
The probability map under the current climate
The probability map under the future climate
1) Climate 2) Soil
3) ClimateSoil 4) Bayesian
19
Consensus model (PCA)
The predicted distribution under current climate
GBM
GAM
GLM
ANN
The predicted distribution under future climate
CART
20
Model performance (AUC)
      Mean LCL UCL
GLM 0.895 0.869 0.921
GBM 0.979 0.969 0.988
Climate Climate GAM 0.951 0.935 0.966
CTA 0.931 0.915 0.948
ANN 0.951 0.936 0.966
PCA 0.962 0.951 0.972
GLM 0.776 0.742 0.809
GBM 0.948 0.934 0.963
Soil Soil GAM 0.869 0.84 0.899
CTA 0.928 0.905 0.952
ANN 0.881 0.855 0.906
PCA 0.925 0.908 0.948
GLM 0.904 0.882 0.926
GBM 0.978 0.969 0.987
Bayesian Bayesian GAM 0.951 0.936 0.965
CTA 0.960 0.947 0.973
ANN 0.955 0.943 0.968
PCA 0.965 0.953 0.976
GLM 0.911 0.889 0.933
GBM 0.983 0.974 0.991
Climate Soil Climate Soil GAM 0.958 0.945 0.972
CTA 0.934 0.916 0.952
ANN 0.964 0.953 0.975
    PCA 0.971 0.961 0.982
21
Range shift
Range Shift Q. douglasii
Present
HADCA2A 2080
Change
22
Range Shift Q. garryana
23
Range Shift P. contorta
24
Scenario Divergence HadCM3
A2
2050
2080
B2
Present
27 Endemic Tree Species
25
Discussion
  • 1. GBM was the best model among 5 individual
    models
  • 2. Bayesian model improved model performance and
    predicted realistic species range compared to
    individual bioclimatic and soil model
  • 3. Consensus model was a reasonable approach to
    reduce the uncertainty of individual model outputs
About PowerShow.com