Detecting desertification
in savanna systems
RJ (Bob) Scholes
Council for Scientific and Industrial Research
Natural Resources and Environment
Pretoria, South Africa
[email protected]
What is desertification?
[Hyperarid]
Arid
Semiarid
Subhumid
Desertification is degradation in dry lands
United Nations Convention on Combatting Desertification
Global distribution of savannas
(basically the same as the summer-rainfall drylands)
Savannas are mixed tree-grass systems
Some facts about savannas
• World’s largest land biome (~12%)
• Second largest Net Primary Productivity and
carbon store (16.8 x 1015 gC/y)
• Large ‘natural’ impact on atmosphere through fire,
carbon cycle and dust
• Home to ~1 billion people; source of food and
energy to most of them
• Centre of biodiversity (7000 species in Africa
alone)
Desertification sucks!
The single environmental issue affecting the livelihoods of
the largest number of people worldwide
41% of Earth’s land surface
2 billion inhabitants
90% in developing countries
What is degradation?
Does not spontaneously return to historical
average within a reasonable time when you
reduce the cause of change
Degradation is a persistent decrease in
the capacity of an ecosystem to deliver
ecosystem services
The benefits that people get from
ecosystems
-Millennium Ecosystem Assessment
Ecosystem Services
Millennium Ecosystem Assessment Classification
Provisioning
Regulating
•Crops
•Climate
•Floods
•Pest regulation
•Disease control
•Grazing and livestock
•Timber and fuel
•Fisheries
•Water
•Medicine
Cultural
•Aesthetic
•Recreation
•Amenity
•Spiritual
•Educational
Supporting
• Net Primary Production
• Nutrient cycling
Biodiversity: the variety of Life on Earth
Summary so far…
To detect desertification, we are looking for
a change in state in savanna ecosystems
Ecohydrological state change
(typically on silty soils)
runoff
-
rainfall
infiltration
+
+
soil water
+
plant production
plant cover
+
-
+
herbivory
Nutrient state change
(typically on less-fertile, erodible soils eg sandy loams or loess)
harvest
rainfall
+
+
+
plant production
-
nutrients
+
-
-
herbivory
-
+
soil erosion
fire
export from the
landscape
vegetation and
litter cover
Change in the rain use efficiency
Measure of plant production
Eg ΣFAPAR or similar
undegraded
Decrease in slope
or position
degraded
Rainfall in growing season
Production in water-limited
pastoral systems
Tree growth
competition
Rainfall
soils
Grass growth
Animal
production
An electrical analog
What controls the brightness of the lamp?
productivity
Sun
nutrients
water
In pulsed systems, water availability controls
the duration of growth opportunity
Linear relation between
grass production and rainfall
Grass AG NPP (g/m2/y)
500
Slope=
‘rain use efficiency’
400
300
clay soil
sand soil
200
100
Intercept = index of un
productive water loss
0
0
200
400
600
Annual rainfall
800
1000
The ‘inverse texture
hypothesis’ of
Noy-Meir
Interannual variability of grass
production is higher on clays than
sandy soils
Grass AG NPP (g/m2/y)
500
400
σclay
300
clay soil
sand soil
200
σsand
100
σrain
0
0
200
400
600
Annual rainfall
800
1000
Rain Use Efficiency
(g/m2/mm)
Rain use efficiency (kg/ha/mm)
12
10
8
6
4
2
0
50
60
70
80
Sand %
90
100
Intercept
dependent on soil water holding capacity
1500
1000
500
(kg/ha/y)
Intercept of AGGNPPvs Rain
2000
0
-500 0
5
10
-1000
-1500
-2000
-2500
-3000
Rain us e e fficie ncy (k g/ha/m m )
15
Changes in vegetation composition
(to species less productive or less useful)
+
+
-
browsers
-
rain
+
Woody
plants
grass
fire
+
+
-
grazers
Annual grass aboveground
production (g/m2/y)
Tree mass has a non-linear inverse effect on
grass production
1200
Walker et al
Beale
Donaldson & Kelk
Aucamp et al
900
600
300
0
0
0.2
0.4
0.6
0.8
1
Fraction of maximum tree quantity
Scholes RJ 2003 Convex relationships in ecosystems containing trees and grass.
Environment and Resource Economics 26, 559-574
Graphical stability analysis
Grass production
Open, grassy ‘parkland’
Low resilience
∆w=0
∆g=0
high resilience
Tree cover
Dense tree cover thicket
‘bush encroached’
Scholes RJ 2003 Convex relationships in ecosystems containing trees and grass.
Environment and Resource Economics 26, 559-574
The actual tree-grass interaction is much more
complex
seeds
Browsers
Flame height
Grazers
Slows growth
Reduces
Supresses
Nutrients
Grass biomass
Flame height = f(grass fuel)
Fire frequency
Rainfall
Relation between rainfall and quantity of trees in savannas
Sankaran et al 2005 Determinants of woody cover in African savannas Nature 438, 846-9
Upward shift detected
by albedo change
or phenology change
Vegetation metrics
Remote sensing
Ecological
SAR interferometry
biomass
height
LIDAR time-to-pulse
BDRF
Albedo
‘degradation’
intra- or interseasonal
variability
greeness
indices
‘structural’
% cover
NDVI
SI
EVI
etc
spectral reflectance
In key bands
spectral features
basal area
phenology
FAPAR
ecosystem
type & area
NPP
LAI
‘functional’
ET
‘compositional’
‘Greeness’ Vegetation Indices
eg NDVI= (Red-NIR)/(Red+NIR)
and many others (SI,SAVI,EVI etc)
Stable reference point
reflectance
Chlorophyll
absorption
400
1000
Wavelength (nm)
Red = reflectance @ 680nm
NIR = reflectance @ 900 nm
Leaf Area Index
• The one-sided leaf area per unit of ground area
• Can go down to near 0 in the dry season
• Theoretical limit about 6
I=I0e –K*LAI (Beer’s Law)
therefore for LAI=6, k=0.5: I=5% of Io
• Hard to measure in forests due to saturation above ~
3
• Seldom above 3 in drylands
– Typically ~1 in savannas
• Useful for estimating evaporation and interception,
but FPAR has fewer assumptions for modelling
productivity
FAPAR
Fraction of Absorbed Photosynthetically-Active Radiation
intercepted by the vegetation canopy (range 0-1)
GPP= ε Σ(PAR*FPAR) *f (stress,nutrients)
NPP=GPP-Ra
GPP = Gross Primary productivity (g/m2) (dry matter = 42% C)
ε
= Radiation use efficiency (gC/MJ),
species dependent (0.3-0.8 gC/MJ)
NPP = Net Primary Productivity (g/m2)
PAR = Photosynthetically-active radiation (400-700 nm) (W/m2)
Ra = respiration by plants (~50% of GPP)
Leaf area index (m2/m2)
Seasonal pattern of LAI
2.5
Tree LAI
Grass LAI
All LAI
2.0
1.5
Skukuza
30% canopy
1.0
0.5
0.0
1
2
3
wet season
4
5
6
7
8
Month of the year
9
10 11 12
wet season
Unscrambling the tree and grass signals
2002
0.40
0.35
0.35
0.30
0.30
0.25
0.25
Greenness
Greenness
2001
0.40
0.20
0.15
0.20
0.15
0.10
0.10
0.05
0.05
0.00
0.00
-0.05
28-Jul-01
16-Sep-01
5-Nov-01
25-Dec-01 13-Feb-02
date
4-Apr-02
24-May-02
-0.05
13-Jul-02
1-Sep-02
21-Oct-02 10-Dec-02 29-Jan-03 20-Mar-03 9-May-03 28-Jun-03
date
Conclusions
• It is possible to define desertification in a
rigorous and operational way, via the concept of
persistent loss of ecosystem services
• There are mechanisms by which state change in
ecosystem service delivery can occur:
desertification is a real and widespread
phenomenon
• These should be detectable using remote
sensing, but require additional non-remote
sensing data (soil and rainfall), time series, and
often the unmixing of the tree and grass signal