Biodiversity Implications of Forest Disturbance and Biodiversity Implications of Forest Disturbance and Related Landscape Dynamics in the Brazilian Related Landscape Dynamics in the Brazilian
AmazonAmazon
Mark A. CochraneMark A. Cochrane1,21,2, David P. Roy, David P. Roy11, Carlos Souza Jr., Carlos Souza Jr.22, Jos Barlow, Jos Barlow33, , Eugenio ArimaEugenio Arima44, Izaya Numata, Izaya Numata11, Christopher P. Barber, Christopher P. Barber11, Luiz Mestre, Luiz Mestre11, ,
Rafael AndradeRafael Andrade11, and Sanath Kumar, and Sanath Kumar11
1 Geographic Information Science Center of Excellence, South Dakota State University, Brookings, SD USA2 IMAZON, Instituto do Homem e Meio Ambiente da Amazônia, Belém, PA Brazil
3 Lancaster University, Lancaster United Kingdom4 Hobart and William Smith Colleges, Geneva NY USA
What is the project?What is the project? The fundamental hypothesis underlying this project
is that the biodiversity levels of Amazonian forests are strongly related to two competing factors: forest disturbance and time since last disturbance
The Brazilian AmazonThe Brazilian Amazon
Amazon humid tropical forest biomes: ~ 6.4 km2
Portion in Brazil: ~ 4 km2
Source: WWF
Human Access to ForestHuman Access to Forest
Source: IBGE, IMAZON
85% of deforestation within 50 km of main roads
73,000 km of official roads in region
240,000 km of unofficial roads
Expansion rates > 40 km / 10,000 km2 / year
Forest wildfires: Interact with ongoing threats to the Amazon
How do we propose to test this?
Our approach is straightforward… 1) Determine recent forest disturbance history
across the Brazilian Amazon (2000-2009); 2) Conduct extensive field studies of indicator
taxa, stratified by disturbance history, to determine biodiversity responses;
3) Model the determinants of fire ignition and fire spread;
4) Predict the current and future levels of biodiversity similarity in disturbed forests spatially across the Brazilian Amazon.
Phase 1. Imagery Acquisition and Processinga) Paragominas,Pará State - 223/62
Soil
GV NDFI
NPV
Image Processing Steps
Shade
Soil
NPV
GV
Image RegistrationRadiance Conversion
CorrectHaze?
AtmosphericCorrection(ACORN)
Yes
No
Estimate Visibilityand water vapor
Apply Carlotto’s
Technique
(1) PRE-PROCESSING
Landsat
ReflectanceSpace
Pixel PurityIndex - (PPI)
VisualizationScatter matrix
Spectral curves
40 million pixels
(2) Build Spectral Library
Generic Image Endmembers
SVDC
(3) SMA
Landsat
NDFI
(4) Enhance and Detect Canopy Damage
ExtractPatios
CCA
CanopyDamage Soil ≥ 10%
1 pixel ≤ Area ≤ 4 pixels
NDFI ≤ 0.75
GV + NPV + Soil + Shade = 1
Souza Jr. et al. (2005), RSE
Haze Correction
Contaminated Image Corrected Image
Ji-Parana, 231/67 – R3, G2, B1
Normalized Difference Fraction Index
SoilNPVGV
Soil)(NPVGVNDFI
Shade
Shade
Shade100
GVGVShade
-1 ≤ NDFI ≤1
NDFI low to moderate
NDFI near 1
High GVLow NPV and Soil
Low to moderate GVModerate to high NPV and Soil
Souza Jr. et al. (2005), RSE
NDFI
226/68 - 2001 (Sinop - MT)
Roads
Logged
Forest
NDFI
226/68 - 2000 (Sinop - MT)
NDFI
226/68 - 2001 (Sinop - MT)
NDFI
226/68 - 2003 (Sinop - MT)
Mapping Forest Damage History
Characterizing Forest FragmentationCharacterizing Forest Fragmentation
Age map
Derived from Landsat time series Used for calculation of fragmentation
features
Forest
Pasture
S.G.Forest
Time series Land cover map Age map
>22 ys
1 y
2 ys
3 ys
4 ys
5 ys
6 ys
7 ys
8 ys
9 ys
10 ys
11 ys
12 ys
13 ys
14 ys
15 ys
16 ys
17 ys
18 ys
19 ys
20 ys
21 ys
1975
2005
Persistence of Forest Edge (Ariquemes)
y = 90.919x-0.522
R² = 0.9771
0
10
20
30
40
50
60
70
80
90
100
0 5 10 15 20
Rem
aini
ng e
dge
(%)
Edge age
Rem
ain
ing
edge
%
Edge length classification using ISO data
0
10
20
30
40
50
60
70
80
90
100
1975 1980 1985 1990 1995 2000 2005
Fore
st e
dge
leng
th (k
m)
Years
Different patterns of forest edge dynamicsClass 1Class 2Class 3Class 4Class 5
1975198419861988199019921994199619982000200220042005
Class 1
1975
Class 2
198419861988199019921994199619982000200220042005 1975
Class 4
1984198519861987198819891990199119921993199419951996199719981999200020012002200320042005
Phase 2. Biodiversity Field Studies
The spatial database of forest disturbance is used to stratify and interpret our field studies investigating the response of 4 major indicator taxa (birds, dung beetles, trees and ants) as a function of disturbance history and time since last disturbance.
Me
an
(±S
E)
corr
ela
tion
co
ee
ffic
ien
t
Butterfl
ies
Large
Mam
mal
s
Lizard
s
Birds
Trees
Arach
nids
Dung bee
tles
Moth
s
Carrio
n flie
s
Fruit
flies
Grass
hoppers
Smal
l mam
mal
s
Orchid
bee
s
Amphib
ians
Bats
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
Barlow et al. 2007 PNAS
Both birds and dung beetles are good indicators of community change in most other faunal groups
Barlow et al. 2007 PNAS
Both birds and dung beetles hold large numbers of species that are restricted to primary forest
% Species unique to primary forest
0 10 20 30 40 50 60 70
TreesBirds
AmphibiansLizards
Small mammalsDung beetles
BatsMoths
ButterfliesGrasshoppers
ArachnidsFruit flies
Scavenger fliesLarge mammals
Orchid bees Many wide ranging species
Mostly territorial or habitat specialists
Standardised survey cost ($)
0 2000 4000 6000 8000
% In
dica
tor
spec
ies
0
5
10
15
20
25
30
35
40
Dung beetles
Birds
Moths Small mammals
Gardner, Barlow et al. 2008 Ecology Letters
Both birds and dung beetles are highly cost effective to sample – you get good information on habitat integrity for a low cost
Selecting effective biodiversity indicators
Wide variability in cost of surveying different taxa Some taxa (e.g. birds and dung beetles) are of high
performance for evaluating forest management systems in the Brazilian Amazon because they: Are sensitive indicators of changes in forest integrity Can be surveyed cost effectively
Ants share similar ecological attributes as birds and dung-beetles They are cheap to sample Species rich, with many different functional groups Most do not move large distances from their colonies (if
you find them in a habitat, they come from that habitat).
Methods – Dung Beetles- Baited pitfall traps (human faeces)
- 5 traps per transect, run for 4 days.
- Follows methods discussed at global Scarabnet meetings, and used effectively in many Amazonian studies
- Complemented by un-baited flight intercept traps
Fire-mediated dieback and compositional cascade
Barlow and Peres 2008
Observation and recording
10 point counts per day per site 06:30 h - 09:00 h
10 min. bird observation and recording,
spaced at least 150m each other, intending to avoid double bird-counting (Parker, 1991).
Along each mist-net transects and other trails.
POINT COUNTS
Less individuals sampled compared to point counts
BUT It is independent of observer accuracy
Capture mainly understory birdsSamples species that are not singing
Possibility to mark (banding) and measure
MIST NETS
MEASURING
We will measure:
Wings
Tarsus
Bill
Weight
Expected Results
A large-scale comparison of effects of fire on Amazonian bird communities.
A long-term comparison of effects of fire on Amazonian bird communities.
One of the best overviews comparing Amazonian bird communities in different
Amazonian States (~500 sp).
Phase 3: Characterizing Fires (Starting June 2008)
Integration of Landsat based analyses of forest disturbance with MODIS-derived fire products will enable us to accurately separate fires into their three main types;
(1) deforestation fires, where slash is burned, creating relatively hot fires that burn for several hours;
2) maintenance fires, which rapidly burn as narrow fire lines through grass and early second growth;
3) forest fires, escaped fires in standing forests which vary from extremely low intensity in previously undisturbed forests to high intensity in previously burned or logged forests
Phase 4: Spatio-socioeconomic modeling (starting June 2008)
We will use spatial regressions of economic (farmgate prices for soybean and beef), physical-geographic (precipitation, soil types, vegetation types, distance to previous deforestation, and land protection status (e.g. indigenous lands, conservation units)) and land cover (disturbance history) factors to model probability surfaces of fire ignition and fire spread.
MODIS fire detections will be used to validate the ignition event model for 2000-2009 and the composite burned area product (Phase 3) will be used to validate the fire spread model over the same time period.
Once validated, the models will be run using likely economic and rainfall scenarios to create spatio-temporal predictions of disturbance frequency and expected biodiversity impacts for the 2010-2019 time period.
Reprise of Project Objectives1) Develop a basin-wide spatial database of all forest disturbance (selective
logging, fragmentation, fire, deforestation) from 2000-2009, based on NDFI analyses of annual Landsat imagery.
2) Derive regional estimation functions of expected biodiversity similarity based on disturbance history (disturbance metric) and time-since-last-disturbance (resilience metric) derived from stratified field data collected for four separate taxa (woody plants, birds, dung beetles and ants).
3) Develop a basin-wide spatial and temporal datasets of all fires by type (1) deforestation fires; 2) maintenance fires; 3) forest fires, using MODIS and Landsat data.
4) Model economic, physical-geographic and land cover factors affecting fire ignition and spread from 2000-2009 to create probability surfaces of fire ignition and fire spread.
5) Create a basin-wide map of probable biodiversity alterations in current standing forests across the Brazilian Amazon and predictions of future changes in these conditions over the next 10 years (2010-2019) based on likely economic and climate scenarios. (Starting late 2009)