Indicators for Sustainable Forest Management in Cultivated Forests

International Conference
Forsee Project final Meeting

Landscape metrics for characterization of forest

landscapes in a sustainable management framework:
Potential application and avoidance of misuse

Emilio Díaz-Varela
Pedro Álvarez Álvarez
Fernando Solla-Gullón
Antonio Rigueiro Rodríguez
Rafael Crecente Maseda
 Overview

GENERAL OVERVIEW OF
PRESENTATION

Use of Landscape Indices

OBJECTIVES and INTRODUCTION
Scale Issues
Basic cartography
MATERIAL and METHODS
Indices and application

RESULTS

DISCUSSION and CONCLUSIONS

 Introduction

Objectives and
Introduction
The use of landscape indices
Scale issues
 Introduction

OBJECTIVES

CHARACTERIZATION OF FOREST LANDSCAPES USING METRICS

Use of landscape metrics for characterising:

 Fragmentation of forest habitats
 Connectivity between forests
 Ecological diversity and heterogeneity in forest
landscapes

ASSESSMENT OF MISUSE ASSOCIATED TO UNSUITABLE SCALES

Analysis of results to check:

 If the observation or cartographic scale is the proper one
 If the scale of analysis suits the objectives of research
 The errors derived from using unsuitable scales
 The use of landscape indices

DEFINITION AND APPLICATIONS

LANDSCAPE INDICES (OR METRICS)

Quantitative indices to describe structures or patterns in
the landscapes (O’Neill et al., 1988)

USES OF LANDSCAPE METRICS

• Diversity in landscapes that suffered wildfires (Romme, 1982)

• Detection of differences in landscape pattern (O’Neill et al., 1988)
• Assessment of destruction and recovery processes in landscapes (Herzog et al., 2001)
• Assessment of urban ecological networks (Cook, 2002)
• Decision making in landscape ecological planning (Botequilha & Ahern, 2002)
• Differentiation between native and introduced forests in Galicia (Saura & Carballal, 2004)
• (…)
 The use of landscape indices

EXAMPLES

Composition metrics Proportion of habitats

Absolute / relative area of each habitat /class
Diversity
…

Configuration metrics Geometric characteristics of patches or landscapes

(perimeter / area relationships, core area, fractal
dimension, etc)
Connectivity and vicinity analysis
…
 The use of landscape indices

LEVELS OF CALCULATION

PATCH LEVEL

Results refer to every single

patch in the landscape

CLASS LEVEL

Results summarize the index

for each class or type of land
cover / habitat

LANDSCAPE LEVEL

Results summarize the index

for the whole set of patches in
the landscape
 Scale issues

MAIN PROBLEMS IN CALCULATION OF INDICES

Degree of dissimilarity between data used and real world

Adequacy of area analyzed to the chosen dataset
Sensibility of indices to changes in grain / extension

Li & Wu, 2004

 Scale issues

Degree of dissimilarity between data used and real world

Adequacy of area analyzed to the chosen dataset
Sensibility of indices to changes in grain / extension

Scale of application of indices is critical for inference of results

Low resolution spatial datasets could hide fine-scaled landscape patterns

Any inferences about scale-dependency in a system are constrained

by the grain and extent of investigation

Li & Wu, 2004

Corry & Nassauer, 2005
Thompson & McGarigal, 2002
 Scale issues

SCALE AND SUSTAINABLE MANAGEMENT

CORRESPONDENCE BETWEEN SCALES

Lack of integration between landscape functional scales, and planning and

management scales is currently one of the main issues in environmental
management (Norton, 1992; Wiens et al., 2002)

DEFINITION OF SUITABLE SCALES FOR MANAGEMENT

Environmental conflicts are to be solved at the ecological scales in which they
occur (Poiani et al., 2000; Saunders & Briggs, 2002)

MANAGEMENT PARADOX
Management efforts more likely create an effect at a fine scale, whereas success
regarding sustainability is more likely to be achieved at a broad scale (Forman, 1995)
 Material and methods

Material and
Methods
 Material and methods

CARTOGRAPHIC SOURCES

Photo-interpretation
National Forest
National Map
Forest Map
Map
(E 1 : 50.000)

Photo-interpretation Map
(E aprox 1 : 2.000)

EUNIS
 Material and methods

SELECTED METRICS

FRAGMENTATION

NP, “Number of patches” AREA_MN, “Mean patch area”

m n
1
NP = ni ∑
i= 1
∑j = 1 ij 10000 )
a ⋅ (
AREA _ MN =
N

ED “Edge Density” SHAPE_AM, “Area-Weighted Mean Shape Index”

  
m
n 
 
∑ eik
SHAPE _ AM = ∑
m

pij
∑j = 1  min p  aij 
ED = k=1
(10000)  m n 
A
i= 1

ij
 ∑ ∑  
  i= 1 j= 1 
 Material and methods

SELECTED METRICS

DIVERSITY

SHDI, “Shannon Diversity Index” SHEI, “Shannon Evenness Index”

m
m
SHDI = − ∑ ( Pi ⋅ ln Pi )
− ∑i= 1 ( Pi ⋅ ln Pi )
i= 1 SHEI =
ln m

CONNECTIVITY

ENN_MN, “Mean Euclidean Nearest Neighbour”

m n

∑ ∑
i= 1 j= 1
hij
ENN _ MN =
N
 Material and methods

SCALES AND LEVELS

Scales: Supra-Local

Local

Stand Buffer

Irixoa

Pilot area

Levels:
Guitiriz
Patch
Aranga
Class

Landscape
 Material and methods

FIELD STANDS

Information was obtained by field work for each 10m radius stand

Geology, lithology
Soil type
Soil depth
Soil chemistry
Biomass
Number of species
Type of species
Species’ diversity
...
 Material and methods

STATISTICAL ANALYSIS

Correlation analysis

Spearman correlation analysis was used for basic analysis of correlation between the
results in the metrics, and the field sampled data

Spearman correlation analysis is a non-parametric method, thus make no assumptions on

the distribution of the values of the variables

6∑ D 2
1−
N ( N 2 − 1)
 Material and methods

STATISTICAL ANALYSIS

CART (Classification and Regression Tree)

Classification of metrics results depending on other variables regarding forest stand

characteristics

Metrics’ results dataset is spilt in successive branches in a binary tree

Each division is made based in the maximum reduction in impurity between mother and
daughter groups (or branches)
Impurity: total sum of squares of the deviations of the individual responses from the mean
response of the group

∆ i ( s, t p ) = i (t p ) − p L i (t L ) − p R i (t R )

n
i (t ) = ∑
i= 1
( yn − y (t ))2
 Results

Results
 Results

ISSUES IDENTIFIED IN DATA

Photo-interpretation National Forest Map

Map
1 : 2.000 1 : 50.000

Modifiable Areal Unit Problem (MAUP)

Openshaw, 1984; Marceau, 1999

Aggregation of spatial data

Coarse scales need to aggregate data (but that can lead to undesired results)
Misclassification
Coarse typological scales, or simply errors, can lead to it
 Results

PATCH-LEVEL METRICS National Forest Map

AREA_MN

SHAPE_MN

ENN_MN
 Results

PATCH-LEVEL METRICS Photo-interpretation

Map
ENN_MN
SHAPE_MN
AREA_MN

AREA_MN

SHAPE_MN

ENN_MN
 Results

LANDSCAPE-LEVEL METRICS Supra-Local and local scales

National Forest Map Photo-interpretation Map

Aranga Aranga
Aranga Guitiriz + Aranga Guitiriz +
Guitiriz Guitiriz

NP 560 + 362 =/ 491 1.897 + 3.943 =/ 5.208

AREA_MN 21,29 80,74 83,81 6,30 7,45 7,93

SHAPE_AM 3,37 5,00 4,64 5,78 6,84 6,81

ENN_MN 155,83 329,22 341,74 171,64 127,68 138,82

SHDI 1,60 1,47 1,62 1,57 1,44 1,52

 Results

LANDSCAPE-LEVEL METRICS Stand – Buffer scale

National Forest Map NP AREA_MN SHAPE_AM ENN_MN SHDI

7 11,22 1,80 215,42 1,22

Photo-interpretation Map

117 0,67 2,38 42,57 1,79

 Results

CORRELATION ANALYSIS Stand – Buffer scale

National Forest Map

Nº species Shannon
Biomass
shrubland index

r Sign r Sign r Sign

. . .
SHAPE_AM + 0,450 0,013 - 0,237 0,208 0,101 0,608

AREA_MN - 0,432 0,017 0,207 0,270 - 0,113 0,566

SHDI 0,333 0,072 - 0,186 0,325 0,137 0,487

 Results

CART ANALYSIS – Landscape level Stand – Buffer scale

Forest dominant species (dep.) - Results for SHAPE_AM (indep.)

National Forest Map Photo-interpretation

Map

Pinus radiata Pinus pinaster Eucalyptus globulus

 Results

CART ANALYSIS – Landscape level Stand – Buffer scale

Forest dominant species (dep.) - Results for SHAPE_MN (indep.)

National Forest Map Photo-interpretation

Map

Pinus radiata Pinus pinaster Eucalyptus globulus

 Discussion and conclusions

Discussion and
conclusions
 Discussion and conclusions

SCALE OF BASIC CARTOGRAPHY CAN CONDITION RESULTS IN THE

APPLICATION OF LANDSCAPE INDICES

Coarse-grained data could be inadequate in cases of severely

fragmented landscapes, as can derive misleading results for the
calculated indices

Adequacy of specific cartography depends on the intrinsic

characteristics of the landscape

However, coarse-grained data could be of use for detecting

general patterns, for which fine-grained data could be too
“noisy”
 Discussion and conclusions

CORRELATION BETWEEN FIELD DATA AND LANDSCAPE METRICS DEPENDS

ON FUNCTIONAL SCALES OF PATTERNS / PROCESSES

Landscape metrics can fail to reflect fine-scaled landscape

patterns if the resolution is too coarse (Thompson & McGarigal, 2002;
Corry & Nassauer, 2005)

Extension of phenomena like plant community distribution (sub-

canopy level) presents site variation that general cartography
could not reflect

The scale of analysis must match the structure or process under

investigation before the search for any correlation between them
 Discussion and conclusions

AVOIDANCE OF SCALE ISSUES MUST ADRESS MULTISCALE ANALYSIS AND

DETECTION OF THRESHOLDS / DOMAINS OF SCALE…
SCALE

Variation of “scale of analysis” (i.e., using different radius in

buffers around sample stands) and comparison of results with
stand data (Thompson & McGarigal, 2002)

Pattern analysis of whole landscapes using “moving window”

analysis (Eiden et al., 2000; Riitters et al., 2002; Diaz, 2005)
 Discussion and conclusions

…AS WELL AS AVOIDANCE OF INADEQUATE CHOICE OF SCALE OR

AGREGATTION METHODS (MAUP ERRORS)

Data filtering and re-sampling must take into account differences

in map information

Legend (i.e., typological scale) must be chosen accordingly to

the spatial scale at which data is presented

Scale of spatial data must match the objectives of research and

analysis
Indicators for Sustainable Forest Management in Cultivated Forests
International Conference
Forsee Project final Meeting

THANK YOU VERY MUCH!!