Using ecological niche models to support tree species ...

14
Using ecological niche models to support tree species selection for forest restoration planning in largely deforested regions Aitor Gastón and Juan I. García-Viñas ECOGESFOR-UPM, Research Group of Ecology and Sustainable Forest Management, Universidad Politécnica de Madrid, Spain

Transcript of Using ecological niche models to support tree species ...

Page 1: Using ecological niche models to support tree species ...

Using ecological niche models to support tree species selection for forest restoration planning in largely deforested regions

Aitor Gastón and Juan I. García-Viñas

ECOGESFOR-UPM, Research Group of Ecology and SustainableForest Management, Universidad Politécnica de Madrid, Spain

Page 2: Using ecological niche models to support tree species ...

The problem: how to select tree species for reforestation?Assessing habitat suitability besides other economic, social and legal criteria

Which tree speciesare suitable for this site?

A usual solution: visit an adjacent reference forest site

Page 3: Using ecological niche models to support tree species ...

Where is the adjacent reference forest site here?

0 20 4010Kilometers

Study areaTierra de Campos (Spain)

8,560 km2

Largely deforested region (92% non forested lands)

The reference site approach is not applicable to largely

deforested regions

Page 4: Using ecological niche models to support tree species ...

An alternative approach is to use ecological niche models

Species occurrences

Ecological factor 1

Ecological factor 2

Ecolog

ical fac

tor 2Ecological factor 1

Hab

itat suitability

Available ecological niche models for the study area focus in dominant tree species

Gandullo & Sánchez Palomares, 1994; Castejón et al., 1998; Gandullo et al., 2004; García López & Allúe Camacho, 2004; Sánchez Palomares et al., 2007; Sánchez Palomares et al., 2008; Alonso Ponce et al., 2010

We present a modeling framework:

· that supports tree species selection for reforestation

· applicable to largely deforested areas in Spain

· that include dominant and non-dominant tree species

Page 5: Using ecological niche models to support tree species ...

Model trainingClimatic predictors

modeled using ESTCLIMA, Sánchez Palomares et al. 1999

Variable reduction:extract first principal 

component of each group

mean water availability

mean thermal conditions

winter thermal conditions

summer thermal conditions

dry season water availability

Coarse scale soil map(1:1.000.00, Van Liedekerke et al. 2006)

Lithology: Calcareous, siliceous or gypsiferous

Mayor FAO soil group

Proved that improves model performance (Gastón & García Viñas, 2011)

Soil related predictors

Species occurrence data

Spanish Forest Map (Ruiz de la Torre, 1990)

120,938 vegetation records in 

continental Spain

70 native tree species

MODEL Ecological niche model• Penalized logistic regression (Harrell, 2001)

• Proved that improves model performance and make it comparable to current standards (e.g. MAXENT, Gastón & García Viñas, 2011)

• 4 knots restricted cubic splines

0 100 200 300 400 500

-6-5

-4-3

PE1

pred

ict(f

, new

data

= P

E1)

0 100 200 300 400 500

0.00

00.

010

0.02

0

PE1pl

ogis

(pre

dict

(f, n

ewda

ta =

PE1

))

0 100 200 300 400 500

-25

-20

-15

-10

-5

PE1

pred

ict(f

, new

data

= P

E1)

Page 6: Using ecological niche models to support tree species ...

Little overlapAcceptable discrimination

Evaluation of the predictive performance (I): Discrimination

Absences

Presences

Large overlapPoor discrimination

No overlapPerfect discrimination

Predictions

Evaluation of the predictive performance (II): Calibration

Mean AUC for tree species

models = 0.90

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Predicted probability

10.90.80.70.60.50.40.30.20.1

0Obs

erve

d fr

eque

ncy

Fit  a  calibration  curve (predicted vs. observed)

Calculate  the  slope  of the curve

Good  calibration  when the slope is near 1

Mean calibration

slope for tree species

models = 0.93

Page 7: Using ecological niche models to support tree species ...

Applications to forest restoration planning in largely deforested regions

At the project scale

At the regional scale

Page 8: Using ecological niche models to support tree species ...

Near Izagre (León)Altitude 800 m

Mean annual temperature 11.2 ºCMean max. temp. warmest month 28.9 ºCMean min. temp. coldest month ‐1.4 ºC

Mean annual rainfall 600 mmMean summer rainfall 80 mmSiliceous parent material

0.25Q. pyrenaica

0.07Juniperus thurifera

0.07P. pinea

0.08P. nigra

0.18Q. faginea

0.22Pinus pinaster

0.61Querus ilex

Occurrence probabilitySpecies

Dominant species

0.10Crataegus monogyna

0.01Rhamnus catharticus

0.02Prunus spinosa

0.06J.communis

Occurrence probabilitySpecies

Non‐dominant species

Lists of suitable tree species(if the predicted probability is higher than the prevalence, Liu et al. 2005)

Restoring a single site, an example

Page 9: Using ecological niche models to support tree species ...

Planning forest restoration at the regional scalee.g., mapping the two most suitable dominant tree species

Mixed forests predicted in 49% of the area

Page 10: Using ecological niche models to support tree species ...

Coniferous forests predicted in 24% of the area

Mixed forests predicted in 49% of the area

Planning forest restoration at the regional scalee.g., mapping the two most suitable dominant tree species

Page 11: Using ecological niche models to support tree species ...

Broadleaf forests predicted in 27% of the area

Coniferous forests predicted in 24% of the area

Mixed forests predicted in 49% of the area

Planning forest restoration at the regional scalee.g., mapping the two most suitable dominant tree species

Page 12: Using ecological niche models to support tree species ...

Planning forest restoration at the regional scalee.g., identify the suitable non‐dominant species for biodiversity oriented reforestation in the region an their potential distribution

0.05Tilia cordata

0.19Taxusbaccata

1.97Sorbus torminalis

2.10Sorbus domestica

0.01Sorbus aria

8.95Sambucus nigra

31.69Rhamus catharticus

45.28P. spinosa

0.52Prunus avium

2.03Malus sylvestris

1.66Frangula alnus

15.67Cornus sanguinea

5.97A. monspessulanum

0.06Acer campestre

0.23Quercus coccifera

3.71Populus tremula

59.44J. oxycedrus

38.01Juniperus communis

1.84Euonymus europaeus

52.93Crataegus monogyna

3.95Amelanchier ovalis

% areaSpecies

Page 13: Using ecological niche models to support tree species ...

Conclusions• Using ecological niche models to support tree species 

selection for forest restoration is crucial in largely deforested regions because adjacent reference forest sites are lacking.

• The proposed methodology allows fitting ecological niche models for every native tree species potentially suitable for forest restoration in Spain.

• The adopted modeling strategy takes advantage of freely available data of species distributions and environmental predictors.

• The models provide occurrence probabilities for every tree species in every site, allowing further analysis that may support forest restoration planning.

Page 14: Using ecological niche models to support tree species ...

Thank you for your attention