Cocoa agroforestry: A viable practice for Cocoa landscape rehabilitation in W...
Mapping Soil and Ecosystem Health in Africa
1. Mapping Soil and
Ecosystem Health in the Land Degradation
Africa Surveillance Framework
(LDSF)
Tor-G. Vågen
World Agroforestry Centre (ICRAF), Nairobi, KENYA
Tuesday, April 12, 2011
2. Land degradation has implications beyond the land
Soahany, Madagascar
Tuesday, April 12, 2011
3. Since landscapes are known to exhibit
hierarchically scaled patterns,
a desirable property of
landscape models
is that they simulate or predict
patterns at different scales
Tuesday, April 12, 2011
4. Survey Sampling
by a survey we mean the process of measuring characteristics of some or all members of an actual population
-
the purpose of which is to make quantitative generalizations about the population as a whole, or its subpopulations (or in
some cases its super-populations)
Probability sampling Non-probability sampling
random systematic stratified convenience judgement quota snowball
sampling sampling sampling sampling sampling sampling sampling
purest form, but reduces sampling
with very large error by first
populations pool simple, also stratifying and
tends to become referred to as the then applying may be used in the nonprobability
biased Nth name selection random sampling exploratory phase equivalent of
technique of research stratified
sampling. first
stratification then
convenience or
judgement
sampling of strata
Tuesday, April 12, 2011
5. AfSIS Sentinel Sites
Probability sampling approach.
Stratified random sample of
African landscapes.
Built on the Land Degradation
Surveillance Framework (LDSF).
Unbiased sample of landscapes
across sub-Saharan Africa.
Initially (“phase I”) 60 sentinel
sites and 60 alternate sites.
Target in this phase - 60 sites
characterized and sampled.
Tuesday, April 12, 2011
6. AfSIS Sentinel Sites
Site = 100 km 2
Cluster = 1 km2
Plot 1
Plot = 0.1 ha
Sub-plot = 0.01 ha
Tuesday, April 12, 2011
10. AfSIS Sentinel Site baseline information
2000 Site averages
Kontela Infiltration testing 2000 Average curves for areas with/
TRUE
Chica_b without root-depth restrictions FALSE
Mbinga (TRUE/FALSE)
1500 1500
1000 1000
IR
IR
500 500
0 0
0 50 100 150 200 0 50 100 150 200
Time Time
2000 2000 Average curves for areas with
Average curves for cultivated (1) 1 TRUE
0 dense woody cover (>40%) FALSE
and natural/semi-natural areas (0)
1500 1500
1000 1000
IR
IR
500 500
0 0
0 50 100 150 200 0 50 100 150 200
Time Time
Tuesday, April 12, 2011
11. IR spectroscopy of soils
Regional network of NIR
MPA (NIR) spectrometer in Arusha
spectral laboratories and
spectral libraries
Nairobi
MPA (NIR) spectrometer in Bamako Construction of IR lab in Lilongwe
NIR training, Arusha Field testing of new spectrometer
Tuesday, April 12, 2011
13. IR spectroscopy
has a wide range of applications, not limited to soils
Baboon
10
Black Rhino
10
Buffalo
11
Bush buck
12
Cape Hare
10
Elephant
17
Giant Forest
Hog
10
Hyena
5
Leopard
2
Mongoose
15
Reedbuck
10
Suni
2
Unknown
3
Warthog
17
Water buck
9
Zebra
12
Partner: KWS
Tuesday, April 12, 2011
16. Scientific workflows
Scalability. Parallel execution on
multi-core systems
Simple extensibility via
a well-defined API for Command line version
plugin extensions for "headless" batch
executions
R integration
Mining of NIR and MIR spectral data
Classification
Clustering
Processing and development of models from MIR spectra
Predictive models
Meta workflows (e.g. cross validation)
Data preprocessing
Databases (data management)
Reporting
Cluster execution
Data management
Sentinel site baselines
Tuesday, April 12, 2011
17. Development of prediction models for soil organic
carbon (SOC) using scientific workflows and R
Tuesday, April 12, 2011
18. Mapping soil carbon
Ol Lentille and Kipsing, northern Laikipia, Kenya
Tuesday, April 12, 2011
23. Automated reporting on soil properties
soil chemical and physical reference values
Tuesday, April 12, 2011
24. Documentation of AfSIS / LDSF methods and
guidelines for implementation
Tuesday, April 12, 2011
25. Documentation of AfSIS / LDSF methods and
guidelines for implementation
“Toolkits”
sentinel site randomization / modeling / ++
Tuesday, April 12, 2011
27. Filled DEM Slope Hydrology
Satellite images and other
spatial covariates
Aspect Specific Wetness
catchment Index
area
Tuesday, April 12, 2011
28. Mapping land cover / vegetation
Thematic layers;
• De-vegetation to enhance soil
background signal
• Soil adjusted vegetation index
• Terrain corrections
• Forest index calculations
• Water index calculations
• Automatic generation of water masks
• Automatic cloud masking
Statistically derived;
• Tree density
Terrain-corrected vegetation index (GRUVI) map
Kwadihombo - north of Morogoro, Tanzania
Tuesday, April 12, 2011
29. Mapping land cover and land use
Tanzania
p(Cultivated)
Tuesday, April 12, 2011
31. Modeling land degradation risk factors
and crop performance
Co-locating trials at cluster level Relating maps to crop performance
Kiberashi Sentinel Site, Tanzania
Percent of Total
10
5
0
1000 1100 1200 1300 1400
Elevation (m)
Tuesday, April 12, 2011
33. Modeling land degradation risk
factors and crop growth response Presence / absence of trees
Presence / absence of erosion
Presence / absence of
root-depth restrictions
Kiberashi sentinel site (Tanzania)
Thuchila sentinel site (Malawi)
Tuesday, April 12, 2011
34. Mapping eroded landscapes
Kiberashi sentinel site (Tanzania)
1987 (left); 2006 (right)
Tuesday, April 12, 2011
35. Mapping eroded landscapes
Yij ! Bernoulli(pij)
logit(pij) = µ+xij!+Vi
Vi ! iid N(0,"2)
Yij indicates presence/absence of
for example erosion in the ith site
and the jth cluster
Mt. Meru / Arusha / Moshi, Tanzania
Tuesday, April 12, 2011