Transcript Observing Kalahari ecosystems at local to regional scales

Observing Kalahari ecosystems at
local to regional scales: a remote
sensing perspective
Nigel Trodd
Coventry University
Vegetation dynamics of Kalahari
ecosystems
savannas, shrublands & grasslands occupy 53 million km2
... including 60% of Africa
state and transition mosaics support sustainable agriculture
Reducing / removing uncertainties
• Vegetation community distribution
• Timing and nature of change
• Vegetation dynamics at regional scales
Aims:
• to develop remote sensing method(s)
to characterise vegetation community
dynamics
• to understand the limits on those
methods at local to regional scales
Method
1
Analyse reflectance properties of individual
landscape components
– H0: there is no difference in the reflectances
2
Simulate the composite reflectance of vegetation
communities
– Develop a landscape reflectance model
– H0: vegetation community structure is not related to
reflectance
3
Apply model to analyse regional vegetation
dynamics
Study areas
Southern Kalahari
• Tshane
• Kalaghadi
Transfrontier Park
• Tsabong
• Severn
Eastern Kalahari
• Makoba
<40% vegetation
cover
Field & lab measurements
Landscape components
Field & Lab
spectroscopy
Experiment 1A: differences in the reflectances of
landscape components
Note: data normalised for
% cover
• grass > bush
visible ~ 3% - 5%
near-infrared 0%
• soil > vegetation
Landsat TM4
Experiment 1b: variation in soil reflectance
principal plane
62
55
t1
t1
m2
m2
tg1
tg1
tg3
tg3
50
Reflectance (%)
Reflectance (%)
56
45
40
50
44
35
30
-70
-60
-50
-40
-30
-20
-10
0
10
20
30
40
50
60
38
-70
70
-60
-50
-40
-30
-20
-10
0
10
20
30
40
50
60
70
50
60
70
View Zenith Angle (degrees)
View Zenith Angle (degrees)
680 nm
850nm
58
50
t1
t1
m2
m2
tg1
tg1
tg3
tg3
54
Reflectance (%)
Reflectance (%)
45
50
40
46
35
-70
-60
-50
-40
-30
-20
-10
0
10
20
30
40
50
60
70
42
-70
-60
orthogonal plane
• Differences of ~2%
• Soil crust ~6% increase
-50
-40
-30
-20
-10
0
10
View Zenith Angle (degrees)
View Zenith Angle (degrees)
20
30
40
Experiment 2: simulating reflectance using a
canopy reflectance model
Simulated reflectance =
 landscape component x reflectance
Bush cover
Grass
cover
0
10%
20%
30%
40%
0
27.3
23.2
19.6
16.2
13.2
30%
22.7
19.0
15.7
12.7
10.0
60%
18.1
14.8
11.8
9.2
*
Results: part 1
• Significant monospectral differences
between soil and vegetation
• Vegetation community structure not
related to reflectance
• Dimensionality of single-date imagery
limits local scale applications
Moving forward...
3
Analyse time series of Earth observation data
– H0: there is no difference in the temporal profiles of vegetation
communities
4
Apply model to analyse regional vegetation dynamics
NOAAAVHRR
G1K - dekad
composite
Results: part 2
• Key periods in the time series are prone
to cloud cover
– Few systems provide data at required temporal
frequency
• Vegetation response to rainfall and temp
is complex
– statistical relationships not reliable
– process-response models immature
• Spatial precision, registration and
spectral calibration
– only output regional scale assessments
Conclusions: problems or
prospects?
• Local scale mapping of key ecosystem variables
is challenging
– limited by dimensionality of spectral data and
variations in soil properties
• Regional scale mapping is feasible but monitoring
is more challenging
• New systems (always) promise more
• but success will depend on integrating data
– multi-sensor
– constraining the landscape reflectance model
(with GI)